diff --git a/bench/abc/optimized/cuddZddLin.ll b/bench/abc/optimized/cuddZddLin.ll
index 3aa46b8c9ed..a2c0642925a 100644
--- a/bench/abc/optimized/cuddZddLin.ll
+++ b/bench/abc/optimized/cuddZddLin.ll
@@ -1012,13 +1012,13 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   br i1 %exitcond439.not, label %.preheader357, label %.lr.ph381, !llvm.loop !69
 
 .preheader355:                                    ; preds = %.lr.ph385, %.preheader357
-  %.0320.lcssa = phi i32 [ %27, %.preheader357 ], [ %107, %.lr.ph385 ]
+  %.0320.lcssa = phi i32 [ %27, %.preheader357 ], [ %100, %.lr.ph385 ]
   %.not338399 = icmp eq ptr %.0299.lcssa, null
   br i1 %.not338399, label %.preheader, label %.lr.ph403
 
 .lr.ph385:                                        ; preds = %.preheader357, %.lr.ph385
   %.2315384 = phi ptr [ %69, %.lr.ph385 ], [ %.0.lcssa, %.preheader357 ]
-  %.0320383 = phi i32 [ %107, %.lr.ph385 ], [ %27, %.preheader357 ]
+  %.0320383 = phi i32 [ %100, %.lr.ph385 ], [ %27, %.preheader357 ]
   %68 = getelementptr inbounds nuw i8, ptr %.2315384, i64 8
   %69 = load ptr, ptr %68, align 8, !tbaa !48
   %70 = getelementptr inbounds nuw i8, ptr %.2315384, i64 16
@@ -1042,29 +1042,24 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   %84 = inttoptr i64 %83 to ptr
   %85 = getelementptr inbounds nuw i8, ptr %84, i64 32
   %86 = load i64, ptr %85, align 8, !tbaa !70
-  %87 = shl i64 %86, 1
-  %88 = and i64 %82, 1
-  %89 = or disjoint i64 %87, %88
-  %90 = trunc i64 %89 to i32
-  %91 = mul i32 %90, 12582917
-  %92 = ptrtoint ptr %78 to i64
-  %93 = and i64 %92, -2
-  %94 = inttoptr i64 %93 to ptr
-  %95 = getelementptr inbounds nuw i8, ptr %94, i64 32
-  %96 = load i64, ptr %95, align 8, !tbaa !70
-  %97 = shl i64 %96, 1
-  %98 = and i64 %92, 1
-  %99 = or disjoint i64 %97, %98
-  %100 = trunc i64 %99 to i32
-  %101 = add i32 %91, %100
-  %102 = mul i32 %101, 4256249
-  %103 = lshr i32 %102, %31
-  %104 = sext i32 %103 to i64
-  %105 = getelementptr inbounds ptr, ptr %25, i64 %104
-  %106 = load ptr, ptr %105, align 8, !tbaa !25
-  store ptr %106, ptr %68, align 8, !tbaa !48
-  store ptr %.2315384, ptr %105, align 8, !tbaa !25
-  %107 = add nsw i32 %.0320383, 1
+  %.tr = trunc i64 %86 to i32
+  %87 = mul i32 %.tr, 25165834
+  %88 = ptrtoint ptr %78 to i64
+  %89 = and i64 %88, -2
+  %90 = inttoptr i64 %89 to ptr
+  %91 = getelementptr inbounds nuw i8, ptr %90, i64 32
+  %92 = load i64, ptr %91, align 8, !tbaa !70
+  %.tr472 = trunc i64 %92 to i32
+  %93 = shl i32 %.tr472, 1
+  %94 = add i32 %87, %93
+  %95 = mul i32 %94, 4256249
+  %96 = lshr i32 %95, %31
+  %97 = sext i32 %96 to i64
+  %98 = getelementptr inbounds ptr, ptr %25, i64 %97
+  %99 = load ptr, ptr %98, align 8, !tbaa !25
+  store ptr %99, ptr %68, align 8, !tbaa !48
+  store ptr %.2315384, ptr %98, align 8, !tbaa !25
+  %100 = add nsw i32 %.0320383, 1
   %.not = icmp eq ptr %69, null
   br i1 %.not, label %.preheader355, label %.lr.ph385, !llvm.loop !71
 
@@ -1085,46 +1080,46 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   %109 = getelementptr inbounds nuw i8, ptr %.3316402, i64 8
   %110 = load ptr, ptr %109, align 8, !tbaa !48
   %111 = getelementptr inbounds nuw i8, ptr %.3316402, i64 16
+  %105 = load ptr, ptr %104, align 8, !tbaa !64
+  %106 = load i32, ptr %105, align 8, !tbaa !65
+  %107 = icmp eq i32 %106, %23
+  %108 = icmp eq i32 %106, %10
+  %or.cond347 = select i1 %107, i1 true, i1 %108
+  br i1 %or.cond347, label %109, label %113
+
+109:                                              ; preds = %.lr.ph403
+  %110 = getelementptr inbounds nuw i8, ptr %105, i64 16
+  %111 = getelementptr inbounds nuw i8, ptr %105, i64 24
   %112 = load ptr, ptr %111, align 8, !tbaa !64
-  %113 = load i32, ptr %112, align 8, !tbaa !65
-  %114 = icmp eq i32 %113, %23
-  %115 = icmp eq i32 %113, %10
-  %or.cond347 = select i1 %114, i1 true, i1 %115
-  br i1 %or.cond347, label %116, label %120
-
-116:                                              ; preds = %.lr.ph403
-  %117 = getelementptr inbounds nuw i8, ptr %112, i64 16
-  %118 = getelementptr inbounds nuw i8, ptr %112, i64 24
-  %119 = load ptr, ptr %118, align 8, !tbaa !64
-  br label %120
-
-120:                                              ; preds = %.lr.ph403, %116
-  %.0312.in = phi ptr [ %117, %116 ], [ @empty, %.lr.ph403 ]
-  %.0311 = phi ptr [ %119, %116 ], [ %112, %.lr.ph403 ]
+  br label %113
+
+113:; preds = %.lr.ph403, %109
+  %.0312.in = phi ptr [ %110, %109 ], [ @empty, %.lr.ph403 ]
+  %.0311 = phi ptr [ %112, %109 ], [ %105, %.lr.ph403 ]
   %.0312 = load ptr, ptr %.0312.in, align 8, !tbaa !64
   %121 = getelementptr inbounds nuw i8, ptr %.3316402, i64 24
+  %115 = load ptr, ptr %114, align 8, !tbaa !64
+  %116 = load i32, ptr %115, align 8, !tbaa !65
+  %117 = icmp eq i32 %116, %23
+  %118 = icmp eq i32 %116, %10
+  %or.cond348 = select i1 %117, i1 true, i1 %118
+  br i1 %or.cond348, label %119, label %123
+
+119:                                              ; preds = %113
+  %120 = getelementptr inbounds nuw i8, ptr %115, i64 16
+  %121 = getelementptr inbounds nuw i8, ptr %115, i64 24
   %122 = load ptr, ptr %121, align 8, !tbaa !64
-  %123 = load i32, ptr %122, align 8, !tbaa !65
-  %124 = icmp eq i32 %123, %23
-  %125 = icmp eq i32 %123, %10
-  %or.cond348 = select i1 %124, i1 true, i1 %125
-  br i1 %or.cond348, label %126, label %130
-
-126:                                              ; preds = %120
-  %127 = getelementptr inbounds nuw i8, ptr %122, i64 16
-  %128 = getelementptr inbounds nuw i8, ptr %122, i64 24
-  %129 = load ptr, ptr %128, align 8, !tbaa !64
-  br label %130
-
-130:                                              ; preds = %120, %126
-  %.0310.in = phi ptr [ %127, %126 ], [ @empty, %120 ]
-  %.0309 = phi ptr [ %129, %126 ], [ %122, %120 ]
+  br label %123
+
+123:; preds = %113, %119
+  %.0310.in = phi ptr [ %120, %119 ], [ @empty, %113 ]
+  %.0309 = phi ptr [ %122, %119 ], [ %115, %113 ]
   %.0310 = load ptr, ptr %.0310.in, align 8, !tbaa !64
   %131 = load ptr, ptr @empty, align 8, !tbaa !25
   %132 = icmp eq ptr %.0310, %131
   br i1 %132, label %.loopexit466, label %133
 
-133:                                              ; preds = %130
+133:                                              ; preds = %123
   %134 = ptrtoint ptr %.0310 to i64
   %135 = and i64 %134, -2
   %136 = inttoptr i64 %135 to ptr
@@ -1154,7 +1149,7 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   br i1 %cond388, label %._crit_edge392, label %.lr.ph391
 
 .lr.ph391:                                        ; preds = %133, %168
-  %.1306389 = phi ptr [ %.1306, %168 ], [ %.1306387, %133 ]
+  %.1306389 = phi ptr [ %.1306, %161 ], [ %.1306387, %126 ]
   %158 = getelementptr inbounds nuw i8, ptr %.1306389, i64 16
   %159 = load ptr, ptr %158, align 8, !tbaa !64
   %160 = icmp eq ptr %159, %.0310
@@ -1201,10 +1196,10 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   store i32 %181, ptr %179, align 4, !tbaa !47
   br label %.loopexit466
 
-.loopexit466:                                     ; preds = %165, %130, %172
-  %.1306389.lcssa.sink = phi ptr [ %.0311, %172 ], [ %.0311, %130 ], [ %.1306389, %165 ]
-  %.2322 = phi i32 [ %176, %172 ], [ %.1321401, %130 ], [ %.1321401, %165 ]
-  %.0305 = phi ptr [ %170, %172 ], [ %.0311, %130 ], [ %.1306389, %165 ]
+.loopexit466:                                     ; preds = %165, %123, %172
+  %.1306389.lcssa.sink = phi ptr [ %.0311, %165 ], [ %.0311, %123 ], [ %.1306389, %158 ]
+  %.2322 = phi i32 [ %176, %165 ], [ %.1321401, %123 ], [ %.1321401, %158 ]
+  %.0305 = phi ptr [ %170, %165 ], [ %.0311, %123 ], [ %.1306389, %158 ]
   %182 = getelementptr inbounds nuw i8, ptr %.1306389.lcssa.sink, i64 4
   %183 = load i32, ptr %182, align 4, !tbaa !47
   %184 = add i32 %183, 1
@@ -1244,7 +1239,7 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   br i1 %cond351394, label %._crit_edge398, label %.lr.ph397
 
 .lr.ph397:                                        ; preds = %187, %222
-  %.1304395 = phi ptr [ %.1304, %222 ], [ %.1304393, %187 ]
+  %.1304395 = phi ptr [ %.1304, %215 ], [ %.1304393, %180 ]
   %212 = getelementptr inbounds nuw i8, ptr %.1304395, i64 16
   %213 = load ptr, ptr %212, align 8, !tbaa !64
   %214 = icmp eq ptr %213, %.0312
@@ -1292,9 +1287,9 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   br label %.loopexit465
 
 .loopexit465:                                     ; preds = %219, %.loopexit466, %226
-  %.1304395.lcssa.sink = phi ptr [ %.0309, %226 ], [ %.0309, %.loopexit466 ], [ %.1304395, %219 ]
-  %.3323 = phi i32 [ %230, %226 ], [ %.2322, %.loopexit466 ], [ %.2322, %219 ]
-  %.0303 = phi ptr [ %224, %226 ], [ %.0309, %.loopexit466 ], [ %.1304395, %219 ]
+  %.1304395.lcssa.sink = phi ptr [ %.0309, %219 ], [ %.0309, %.loopexit466 ], [ %.1304395, %212 ]
+  %.3323 = phi i32 [ %230, %219 ], [ %.2322, %.loopexit466 ], [ %.2322, %212 ]
+  %.0303 = phi ptr [ %224, %219 ], [ %.0309, %.loopexit466 ], [ %.1304395, %212 ]
   %236 = getelementptr inbounds nuw i8, ptr %.1304395.lcssa.sink, i64 4
   %237 = load i32, ptr %236, align 4, !tbaa !47
   %238 = add i32 %237, 1
@@ -1341,10 +1336,10 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   br i1 %.not339412, label %._crit_edge419, label %.lr.ph418
 
 .lr.ph418:                                        ; preds = %265, %392
-  %.0295416 = phi ptr [ %.1296, %392 ], [ null, %265 ]
-  %.4415 = phi ptr [ %269, %392 ], [ %267, %265 ]
-  %.5414 = phi i32 [ %.6, %392 ], [ %.4324423, %265 ]
-  %.2327413 = phi i32 [ %.3328, %392 ], [ %.1326422, %265 ]
+  %.0295416 = phi ptr [ %.1296, %381 ], [ null, %258 ]
+  %.4415 = phi ptr [ %269, %381 ], [ %267, %258 ]
+  %.5414 = phi i32 [ %.6, %381 ], [ %.4324423, %258 ]
+  %.2327413 = phi i32 [ %.3328, %381 ], [ %.1326422, %258 ]
   %268 = getelementptr inbounds nuw i8, ptr %.4415, i64 8
   %269 = load ptr, ptr %268, align 8, !tbaa !48
   %270 = getelementptr inbounds nuw i8, ptr %.4415, i64 4
@@ -1407,6 +1402,14 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   %303 = add i32 %302, -1
   store i32 %303, ptr %301, align 4, !tbaa !47
   %304 = ptrtoint ptr %300 to i64
+  %298 = and i64 %297, -2
+  %299 = inttoptr i64 %298 to ptr
+  %300 = getelementptr inbounds nuw i8, ptr %299, i64 32
+  %301 = load i64, ptr %300, align 8, !tbaa !70
+  %.tr473 = trunc i64 %301 to i32
+  %302 = mul i32 %.tr473, 25165834
+  %303 = load ptr, ptr @empty, align 8, !tbaa !25
+  %304 = ptrtoint ptr %303 to i64
   %305 = and i64 %304, -2
   %306 = inttoptr i64 %305 to ptr
   %307 = getelementptr inbounds nuw i8, ptr %306, i64 32
@@ -1415,28 +1418,17 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   %310 = and i64 %304, 1
   %311 = or disjoint i64 %309, %310
   %312 = trunc i64 %311 to i32
-  %313 = mul i32 %312, 12582917
-  %314 = load ptr, ptr @empty, align 8, !tbaa !25
-  %315 = ptrtoint ptr %314 to i64
-  %316 = and i64 %315, -2
-  %317 = inttoptr i64 %316 to ptr
-  %318 = getelementptr inbounds nuw i8, ptr %317, i64 32
-  %319 = load i64, ptr %318, align 8, !tbaa !70
-  %320 = shl i64 %319, 1
-  %321 = and i64 %315, 1
-  %322 = or disjoint i64 %320, %321
-  %323 = trunc i64 %322 to i32
-  %324 = add i32 %313, %323
-  %325 = mul i32 %324, 4256249
-  %326 = lshr i32 %325, %31
-  %327 = sext i32 %326 to i64
-  %328 = getelementptr inbounds ptr, ptr %25, i64 %327
-  %.2307406 = load ptr, ptr %328, align 8, !tbaa !25
+  %313 = add i32 %302, %312
+  %314 = mul i32 %313, 4256249
+  %315 = lshr i32 %314, %31
+  %316 = sext i32 %315 to i64
+  %317 = getelementptr inbounds ptr, ptr %25, i64 %316
+  %.2307406 = load ptr, ptr %317, align 8, !tbaa !25
   %cond352407 = icmp eq ptr %.2307406, null
   br i1 %cond352407, label %._crit_edge411, label %.lr.ph410
 
 .lr.ph410:                                        ; preds = %298, %344
-  %.2307408 = phi ptr [ %.2307, %344 ], [ %.2307406, %298 ]
+  %.2307408 = phi ptr [ %.2307, %333 ], [ %.2307406, %291 ]
   %329 = getelementptr inbounds nuw i8, ptr %.2307408, i64 16
   %330 = load ptr, ptr %329, align 8, !tbaa !64
   %331 = icmp eq ptr %330, %300
@@ -1445,7 +1437,7 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
 332:                                              ; preds = %.lr.ph410
   %333 = getelementptr inbounds nuw i8, ptr %.2307408, i64 24
   %334 = load ptr, ptr %333, align 8, !tbaa !64
-  %335 = icmp eq ptr %334, %314
+  %335 = icmp eq ptr %334, %303
   br i1 %335, label %336, label %344
 
 336:                                              ; preds = %332
@@ -1481,11 +1473,11 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   %351 = load ptr, ptr @empty, align 8, !tbaa !25
   %352 = getelementptr inbounds nuw i8, ptr %346, i64 24
   store ptr %351, ptr %352, align 8, !tbaa !64
-  %353 = load ptr, ptr %328, align 8, !tbaa !25
+  %353 = load ptr, ptr %317, align 8, !tbaa !25
   %354 = getelementptr inbounds nuw i8, ptr %346, i64 8
   store ptr %353, ptr %354, align 8, !tbaa !48
-  store ptr %346, ptr %328, align 8, !tbaa !25
-  %355 = zext i32 %326 to i64
+  store ptr %346, ptr %317, align 8, !tbaa !25
+  %355 = zext i32 %315 to i64
   %356 = icmp eq i64 %indvars.iv440, %355
   %or.cond = and i1 %294, %356
   %spec.select = select i1 %or.cond, ptr %346, ptr %.0295416
@@ -1500,9 +1492,9 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   br label %363
 
 363:                                              ; preds = %339, %348
-  %.7 = phi i32 [ %.5414, %348 ], [ %343, %339 ]
-  %.3308 = phi ptr [ %346, %348 ], [ %.2307408, %339 ]
-  %.2297 = phi ptr [ %spec.select, %348 ], [ %.0295416, %339 ]
+  %.7 = phi i32 [ %.5414, %337 ], [ %343, %328 ]
+  %.3308 = phi ptr [ %346, %337 ], [ %.2307408, %328 ]
+  %.2297 = phi ptr [ %spec.select, %337 ], [ %.0295416, %328 ]
   store ptr %.3308, ptr %299, align 8, !tbaa !64
   %364 = getelementptr inbounds nuw i8, ptr %.4415, i64 24
   %365 = load ptr, ptr %364, align 8, !tbaa !64
@@ -1537,15 +1529,15 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   br label %392
 
 392:                                              ; preds = %290, %363, %287, %288
-  %.3328 = phi i32 [ %.2327413, %287 ], [ %.2327413, %288 ], [ %388, %363 ], [ %.2327413, %290 ]
-  %.6 = phi i32 [ %285, %287 ], [ %285, %288 ], [ %.7, %363 ], [ %.5414, %290 ]
-  %.1296 = phi ptr [ null, %287 ], [ %.0295416, %288 ], [ %.2297, %363 ], [ %.4415, %290 ]
+  %.3328 = phi i32 [ %.2327413, %280 ], [ %.2327413, %281 ], [ %388, %352 ], [ %.2327413, %283 ]
+  %.6 = phi i32 [ %285, %280 ], [ %285, %281 ], [ %.7, %352 ], [ %.5414, %283 ]
+  %.1296 = phi ptr [ null, %280 ], [ %.0295416, %281 ], [ %.2297, %352 ], [ %.4415, %283 ]
   %.not339 = icmp eq ptr %269, null
   br i1 %.not339, label %._crit_edge419, label %.lr.ph418, !llvm.loop !76
 
 ._crit_edge419:                                   ; preds = %392, %265
-  %.2327.lcssa = phi i32 [ %.1326422, %265 ], [ %.3328, %392 ]
-  %.5.lcssa = phi i32 [ %.4324423, %265 ], [ %.6, %392 ]
+  %.2327.lcssa = phi i32 [ %.1326422, %258 ], [ %.3328, %381 ]
+  %.5.lcssa = phi i32 [ %.4324423, %258 ], [ %.6, %381 ]
   %indvars.iv.next441 = add nuw nsw i64 %indvars.iv440, 1
   %exitcond444.not = icmp eq i64 %indvars.iv.next441, %wide.trip.count443
   br i1 %exitcond444.not, label %._crit_edge426, label %265, !llvm.loop !77
@@ -1558,23 +1550,23 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   %395 = getelementptr inbounds nuw i8, ptr %394, i64 16
   store i32 %.1326.lcssa, ptr %395, align 8, !tbaa !31
   %396 = getelementptr inbounds %struct.DdSubtable, ptr %393, i64 %21
+  %386 = getelementptr inbounds nuw i8, ptr %385, i64 16
+  store i32 %.4324.lcssa, ptr %386, align 8, !tbaa !31
+  %387 = add i32 %27, %16
+  %388 = getelementptr inbounds nuw i8, ptr %0, i64 232
+  %389 = load i32, ptr %388, align 8, !tbaa !36
+  %390 = sub i32 %.1326.lcssa, %387
+  %391 = add i32 %390, %.4324.lcssa
+  %392 = add i32 %391, %389
+  store i32 %392, ptr %388, align 8, !tbaa !36
+  %393 = getelementptr inbounds nuw i8, ptr %0, i64 360
+  %394 = load ptr, ptr %393, align 8, !tbaa !78
+  %395 = getelementptr inbounds ptr, ptr %394, i64 %8
+  %396 = load ptr, ptr %395, align 8, !tbaa !25
   %397 = getelementptr inbounds nuw i8, ptr %396, i64 16
-  store i32 %.4324.lcssa, ptr %397, align 8, !tbaa !31
-  %398 = add i32 %27, %16
-  %399 = getelementptr inbounds nuw i8, ptr %0, i64 232
-  %400 = load i32, ptr %399, align 8, !tbaa !36
-  %401 = sub i32 %.1326.lcssa, %398
-  %402 = add i32 %401, %.4324.lcssa
-  %403 = add i32 %402, %400
-  store i32 %403, ptr %399, align 8, !tbaa !36
-  %404 = getelementptr inbounds nuw i8, ptr %0, i64 360
-  %405 = load ptr, ptr %404, align 8, !tbaa !78
-  %406 = getelementptr inbounds ptr, ptr %405, i64 %8
-  %407 = load ptr, ptr %406, align 8, !tbaa !25
-  %408 = getelementptr inbounds nuw i8, ptr %407, i64 16
-  %409 = load ptr, ptr %408, align 8, !tbaa !64
-  %410 = getelementptr inbounds ptr, ptr %405, i64 %21
-  store ptr %409, ptr %410, align 8, !tbaa !25
+  %398 = load ptr, ptr %397, align 8, !tbaa !64
+  %399 = getelementptr inbounds ptr, ptr %394, i64 %21
+  store ptr %398, ptr %399, align 8, !tbaa !25
   br label %414
 
 .loopexit:                                        ; preds = %._crit_edge398, %._crit_edge392, %._crit_edge411
@@ -1584,7 +1576,7 @@ define internal fastcc i32 @cuddZddLinearInPlace(ptr noundef %0, i32 noundef %1,
   br label %414
 
 414:                                              ; preds = %.loopexit, %._crit_edge426
-  %.0329 = phi i32 [ 0, %.loopexit ], [ %403, %._crit_edge426 ]
+  %.0329 = phi i32 [ 0, %.loopexit ], [ %392, %._crit_edge426 ]
   ret i32 %.0329
 }
 
diff --git a/bench/lean4/optimized/AndFlatten.ll b/bench/lean4/optimized/AndFlatten.ll
index ab20e4f613b..5555dc308a9 100644
--- a/bench/lean4/optimized/AndFlatten.ll
+++ b/bench/lean4/optimized/AndFlatten.ll
@@ -13070,56 +13070,44 @@ lean_alloc_closure.exit:                          ; preds = %7
 define ptr @l_Array_foldlMUnsafe_fold___at_Lean_Elab_Tactic_BVDecide_Frontend_Normalize_andFlatteningPass_processGoal___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9) local_unnamed_addr #1 {
   %11 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %11, align 8, !tbaa !12
-  %12 = ptrtoint ptr %1 to i64
-  %13 = and i64 %12, 1
-  %.not = icmp eq i64 %13, 0
-  br i1 %.not, label %14, label %lean_dec.exit16
+  %12 = load i32, ptr %1, align 8, !tbaa !4
+  %13 = icmp sgt i32 %12, 1
+  br i1 %13, label %14, label %16, !prof !11
 
 14:                                               ; preds = %10
-  %15 = load i32, ptr %1, align 4, !tbaa !4
-  %16 = icmp sgt i32 %15, 1
-  br i1 %16, label %17, label %19, !prof !11
-
-17:                                               ; preds = %14
-  %18 = add nsw i32 %15, -1
-  store i32 %18, ptr %1, align 4, !tbaa !4
+  %15 = add nsw i32 %12, -1
+  store i32 %15, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit16
 
-19:                                               ; preds = %14
-  %.not.i = icmp eq i32 %15, 0
+19:                                               ; preds = %10
+  %.not.i = icmp eq i32 %12, 0
   br i1 %.not.i, label %lean_dec.exit16, label %20
 
 20:                                               ; preds = %19
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit16
 
-lean_dec.exit16:                                  ; preds = %20, %19, %17, %10
+lean_dec.exit16:                                  ; preds = %20, %19, %14
   %21 = getelementptr i8, ptr %2, i64 8
   %.val21 = load i64, ptr %21, align 8, !tbaa !12
-  %22 = ptrtoint ptr %2 to i64
-  %23 = and i64 %22, 1
-  %.not22 = icmp eq i64 %23, 0
-  br i1 %.not22, label %24, label %lean_dec.exit15
-
-24:                                               ; preds = %lean_dec.exit16
-  %25 = load i32, ptr %2, align 4, !tbaa !4
-  %26 = icmp sgt i32 %25, 1
-  br i1 %26, label %27, label %29, !prof !11
+  %19 = load i32, ptr %2, align 8, !tbaa !4
+  %20 = icmp sgt i32 %19, 1
+  br i1 %20, label %21, label %23, !prof !11
 
-27:                                               ; preds = %24
-  %28 = add nsw i32 %25, -1
+27:                                               ; preds = %lean_dec.exit16
+  %28 = add nsw i32 %19, -1
   store i32 %28, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit15
 
-29:                                               ; preds = %24
-  %.not.i17 = icmp eq i32 %25, 0
+29:                                               ; preds = %lean_dec.exit16
+  %.not.i17 = icmp eq i32 %19, 0
   br i1 %.not.i17, label %lean_dec.exit15, label %30
 
 30:                                               ; preds = %29
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #4
   br label %lean_dec.exit15
 
-lean_dec.exit15:                                  ; preds = %30, %29, %27, %lean_dec.exit16
+lean_dec.exit15:                                  ; preds = %30, %29, %21
   %31 = tail call ptr @l_Array_foldlMUnsafe_fold___at_Lean_Elab_Tactic_BVDecide_Frontend_Normalize_andFlatteningPass_processGoal___spec__1(ptr noundef %0, i64 noundef %.val, i64 noundef %.val21, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9)
   %32 = ptrtoint ptr %0 to i64
   %33 = and i64 %32, 1
diff --git a/bench/lean4/optimized/Assumption.ll b/bench/lean4/optimized/Assumption.ll
index c905357cc9d..8ce1f9c6902 100644
--- a/bench/lean4/optimized/Assumption.ll
+++ b/bench/lean4/optimized/Assumption.ll
@@ -6169,30 +6169,24 @@ define ptr @l_Lean_LocalContext_findDeclRevM_x3f___at_Lean_Meta_findLocalDeclWit
   %8 = getelementptr inbounds nuw i8, ptr %1, i64 16
   %9 = load ptr, ptr %8, align 8, !tbaa !11
   %10 = tail call ptr @l_Lean_PersistentArray_findSomeRevM_x3f___at_Lean_Meta_findLocalDeclWithType_x3f___spec__2(ptr noundef %0, ptr noundef %9, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6)
-  %11 = ptrtoint ptr %1 to i64
-  %12 = and i64 %11, 1
-  %.not = icmp eq i64 %12, 0
-  br i1 %.not, label %13, label %lean_dec.exit
+  %11 = load i32, ptr %1, align 8, !tbaa !5
+  %12 = icmp sgt i32 %11, 1
+  br i1 %12, label %13, label %15, !prof !10
 
 13:                                               ; preds = %7
-  %14 = load i32, ptr %1, align 4, !tbaa !5
-  %15 = icmp sgt i32 %14, 1
-  br i1 %15, label %16, label %18, !prof !10
-
-16:                                               ; preds = %13
-  %17 = add nsw i32 %14, -1
-  store i32 %17, ptr %1, align 4, !tbaa !5
+  %14 = add nsw i32 %11, -1
+  store i32 %14, ptr %1, align 4, !tbaa !5
   br label %lean_dec.exit
 
-18:                                               ; preds = %13
-  %.not.i = icmp eq i32 %14, 0
+18:                                               ; preds = %7
+  %.not.i = icmp eq i32 %11, 0
   br i1 %.not.i, label %lean_dec.exit, label %19
 
 19:                                               ; preds = %18
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %19, %18, %16, %7
+lean_dec.exit:                                    ; preds = %19, %18, %13
   ret ptr %10
 }
 
diff --git a/bench/lean4/optimized/Attach.ll b/bench/lean4/optimized/Attach.ll
index 046da890f73..200ff5c5c01 100644
--- a/bench/lean4/optimized/Attach.ll
+++ b/bench/lean4/optimized/Attach.ll
@@ -205,56 +205,44 @@ lean_alloc_closure.exit:                          ; preds = %3
 define ptr @l_Array_mapMUnsafe_map___at_Vector_pmap___spec__1___rarg___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3) #0 {
   %5 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %5, align 8, !tbaa !14
-  %6 = ptrtoint ptr %1 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit8
+  %6 = load i32, ptr %1, align 8, !tbaa !4
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !9
 
 8:                                                ; preds = %4
-  %9 = load i32, ptr %1, align 4, !tbaa !4
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !9
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %1, align 4, !tbaa !4
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit8
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %4
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit8, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_dec.exit8
 
-lean_dec.exit8:                                   ; preds = %14, %13, %11, %4
+lean_dec.exit8:                                   ; preds = %14, %13, %8
   %15 = getelementptr i8, ptr %2, i64 8
   %.val11 = load i64, ptr %15, align 8, !tbaa !14
-  %16 = ptrtoint ptr %2 to i64
-  %17 = and i64 %16, 1
-  %.not12 = icmp eq i64 %17, 0
-  br i1 %.not12, label %18, label %lean_dec.exit
-
-18:                                               ; preds = %lean_dec.exit8
-  %19 = load i32, ptr %2, align 4, !tbaa !4
-  %20 = icmp sgt i32 %19, 1
-  br i1 %20, label %21, label %23, !prof !9
-
-21:                                               ; preds = %18
-  %22 = add nsw i32 %19, -1
-  store i32 %22, ptr %2, align 4, !tbaa !4
+  %13 = load i32, ptr %2, align 8, !tbaa !4
+  %14 = icmp sgt i32 %13, 1
+  br i1 %14, label %15, label %17, !prof !9
+
+15:                                               ; preds = %lean_dec.exit8
+  %16 = add nsw i32 %13, -1
+  store i32 %16, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit
 
-23:                                               ; preds = %18
-  %.not.i9 = icmp eq i32 %19, 0
+23:                                               ; preds = %lean_dec.exit8
+  %.not.i9 = icmp eq i32 %13, 0
   br i1 %.not.i9, label %lean_dec.exit, label %24
 
 24:                                               ; preds = %23
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %24, %23, %21, %lean_dec.exit8
+lean_dec.exit:                                    ; preds = %24, %23, %15
   %25 = tail call ptr @l_Array_mapMUnsafe_map___at_Vector_pmap___spec__1___rarg(ptr noundef %0, i64 noundef %.val, i64 noundef %.val11, ptr noundef %3)
   ret ptr %25
 }
@@ -835,56 +823,44 @@ lean_alloc_closure.exit:                          ; preds = %3
 define ptr @l_Array_mapMUnsafe_map___at___private_Init_Data_Vector_Attach_0__Vector_pmapImpl___spec__1___rarg___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3) #0 {
   %5 = getelementptr i8, ptr %1, i64 8
   %.val11 = load i64, ptr %5, align 8, !tbaa !14
-  %6 = ptrtoint ptr %1 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit8
+  %6 = load i32, ptr %1, align 8, !tbaa !4
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !9
 
 8:                                                ; preds = %4
-  %9 = load i32, ptr %1, align 4, !tbaa !4
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !9
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %1, align 4, !tbaa !4
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit8
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %4
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit8, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_dec.exit8
 
-lean_dec.exit8:                                   ; preds = %14, %13, %11, %4
+lean_dec.exit8:                                   ; preds = %14, %13, %8
   %15 = getelementptr i8, ptr %2, i64 8
   %.val = load i64, ptr %15, align 8, !tbaa !14
-  %16 = ptrtoint ptr %2 to i64
-  %17 = and i64 %16, 1
-  %.not12 = icmp eq i64 %17, 0
-  br i1 %.not12, label %18, label %lean_dec.exit
-
-18:                                               ; preds = %lean_dec.exit8
-  %19 = load i32, ptr %2, align 4, !tbaa !4
-  %20 = icmp sgt i32 %19, 1
-  br i1 %20, label %21, label %23, !prof !9
-
-21:                                               ; preds = %18
-  %22 = add nsw i32 %19, -1
-  store i32 %22, ptr %2, align 4, !tbaa !4
+  %13 = load i32, ptr %2, align 8, !tbaa !4
+  %14 = icmp sgt i32 %13, 1
+  br i1 %14, label %15, label %17, !prof !9
+
+15:                                               ; preds = %lean_dec.exit8
+  %16 = add nsw i32 %13, -1
+  store i32 %16, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit
 
-23:                                               ; preds = %18
-  %.not.i9 = icmp eq i32 %19, 0
+23:                                               ; preds = %lean_dec.exit8
+  %.not.i9 = icmp eq i32 %13, 0
   br i1 %.not.i9, label %lean_dec.exit, label %24
 
 24:                                               ; preds = %23
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %24, %23, %21, %lean_dec.exit8
+lean_dec.exit:                                    ; preds = %24, %23, %15
   %25 = tail call ptr @l_Array_mapMUnsafe_map___at___private_Init_Data_Vector_Attach_0__Vector_pmapImpl___spec__1___rarg(ptr noundef %0, i64 noundef %.val11, i64 noundef %.val, ptr noundef %3)
   ret ptr %25
 }
@@ -1120,56 +1096,44 @@ lean_alloc_closure.exit:                          ; preds = %2
 define ptr @l_Array_mapMUnsafe_map___at_Vector_unattach___spec__1___rarg___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) #0 {
   %4 = getelementptr i8, ptr %0, i64 8
   %.val10 = load i64, ptr %4, align 8, !tbaa !14
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit7
+  %5 = load i32, ptr %0, align 8, !tbaa !4
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !9
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %0, align 4, !tbaa !4
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !9
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !4
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !4
   br label %lean_dec.exit7
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit7, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_dec.exit7
 
-lean_dec.exit7:                                   ; preds = %13, %12, %10, %3
+lean_dec.exit7:                                   ; preds = %13, %12, %7
   %14 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %14, align 8, !tbaa !14
-  %15 = ptrtoint ptr %1 to i64
-  %16 = and i64 %15, 1
-  %.not11 = icmp eq i64 %16, 0
-  br i1 %.not11, label %17, label %lean_dec.exit
-
-17:                                               ; preds = %lean_dec.exit7
-  %18 = load i32, ptr %1, align 4, !tbaa !4
-  %19 = icmp sgt i32 %18, 1
-  br i1 %19, label %20, label %22, !prof !9
-
-20:                                               ; preds = %17
-  %21 = add nsw i32 %18, -1
-  store i32 %21, ptr %1, align 4, !tbaa !4
+  %15 = load i32, ptr %1, align 8, !tbaa !4
+  %13 = icmp sgt i32 %15, 1
+  br i1 %13, label %14, label %16, !prof !9
+
+14:                                               ; preds = %lean_dec.exit7
+  %15 = add nsw i32 %12, -1
+  store i32 %15, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit
 
-22:                                               ; preds = %17
-  %.not.i8 = icmp eq i32 %18, 0
+22:                                               ; preds = %lean_dec.exit7
+  %.not.i8 = icmp eq i32 %12, 0
   br i1 %.not.i8, label %lean_dec.exit, label %23
 
 23:                                               ; preds = %22
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %23, %22, %20, %lean_dec.exit7
+lean_dec.exit:                                    ; preds = %23, %22, %14
   %24 = tail call ptr @l_Array_mapMUnsafe_map___at_Vector_unattach___spec__1___rarg(i64 noundef %.val10, i64 noundef %.val, ptr noundef %2)
   ret ptr %24
 }
diff --git a/bench/lean4/optimized/AutoBound.ll b/bench/lean4/optimized/AutoBound.ll
index c961dfdcfe6..5241c6f97ad 100644
--- a/bench/lean4/optimized/AutoBound.ll
+++ b/bench/lean4/optimized/AutoBound.ll
@@ -774,22 +774,16 @@ lean_inc.exit24:
   %2 = shl i64 %.val, 1
   %3 = add i64 %2, -1
   %4 = inttoptr i64 %3 to ptr
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_inc.exit
-
-7:                                                ; preds = %lean_inc.exit24
-  %.val.i35 = load i32, ptr %0, align 4, !tbaa !10
-  %8 = icmp sgt i32 %.val.i35, 0
-  br i1 %8, label %9, label %11, !prof !14
+  %.val.i35 = load i32, ptr %0, align 8, !tbaa !10
+  %5 = icmp sgt i32 %.val.i35, 0
+  br i1 %5, label %6, label %8, !prof !14
 
-9:                                                ; preds = %7
+9:                                                ; preds = %lean_inc.exit24
   %10 = add nuw i32 %.val.i35, 1
   store i32 %10, ptr %0, align 4, !tbaa !10
   br label %lean_inc.exit
 
-11:                                               ; preds = %7
+11:                                               ; preds = %lean_inc.exit24
   %.not.i36 = icmp eq i32 %.val.i35, 0
   br i1 %.not.i36, label %lean_inc.exit, label %12
 
@@ -797,7 +791,7 @@ lean_inc.exit24:
   tail call void @lean_inc_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_inc.exit
 
-lean_inc.exit:                                    ; preds = %12, %11, %9, %lean_inc.exit24
+lean_inc.exit:                                    ; preds = %12, %11, %9
   tail call void @lean_inc_heartbeat() #3
   %13 = tail call noalias ptr @mi_malloc_small(i64 noundef 32) #3
   %14 = icmp eq ptr %13, null
@@ -818,30 +812,24 @@ lean_alloc_ctor.exit:                             ; preds = %lean_inc.exit
   %19 = getelementptr inbounds nuw i8, ptr %13, i64 24
   store ptr %4, ptr %19, align 8, !tbaa !4
   %20 = tail call ptr @l_Substring_nextn(ptr noundef nonnull %13, ptr noundef nonnull inttoptr (i64 3 to ptr), ptr noundef nonnull inttoptr (i64 1 to ptr)) #3
-  %21 = ptrtoint ptr %13 to i64
-  %22 = and i64 %21, 1
-  %.not39 = icmp eq i64 %22, 0
-  br i1 %.not39, label %23, label %lean_dec.exit27
-
-23:                                               ; preds = %lean_alloc_ctor.exit
-  %24 = load i32, ptr %13, align 4, !tbaa !10
-  %25 = icmp sgt i32 %24, 1
-  br i1 %25, label %26, label %28, !prof !14
-
-26:                                               ; preds = %23
-  %27 = add nsw i32 %24, -1
-  store i32 %27, ptr %13, align 4, !tbaa !10
+  %18 = load i32, ptr %13, align 8, !tbaa !10
+  %19 = icmp sgt i32 %18, 1
+  br i1 %19, label %20, label %22, !prof !14
+
+20:                                               ; preds = %lean_alloc_ctor.exit
+  %21 = add nsw i32 %18, -1
+  store i32 %21, ptr %13, align 4, !tbaa !10
   br label %lean_dec.exit27
 
-28:                                               ; preds = %23
-  %.not.i = icmp eq i32 %24, 0
+28:                                               ; preds = %lean_alloc_ctor.exit
+  %.not.i = icmp eq i32 %18, 0
   br i1 %.not.i, label %lean_dec.exit27, label %29
 
 29:                                               ; preds = %28
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %13) #3
   br label %lean_dec.exit27
 
-lean_dec.exit27:                                  ; preds = %lean_alloc_ctor.exit, %26, %28, %29
+lean_dec.exit27:                                  ; preds = %20, %28, %29
   %30 = ptrtoint ptr %20 to i64
   %31 = and i64 %30, 1
   %.not40 = icmp eq i64 %31, 0
@@ -867,29 +855,26 @@ lean_dec.exit27:                                  ; preds = %lean_alloc_ctor.exi
   br label %lean_dec.exit25
 
 lean_dec.exit25:                                  ; preds = %lean_dec.exit27, %36, %38, %39
-  %.0.i48 = phi ptr [ %33, %36 ], [ %33, %38 ], [ %33, %39 ], [ %20, %lean_dec.exit27 ]
+  %.0.i48 = phi ptr [ %33, %30 ], [ %33, %32 ], [ %33, %33 ], [ %20, %lean_dec.exit27 ]
   %40 = tail call zeroext i8 @l_String_anyAux___at___private_Lean_Elab_AutoBound_0__Lean_Elab_isValidAutoBoundSuffix___spec__1(ptr noundef nonnull %0, ptr noundef nonnull %4, ptr noundef %.0.i48)
-  br i1 %.not, label %41, label %lean_dec.exit
-
-41:                                               ; preds = %lean_dec.exit25
-  %42 = load i32, ptr %0, align 4, !tbaa !10
-  %43 = icmp sgt i32 %42, 1
-  br i1 %43, label %44, label %46, !prof !14
+  %35 = load i32, ptr %0, align 4, !tbaa !10
+  %36 = icmp sgt i32 %35, 1
+  br i1 %36, label %37, label %312, !prof !14
 
-44:                                               ; preds = %41
-  %45 = add nsw i32 %42, -1
+44:                                               ; preds = %lean_dec.exit25
+  %45 = add nsw i32 %35, -1
   store i32 %45, ptr %0, align 4, !tbaa !10
   br label %lean_dec.exit
 
-46:                                               ; preds = %41
-  %.not.i32 = icmp eq i32 %42, 0
+46:                                               ; preds = %lean_dec.exit25
+  %.not.i32 = icmp eq i32 %35, 0
   br i1 %.not.i32, label %lean_dec.exit, label %47
 
 47:                                               ; preds = %46
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %47, %46, %44, %lean_dec.exit25
+lean_dec.exit:                                    ; preds = %47, %46, %44
   %. = xor i8 %40, 1
   ret i8 %.
 }
diff --git a/bench/lean4/optimized/AuxDeclCache.ll b/bench/lean4/optimized/AuxDeclCache.ll
index 2e69ceffa0a..55d2d06dc92 100644
--- a/bench/lean4/optimized/AuxDeclCache.ll
+++ b/bench/lean4/optimized/AuxDeclCache.ll
@@ -5145,30 +5145,24 @@ lean_dec.exit:                                    ; preds = %24, %23, %21, %lean
 define ptr @l_Lean_PersistentHashMap_findAux___at_Lean_Compiler_LCNF_initFn____x40_Lean_Compiler_LCNF_AuxDeclCache___hyg_3____spec__3___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) local_unnamed_addr #1 {
   %4 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %4, align 8, !tbaa !12
-  %5 = ptrtoint ptr %1 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit
+  %5 = load i32, ptr %1, align 8, !tbaa !4
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !11
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %1, align 4, !tbaa !4
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !11
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %1, align 4, !tbaa !4
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %13, %12, %10, %3
+lean_dec.exit:                                    ; preds = %13, %12, %7
   %14 = tail call ptr @l_Lean_PersistentHashMap_findAux___at_Lean_Compiler_LCNF_initFn____x40_Lean_Compiler_LCNF_AuxDeclCache___hyg_3____spec__3(ptr noundef %0, i64 noundef %.val, ptr noundef %2)
   ret ptr %14
 }
@@ -5177,30 +5171,24 @@ lean_dec.exit:                                    ; preds = %13, %12, %10, %3
 define ptr @l_Lean_PersistentHashMap_insertAux_traverse___at_Lean_Compiler_LCNF_initFn____x40_Lean_Compiler_LCNF_AuxDeclCache___hyg_3____spec__7___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef readnone captures(none) %3, ptr noundef %4, ptr noundef %5) local_unnamed_addr #1 {
   %7 = getelementptr i8, ptr %0, i64 8
   %.val = load i64, ptr %7, align 8, !tbaa !12
-  %8 = ptrtoint ptr %0 to i64
-  %9 = and i64 %8, 1
-  %.not = icmp eq i64 %9, 0
-  br i1 %.not, label %10, label %lean_dec.exit11
+  %8 = load i32, ptr %0, align 8, !tbaa !4
+  %9 = icmp sgt i32 %8, 1
+  br i1 %9, label %10, label %12, !prof !11
 
 10:                                               ; preds = %6
-  %11 = load i32, ptr %0, align 4, !tbaa !4
-  %12 = icmp sgt i32 %11, 1
-  br i1 %12, label %13, label %15, !prof !11
-
-13:                                               ; preds = %10
-  %14 = add nsw i32 %11, -1
-  store i32 %14, ptr %0, align 4, !tbaa !4
+  %11 = add nsw i32 %8, -1
+  store i32 %11, ptr %0, align 4, !tbaa !4
   br label %lean_dec.exit11
 
-15:                                               ; preds = %10
-  %.not.i = icmp eq i32 %11, 0
+15:                                               ; preds = %6
+  %.not.i = icmp eq i32 %8, 0
   br i1 %.not.i, label %lean_dec.exit11, label %16
 
 16:                                               ; preds = %15
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #4
   br label %lean_dec.exit11
 
-lean_dec.exit11:                                  ; preds = %16, %15, %13, %6
+lean_dec.exit11:                                  ; preds = %16, %15, %13
   %17 = tail call ptr @l_Lean_PersistentHashMap_insertAux_traverse___at_Lean_Compiler_LCNF_initFn____x40_Lean_Compiler_LCNF_AuxDeclCache___hyg_3____spec__7(i64 noundef %.val, ptr noundef %1, ptr noundef %2, ptr poison, ptr noundef %4, ptr noundef %5)
   %18 = ptrtoint ptr %2 to i64
   %19 = and i64 %18, 1
@@ -5210,7 +5198,7 @@ lean_dec.exit11:                                  ; preds = %16, %15, %13, %6
 20:                                               ; preds = %lean_dec.exit11
   %21 = load i32, ptr %2, align 4, !tbaa !4
   %22 = icmp sgt i32 %21, 1
-  br i1 %22, label %23, label %25, !prof !11
+  br i1 %112, label %23, label %25, !prof !11
 
 23:                                               ; preds = %20
   %24 = add nsw i32 %21, -1
@@ -5234,7 +5222,7 @@ lean_dec.exit10:                                  ; preds = %26, %25, %23, %lean
 29:                                               ; preds = %lean_dec.exit10
   %30 = load i32, ptr %1, align 4, !tbaa !4
   %31 = icmp sgt i32 %30, 1
-  br i1 %31, label %32, label %34, !prof !11
+  br i1 %31, label %212, label %34, !prof !11
 
 32:                                               ; preds = %29
   %33 = add nsw i32 %30, -1
@@ -5257,56 +5245,44 @@ lean_dec.exit:                                    ; preds = %35, %34, %32, %lean
 define ptr @l_Lean_PersistentHashMap_insertAux___at_Lean_Compiler_LCNF_initFn____x40_Lean_Compiler_LCNF_AuxDeclCache___hyg_3____spec__6___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4) local_unnamed_addr #1 {
   %6 = getelementptr i8, ptr %1, i64 8
   %.val12 = load i64, ptr %6, align 8, !tbaa !12
-  %7 = ptrtoint ptr %1 to i64
-  %8 = and i64 %7, 1
-  %.not = icmp eq i64 %8, 0
-  br i1 %.not, label %9, label %lean_dec.exit9
+  %7 = load i32, ptr %1, align 8, !tbaa !4
+  %8 = icmp sgt i32 %7, 1
+  br i1 %8, label %9, label %11, !prof !11
 
 9:                                                ; preds = %5
-  %10 = load i32, ptr %1, align 4, !tbaa !4
-  %11 = icmp sgt i32 %10, 1
-  br i1 %11, label %12, label %14, !prof !11
-
-12:                                               ; preds = %9
-  %13 = add nsw i32 %10, -1
-  store i32 %13, ptr %1, align 4, !tbaa !4
+  %10 = add nsw i32 %7, -1
+  store i32 %10, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit9
 
-14:                                               ; preds = %9
-  %.not.i = icmp eq i32 %10, 0
+14:                                               ; preds = %5
+  %.not.i = icmp eq i32 %7, 0
   br i1 %.not.i, label %lean_dec.exit9, label %15
 
 15:                                               ; preds = %14
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit9
 
-lean_dec.exit9:                                   ; preds = %15, %14, %12, %5
+lean_dec.exit9:                                   ; preds = %15, %14, %12
   %16 = getelementptr i8, ptr %2, i64 8
   %.val = load i64, ptr %16, align 8, !tbaa !12
-  %17 = ptrtoint ptr %2 to i64
-  %18 = and i64 %17, 1
-  %.not13 = icmp eq i64 %18, 0
-  br i1 %.not13, label %19, label %lean_dec.exit
+  %17 = load i32, ptr %2, align 8, !tbaa !4
+  %18 = icmp sgt i32 %17, 1
+  br i1 %15, label %16, label %18, !prof !11
 
 19:                                               ; preds = %lean_dec.exit9
-  %20 = load i32, ptr %2, align 4, !tbaa !4
-  %21 = icmp sgt i32 %20, 1
-  br i1 %21, label %22, label %24, !prof !11
-
-22:                                               ; preds = %19
-  %23 = add nsw i32 %20, -1
-  store i32 %23, ptr %2, align 4, !tbaa !4
+  %17 = add nsw i32 %14, -1
+  store i32 %20, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit
 
-24:                                               ; preds = %19
-  %.not.i10 = icmp eq i32 %20, 0
-  br i1 %.not.i10, label %lean_dec.exit, label %25
+24:                                               ; preds = %lean_dec.exit9
+  %.not.i10 = icmp eq i32 %14, 0
+  br i1 %.not.i10, label %lean_dec.exit, label %112
 
-25:                                               ; preds = %24
+112:                                               ; preds = %24
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #4
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %25, %24, %22, %lean_dec.exit9
+lean_dec.exit:                                    ; preds = %25, %24, %22
   %26 = tail call ptr @l_Lean_PersistentHashMap_insertAux___at_Lean_Compiler_LCNF_initFn____x40_Lean_Compiler_LCNF_AuxDeclCache___hyg_3____spec__6(ptr noundef %0, i64 noundef %.val12, i64 noundef %.val, ptr noundef %3, ptr noundef %4)
   ret ptr %26
 }
diff --git a/bench/lean4/optimized/AuxLemma.ll b/bench/lean4/optimized/AuxLemma.ll
index e0008246198..fc9b9d4861d 100644
--- a/bench/lean4/optimized/AuxLemma.ll
+++ b/bench/lean4/optimized/AuxLemma.ll
@@ -9989,30 +9989,24 @@ lean_dec.exit:                                    ; preds = %43, %42, %40, %lean
 define ptr @l_Lean_PersistentHashMap_insertAux_traverse___at_Lean_Meta_mkAuxLemma___spec__4___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef readnone captures(none) %3, ptr noundef %4, ptr noundef %5) local_unnamed_addr #1 {
   %7 = getelementptr i8, ptr %0, i64 8
   %.val = load i64, ptr %7, align 8, !tbaa !12
-  %8 = ptrtoint ptr %0 to i64
-  %9 = and i64 %8, 1
-  %.not = icmp eq i64 %9, 0
-  br i1 %.not, label %10, label %lean_dec.exit11
+  %8 = load i32, ptr %0, align 8, !tbaa !4
+  %9 = icmp sgt i32 %8, 1
+  br i1 %9, label %10, label %12, !prof !11
 
 10:                                               ; preds = %6
-  %11 = load i32, ptr %0, align 4, !tbaa !4
-  %12 = icmp sgt i32 %11, 1
-  br i1 %12, label %13, label %15, !prof !11
-
-13:                                               ; preds = %10
-  %14 = add nsw i32 %11, -1
-  store i32 %14, ptr %0, align 4, !tbaa !4
+  %11 = add nsw i32 %8, -1
+  store i32 %11, ptr %0, align 4, !tbaa !4
   br label %lean_dec.exit11
 
-15:                                               ; preds = %10
-  %.not.i = icmp eq i32 %11, 0
+15:                                               ; preds = %6
+  %.not.i = icmp eq i32 %8, 0
   br i1 %.not.i, label %lean_dec.exit11, label %16
 
 16:                                               ; preds = %15
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #4
   br label %lean_dec.exit11
 
-lean_dec.exit11:                                  ; preds = %16, %15, %13, %6
+lean_dec.exit11:                                  ; preds = %16, %15, %13
   %17 = tail call ptr @l_Lean_PersistentHashMap_insertAux_traverse___at_Lean_Meta_mkAuxLemma___spec__4(i64 noundef %.val, ptr noundef %1, ptr noundef %2, ptr poison, ptr noundef %4, ptr noundef %5)
   %18 = ptrtoint ptr %2 to i64
   %19 = and i64 %18, 1
@@ -10069,56 +10063,44 @@ lean_dec.exit:                                    ; preds = %35, %34, %32, %lean
 define ptr @l_Lean_PersistentHashMap_insertAux___at_Lean_Meta_mkAuxLemma___spec__3___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4) local_unnamed_addr #1 {
   %6 = getelementptr i8, ptr %1, i64 8
   %.val12 = load i64, ptr %6, align 8, !tbaa !12
-  %7 = ptrtoint ptr %1 to i64
-  %8 = and i64 %7, 1
-  %.not = icmp eq i64 %8, 0
-  br i1 %.not, label %9, label %lean_dec.exit9
+  %7 = load i32, ptr %1, align 8, !tbaa !4
+  %8 = icmp sgt i32 %7, 1
+  br i1 %8, label %9, label %11, !prof !11
 
 9:                                                ; preds = %5
-  %10 = load i32, ptr %1, align 4, !tbaa !4
-  %11 = icmp sgt i32 %10, 1
-  br i1 %11, label %12, label %14, !prof !11
-
-12:                                               ; preds = %9
-  %13 = add nsw i32 %10, -1
-  store i32 %13, ptr %1, align 4, !tbaa !4
+  %10 = add nsw i32 %7, -1
+  store i32 %10, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit9
 
-14:                                               ; preds = %9
-  %.not.i = icmp eq i32 %10, 0
+14:                                               ; preds = %5
+  %.not.i = icmp eq i32 %7, 0
   br i1 %.not.i, label %lean_dec.exit9, label %15
 
 15:                                               ; preds = %14
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit9
 
-lean_dec.exit9:                                   ; preds = %15, %14, %12, %5
+lean_dec.exit9:                                   ; preds = %15, %14, %12
   %16 = getelementptr i8, ptr %2, i64 8
   %.val = load i64, ptr %16, align 8, !tbaa !12
-  %17 = ptrtoint ptr %2 to i64
-  %18 = and i64 %17, 1
-  %.not13 = icmp eq i64 %18, 0
-  br i1 %.not13, label %19, label %lean_dec.exit
-
-19:                                               ; preds = %lean_dec.exit9
-  %20 = load i32, ptr %2, align 4, !tbaa !4
-  %21 = icmp sgt i32 %20, 1
-  br i1 %21, label %22, label %24, !prof !11
+  %17 = load i32, ptr %2, align 8, !tbaa !4
+  %18 = icmp sgt i32 %17, 1
+  br i1 %15, label %16, label %18, !prof !11
 
-22:                                               ; preds = %19
-  %23 = add nsw i32 %20, -1
+22:                                               ; preds = %lean_dec.exit9
+  %23 = add nsw i32 %14, -1
   store i32 %23, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit
 
-24:                                               ; preds = %19
-  %.not.i10 = icmp eq i32 %20, 0
+24:                                               ; preds = %lean_dec.exit9
+  %.not.i10 = icmp eq i32 %14, 0
   br i1 %.not.i10, label %lean_dec.exit, label %25
 
 25:                                               ; preds = %24
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #4
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %25, %24, %22, %lean_dec.exit9
+lean_dec.exit:                                    ; preds = %25, %24, %22
   %26 = tail call ptr @l_Lean_PersistentHashMap_insertAux___at_Lean_Meta_mkAuxLemma___spec__3(ptr noundef %0, i64 noundef %.val12, i64 noundef %.val, ptr noundef %3, ptr noundef %4)
   ret ptr %26
 }
@@ -10205,30 +10187,24 @@ lean_dec.exit:                                    ; preds = %33, %32, %30, %lean
 define ptr @l_Lean_PersistentHashMap_findAux___at_Lean_Meta_mkAuxLemma___spec__7___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) local_unnamed_addr #1 {
   %4 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %4, align 8, !tbaa !12
-  %5 = ptrtoint ptr %1 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit6
+  %5 = load i32, ptr %1, align 8, !tbaa !4
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !11
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %1, align 4, !tbaa !4
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !11
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %1, align 4, !tbaa !4
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit6
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit6, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit6
 
-lean_dec.exit6:                                   ; preds = %13, %12, %10, %3
+lean_dec.exit6:                                   ; preds = %13, %12, %7
   %14 = tail call ptr @l_Lean_PersistentHashMap_findAux___at_Lean_Meta_mkAuxLemma___spec__7(ptr noundef %0, i64 noundef %.val, ptr noundef %2)
   %15 = ptrtoint ptr %2 to i64
   %16 = and i64 %15, 1
diff --git a/bench/lean4/optimized/ByteArray.ll b/bench/lean4/optimized/ByteArray.ll
index 013251bd4f8..398f749d654 100644
--- a/bench/lean4/optimized/ByteArray.ll
+++ b/bench/lean4/optimized/ByteArray.ll
@@ -729,22 +729,16 @@ define nonnull ptr @l_Std_Internal_Parsec_ByteArray_instInputIteratorUInt8Nat___
   %9 = getelementptr inbounds nuw i8, ptr %4, i64 24
   %10 = getelementptr inbounds nuw i8, ptr %9, i64 %8
   %11 = load i8, ptr %10, align 1, !tbaa !15
-  %12 = ptrtoint ptr %0 to i64
-  %13 = and i64 %12, 1
-  %.not = icmp eq i64 %13, 0
-  br i1 %.not, label %14, label %lean_dec.exit
+  %12 = load i32, ptr %0, align 8, !tbaa !8
+  %13 = icmp sgt i32 %12, 1
+  br i1 %13, label %14, label %16, !prof !11
 
 14:                                               ; preds = %2
-  %15 = load i32, ptr %0, align 4, !tbaa !8
-  %16 = icmp sgt i32 %15, 1
-  br i1 %16, label %17, label %19, !prof !11
-
-17:                                               ; preds = %14
-  %18 = add nsw i32 %15, -1
-  store i32 %18, ptr %0, align 4, !tbaa !8
+  %15 = add nsw i32 %12, -1
+  store i32 %15, ptr %0, align 4, !tbaa !8
   br label %lean_dec.exit
 
-19:                                               ; preds = %14
+19:                                               ; preds = %2
   %.not.i = icmp eq i32 %15, 0
   br i1 %.not.i, label %lean_dec.exit, label %20
 
@@ -752,7 +746,7 @@ define nonnull ptr @l_Std_Internal_Parsec_ByteArray_instInputIteratorUInt8Nat___
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #6
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %20, %19, %17, %2
+lean_dec.exit:                                    ; preds = %20, %19, %14
   %21 = zext i8 %11 to i64
   %22 = shl nuw nsw i64 %21, 1
   %23 = or disjoint i64 %22, 1
@@ -2473,56 +2467,44 @@ lean_alloc_ctor.exit54:                           ; preds = %lean_dec.exit
 define noalias noundef nonnull ptr @l_ByteArray_forInUnsafe_loop___at_Std_Internal_Parsec_ByteArray_skipBytes___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4) local_unnamed_addr #0 {
   %6 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %6, align 8, !tbaa !13
-  %7 = ptrtoint ptr %1 to i64
-  %8 = and i64 %7, 1
-  %.not = icmp eq i64 %8, 0
-  br i1 %.not, label %9, label %lean_dec.exit11
+  %7 = load i32, ptr %1, align 8, !tbaa !8
+  %8 = icmp sgt i32 %7, 1
+  br i1 %8, label %9, label %11, !prof !11
 
 9:                                                ; preds = %5
-  %10 = load i32, ptr %1, align 4, !tbaa !8
-  %11 = icmp sgt i32 %10, 1
-  br i1 %11, label %12, label %14, !prof !11
-
-12:                                               ; preds = %9
-  %13 = add nsw i32 %10, -1
-  store i32 %13, ptr %1, align 4, !tbaa !8
+  %10 = add nsw i32 %7, -1
+  store i32 %10, ptr %1, align 4, !tbaa !8
   br label %lean_dec.exit11
 
-14:                                               ; preds = %9
-  %.not.i = icmp eq i32 %10, 0
+14:                                               ; preds = %5
+  %.not.i = icmp eq i32 %7, 0
   br i1 %.not.i, label %lean_dec.exit11, label %15
 
 15:                                               ; preds = %14
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #6
   br label %lean_dec.exit11
 
-lean_dec.exit11:                                  ; preds = %15, %14, %12, %5
+lean_dec.exit11:                                  ; preds = %15, %14, %9
   %16 = getelementptr i8, ptr %2, i64 8
   %.val16 = load i64, ptr %16, align 8, !tbaa !13
-  %17 = ptrtoint ptr %2 to i64
-  %18 = and i64 %17, 1
-  %.not17 = icmp eq i64 %18, 0
-  br i1 %.not17, label %19, label %lean_dec.exit10
-
-19:                                               ; preds = %lean_dec.exit11
-  %20 = load i32, ptr %2, align 4, !tbaa !8
-  %21 = icmp sgt i32 %20, 1
-  br i1 %21, label %22, label %24, !prof !11
+  %14 = load i32, ptr %2, align 8, !tbaa !8
+  %15 = icmp sgt i32 %14, 1
+  br i1 %15, label %16, label %18, !prof !11
 
-22:                                               ; preds = %19
-  %23 = add nsw i32 %20, -1
+22:                                               ; preds = %lean_dec.exit11
+  %23 = add nsw i32 %14, -1
   store i32 %23, ptr %2, align 4, !tbaa !8
   br label %lean_dec.exit10
 
-24:                                               ; preds = %19
-  %.not.i12 = icmp eq i32 %20, 0
-  br i1 %.not.i12, label %lean_dec.exit10, label %25
+24:                                               ; preds = %lean_dec.exit11
+  %.not.i12 = icmp eq i32 %14, 0
+  br i1 %.not.i12, label %lean_dec.exit10, label %22
 
 25:                                               ; preds = %24
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #6
   br label %lean_dec.exit10
 
-lean_dec.exit10:                                  ; preds = %25, %24, %22, %lean_dec.exit11
+lean_dec.exit10:                                  ; preds = %22, %24, %22
   %26 = tail call ptr @l_ByteArray_forInUnsafe_loop___at_Std_Internal_Parsec_ByteArray_skipBytes___spec__1(ptr noundef %0, i64 noundef %.val, i64 noundef %.val16, ptr noundef %3, ptr noundef %4)
   %27 = ptrtoint ptr %0 to i64
   %28 = and i64 %27, 1
diff --git a/bench/lean4/optimized/CSE.ll b/bench/lean4/optimized/CSE.ll
index 9d2ffe7457b..195c4001ab0 100644
--- a/bench/lean4/optimized/CSE.ll
+++ b/bench/lean4/optimized/CSE.ll
@@ -4839,30 +4839,24 @@ lean_alloc_ctor.exit103:                          ; preds = %161
 define ptr @l_Lean_PersistentHashMap_insertAux_traverse___at_Lean_Compiler_LCNF_CSE_addEntry___spec__3___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef readnone captures(none) %3, ptr noundef %4, ptr noundef %5) local_unnamed_addr #1 {
   %7 = getelementptr i8, ptr %0, i64 8
   %.val = load i64, ptr %7, align 8, !tbaa !12
-  %8 = ptrtoint ptr %0 to i64
-  %9 = and i64 %8, 1
-  %.not = icmp eq i64 %9, 0
-  br i1 %.not, label %10, label %lean_dec.exit11
+  %8 = load i32, ptr %0, align 8, !tbaa !4
+  %9 = icmp sgt i32 %8, 1
+  br i1 %9, label %10, label %12, !prof !11
 
 10:                                               ; preds = %6
-  %11 = load i32, ptr %0, align 4, !tbaa !4
-  %12 = icmp sgt i32 %11, 1
-  br i1 %12, label %13, label %15, !prof !11
-
-13:                                               ; preds = %10
-  %14 = add nsw i32 %11, -1
-  store i32 %14, ptr %0, align 4, !tbaa !4
+  %11 = add nsw i32 %8, -1
+  store i32 %11, ptr %0, align 4, !tbaa !4
   br label %lean_dec.exit11
 
-15:                                               ; preds = %10
-  %.not.i = icmp eq i32 %11, 0
+15:                                               ; preds = %6
+  %.not.i = icmp eq i32 %8, 0
   br i1 %.not.i, label %lean_dec.exit11, label %16
 
 16:                                               ; preds = %15
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #4
   br label %lean_dec.exit11
 
-lean_dec.exit11:                                  ; preds = %16, %15, %13, %6
+lean_dec.exit11:                                  ; preds = %16, %15, %13
   %17 = tail call ptr @l_Lean_PersistentHashMap_insertAux_traverse___at_Lean_Compiler_LCNF_CSE_addEntry___spec__3(i64 noundef %.val, ptr noundef %1, ptr noundef %2, ptr poison, ptr noundef %4, ptr noundef %5)
   %18 = ptrtoint ptr %2 to i64
   %19 = and i64 %18, 1
@@ -4919,56 +4913,44 @@ lean_dec.exit:                                    ; preds = %35, %34, %32, %lean
 define ptr @l_Lean_PersistentHashMap_insertAux___at_Lean_Compiler_LCNF_CSE_addEntry___spec__2___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4) local_unnamed_addr #1 {
   %6 = getelementptr i8, ptr %1, i64 8
   %.val12 = load i64, ptr %6, align 8, !tbaa !12
-  %7 = ptrtoint ptr %1 to i64
-  %8 = and i64 %7, 1
-  %.not = icmp eq i64 %8, 0
-  br i1 %.not, label %9, label %lean_dec.exit9
+  %7 = load i32, ptr %1, align 8, !tbaa !4
+  %8 = icmp sgt i32 %7, 1
+  br i1 %8, label %9, label %11, !prof !11
 
 9:                                                ; preds = %5
-  %10 = load i32, ptr %1, align 4, !tbaa !4
-  %11 = icmp sgt i32 %10, 1
-  br i1 %11, label %12, label %14, !prof !11
-
-12:                                               ; preds = %9
-  %13 = add nsw i32 %10, -1
-  store i32 %13, ptr %1, align 4, !tbaa !4
+  %10 = add nsw i32 %7, -1
+  store i32 %10, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit9
 
-14:                                               ; preds = %9
-  %.not.i = icmp eq i32 %10, 0
+14:                                               ; preds = %5
+  %.not.i = icmp eq i32 %7, 0
   br i1 %.not.i, label %lean_dec.exit9, label %15
 
 15:                                               ; preds = %14
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit9
 
-lean_dec.exit9:                                   ; preds = %15, %14, %12, %5
+lean_dec.exit9:                                   ; preds = %15, %14, %12
   %16 = getelementptr i8, ptr %2, i64 8
   %.val = load i64, ptr %16, align 8, !tbaa !12
-  %17 = ptrtoint ptr %2 to i64
-  %18 = and i64 %17, 1
-  %.not13 = icmp eq i64 %18, 0
-  br i1 %.not13, label %19, label %lean_dec.exit
-
-19:                                               ; preds = %lean_dec.exit9
-  %20 = load i32, ptr %2, align 4, !tbaa !4
-  %21 = icmp sgt i32 %20, 1
-  br i1 %21, label %22, label %24, !prof !11
+  %17 = load i32, ptr %2, align 8, !tbaa !4
+  %18 = icmp sgt i32 %17, 1
+  br i1 %15, label %16, label %18, !prof !11
 
-22:                                               ; preds = %19
-  %23 = add nsw i32 %20, -1
+22:                                               ; preds = %lean_dec.exit9
+  %23 = add nsw i32 %14, -1
   store i32 %23, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit
 
-24:                                               ; preds = %19
-  %.not.i10 = icmp eq i32 %20, 0
+24:                                               ; preds = %lean_dec.exit9
+  %.not.i10 = icmp eq i32 %14, 0
   br i1 %.not.i10, label %lean_dec.exit, label %25
 
 25:                                               ; preds = %24
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #4
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %25, %24, %22, %lean_dec.exit9
+lean_dec.exit:                                    ; preds = %25, %24, %22
   %26 = tail call ptr @l_Lean_PersistentHashMap_insertAux___at_Lean_Compiler_LCNF_CSE_addEntry___spec__2(ptr noundef %0, i64 noundef %.val12, i64 noundef %.val, ptr noundef %3, ptr noundef %4)
   ret ptr %26
 }
@@ -35747,30 +35729,24 @@ lean_dec.exit:                                    ; preds = %33, %32, %30, %lean
 define ptr @l_Lean_PersistentHashMap_findAux___at_Lean_Compiler_LCNF_Code_cse_go___spec__3___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) local_unnamed_addr #1 {
   %4 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %4, align 8, !tbaa !12
-  %5 = ptrtoint ptr %1 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit6
+  %5 = load i32, ptr %1, align 8, !tbaa !4
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !11
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %1, align 4, !tbaa !4
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !11
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %1, align 4, !tbaa !4
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit6
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit6, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit6
 
-lean_dec.exit6:                                   ; preds = %13, %12, %10, %3
+lean_dec.exit6:                                   ; preds = %13, %12, %7
   %14 = tail call ptr @l_Lean_PersistentHashMap_findAux___at_Lean_Compiler_LCNF_Code_cse_go___spec__3(ptr noundef %0, i64 noundef %.val, ptr noundef %2)
   %15 = ptrtoint ptr %2 to i64
   %16 = and i64 %15, 1
diff --git a/bench/lean4/optimized/Calc.ll b/bench/lean4/optimized/Calc.ll
index 51de904d7f5..e9a27ebe05e 100644
--- a/bench/lean4/optimized/Calc.ll
+++ b/bench/lean4/optimized/Calc.ll
@@ -3943,18 +3943,9 @@ lean_alloc_closure.exit:                          ; preds = %lean_alloc_ctor.exi
   store i16 1, ptr %23, align 2, !tbaa !15
   %24 = getelementptr inbounds nuw i8, ptr %17, i64 24
   store ptr %1, ptr %24, align 8, !tbaa !4
-  %25 = ptrtoint ptr %17 to i64
-  %26 = and i64 %25, 1
-  %.not = icmp eq i64 %26, 0
-  br i1 %.not, label %lean_inc_ref.exit, label %lean_inc.exit
-
-lean_inc_ref.exit:                                ; preds = %lean_alloc_closure.exit
-  store i32 2, ptr %17, align 4, !tbaa !8
-  br label %lean_inc.exit
-
-lean_inc.exit:                                    ; preds = %lean_alloc_closure.exit, %lean_inc_ref.exit
-  %27 = tail call ptr @l_Lean_Elab_Term_ensureHasTypeWithErrorMsgs(ptr noundef nonnull %12, ptr noundef %2, ptr noundef nonnull %17, ptr noundef nonnull %17, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9, ptr noundef %10) #4
-  ret ptr %27
+  store i32 2, ptr %17, align 8, !tbaa !8
+  %25 = tail call ptr @l_Lean_Elab_Term_ensureHasTypeWithErrorMsgs(ptr noundef nonnull %12, ptr noundef %2, ptr noundef nonnull %17, ptr noundef nonnull %17, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9, ptr noundef %10) #4
+  ret ptr %25
 }
 
 ; Function Attrs: nounwind uwtable
diff --git a/bench/lean4/optimized/CaseValues.ll b/bench/lean4/optimized/CaseValues.ll
index c1233cec6b6..fcbc3ecff1a 100644
--- a/bench/lean4/optimized/CaseValues.ll
+++ b/bench/lean4/optimized/CaseValues.ll
@@ -19209,56 +19209,44 @@ declare ptr @l_Lean_Meta_throwTacticEx___rarg(ptr noundef, ptr noundef, ptr noun
 define ptr @l_Array_foldlMUnsafe_fold___at_Lean_Meta_caseValues_loop___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9) local_unnamed_addr #0 {
   %11 = getelementptr i8, ptr %2, i64 8
   %.val = load i64, ptr %11, align 8, !tbaa !16
-  %12 = ptrtoint ptr %2 to i64
-  %13 = and i64 %12, 1
-  %.not = icmp eq i64 %13, 0
-  br i1 %.not, label %14, label %lean_dec.exit18
+  %12 = load i32, ptr %2, align 8, !tbaa !4
+  %13 = icmp sgt i32 %12, 1
+  br i1 %13, label %14, label %16, !prof !11
 
 14:                                               ; preds = %10
-  %15 = load i32, ptr %2, align 4, !tbaa !4
-  %16 = icmp sgt i32 %15, 1
-  br i1 %16, label %17, label %19, !prof !11
-
-17:                                               ; preds = %14
-  %18 = add nsw i32 %15, -1
-  store i32 %18, ptr %2, align 4, !tbaa !4
+  %15 = add nsw i32 %12, -1
+  store i32 %15, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit18
 
-19:                                               ; preds = %14
-  %.not.i = icmp eq i32 %15, 0
+19:                                               ; preds = %10
+  %.not.i = icmp eq i32 %12, 0
   br i1 %.not.i, label %lean_dec.exit18, label %20
 
 20:                                               ; preds = %19
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #5
   br label %lean_dec.exit18
 
-lean_dec.exit18:                                  ; preds = %20, %19, %17, %10
+lean_dec.exit18:                                  ; preds = %20, %19, %14
   %21 = getelementptr i8, ptr %3, i64 8
   %.val25 = load i64, ptr %21, align 8, !tbaa !16
-  %22 = ptrtoint ptr %3 to i64
-  %23 = and i64 %22, 1
-  %.not26 = icmp eq i64 %23, 0
-  br i1 %.not26, label %24, label %lean_dec.exit17
-
-24:                                               ; preds = %lean_dec.exit18
-  %25 = load i32, ptr %3, align 4, !tbaa !4
-  %26 = icmp sgt i32 %25, 1
-  br i1 %26, label %27, label %29, !prof !11
-
-27:                                               ; preds = %24
-  %28 = add nsw i32 %25, -1
-  store i32 %28, ptr %3, align 4, !tbaa !4
+  %19 = load i32, ptr %3, align 8, !tbaa !4
+  %20 = icmp sgt i32 %19, 1
+  br i1 %20, label %21, label %23, !prof !11
+
+21:                                               ; preds = %lean_dec.exit18
+  %22 = add nsw i32 %19, -1
+  store i32 %22, ptr %3, align 4, !tbaa !4
   br label %lean_dec.exit17
 
-29:                                               ; preds = %24
-  %.not.i19 = icmp eq i32 %25, 0
+29:                                               ; preds = %lean_dec.exit18
+  %.not.i19 = icmp eq i32 %19, 0
   br i1 %.not.i19, label %lean_dec.exit17, label %30
 
 30:                                               ; preds = %29
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %3) #5
   br label %lean_dec.exit17
 
-lean_dec.exit17:                                  ; preds = %30, %29, %27, %lean_dec.exit18
+lean_dec.exit17:                                  ; preds = %30, %29, %21
   %31 = tail call ptr @l_Array_foldlMUnsafe_fold___at_Lean_Meta_caseValues_loop___spec__1(ptr noundef %0, ptr noundef %1, i64 noundef %.val, i64 noundef %.val25, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9)
   %32 = ptrtoint ptr %1 to i64
   %33 = and i64 %32, 1
diff --git a/bench/lean4/optimized/CheckTactic.ll b/bench/lean4/optimized/CheckTactic.ll
index d29a08ca3e8..03a99bfcad3 100644
--- a/bench/lean4/optimized/CheckTactic.ll
+++ b/bench/lean4/optimized/CheckTactic.ll
@@ -37531,29 +37531,20 @@ lean_alloc_ctor.exit104:                          ; preds = %lean_inc.exit75
   %108 = getelementptr inbounds nuw i8, ptr %102, i64 24
   store ptr %95, ptr %108, align 8, !tbaa !4
   %109 = load ptr, ptr @l_Lean_Elab_CheckTactic_expandCheckSimp___closed__10, align 8, !tbaa !4
-  %110 = ptrtoint ptr %102 to i64
-  %111 = and i64 %110, 1
-  %.not117 = icmp eq i64 %111, 0
-  br i1 %.not117, label %lean_inc_ref.exit107, label %lean_inc.exit74
+  store i32 2, ptr %102, align 8, !tbaa !8
+  br i1 %.not116, label %110, label %lean_inc.exit73
 
 lean_inc_ref.exit107:                             ; preds = %lean_alloc_ctor.exit104
-  store i32 2, ptr %102, align 4, !tbaa !8
-  br label %lean_inc.exit74
-
-lean_inc.exit74:                                  ; preds = %lean_alloc_ctor.exit104, %lean_inc_ref.exit107
-  br i1 %.not116, label %112, label %lean_inc.exit73
-
-112:                                              ; preds = %lean_inc.exit74
   %.val.i108 = load i32, ptr %39, align 4, !tbaa !8
   %113 = icmp sgt i32 %.val.i108, 0
   br i1 %113, label %114, label %116, !prof !15
 
-114:                                              ; preds = %112
+114:                                              ; preds = %110
   %115 = add nuw i32 %.val.i108, 1
   store i32 %115, ptr %39, align 4, !tbaa !8
   br label %lean_inc.exit73
 
-116:                                              ; preds = %112
+116:                                              ; preds = %110
   %.not.i109 = icmp eq i32 %.val.i108, 0
   br i1 %.not.i109, label %lean_inc.exit73, label %117
 
@@ -37561,22 +37552,19 @@ lean_inc.exit74:                                  ; preds = %lean_alloc_ctor.exi
   tail call void @lean_inc_ref_cold(ptr noundef nonnull %39) #6
   br label %lean_inc.exit73
 
-lean_inc.exit73:                                  ; preds = %117, %116, %114, %lean_inc.exit74
+lean_inc.exit73:                                  ; preds = %117, %116, %114, %lean_alloc_ctor.exit104
   %118 = tail call ptr @l_Lean_Syntax_node1(ptr noundef %39, ptr noundef %109, ptr noundef nonnull %102) #6
   %119 = load ptr, ptr @l_Lean_Elab_CheckTactic_expandCheckSimp___closed__8, align 8, !tbaa !4
-  br i1 %.not117, label %120, label %lean_inc_n.exit
-
-120:                                              ; preds = %lean_inc.exit73
   %.val.i.i = load i32, ptr %102, align 4, !tbaa !8
   %121 = icmp sgt i32 %.val.i.i, 0
   br i1 %121, label %122, label %124, !prof !15
 
-122:                                              ; preds = %120
+122:                                              ; preds = %lean_inc.exit73
   %123 = add nuw i32 %.val.i.i, 3
   store i32 %123, ptr %102, align 4, !tbaa !8
   br label %lean_inc_n.exit
 
-124:                                              ; preds = %120
+124:                                              ; preds = %lean_inc.exit73
   %.not.i.i = icmp eq i32 %.val.i.i, 0
   br i1 %.not.i.i, label %lean_inc_n.exit, label %125
 
@@ -37584,7 +37572,7 @@ lean_inc.exit73:                                  ; preds = %117, %116, %114, %l
   tail call void @lean_inc_ref_n_cold(ptr noundef nonnull %102, i32 noundef 3) #6
   br label %lean_inc_n.exit
 
-lean_inc_n.exit:                                  ; preds = %lean_inc.exit73, %122, %124, %125
+lean_inc_n.exit:                                  ; preds = %122, %124, %125
   br i1 %.not116, label %126, label %lean_inc.exit
 
 126:                                              ; preds = %lean_inc_n.exit
@@ -37941,29 +37929,20 @@ lean_alloc_ctor.exit90:                           ; preds = %lean_inc.exit66
   %95 = getelementptr inbounds nuw i8, ptr %89, i64 24
   store ptr %82, ptr %95, align 8, !tbaa !4
   %96 = load ptr, ptr @l_Lean_Elab_CheckTactic_expandCheckSimp___closed__10, align 8, !tbaa !4
-  %97 = ptrtoint ptr %89 to i64
-  %98 = and i64 %97, 1
-  %.not104 = icmp eq i64 %98, 0
-  br i1 %.not104, label %lean_inc_ref.exit93, label %lean_inc.exit65
+  store i32 2, ptr %89, align 8, !tbaa !8
+  br i1 %.not103, label %97, label %lean_inc.exit64
 
 lean_inc_ref.exit93:                              ; preds = %lean_alloc_ctor.exit90
-  store i32 2, ptr %89, align 4, !tbaa !8
-  br label %lean_inc.exit65
-
-lean_inc.exit65:                                  ; preds = %lean_alloc_ctor.exit90, %lean_inc_ref.exit93
-  br i1 %.not103, label %99, label %lean_inc.exit64
-
-99:                                               ; preds = %lean_inc.exit65
   %.val.i94 = load i32, ptr %39, align 4, !tbaa !8
   %100 = icmp sgt i32 %.val.i94, 0
   br i1 %100, label %101, label %103, !prof !15
 
-101:                                              ; preds = %99
+101:                                              ; preds = %97
   %102 = add nuw i32 %.val.i94, 1
   store i32 %102, ptr %39, align 4, !tbaa !8
   br label %lean_inc.exit64
 
-103:                                              ; preds = %99
+103:                                              ; preds = %97
   %.not.i95 = icmp eq i32 %.val.i94, 0
   br i1 %.not.i95, label %lean_inc.exit64, label %104
 
@@ -37971,22 +37950,19 @@ lean_inc.exit65:                                  ; preds = %lean_alloc_ctor.exi
   tail call void @lean_inc_ref_cold(ptr noundef nonnull %39) #6
   br label %lean_inc.exit64
 
-lean_inc.exit64:                                  ; preds = %104, %103, %101, %lean_inc.exit65
+lean_inc.exit64:                                  ; preds = %104, %103, %101, %lean_alloc_ctor.exit90
   %105 = tail call ptr @l_Lean_Syntax_node1(ptr noundef %39, ptr noundef %96, ptr noundef nonnull %89) #6
   %106 = load ptr, ptr @l_Lean_Elab_CheckTactic_expandCheckSimp___closed__8, align 8, !tbaa !4
-  br i1 %.not104, label %107, label %lean_inc_n.exit
-
-107:                                              ; preds = %lean_inc.exit64
   %.val.i.i = load i32, ptr %89, align 4, !tbaa !8
   %108 = icmp sgt i32 %.val.i.i, 0
   br i1 %108, label %109, label %111, !prof !15
 
-109:                                              ; preds = %107
+109:                                              ; preds = %lean_inc.exit64
   %110 = add nuw i32 %.val.i.i, 3
   store i32 %110, ptr %89, align 4, !tbaa !8
   br label %lean_inc_n.exit
 
-111:                                              ; preds = %107
+111:                                              ; preds = %lean_inc.exit64
   %.not.i.i = icmp eq i32 %.val.i.i, 0
   br i1 %.not.i.i, label %lean_inc_n.exit, label %112
 
@@ -37994,7 +37970,7 @@ lean_inc.exit64:                                  ; preds = %104, %103, %101, %l
   tail call void @lean_inc_ref_n_cold(ptr noundef nonnull %89, i32 noundef 3) #6
   br label %lean_inc_n.exit
 
-lean_inc_n.exit:                                  ; preds = %lean_inc.exit64, %109, %111, %112
+lean_inc_n.exit:                                  ; preds = %109, %111, %112
   br i1 %.not103, label %113, label %lean_inc.exit
 
 113:                                              ; preds = %lean_inc_n.exit
diff --git a/bench/lean4/optimized/Client.ll b/bench/lean4/optimized/Client.ll
index 9e9257836cf..65fb4d1c702 100644
--- a/bench/lean4/optimized/Client.ll
+++ b/bench/lean4/optimized/Client.ll
@@ -1532,56 +1532,44 @@ lean_alloc_ctor.exit18:                           ; preds = %lean_alloc_ctor.exi
 define ptr @l_Array_mapMUnsafe_map___at___private_Lean_Data_Lsp_Client_0__Lean_Lsp_toJsonRegistrationParams____x40_Lean_Data_Lsp_Client___hyg_258____spec__2___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) local_unnamed_addr #0 {
   %4 = getelementptr i8, ptr %0, i64 8
   %.val = load i64, ptr %4, align 8, !tbaa !12
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit7
+  %5 = load i32, ptr %0, align 8, !tbaa !8
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !11
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %0, align 4, !tbaa !8
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !11
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !8
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !8
   br label %lean_dec.exit7
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit7, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #4
   br label %lean_dec.exit7
 
-lean_dec.exit7:                                   ; preds = %13, %12, %10, %3
+lean_dec.exit7:                                   ; preds = %13, %12, %7
   %14 = getelementptr i8, ptr %1, i64 8
   %.val10 = load i64, ptr %14, align 8, !tbaa !12
-  %15 = ptrtoint ptr %1 to i64
-  %16 = and i64 %15, 1
-  %.not11 = icmp eq i64 %16, 0
-  br i1 %.not11, label %17, label %lean_dec.exit
-
-17:                                               ; preds = %lean_dec.exit7
-  %18 = load i32, ptr %1, align 4, !tbaa !8
-  %19 = icmp sgt i32 %18, 1
-  br i1 %19, label %20, label %22, !prof !11
-
-20:                                               ; preds = %17
-  %21 = add nsw i32 %18, -1
-  store i32 %21, ptr %1, align 4, !tbaa !8
+  %12 = load i32, ptr %1, align 8, !tbaa !8
+  %13 = icmp sgt i32 %12, 1
+  br i1 %13, label %14, label %16, !prof !11
+
+14:                                               ; preds = %lean_dec.exit7
+  %15 = add nsw i32 %12, -1
+  store i32 %15, ptr %1, align 4, !tbaa !8
   br label %lean_dec.exit
 
-22:                                               ; preds = %17
-  %.not.i8 = icmp eq i32 %18, 0
+22:                                               ; preds = %lean_dec.exit7
+  %.not.i8 = icmp eq i32 %12, 0
   br i1 %.not.i8, label %lean_dec.exit, label %23
 
 23:                                               ; preds = %22
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %23, %22, %20, %lean_dec.exit7
+lean_dec.exit:                                    ; preds = %23, %22, %14
   %24 = tail call ptr @l_Array_mapMUnsafe_map___at___private_Lean_Data_Lsp_Client_0__Lean_Lsp_toJsonRegistrationParams____x40_Lean_Data_Lsp_Client___hyg_258____spec__2(i64 noundef %.val, i64 noundef %.val10, ptr noundef %2)
   ret ptr %24
 }
@@ -2457,56 +2445,44 @@ lean_alloc_ctor.exit47:                           ; preds = %lean_dec.exit
 define ptr @l_Array_mapMUnsafe_map___at___private_Lean_Data_Lsp_Client_0__Lean_Lsp_fromJsonRegistrationParams____x40_Lean_Data_Lsp_Client___hyg_296____spec__3___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) local_unnamed_addr #0 {
   %4 = getelementptr i8, ptr %0, i64 8
   %.val10 = load i64, ptr %4, align 8, !tbaa !12
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit7
+  %5 = load i32, ptr %0, align 8, !tbaa !8
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !11
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %0, align 4, !tbaa !8
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !11
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !8
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !8
   br label %lean_dec.exit7
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit7, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #4
   br label %lean_dec.exit7
 
-lean_dec.exit7:                                   ; preds = %13, %12, %10, %3
+lean_dec.exit7:                                   ; preds = %13, %12, %7
   %14 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %14, align 8, !tbaa !12
-  %15 = ptrtoint ptr %1 to i64
-  %16 = and i64 %15, 1
-  %.not11 = icmp eq i64 %16, 0
-  br i1 %.not11, label %17, label %lean_dec.exit
-
-17:                                               ; preds = %lean_dec.exit7
-  %18 = load i32, ptr %1, align 4, !tbaa !8
-  %19 = icmp sgt i32 %18, 1
-  br i1 %19, label %20, label %22, !prof !11
-
-20:                                               ; preds = %17
-  %21 = add nsw i32 %18, -1
-  store i32 %21, ptr %1, align 4, !tbaa !8
+  %12 = load i32, ptr %1, align 8, !tbaa !8
+  %13 = icmp sgt i32 %12, 1
+  br i1 %13, label %14, label %16, !prof !11
+
+14:                                               ; preds = %lean_dec.exit7
+  %15 = add nsw i32 %12, -1
+  store i32 %15, ptr %1, align 4, !tbaa !8
   br label %lean_dec.exit
 
-22:                                               ; preds = %17
-  %.not.i8 = icmp eq i32 %18, 0
+22:                                               ; preds = %lean_dec.exit7
+  %.not.i8 = icmp eq i32 %12, 0
   br i1 %.not.i8, label %lean_dec.exit, label %23
 
 23:                                               ; preds = %22
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %23, %22, %20, %lean_dec.exit7
+lean_dec.exit:                                    ; preds = %23, %22, %14
   %24 = tail call ptr @l_Array_mapMUnsafe_map___at___private_Lean_Data_Lsp_Client_0__Lean_Lsp_fromJsonRegistrationParams____x40_Lean_Data_Lsp_Client___hyg_296____spec__3(i64 noundef %.val10, i64 noundef %.val, ptr noundef %2)
   ret ptr %24
 }
diff --git a/bench/lean4/optimized/CollectAxioms.ll b/bench/lean4/optimized/CollectAxioms.ll
index c91471b5800..2872e79faee 100644
--- a/bench/lean4/optimized/CollectAxioms.ll
+++ b/bench/lean4/optimized/CollectAxioms.ll
@@ -6003,56 +6003,44 @@ declare ptr @l_Lean_Expr_getUsedConstants(ptr noundef) local_unnamed_addr #2
 define noalias nonnull ptr @l_Array_foldlMUnsafe_fold___at_Lean_CollectAxioms_collect___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5) local_unnamed_addr #0 {
   %7 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %7, align 8, !tbaa !12
-  %8 = ptrtoint ptr %1 to i64
-  %9 = and i64 %8, 1
-  %.not = icmp eq i64 %9, 0
-  br i1 %.not, label %10, label %lean_dec.exit12
+  %8 = load i32, ptr %1, align 8, !tbaa !4
+  %9 = icmp sgt i32 %8, 1
+  br i1 %9, label %10, label %12, !prof !9
 
 10:                                               ; preds = %6
-  %11 = load i32, ptr %1, align 4, !tbaa !4
-  %12 = icmp sgt i32 %11, 1
-  br i1 %12, label %13, label %15, !prof !9
-
-13:                                               ; preds = %10
-  %14 = add nsw i32 %11, -1
-  store i32 %14, ptr %1, align 4, !tbaa !4
+  %11 = add nsw i32 %8, -1
+  store i32 %11, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit12
 
-15:                                               ; preds = %10
-  %.not.i = icmp eq i32 %11, 0
+15:                                               ; preds = %6
+  %.not.i = icmp eq i32 %8, 0
   br i1 %.not.i, label %lean_dec.exit12, label %16
 
 16:                                               ; preds = %15
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit12
 
-lean_dec.exit12:                                  ; preds = %16, %15, %13, %6
+lean_dec.exit12:                                  ; preds = %16, %15, %10
   %17 = getelementptr i8, ptr %2, i64 8
   %.val17 = load i64, ptr %17, align 8, !tbaa !12
-  %18 = ptrtoint ptr %2 to i64
-  %19 = and i64 %18, 1
-  %.not18 = icmp eq i64 %19, 0
-  br i1 %.not18, label %20, label %lean_dec.exit11
-
-20:                                               ; preds = %lean_dec.exit12
-  %21 = load i32, ptr %2, align 4, !tbaa !4
-  %22 = icmp sgt i32 %21, 1
-  br i1 %22, label %23, label %25, !prof !9
-
-23:                                               ; preds = %20
-  %24 = add nsw i32 %21, -1
-  store i32 %24, ptr %2, align 4, !tbaa !4
+  %15 = load i32, ptr %2, align 8, !tbaa !4
+  %16 = icmp sgt i32 %15, 1
+  br i1 %16, label %17, label %19, !prof !9
+
+17:                                               ; preds = %lean_dec.exit12
+  %18 = add nsw i32 %15, -1
+  store i32 %18, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit11
 
-25:                                               ; preds = %20
-  %.not.i13 = icmp eq i32 %21, 0
+25:                                               ; preds = %lean_dec.exit12
+  %.not.i13 = icmp eq i32 %15, 0
   br i1 %.not.i13, label %lean_dec.exit11, label %26
 
 26:                                               ; preds = %25
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #4
   br label %lean_dec.exit11
 
-lean_dec.exit11:                                  ; preds = %26, %25, %23, %lean_dec.exit12
+lean_dec.exit11:                                  ; preds = %26, %25, %17
   %27 = tail call ptr @l_Array_foldlMUnsafe_fold___at_Lean_CollectAxioms_collect___spec__1(ptr noundef %0, i64 noundef %.val, i64 noundef %.val17, ptr noundef %3, ptr noundef %4, ptr noundef %5)
   %28 = ptrtoint ptr %0 to i64
   %29 = and i64 %28, 1
diff --git a/bench/lean4/optimized/CollectFVars.ll b/bench/lean4/optimized/CollectFVars.ll
index 6780a85a9cc..bd0d7554343 100644
--- a/bench/lean4/optimized/CollectFVars.ll
+++ b/bench/lean4/optimized/CollectFVars.ll
@@ -10857,56 +10857,44 @@ lean_dec.exit:                                    ; preds = %21, %20, %18, %lean
 define ptr @l_Array_foldrMUnsafe_fold___at_Lean_Meta_removeUnused___spec__2___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8) local_unnamed_addr #0 {
   %10 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %10, align 8, !tbaa !12
-  %11 = ptrtoint ptr %1 to i64
-  %12 = and i64 %11, 1
-  %.not = icmp eq i64 %12, 0
-  br i1 %.not, label %13, label %lean_dec.exit15
+  %11 = load i32, ptr %1, align 8, !tbaa !4
+  %12 = icmp sgt i32 %11, 1
+  br i1 %12, label %13, label %15, !prof !11
 
 13:                                               ; preds = %9
-  %14 = load i32, ptr %1, align 4, !tbaa !4
-  %15 = icmp sgt i32 %14, 1
-  br i1 %15, label %16, label %18, !prof !11
-
-16:                                               ; preds = %13
-  %17 = add nsw i32 %14, -1
-  store i32 %17, ptr %1, align 4, !tbaa !4
+  %14 = add nsw i32 %11, -1
+  store i32 %14, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit15
 
-18:                                               ; preds = %13
-  %.not.i = icmp eq i32 %14, 0
+18:                                               ; preds = %9
+  %.not.i = icmp eq i32 %11, 0
   br i1 %.not.i, label %lean_dec.exit15, label %19
 
 19:                                               ; preds = %18
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_dec.exit15
 
-lean_dec.exit15:                                  ; preds = %19, %18, %16, %9
+lean_dec.exit15:                                  ; preds = %19, %18, %13
   %20 = getelementptr i8, ptr %2, i64 8
   %.val20 = load i64, ptr %20, align 8, !tbaa !12
-  %21 = ptrtoint ptr %2 to i64
-  %22 = and i64 %21, 1
-  %.not21 = icmp eq i64 %22, 0
-  br i1 %.not21, label %23, label %lean_dec.exit14
+  %18 = load i32, ptr %2, align 8, !tbaa !4
+  %19 = icmp sgt i32 %18, 1
+  br i1 %19, label %26, label %22, !prof !11
 
-23:                                               ; preds = %lean_dec.exit15
-  %24 = load i32, ptr %2, align 4, !tbaa !4
-  %25 = icmp sgt i32 %24, 1
-  br i1 %25, label %26, label %28, !prof !11
-
-26:                                               ; preds = %23
-  %27 = add nsw i32 %24, -1
+26:                                               ; preds = %lean_dec.exit15
+  %27 = add nsw i32 %18, -1
   store i32 %27, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit14
 
-28:                                               ; preds = %23
-  %.not.i16 = icmp eq i32 %24, 0
+28:                                               ; preds = %lean_dec.exit15
+  %.not.i16 = icmp eq i32 %18, 0
   br i1 %.not.i16, label %lean_dec.exit14, label %29
 
 29:                                               ; preds = %28
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #3
   br label %lean_dec.exit14
 
-lean_dec.exit14:                                  ; preds = %29, %28, %26, %lean_dec.exit15
+lean_dec.exit14:                                  ; preds = %29, %28, %26
   %30 = tail call ptr @l_Array_foldrMUnsafe_fold___at_Lean_Meta_removeUnused___spec__2(ptr noundef %0, i64 noundef %.val, i64 noundef %.val20, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8)
   %31 = ptrtoint ptr %0 to i64
   %32 = and i64 %31, 1
diff --git a/bench/lean4/optimized/ConstFolding.ll b/bench/lean4/optimized/ConstFolding.ll
index 29609c81981..afdaa810b96 100644
--- a/bench/lean4/optimized/ConstFolding.ll
+++ b/bench/lean4/optimized/ConstFolding.ll
@@ -321,30 +321,24 @@ define nonnull ptr @l_Lean_Compiler_isOfNat___lambda__1___boxed(ptr noundef %0,
   %3 = getelementptr inbounds nuw i8, ptr %1, i64 24
   %4 = load ptr, ptr %3, align 8, !tbaa !4
   %5 = tail call zeroext i8 @lean_name_eq(ptr noundef %4, ptr noundef %0) #5
-  %6 = ptrtoint ptr %1 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit5
+  %6 = load i32, ptr %1, align 8, !tbaa !8
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !11
 
 8:                                                ; preds = %2
-  %9 = load i32, ptr %1, align 4, !tbaa !8
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !11
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %1, align 4, !tbaa !8
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %1, align 4, !tbaa !8
   br label %lean_dec.exit5
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %2
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit5, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #5
   br label %lean_dec.exit5
 
-lean_dec.exit5:                                   ; preds = %14, %13, %11, %2
+lean_dec.exit5:                                   ; preds = %14, %13, %8
   %15 = ptrtoint ptr %0 to i64
   %16 = and i64 %15, 1
   %.not8 = icmp eq i64 %16, 0
@@ -451,30 +445,24 @@ define nonnull ptr @l_Lean_Compiler_isToNat___lambda__1___boxed(ptr noundef %0,
   %3 = getelementptr inbounds nuw i8, ptr %1, i64 32
   %4 = load ptr, ptr %3, align 8, !tbaa !4
   %5 = tail call zeroext i8 @lean_name_eq(ptr noundef %4, ptr noundef %0) #5
-  %6 = ptrtoint ptr %1 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit5
+  %6 = load i32, ptr %1, align 8, !tbaa !8
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !11
 
 8:                                                ; preds = %2
-  %9 = load i32, ptr %1, align 4, !tbaa !8
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !11
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %1, align 4, !tbaa !8
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %1, align 4, !tbaa !8
   br label %lean_dec.exit5
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %2
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit5, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #5
   br label %lean_dec.exit5
 
-lean_dec.exit5:                                   ; preds = %14, %13, %11, %2
+lean_dec.exit5:                                   ; preds = %14, %13, %8
   %15 = ptrtoint ptr %0 to i64
   %16 = and i64 %15, 1
   %.not8 = icmp eq i64 %16, 0
diff --git a/bench/lean4/optimized/Database.ll b/bench/lean4/optimized/Database.ll
index 40a199bb01d..57e35d6e64f 100644
--- a/bench/lean4/optimized/Database.ll
+++ b/bench/lean4/optimized/Database.ll
@@ -2107,56 +2107,44 @@ declare ptr @l_Std_Time_Database_Windows_getZoneRules(ptr noundef, ptr noundef)
 define ptr @l_Array_forIn_x27Unsafe_loop___at_Std_Time_Database_defaultGetZoneRules___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8) local_unnamed_addr #0 {
   %10 = getelementptr i8, ptr %5, i64 8
   %.val = load i64, ptr %10, align 8, !tbaa !13
-  %11 = ptrtoint ptr %5 to i64
-  %12 = and i64 %11, 1
-  %.not = icmp eq i64 %12, 0
-  br i1 %.not, label %13, label %lean_dec.exit19
+  %11 = load i32, ptr %5, align 8, !tbaa !4
+  %12 = icmp sgt i32 %11, 1
+  br i1 %12, label %13, label %15, !prof !9
 
 13:                                               ; preds = %9
-  %14 = load i32, ptr %5, align 4, !tbaa !4
-  %15 = icmp sgt i32 %14, 1
-  br i1 %15, label %16, label %18, !prof !9
-
-16:                                               ; preds = %13
-  %17 = add nsw i32 %14, -1
-  store i32 %17, ptr %5, align 4, !tbaa !4
+  %14 = add nsw i32 %11, -1
+  store i32 %14, ptr %5, align 4, !tbaa !4
   br label %lean_dec.exit19
 
-18:                                               ; preds = %13
-  %.not.i = icmp eq i32 %14, 0
+18:                                               ; preds = %9
+  %.not.i = icmp eq i32 %11, 0
   br i1 %.not.i, label %lean_dec.exit19, label %19
 
 19:                                               ; preds = %18
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %5) #4
   br label %lean_dec.exit19
 
-lean_dec.exit19:                                  ; preds = %19, %18, %16, %9
+lean_dec.exit19:                                  ; preds = %19, %18, %13
   %20 = getelementptr i8, ptr %6, i64 8
   %.val28 = load i64, ptr %20, align 8, !tbaa !13
-  %21 = ptrtoint ptr %6 to i64
-  %22 = and i64 %21, 1
-  %.not29 = icmp eq i64 %22, 0
-  br i1 %.not29, label %23, label %lean_dec.exit18
-
-23:                                               ; preds = %lean_dec.exit19
-  %24 = load i32, ptr %6, align 4, !tbaa !4
-  %25 = icmp sgt i32 %24, 1
-  br i1 %25, label %26, label %28, !prof !9
-
-26:                                               ; preds = %23
-  %27 = add nsw i32 %24, -1
-  store i32 %27, ptr %6, align 4, !tbaa !4
+  %18 = load i32, ptr %6, align 8, !tbaa !4
+  %19 = icmp sgt i32 %18, 1
+  br i1 %19, label %20, label %22, !prof !9
+
+20:                                               ; preds = %lean_dec.exit19
+  %21 = add nsw i32 %18, -1
+  store i32 %21, ptr %6, align 4, !tbaa !4
   br label %lean_dec.exit18
 
-28:                                               ; preds = %23
-  %.not.i20 = icmp eq i32 %24, 0
+28:                                               ; preds = %lean_dec.exit19
+  %.not.i20 = icmp eq i32 %18, 0
   br i1 %.not.i20, label %lean_dec.exit18, label %29
 
 29:                                               ; preds = %28
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %6) #4
   br label %lean_dec.exit18
 
-lean_dec.exit18:                                  ; preds = %29, %28, %26, %lean_dec.exit19
+lean_dec.exit18:                                  ; preds = %29, %28, %20
   %30 = tail call ptr @l_Array_forIn_x27Unsafe_loop___at_Std_Time_Database_defaultGetZoneRules___spec__1(ptr noundef %0, ptr poison, ptr poison, ptr noundef %3, ptr noundef %4, i64 noundef %.val, i64 noundef %.val28, ptr noundef %7, ptr noundef %8)
   %31 = ptrtoint ptr %4 to i64
   %32 = and i64 %31, 1
diff --git a/bench/lean4/optimized/Delta.ll b/bench/lean4/optimized/Delta.ll
index 6976565abec..68b1a04162c 100644
--- a/bench/lean4/optimized/Delta.ll
+++ b/bench/lean4/optimized/Delta.ll
@@ -1564,56 +1564,44 @@ declare ptr @l_Lean_Syntax_getArgs(ptr noundef) local_unnamed_addr #1
 define ptr @l_Lean_Elab_Tactic_Conv_evalDelta___lambda__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9, ptr noundef %10, ptr noundef %11) #0 {
   %13 = getelementptr i8, ptr %0, i64 8
   %.val = load i64, ptr %13, align 8, !tbaa !14
-  %14 = ptrtoint ptr %0 to i64
-  %15 = and i64 %14, 1
-  %.not = icmp eq i64 %15, 0
-  br i1 %.not, label %16, label %lean_dec.exit16
+  %14 = load i32, ptr %0, align 8, !tbaa !4
+  %15 = icmp sgt i32 %14, 1
+  br i1 %15, label %16, label %18, !prof !9
 
 16:                                               ; preds = %12
-  %17 = load i32, ptr %0, align 4, !tbaa !4
-  %18 = icmp sgt i32 %17, 1
-  br i1 %18, label %19, label %21, !prof !9
-
-19:                                               ; preds = %16
-  %20 = add nsw i32 %17, -1
-  store i32 %20, ptr %0, align 4, !tbaa !4
+  %17 = add nsw i32 %14, -1
+  store i32 %17, ptr %0, align 4, !tbaa !4
   br label %lean_dec.exit16
 
-21:                                               ; preds = %16
-  %.not.i = icmp eq i32 %17, 0
+21:                                               ; preds = %12
+  %.not.i = icmp eq i32 %14, 0
   br i1 %.not.i, label %lean_dec.exit16, label %22
 
 22:                                               ; preds = %21
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_dec.exit16
 
-lean_dec.exit16:                                  ; preds = %22, %21, %19, %12
+lean_dec.exit16:                                  ; preds = %22, %21, %16
   %23 = getelementptr i8, ptr %1, i64 8
   %.val19 = load i64, ptr %23, align 8, !tbaa !14
-  %24 = ptrtoint ptr %1 to i64
-  %25 = and i64 %24, 1
-  %.not20 = icmp eq i64 %25, 0
-  br i1 %.not20, label %26, label %lean_dec.exit
-
-26:                                               ; preds = %lean_dec.exit16
-  %27 = load i32, ptr %1, align 4, !tbaa !4
-  %28 = icmp sgt i32 %27, 1
-  br i1 %28, label %29, label %31, !prof !9
-
-29:                                               ; preds = %26
-  %30 = add nsw i32 %27, -1
-  store i32 %30, ptr %1, align 4, !tbaa !4
+  %21 = load i32, ptr %1, align 8, !tbaa !4
+  %22 = icmp sgt i32 %21, 1
+  br i1 %22, label %23, label %25, !prof !9
+
+23:                                               ; preds = %lean_dec.exit16
+  %24 = add nsw i32 %21, -1
+  store i32 %24, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit
 
-31:                                               ; preds = %26
-  %.not.i17 = icmp eq i32 %27, 0
+31:                                               ; preds = %lean_dec.exit16
+  %.not.i17 = icmp eq i32 %21, 0
   br i1 %.not.i17, label %lean_dec.exit, label %32
 
 32:                                               ; preds = %31
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %32, %31, %29, %lean_dec.exit16
+lean_dec.exit:                                    ; preds = %32, %31, %23
   %33 = tail call ptr @l_Lean_Elab_Tactic_Conv_evalDelta___lambda__1(i64 noundef %.val, i64 noundef %.val19, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9, ptr noundef %10, ptr noundef %11)
   ret ptr %33
 }
diff --git a/bench/lean4/optimized/EligibleHeaderDecls.ll b/bench/lean4/optimized/EligibleHeaderDecls.ll
index a8172459c1c..16d8b113815 100644
--- a/bench/lean4/optimized/EligibleHeaderDecls.ll
+++ b/bench/lean4/optimized/EligibleHeaderDecls.ll
@@ -3074,56 +3074,44 @@ declare ptr @lean_st_ref_set(ptr noundef, ptr noundef, ptr noundef) local_unname
 define noalias nonnull ptr @l_Array_foldlMUnsafe_fold___at_Lean_Server_Completion_getEligibleHeaderDecls___spec__2___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5) local_unnamed_addr #0 {
   %7 = getelementptr i8, ptr %2, i64 8
   %.val = load i64, ptr %7, align 8, !tbaa !12
-  %8 = ptrtoint ptr %2 to i64
-  %9 = and i64 %8, 1
-  %.not = icmp eq i64 %9, 0
-  br i1 %.not, label %10, label %lean_dec.exit12
+  %8 = load i32, ptr %2, align 8, !tbaa !4
+  %9 = icmp sgt i32 %8, 1
+  br i1 %9, label %10, label %12, !prof !11
 
 10:                                               ; preds = %6
-  %11 = load i32, ptr %2, align 4, !tbaa !4
-  %12 = icmp sgt i32 %11, 1
-  br i1 %12, label %13, label %15, !prof !11
-
-13:                                               ; preds = %10
-  %14 = add nsw i32 %11, -1
-  store i32 %14, ptr %2, align 4, !tbaa !4
+  %11 = add nsw i32 %8, -1
+  store i32 %11, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit12
 
-15:                                               ; preds = %10
-  %.not.i = icmp eq i32 %11, 0
+15:                                               ; preds = %6
+  %.not.i = icmp eq i32 %8, 0
   br i1 %.not.i, label %lean_dec.exit12, label %16
 
 16:                                               ; preds = %15
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #3
   br label %lean_dec.exit12
 
-lean_dec.exit12:                                  ; preds = %16, %15, %13, %6
+lean_dec.exit12:                                  ; preds = %16, %15, %10
   %17 = getelementptr i8, ptr %3, i64 8
   %.val17 = load i64, ptr %17, align 8, !tbaa !12
-  %18 = ptrtoint ptr %3 to i64
-  %19 = and i64 %18, 1
-  %.not18 = icmp eq i64 %19, 0
-  br i1 %.not18, label %20, label %lean_dec.exit11
-
-20:                                               ; preds = %lean_dec.exit12
-  %21 = load i32, ptr %3, align 4, !tbaa !4
-  %22 = icmp sgt i32 %21, 1
-  br i1 %22, label %23, label %25, !prof !11
+  %15 = load i32, ptr %3, align 8, !tbaa !4
+  %16 = icmp sgt i32 %15, 1
+  br i1 %16, label %17, label %19, !prof !11
 
-23:                                               ; preds = %20
-  %24 = add nsw i32 %21, -1
+23:                                               ; preds = %lean_dec.exit12
+  %24 = add nsw i32 %15, -1
   store i32 %24, ptr %3, align 4, !tbaa !4
   br label %lean_dec.exit11
 
-25:                                               ; preds = %20
-  %.not.i13 = icmp eq i32 %21, 0
+25:                                               ; preds = %lean_dec.exit12
+  %.not.i13 = icmp eq i32 %15, 0
   br i1 %.not.i13, label %lean_dec.exit11, label %26
 
 26:                                               ; preds = %25
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %3) #3
   br label %lean_dec.exit11
 
-lean_dec.exit11:                                  ; preds = %26, %25, %23, %lean_dec.exit12
+lean_dec.exit11:                                  ; preds = %26, %25, %23
   %27 = tail call ptr @l_Array_foldlMUnsafe_fold___at_Lean_Server_Completion_getEligibleHeaderDecls___spec__2(ptr noundef %0, ptr noundef %1, i64 noundef %.val, i64 noundef %.val17, ptr noundef %4, ptr noundef %5)
   %28 = ptrtoint ptr %1 to i64
   %29 = and i64 %28, 1
@@ -4020,56 +4008,44 @@ lean_dec.exit:                                    ; preds = %132, %131, %129, %l
 define ptr @l_Array_foldlMUnsafe_fold___at_Lean_Server_Completion_forEligibleDeclsM___spec__4___rarg___lambda__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5) #0 {
   %7 = getelementptr i8, ptr %0, i64 8
   %.val13 = load i64, ptr %7, align 8, !tbaa !12
-  %8 = ptrtoint ptr %0 to i64
-  %9 = and i64 %8, 1
-  %.not = icmp eq i64 %9, 0
-  br i1 %.not, label %10, label %lean_dec.exit10
+  %8 = load i32, ptr %0, align 8, !tbaa !4
+  %9 = icmp sgt i32 %8, 1
+  br i1 %9, label %10, label %12, !prof !11
 
 10:                                               ; preds = %6
-  %11 = load i32, ptr %0, align 4, !tbaa !4
-  %12 = icmp sgt i32 %11, 1
-  br i1 %12, label %13, label %15, !prof !11
-
-13:                                               ; preds = %10
-  %14 = add nsw i32 %11, -1
-  store i32 %14, ptr %0, align 4, !tbaa !4
+  %11 = add nsw i32 %8, -1
+  store i32 %11, ptr %0, align 4, !tbaa !4
   br label %lean_dec.exit10
 
-15:                                               ; preds = %10
-  %.not.i = icmp eq i32 %11, 0
+15:                                               ; preds = %6
+  %.not.i = icmp eq i32 %8, 0
   br i1 %.not.i, label %lean_dec.exit10, label %16
 
 16:                                               ; preds = %15
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_dec.exit10
 
-lean_dec.exit10:                                  ; preds = %16, %15, %13, %6
+lean_dec.exit10:                                  ; preds = %16, %15, %10
   %17 = getelementptr i8, ptr %4, i64 8
   %.val = load i64, ptr %17, align 8, !tbaa !12
-  %18 = ptrtoint ptr %4 to i64
-  %19 = and i64 %18, 1
-  %.not14 = icmp eq i64 %19, 0
-  br i1 %.not14, label %20, label %lean_dec.exit
+  %15 = load i32, ptr %4, align 8, !tbaa !4
+  %16 = icmp sgt i32 %15, 1
+  br i1 %16, label %17, label %19, !prof !11
 
-20:                                               ; preds = %lean_dec.exit10
-  %21 = load i32, ptr %4, align 4, !tbaa !4
-  %22 = icmp sgt i32 %21, 1
-  br i1 %22, label %23, label %25, !prof !11
-
-23:                                               ; preds = %20
-  %24 = add nsw i32 %21, -1
+23:                                               ; preds = %lean_dec.exit10
+  %24 = add nsw i32 %15, -1
   store i32 %24, ptr %4, align 4, !tbaa !4
   br label %lean_dec.exit
 
-25:                                               ; preds = %20
-  %.not.i11 = icmp eq i32 %21, 0
+25:                                               ; preds = %lean_dec.exit10
+  %.not.i11 = icmp eq i32 %15, 0
   br i1 %.not.i11, label %lean_dec.exit, label %26
 
 26:                                               ; preds = %25
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %4) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %26, %25, %23, %lean_dec.exit10
+lean_dec.exit:                                    ; preds = %26, %25, %23
   %27 = add i64 %.val13, 1
   %28 = tail call ptr @l_Array_foldlMUnsafe_fold___at_Lean_Server_Completion_forEligibleDeclsM___spec__4___rarg(ptr noundef %1, ptr noundef %2, ptr noundef %3, i64 noundef %27, i64 noundef %.val, ptr noundef %5)
   ret ptr %28
@@ -4103,56 +4079,44 @@ lean_alloc_closure.exit:                          ; preds = %1
 define ptr @l_Array_foldlMUnsafe_fold___at_Lean_Server_Completion_forEligibleDeclsM___spec__4___rarg___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5) #0 {
   %7 = getelementptr i8, ptr %3, i64 8
   %.val13 = load i64, ptr %7, align 8, !tbaa !12
-  %8 = ptrtoint ptr %3 to i64
-  %9 = and i64 %8, 1
-  %.not = icmp eq i64 %9, 0
-  br i1 %.not, label %10, label %lean_dec.exit10
+  %8 = load i32, ptr %3, align 8, !tbaa !4
+  %9 = icmp sgt i32 %8, 1
+  br i1 %9, label %10, label %12, !prof !11
 
 10:                                               ; preds = %6
-  %11 = load i32, ptr %3, align 4, !tbaa !4
-  %12 = icmp sgt i32 %11, 1
-  br i1 %12, label %13, label %15, !prof !11
-
-13:                                               ; preds = %10
-  %14 = add nsw i32 %11, -1
-  store i32 %14, ptr %3, align 4, !tbaa !4
+  %11 = add nsw i32 %8, -1
+  store i32 %11, ptr %3, align 4, !tbaa !4
   br label %lean_dec.exit10
 
-15:                                               ; preds = %10
-  %.not.i = icmp eq i32 %11, 0
+15:                                               ; preds = %6
+  %.not.i = icmp eq i32 %8, 0
   br i1 %.not.i, label %lean_dec.exit10, label %16
 
 16:                                               ; preds = %15
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %3) #3
   br label %lean_dec.exit10
 
-lean_dec.exit10:                                  ; preds = %16, %15, %13, %6
+lean_dec.exit10:                                  ; preds = %16, %15, %10
   %17 = getelementptr i8, ptr %4, i64 8
   %.val = load i64, ptr %17, align 8, !tbaa !12
-  %18 = ptrtoint ptr %4 to i64
-  %19 = and i64 %18, 1
-  %.not14 = icmp eq i64 %19, 0
-  br i1 %.not14, label %20, label %lean_dec.exit
-
-20:                                               ; preds = %lean_dec.exit10
-  %21 = load i32, ptr %4, align 4, !tbaa !4
-  %22 = icmp sgt i32 %21, 1
-  br i1 %22, label %23, label %25, !prof !11
+  %15 = load i32, ptr %4, align 8, !tbaa !4
+  %16 = icmp sgt i32 %15, 1
+  br i1 %16, label %17, label %19, !prof !11
 
-23:                                               ; preds = %20
-  %24 = add nsw i32 %21, -1
+23:                                               ; preds = %lean_dec.exit10
+  %24 = add nsw i32 %15, -1
   store i32 %24, ptr %4, align 4, !tbaa !4
   br label %lean_dec.exit
 
-25:                                               ; preds = %20
-  %.not.i11 = icmp eq i32 %21, 0
+25:                                               ; preds = %lean_dec.exit10
+  %.not.i11 = icmp eq i32 %15, 0
   br i1 %.not.i11, label %lean_dec.exit, label %26
 
 26:                                               ; preds = %25
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %4) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %26, %25, %23, %lean_dec.exit10
+lean_dec.exit:                                    ; preds = %26, %25, %23
   %27 = tail call ptr @l_Array_foldlMUnsafe_fold___at_Lean_Server_Completion_forEligibleDeclsM___spec__4___rarg(ptr noundef %0, ptr noundef %1, ptr noundef %2, i64 noundef %.val13, i64 noundef %.val, ptr noundef %5)
   ret ptr %27
 }
diff --git a/bench/lean4/optimized/ElimDeadVars.ll b/bench/lean4/optimized/ElimDeadVars.ll
index f08957a144d..607ea73ef69 100644
--- a/bench/lean4/optimized/ElimDeadVars.ll
+++ b/bench/lean4/optimized/ElimDeadVars.ll
@@ -1143,13 +1143,7 @@ lean_alloc_ctor.exit:                             ; preds = %lean_dec.exit
   store ptr %57, ptr %107, align 8, !tbaa !4
   %108 = getelementptr inbounds nuw i8, ptr %101, i64 32
   store ptr %100, ptr %108, align 8, !tbaa !4
-  %109 = ptrtoint ptr %101 to i64
-  %110 = and i64 %109, 1
-  %.not.i80 = icmp eq i64 %110, 0
-  br i1 %.not.i80, label %111, label %l_Lean_IR_reshapeWithoutDead.exit
-
-111:                                              ; preds = %lean_alloc_ctor.exit
-  store i32 2, ptr %101, align 4, !tbaa !8
+  store i32 2, ptr %101, align 8, !tbaa !8
   br label %l_Lean_IR_reshapeWithoutDead.exit
 
 112:                                              ; preds = %lean_obj_tag.exit
@@ -1170,8 +1164,8 @@ lean_alloc_ctor.exit:                             ; preds = %lean_dec.exit
   tail call void @lean_inc_ref_cold(ptr noundef nonnull %14) #3
   br label %l_Lean_IR_reshapeWithoutDead.exit
 
-l_Lean_IR_reshapeWithoutDead.exit:                ; preds = %117, %116, %114, %lean_obj_tag.exit.thread, %111, %lean_alloc_ctor.exit, %50, %49, %47, %40
-  %.sink100 = phi ptr [ %14, %40 ], [ %14, %47 ], [ %14, %49 ], [ %14, %50 ], [ %101, %lean_alloc_ctor.exit ], [ %101, %111 ], [ %14, %lean_obj_tag.exit.thread ], [ %14, %114 ], [ %14, %116 ], [ %14, %117 ]
+l_Lean_IR_reshapeWithoutDead.exit:                ; preds = %117, %116, %114, %lean_obj_tag.exit.thread, %lean_alloc_ctor.exit, %50, %49, %47, %40
+  %.sink100 = phi ptr [ %14, %40 ], [ %14, %47 ], [ %14, %49 ], [ %14, %50 ], [ %101, %lean_alloc_ctor.exit ], [ %14, %lean_obj_tag.exit.thread ], [ %14, %111 ], [ %14, %113 ], [ %14, %114 ]
   %118 = tail call ptr @l_Lean_IR_FnBody_freeIndices(ptr noundef %.sink100) #3
   %119 = tail call ptr @l_Lean_IR_reshapeWithoutDead_reshape(ptr noundef %33, ptr noundef %.sink100, ptr noundef %118)
   ret ptr %119
@@ -1617,56 +1611,44 @@ declare ptr @l_Lean_IR_FnBody_flatten(ptr noundef) local_unnamed_addr #1
 define ptr @l_Array_mapMUnsafe_map___at_Lean_IR_FnBody_elimDead___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) local_unnamed_addr #0 {
   %4 = getelementptr i8, ptr %0, i64 8
   %.val = load i64, ptr %4, align 8, !tbaa !11
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit7
+  %5 = load i32, ptr %0, align 8, !tbaa !8
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !13
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %0, align 4, !tbaa !8
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !13
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !8
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !8
   br label %lean_dec.exit7
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit7, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_dec.exit7
 
-lean_dec.exit7:                                   ; preds = %13, %12, %10, %3
+lean_dec.exit7:                                   ; preds = %13, %12, %7
   %14 = getelementptr i8, ptr %1, i64 8
   %.val10 = load i64, ptr %14, align 8, !tbaa !11
-  %15 = ptrtoint ptr %1 to i64
-  %16 = and i64 %15, 1
-  %.not11 = icmp eq i64 %16, 0
-  br i1 %.not11, label %17, label %lean_dec.exit
-
-17:                                               ; preds = %lean_dec.exit7
-  %18 = load i32, ptr %1, align 4, !tbaa !8
-  %19 = icmp sgt i32 %18, 1
-  br i1 %19, label %20, label %22, !prof !13
+  %12 = load i32, ptr %1, align 8, !tbaa !8
+  %13 = icmp sgt i32 %12, 1
+  br i1 %13, label %14, label %16, !prof !13
 
-20:                                               ; preds = %17
-  %21 = add nsw i32 %18, -1
+20:                                               ; preds = %lean_dec.exit7
+  %21 = add nsw i32 %12, -1
   store i32 %21, ptr %1, align 4, !tbaa !8
   br label %lean_dec.exit
 
-22:                                               ; preds = %17
-  %.not.i8 = icmp eq i32 %18, 0
+22:                                               ; preds = %lean_dec.exit7
+  %.not.i8 = icmp eq i32 %12, 0
   br i1 %.not.i8, label %lean_dec.exit, label %23
 
 23:                                               ; preds = %22
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %23, %22, %20, %lean_dec.exit7
+lean_dec.exit:                                    ; preds = %23, %22, %20
   %24 = tail call ptr @l_Array_mapMUnsafe_map___at_Lean_IR_FnBody_elimDead___spec__1(i64 noundef %.val, i64 noundef %.val10, ptr noundef %2)
   ret ptr %24
 }
@@ -1675,56 +1657,44 @@ lean_dec.exit:                                    ; preds = %23, %22, %20, %lean
 define ptr @l_Array_mapMUnsafe_map___at_Lean_IR_FnBody_elimDead___spec__2___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) local_unnamed_addr #0 {
   %4 = getelementptr i8, ptr %0, i64 8
   %.val10 = load i64, ptr %4, align 8, !tbaa !11
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit7
+  %5 = load i32, ptr %0, align 8, !tbaa !8
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !13
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %0, align 4, !tbaa !8
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !13
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !8
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !8
   br label %lean_dec.exit7
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit7, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_dec.exit7
 
-lean_dec.exit7:                                   ; preds = %13, %12, %10, %3
+lean_dec.exit7:                                   ; preds = %13, %12, %7
   %14 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %14, align 8, !tbaa !11
-  %15 = ptrtoint ptr %1 to i64
-  %16 = and i64 %15, 1
-  %.not11 = icmp eq i64 %16, 0
-  br i1 %.not11, label %17, label %lean_dec.exit
-
-17:                                               ; preds = %lean_dec.exit7
-  %18 = load i32, ptr %1, align 4, !tbaa !8
-  %19 = icmp sgt i32 %18, 1
-  br i1 %19, label %20, label %22, !prof !13
+  %12 = load i32, ptr %1, align 8, !tbaa !8
+  %13 = icmp sgt i32 %12, 1
+  br i1 %13, label %14, label %16, !prof !13
 
-20:                                               ; preds = %17
-  %21 = add nsw i32 %18, -1
+20:                                               ; preds = %lean_dec.exit7
+  %21 = add nsw i32 %12, -1
   store i32 %21, ptr %1, align 4, !tbaa !8
   br label %lean_dec.exit
 
-22:                                               ; preds = %17
-  %.not.i8 = icmp eq i32 %18, 0
+22:                                               ; preds = %lean_dec.exit7
+  %.not.i8 = icmp eq i32 %12, 0
   br i1 %.not.i8, label %lean_dec.exit, label %23
 
 23:                                               ; preds = %22
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %23, %22, %20, %lean_dec.exit7
+lean_dec.exit:                                    ; preds = %23, %22, %20
   %24 = tail call ptr @l_Array_mapMUnsafe_map___at_Lean_IR_FnBody_elimDead___spec__2(i64 noundef %.val10, i64 noundef %.val, ptr noundef %2)
   ret ptr %24
 }
diff --git a/bench/lean4/optimized/EmbeddedConstraint.ll b/bench/lean4/optimized/EmbeddedConstraint.ll
index 6ab461974f9..86fd2c85611 100644
--- a/bench/lean4/optimized/EmbeddedConstraint.ll
+++ b/bench/lean4/optimized/EmbeddedConstraint.ll
@@ -10546,56 +10546,44 @@ lean_dec.exit:                                    ; preds = %32, %31, %29, %lean
 define ptr @l_Array_forIn_x27Unsafe_loop___at_Lean_Elab_Tactic_BVDecide_Frontend_Normalize_embeddedConstraintPass___elambda__1___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9, ptr noundef %10, ptr noundef %11, ptr noundef %12) local_unnamed_addr #0 {
   %14 = getelementptr i8, ptr %3, i64 8
   %.val = load i64, ptr %14, align 8, !tbaa !12
-  %15 = ptrtoint ptr %3 to i64
-  %16 = and i64 %15, 1
-  %.not = icmp eq i64 %16, 0
-  br i1 %.not, label %17, label %lean_dec.exit27
+  %15 = load i32, ptr %3, align 8, !tbaa !8
+  %16 = icmp sgt i32 %15, 1
+  br i1 %16, label %17, label %19, !prof !11
 
 17:                                               ; preds = %13
-  %18 = load i32, ptr %3, align 4, !tbaa !8
-  %19 = icmp sgt i32 %18, 1
-  br i1 %19, label %20, label %22, !prof !11
-
-20:                                               ; preds = %17
-  %21 = add nsw i32 %18, -1
-  store i32 %21, ptr %3, align 4, !tbaa !8
+  %18 = add nsw i32 %15, -1
+  store i32 %18, ptr %3, align 4, !tbaa !8
   br label %lean_dec.exit27
 
-22:                                               ; preds = %17
-  %.not.i = icmp eq i32 %18, 0
+22:                                               ; preds = %13
+  %.not.i = icmp eq i32 %15, 0
   br i1 %.not.i, label %lean_dec.exit27, label %23
 
 23:                                               ; preds = %22
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %3) #4
   br label %lean_dec.exit27
 
-lean_dec.exit27:                                  ; preds = %23, %22, %20, %13
+lean_dec.exit27:                                  ; preds = %23, %22, %17
   %24 = getelementptr i8, ptr %4, i64 8
   %.val40 = load i64, ptr %24, align 8, !tbaa !12
-  %25 = ptrtoint ptr %4 to i64
-  %26 = and i64 %25, 1
-  %.not41 = icmp eq i64 %26, 0
-  br i1 %.not41, label %27, label %lean_dec.exit26
-
-27:                                               ; preds = %lean_dec.exit27
-  %28 = load i32, ptr %4, align 4, !tbaa !8
-  %29 = icmp sgt i32 %28, 1
-  br i1 %29, label %30, label %32, !prof !11
+  %22 = load i32, ptr %4, align 8, !tbaa !8
+  %23 = icmp sgt i32 %22, 1
+  br i1 %23, label %24, label %26, !prof !11
 
-30:                                               ; preds = %27
-  %31 = add nsw i32 %28, -1
+30:                                               ; preds = %lean_dec.exit27
+  %31 = add nsw i32 %22, -1
   store i32 %31, ptr %4, align 4, !tbaa !8
   br label %lean_dec.exit26
 
-32:                                               ; preds = %27
-  %.not.i28 = icmp eq i32 %28, 0
+32:                                               ; preds = %lean_dec.exit27
+  %.not.i28 = icmp eq i32 %22, 0
   br i1 %.not.i28, label %lean_dec.exit26, label %33
 
 33:                                               ; preds = %32
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %4) #4
   br label %lean_dec.exit26
 
-lean_dec.exit26:                                  ; preds = %33, %32, %30, %lean_dec.exit27
+lean_dec.exit26:                                  ; preds = %33, %32, %30
   %34 = tail call ptr @l_Array_forIn_x27Unsafe_loop___at_Lean_Elab_Tactic_BVDecide_Frontend_Normalize_embeddedConstraintPass___elambda__1___spec__1(ptr poison, ptr poison, ptr noundef %2, i64 noundef %.val, i64 noundef %.val40, ptr noundef %5, ptr poison, ptr poison, ptr noundef %8, ptr noundef %9, ptr noundef %10, ptr noundef %11, ptr noundef %12)
   %35 = ptrtoint ptr %7 to i64
   %36 = and i64 %35, 1
diff --git a/bench/lean4/optimized/Encode.ll b/bench/lean4/optimized/Encode.ll
index 8c2f4898abd..5c8c491a7b2 100644
--- a/bench/lean4/optimized/Encode.ll
+++ b/bench/lean4/optimized/Encode.ll
@@ -204,30 +204,24 @@ lean_alloc_ctor.exit:                             ; preds = %1
 define noalias nonnull ptr @l_Lake_instToTomlFloat___boxed(ptr noundef %0) local_unnamed_addr #0 {
   %2 = getelementptr i8, ptr %0, i64 8
   %.val = load double, ptr %2, align 8, !tbaa !12
-  %3 = ptrtoint ptr %0 to i64
-  %4 = and i64 %3, 1
-  %.not = icmp eq i64 %4, 0
-  br i1 %.not, label %5, label %lean_dec.exit
+  %3 = load i32, ptr %0, align 8, !tbaa !4
+  %4 = icmp sgt i32 %3, 1
+  br i1 %4, label %5, label %10, !prof !11
 
 5:                                                ; preds = %1
-  %6 = load i32, ptr %0, align 4, !tbaa !4
-  %7 = icmp sgt i32 %6, 1
-  br i1 %7, label %8, label %10, !prof !11
-
-8:                                                ; preds = %5
-  %9 = add nsw i32 %6, -1
-  store i32 %9, ptr %0, align 4, !tbaa !4
+  %6 = add nsw i32 %3, -1
+  store i32 %6, ptr %0, align 4, !tbaa !4
   br label %lean_dec.exit
 
-10:                                               ; preds = %5
-  %.not.i = icmp eq i32 %6, 0
+10:                                               ; preds = %1
+  %.not.i = icmp eq i32 %3, 0
   br i1 %.not.i, label %lean_dec.exit, label %11
 
 11:                                               ; preds = %10
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #4
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %11, %10, %8, %1
+lean_dec.exit:                                    ; preds = %11, %10, %5
   tail call void @lean_inc_heartbeat() #4
   %12 = tail call noalias ptr @mi_malloc_small(i64 noundef 24) #4
   %13 = icmp eq ptr %12, null
@@ -520,56 +514,44 @@ lean_alloc_closure.exit:                          ; preds = %1
 define ptr @l_Array_mapMUnsafe_map___at_Lake_instToTomlArray___spec__1___rarg___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3) #0 {
   %5 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %5, align 8, !tbaa !14
-  %6 = ptrtoint ptr %1 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit8
+  %6 = load i32, ptr %1, align 8, !tbaa !4
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !11
 
 8:                                                ; preds = %4
-  %9 = load i32, ptr %1, align 4, !tbaa !4
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !11
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %1, align 4, !tbaa !4
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit8
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %4
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit8, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit8
 
-lean_dec.exit8:                                   ; preds = %14, %13, %11, %4
+lean_dec.exit8:                                   ; preds = %14, %13, %11
   %15 = getelementptr i8, ptr %2, i64 8
   %.val11 = load i64, ptr %15, align 8, !tbaa !14
-  %16 = ptrtoint ptr %2 to i64
-  %17 = and i64 %16, 1
-  %.not12 = icmp eq i64 %17, 0
-  br i1 %.not12, label %18, label %lean_dec.exit
+  %16 = load i32, ptr %2, align 8, !tbaa !4
+  %17 = icmp sgt i32 %16, 1
+  br i1 %14, label %15, label %17, !prof !11
 
 18:                                               ; preds = %lean_dec.exit8
-  %19 = load i32, ptr %2, align 4, !tbaa !4
-  %20 = icmp sgt i32 %19, 1
-  br i1 %20, label %21, label %23, !prof !11
-
-21:                                               ; preds = %18
-  %22 = add nsw i32 %19, -1
-  store i32 %22, ptr %2, align 4, !tbaa !4
+  %16 = add nsw i32 %13, -1
+  store i32 %16, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit
 
-23:                                               ; preds = %18
-  %.not.i9 = icmp eq i32 %19, 0
+23:                                               ; preds = %lean_dec.exit8
+  %.not.i9 = icmp eq i32 %13, 0
   br i1 %.not.i9, label %lean_dec.exit, label %24
 
 24:                                               ; preds = %23
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #4
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %24, %23, %21, %lean_dec.exit8
+lean_dec.exit:                                    ; preds = %24, %23, %18
   %25 = tail call ptr @l_Array_mapMUnsafe_map___at_Lake_instToTomlArray___spec__1___rarg(ptr noundef %0, i64 noundef %.val, i64 noundef %.val11, ptr noundef %3)
   ret ptr %25
 }
@@ -1030,22 +1012,16 @@ lean_alloc_closure.exit:                          ; preds = %1
 define ptr @l_Array_foldlMUnsafe_fold___at_Lake_Toml_encodeArray_x3f___spec__1___rarg___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4) #0 {
   %6 = getelementptr i8, ptr %2, i64 8
   %.val16 = load i64, ptr %6, align 8, !tbaa !14
-  %7 = ptrtoint ptr %2 to i64
-  %8 = and i64 %7, 1
-  %.not = icmp eq i64 %8, 0
-  br i1 %.not, label %9, label %lean_dec.exit11
+  %7 = load i32, ptr %2, align 8, !tbaa !4
+  %8 = icmp sgt i32 %7, 1
+  br i1 %8, label %9, label %11, !prof !11
 
 9:                                                ; preds = %5
-  %10 = load i32, ptr %2, align 4, !tbaa !4
-  %11 = icmp sgt i32 %10, 1
-  br i1 %11, label %12, label %14, !prof !11
-
-12:                                               ; preds = %9
-  %13 = add nsw i32 %10, -1
-  store i32 %13, ptr %2, align 4, !tbaa !4
+  %10 = add nsw i32 %7, -1
+  store i32 %10, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit11
 
-14:                                               ; preds = %9
+14:                                               ; preds = %5
   %.not.i = icmp eq i32 %10, 0
   br i1 %.not.i, label %lean_dec.exit11, label %15
 
@@ -1053,33 +1029,27 @@ define ptr @l_Array_foldlMUnsafe_fold___at_Lake_Toml_encodeArray_x3f___spec__1__
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #4
   br label %lean_dec.exit11
 
-lean_dec.exit11:                                  ; preds = %15, %14, %12, %5
+lean_dec.exit11:                                  ; preds = %15, %14, %12
   %16 = getelementptr i8, ptr %3, i64 8
   %.val = load i64, ptr %16, align 8, !tbaa !14
-  %17 = ptrtoint ptr %3 to i64
-  %18 = and i64 %17, 1
-  %.not17 = icmp eq i64 %18, 0
-  br i1 %.not17, label %19, label %lean_dec.exit10
-
-19:                                               ; preds = %lean_dec.exit11
-  %20 = load i32, ptr %3, align 4, !tbaa !4
-  %21 = icmp sgt i32 %20, 1
-  br i1 %21, label %22, label %24, !prof !11
+  %17 = load i32, ptr %3, align 8, !tbaa !4
+  %15 = icmp sgt i32 %17, 1
+  br i1 %15, label %16, label %18, !prof !11
 
-22:                                               ; preds = %19
-  %23 = add nsw i32 %20, -1
+22:                                               ; preds = %lean_dec.exit11
+  %23 = add nsw i32 %14, -1
   store i32 %23, ptr %3, align 4, !tbaa !4
   br label %lean_dec.exit10
 
-24:                                               ; preds = %19
-  %.not.i12 = icmp eq i32 %20, 0
+24:                                               ; preds = %lean_dec.exit11
+  %.not.i12 = icmp eq i32 %14, 0
   br i1 %.not.i12, label %lean_dec.exit10, label %25
 
 25:                                               ; preds = %24
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %3) #4
   br label %lean_dec.exit10
 
-lean_dec.exit10:                                  ; preds = %25, %24, %22, %lean_dec.exit11
+lean_dec.exit10:                                  ; preds = %25, %24, %22
   %26 = tail call ptr @l_Array_foldlMUnsafe_fold___at_Lake_Toml_encodeArray_x3f___spec__1___rarg(ptr noundef %0, ptr noundef %1, i64 noundef %.val16, i64 noundef %.val, ptr noundef %4)
   %27 = ptrtoint ptr %1 to i64
   %28 = and i64 %27, 1
diff --git a/bench/lean4/optimized/ExportAttr.ll b/bench/lean4/optimized/ExportAttr.ll
index 9eb82b706f6..75b4d9d78d1 100644
--- a/bench/lean4/optimized/ExportAttr.ll
+++ b/bench/lean4/optimized/ExportAttr.ll
@@ -341,22 +341,16 @@ lean_inc.exit:                                    ; preds = %13, %26
   %39 = shl i64 %.val, 1
   %40 = add i64 %39, -1
   %41 = inttoptr i64 %40 to ptr
-  %42 = ptrtoint ptr %0 to i64
-  %43 = and i64 %42, 1
-  %.not104 = icmp eq i64 %43, 0
-  br i1 %.not104, label %44, label %lean_inc.exit80
-
-44:                                               ; preds = %lean_inc.exit
-  %.val.i98 = load i32, ptr %0, align 4, !tbaa !5
-  %45 = icmp sgt i32 %.val.i98, 0
-  br i1 %45, label %46, label %48, !prof !10
-
-46:                                               ; preds = %44
-  %47 = add nuw i32 %.val.i98, 1
-  store i32 %47, ptr %0, align 4, !tbaa !5
+  %.val.i98 = load i32, ptr %0, align 8, !tbaa !5
+  %42 = icmp sgt i32 %.val.i98, 0
+  br i1 %42, label %43, label %45, !prof !10
+
+43:                                               ; preds = %lean_inc.exit
+  %44 = add nuw i32 %.val.i98, 1
+  store i32 %44, ptr %0, align 4, !tbaa !5
   br label %lean_inc.exit80
 
-48:                                               ; preds = %44
+48:                                               ; preds = %lean_inc.exit
   %.not.i99 = icmp eq i32 %.val.i98, 0
   br i1 %.not.i99, label %lean_inc.exit80, label %49
 
@@ -364,7 +358,7 @@ lean_inc.exit:                                    ; preds = %13, %26
   tail call void @lean_inc_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_inc.exit80
 
-lean_inc.exit80:                                  ; preds = %49, %48, %46, %lean_inc.exit
+lean_inc.exit80:                                  ; preds = %49, %48, %43
   tail call void @lean_inc_heartbeat() #3
   %50 = tail call noalias ptr @mi_malloc_small(i64 noundef 32) #3
   %51 = icmp eq ptr %50, null
@@ -385,79 +379,70 @@ lean_alloc_ctor.exit:                             ; preds = %lean_inc.exit80
   %56 = getelementptr inbounds nuw i8, ptr %50, i64 24
   store ptr %41, ptr %56, align 8, !tbaa !13
   %57 = tail call ptr @l_Substring_nextn(ptr noundef nonnull %50, ptr noundef nonnull inttoptr (i64 3 to ptr), ptr noundef nonnull inttoptr (i64 1 to ptr)) #3
-  %58 = ptrtoint ptr %50 to i64
-  %59 = and i64 %58, 1
-  %.not105 = icmp eq i64 %59, 0
-  br i1 %.not105, label %60, label %lean_dec.exit74
-
-60:                                               ; preds = %lean_alloc_ctor.exit
-  %61 = load i32, ptr %50, align 4, !tbaa !5
-  %62 = icmp sgt i32 %61, 1
-  br i1 %62, label %63, label %65, !prof !10
-
-63:                                               ; preds = %60
-  %64 = add nsw i32 %61, -1
-  store i32 %64, ptr %50, align 4, !tbaa !5
+  %55 = load i32, ptr %50, align 8, !tbaa !5
+  %56 = icmp sgt i32 %55, 1
+  br i1 %56, label %57, label %59, !prof !10
+
+57:                                               ; preds = %lean_alloc_ctor.exit
+  %58 = add nsw i32 %55, -1
+  store i32 %58, ptr %50, align 4, !tbaa !5
   br label %lean_dec.exit74
 
-65:                                               ; preds = %60
-  %.not.i89 = icmp eq i32 %61, 0
+65:                                               ; preds = %lean_alloc_ctor.exit
+  %.not.i89 = icmp eq i32 %55, 0
   br i1 %.not.i89, label %lean_dec.exit74, label %66
 
 66:                                               ; preds = %65
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %50) #3
   br label %lean_dec.exit74
 
-lean_dec.exit74:                                  ; preds = %lean_alloc_ctor.exit, %63, %65, %66
+lean_dec.exit74:                                  ; preds = %57, %65, %66
   %67 = ptrtoint ptr %57 to i64
   %68 = and i64 %67, 1
   %.not106 = icmp eq i64 %68, 0
-  br i1 %.not106, label %69, label %lean_dec.exit72, !prof !4
+  br i1 %.not106, label %63, label %lean_dec.exit72, !prof !4
 
-69:                                               ; preds = %lean_dec.exit74
-  %70 = tail call ptr @lean_nat_big_add(ptr noundef nonnull inttoptr (i64 1 to ptr), ptr noundef %57) #3
-  %71 = load i32, ptr %57, align 4, !tbaa !5
-  %72 = icmp sgt i32 %71, 1
-  br i1 %72, label %73, label %75, !prof !10
+75:                                               ; preds = %lean_dec.exit74
+  %64 = tail call ptr @lean_nat_big_add(ptr noundef nonnull inttoptr (i64 1 to ptr), ptr noundef %54) #3
+  %65 = load i32, ptr %57, align 4, !tbaa !5
+  %66 = icmp sgt i32 %65, 1
+  br i1 %66, label %67, label %69, !prof !10
 
-73:                                               ; preds = %69
-  %74 = add nsw i32 %71, -1
-  store i32 %74, ptr %57, align 4, !tbaa !5
+76:                                               ; preds = %75
+  %68 = add nsw i32 %65, -1
+  store i32 %68, ptr %54, align 4, !tbaa !5
   br label %lean_dec.exit72
 
-75:                                               ; preds = %69
-  %.not.i91 = icmp eq i32 %71, 0
-  br i1 %.not.i91, label %lean_dec.exit72, label %76
+lean_dec.exit72:                                  ; preds = %75
+  %.not.i91 = icmp eq i32 %65, 0
+  br i1 %.not.i91, label %lean_dec.exit72, label %70
 
-76:                                               ; preds = %75
-  tail call void @lean_dec_ref_cold(ptr noundef nonnull %57) #3
+70:                                               ; preds = %69
+  tail call void @lean_dec_ref_cold(ptr noundef nonnull %54) #3
   br label %lean_dec.exit72
 
-lean_dec.exit72:                                  ; preds = %lean_dec.exit74, %73, %75, %76
-  %.0.i122 = phi ptr [ %70, %73 ], [ %70, %75 ], [ %70, %76 ], [ %57, %lean_dec.exit74 ]
-  %77 = tail call zeroext i8 @l_String_anyAux___at___private_Lean_Compiler_ExportAttr_0__Lean_isValidCppId___spec__1(ptr noundef nonnull %0, ptr noundef nonnull %41, ptr noundef %.0.i122)
-  br i1 %.not104, label %78, label %lean_dec.exit
-
-78:                                               ; preds = %lean_dec.exit72
-  %79 = load i32, ptr %0, align 4, !tbaa !5
-  %80 = icmp sgt i32 %79, 1
-  br i1 %80, label %81, label %83, !prof !10
+78:                                               ; preds = %lean_dec.exit74, %67, %69, %70
+  %.0.i122 = phi ptr [ %64, %67 ], [ %64, %69 ], [ %64, %70 ], [ %54, %lean_dec.exit74 ]
+  %71 = tail call zeroext i8 @l_String_anyAux___at___private_Lean_Compiler_ExportAttr_0__Lean_isValidCppId___spec__1(ptr noundef nonnull %0, ptr noundef nonnull %41, ptr noundef %.0.i122)
+  %72 = load i32, ptr %0, align 4, !tbaa !5
+  %73 = icmp sgt i32 %72, 1
+  br i1 %73, label %74, label %76, !prof !10
 
 81:                                               ; preds = %78
-  %82 = add nsw i32 %79, -1
+  %82 = add nsw i32 %72, -1
   store i32 %82, ptr %0, align 4, !tbaa !5
   br label %lean_dec.exit
 
 83:                                               ; preds = %78
-  %.not.i95 = icmp eq i32 %79, 0
+  %.not.i95 = icmp eq i32 %72, 0
   br i1 %.not.i95, label %lean_dec.exit, label %84
 
 84:                                               ; preds = %83
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %84, %83, %81, %lean_dec.exit72
-  %. = xor i8 %77, 1
+lean_dec.exit:                                    ; preds = %84, %83, %81
+  %. = xor i8 %71, 1
   br label %lean_dec.exit79
 
 lean_dec.exit79:                                  ; preds = %28, %34, %36, %37, %.critedge71, %22, %24, %25, %.critedge, %9, %11, %12, %lean_dec.exit
diff --git a/bench/lean4/optimized/External.ll b/bench/lean4/optimized/External.ll
index a5461d3b221..b680ba398fc 100644
--- a/bench/lean4/optimized/External.ll
+++ b/bench/lean4/optimized/External.ll
@@ -6676,56 +6676,44 @@ lean_dec.exit:                                    ; preds = %15, %14, %12, %2
 define ptr @l_Array_mapMUnsafe_map___at_Lean_Elab_Tactic_BVDecide_External_ModelParser_parsePartialAssignment___spec__2___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) local_unnamed_addr #0 {
   %4 = getelementptr i8, ptr %0, i64 8
   %.val = load i64, ptr %4, align 8, !tbaa !12
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit7
+  %5 = load i32, ptr %0, align 8, !tbaa !8
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !11
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %0, align 4, !tbaa !8
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !11
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !8
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !8
   br label %lean_dec.exit7
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit7, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #5
   br label %lean_dec.exit7
 
-lean_dec.exit7:                                   ; preds = %13, %12, %10, %3
+lean_dec.exit7:                                   ; preds = %13, %12, %7
   %14 = getelementptr i8, ptr %1, i64 8
   %.val10 = load i64, ptr %14, align 8, !tbaa !12
-  %15 = ptrtoint ptr %1 to i64
-  %16 = and i64 %15, 1
-  %.not11 = icmp eq i64 %16, 0
-  br i1 %.not11, label %17, label %lean_dec.exit
-
-17:                                               ; preds = %lean_dec.exit7
-  %18 = load i32, ptr %1, align 4, !tbaa !8
-  %19 = icmp sgt i32 %18, 1
-  br i1 %19, label %20, label %22, !prof !11
+  %12 = load i32, ptr %1, align 8, !tbaa !8
+  %13 = icmp sgt i32 %12, 1
+  br i1 %13, label %14, label %16, !prof !11
 
-20:                                               ; preds = %17
-  %21 = add nsw i32 %18, -1
+20:                                               ; preds = %lean_dec.exit7
+  %21 = add nsw i32 %12, -1
   store i32 %21, ptr %1, align 4, !tbaa !8
   br label %lean_dec.exit
 
-22:                                               ; preds = %17
-  %.not.i8 = icmp eq i32 %18, 0
+22:                                               ; preds = %lean_dec.exit7
+  %.not.i8 = icmp eq i32 %12, 0
   br i1 %.not.i8, label %lean_dec.exit, label %23
 
 23:                                               ; preds = %22
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #5
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %23, %22, %20, %lean_dec.exit7
+lean_dec.exit:                                    ; preds = %23, %22, %20
   %24 = tail call ptr @l_Array_mapMUnsafe_map___at_Lean_Elab_Tactic_BVDecide_External_ModelParser_parsePartialAssignment___spec__2(i64 noundef %.val, i64 noundef %.val10, ptr noundef %2)
   ret ptr %24
 }
diff --git a/bench/lean4/optimized/FVarSubset.ll b/bench/lean4/optimized/FVarSubset.ll
index c2f4d3d8f36..a500638e77d 100644
--- a/bench/lean4/optimized/FVarSubset.ll
+++ b/bench/lean4/optimized/FVarSubset.ll
@@ -319,30 +319,24 @@ lean_alloc_closure.exit:                          ; preds = %28
   %38 = getelementptr inbounds nuw i8, ptr %31, i64 24
   store ptr %30, ptr %38, align 8, !tbaa !4
   %39 = tail call ptr @lean_find_ext_expr(ptr noundef nonnull %31, ptr noundef %0) #3
-  %40 = ptrtoint ptr %31 to i64
-  %41 = and i64 %40, 1
-  %.not = icmp eq i64 %41, 0
-  br i1 %.not, label %42, label %lean_dec.exit19
+  %40 = load i32, ptr %31, align 8, !tbaa !8
+  %41 = icmp sgt i32 %40, 1
+  br i1 %41, label %42, label %44, !prof !11
 
 42:                                               ; preds = %lean_alloc_closure.exit
-  %43 = load i32, ptr %31, align 4, !tbaa !8
-  %44 = icmp sgt i32 %43, 1
-  br i1 %44, label %45, label %47, !prof !11
-
-45:                                               ; preds = %42
-  %46 = add nsw i32 %43, -1
-  store i32 %46, ptr %31, align 4, !tbaa !8
+  %43 = add nsw i32 %40, -1
+  store i32 %43, ptr %31, align 4, !tbaa !8
   br label %lean_dec.exit19
 
-47:                                               ; preds = %42
-  %.not.i24 = icmp eq i32 %43, 0
+47:                                               ; preds = %lean_alloc_closure.exit
+  %.not.i24 = icmp eq i32 %40, 0
   br i1 %.not.i24, label %lean_dec.exit19, label %48
 
 48:                                               ; preds = %47
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %31) #3
   br label %lean_dec.exit19
 
-lean_dec.exit19:                                  ; preds = %48, %47, %45, %lean_alloc_closure.exit
+lean_dec.exit19:                                  ; preds = %48, %47, %42
   %49 = ptrtoint ptr %39 to i64
   %50 = and i64 %49, 1
   %.not.i28 = icmp eq i64 %50, 0
@@ -379,7 +373,7 @@ lean_obj_tag.exit.thread:                         ; preds = %lean_dec.exit19
   br label %lean_dec.exit21
 
 lean_dec.exit21:                                  ; preds = %lean_obj_tag.exit, %57, %59, %60, %lean_obj_tag.exit.thread, %18, %24, %26, %27, %5, %11, %13, %14
-  %.0 = phi i8 [ 1, %14 ], [ 1, %13 ], [ 1, %11 ], [ 1, %5 ], [ 0, %27 ], [ 0, %26 ], [ 0, %24 ], [ 0, %18 ], [ 1, %lean_obj_tag.exit.thread ], [ 0, %60 ], [ 0, %59 ], [ 0, %57 ], [ %spec.select, %lean_obj_tag.exit ]
+  %.0 = phi i8 [ 1, %14 ], [ 1, %13 ], [ 1, %11 ], [ 1, %5 ], [ 0, %27 ], [ 0, %26 ], [ 0, %24 ], [ 0, %18 ], [ 1, %lean_obj_tag.exit.thread ], [ 0, %57 ], [ 0, %56 ], [ 0, %54 ], [ %spec.select, %lean_obj_tag.exit ]
   ret i8 %.0
 }
 
diff --git a/bench/lean4/optimized/FVarSubst.ll b/bench/lean4/optimized/FVarSubst.ll
index 4d610c97790..b743d72aa72 100644
--- a/bench/lean4/optimized/FVarSubst.ll
+++ b/bench/lean4/optimized/FVarSubst.ll
@@ -1529,23 +1529,17 @@ lean_alloc_closure.exit:                          ; preds = %26
   %34 = getelementptr inbounds nuw i8, ptr %27, i64 24
   store ptr %0, ptr %34, align 8, !tbaa !10
   %35 = tail call ptr @lean_replace_expr(ptr noundef nonnull %27, ptr noundef %1) #3
-  %36 = ptrtoint ptr %27 to i64
-  %37 = and i64 %36, 1
-  %.not28 = icmp eq i64 %37, 0
-  br i1 %.not28, label %38, label %lean_inc.exit18
+  %36 = load i32, ptr %27, align 8, !tbaa !4
+  %37 = icmp sgt i32 %36, 1
+  br i1 %37, label %38, label %40, !prof !9
 
 38:                                               ; preds = %lean_alloc_closure.exit
-  %39 = load i32, ptr %27, align 4, !tbaa !4
-  %40 = icmp sgt i32 %39, 1
-  br i1 %40, label %41, label %43, !prof !9
-
-41:                                               ; preds = %38
-  %42 = add nsw i32 %39, -1
-  store i32 %42, ptr %27, align 4, !tbaa !4
+  %39 = add nsw i32 %36, -1
+  store i32 %39, ptr %27, align 4, !tbaa !4
   br label %lean_inc.exit18
 
-43:                                               ; preds = %38
-  %.not.i19 = icmp eq i32 %39, 0
+43:                                               ; preds = %lean_alloc_closure.exit
+  %.not.i19 = icmp eq i32 %36, 0
   br i1 %.not.i19, label %lean_inc.exit18, label %44
 
 44:                                               ; preds = %43
@@ -1600,8 +1594,8 @@ lean_dec.exit:                                    ; preds = %54, %53, %51, %45
   tail call void @lean_inc_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_inc.exit18
 
-lean_inc.exit18:                                  ; preds = %lean_dec.exit, %59, %61, %62, %lean_alloc_closure.exit, %41, %43, %44, %lean_dec.exit17, %22, %24, %25
-  %.1 = phi ptr [ %1, %25 ], [ %1, %24 ], [ %1, %22 ], [ %1, %lean_dec.exit17 ], [ %35, %44 ], [ %35, %43 ], [ %35, %41 ], [ %35, %lean_alloc_closure.exit ], [ %1, %62 ], [ %1, %61 ], [ %1, %59 ], [ %1, %lean_dec.exit ]
+lean_inc.exit18:                                  ; preds = %lean_dec.exit, %59, %61, %62, %38, %43, %44, %lean_dec.exit17, %22, %24, %25
+  %.1 = phi ptr [ %1, %25 ], [ %1, %24 ], [ %1, %22 ], [ %1, %lean_dec.exit17 ], [ %35, %41 ], [ %35, %40 ], [ %35, %38 ], [ %1, %59 ], [ %1, %58 ], [ %1, %56 ], [ %1, %lean_dec.exit ]
   ret ptr %.1
 }
 
diff --git a/bench/lean4/optimized/FilePath.ll b/bench/lean4/optimized/FilePath.ll
index a6a759d10d2..b5424a96454 100644
--- a/bench/lean4/optimized/FilePath.ll
+++ b/bench/lean4/optimized/FilePath.ll
@@ -767,30 +767,24 @@ lean_inc.exit:
   %4 = add i64 %3, -1
   %5 = inttoptr i64 %4 to ptr
   %6 = tail call ptr @l_Lake_modOfFilePath_removeExts(ptr noundef %1, ptr noundef nonnull %5, ptr noundef nonnull %5)
-  %7 = ptrtoint ptr %1 to i64
-  %8 = and i64 %7, 1
-  %.not = icmp eq i64 %8, 0
-  br i1 %.not, label %9, label %lean_dec.exit
+  %7 = load i32, ptr %1, align 8, !tbaa !8
+  %8 = icmp sgt i32 %7, 1
+  br i1 %8, label %9, label %11, !prof !11
 
 9:                                                ; preds = %lean_inc.exit
-  %10 = load i32, ptr %1, align 4, !tbaa !8
-  %11 = icmp sgt i32 %10, 1
-  br i1 %11, label %12, label %14, !prof !11
-
-12:                                               ; preds = %9
-  %13 = add nsw i32 %10, -1
-  store i32 %13, ptr %1, align 4, !tbaa !8
+  %10 = add nsw i32 %7, -1
+  store i32 %10, ptr %1, align 4, !tbaa !8
   br label %lean_dec.exit
 
-14:                                               ; preds = %9
-  %.not.i = icmp eq i32 %10, 0
+14:                                               ; preds = %lean_inc.exit
+  %.not.i = icmp eq i32 %7, 0
   br i1 %.not.i, label %lean_dec.exit, label %15
 
 15:                                               ; preds = %14
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %15, %14, %12, %lean_inc.exit
+lean_dec.exit:                                    ; preds = %15, %14, %9
   %16 = load ptr, ptr @l_Lake_joinRelative___closed__3, align 8, !tbaa !13
   %17 = tail call ptr @l_String_stripSuffix(ptr noundef %6, ptr noundef %16) #3
   %18 = tail call ptr @l_System_FilePath_components(ptr noundef %17) #3
diff --git a/bench/lean4/optimized/FindLevelMVar.ll b/bench/lean4/optimized/FindLevelMVar.ll
index 8ee7631e28b..814e2d204d9 100644
--- a/bench/lean4/optimized/FindLevelMVar.ll
+++ b/bench/lean4/optimized/FindLevelMVar.ll
@@ -2246,56 +2246,44 @@ lean_dec.exit45:                                  ; preds = %lean_dec.exit39, %3
 define ptr @l_Array_foldrMUnsafe_fold___at_Lean_FindLevelMVar_main___spec__2___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5) local_unnamed_addr #0 {
   %7 = getelementptr i8, ptr %2, i64 8
   %.val = load i64, ptr %7, align 8, !tbaa !14
-  %8 = ptrtoint ptr %2 to i64
-  %9 = and i64 %8, 1
-  %.not = icmp eq i64 %9, 0
-  br i1 %.not, label %10, label %lean_dec.exit12
+  %8 = load i32, ptr %2, align 8, !tbaa !4
+  %9 = icmp sgt i32 %8, 1
+  br i1 %9, label %10, label %12, !prof !9
 
 10:                                               ; preds = %6
-  %11 = load i32, ptr %2, align 4, !tbaa !4
-  %12 = icmp sgt i32 %11, 1
-  br i1 %12, label %13, label %15, !prof !9
-
-13:                                               ; preds = %10
-  %14 = add nsw i32 %11, -1
-  store i32 %14, ptr %2, align 4, !tbaa !4
+  %11 = add nsw i32 %8, -1
+  store i32 %11, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit12
 
-15:                                               ; preds = %10
-  %.not.i = icmp eq i32 %11, 0
+15:                                               ; preds = %6
+  %.not.i = icmp eq i32 %8, 0
   br i1 %.not.i, label %lean_dec.exit12, label %16
 
 16:                                               ; preds = %15
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #3
   br label %lean_dec.exit12
 
-lean_dec.exit12:                                  ; preds = %16, %15, %13, %6
+lean_dec.exit12:                                  ; preds = %16, %15, %10
   %17 = getelementptr i8, ptr %3, i64 8
   %.val17 = load i64, ptr %17, align 8, !tbaa !14
-  %18 = ptrtoint ptr %3 to i64
-  %19 = and i64 %18, 1
-  %.not18 = icmp eq i64 %19, 0
-  br i1 %.not18, label %20, label %lean_dec.exit11
-
-20:                                               ; preds = %lean_dec.exit12
-  %21 = load i32, ptr %3, align 4, !tbaa !4
-  %22 = icmp sgt i32 %21, 1
-  br i1 %22, label %23, label %25, !prof !9
-
-23:                                               ; preds = %20
-  %24 = add nsw i32 %21, -1
-  store i32 %24, ptr %3, align 4, !tbaa !4
+  %15 = load i32, ptr %3, align 8, !tbaa !4
+  %16 = icmp sgt i32 %15, 1
+  br i1 %16, label %17, label %19, !prof !9
+
+17:                                               ; preds = %lean_dec.exit12
+  %18 = add nsw i32 %15, -1
+  store i32 %18, ptr %3, align 4, !tbaa !4
   br label %lean_dec.exit11
 
-25:                                               ; preds = %20
-  %.not.i13 = icmp eq i32 %21, 0
+25:                                               ; preds = %lean_dec.exit12
+  %.not.i13 = icmp eq i32 %15, 0
   br i1 %.not.i13, label %lean_dec.exit11, label %26
 
 26:                                               ; preds = %25
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %3) #3
   br label %lean_dec.exit11
 
-lean_dec.exit11:                                  ; preds = %26, %25, %23, %lean_dec.exit12
+lean_dec.exit11:                                  ; preds = %26, %25, %17
   %27 = tail call ptr @l_Array_foldrMUnsafe_fold___at_Lean_FindLevelMVar_main___spec__2(ptr noundef %0, ptr noundef %1, i64 noundef %.val, i64 noundef %.val17, ptr noundef %4, ptr noundef %5)
   %28 = ptrtoint ptr %1 to i64
   %29 = and i64 %28, 1
@@ -2328,56 +2316,44 @@ lean_dec.exit:                                    ; preds = %36, %35, %33, %lean
 define ptr @l_Array_foldrMUnsafe_fold___at_Lean_FindLevelMVar_main___spec__3___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5) local_unnamed_addr #0 {
   %7 = getelementptr i8, ptr %2, i64 8
   %.val17 = load i64, ptr %7, align 8, !tbaa !14
-  %8 = ptrtoint ptr %2 to i64
-  %9 = and i64 %8, 1
-  %.not = icmp eq i64 %9, 0
-  br i1 %.not, label %10, label %lean_dec.exit12
+  %8 = load i32, ptr %2, align 8, !tbaa !4
+  %9 = icmp sgt i32 %8, 1
+  br i1 %9, label %10, label %12, !prof !9
 
 10:                                               ; preds = %6
-  %11 = load i32, ptr %2, align 4, !tbaa !4
-  %12 = icmp sgt i32 %11, 1
-  br i1 %12, label %13, label %15, !prof !9
-
-13:                                               ; preds = %10
-  %14 = add nsw i32 %11, -1
-  store i32 %14, ptr %2, align 4, !tbaa !4
+  %11 = add nsw i32 %8, -1
+  store i32 %11, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit12
 
-15:                                               ; preds = %10
-  %.not.i = icmp eq i32 %11, 0
+15:                                               ; preds = %6
+  %.not.i = icmp eq i32 %8, 0
   br i1 %.not.i, label %lean_dec.exit12, label %16
 
 16:                                               ; preds = %15
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #3
   br label %lean_dec.exit12
 
-lean_dec.exit12:                                  ; preds = %16, %15, %13, %6
+lean_dec.exit12:                                  ; preds = %16, %15, %10
   %17 = getelementptr i8, ptr %3, i64 8
   %.val = load i64, ptr %17, align 8, !tbaa !14
-  %18 = ptrtoint ptr %3 to i64
-  %19 = and i64 %18, 1
-  %.not18 = icmp eq i64 %19, 0
-  br i1 %.not18, label %20, label %lean_dec.exit11
-
-20:                                               ; preds = %lean_dec.exit12
-  %21 = load i32, ptr %3, align 4, !tbaa !4
-  %22 = icmp sgt i32 %21, 1
-  br i1 %22, label %23, label %25, !prof !9
-
-23:                                               ; preds = %20
-  %24 = add nsw i32 %21, -1
-  store i32 %24, ptr %3, align 4, !tbaa !4
+  %15 = load i32, ptr %3, align 8, !tbaa !4
+  %16 = icmp sgt i32 %15, 1
+  br i1 %16, label %17, label %19, !prof !9
+
+17:                                               ; preds = %lean_dec.exit12
+  %18 = add nsw i32 %15, -1
+  store i32 %18, ptr %3, align 4, !tbaa !4
   br label %lean_dec.exit11
 
-25:                                               ; preds = %20
-  %.not.i13 = icmp eq i32 %21, 0
+25:                                               ; preds = %lean_dec.exit12
+  %.not.i13 = icmp eq i32 %15, 0
   br i1 %.not.i13, label %lean_dec.exit11, label %26
 
 26:                                               ; preds = %25
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %3) #3
   br label %lean_dec.exit11
 
-lean_dec.exit11:                                  ; preds = %26, %25, %23, %lean_dec.exit12
+lean_dec.exit11:                                  ; preds = %26, %25, %17
   %27 = tail call ptr @l_Array_foldrMUnsafe_fold___at_Lean_FindLevelMVar_main___spec__3(ptr noundef %0, ptr noundef %1, i64 noundef %.val17, i64 noundef %.val, ptr noundef %4, ptr noundef %5)
   %28 = ptrtoint ptr %1 to i64
   %29 = and i64 %28, 1
diff --git a/bench/lean4/optimized/ForEachExprWhere.ll b/bench/lean4/optimized/ForEachExprWhere.ll
index 941ae5dc21a..a68a80c580a 100644
--- a/bench/lean4/optimized/ForEachExprWhere.ll
+++ b/bench/lean4/optimized/ForEachExprWhere.ll
@@ -675,30 +675,24 @@ lean_dec.exit:                                    ; preds = %134, %133, %131, %l
 define noalias noundef nonnull ptr @l_Lean_ForEachExprWhere_visited___rarg___lambda__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) #0 {
   %4 = getelementptr i8, ptr %0, i64 8
   %.val = load i64, ptr %4, align 8, !tbaa !13
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit
+  %5 = load i32, ptr %0, align 8, !tbaa !4
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !11
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %0, align 4, !tbaa !4
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !11
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !4
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !4
   br label %lean_dec.exit
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %13, %12, %10, %3
+lean_dec.exit:                                    ; preds = %13, %12, %7
   %14 = tail call ptr @l_Lean_ForEachExprWhere_visited___rarg___lambda__1(i64 noundef %.val, ptr noundef %1, ptr noundef %2)
   ret ptr %14
 }
@@ -3582,58 +3576,46 @@ lean_dec.exit:                                    ; preds = %161, %160, %158, %l
 define noalias noundef nonnull ptr @l_Lean_ForEachExprWhere_checked___rarg___lambda__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3) #0 {
   %5 = getelementptr i8, ptr %0, i64 8
   %.val11 = load i64, ptr %5, align 8, !tbaa !13
-  %6 = ptrtoint ptr %0 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit8
+  %6 = load i32, ptr %0, align 8, !tbaa !4
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !11
 
 8:                                                ; preds = %4
-  %9 = load i32, ptr %0, align 4, !tbaa !4
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !11
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %0, align 4, !tbaa !4
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %0, align 4, !tbaa !4
   br label %lean_dec.exit8
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %4
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit8, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_dec.exit8
 
-lean_dec.exit8:                                   ; preds = %14, %13, %11, %4
+lean_dec.exit8:                                   ; preds = %14, %13, %8
   %15 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %15, align 8, !tbaa !13
-  %16 = ptrtoint ptr %1 to i64
-  %17 = and i64 %16, 1
-  %.not12 = icmp eq i64 %17, 0
-  br i1 %.not12, label %18, label %lean_dec.exit
-
-18:                                               ; preds = %lean_dec.exit8
-  %19 = load i32, ptr %1, align 4, !tbaa !4
-  %20 = icmp sgt i32 %19, 1
-  br i1 %20, label %21, label %23, !prof !11
+  %13 = load i32, ptr %1, align 8, !tbaa !4
+  %14 = icmp sgt i32 %13, 1
+  br i1 %14, label %15, label %17, !prof !11
 
-21:                                               ; preds = %18
-  %22 = add nsw i32 %19, -1
+21:                                               ; preds = %lean_dec.exit8
+  %22 = add nsw i32 %13, -1
   store i32 %22, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit
 
-23:                                               ; preds = %18
-  %.not.i9 = icmp eq i32 %19, 0
+23:                                               ; preds = %lean_dec.exit8
+  %.not.i9 = icmp eq i32 %13, 0
   br i1 %.not.i9, label %lean_dec.exit, label %24
 
 24:                                               ; preds = %23
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %24, %23, %21, %lean_dec.exit8
+lean_dec.exit:                                    ; preds = %24, %23, %21
   %25 = tail call ptr @l_Lean_ForEachExprWhere_checked___rarg___lambda__1(i64 noundef %.val11, i64 noundef %.val, ptr noundef %2, ptr noundef %3)
-  ret ptr %25
+  ret ptr %112
 }
 
 ; Function Attrs: nounwind uwtable
diff --git a/bench/lean4/optimized/GeneralizeTelescope.ll b/bench/lean4/optimized/GeneralizeTelescope.ll
index ccc89cf2394..dd40718305b 100644
--- a/bench/lean4/optimized/GeneralizeTelescope.ll
+++ b/bench/lean4/optimized/GeneralizeTelescope.ll
@@ -5586,56 +5586,44 @@ lean_alloc_closure.exit:                          ; preds = %1
 define ptr @l_Array_mapMUnsafe_map___at_Lean_Meta_GeneralizeTelescope_generalizeTelescopeAux___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) local_unnamed_addr #0 {
   %4 = getelementptr i8, ptr %0, i64 8
   %.val = load i64, ptr %4, align 8, !tbaa !4
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit7
+  %5 = load i32, ptr %0, align 8, !tbaa !9
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !12
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %0, align 4, !tbaa !9
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !12
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !9
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !12
   br label %lean_dec.exit7
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit7, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_dec.exit7
 
-lean_dec.exit7:                                   ; preds = %13, %12, %10, %3
+lean_dec.exit7:                                   ; preds = %13, %12, %7
   %14 = getelementptr i8, ptr %1, i64 8
   %.val10 = load i64, ptr %14, align 8, !tbaa !4
-  %15 = ptrtoint ptr %1 to i64
-  %16 = and i64 %15, 1
-  %.not11 = icmp eq i64 %16, 0
-  br i1 %.not11, label %17, label %lean_dec.exit
-
-17:                                               ; preds = %lean_dec.exit7
-  %18 = load i32, ptr %1, align 4, !tbaa !9
-  %19 = icmp sgt i32 %18, 1
-  br i1 %19, label %20, label %22, !prof !12
+  %12 = load i32, ptr %1, align 8, !tbaa !9
+  %13 = icmp sgt i32 %12, 1
+  br i1 %13, label %14, label %16, !prof !12
 
-20:                                               ; preds = %17
-  %21 = add nsw i32 %18, -1
+20:                                               ; preds = %lean_dec.exit7
+  %21 = add nsw i32 %12, -1
   store i32 %21, ptr %1, align 4, !tbaa !9
   br label %lean_dec.exit
 
-22:                                               ; preds = %17
-  %.not.i8 = icmp eq i32 %18, 0
+22:                                               ; preds = %lean_dec.exit7
+  %.not.i8 = icmp eq i32 %12, 0
   br i1 %.not.i8, label %lean_dec.exit, label %23
 
 23:                                               ; preds = %22
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %23, %22, %20, %lean_dec.exit7
+lean_dec.exit:                                    ; preds = %23, %22, %20
   %24 = tail call ptr @l_Array_mapMUnsafe_map___at_Lean_Meta_GeneralizeTelescope_generalizeTelescopeAux___spec__1(i64 noundef %.val, i64 noundef %.val10, ptr noundef %2)
   ret ptr %24
 }
@@ -6883,22 +6871,16 @@ lean_alloc_closure.exit:                          ; preds = %1
 define ptr @l_Array_mapMUnsafe_map___at_Lean_Meta_generalizeTelescope___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7) local_unnamed_addr #0 {
   %9 = getelementptr i8, ptr %0, i64 8
   %.val15 = load i64, ptr %9, align 8, !tbaa !4
-  %10 = ptrtoint ptr %0 to i64
-  %11 = and i64 %10, 1
-  %.not = icmp eq i64 %11, 0
-  br i1 %.not, label %12, label %lean_dec.exit12
+  %10 = load i32, ptr %0, align 8, !tbaa !9
+  %11 = icmp sgt i32 %10, 1
+  br i1 %11, label %12, label %20, !prof !12
 
 12:                                               ; preds = %8
-  %13 = load i32, ptr %0, align 4, !tbaa !9
-  %14 = icmp sgt i32 %13, 1
-  br i1 %14, label %15, label %17, !prof !12
-
-15:                                               ; preds = %12
-  %16 = add nsw i32 %13, -1
-  store i32 %16, ptr %0, align 4, !tbaa !9
+  %13 = add nsw i32 %10, -1
+  store i32 %13, ptr %0, align 4, !tbaa !12
   br label %lean_dec.exit12
 
-17:                                               ; preds = %12
+17:                                               ; preds = %8
   %.not.i = icmp eq i32 %13, 0
   br i1 %.not.i, label %lean_dec.exit12, label %18
 
@@ -6906,25 +6888,19 @@ define ptr @l_Array_mapMUnsafe_map___at_Lean_Meta_generalizeTelescope___spec__1_
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_dec.exit12
 
-lean_dec.exit12:                                  ; preds = %18, %17, %15, %8
+lean_dec.exit12:                                  ; preds = %18, %17, %12
   %19 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %19, align 8, !tbaa !4
-  %20 = ptrtoint ptr %1 to i64
-  %21 = and i64 %20, 1
-  %.not16 = icmp eq i64 %21, 0
-  br i1 %.not16, label %22, label %lean_dec.exit
-
-22:                                               ; preds = %lean_dec.exit12
-  %23 = load i32, ptr %1, align 4, !tbaa !9
-  %24 = icmp sgt i32 %23, 1
-  br i1 %24, label %25, label %27, !prof !12
-
-25:                                               ; preds = %22
-  %26 = add nsw i32 %23, -1
-  store i32 %26, ptr %1, align 4, !tbaa !9
+  %17 = load i32, ptr %1, align 8, !tbaa !9
+  %18 = icmp sgt i32 %23, 1
+  br i1 %18, label %19, label %21, !prof !12
+
+19:                                               ; preds = %lean_dec.exit12
+  %20 = add nsw i32 %17, -1
+  store i32 %20, ptr %1, align 4, !tbaa !9
   br label %lean_dec.exit
 
-27:                                               ; preds = %22
+27:                                               ; preds = %lean_dec.exit12
   %.not.i13 = icmp eq i32 %23, 0
   br i1 %.not.i13, label %lean_dec.exit, label %28
 
@@ -6932,7 +6908,7 @@ lean_dec.exit12:                                  ; preds = %18, %17, %15, %8
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %28, %27, %25, %lean_dec.exit12
+lean_dec.exit:                                    ; preds = %28, %27, %112
   %29 = tail call ptr @l_Array_mapMUnsafe_map___at_Lean_Meta_generalizeTelescope___spec__1(i64 noundef %.val15, i64 noundef %.val, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7)
   ret ptr %29
 }
diff --git a/bench/lean4/optimized/HasConstCache.ll b/bench/lean4/optimized/HasConstCache.ll
index 6fbdc860109..1ccebf3001b 100644
--- a/bench/lean4/optimized/HasConstCache.ll
+++ b/bench/lean4/optimized/HasConstCache.ll
@@ -5000,56 +5000,44 @@ lean_alloc_ctor.exit610:                          ; preds = %lean_dec.exit
 define noundef nonnull ptr @l_Array_anyMUnsafe_any___at_Lean_HasConstCache_containsUnsafe___spec__2___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3) local_unnamed_addr #1 {
   %5 = getelementptr i8, ptr %2, i64 8
   %.val = load i64, ptr %5, align 8, !tbaa !11
-  %6 = ptrtoint ptr %2 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit13
+  %6 = load i32, ptr %2, align 8, !tbaa !8
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !13
 
 8:                                                ; preds = %4
-  %9 = load i32, ptr %2, align 4, !tbaa !8
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !13
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %2, align 4, !tbaa !8
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %2, align 4, !tbaa !8
   br label %lean_dec.exit13
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %4
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit13, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #4
   br label %lean_dec.exit13
 
-lean_dec.exit13:                                  ; preds = %14, %13, %11, %4
+lean_dec.exit13:                                  ; preds = %14, %13, %8
   %15 = getelementptr i8, ptr %3, i64 8
   %.val20 = load i64, ptr %15, align 8, !tbaa !11
-  %16 = ptrtoint ptr %3 to i64
-  %17 = and i64 %16, 1
-  %.not22 = icmp eq i64 %17, 0
-  br i1 %.not22, label %18, label %lean_dec.exit12
-
-18:                                               ; preds = %lean_dec.exit13
-  %19 = load i32, ptr %3, align 4, !tbaa !8
-  %20 = icmp sgt i32 %19, 1
-  br i1 %20, label %21, label %23, !prof !13
-
-21:                                               ; preds = %18
-  %22 = add nsw i32 %19, -1
-  store i32 %22, ptr %3, align 4, !tbaa !8
+  %13 = load i32, ptr %3, align 8, !tbaa !8
+  %14 = icmp sgt i32 %13, 1
+  br i1 %14, label %15, label %17, !prof !13
+
+15:                                               ; preds = %lean_dec.exit13
+  %16 = add nsw i32 %13, -1
+  store i32 %16, ptr %3, align 4, !tbaa !8
   br label %lean_dec.exit12
 
-23:                                               ; preds = %18
-  %.not.i14 = icmp eq i32 %19, 0
+23:                                               ; preds = %lean_dec.exit13
+  %.not.i14 = icmp eq i32 %13, 0
   br i1 %.not.i14, label %lean_dec.exit12, label %24
 
 24:                                               ; preds = %23
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %3) #4
   br label %lean_dec.exit12
 
-lean_dec.exit12:                                  ; preds = %24, %23, %21, %lean_dec.exit13
+lean_dec.exit12:                                  ; preds = %24, %23, %15
   %.not25.i = icmp eq i64 %.val, %.val20
   br i1 %.not25.i, label %l_Array_anyMUnsafe_any___at_Lean_HasConstCache_containsUnsafe___spec__2.exit, label %.lr.ph.i
 
@@ -5063,7 +5051,7 @@ lean_dec.exit12:                                  ; preds = %24, %23, %21, %lean
   br i1 %.not.i21, label %l_Array_anyMUnsafe_any___at_Lean_HasConstCache_containsUnsafe___spec__2.exit, label %28
 
 28:                                               ; preds = %26, %.lr.ph.i
-  %.01526.i = phi i64 [ %.val, %.lr.ph.i ], [ %27, %26 ]
+  %.01526.i = phi i64 [ %.val, %.lr.ph.i ], [ %27, %20 ]
   %29 = getelementptr inbounds nuw ptr, ptr %25, i64 %.01526.i
   %30 = load ptr, ptr %29, align 8, !tbaa !4
   %31 = ptrtoint ptr %30 to i64
@@ -5113,12 +5101,12 @@ lean_array_uget.exit.i:                           ; preds = %28
   br label %lean_dec.exit.i
 
 lean_dec.exit.i:                                  ; preds = %47, %46, %44, %lean_array_uget.exit.i
-  %48 = phi i8 [ %39, %lean_array_uget.exit.i ], [ %41, %44 ], [ %41, %46 ], [ %41, %47 ]
+  %48 = phi i8 [ %39, %lean_array_uget.exit.i ], [ %41, %38 ], [ %41, %40 ], [ %41, %41 ]
   %49 = icmp eq i8 %48, 0
   br i1 %49, label %26, label %l_Array_anyMUnsafe_any___at_Lean_HasConstCache_containsUnsafe___spec__2.exit
 
 l_Array_anyMUnsafe_any___at_Lean_HasConstCache_containsUnsafe___spec__2.exit: ; preds = %26, %lean_dec.exit.i, %lean_dec.exit12
-  %.2.ph.i = phi i64 [ 1, %lean_dec.exit12 ], [ 1, %26 ], [ 3, %lean_dec.exit.i ]
+  %.2.ph.i = phi i64 [ 1, %lean_dec.exit12 ], [ 1, %20 ], [ 3, %lean_dec.exit.i ]
   %50 = ptrtoint ptr %1 to i64
   %51 = and i64 %50, 1
   %.not23 = icmp eq i64 %51, 0
diff --git a/bench/lean4/optimized/IndGroupInfo.ll b/bench/lean4/optimized/IndGroupInfo.ll
index c00e7e89859..aa08291790d 100644
--- a/bench/lean4/optimized/IndGroupInfo.ll
+++ b/bench/lean4/optimized/IndGroupInfo.ll
@@ -5415,56 +5415,44 @@ lean_alloc_ctor.exit:                             ; preds = %lean_dec.exit
 define ptr @l_Array_anyMUnsafe_any___at_Lean_Elab_Structural_IndGroupInst_isDefEq___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7) local_unnamed_addr #0 {
   %9 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %9, align 8, !tbaa !13
-  %10 = ptrtoint ptr %1 to i64
-  %11 = and i64 %10, 1
-  %.not = icmp eq i64 %11, 0
-  br i1 %.not, label %12, label %lean_dec.exit14
+  %10 = load i32, ptr %1, align 8, !tbaa !5
+  %11 = icmp sgt i32 %10, 1
+  br i1 %11, label %12, label %14, !prof !10
 
 12:                                               ; preds = %8
-  %13 = load i32, ptr %1, align 4, !tbaa !5
-  %14 = icmp sgt i32 %13, 1
-  br i1 %14, label %15, label %17, !prof !10
-
-15:                                               ; preds = %12
-  %16 = add nsw i32 %13, -1
-  store i32 %16, ptr %1, align 4, !tbaa !5
+  %13 = add nsw i32 %10, -1
+  store i32 %13, ptr %1, align 4, !tbaa !5
   br label %lean_dec.exit14
 
-17:                                               ; preds = %12
-  %.not.i = icmp eq i32 %13, 0
+17:                                               ; preds = %8
+  %.not.i = icmp eq i32 %10, 0
   br i1 %.not.i, label %lean_dec.exit14, label %18
 
 18:                                               ; preds = %17
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit14
 
-lean_dec.exit14:                                  ; preds = %18, %17, %15, %8
+lean_dec.exit14:                                  ; preds = %18, %17, %15
   %19 = getelementptr i8, ptr %2, i64 8
   %.val19 = load i64, ptr %19, align 8, !tbaa !13
-  %20 = ptrtoint ptr %2 to i64
-  %21 = and i64 %20, 1
-  %.not20 = icmp eq i64 %21, 0
-  br i1 %.not20, label %22, label %lean_dec.exit13
-
-22:                                               ; preds = %lean_dec.exit14
-  %23 = load i32, ptr %2, align 4, !tbaa !5
-  %24 = icmp sgt i32 %23, 1
-  br i1 %24, label %25, label %27, !prof !10
+  %17 = load i32, ptr %2, align 8, !tbaa !5
+  %18 = icmp sgt i32 %17, 1
+  br i1 %18, label %19, label %21, !prof !10
 
-25:                                               ; preds = %22
-  %26 = add nsw i32 %23, -1
+25:                                               ; preds = %lean_dec.exit14
+  %26 = add nsw i32 %17, -1
   store i32 %26, ptr %2, align 4, !tbaa !5
   br label %lean_dec.exit13
 
-27:                                               ; preds = %22
-  %.not.i15 = icmp eq i32 %23, 0
+27:                                               ; preds = %lean_dec.exit14
+  %.not.i15 = icmp eq i32 %17, 0
   br i1 %.not.i15, label %lean_dec.exit13, label %28
 
 28:                                               ; preds = %27
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #4
   br label %lean_dec.exit13
 
-lean_dec.exit13:                                  ; preds = %28, %27, %25, %lean_dec.exit14
+lean_dec.exit13:                                  ; preds = %28, %27, %25
   %29 = tail call ptr @l_Array_anyMUnsafe_any___at_Lean_Elab_Structural_IndGroupInst_isDefEq___spec__1(ptr noundef %0, i64 noundef %.val, i64 noundef %.val19, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7)
   %30 = ptrtoint ptr %0 to i64
   %31 = and i64 %30, 1
@@ -5798,30 +5786,24 @@ define nonnull ptr @l_Lean_Elab_Structural_IndGroupInst_isDefEq___lambda__4___bo
   %4 = getelementptr inbounds nuw i8, ptr %0, i64 16
   %5 = load ptr, ptr %4, align 8, !tbaa !11
   %6 = tail call zeroext i8 @l_Lean_Level_isEquiv(ptr noundef %3, ptr noundef %5) #4
-  %7 = ptrtoint ptr %0 to i64
-  %8 = and i64 %7, 1
-  %.not = icmp eq i64 %8, 0
-  br i1 %.not, label %9, label %lean_dec.exit
+  %7 = load i32, ptr %0, align 8, !tbaa !5
+  %8 = icmp sgt i32 %7, 1
+  br i1 %8, label %9, label %11, !prof !10
 
 9:                                                ; preds = %1
-  %10 = load i32, ptr %0, align 4, !tbaa !5
-  %11 = icmp sgt i32 %10, 1
-  br i1 %11, label %12, label %14, !prof !10
-
-12:                                               ; preds = %9
-  %13 = add nsw i32 %10, -1
-  store i32 %13, ptr %0, align 4, !tbaa !5
+  %10 = add nsw i32 %7, -1
+  store i32 %10, ptr %0, align 4, !tbaa !5
   br label %lean_dec.exit
 
-14:                                               ; preds = %9
-  %.not.i = icmp eq i32 %10, 0
+14:                                               ; preds = %1
+  %.not.i = icmp eq i32 %7, 0
   br i1 %.not.i, label %lean_dec.exit, label %15
 
 15:                                               ; preds = %14
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #4
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %15, %14, %12, %1
+lean_dec.exit:                                    ; preds = %15, %14, %9
   %16 = zext i8 %6 to i64
   %17 = shl nuw nsw i64 %16, 1
   %18 = or disjoint i64 %17, 1
@@ -9595,56 +9577,44 @@ lean_dec.exit:                                    ; preds = %17, %16, %14, %7
 define ptr @l_Array_mapMUnsafe_map___at_Lean_Elab_Structural_IndGroupInst_nestedTypeFormers___spec__5___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7) local_unnamed_addr #0 {
   %9 = getelementptr i8, ptr %0, i64 8
   %.val15 = load i64, ptr %9, align 8, !tbaa !13
-  %10 = ptrtoint ptr %0 to i64
-  %11 = and i64 %10, 1
-  %.not = icmp eq i64 %11, 0
-  br i1 %.not, label %12, label %lean_dec.exit12
+  %10 = load i32, ptr %0, align 8, !tbaa !5
+  %11 = icmp sgt i32 %10, 1
+  br i1 %11, label %12, label %14, !prof !10
 
 12:                                               ; preds = %8
-  %13 = load i32, ptr %0, align 4, !tbaa !5
-  %14 = icmp sgt i32 %13, 1
-  br i1 %14, label %15, label %17, !prof !10
-
-15:                                               ; preds = %12
-  %16 = add nsw i32 %13, -1
-  store i32 %16, ptr %0, align 4, !tbaa !5
+  %13 = add nsw i32 %10, -1
+  store i32 %13, ptr %0, align 4, !tbaa !5
   br label %lean_dec.exit12
 
-17:                                               ; preds = %12
-  %.not.i = icmp eq i32 %13, 0
+17:                                               ; preds = %8
+  %.not.i = icmp eq i32 %10, 0
   br i1 %.not.i, label %lean_dec.exit12, label %18
 
 18:                                               ; preds = %17
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #4
   br label %lean_dec.exit12
 
-lean_dec.exit12:                                  ; preds = %18, %17, %15, %8
+lean_dec.exit12:                                  ; preds = %18, %17, %15
   %19 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %19, align 8, !tbaa !13
-  %20 = ptrtoint ptr %1 to i64
-  %21 = and i64 %20, 1
-  %.not16 = icmp eq i64 %21, 0
-  br i1 %.not16, label %22, label %lean_dec.exit
-
-22:                                               ; preds = %lean_dec.exit12
-  %23 = load i32, ptr %1, align 4, !tbaa !5
-  %24 = icmp sgt i32 %23, 1
-  br i1 %24, label %25, label %27, !prof !10
+  %17 = load i32, ptr %1, align 8, !tbaa !5
+  %18 = icmp sgt i32 %17, 1
+  br i1 %18, label %19, label %21, !prof !10
 
-25:                                               ; preds = %22
-  %26 = add nsw i32 %23, -1
+25:                                               ; preds = %lean_dec.exit12
+  %26 = add nsw i32 %17, -1
   store i32 %26, ptr %1, align 4, !tbaa !5
   br label %lean_dec.exit
 
-27:                                               ; preds = %22
-  %.not.i13 = icmp eq i32 %23, 0
+27:                                               ; preds = %lean_dec.exit12
+  %.not.i13 = icmp eq i32 %17, 0
   br i1 %.not.i13, label %lean_dec.exit, label %28
 
 28:                                               ; preds = %27
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #4
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %28, %27, %25, %lean_dec.exit12
+lean_dec.exit:                                    ; preds = %28, %27, %25
   %29 = tail call ptr @l_Array_mapMUnsafe_map___at_Lean_Elab_Structural_IndGroupInst_nestedTypeFormers___spec__5(i64 noundef %.val15, i64 noundef %.val, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7)
   ret ptr %29
 }
diff --git a/bench/lean4/optimized/InitShutdown.ll b/bench/lean4/optimized/InitShutdown.ll
index 9a75cf29c63..b62090bc4d2 100644
--- a/bench/lean4/optimized/InitShutdown.ll
+++ b/bench/lean4/optimized/InitShutdown.ll
@@ -3287,56 +3287,44 @@ declare ptr @l___private_Lean_Data_Lsp_Capabilities_0__Lean_Lsp_toJsonClientCapa
 define ptr @l_Array_mapMUnsafe_map___at___private_Lean_Data_Lsp_InitShutdown_0__Lean_Lsp_toJsonInitializeParams____x40_Lean_Data_Lsp_InitShutdown___hyg_523____spec__5___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) local_unnamed_addr #0 {
   %4 = getelementptr i8, ptr %0, i64 8
   %.val = load i64, ptr %4, align 8, !tbaa !12
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit7
+  %5 = load i32, ptr %0, align 8, !tbaa !4
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !11
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %0, align 4, !tbaa !4
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !11
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !4
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !4
   br label %lean_dec.exit7
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit7, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #6
   br label %lean_dec.exit7
 
-lean_dec.exit7:                                   ; preds = %13, %12, %10, %3
+lean_dec.exit7:                                   ; preds = %13, %12, %7
   %14 = getelementptr i8, ptr %1, i64 8
   %.val10 = load i64, ptr %14, align 8, !tbaa !12
-  %15 = ptrtoint ptr %1 to i64
-  %16 = and i64 %15, 1
-  %.not11 = icmp eq i64 %16, 0
-  br i1 %.not11, label %17, label %lean_dec.exit
-
-17:                                               ; preds = %lean_dec.exit7
-  %18 = load i32, ptr %1, align 4, !tbaa !4
-  %19 = icmp sgt i32 %18, 1
-  br i1 %19, label %20, label %22, !prof !11
+  %12 = load i32, ptr %1, align 8, !tbaa !4
+  %13 = icmp sgt i32 %12, 1
+  br i1 %13, label %14, label %16, !prof !11
 
-20:                                               ; preds = %17
-  %21 = add nsw i32 %18, -1
+20:                                               ; preds = %lean_dec.exit7
+  %21 = add nsw i32 %12, -1
   store i32 %21, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit
 
-22:                                               ; preds = %17
-  %.not.i8 = icmp eq i32 %18, 0
+22:                                               ; preds = %lean_dec.exit7
+  %.not.i8 = icmp eq i32 %12, 0
   br i1 %.not.i8, label %lean_dec.exit, label %23
 
 23:                                               ; preds = %22
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #6
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %23, %22, %20, %lean_dec.exit7
+lean_dec.exit:                                    ; preds = %23, %22, %20
   %24 = tail call ptr @l_Array_mapMUnsafe_map___at___private_Lean_Data_Lsp_InitShutdown_0__Lean_Lsp_toJsonInitializeParams____x40_Lean_Data_Lsp_InitShutdown___hyg_523____spec__5(i64 noundef %.val, i64 noundef %.val10, ptr noundef %2)
   ret ptr %24
 }
@@ -11731,56 +11719,44 @@ lean_dec.exit:                                    ; preds = %12, %11, %9, %2
 define ptr @l_Array_mapMUnsafe_map___at_Lean_Lsp_instFromJsonInitializeParams___spec__8___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) local_unnamed_addr #0 {
   %4 = getelementptr i8, ptr %0, i64 8
   %.val10 = load i64, ptr %4, align 8, !tbaa !12
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit7
+  %5 = load i32, ptr %0, align 8, !tbaa !4
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !11
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %0, align 4, !tbaa !4
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !11
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !4
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !4
   br label %lean_dec.exit7
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit7, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #6
   br label %lean_dec.exit7
 
-lean_dec.exit7:                                   ; preds = %13, %12, %10, %3
+lean_dec.exit7:                                   ; preds = %13, %12, %7
   %14 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %14, align 8, !tbaa !12
-  %15 = ptrtoint ptr %1 to i64
-  %16 = and i64 %15, 1
-  %.not11 = icmp eq i64 %16, 0
-  br i1 %.not11, label %17, label %lean_dec.exit
-
-17:                                               ; preds = %lean_dec.exit7
-  %18 = load i32, ptr %1, align 4, !tbaa !4
-  %19 = icmp sgt i32 %18, 1
-  br i1 %19, label %20, label %22, !prof !11
+  %12 = load i32, ptr %1, align 8, !tbaa !4
+  %13 = icmp sgt i32 %12, 1
+  br i1 %13, label %14, label %16, !prof !11
 
-20:                                               ; preds = %17
-  %21 = add nsw i32 %18, -1
+20:                                               ; preds = %lean_dec.exit7
+  %21 = add nsw i32 %12, -1
   store i32 %21, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit
 
-22:                                               ; preds = %17
-  %.not.i8 = icmp eq i32 %18, 0
+22:                                               ; preds = %lean_dec.exit7
+  %.not.i8 = icmp eq i32 %12, 0
   br i1 %.not.i8, label %lean_dec.exit, label %23
 
 23:                                               ; preds = %22
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #6
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %23, %22, %20, %lean_dec.exit7
+lean_dec.exit:                                    ; preds = %23, %22, %20
   %24 = tail call ptr @l_Array_mapMUnsafe_map___at_Lean_Lsp_instFromJsonInitializeParams___spec__8(i64 noundef %.val10, i64 noundef %.val, ptr noundef %2)
   ret ptr %24
 }
diff --git a/bench/lean4/optimized/InlineCandidate.ll b/bench/lean4/optimized/InlineCandidate.ll
index ee2f949eb36..68486ab152a 100644
--- a/bench/lean4/optimized/InlineCandidate.ll
+++ b/bench/lean4/optimized/InlineCandidate.ll
@@ -62,30 +62,24 @@ define nonnull ptr @l_Lean_Compiler_LCNF_Simp_InlineCandidateInfo_arity___boxed(
   %3 = load ptr, ptr %2, align 8, !tbaa !4
   %4 = getelementptr i8, ptr %3, i64 8
   %.val.i = load i64, ptr %4, align 8, !tbaa !8
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit
+  %5 = load i32, ptr %0, align 8, !tbaa !10
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !13
 
 7:                                                ; preds = %1
-  %8 = load i32, ptr %0, align 4, !tbaa !10
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !13
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !10
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !10
   br label %lean_dec.exit
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %1
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #5
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %13, %12, %10, %1
+lean_dec.exit:                                    ; preds = %13, %12, %7
   %14 = shl i64 %.val.i, 1
   %15 = or disjoint i64 %14, 1
   %16 = inttoptr i64 %15 to ptr
diff --git a/bench/lean4/optimized/InputFileConfig.ll b/bench/lean4/optimized/InputFileConfig.ll
index 828b7c8223c..813976d602a 100644
--- a/bench/lean4/optimized/InputFileConfig.ll
+++ b/bench/lean4/optimized/InputFileConfig.ll
@@ -1449,56 +1449,44 @@ declare ptr @l_Lean_RBNode_insert___at_Lean_NameMap_insert___spec__1___rarg(ptr
 define ptr @l_Array_foldlMUnsafe_fold___at_Lake_InputFileConfig_instConfigMeta___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3) local_unnamed_addr #1 {
   %5 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %5, align 8, !tbaa !13
-  %6 = ptrtoint ptr %1 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit10
+  %6 = load i32, ptr %1, align 8, !tbaa !8
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !12
 
 8:                                                ; preds = %4
-  %9 = load i32, ptr %1, align 4, !tbaa !8
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !12
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %1, align 4, !tbaa !8
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %1, align 4, !tbaa !8
   br label %lean_dec.exit10
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %4
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit10, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #7
   br label %lean_dec.exit10
 
-lean_dec.exit10:                                  ; preds = %14, %13, %11, %4
+lean_dec.exit10:                                  ; preds = %14, %13, %8
   %15 = getelementptr i8, ptr %2, i64 8
   %.val15 = load i64, ptr %15, align 8, !tbaa !13
-  %16 = ptrtoint ptr %2 to i64
-  %17 = and i64 %16, 1
-  %.not17 = icmp eq i64 %17, 0
-  br i1 %.not17, label %18, label %lean_dec.exit9
+  %13 = load i32, ptr %2, align 8, !tbaa !8
+  %14 = icmp sgt i32 %13, 1
+  br i1 %14, label %15, label %17, !prof !12
 
-18:                                               ; preds = %lean_dec.exit10
-  %19 = load i32, ptr %2, align 4, !tbaa !8
-  %20 = icmp sgt i32 %19, 1
-  br i1 %20, label %21, label %23, !prof !12
-
-21:                                               ; preds = %18
-  %22 = add nsw i32 %19, -1
+21:                                               ; preds = %lean_dec.exit10
+  %22 = add nsw i32 %13, -1
   store i32 %22, ptr %2, align 4, !tbaa !8
   br label %lean_dec.exit9
 
-23:                                               ; preds = %18
-  %.not.i11 = icmp eq i32 %19, 0
+23:                                               ; preds = %lean_dec.exit10
+  %.not.i11 = icmp eq i32 %13, 0
   br i1 %.not.i11, label %lean_dec.exit9, label %24
 
 24:                                               ; preds = %23
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #7
   br label %lean_dec.exit9
 
-lean_dec.exit9:                                   ; preds = %24, %23, %21, %lean_dec.exit10
+lean_dec.exit9:                                   ; preds = %24, %23, %21
   %.not28.i = icmp eq i64 %.val, %.val15
   br i1 %.not28.i, label %l_Array_foldlMUnsafe_fold___at_Lake_InputFileConfig_instConfigMeta___spec__1.exit, label %.lr.ph.i
 
@@ -1507,8 +1495,8 @@ lean_dec.exit9:                                   ; preds = %24, %23, %21, %lean
   br label %26
 
 26:                                               ; preds = %47, %.lr.ph.i
-  %.01730.i = phi i64 [ %.val, %.lr.ph.i ], [ %49, %47 ]
-  %.01929.i = phi ptr [ %3, %.lr.ph.i ], [ %48, %47 ]
+  %.01730.i = phi i64 [ %.val, %.lr.ph.i ], [ %49, %41 ]
+  %.01929.i = phi ptr [ %3, %.lr.ph.i ], [ %48, %41 ]
   %27 = getelementptr inbounds nuw ptr, ptr %25, i64 %.01730.i
   %28 = load ptr, ptr %27, align 8, !tbaa !4
   %29 = ptrtoint ptr %28 to i64
@@ -1567,7 +1555,7 @@ lean_array_uget.exit.i:                           ; preds = %36, %35, %33, %26
   br i1 %.not.i16, label %l_Array_foldlMUnsafe_fold___at_Lake_InputFileConfig_instConfigMeta___spec__1.exit, label %26
 
 l_Array_foldlMUnsafe_fold___at_Lake_InputFileConfig_instConfigMeta___spec__1.exit: ; preds = %47, %lean_dec.exit9
-  %.019.lcssa.i = phi ptr [ %3, %lean_dec.exit9 ], [ %48, %47 ]
+  %.019.lcssa.i = phi ptr [ %3, %lean_dec.exit9 ], [ %48, %41 ]
   %50 = ptrtoint ptr %0 to i64
   %51 = and i64 %50, 1
   %.not18 = icmp eq i64 %51, 0
diff --git a/bench/lean4/optimized/InternalExceptionId.ll b/bench/lean4/optimized/InternalExceptionId.ll
index eeb157ba0c3..547265d81d3 100644
--- a/bench/lean4/optimized/InternalExceptionId.ll
+++ b/bench/lean4/optimized/InternalExceptionId.ll
@@ -1041,56 +1041,44 @@ declare ptr @lean_string_append(ptr noundef, ptr noundef) local_unnamed_addr #1
 define noundef nonnull ptr @l_Array_anyMUnsafe_any___at_Lean_registerInternalExceptionId___spec__2___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3) local_unnamed_addr #0 {
   %5 = getelementptr i8, ptr %2, i64 8
   %.val = load i64, ptr %5, align 8, !tbaa !13
-  %6 = ptrtoint ptr %2 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit13
+  %6 = load i32, ptr %2, align 8, !tbaa !5
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !10
 
 8:                                                ; preds = %4
-  %9 = load i32, ptr %2, align 4, !tbaa !5
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !10
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %2, align 4, !tbaa !5
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %2, align 4, !tbaa !5
   br label %lean_dec.exit13
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %4
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit13, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #4
   br label %lean_dec.exit13
 
-lean_dec.exit13:                                  ; preds = %14, %13, %11, %4
+lean_dec.exit13:                                  ; preds = %14, %13, %8
   %15 = getelementptr i8, ptr %3, i64 8
   %.val20 = load i64, ptr %15, align 8, !tbaa !13
-  %16 = ptrtoint ptr %3 to i64
-  %17 = and i64 %16, 1
-  %.not22 = icmp eq i64 %17, 0
-  br i1 %.not22, label %18, label %lean_dec.exit12
-
-18:                                               ; preds = %lean_dec.exit13
-  %19 = load i32, ptr %3, align 4, !tbaa !5
-  %20 = icmp sgt i32 %19, 1
-  br i1 %20, label %21, label %23, !prof !10
+  %13 = load i32, ptr %3, align 8, !tbaa !5
+  %14 = icmp sgt i32 %13, 1
+  br i1 %14, label %15, label %17, !prof !10
 
-21:                                               ; preds = %18
-  %22 = add nsw i32 %19, -1
+21:                                               ; preds = %lean_dec.exit13
+  %22 = add nsw i32 %13, -1
   store i32 %22, ptr %3, align 4, !tbaa !5
   br label %lean_dec.exit12
 
-23:                                               ; preds = %18
-  %.not.i14 = icmp eq i32 %19, 0
+23:                                               ; preds = %lean_dec.exit13
+  %.not.i14 = icmp eq i32 %13, 0
   br i1 %.not.i14, label %lean_dec.exit12, label %24
 
 24:                                               ; preds = %23
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %3) #4
   br label %lean_dec.exit12
 
-lean_dec.exit12:                                  ; preds = %24, %23, %21, %lean_dec.exit13
+lean_dec.exit12:                                  ; preds = %24, %23, %21
   %.not25.i = icmp eq i64 %.val, %.val20
   br i1 %.not25.i, label %l_Array_anyMUnsafe_any___at_Lean_registerInternalExceptionId___spec__2.exit, label %.lr.ph.i
 
@@ -1104,29 +1092,29 @@ lean_dec.exit12:                                  ; preds = %24, %23, %21, %lean
   br i1 %.not.i21, label %l_Array_anyMUnsafe_any___at_Lean_registerInternalExceptionId___spec__2.exit, label %28
 
 28:                                               ; preds = %26, %.lr.ph.i
-  %.01526.i = phi i64 [ %.val, %.lr.ph.i ], [ %27, %26 ]
+  %.01526.i = phi i64 [ %.val, %.lr.ph.i ], [ %27, %20 ]
   %29 = getelementptr inbounds nuw ptr, ptr %25, i64 %.01526.i
   %30 = load ptr, ptr %29, align 8, !tbaa !11
   %31 = ptrtoint ptr %30 to i64
   %32 = and i64 %31, 1
   %.not.i20.i = icmp eq i64 %32, 0
-  br i1 %.not.i20.i, label %33, label %lean_array_uget.exit.i
+  br i1 %.not.i20.i, label %27, label %lean_array_uget.exit.i
 
-33:                                               ; preds = %28
-  %.val.i.i.i = load i32, ptr %30, align 4, !tbaa !5
-  %34 = icmp sgt i32 %.val.i.i.i, 0
-  br i1 %34, label %35, label %37, !prof !10
+27:                                               ; preds = %22
+  %.val.i.i.i = load i32, ptr %24, align 4, !tbaa !5
+  %28 = icmp sgt i32 %.val.i.i.i, 0
+  br i1 %28, label %29, label %31, !prof !10
 
-35:                                               ; preds = %33
-  %36 = add nuw i32 %.val.i.i.i, 1
-  store i32 %36, ptr %30, align 4, !tbaa !5
-  br label %40
+29:                                               ; preds = %27
+  %30 = add nuw i32 %.val.i.i.i, 1
+  store i32 %30, ptr %24, align 4, !tbaa !5
+  br label %34
 
-37:                                               ; preds = %33
+33:                                               ; preds = %27
   %.not.i.i.i = icmp eq i32 %.val.i.i.i, 0
   br i1 %.not.i.i.i, label %40, label %38
 
-38:                                               ; preds = %37
+38:                                               ; preds = %31
   tail call void @lean_inc_ref_cold(ptr noundef nonnull %30) #4
   br label %40
 
@@ -1134,7 +1122,7 @@ lean_array_uget.exit.i:                           ; preds = %28
   %39 = tail call zeroext i8 @lean_name_eq(ptr noundef %0, ptr noundef %30) #4
   br label %lean_dec.exit.i
 
-40:                                               ; preds = %38, %37, %35
+40:                                               ; preds = %38, %31, %29
   %41 = tail call zeroext i8 @lean_name_eq(ptr noundef %0, ptr noundef nonnull %30) #4
   %42 = load i32, ptr %30, align 4, !tbaa !5
   %43 = icmp sgt i32 %42, 1
@@ -1154,12 +1142,12 @@ lean_array_uget.exit.i:                           ; preds = %28
   br label %lean_dec.exit.i
 
 lean_dec.exit.i:                                  ; preds = %47, %46, %44, %lean_array_uget.exit.i
-  %48 = phi i8 [ %39, %lean_array_uget.exit.i ], [ %41, %44 ], [ %41, %46 ], [ %41, %47 ]
+  %48 = phi i8 [ %39, %lean_array_uget.exit.i ], [ %41, %38 ], [ %41, %40 ], [ %41, %41 ]
   %49 = icmp eq i8 %48, 0
   br i1 %49, label %26, label %l_Array_anyMUnsafe_any___at_Lean_registerInternalExceptionId___spec__2.exit
 
 l_Array_anyMUnsafe_any___at_Lean_registerInternalExceptionId___spec__2.exit: ; preds = %26, %lean_dec.exit.i, %lean_dec.exit12
-  %.2.ph.i = phi i64 [ 1, %lean_dec.exit12 ], [ 1, %26 ], [ 3, %lean_dec.exit.i ]
+  %.2.ph.i = phi i64 [ 1, %lean_dec.exit12 ], [ 1, %20 ], [ 3, %lean_dec.exit.i ]
   %50 = ptrtoint ptr %1 to i64
   %51 = and i64 %50, 1
   %.not23 = icmp eq i64 %51, 0
diff --git a/bench/lean4/optimized/Links.ll b/bench/lean4/optimized/Links.ll
index 35aea76f649..03f73e20103 100644
--- a/bench/lean4/optimized/Links.ll
+++ b/bench/lean4/optimized/Links.ll
@@ -84,22 +84,16 @@ lean_inc.exit:
   %3 = shl i64 %.val, 1
   %4 = add i64 %3, -1
   %5 = inttoptr i64 %4 to ptr
-  %6 = ptrtoint ptr %0 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_inc.exit41
-
-8:                                                ; preds = %lean_inc.exit
-  %.val.i47 = load i32, ptr %0, align 4, !tbaa !8
-  %9 = icmp sgt i32 %.val.i47, 0
-  br i1 %9, label %10, label %12, !prof !11
+  %.val.i47 = load i32, ptr %0, align 8, !tbaa !8
+  %6 = icmp sgt i32 %.val.i47, 0
+  br i1 %6, label %7, label %9, !prof !11
 
-10:                                               ; preds = %8
+10:                                               ; preds = %lean_inc.exit
   %11 = add nuw i32 %.val.i47, 1
   store i32 %11, ptr %0, align 4, !tbaa !8
   br label %lean_inc.exit41
 
-12:                                               ; preds = %8
+12:                                               ; preds = %lean_inc.exit
   %.not.i48 = icmp eq i32 %.val.i47, 0
   br i1 %.not.i48, label %lean_inc.exit41, label %13
 
@@ -107,7 +101,7 @@ lean_inc.exit:
   tail call void @lean_inc_ref_cold(ptr noundef nonnull %0) #4
   br label %lean_inc.exit41
 
-lean_inc.exit41:                                  ; preds = %13, %12, %10, %lean_inc.exit
+lean_inc.exit41:                                  ; preds = %13, %12, %10
   tail call void @lean_inc_heartbeat() #4
   %14 = tail call noalias ptr @mi_malloc_small(i64 noundef 32) #4
   %15 = icmp eq ptr %14, null
@@ -126,62 +120,53 @@ lean_nat_sub.exit:                                ; preds = %lean_inc.exit41
   %19 = getelementptr inbounds nuw i8, ptr %14, i64 16
   store ptr inttoptr (i64 1 to ptr), ptr %19, align 8, !tbaa !12
   %20 = getelementptr inbounds nuw i8, ptr %14, i64 24
-  store ptr %5, ptr %20, align 8, !tbaa !12
+  store ptr %5, ptr %110, align 8, !tbaa !12
   %21 = load ptr, ptr @l_Lean_manualRoot___lambda__1___closed__2, align 8, !tbaa !12
   %22 = tail call ptr @l_Substring_prevn(ptr noundef nonnull %14, ptr noundef %21, ptr noundef nonnull %5) #4
-  %23 = ptrtoint ptr %14 to i64
-  %24 = and i64 %23, 1
-  %.not57 = icmp eq i64 %24, 0
-  br i1 %.not57, label %25, label %lean_dec.exit
-
-25:                                               ; preds = %lean_nat_sub.exit
-  %26 = load i32, ptr %14, align 4, !tbaa !8
-  %27 = icmp sgt i32 %26, 1
-  br i1 %27, label %28, label %30, !prof !11
+  %20 = load i32, ptr %14, align 8, !tbaa !8
+  %21 = icmp sgt i32 %20, 1
+  br i1 %21, label %22, label %24, !prof !11
 
-28:                                               ; preds = %25
-  %29 = add nsw i32 %26, -1
+28:                                               ; preds = %lean_nat_sub.exit
+  %29 = add nsw i32 %20, -1
   store i32 %29, ptr %14, align 4, !tbaa !8
   br label %lean_dec.exit
 
-30:                                               ; preds = %25
-  %.not.i44 = icmp eq i32 %26, 0
+30:                                               ; preds = %lean_nat_sub.exit
+  %.not.i44 = icmp eq i32 %20, 0
   br i1 %.not.i44, label %lean_dec.exit, label %31
 
 31:                                               ; preds = %30
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %14) #4
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %lean_nat_sub.exit, %28, %30, %31
+lean_dec.exit:                                    ; preds = %28, %30, %31
   %32 = ptrtoint ptr %22 to i64
   %33 = and i64 %32, 1
   %.not58 = icmp eq i64 %33, 0
-  br i1 %.not58, label %34, label %lean_dec.exit43, !prof !14
+  br i1 %.not58, label %40, label %lean_dec.exit43, !prof !14
 
-34:                                               ; preds = %lean_dec.exit
-  %35 = tail call ptr @lean_nat_big_add(ptr noundef nonnull inttoptr (i64 1 to ptr), ptr noundef %22) #4
-  %36 = load i32, ptr %22, align 4, !tbaa !8
-  %37 = icmp sgt i32 %36, 1
-  br i1 %37, label %38, label %40, !prof !11
+40:                                               ; preds = %lean_dec.exit
+  %29 = tail call ptr @lean_nat_big_add(ptr noundef nonnull inttoptr (i64 1 to ptr), ptr noundef %19) #4
+  %30 = load i32, ptr %19, align 4, !tbaa !8
+  %31 = icmp sgt i32 %30, 1
+  br i1 %31, label %32, label %34, !prof !11
 
-38:                                               ; preds = %34
-  %39 = add nsw i32 %36, -1
-  store i32 %39, ptr %22, align 4, !tbaa !8
+41:                                               ; preds = %40
+  %33 = add nsw i32 %30, -1
+  store i32 %33, ptr %19, align 4, !tbaa !8
   br label %lean_dec.exit43
 
-40:                                               ; preds = %34
-  %.not.i = icmp eq i32 %36, 0
-  br i1 %.not.i, label %lean_dec.exit43, label %41
+lean_dec.exit43:                                  ; preds = %40
+  %.not.i = icmp eq i32 %30, 0
+  br i1 %.not.i, label %42, label %35
 
-41:                                               ; preds = %40
-  tail call void @lean_dec_ref_cold(ptr noundef nonnull %22) #4
+35:                                               ; preds = %34
+  tail call void @lean_dec_ref_cold(ptr noundef nonnull %19) #4
   br label %lean_dec.exit43
 
-lean_dec.exit43:                                  ; preds = %lean_dec.exit, %41, %40, %38
-  %.0.i68 = phi ptr [ %35, %41 ], [ %35, %40 ], [ %35, %38 ], [ %22, %lean_dec.exit ]
-  br i1 %.not, label %42, label %lean_inc.exit42
-
-42:                                               ; preds = %lean_dec.exit43
+42:                                               ; preds = %32, %34, %35, %lean_dec.exit
+  %.0.i68 = phi ptr [ %29, %35 ], [ %29, %34 ], [ %29, %32 ], [ %19, %lean_dec.exit ]
   %.val.i51 = load i32, ptr %0, align 4, !tbaa !8
   %43 = icmp sgt i32 %.val.i51, 0
   br i1 %43, label %44, label %46, !prof !11
@@ -199,7 +184,7 @@ lean_dec.exit43:                                  ; preds = %lean_dec.exit, %41,
   tail call void @lean_inc_ref_cold(ptr noundef nonnull %0) #4
   br label %lean_inc.exit42
 
-lean_inc.exit42:                                  ; preds = %47, %46, %44, %lean_dec.exit43
+lean_inc.exit42:                                  ; preds = %47, %46, %44
   tail call void @lean_inc_heartbeat() #4
   %48 = tail call noalias ptr @mi_malloc_small(i64 noundef 32) #4
   %49 = icmp eq ptr %48, null
@@ -247,8 +232,8 @@ lean_alloc_ctor.exit54:                           ; preds = %lean_inc.exit42
   unreachable
 
 lean_alloc_ctor.exit55:                           ; preds = %64, %58
-  %.sink75 = phi ptr [ %61, %58 ], [ %65, %64 ]
-  %.sink = phi ptr [ %60, %58 ], [ %0, %64 ]
+  %.sink75 = phi ptr [ %61, %51 ], [ %65, %57 ]
+  %.sink = phi ptr [ %60, %51 ], [ %0, %57 ]
   %68 = getelementptr inbounds nuw i8, ptr %.sink75, i64 4
   store i32 1, ptr %.sink75, align 4, !tbaa !8
   store i32 131096, ptr %68, align 4
diff --git a/bench/lean4/optimized/MVarRenaming.ll b/bench/lean4/optimized/MVarRenaming.ll
index 54972ea8243..24aa3f45eb7 100644
--- a/bench/lean4/optimized/MVarRenaming.ll
+++ b/bench/lean4/optimized/MVarRenaming.ll
@@ -859,31 +859,25 @@ lean_alloc_closure.exit:                          ; preds = %40
   %48 = getelementptr inbounds nuw i8, ptr %41, i64 24
   store ptr %0, ptr %48, align 8, !tbaa !10
   %49 = tail call ptr @lean_replace_expr(ptr noundef nonnull %41, ptr noundef %1) #4
-  %50 = ptrtoint ptr %41 to i64
-  %51 = and i64 %50, 1
-  %.not = icmp eq i64 %51, 0
-  br i1 %.not, label %52, label %lean_inc.exit15
+  %50 = load i32, ptr %41, align 8, !tbaa !4
+  %51 = icmp sgt i32 %50, 1
+  br i1 %51, label %52, label %54, !prof !9
 
 52:                                               ; preds = %lean_alloc_closure.exit
-  %53 = load i32, ptr %41, align 4, !tbaa !4
-  %54 = icmp sgt i32 %53, 1
-  br i1 %54, label %55, label %57, !prof !9
-
-55:                                               ; preds = %52
-  %56 = add nsw i32 %53, -1
-  store i32 %56, ptr %41, align 4, !tbaa !4
+  %53 = add nsw i32 %50, -1
+  store i32 %53, ptr %41, align 4, !tbaa !4
   br label %lean_inc.exit15
 
-57:                                               ; preds = %52
-  %.not.i16 = icmp eq i32 %53, 0
+57:                                               ; preds = %lean_alloc_closure.exit
+  %.not.i16 = icmp eq i32 %50, 0
   br i1 %.not.i16, label %lean_inc.exit15, label %58
 
 58:                                               ; preds = %57
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %41) #4
   br label %lean_inc.exit15
 
-lean_inc.exit15:                                  ; preds = %lean_alloc_closure.exit, %55, %57, %58, %31, %36, %38, %39, %lean_dec.exit14, %19, %21, %22
-  %.0 = phi ptr [ %1, %22 ], [ %1, %21 ], [ %1, %19 ], [ %1, %lean_dec.exit14 ], [ %1, %39 ], [ %1, %38 ], [ %1, %36 ], [ %1, %31 ], [ %49, %58 ], [ %49, %57 ], [ %49, %55 ], [ %49, %lean_alloc_closure.exit ]
+lean_inc.exit15:                                  ; preds = %52, %57, %58, %31, %36, %38, %39, %lean_dec.exit14, %19, %21, %22
+  %.0 = phi ptr [ %1, %22 ], [ %1, %21 ], [ %1, %19 ], [ %1, %lean_dec.exit14 ], [ %1, %39 ], [ %1, %38 ], [ %1, %36 ], [ %1, %31 ], [ %49, %55 ], [ %49, %54 ], [ %49, %52 ]
   ret ptr %.0
 }
 
diff --git a/bench/lean4/optimized/PlainDateTime.ll b/bench/lean4/optimized/PlainDateTime.ll
index b89c1666e37..c1d130c0c47 100644
--- a/bench/lean4/optimized/PlainDateTime.ll
+++ b/bench/lean4/optimized/PlainDateTime.ll
@@ -31691,30 +31691,24 @@ define nonnull ptr @l_Std_Time_PlainDateTime_era___boxed(ptr noundef %0) local_u
   %4 = getelementptr inbounds nuw i8, ptr %3, i64 8
   %5 = load ptr, ptr %4, align 8, !tbaa !5
   %6 = tail call zeroext i8 @l_Std_Time_Year_Offset_era(ptr noundef %5) #5
-  %7 = ptrtoint ptr %0 to i64
-  %8 = and i64 %7, 1
-  %.not = icmp eq i64 %8, 0
-  br i1 %.not, label %9, label %lean_dec.exit
+  %7 = load i32, ptr %0, align 8, !tbaa !9
+  %8 = icmp sgt i32 %7, 1
+  br i1 %8, label %9, label %11, !prof !12
 
 9:                                                ; preds = %1
-  %10 = load i32, ptr %0, align 4, !tbaa !9
-  %11 = icmp sgt i32 %10, 1
-  br i1 %11, label %12, label %14, !prof !12
-
-12:                                               ; preds = %9
-  %13 = add nsw i32 %10, -1
-  store i32 %13, ptr %0, align 4, !tbaa !9
+  %10 = add nsw i32 %7, -1
+  store i32 %10, ptr %0, align 4, !tbaa !12
   br label %lean_dec.exit
 
-14:                                               ; preds = %9
-  %.not.i = icmp eq i32 %10, 0
+14:                                               ; preds = %1
+  %.not.i = icmp eq i32 %7, 0
   br i1 %.not.i, label %lean_dec.exit, label %15
 
 15:                                               ; preds = %14
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #5
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %15, %14, %12, %1
+lean_dec.exit:                                    ; preds = %15, %14, %12
   %16 = zext i8 %6 to i64
   %17 = shl nuw nsw i64 %16, 1
   %18 = or disjoint i64 %17, 1
@@ -32061,30 +32055,24 @@ define ptr @l_Std_Time_PlainDateTime_weekOfMonth___boxed(ptr noundef %0) local_u
   %2 = getelementptr inbounds nuw i8, ptr %0, i64 8
   %3 = load ptr, ptr %2, align 8, !tbaa !5
   %4 = tail call ptr @l_Std_Time_PlainDate_weekOfMonth(ptr noundef %3) #5
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit
+  %5 = load i32, ptr %0, align 8, !tbaa !9
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !12
 
 7:                                                ; preds = %1
-  %8 = load i32, ptr %0, align 4, !tbaa !9
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !12
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !9
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !12
   br label %lean_dec.exit
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %1
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #5
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %13, %12, %10, %1
+lean_dec.exit:                                    ; preds = %13, %12, %7
   ret ptr %4
 }
 
@@ -32593,30 +32581,24 @@ define ptr @l_Std_Time_PlainDateTime_quarter___boxed(ptr noundef %0) local_unnam
   %2 = getelementptr inbounds nuw i8, ptr %0, i64 8
   %3 = load ptr, ptr %2, align 8, !tbaa !5
   %4 = tail call ptr @l_Std_Time_PlainDate_quarter(ptr noundef %3) #5
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit
+  %5 = load i32, ptr %0, align 8, !tbaa !9
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !12
 
 7:                                                ; preds = %1
-  %8 = load i32, ptr %0, align 4, !tbaa !9
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !12
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !9
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !12
   br label %lean_dec.exit
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %1
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #5
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %13, %12, %10, %1
+lean_dec.exit:                                    ; preds = %13, %12, %7
   ret ptr %4
 }
 
diff --git a/bench/lean4/optimized/Position.ll b/bench/lean4/optimized/Position.ll
index f5fc89caa95..1f899406a5b 100644
--- a/bench/lean4/optimized/Position.ll
+++ b/bench/lean4/optimized/Position.ll
@@ -2711,30 +2711,24 @@ define nonnull ptr @l_Lean_FileMap_getLastLine___boxed(ptr noundef %0) local_unn
   %3 = load ptr, ptr %2, align 8, !tbaa !4
   %4 = getelementptr i8, ptr %3, i64 8
   %.val.i = load i64, ptr %4, align 8, !tbaa !13
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit
+  %5 = load i32, ptr %0, align 8, !tbaa !9
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !12
 
 7:                                                ; preds = %1
-  %8 = load i32, ptr %0, align 4, !tbaa !9
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !12
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !9
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !12
   br label %lean_dec.exit
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %1
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #6
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %13, %12, %10, %1
+lean_dec.exit:                                    ; preds = %13, %12, %7
   %14 = and i64 %.val.i, 9223372036854775807
   %15 = icmp eq i64 %14, 0
   %16 = shl i64 %.val.i, 1
diff --git a/bench/lean4/optimized/PrettyPrinter.ll b/bench/lean4/optimized/PrettyPrinter.ll
index d67b3ca8e41..6f14e441ec3 100644
--- a/bench/lean4/optimized/PrettyPrinter.ll
+++ b/bench/lean4/optimized/PrettyPrinter.ll
@@ -23616,56 +23616,44 @@ lean_alloc_ctor.exit210:                          ; preds = %lean_dec.exit
 define ptr @l_Array_mapMUnsafe_map___at___private_Lean_PrettyPrinter_0__Lean_PrettyPrinter_noContext___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2) local_unnamed_addr #0 {
   %4 = getelementptr i8, ptr %0, i64 8
   %.val = load i64, ptr %4, align 8, !tbaa !13
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit7
+  %5 = load i32, ptr %0, align 8, !tbaa !4
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !11
 
 7:                                                ; preds = %3
-  %8 = load i32, ptr %0, align 4, !tbaa !4
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !11
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !4
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !4
   br label %lean_dec.exit7
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %3
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit7, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #6
   br label %lean_dec.exit7
 
-lean_dec.exit7:                                   ; preds = %13, %12, %10, %3
+lean_dec.exit7:                                   ; preds = %13, %12, %7
   %14 = getelementptr i8, ptr %1, i64 8
   %.val10 = load i64, ptr %14, align 8, !tbaa !13
-  %15 = ptrtoint ptr %1 to i64
-  %16 = and i64 %15, 1
-  %.not11 = icmp eq i64 %16, 0
-  br i1 %.not11, label %17, label %lean_dec.exit
-
-17:                                               ; preds = %lean_dec.exit7
-  %18 = load i32, ptr %1, align 4, !tbaa !4
-  %19 = icmp sgt i32 %18, 1
-  br i1 %19, label %20, label %22, !prof !11
+  %15 = load i32, ptr %1, align 8, !tbaa !4
+  %13 = icmp sgt i32 %15, 1
+  br i1 %13, label %14, label %16, !prof !11
 
-20:                                               ; preds = %17
-  %21 = add nsw i32 %18, -1
+20:                                               ; preds = %lean_dec.exit7
+  %21 = add nsw i32 %12, -1
   store i32 %21, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit
 
-22:                                               ; preds = %17
-  %.not.i8 = icmp eq i32 %18, 0
+22:                                               ; preds = %lean_dec.exit7
+  %.not.i8 = icmp eq i32 %12, 0
   br i1 %.not.i8, label %lean_dec.exit, label %23
 
 23:                                               ; preds = %22
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #6
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %23, %22, %20, %lean_dec.exit7
+lean_dec.exit:                                    ; preds = %23, %22, %20
   %24 = tail call ptr @l_Array_mapMUnsafe_map___at___private_Lean_PrettyPrinter_0__Lean_PrettyPrinter_noContext___spec__1(i64 noundef %.val, i64 noundef %.val10, ptr noundef %2)
   ret ptr %24
 }
@@ -26993,56 +26981,44 @@ lean_dec.exit:                                    ; preds = %25, %24, %22, %l_Le
 define nonnull ptr @l_Array_anyMUnsafe_any___at_Lean_MessageData_ofLazyM___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3) local_unnamed_addr #0 {
   %5 = getelementptr i8, ptr %2, i64 8
   %.val16 = load i64, ptr %5, align 8, !tbaa !13
-  %6 = ptrtoint ptr %2 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit11
+  %6 = load i32, ptr %2, align 8, !tbaa !4
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !11
 
 8:                                                ; preds = %4
-  %9 = load i32, ptr %2, align 4, !tbaa !4
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !11
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %2, align 4, !tbaa !4
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit11
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %4
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit11, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #6
   br label %lean_dec.exit11
 
-lean_dec.exit11:                                  ; preds = %14, %13, %11, %4
+lean_dec.exit11:                                  ; preds = %14, %13, %8
   %15 = getelementptr i8, ptr %3, i64 8
   %.val = load i64, ptr %15, align 8, !tbaa !13
-  %16 = ptrtoint ptr %3 to i64
-  %17 = and i64 %16, 1
-  %.not17 = icmp eq i64 %17, 0
-  br i1 %.not17, label %18, label %lean_dec.exit10
-
-18:                                               ; preds = %lean_dec.exit11
-  %19 = load i32, ptr %3, align 4, !tbaa !4
-  %20 = icmp sgt i32 %19, 1
-  br i1 %20, label %21, label %23, !prof !11
+  %13 = load i32, ptr %3, align 8, !tbaa !4
+  %14 = icmp sgt i32 %13, 1
+  br i1 %14, label %15, label %17, !prof !11
 
-21:                                               ; preds = %18
-  %22 = add nsw i32 %19, -1
+21:                                               ; preds = %lean_dec.exit11
+  %22 = add nsw i32 %13, -1
   store i32 %22, ptr %3, align 4, !tbaa !4
   br label %lean_dec.exit10
 
-23:                                               ; preds = %18
-  %.not.i12 = icmp eq i32 %19, 0
+23:                                               ; preds = %lean_dec.exit11
+  %.not.i12 = icmp eq i32 %13, 0
   br i1 %.not.i12, label %lean_dec.exit10, label %24
 
 24:                                               ; preds = %23
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %3) #6
   br label %lean_dec.exit10
 
-lean_dec.exit10:                                  ; preds = %24, %23, %21, %lean_dec.exit11
+lean_dec.exit10:                                  ; preds = %24, %23, %21
   %25 = tail call zeroext i8 @l_Array_anyMUnsafe_any___at_Lean_MessageData_ofLazyM___spec__1(ptr noundef %0, ptr noundef %1, i64 noundef %.val16, i64 noundef %.val)
   %26 = ptrtoint ptr %1 to i64
   %27 = and i64 %26, 1
@@ -27068,7 +27044,7 @@ lean_dec.exit10:                                  ; preds = %24, %23, %21, %lean
   br label %lean_dec.exit
 
 lean_dec.exit:                                    ; preds = %34, %33, %31, %lean_dec.exit10
-  %35 = shl nuw nsw i8 %25, 1
+  %35 = shl nuw nsw i8 %112, 1
   %36 = or disjoint i8 %35, 1
   %37 = zext nneg i8 %36 to i64
   %38 = inttoptr i64 %37 to ptr
diff --git a/bench/lean4/optimized/Propagate.ll b/bench/lean4/optimized/Propagate.ll
index 309da0be281..13a236309ed 100644
--- a/bench/lean4/optimized/Propagate.ll
+++ b/bench/lean4/optimized/Propagate.ll
@@ -62052,56 +62052,44 @@ lean_alloc_ctor.exit1219:                         ; preds = %lean_dec.exit
 define ptr @l_Array_forIn_x27Unsafe_loop___at_Lean_Meta_Grind_propagateEqDown___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9, ptr noundef %10, ptr noundef %11, ptr noundef %12, ptr noundef %13, ptr noundef %14, ptr noundef %15) local_unnamed_addr #0 {
   %17 = getelementptr i8, ptr %4, i64 8
   %.val = load i64, ptr %17, align 8, !tbaa !16
-  %18 = ptrtoint ptr %4 to i64
-  %19 = and i64 %18, 1
-  %.not = icmp eq i64 %19, 0
-  br i1 %.not, label %20, label %lean_dec.exit26
+  %18 = load i32, ptr %4, align 8, !tbaa !4
+  %19 = icmp sgt i32 %18, 1
+  br i1 %19, label %20, label %22, !prof !9
 
 20:                                               ; preds = %16
-  %21 = load i32, ptr %4, align 4, !tbaa !4
-  %22 = icmp sgt i32 %21, 1
-  br i1 %22, label %23, label %25, !prof !9
-
-23:                                               ; preds = %20
-  %24 = add nsw i32 %21, -1
-  store i32 %24, ptr %4, align 4, !tbaa !4
+  %21 = add nsw i32 %18, -1
+  store i32 %21, ptr %4, align 4, !tbaa !4
   br label %lean_dec.exit26
 
-25:                                               ; preds = %20
-  %.not.i = icmp eq i32 %21, 0
+25:                                               ; preds = %16
+  %.not.i = icmp eq i32 %18, 0
   br i1 %.not.i, label %lean_dec.exit26, label %26
 
 26:                                               ; preds = %25
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %4) #4
   br label %lean_dec.exit26
 
-lean_dec.exit26:                                  ; preds = %26, %25, %23, %16
+lean_dec.exit26:                                  ; preds = %26, %25, %20
   %27 = getelementptr i8, ptr %5, i64 8
   %.val35 = load i64, ptr %27, align 8, !tbaa !16
-  %28 = ptrtoint ptr %5 to i64
-  %29 = and i64 %28, 1
-  %.not36 = icmp eq i64 %29, 0
-  br i1 %.not36, label %30, label %lean_dec.exit25
-
-30:                                               ; preds = %lean_dec.exit26
-  %31 = load i32, ptr %5, align 4, !tbaa !4
-  %32 = icmp sgt i32 %31, 1
-  br i1 %32, label %33, label %35, !prof !9
+  %25 = load i32, ptr %5, align 8, !tbaa !4
+  %26 = icmp sgt i32 %25, 1
+  br i1 %26, label %27, label %29, !prof !9
 
-33:                                               ; preds = %30
-  %34 = add nsw i32 %31, -1
+33:                                               ; preds = %lean_dec.exit26
+  %34 = add nsw i32 %25, -1
   store i32 %34, ptr %5, align 4, !tbaa !4
   br label %lean_dec.exit25
 
-35:                                               ; preds = %30
-  %.not.i27 = icmp eq i32 %31, 0
+35:                                               ; preds = %lean_dec.exit26
+  %.not.i27 = icmp eq i32 %25, 0
   br i1 %.not.i27, label %lean_dec.exit25, label %36
 
 36:                                               ; preds = %35
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %5) #4
   br label %lean_dec.exit25
 
-lean_dec.exit25:                                  ; preds = %36, %35, %33, %lean_dec.exit26
+lean_dec.exit25:                                  ; preds = %36, %35, %33
   %37 = tail call ptr @l_Array_forIn_x27Unsafe_loop___at_Lean_Meta_Grind_propagateEqDown___spec__1(ptr noundef %0, ptr poison, ptr poison, ptr noundef %3, i64 noundef %.val, i64 noundef %.val35, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9, ptr noundef %10, ptr noundef %11, ptr noundef %12, ptr noundef %13, ptr noundef %14, ptr noundef %15)
   %38 = ptrtoint ptr %3 to i64
   %39 = and i64 %38, 1
diff --git a/bench/lean4/optimized/ReplaceLevel.ll b/bench/lean4/optimized/ReplaceLevel.ll
index dded5bb9cef..135ea05b531 100644
--- a/bench/lean4/optimized/ReplaceLevel.ll
+++ b/bench/lean4/optimized/ReplaceLevel.ll
@@ -702,30 +702,24 @@ lean_alloc_ctor.exit36:                           ; preds = %lean_alloc_ctor.exi
 define noalias nonnull ptr @l_Lean_Expr_ReplaceLevelImpl_cache___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3) local_unnamed_addr #0 {
   %5 = getelementptr i8, ptr %0, i64 8
   %.val = load i64, ptr %5, align 8, !tbaa !12
-  %6 = ptrtoint ptr %0 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit
+  %6 = load i32, ptr %0, align 8, !tbaa !4
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !9
 
 8:                                                ; preds = %4
-  %9 = load i32, ptr %0, align 4, !tbaa !4
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !9
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %0, align 4, !tbaa !4
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %0, align 4, !tbaa !4
   br label %lean_dec.exit
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %4
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %14, %13, %11, %4
+lean_dec.exit:                                    ; preds = %14, %13, %8
   %15 = tail call ptr @l_Lean_Expr_ReplaceLevelImpl_cache(i64 noundef %.val, ptr noundef %1, ptr noundef %2, ptr noundef %3)
   ret ptr %15
 }
@@ -4438,30 +4432,24 @@ declare ptr @l_Lean_Expr_proj___override(ptr noundef, ptr noundef, ptr noundef)
 define noalias nonnull ptr @l_Lean_Expr_ReplaceLevelImpl_replaceUnsafeM_visit___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3) local_unnamed_addr #0 {
   %5 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %5, align 8, !tbaa !12
-  %6 = ptrtoint ptr %1 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit
+  %6 = load i32, ptr %1, align 8, !tbaa !4
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !9
 
 8:                                                ; preds = %4
-  %9 = load i32, ptr %1, align 4, !tbaa !4
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !9
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %1, align 4, !tbaa !4
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %4
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %14, %13, %11, %4
+lean_dec.exit:                                    ; preds = %14, %13, %8
   %15 = tail call ptr @l_Lean_Expr_ReplaceLevelImpl_replaceUnsafeM_visit(ptr noundef %0, i64 noundef %.val, ptr noundef %2, ptr noundef %3)
   ret ptr %15
 }
@@ -4476,30 +4464,24 @@ define noalias nonnull ptr @l_Lean_Expr_ReplaceLevelImpl_replaceUnsafeM(ptr noun
 define noalias nonnull ptr @l_Lean_Expr_ReplaceLevelImpl_replaceUnsafeM___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3) local_unnamed_addr #0 {
   %5 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %5, align 8, !tbaa !12
-  %6 = ptrtoint ptr %1 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit
+  %6 = load i32, ptr %1, align 8, !tbaa !4
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !9
 
 8:                                                ; preds = %4
-  %9 = load i32, ptr %1, align 4, !tbaa !4
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !9
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %1, align 4, !tbaa !4
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %4
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #3
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %14, %13, %11, %4
+lean_dec.exit:                                    ; preds = %14, %13, %8
   %15 = tail call noalias nonnull ptr @l_Lean_Expr_ReplaceLevelImpl_replaceUnsafeM_visit(ptr noundef %0, i64 noundef %.val, ptr noundef %2, ptr noundef %3)
   ret ptr %15
 }
diff --git a/bench/lean4/optimized/RequestCancellation.ll b/bench/lean4/optimized/RequestCancellation.ll
index 3db35ae675a..e319b03c1b7 100644
--- a/bench/lean4/optimized/RequestCancellation.ll
+++ b/bench/lean4/optimized/RequestCancellation.ll
@@ -1413,30 +1413,24 @@ define ptr @l_Lean_Server_RequestCancellationToken_requestCancellationTask___box
   %2 = getelementptr inbounds nuw i8, ptr %0, i64 24
   %3 = load ptr, ptr %2, align 8, !tbaa !4
   %4 = tail call ptr @l_IO_Promise_result_x21___rarg(ptr noundef %3) #5
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit
+  %5 = load i32, ptr %0, align 8, !tbaa !8
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !11
 
 7:                                                ; preds = %1
-  %8 = load i32, ptr %0, align 4, !tbaa !8
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !11
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !8
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !8
   br label %lean_dec.exit
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %1
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #5
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %13, %12, %10, %1
+lean_dec.exit:                                    ; preds = %13, %12, %7
   ret ptr %4
 }
 
@@ -1453,30 +1447,24 @@ define ptr @l_Lean_Server_RequestCancellationToken_editCancellationTask___boxed(
   %2 = getelementptr inbounds nuw i8, ptr %0, i64 32
   %3 = load ptr, ptr %2, align 8, !tbaa !4
   %4 = tail call ptr @l_IO_Promise_result_x21___rarg(ptr noundef %3) #5
-  %5 = ptrtoint ptr %0 to i64
-  %6 = and i64 %5, 1
-  %.not = icmp eq i64 %6, 0
-  br i1 %.not, label %7, label %lean_dec.exit
+  %5 = load i32, ptr %0, align 8, !tbaa !8
+  %6 = icmp sgt i32 %5, 1
+  br i1 %6, label %7, label %9, !prof !11
 
 7:                                                ; preds = %1
-  %8 = load i32, ptr %0, align 4, !tbaa !8
-  %9 = icmp sgt i32 %8, 1
-  br i1 %9, label %10, label %12, !prof !11
-
-10:                                               ; preds = %7
-  %11 = add nsw i32 %8, -1
-  store i32 %11, ptr %0, align 4, !tbaa !8
+  %8 = add nsw i32 %5, -1
+  store i32 %8, ptr %0, align 4, !tbaa !8
   br label %lean_dec.exit
 
-12:                                               ; preds = %7
-  %.not.i = icmp eq i32 %8, 0
+12:                                               ; preds = %1
+  %.not.i = icmp eq i32 %5, 0
   br i1 %.not.i, label %lean_dec.exit, label %13
 
 13:                                               ; preds = %12
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #5
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %13, %12, %10, %1
+lean_dec.exit:                                    ; preds = %13, %12, %7
   ret ptr %4
 }
 
@@ -1570,30 +1558,24 @@ define ptr @l_Lean_Server_RequestCancellationToken_wasCancelledByCancelRequest__
   %3 = getelementptr inbounds nuw i8, ptr %0, i64 8
   %4 = load ptr, ptr %3, align 8, !tbaa !4
   %5 = tail call ptr @lean_st_ref_get(ptr noundef %4, ptr noundef %1) #5
-  %6 = ptrtoint ptr %0 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit
+  %6 = load i32, ptr %0, align 8, !tbaa !8
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !11
 
 8:                                                ; preds = %2
-  %9 = load i32, ptr %0, align 4, !tbaa !8
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !11
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %0, align 4, !tbaa !8
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %0, align 4, !tbaa !8
   br label %lean_dec.exit
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %2
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #5
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %14, %13, %11, %2
+lean_dec.exit:                                    ; preds = %14, %13, %8
   ret ptr %5
 }
 
@@ -1610,30 +1592,24 @@ define ptr @l_Lean_Server_RequestCancellationToken_wasCancelledByEdit___boxed(pt
   %3 = getelementptr inbounds nuw i8, ptr %0, i64 16
   %4 = load ptr, ptr %3, align 8, !tbaa !4
   %5 = tail call ptr @lean_st_ref_get(ptr noundef %4, ptr noundef %1) #5
-  %6 = ptrtoint ptr %0 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit
+  %6 = load i32, ptr %0, align 8, !tbaa !8
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !11
 
 8:                                                ; preds = %2
-  %9 = load i32, ptr %0, align 4, !tbaa !8
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !11
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %0, align 4, !tbaa !8
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %0, align 4, !tbaa !8
   br label %lean_dec.exit
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %2
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %0) #5
   br label %lean_dec.exit
 
-lean_dec.exit:                                    ; preds = %14, %13, %11, %2
+lean_dec.exit:                                    ; preds = %14, %13, %8
   ret ptr %5
 }
 
diff --git a/bench/lean4/optimized/Simp.ll b/bench/lean4/optimized/Simp.ll
index e1c4b6eafa4..ed452f92713 100644
--- a/bench/lean4/optimized/Simp.ll
+++ b/bench/lean4/optimized/Simp.ll
@@ -9410,23 +9410,17 @@ lean_alloc_ctor.exit:                             ; preds = %lean_dec.exit84
   %275 = getelementptr inbounds nuw i8, ptr %271, i64 8
   store ptr %261, ptr %275, align 8, !tbaa !4
   %276 = tail call ptr @l_Lean_Elab_Tactic_Conv_evalSimpTrace___lambda__1(ptr noundef %1, ptr noundef %2, ptr noundef %5, ptr noundef %3, ptr nonnull poison, ptr noundef nonnull %271, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9, ptr noundef %10, ptr noundef %11, ptr noundef %12, ptr noundef %13, ptr noundef %14)
-  %277 = ptrtoint ptr %271 to i64
-  %278 = and i64 %277, 1
-  %.not171 = icmp eq i64 %278, 0
-  br i1 %.not171, label %279, label %lean_dec.exit83
+  %277 = load i32, ptr %271, align 8, !tbaa !8
+  %278 = icmp sgt i32 %277, 1
+  br i1 %278, label %279, label %281, !prof !11
 
 279:                                              ; preds = %lean_alloc_ctor.exit
-  %280 = load i32, ptr %271, align 4, !tbaa !8
-  %281 = icmp sgt i32 %280, 1
-  br i1 %281, label %282, label %284, !prof !11
-
-282:                                              ; preds = %279
-  %283 = add nsw i32 %280, -1
-  store i32 %283, ptr %271, align 4, !tbaa !8
+  %280 = add nsw i32 %277, -1
+  store i32 %280, ptr %271, align 4, !tbaa !8
   br label %lean_dec.exit83
 
-284:                                              ; preds = %279
-  %.not.i157 = icmp eq i32 %280, 0
+284:                                              ; preds = %lean_alloc_ctor.exit
+  %.not.i157 = icmp eq i32 %277, 0
   br i1 %.not.i157, label %lean_dec.exit83, label %285
 
 285:                                              ; preds = %284
@@ -9458,8 +9452,8 @@ lean_dec.exit:                                    ; preds = %293, %292, %290, %2
   %294 = tail call ptr @l_Lean_Elab_Tactic_Conv_evalSimpTrace___lambda__1(ptr noundef %1, ptr noundef %2, ptr noundef %5, ptr noundef %3, ptr nonnull poison, ptr noundef nonnull inttoptr (i64 1 to ptr), ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9, ptr noundef %10, ptr noundef %11, ptr noundef %12, ptr noundef %13, ptr noundef %14)
   br label %lean_dec.exit83
 
-lean_dec.exit83:                                  ; preds = %lean_alloc_ctor.exit, %282, %284, %285, %lean_dec.exit98, %lean_dec.exit86, %lean_dec.exit
-  %.2 = phi ptr [ %294, %lean_dec.exit ], [ %130, %lean_dec.exit98 ], [ %251, %lean_dec.exit86 ], [ %276, %285 ], [ %276, %284 ], [ %276, %282 ], [ %276, %lean_alloc_ctor.exit ]
+lean_dec.exit83:                                  ; preds = %279, %284, %285, %lean_dec.exit98, %lean_dec.exit86, %lean_dec.exit
+  %.2 = phi ptr [ %294, %lean_dec.exit ], [ %130, %lean_dec.exit98 ], [ %251, %lean_dec.exit86 ], [ %276, %282 ], [ %276, %281 ], [ %276, %279 ]
   ret ptr %.2
 }
 
@@ -18845,23 +18839,17 @@ lean_alloc_ctor.exit:                             ; preds = %lean_dec.exit82
   %274 = getelementptr inbounds nuw i8, ptr %270, i64 8
   store ptr %260, ptr %274, align 8, !tbaa !4
   %275 = tail call ptr @l_Lean_Elab_Tactic_Conv_evalDSimpTrace___lambda__1(ptr noundef %1, ptr noundef %4, ptr noundef %2, ptr nonnull poison, ptr noundef nonnull %270, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9, ptr noundef %10, ptr noundef %11, ptr noundef %12, ptr noundef %13)
-  %276 = ptrtoint ptr %270 to i64
-  %277 = and i64 %276, 1
-  %.not169 = icmp eq i64 %277, 0
-  br i1 %.not169, label %278, label %lean_dec.exit81
+  %276 = load i32, ptr %270, align 8, !tbaa !8
+  %277 = icmp sgt i32 %276, 1
+  br i1 %277, label %278, label %280, !prof !11
 
 278:                                              ; preds = %lean_alloc_ctor.exit
-  %279 = load i32, ptr %270, align 4, !tbaa !8
-  %280 = icmp sgt i32 %279, 1
-  br i1 %280, label %281, label %283, !prof !11
-
-281:                                              ; preds = %278
-  %282 = add nsw i32 %279, -1
-  store i32 %282, ptr %270, align 4, !tbaa !8
+  %279 = add nsw i32 %276, -1
+  store i32 %279, ptr %270, align 4, !tbaa !8
   br label %lean_dec.exit81
 
-283:                                              ; preds = %278
-  %.not.i155 = icmp eq i32 %279, 0
+283:                                              ; preds = %lean_alloc_ctor.exit
+  %.not.i155 = icmp eq i32 %276, 0
   br i1 %.not.i155, label %lean_dec.exit81, label %284
 
 284:                                              ; preds = %283
@@ -18893,8 +18881,8 @@ lean_dec.exit:                                    ; preds = %292, %291, %289, %2
   %293 = tail call ptr @l_Lean_Elab_Tactic_Conv_evalDSimpTrace___lambda__1(ptr noundef %1, ptr noundef %4, ptr noundef %2, ptr nonnull poison, ptr noundef nonnull inttoptr (i64 1 to ptr), ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8, ptr noundef %9, ptr noundef %10, ptr noundef %11, ptr noundef %12, ptr noundef %13)
   br label %lean_dec.exit81
 
-lean_dec.exit81:                                  ; preds = %lean_alloc_ctor.exit, %281, %283, %284, %lean_dec.exit96, %lean_dec.exit84, %lean_dec.exit
-  %.2 = phi ptr [ %293, %lean_dec.exit ], [ %129, %lean_dec.exit96 ], [ %250, %lean_dec.exit84 ], [ %275, %284 ], [ %275, %283 ], [ %275, %281 ], [ %275, %lean_alloc_ctor.exit ]
+lean_dec.exit81:                                  ; preds = %278, %283, %284, %lean_dec.exit96, %lean_dec.exit84, %lean_dec.exit
+  %.2 = phi ptr [ %293, %lean_dec.exit ], [ %129, %lean_dec.exit96 ], [ %250, %lean_dec.exit84 ], [ %275, %281 ], [ %275, %280 ], [ %275, %278 ]
   ret ptr %.2
 }
 
diff --git a/bench/lean4/optimized/TZdb.ll b/bench/lean4/optimized/TZdb.ll
index 956e2f3887c..440da13d0ee 100644
--- a/bench/lean4/optimized/TZdb.ll
+++ b/bench/lean4/optimized/TZdb.ll
@@ -4020,56 +4020,44 @@ lean_dec.exit:                                    ; preds = %21, %20, %18, %lean
 define ptr @l_Array_forIn_x27Unsafe_loop___at_Std_Time_Database_TZdb_inst___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8) local_unnamed_addr #0 {
   %10 = getelementptr i8, ptr %5, i64 8
   %.val = load i64, ptr %10, align 8, !tbaa !12
-  %11 = ptrtoint ptr %5 to i64
-  %12 = and i64 %11, 1
-  %.not = icmp eq i64 %12, 0
-  br i1 %.not, label %13, label %lean_dec.exit19
+  %11 = load i32, ptr %5, align 8, !tbaa !8
+  %12 = icmp sgt i32 %11, 1
+  br i1 %12, label %13, label %15, !prof !11
 
 13:                                               ; preds = %9
-  %14 = load i32, ptr %5, align 4, !tbaa !8
-  %15 = icmp sgt i32 %14, 1
-  br i1 %15, label %16, label %18, !prof !11
-
-16:                                               ; preds = %13
-  %17 = add nsw i32 %14, -1
-  store i32 %17, ptr %5, align 4, !tbaa !8
+  %14 = add nsw i32 %11, -1
+  store i32 %14, ptr %5, align 4, !tbaa !8
   br label %lean_dec.exit19
 
-18:                                               ; preds = %13
-  %.not.i = icmp eq i32 %14, 0
+18:                                               ; preds = %9
+  %.not.i = icmp eq i32 %11, 0
   br i1 %.not.i, label %lean_dec.exit19, label %19
 
 19:                                               ; preds = %18
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %5) #3
   br label %lean_dec.exit19
 
-lean_dec.exit19:                                  ; preds = %19, %18, %16, %9
+lean_dec.exit19:                                  ; preds = %19, %18, %13
   %20 = getelementptr i8, ptr %6, i64 8
   %.val28 = load i64, ptr %20, align 8, !tbaa !12
-  %21 = ptrtoint ptr %6 to i64
-  %22 = and i64 %21, 1
-  %.not29 = icmp eq i64 %22, 0
-  br i1 %.not29, label %23, label %lean_dec.exit18
-
-23:                                               ; preds = %lean_dec.exit19
-  %24 = load i32, ptr %6, align 4, !tbaa !8
-  %25 = icmp sgt i32 %24, 1
-  br i1 %25, label %26, label %28, !prof !11
-
-26:                                               ; preds = %23
-  %27 = add nsw i32 %24, -1
-  store i32 %27, ptr %6, align 4, !tbaa !8
+  %18 = load i32, ptr %6, align 8, !tbaa !8
+  %19 = icmp sgt i32 %18, 1
+  br i1 %19, label %20, label %22, !prof !11
+
+20:                                               ; preds = %lean_dec.exit19
+  %21 = add nsw i32 %18, -1
+  store i32 %21, ptr %6, align 4, !tbaa !8
   br label %lean_dec.exit18
 
-28:                                               ; preds = %23
-  %.not.i20 = icmp eq i32 %24, 0
+28:                                               ; preds = %lean_dec.exit19
+  %.not.i20 = icmp eq i32 %18, 0
   br i1 %.not.i20, label %lean_dec.exit18, label %29
 
 29:                                               ; preds = %28
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %6) #3
   br label %lean_dec.exit18
 
-lean_dec.exit18:                                  ; preds = %29, %28, %26, %lean_dec.exit19
+lean_dec.exit18:                                  ; preds = %29, %28, %20
   %30 = tail call ptr @l_Array_forIn_x27Unsafe_loop___at_Std_Time_Database_TZdb_inst___spec__1(ptr noundef %0, ptr poison, ptr poison, ptr noundef %3, ptr noundef %4, i64 noundef %.val, i64 noundef %.val28, ptr noundef %7, ptr noundef %8)
   %31 = ptrtoint ptr %4 to i64
   %32 = and i64 %31, 1
diff --git a/bench/lean4/optimized/Topological.ll b/bench/lean4/optimized/Topological.ll
index c36be0a7629..e880a8551d9 100644
--- a/bench/lean4/optimized/Topological.ll
+++ b/bench/lean4/optimized/Topological.ll
@@ -987,26 +987,17 @@ lean_alloc_closure.exit:                          ; preds = %6
   store ptr %4, ptr %17, align 8, !tbaa !10
   %18 = getelementptr inbounds nuw i8, ptr %7, i64 56
   store ptr %0, ptr %18, align 8, !tbaa !10
-  %19 = ptrtoint ptr %7 to i64
-  %20 = and i64 %19, 1
-  %.not.i = icmp eq i64 %20, 0
-  br i1 %.not.i, label %21, label %lean_inc.exit.i
-
-21:                                               ; preds = %lean_alloc_closure.exit
-  store i32 2, ptr %7, align 4, !tbaa !4
-  br label %lean_inc.exit.i
-
-lean_inc.exit.i:                                  ; preds = %21, %lean_alloc_closure.exit
+  store i32 2, ptr %7, align 8, !tbaa !4
   tail call void @lean_inc_heartbeat() #3
   %22 = tail call noalias ptr @mi_malloc_small(i64 noundef 32) #3
   %23 = icmp eq ptr %22, null
   br i1 %23, label %24, label %l_Lake_recFetch___rarg.exit
 
-24:                                               ; preds = %lean_inc.exit.i
+24:                                               ; preds = %lean_alloc_closure.exit
   tail call void @lean_internal_panic_out_of_memory() #4
   unreachable
 
-l_Lake_recFetch___rarg.exit:                      ; preds = %lean_inc.exit.i
+l_Lake_recFetch___rarg.exit:                      ; preds = %lean_alloc_closure.exit
   %25 = getelementptr inbounds nuw i8, ptr %22, i64 4
   store i32 1, ptr %22, align 4, !tbaa !4
   store i32 -184549344, ptr %25, align 4
@@ -2061,26 +2052,17 @@ lean_alloc_closure.exit:                          ; preds = %7
   store ptr %3, ptr %19, align 8, !tbaa !10
   %20 = getelementptr inbounds nuw i8, ptr %8, i64 64
   store ptr %0, ptr %20, align 8, !tbaa !10
-  %21 = ptrtoint ptr %8 to i64
-  %22 = and i64 %21, 1
-  %.not.i = icmp eq i64 %22, 0
-  br i1 %.not.i, label %23, label %lean_inc.exit.i
-
-23:                                               ; preds = %lean_alloc_closure.exit
-  store i32 2, ptr %8, align 4, !tbaa !4
-  br label %lean_inc.exit.i
-
-lean_inc.exit.i:                                  ; preds = %23, %lean_alloc_closure.exit
+  store i32 2, ptr %8, align 8, !tbaa !4
   tail call void @lean_inc_heartbeat() #3
   %24 = tail call noalias ptr @mi_malloc_small(i64 noundef 32) #3
   %25 = icmp eq ptr %24, null
   br i1 %25, label %26, label %l_Lake_recFetch___rarg.exit
 
-26:                                               ; preds = %lean_inc.exit.i
+26:                                               ; preds = %lean_alloc_closure.exit
   tail call void @lean_internal_panic_out_of_memory() #4
   unreachable
 
-l_Lake_recFetch___rarg.exit:                      ; preds = %lean_inc.exit.i
+l_Lake_recFetch___rarg.exit:                      ; preds = %lean_alloc_closure.exit
   %27 = getelementptr inbounds nuw i8, ptr %24, i64 4
   store i32 1, ptr %24, align 4, !tbaa !4
   store i32 -184549344, ptr %27, align 4
diff --git a/bench/lean4/optimized/TypeName.ll b/bench/lean4/optimized/TypeName.ll
index 80fc6deca17..8637a84afac 100644
--- a/bench/lean4/optimized/TypeName.ll
+++ b/bench/lean4/optimized/TypeName.ll
@@ -14220,56 +14220,44 @@ lean_dec.exit:                                    ; preds = %15, %14, %12, %5
 define ptr @l_Array_forIn_x27Unsafe_loop___at___private_Lean_Elab_Deriving_TypeName_0__Lean_Elab_deriveTypeNameInstance___spec__3___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3, ptr noundef %4, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8) local_unnamed_addr #0 {
   %10 = getelementptr i8, ptr %3, i64 8
   %.val = load i64, ptr %10, align 8, !tbaa !15
-  %11 = ptrtoint ptr %3 to i64
-  %12 = and i64 %11, 1
-  %.not = icmp eq i64 %12, 0
-  br i1 %.not, label %13, label %lean_dec.exit19
+  %11 = load i32, ptr %3, align 8, !tbaa !8
+  %12 = icmp sgt i32 %11, 1
+  br i1 %12, label %13, label %15, !prof !11
 
 13:                                               ; preds = %9
-  %14 = load i32, ptr %3, align 4, !tbaa !8
-  %15 = icmp sgt i32 %14, 1
-  br i1 %15, label %16, label %18, !prof !11
-
-16:                                               ; preds = %13
-  %17 = add nsw i32 %14, -1
-  store i32 %17, ptr %3, align 4, !tbaa !8
+  %14 = add nsw i32 %11, -1
+  store i32 %14, ptr %3, align 4, !tbaa !8
   br label %lean_dec.exit19
 
-18:                                               ; preds = %13
-  %.not.i = icmp eq i32 %14, 0
+18:                                               ; preds = %9
+  %.not.i = icmp eq i32 %11, 0
   br i1 %.not.i, label %lean_dec.exit19, label %19
 
 19:                                               ; preds = %18
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %3) #4
   br label %lean_dec.exit19
 
-lean_dec.exit19:                                  ; preds = %19, %18, %16, %9
+lean_dec.exit19:                                  ; preds = %19, %18, %13
   %20 = getelementptr i8, ptr %4, i64 8
   %.val28 = load i64, ptr %20, align 8, !tbaa !15
-  %21 = ptrtoint ptr %4 to i64
-  %22 = and i64 %21, 1
-  %.not29 = icmp eq i64 %22, 0
-  br i1 %.not29, label %23, label %lean_dec.exit18
-
-23:                                               ; preds = %lean_dec.exit19
-  %24 = load i32, ptr %4, align 4, !tbaa !8
-  %25 = icmp sgt i32 %24, 1
-  br i1 %25, label %26, label %28, !prof !11
+  %18 = load i32, ptr %4, align 8, !tbaa !8
+  %19 = icmp sgt i32 %18, 1
+  br i1 %19, label %26, label %22, !prof !11
 
-26:                                               ; preds = %23
-  %27 = add nsw i32 %24, -1
+26:                                               ; preds = %lean_dec.exit19
+  %27 = add nsw i32 %18, -1
   store i32 %27, ptr %4, align 4, !tbaa !8
   br label %lean_dec.exit18
 
-28:                                               ; preds = %23
-  %.not.i20 = icmp eq i32 %24, 0
+28:                                               ; preds = %lean_dec.exit19
+  %.not.i20 = icmp eq i32 %18, 0
   br i1 %.not.i20, label %lean_dec.exit18, label %29
 
 29:                                               ; preds = %28
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %4) #4
   br label %lean_dec.exit18
 
-lean_dec.exit18:                                  ; preds = %29, %28, %26, %lean_dec.exit19
+lean_dec.exit18:                                  ; preds = %29, %28, %26
   %30 = tail call ptr @l_Array_forIn_x27Unsafe_loop___at___private_Lean_Elab_Deriving_TypeName_0__Lean_Elab_deriveTypeNameInstance___spec__3(ptr poison, ptr poison, ptr noundef %2, i64 noundef %.val, i64 noundef %.val28, ptr noundef %5, ptr noundef %6, ptr noundef %7, ptr noundef %8)
   %31 = ptrtoint ptr %2 to i64
   %32 = and i64 %31, 1
diff --git a/bench/lean4/optimized/WorkspaceConfig.ll b/bench/lean4/optimized/WorkspaceConfig.ll
index 6a1ed338d3c..c514fdf8b7b 100644
--- a/bench/lean4/optimized/WorkspaceConfig.ll
+++ b/bench/lean4/optimized/WorkspaceConfig.ll
@@ -581,93 +581,81 @@ declare ptr @l_Lean_RBNode_insert___at_Lean_NameMap_insert___spec__1___rarg(ptr
 define ptr @l_Array_foldlMUnsafe_fold___at_Lake_WorkspaceConfig_instConfigMeta___spec__1___boxed(ptr noundef %0, ptr noundef %1, ptr noundef %2, ptr noundef %3) local_unnamed_addr #0 {
   %5 = getelementptr i8, ptr %1, i64 8
   %.val = load i64, ptr %5, align 8, !tbaa !11
-  %6 = ptrtoint ptr %1 to i64
-  %7 = and i64 %6, 1
-  %.not = icmp eq i64 %7, 0
-  br i1 %.not, label %8, label %lean_dec.exit10
+  %6 = load i32, ptr %1, align 8, !tbaa !4
+  %7 = icmp sgt i32 %6, 1
+  br i1 %7, label %8, label %10, !prof !13
 
 8:                                                ; preds = %4
-  %9 = load i32, ptr %1, align 4, !tbaa !4
-  %10 = icmp sgt i32 %9, 1
-  br i1 %10, label %11, label %13, !prof !13
-
-11:                                               ; preds = %8
-  %12 = add nsw i32 %9, -1
-  store i32 %12, ptr %1, align 4, !tbaa !4
+  %9 = add nsw i32 %6, -1
+  store i32 %9, ptr %1, align 4, !tbaa !4
   br label %lean_dec.exit10
 
-13:                                               ; preds = %8
-  %.not.i = icmp eq i32 %9, 0
+13:                                               ; preds = %4
+  %.not.i = icmp eq i32 %6, 0
   br i1 %.not.i, label %lean_dec.exit10, label %14
 
 14:                                               ; preds = %13
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %1) #5
   br label %lean_dec.exit10
 
-lean_dec.exit10:                                  ; preds = %14, %13, %11, %4
+lean_dec.exit10:                                  ; preds = %14, %13, %8
   %15 = getelementptr i8, ptr %2, i64 8
   %.val15 = load i64, ptr %15, align 8, !tbaa !11
-  %16 = ptrtoint ptr %2 to i64
-  %17 = and i64 %16, 1
-  %.not17 = icmp eq i64 %17, 0
-  br i1 %.not17, label %18, label %lean_dec.exit9
-
-18:                                               ; preds = %lean_dec.exit10
-  %19 = load i32, ptr %2, align 4, !tbaa !4
-  %20 = icmp sgt i32 %19, 1
-  br i1 %20, label %21, label %23, !prof !13
-
-21:                                               ; preds = %18
-  %22 = add nsw i32 %19, -1
-  store i32 %22, ptr %2, align 4, !tbaa !4
+  %13 = load i32, ptr %2, align 8, !tbaa !4
+  %14 = icmp sgt i32 %13, 1
+  br i1 %14, label %15, label %17, !prof !13
+
+15:                                               ; preds = %lean_dec.exit10
+  %16 = add nsw i32 %13, -1
+  store i32 %16, ptr %2, align 4, !tbaa !4
   br label %lean_dec.exit9
 
-23:                                               ; preds = %18
-  %.not.i11 = icmp eq i32 %19, 0
+23:                                               ; preds = %lean_dec.exit10
+  %.not.i11 = icmp eq i32 %13, 0
   br i1 %.not.i11, label %lean_dec.exit9, label %24
 
 24:                                               ; preds = %23
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %2) #5
   br label %lean_dec.exit9
 
-lean_dec.exit9:                                   ; preds = %24, %23, %21, %lean_dec.exit10
+lean_dec.exit9:                                   ; preds = %24, %23, %15
   %.not28.i = icmp eq i64 %.val, %.val15
   br i1 %.not28.i, label %l_Array_foldlMUnsafe_fold___at_Lake_WorkspaceConfig_instConfigMeta___spec__1.exit, label %.lr.ph.i
 
 .lr.ph.i:                                         ; preds = %lean_dec.exit9
   %25 = getelementptr inbounds nuw i8, ptr %0, i64 24
-  br label %26
+  br label %20
+
+20:                                               ; preds = %40, %.lr.ph.i
+  %.01730.i = phi i64 [ %.val, %.lr.ph.i ], [ %42, %40 ]
+  %.01929.i = phi ptr [ %3, %.lr.ph.i ], [ %41, %40 ]
+  %21 = getelementptr inbounds nuw ptr, ptr %19, i64 %.01730.i
+  %22 = load ptr, ptr %21, align 8, !tbaa !9
+  %23 = ptrtoint ptr %22 to i64
+  %24 = and i64 %23, 1
+  %.not.i.i = icmp eq i64 %24, 0
+  br i1 %.not.i.i, label %25, label %lean_array_uget.exit.i
 
-26:                                               ; preds = %46, %.lr.ph.i
-  %.01730.i = phi i64 [ %.val, %.lr.ph.i ], [ %48, %46 ]
-  %.01929.i = phi ptr [ %3, %.lr.ph.i ], [ %47, %46 ]
-  %27 = getelementptr inbounds nuw ptr, ptr %25, i64 %.01730.i
-  %28 = load ptr, ptr %27, align 8, !tbaa !9
-  %29 = ptrtoint ptr %28 to i64
-  %30 = and i64 %29, 1
-  %.not.i.i = icmp eq i64 %30, 0
-  br i1 %.not.i.i, label %31, label %lean_array_uget.exit.i
-
-31:                                               ; preds = %26
-  %.val.i.i.i = load i32, ptr %28, align 4, !tbaa !4
-  %32 = icmp sgt i32 %.val.i.i.i, 0
-  br i1 %32, label %33, label %35, !prof !13
-
-33:                                               ; preds = %31
-  %34 = add nuw i32 %.val.i.i.i, 1
-  store i32 %34, ptr %28, align 4, !tbaa !4
+25:                                               ; preds = %20
+  %.val.i.i.i = load i32, ptr %22, align 4, !tbaa !4
+  %26 = icmp sgt i32 %.val.i.i.i, 0
+  br i1 %26, label %27, label %29, !prof !13
+
+26:                                               ; preds = %25
+  %28 = add nuw i32 %.val.i.i.i, 1
+  store i32 %28, ptr %22, align 4, !tbaa !4
   br label %lean_array_uget.exit.i
 
-35:                                               ; preds = %31
+35:                                               ; preds = %25
   %.not.i.i.i = icmp eq i32 %.val.i.i.i, 0
   br i1 %.not.i.i.i, label %lean_array_uget.exit.i, label %36
 
 36:                                               ; preds = %35
-  tail call void @lean_inc_ref_cold(ptr noundef nonnull %28) #5
+  tail call void @lean_inc_ref_cold(ptr noundef nonnull %22) #5
   br label %lean_array_uget.exit.i
 
-lean_array_uget.exit.i:                           ; preds = %36, %35, %33, %26
-  %37 = getelementptr i8, ptr %28, i64 8
+lean_array_uget.exit.i:                           ; preds = %36, %35, %26, %20
+  %37 = getelementptr i8, ptr %22, i64 8
   %.val.i = load ptr, ptr %37, align 8, !tbaa !9
   %38 = ptrtoint ptr %.val.i to i64
   %39 = and i64 %38, 1
@@ -693,13 +681,13 @@ lean_array_uget.exit.i:                           ; preds = %36, %35, %33, %26
   br label %46
 
 46:                                               ; preds = %45, %44, %42, %lean_array_uget.exit.i
-  %47 = tail call ptr @l_Lean_RBNode_insert___at_Lean_NameMap_insert___spec__1___rarg(ptr noundef %.01929.i, ptr noundef %.val.i, ptr noundef nonnull %28) #5
+  %47 = tail call ptr @l_Lean_RBNode_insert___at_Lean_NameMap_insert___spec__1___rarg(ptr noundef %.01929.i, ptr noundef %.val.i, ptr noundef nonnull %22) #5
   %48 = add i64 %.01730.i, 1
   %.not.i16 = icmp eq i64 %48, %.val15
-  br i1 %.not.i16, label %l_Array_foldlMUnsafe_fold___at_Lake_WorkspaceConfig_instConfigMeta___spec__1.exit, label %26
+  br i1 %.not.i16, label %l_Array_foldlMUnsafe_fold___at_Lake_WorkspaceConfig_instConfigMeta___spec__1.exit, label %20
 
 l_Array_foldlMUnsafe_fold___at_Lake_WorkspaceConfig_instConfigMeta___spec__1.exit: ; preds = %46, %lean_dec.exit9
-  %.019.lcssa.i = phi ptr [ %3, %lean_dec.exit9 ], [ %47, %46 ]
+  %.019.lcssa.i = phi ptr [ %3, %lean_dec.exit9 ], [ %47, %40 ]
   %49 = ptrtoint ptr %0 to i64
   %50 = and i64 %49, 1
   %.not18 = icmp eq i64 %50, 0
diff --git a/bench/lean4/optimized/expr.ll b/bench/lean4/optimized/expr.ll
index 7643fc495b9..630b8b0d585 100644
--- a/bench/lean4/optimized/expr.ll
+++ b/bench/lean4/optimized/expr.ll
@@ -3289,22 +3289,16 @@ define hidden void @_ZN4lean14update_bindingERKNS_4exprES2_S2_(ptr dead_on_unwin
   %16 = getelementptr i8, ptr %5, i64 4
   %.val.i.i.i = load i32, ptr %16, align 4
   %17 = getelementptr inbounds nuw i8, ptr %5, i64 8
-  %18 = ptrtoint ptr %5 to i64
-  %19 = and i64 %18, 1
-  %.not.i.i.i = icmp eq i64 %19, 0
-  br i1 %.not.i.i.i, label %20, label %_ZN4lean12binding_infoERKNS_4exprE.exit
-
-20:                                               ; preds = %15
   %.val.i.i.i.i = load i32, ptr %5, align 4, !tbaa !21
   %21 = icmp sgt i32 %.val.i.i.i.i, 0
   br i1 %21, label %22, label %24, !prof !24
 
-22:                                               ; preds = %20
+22:                                               ; preds = %15
   %23 = add nuw nsw i32 %.val.i.i.i.i, 1
   store i32 %23, ptr %5, align 4, !tbaa !21
   br label %_ZN4lean12binding_infoERKNS_4exprE.exit
 
-24:                                               ; preds = %20
+24:                                               ; preds = %15
   %.not.i.i.i.i = icmp eq i32 %.val.i.i.i.i, 0
   br i1 %.not.i.i.i.i, label %_ZN4lean12binding_infoERKNS_4exprE.exit, label %25
 
@@ -3313,8 +3307,8 @@ define hidden void @_ZN4lean14update_bindingERKNS_4exprES2_S2_(ptr dead_on_unwin
   %.pre.i.i = load ptr, ptr %1, align 8, !tbaa !10
   br label %_ZN4lean12binding_infoERKNS_4exprE.exit
 
-_ZN4lean12binding_infoERKNS_4exprE.exit:          ; preds = %15, %22, %24, %25
-  %26 = phi ptr [ %5, %15 ], [ %5, %22 ], [ %5, %24 ], [ %.pre.i.i, %25 ]
+_ZN4lean12binding_infoERKNS_4exprE.exit:          ; preds = %22, %24, %25
+  %26 = phi ptr [ %5, %19 ], [ %5, %21 ], [ %.pre.i.i, %22 ]
   %27 = tail call zeroext i8 @lean_expr_binder_info(ptr noundef %26)
   %28 = zext i8 %27 to i32
   %.mask = and i32 %.val.i.i.i, -16777216
diff --git a/bench/lean4/optimized/inductive.ll b/bench/lean4/optimized/inductive.ll
index 3a4bcecf7a1..e1146681b1a 100644
--- a/bench/lean4/optimized/inductive.ll
+++ b/bench/lean4/optimized/inductive.ll
@@ -780,23 +780,17 @@ define hidden noundef zeroext i1 @_ZN4lean12is_inductiveERKNS_11environmentERKNS
   %.val.i.i.i = load i32, ptr %9, align 4
   %.mask.i = and i32 %.val.i.i.i, -16777216
   %10 = icmp eq i32 %.mask.i, 83886080
-  %11 = ptrtoint ptr %8 to i64
-  %12 = and i64 %11, 1
-  %.not.i.i.i = icmp eq i64 %12, 0
-  br i1 %.not.i.i.i, label %13, label %_ZN4lean8optionalINS_13constant_infoEED2Ev.exit
+  %11 = load i32, ptr %8, align 4, !tbaa !8
+  %12 = icmp sgt i32 %11, 1
+  br i1 %12, label %13, label %15, !prof !11
 
 13:                                               ; preds = %6
-  %14 = load i32, ptr %8, align 4, !tbaa !8
-  %15 = icmp sgt i32 %14, 1
-  br i1 %15, label %16, label %18, !prof !11
-
-16:                                               ; preds = %13
-  %17 = add nsw i32 %14, -1
-  store i32 %17, ptr %8, align 4, !tbaa !8
+  %14 = add nsw i32 %11, -1
+  store i32 %14, ptr %8, align 4, !tbaa !8
   br label %_ZN4lean8optionalINS_13constant_infoEED2Ev.exit
 
-18:                                               ; preds = %13
-  %.not.i.i.i.i = icmp eq i32 %14, 0
+18:                                               ; preds = %6
+  %.not.i.i.i.i = icmp eq i32 %11, 0
   br i1 %.not.i.i.i.i, label %_ZN4lean8optionalINS_13constant_infoEED2Ev.exit, label %19
 
 19:                                               ; preds = %18
@@ -810,8 +804,8 @@ define hidden noundef zeroext i1 @_ZN4lean12is_inductiveERKNS_11environmentERKNS
   call void @__clang_call_terminate(ptr %22) #15
   unreachable
 
-_ZN4lean8optionalINS_13constant_infoEED2Ev.exit:  ; preds = %2, %6, %16, %18, %19
-  %spec.select = phi i1 [ %10, %6 ], [ %10, %16 ], [ %10, %18 ], [ %10, %19 ], [ false, %2 ]
+_ZN4lean8optionalINS_13constant_infoEED2Ev.exit:  ; preds = %2, %13, %18, %19
+  %spec.select = phi i1 [ %10, %13 ], [ %10, %15 ], [ %10, %16 ], [ false, %2 ]
   call void @llvm.lifetime.end.p0(ptr nonnull %3)
   ret i1 %spec.select
 }
@@ -877,23 +871,17 @@ define hidden noundef zeroext i1 @_ZN4lean14is_constructorERKNS_11environmentERK
   %.val.i.i.i = load i32, ptr %9, align 4
   %.mask.i = and i32 %.val.i.i.i, -16777216
   %10 = icmp eq i32 %.mask.i, 100663296
-  %11 = ptrtoint ptr %8 to i64
-  %12 = and i64 %11, 1
-  %.not.i.i.i = icmp eq i64 %12, 0
-  br i1 %.not.i.i.i, label %13, label %_ZN4lean8optionalINS_13constant_infoEED2Ev.exit
+  %11 = load i32, ptr %8, align 4, !tbaa !8
+  %12 = icmp sgt i32 %11, 1
+  br i1 %12, label %13, label %15, !prof !11
 
 13:                                               ; preds = %6
-  %14 = load i32, ptr %8, align 4, !tbaa !8
-  %15 = icmp sgt i32 %14, 1
-  br i1 %15, label %16, label %18, !prof !11
-
-16:                                               ; preds = %13
-  %17 = add nsw i32 %14, -1
-  store i32 %17, ptr %8, align 4, !tbaa !8
+  %14 = add nsw i32 %11, -1
+  store i32 %14, ptr %8, align 4, !tbaa !8
   br label %_ZN4lean8optionalINS_13constant_infoEED2Ev.exit
 
-18:                                               ; preds = %13
-  %.not.i.i.i.i = icmp eq i32 %14, 0
+18:                                               ; preds = %6
+  %.not.i.i.i.i = icmp eq i32 %11, 0
   br i1 %.not.i.i.i.i, label %_ZN4lean8optionalINS_13constant_infoEED2Ev.exit, label %19
 
 19:                                               ; preds = %18
@@ -907,8 +895,8 @@ define hidden noundef zeroext i1 @_ZN4lean14is_constructorERKNS_11environmentERK
   call void @__clang_call_terminate(ptr %22) #15
   unreachable
 
-_ZN4lean8optionalINS_13constant_infoEED2Ev.exit:  ; preds = %2, %6, %16, %18, %19
-  %spec.select = phi i1 [ %10, %6 ], [ %10, %16 ], [ %10, %18 ], [ %10, %19 ], [ false, %2 ]
+_ZN4lean8optionalINS_13constant_infoEED2Ev.exit:  ; preds = %2, %13, %18, %19
+  %spec.select = phi i1 [ %10, %13 ], [ %10, %15 ], [ %10, %16 ], [ false, %2 ]
   call void @llvm.lifetime.end.p0(ptr nonnull %3)
   ret i1 %spec.select
 }
@@ -929,23 +917,17 @@ define hidden noundef zeroext i1 @_ZN4lean11is_recursorERKNS_11environmentERKNS_
   %.val.i.i.i = load i32, ptr %9, align 4
   %.mask.i = and i32 %.val.i.i.i, -16777216
   %10 = icmp eq i32 %.mask.i, 117440512
-  %11 = ptrtoint ptr %8 to i64
-  %12 = and i64 %11, 1
-  %.not.i.i.i = icmp eq i64 %12, 0
-  br i1 %.not.i.i.i, label %13, label %_ZN4lean8optionalINS_13constant_infoEED2Ev.exit
+  %11 = load i32, ptr %8, align 4, !tbaa !8
+  %12 = icmp sgt i32 %11, 1
+  br i1 %12, label %13, label %15, !prof !11
 
 13:                                               ; preds = %6
-  %14 = load i32, ptr %8, align 4, !tbaa !8
-  %15 = icmp sgt i32 %14, 1
-  br i1 %15, label %16, label %18, !prof !11
-
-16:                                               ; preds = %13
-  %17 = add nsw i32 %14, -1
-  store i32 %17, ptr %8, align 4, !tbaa !8
+  %14 = add nsw i32 %11, -1
+  store i32 %14, ptr %8, align 4, !tbaa !8
   br label %_ZN4lean8optionalINS_13constant_infoEED2Ev.exit
 
-18:                                               ; preds = %13
-  %.not.i.i.i.i = icmp eq i32 %14, 0
+18:                                               ; preds = %6
+  %.not.i.i.i.i = icmp eq i32 %11, 0
   br i1 %.not.i.i.i.i, label %_ZN4lean8optionalINS_13constant_infoEED2Ev.exit, label %19
 
 19:                                               ; preds = %18
@@ -959,8 +941,8 @@ define hidden noundef zeroext i1 @_ZN4lean11is_recursorERKNS_11environmentERKNS_
   call void @__clang_call_terminate(ptr %22) #15
   unreachable
 
-_ZN4lean8optionalINS_13constant_infoEED2Ev.exit:  ; preds = %2, %6, %16, %18, %19
-  %spec.select = phi i1 [ %10, %6 ], [ %10, %16 ], [ %10, %18 ], [ %10, %19 ], [ false, %2 ]
+_ZN4lean8optionalINS_13constant_infoEED2Ev.exit:  ; preds = %2, %13, %18, %19
+  %spec.select = phi i1 [ %10, %13 ], [ %10, %15 ], [ %10, %16 ], [ false, %2 ]
   call void @llvm.lifetime.end.p0(ptr nonnull %3)
   ret i1 %spec.select
 }
@@ -995,23 +977,17 @@ _ZN4lean14is_constructorERKNS_11environmentERKNS_4nameE.exit.thread: ; preds = %
   %.val.i.i.i.i = load i32, ptr %16, align 4
   %.mask.i.i5 = and i32 %.val.i.i.i.i, -16777216
   %17 = icmp eq i32 %.mask.i.i5, 100663296
-  %18 = ptrtoint ptr %15 to i64
-  %19 = and i64 %18, 1
-  %.not.i.i.i.i = icmp eq i64 %19, 0
-  br i1 %.not.i.i.i.i, label %20, label %_ZN4lean14is_constructorERKNS_11environmentERKNS_4nameE.exit
+  %18 = load i32, ptr %15, align 4, !tbaa !8
+  %19 = icmp sgt i32 %18, 1
+  br i1 %19, label %20, label %22, !prof !11
 
 20:                                               ; preds = %13
-  %21 = load i32, ptr %15, align 4, !tbaa !8
-  %22 = icmp sgt i32 %21, 1
-  br i1 %22, label %23, label %25, !prof !11
-
-23:                                               ; preds = %20
-  %24 = add nsw i32 %21, -1
-  store i32 %24, ptr %15, align 4, !tbaa !8
+  %21 = add nsw i32 %18, -1
+  store i32 %21, ptr %15, align 4, !tbaa !8
   call void @llvm.lifetime.end.p0(ptr nonnull %4)
   br i1 %17, label %30, label %43
 
-25:                                               ; preds = %20
+25:                                               ; preds = %13
   %.not.i.i.i.i.i = icmp eq i32 %21, 0
   br i1 %.not.i.i.i.i.i, label %_ZN4lean14is_constructorERKNS_11environmentERKNS_4nameE.exit, label %26
 
@@ -1026,11 +1002,11 @@ _ZN4lean14is_constructorERKNS_11environmentERKNS_4nameE.exit.thread: ; preds = %
   call void @__clang_call_terminate(ptr %29) #15
   unreachable
 
-_ZN4lean14is_constructorERKNS_11environmentERKNS_4nameE.exit: ; preds = %13, %25, %26
+_ZN4lean14is_constructorERKNS_11environmentERKNS_4nameE.exit: ; preds = %25, %26
   call void @llvm.lifetime.end.p0(ptr nonnull %4)
   br i1 %17, label %30, label %43
 
-30:                                               ; preds = %23, %_ZN4lean14is_constructorERKNS_11environmentERKNS_4nameE.exit
+30:                                               ; preds = %20, %_ZN4lean14is_constructorERKNS_11environmentERKNS_4nameE.exit
   %31 = load ptr, ptr %5, align 8, !tbaa !3
   %32 = getelementptr inbounds nuw i8, ptr %31, i64 8
   store i8 1, ptr %0, align 8, !tbaa !20
@@ -1060,7 +1036,7 @@ _ZN4lean14is_constructorERKNS_11environmentERKNS_4nameE.exit: ; preds = %13, %25
   call void @lean_inc_ref_cold(ptr noundef nonnull %34)
   br label %_ZN4lean8optionalINS_4nameEEC2ERKS1_.exit
 
-43:                                               ; preds = %23, %_ZN4lean14is_constructorERKNS_11environmentERKNS_4nameE.exit.thread, %_ZN4lean14is_constructorERKNS_11environmentERKNS_4nameE.exit, %3
+43:                                               ; preds = %20, %_ZN4lean14is_constructorERKNS_11environmentERKNS_4nameE.exit.thread, %_ZN4lean14is_constructorERKNS_11environmentERKNS_4nameE.exit, %3
   store i8 0, ptr %0, align 8, !tbaa !20
   br label %_ZN4lean8optionalINS_4nameEEC2ERKS1_.exit
 
diff --git a/bench/lean4/optimized/lcnf.ll b/bench/lean4/optimized/lcnf.ll
index 574a767165d..b52b7dad1bf 100644
--- a/bench/lean4/optimized/lcnf.ll
+++ b/bench/lean4/optimized/lcnf.ll
@@ -15106,23 +15106,17 @@ _ZNK4lean16elab_environment4findERKNS_4nameE.exit: ; preds = %5, %12, %14, %15
   %.val.i.i.i = load i32, ptr %26, align 4
   %.mask.i = and i32 %.val.i.i.i, -16777216
   %27 = icmp eq i32 %.mask.i, 67108864
-  %28 = ptrtoint ptr %25 to i64
-  %29 = and i64 %28, 1
-  %.not.i.i.i3 = icmp eq i64 %29, 0
-  br i1 %.not.i.i.i3, label %30, label %_ZN4lean8optionalINS_13constant_infoEED2Ev.exit
+  %28 = load i32, ptr %25, align 4, !tbaa !8
+  %29 = icmp sgt i32 %28, 1
+  br i1 %29, label %30, label %32, !prof !11
 
 30:                                               ; preds = %23
-  %31 = load i32, ptr %25, align 4, !tbaa !8
-  %32 = icmp sgt i32 %31, 1
-  br i1 %32, label %33, label %35, !prof !11
-
-33:                                               ; preds = %30
-  %34 = add nsw i32 %31, -1
-  store i32 %34, ptr %25, align 4, !tbaa !8
+  %31 = add nsw i32 %28, -1
+  store i32 %31, ptr %25, align 4, !tbaa !8
   br label %_ZN4lean8optionalINS_13constant_infoEED2Ev.exit
 
-35:                                               ; preds = %30
-  %.not.i.i.i.i4 = icmp eq i32 %31, 0
+35:                                               ; preds = %23
+  %.not.i.i.i.i4 = icmp eq i32 %28, 0
   br i1 %.not.i.i.i.i4, label %_ZN4lean8optionalINS_13constant_infoEED2Ev.exit, label %36
 
 36:                                               ; preds = %35
@@ -15136,8 +15130,8 @@ _ZNK4lean16elab_environment4findERKNS_4nameE.exit: ; preds = %5, %12, %14, %15
   call void @__clang_call_terminate(ptr %39) #16
   unreachable
 
-_ZN4lean8optionalINS_13constant_infoEED2Ev.exit:  ; preds = %_ZNK4lean16elab_environment4findERKNS_4nameE.exit, %23, %33, %35, %36
-  %40 = phi i1 [ %27, %23 ], [ %27, %33 ], [ %27, %35 ], [ %27, %36 ], [ false, %_ZNK4lean16elab_environment4findERKNS_4nameE.exit ]
+_ZN4lean8optionalINS_13constant_infoEED2Ev.exit:  ; preds = %_ZNK4lean16elab_environment4findERKNS_4nameE.exit, %30, %35, %36
+  %40 = phi i1 [ %27, %30 ], [ %27, %32 ], [ %27, %33 ], [ false, %_ZNK4lean16elab_environment4findERKNS_4nameE.exit ]
   call void @llvm.lifetime.end.p0(ptr nonnull %4)
   ret i1 %40
 }
@@ -16854,23 +16848,17 @@ define linkonce_odr hidden noundef i32 @_ZN4lean10to_lcnf_fn23get_constructor_nf
   %24 = load ptr, ptr %23, align 8, !tbaa !3
   %25 = getelementptr inbounds nuw i8, ptr %24, i64 40
   %26 = load ptr, ptr %25, align 8, !tbaa !3
-  %27 = ptrtoint ptr %22 to i64
-  %28 = and i64 %27, 1
-  %.not.i.i = icmp eq i64 %28, 0
-  br i1 %.not.i.i, label %29, label %_ZN4lean10object_refD2Ev.exit
+  %27 = load i32, ptr %22, align 8, !tbaa !8
+  %28 = icmp sgt i32 %27, 1
+  br i1 %28, label %29, label %31, !prof !11
 
 29:                                               ; preds = %21
-  %30 = load i32, ptr %22, align 4, !tbaa !8
-  %31 = icmp sgt i32 %30, 1
-  br i1 %31, label %32, label %34, !prof !11
-
-32:                                               ; preds = %29
-  %33 = add nsw i32 %30, -1
-  store i32 %33, ptr %22, align 4, !tbaa !8
+  %30 = add nsw i32 %27, -1
+  store i32 %30, ptr %22, align 4, !tbaa !8
   br label %_ZN4lean10object_refD2Ev.exit
 
-34:                                               ; preds = %29
-  %.not.i.i.i3 = icmp eq i32 %30, 0
+34:                                               ; preds = %21
+  %.not.i.i.i3 = icmp eq i32 %27, 0
   br i1 %.not.i.i.i3, label %_ZN4lean10object_refD2Ev.exit, label %35
 
 35:                                               ; preds = %34
@@ -16884,7 +16872,7 @@ define linkonce_odr hidden noundef i32 @_ZN4lean10to_lcnf_fn23get_constructor_nf
   call void @__clang_call_terminate(ptr %38) #16
   unreachable
 
-_ZN4lean10object_refD2Ev.exit:                    ; preds = %21, %32, %34, %35
+_ZN4lean10object_refD2Ev.exit:                    ; preds = %29, %34, %35
   %39 = ptrtoint ptr %26 to i64
   %40 = lshr i64 %39, 1
   %41 = trunc i64 %40 to i32
diff --git a/bench/lean4/optimized/local_ctx.ll b/bench/lean4/optimized/local_ctx.ll
index d8be5956596..d081abd3c22 100644
--- a/bench/lean4/optimized/local_ctx.ll
+++ b/bench/lean4/optimized/local_ctx.ll
@@ -390,22 +390,16 @@ define hidden void @_ZN4lean10local_declC2ERKS0_RKNS_4exprE(ptr noundef nonnull
   %4 = load ptr, ptr %1, align 8, !tbaa !8
   %5 = getelementptr inbounds nuw i8, ptr %4, i64 8
   %6 = load ptr, ptr %5, align 8, !tbaa !8
-  %7 = ptrtoint ptr %4 to i64
-  %8 = and i64 %7, 1
-  %.not.i.i.i = icmp eq i64 %8, 0
-  br i1 %.not.i.i.i, label %9, label %_ZNK4lean10local_decl8get_infoEv.exit
-
-9:                                                ; preds = %3
-  %.val.i.i.i.i = load i32, ptr %4, align 4, !tbaa !10
-  %10 = icmp sgt i32 %.val.i.i.i.i, 0
-  br i1 %10, label %11, label %13, !prof !13
+  %.val.i.i.i.i = load i32, ptr %4, align 8, !tbaa !10
+  %7 = icmp sgt i32 %.val.i.i.i.i, 0
+  br i1 %7, label %8, label %10, !prof !13
 
-11:                                               ; preds = %9
+11:                                               ; preds = %3
   %12 = add nuw nsw i32 %.val.i.i.i.i, 1
   store i32 %12, ptr %4, align 4, !tbaa !10
   br label %_ZNK4lean10local_decl8get_infoEv.exit
 
-13:                                               ; preds = %9
+13:                                               ; preds = %3
   %.not.i.i.i.i = icmp eq i32 %.val.i.i.i.i, 0
   br i1 %.not.i.i.i.i, label %_ZNK4lean10local_decl8get_infoEv.exit, label %14
 
@@ -414,8 +408,8 @@ define hidden void @_ZN4lean10local_declC2ERKS0_RKNS_4exprE(ptr noundef nonnull
   %.pre.i.i = load ptr, ptr %1, align 8, !tbaa !8
   br label %_ZNK4lean10local_decl8get_infoEv.exit
 
-_ZNK4lean10local_decl8get_infoEv.exit:            ; preds = %3, %11, %13, %14
-  %15 = phi ptr [ %4, %3 ], [ %4, %11 ], [ %4, %13 ], [ %.pre.i.i, %14 ]
+_ZNK4lean10local_decl8get_infoEv.exit:            ; preds = %11, %13, %14
+  %15 = phi ptr [ %4, %8 ], [ %4, %10 ], [ %.pre.i.i, %11 ]
   %16 = getelementptr inbounds nuw i8, ptr %4, i64 24
   %17 = getelementptr inbounds nuw i8, ptr %4, i64 16
   %18 = ptrtoint ptr %6 to i64
diff --git a/bench/lean4/optimized/sharecommon.ll b/bench/lean4/optimized/sharecommon.ll
index 2d8fe9a13d6..b6f5fe36b6e 100644
--- a/bench/lean4/optimized/sharecommon.ll
+++ b/bench/lean4/optimized/sharecommon.ll
@@ -2547,30 +2547,24 @@ _ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__
 74:                                               ; preds = %72
   %75 = getelementptr inbounds nuw i8, ptr %73, i64 8
   %76 = load ptr, ptr %75, align 8, !tbaa !3
-  %77 = ptrtoint ptr %73 to i64
-  %78 = and i64 %77, 1
-  %.not = icmp eq i64 %78, 0
-  br i1 %.not, label %79, label %_ZL8lean_decP11lean_object.exit
+  %77 = load i32, ptr %73, align 8, !tbaa !13
+  %78 = icmp sgt i32 %77, 1
+  br i1 %78, label %79, label %81, !prof !16
 
 79:                                               ; preds = %74
-  %80 = load i32, ptr %73, align 4, !tbaa !13
-  %81 = icmp sgt i32 %80, 1
-  br i1 %81, label %82, label %84, !prof !16
-
-82:                                               ; preds = %79
-  %83 = add nsw i32 %80, -1
-  store i32 %83, ptr %73, align 4, !tbaa !13
+  %80 = add nsw i32 %77, -1
+  store i32 %80, ptr %73, align 4, !tbaa !13
   br label %_ZL8lean_decP11lean_object.exit
 
-84:                                               ; preds = %79
-  %.not.i.i = icmp eq i32 %80, 0
+84:                                               ; preds = %74
+  %.not.i.i = icmp eq i32 %77, 0
   br i1 %.not.i.i, label %_ZL8lean_decP11lean_object.exit, label %85
 
 85:                                               ; preds = %84
   tail call void @lean_dec_ref_cold(ptr noundef nonnull %73)
   br label %_ZL8lean_decP11lean_object.exit
 
-_ZL8lean_decP11lean_object.exit:                  ; preds = %82, %84, %85, %74
+_ZL8lean_decP11lean_object.exit:                  ; preds = %79, %84, %85
   %86 = getelementptr inbounds nuw i8, ptr %0, i64 48
   %87 = getelementptr inbounds nuw i8, ptr %0, i64 56
   %88 = load ptr, ptr %87, align 8, !tbaa !73
@@ -2699,7 +2693,7 @@ _ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__
   br label %_ZNSt6vectorIP11lean_objectSaIS1_EE9push_backERKS1_.exit
 
 _ZNSt6vectorIP11lean_objectSaIS1_EE9push_backERKS1_.exit: ; preds = %_ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__normal_iteratorIPS1_S3_EEDpOT_.exit.i29, %120, %_ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__normal_iteratorIPS1_S3_EEDpOT_.exit.i21, %91, %_ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__normal_iteratorIPS1_S3_EEDpOT_.exit.i13, %49, %_ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__normal_iteratorIPS1_S3_EEDpOT_.exit.i, %11
-  %.0 = phi i1 [ true, %11 ], [ true, %_ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__normal_iteratorIPS1_S3_EEDpOT_.exit.i ], [ true, %49 ], [ true, %_ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__normal_iteratorIPS1_S3_EEDpOT_.exit.i13 ], [ true, %91 ], [ true, %_ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__normal_iteratorIPS1_S3_EEDpOT_.exit.i21 ], [ false, %120 ], [ false, %_ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__normal_iteratorIPS1_S3_EEDpOT_.exit.i29 ]
+  %.0 = phi i1 [ true, %11 ], [ true, %_ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__normal_iteratorIPS1_S3_EEDpOT_.exit.i ], [ true, %49 ], [ true, %_ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__normal_iteratorIPS1_S3_EEDpOT_.exit.i13 ], [ true, %88 ], [ true, %_ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__normal_iteratorIPS1_S3_EEDpOT_.exit.i21 ], [ false, %117 ], [ false, %_ZNSt6vectorIP11lean_objectSaIS1_EE17_M_realloc_insertIJRKS1_EEEvN9__gnu_cxx17__normal_iteratorIPS1_S3_EEDpOT_.exit.i29 ]
   ret i1 %.0
 }
 
diff --git a/bench/lean4/optimized/tcp.ll b/bench/lean4/optimized/tcp.ll
index ab511bc0e5a..277e84fe128 100644
--- a/bench/lean4/optimized/tcp.ll
+++ b/bench/lean4/optimized/tcp.ll
@@ -266,22 +266,16 @@ define noalias noundef nonnull ptr @lean_uv_tcp_connect(ptr noundef %0, ptr noun
   %9 = getelementptr inbounds nuw i8, ptr %8, i64 8
   store ptr %0, ptr %9, align 8, !tbaa !47
   store ptr %8, ptr %7, align 8, !tbaa !48
-  %10 = ptrtoint ptr %0 to i64
-  %11 = and i64 %10, 1
-  %.not = icmp eq i64 %11, 0
-  br i1 %.not, label %12, label %_ZL8lean_incP11lean_object.exit
-
-12:                                               ; preds = %3
-  %.val.i = load i32, ptr %0, align 4, !tbaa !40
-  %13 = icmp sgt i32 %.val.i, 0
-  br i1 %13, label %14, label %16, !prof !9
+  %.val.i = load i32, ptr %0, align 8, !tbaa !40
+  %10 = icmp sgt i32 %.val.i, 0
+  br i1 %10, label %11, label %13, !prof !9
 
-14:                                               ; preds = %12
+14:                                               ; preds = %3
   %15 = add nuw nsw i32 %.val.i, 1
   store i32 %15, ptr %0, align 4, !tbaa !40
   br label %_ZL8lean_incP11lean_object.exit
 
-16:                                               ; preds = %12
+16:                                               ; preds = %3
   %.not.i29 = icmp eq i32 %.val.i, 0
   br i1 %.not.i29, label %_ZL8lean_incP11lean_object.exit, label %17
 
@@ -289,7 +283,7 @@ define noalias noundef nonnull ptr @lean_uv_tcp_connect(ptr noundef %0, ptr noun
   call void @lean_inc_ref_cold(ptr noundef nonnull %0)
   br label %_ZL8lean_incP11lean_object.exit
 
-_ZL8lean_incP11lean_object.exit:                  ; preds = %17, %16, %14, %3
+_ZL8lean_incP11lean_object.exit:                  ; preds = %17, %16, %14
   %18 = ptrtoint ptr %6 to i64
   %19 = and i64 %18, 1
   %.not35 = icmp eq i64 %19, 0
@@ -344,7 +338,7 @@ _ZL8lean_incP11lean_object.exit22:                ; preds = %25, %24, %22, %_ZL8
   br label %37
 
 37:                                               ; preds = %36, %33
-  %38 = phi i32 [ %.pr, %36 ], [ %34, %33 ]
+  %38 = phi i32 [ %.pr, %33 ], [ %34, %30 ]
   %39 = icmp sgt i32 %38, 1
   br i1 %39, label %40, label %42, !prof !52
 
@@ -361,28 +355,25 @@ _ZL8lean_incP11lean_object.exit22:                ; preds = %25, %24, %22, %_ZL8
   call void @lean_dec_ref_cold(ptr noundef nonnull %6)
   br label %_ZL8lean_decP11lean_object.exit23
 
-_ZL8lean_decP11lean_object.exit23:                ; preds = %35, %43, %42, %40, %29
-  br i1 %.not, label %44, label %_ZL8lean_decP11lean_object.exit24
-
-44:                                               ; preds = %_ZL8lean_decP11lean_object.exit23
-  %45 = load i32, ptr %0, align 4, !tbaa !40
-  %46 = icmp sgt i32 %45, 1
-  br i1 %46, label %47, label %49, !prof !9
+_ZL8lean_decP11lean_object.exit23:                ; preds = %26, %37, %42, %40, %32
+  %41 = load i32, ptr %0, align 4, !tbaa !40
+  %42 = icmp sgt i32 %41, 1
+  br i1 %42, label %43, label %45, !prof !9
 
-47:                                               ; preds = %44
-  %48 = add nsw i32 %45, -1
+47:                                               ; preds = %_ZL8lean_decP11lean_object.exit23
+  %48 = add nsw i32 %41, -1
   store i32 %48, ptr %0, align 4, !tbaa !40
   br label %_ZL8lean_decP11lean_object.exit24
 
-49:                                               ; preds = %44
-  %.not.i = icmp eq i32 %45, 0
+49:                                               ; preds = %_ZL8lean_decP11lean_object.exit23
+  %.not.i = icmp eq i32 %41, 0
   br i1 %.not.i, label %_ZL8lean_decP11lean_object.exit24, label %50
 
 50:                                               ; preds = %49
   call void @lean_dec_ref_cold(ptr noundef nonnull %0)
   br label %_ZL8lean_decP11lean_object.exit24
 
-_ZL8lean_decP11lean_object.exit24:                ; preds = %50, %49, %47, %_ZL8lean_decP11lean_object.exit23
+_ZL8lean_decP11lean_object.exit24:                ; preds = %50, %49, %43
   %51 = load ptr, ptr %7, align 8, !tbaa !48
   call void @free(ptr noundef %51) #15
   call void @free(ptr noundef nonnull %7) #15
@@ -407,9 +398,9 @@ _ZL8lean_decP11lean_object.exit24:                ; preds = %50, %49, %47, %_ZL8
   unreachable
 
 _ZL23lean_io_result_mk_errorP11lean_object.exit:  ; preds = %56, %_ZL8lean_decP11lean_object.exit24
-  %.sink50 = phi ptr [ %53, %_ZL8lean_decP11lean_object.exit24 ], [ %57, %56 ]
-  %.sink47 = phi i32 [ 16908312, %_ZL8lean_decP11lean_object.exit24 ], [ 131096, %56 ]
-  %.sink = phi ptr [ %52, %_ZL8lean_decP11lean_object.exit24 ], [ %6, %56 ]
+  %.sink50 = phi ptr [ %53, %_ZL8lean_decP11lean_object.exit24 ], [ %57, %52 ]
+  %.sink47 = phi i32 [ 16908312, %_ZL8lean_decP11lean_object.exit24 ], [ 131096, %52 ]
+  %.sink = phi ptr [ %52, %_ZL8lean_decP11lean_object.exit24 ], [ %6, %52 ]
   %60 = getelementptr inbounds nuw i8, ptr %.sink50, i64 4
   store i32 1, ptr %.sink50, align 4, !tbaa !40
   store i32 %.sink47, ptr %60, align 4
diff --git a/bench/libigl/optimized/unproject.ll b/bench/libigl/optimized/unproject.ll
index d73523b0805..e4f675f0dd4 100644
--- a/bench/libigl/optimized/unproject.ll
+++ b/bench/libigl/optimized/unproject.ll
@@ -190,46 +190,46 @@ define weak_odr dso_local void @_ZN3igl9unprojectIdEEN5Eigen6MatrixIT_Li3ELi1ELi
   %27 = fmul <2 x double> %.sroa.0.i.i.sroa.7.24.vec.insert, splat (double 2.000000e+00)
   %28 = fadd <2 x double> %27, splat (double -1.000000e+00)
   %29 = shufflevector <2 x double> %28, <2 x double> poison, <2 x i32> <i32 1, i32 1>
-  %30 = getelementptr inbounds nuw i8, ptr %6, i64 96
+  %30 = getelementptr inbounds nuw i8, ptr %6, i64 112
   %31 = load <2 x double>, ptr %30, align 16, !tbaa !14
   %32 = fmul <2 x double> %29, %31
-  %33 = shufflevector <2 x double> %28, <2 x double> poison, <2 x i32> zeroinitializer
-  %34 = getelementptr inbounds nuw i8, ptr %6, i64 64
+  %33 = shufflevector <2 x double> %26, <2 x double> poison, <2 x i32> zeroinitializer
+  %34 = getelementptr inbounds nuw i8, ptr %6, i64 16
   %35 = load <2 x double>, ptr %34, align 16, !tbaa !14
   %36 = fmul <2 x double> %33, %35
-  %37 = shufflevector <2 x double> %26, <2 x double> poison, <2 x i32> zeroinitializer
-  %38 = load <2 x double>, ptr %6, align 16, !tbaa !14
-  %39 = fmul <2 x double> %37, %38
-  %40 = getelementptr inbounds nuw i8, ptr %6, i64 32
-  %41 = load <2 x double>, ptr %40, align 16, !tbaa !14
   %.sroa.0.i.i.sroa.0.8.vec.extract = extractelement <2 x double> %26, i64 1
   %42 = bitcast double %.sroa.0.i.i.sroa.0.8.vec.extract to <1 x double>
   %43 = shufflevector <1 x double> %42, <1 x double> poison, <2 x i32> zeroinitializer
-  %44 = fmul <2 x double> %41, %43
-  %45 = fadd <2 x double> %39, %44
-  %46 = fadd <2 x double> %36, %45
-  %47 = fadd <2 x double> %32, %46
-  %48 = getelementptr inbounds nuw i8, ptr %6, i64 112
-  %49 = load <2 x double>, ptr %48, align 16, !tbaa !14
-  %50 = fmul <2 x double> %29, %49
-  %51 = getelementptr inbounds nuw i8, ptr %6, i64 16
+  %39 = getelementptr inbounds nuw i8, ptr %6, i64 48
+  %40 = load <2 x double>, ptr %39, align 16, !tbaa !14
+  %41 = fmul <2 x double> %38, %40
+  %47 = fadd <2 x double> %36, %41
+  %43 = shufflevector <2 x double> %28, <2 x double> poison, <2 x i32> zeroinitializer
+  %44 = getelementptr inbounds nuw i8, ptr %6, i64 80
+  %45 = load <2 x double>, ptr %44, align 16, !tbaa !14
+  %46 = fmul <2 x double> %43, %45
+  %47 = fadd <2 x double> %42, %46
+  %48 = fadd <2 x double> %32, %47
+  %49 = shufflevector <2 x double> %48, <2 x double> poison, <2 x i32> <i32 1, i32 1>
+  %50 = fdiv <2 x double> %48, %49
+  %51 = getelementptr inbounds nuw i8, ptr %6, i64 96
   %52 = load <2 x double>, ptr %51, align 16, !tbaa !14
-  %53 = fmul <2 x double> %37, %52
-  %54 = getelementptr inbounds nuw i8, ptr %6, i64 48
+  %53 = fmul <2 x double> %29, %52
+  %54 = getelementptr inbounds nuw i8, ptr %6, i64 64
   %55 = load <2 x double>, ptr %54, align 16, !tbaa !14
   %56 = fmul <2 x double> %43, %55
-  %57 = fadd <2 x double> %53, %56
-  %58 = getelementptr inbounds nuw i8, ptr %6, i64 80
-  %59 = load <2 x double>, ptr %58, align 16, !tbaa !14
-  %60 = fmul <2 x double> %33, %59
-  %61 = fadd <2 x double> %57, %60
-  %62 = fadd <2 x double> %50, %61
-  %63 = shufflevector <2 x double> %62, <2 x double> poison, <2 x i32> <i32 1, i32 1>
-  %64 = fdiv <2 x double> %47, %63
-  %65 = fdiv <2 x double> %62, %63
-  %66 = extractelement <2 x double> %65, i64 0
+  %57 = load <2 x double>, ptr %6, align 16, !tbaa !14
+  %58 = fmul <2 x double> %33, %57
+  %59 = getelementptr inbounds nuw i8, ptr %6, i64 32
+  %60 = load <2 x double>, ptr %59, align 16, !tbaa !14
+  %61 = fmul <2 x double> %60, %38
+  %62 = fadd <2 x double> %58, %61
+  %63 = fadd <2 x double> %56, %62
+  %64 = fadd <2 x double> %53, %63
+  %65 = fdiv <2 x double> %64, %49
+  %66 = extractelement <2 x double> %50, i64 0
   call void @llvm.lifetime.end.p0(ptr nonnull %6)
-  store <2 x double> %64, ptr %0, align 8, !tbaa !14
+  store <2 x double> %65, ptr %0, align 8, !tbaa !14
   %67 = getelementptr inbounds nuw i8, ptr %0, i64 16
   store double %66, ptr %67, align 8, !tbaa !18
   ret void
diff --git a/bench/libigl/optimized/vertex_triangle_adjacency.ll b/bench/libigl/optimized/vertex_triangle_adjacency.ll
index ab5994f2ab9..19a164ed71a 100644
--- a/bench/libigl/optimized/vertex_triangle_adjacency.ll
+++ b/bench/libigl/optimized/vertex_triangle_adjacency.ll
@@ -2937,25 +2937,16 @@ _ZN5Eigen8internal23check_size_for_overflowIiEEvm.exit.i.i.i.i.i: ; preds = %32
   %45 = getelementptr inbounds nuw i8, ptr %.sroa.0.0, i64 4
   %46 = load ptr, ptr %3, align 8, !tbaa !108
   %47 = ptrtoint ptr %45 to i64
-  %48 = and i64 %47, 3
-  %.not.i.i.i.i.i.i.i.i.i.i.i.i = icmp eq i64 %48, 0
-  br i1 %.not.i.i.i.i.i.i.i.i.i.i.i.i, label %49, label %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i
-
-49:                                               ; preds = %42
-  %50 = lshr exact i64 %47, 2
-  %51 = sub nsw i64 0, %50
-  %52 = and i64 %51, 3
-  %53 = call i64 @llvm.smin.i64(i64 %52, i64 %44)
-  br label %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i
-
-_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i: ; preds = %49, %42
-  %.0.i.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %53, %49 ], [ %44, %42 ]
-  %54 = sub nsw i64 %44, %.0.i.i.i.i.i.i.i.i.i.i.i.i
-  %55 = sdiv i64 %54, 4
-  %56 = shl nsw i64 %55, 2
-  %57 = add nsw i64 %56, %.0.i.i.i.i.i.i.i.i.i.i.i.i
-  %58 = icmp sgt i64 %.0.i.i.i.i.i.i.i.i.i.i.i.i, 0
-  br i1 %58, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i
+  %47 = lshr exact i64 %47, 2
+  %48 = sub nsw i64 0, %47
+  %49 = and i64 %48, 3
+  %50 = call i64 @llvm.smin.i64(i64 %49, i64 %43)
+  %51 = sub nsw i64 %43, %50
+  %52 = sdiv i64 %51, 4
+  %53 = shl nsw i64 %52, 2
+  %54 = add nsw i64 %53, %50
+  %55 = icmp sgt i64 %50, 0
+  br i1 %55, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i
 
 .lr.ph.i.i.i.i.i.i.i.i.i.i.i.i:                   ; preds = %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i
   %.05.i.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %62, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i ], [ 0, %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i ]
@@ -2964,19 +2955,19 @@ _ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.
   %61 = load i32, ptr %60, align 4, !tbaa !19
   store i32 %61, ptr %59, align 4, !tbaa !19
   %62 = add nuw nsw i64 %.05.i.i.i.i.i.i.i.i.i.i.i.i, 1
-  %exitcond.not.i.i.i.i.i.i.i.i.i.i.i.i = icmp eq i64 %62, %.0.i.i.i.i.i.i.i.i.i.i.i.i
+  %exitcond.not.i.i.i.i.i.i.i.i.i.i.i.i = icmp eq i64 %62, %50
   br i1 %exitcond.not.i.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i, !llvm.loop !114
 
 _ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i: ; preds = %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i, %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i
-  %63 = icmp sgt i64 %54, 3
+  %63 = icmp sgt i64 %51, 3
   br i1 %63, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i, label %._crit_edge.i.i.i.i.i.i.i.i.i.i.i
 
 ._crit_edge.i.i.i.i.i.i.i.i.i.i.i:                ; preds = %.lr.ph.i.i.i.i.i.i.i.i.i.i.i, %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i
-  %64 = icmp slt i64 %57, %44
+  %64 = icmp slt i64 %54, %44
   br i1 %64, label %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit
 
 .lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i:                 ; preds = %._crit_edge.i.i.i.i.i.i.i.i.i.i.i, %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i
-  %.05.i18.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %68, %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i ], [ %57, %._crit_edge.i.i.i.i.i.i.i.i.i.i.i ]
+  %.05.i18.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %68, %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i ], [ %54, %._crit_edge.i.i.i.i.i.i.i.i.i.i.i ]
   %65 = getelementptr inbounds i32, ptr %45, i64 %.05.i18.i.i.i.i.i.i.i.i.i.i.i
   %66 = getelementptr inbounds i32, ptr %46, i64 %.05.i18.i.i.i.i.i.i.i.i.i.i.i
   %67 = load i32, ptr %66, align 4, !tbaa !19
@@ -2986,13 +2977,13 @@ _ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dens
   br i1 %exitcond.not.i19.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit, label %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i, !llvm.loop !114
 
 .lr.ph.i.i.i.i.i.i.i.i.i.i.i:                     ; preds = %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i
-  %.021.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %72, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i ], [ %.0.i.i.i.i.i.i.i.i.i.i.i.i, %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i ]
+  %.021.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %72, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i ], [ %50, %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i ]
   %69 = getelementptr inbounds i32, ptr %45, i64 %.021.i.i.i.i.i.i.i.i.i.i.i
   %70 = getelementptr inbounds i32, ptr %46, i64 %.021.i.i.i.i.i.i.i.i.i.i.i
   %71 = load <2 x i64>, ptr %70, align 1, !tbaa !115
   store <2 x i64> %71, ptr %69, align 16, !tbaa !115
   %72 = add nsw i64 %.021.i.i.i.i.i.i.i.i.i.i.i, 4
-  %73 = icmp slt i64 %72, %57
+  %73 = icmp slt i64 %72, %54
   br i1 %73, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i, label %._crit_edge.i.i.i.i.i.i.i.i.i.i.i, !llvm.loop !116
 
 _ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit: ; preds = %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i, %._crit_edge.i.i.i.i.i.i.i.i.i.i.i
@@ -3031,7 +3022,7 @@ thread-pre-split.i.i.i.i.i.i.i:                   ; preds = %_ZN5Eigen16CommaIni
   br i1 %exitcond.not.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEaSERKS3_.exit, label %.lr.ph.i.i.i.i.i.i.i.i.i, !llvm.loop !117
 
 .lr.ph.i.i.i.i.i.i.i.i:                           ; preds = %74, %.lr.ph.i.i.i.i.i.i.i.i
-  %.011.i.i.i.i.i.i.i.i = phi i64 [ %88, %.lr.ph.i.i.i.i.i.i.i.i ], [ 0, %74 ]
+  %.011.i.i.i.i.i.i.i.i = phi i64 [ %88, %.lr.ph.i.i.i.i.i.i.i.i ], [ 0, %71 ]
   %85 = getelementptr inbounds nuw i32, ptr %75, i64 %.011.i.i.i.i.i.i.i.i
   %86 = getelementptr inbounds nuw i32, ptr %.sroa.0.0, i64 %.011.i.i.i.i.i.i.i.i
   %87 = load <2 x i64>, ptr %86, align 16, !tbaa !115
@@ -3079,7 +3070,7 @@ thread-pre-split.i.i.i.i.i.i.i44:                 ; preds = %_ZN5Eigen15PlainObj
   br i1 %exitcond.not.i.i.i.i.i.i.i.i.i49, label %_ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEaSIS1_EERS1_RKNS_9DenseBaseIT_EE.exit, label %.lr.ph.i.i.i.i.i.i.i.i.i47, !llvm.loop !117
 
 .lr.ph.i.i.i.i.i.i.i.i50:                         ; preds = %93, %.lr.ph.i.i.i.i.i.i.i.i50
-  %.011.i.i.i.i.i.i.i.i51 = phi i64 [ %107, %.lr.ph.i.i.i.i.i.i.i.i50 ], [ 0, %93 ]
+  %.011.i.i.i.i.i.i.i.i51 = phi i64 [ %107, %.lr.ph.i.i.i.i.i.i.i.i50 ], [ 0, %90 ]
   %104 = getelementptr inbounds nuw i32, ptr %95, i64 %.011.i.i.i.i.i.i.i.i51
   %105 = getelementptr inbounds nuw i32, ptr %90, i64 %.011.i.i.i.i.i.i.i.i51
   %106 = load <2 x i64>, ptr %105, align 16, !tbaa !115
@@ -3128,7 +3119,7 @@ _ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEC2IliEERKT_RKT0_.exit: ; preds = %_ZN5E
   br label %.preheader
 
 .preheader:                                       ; preds = %.preheader.lr.ph, %129
-  %indvars.iv89 = phi i64 [ 0, %.preheader.lr.ph ], [ %indvars.iv.next90, %129 ]
+  %indvars.iv89 = phi i64 [ 0, %.preheader.lr.ph ], [ %indvars.iv.next90, %126 ]
   %123 = getelementptr i32, ptr %121, i64 %indvars.iv89
   %124 = trunc nuw nsw i64 %indvars.iv89 to i32
   br label %130
@@ -3155,28 +3146,28 @@ _ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEC2IliEERKT_RKT0_.exit: ; preds = %_ZN5E
   br i1 %exitcond92.not, label %._crit_edge75, label %.preheader, !llvm.loop !120
 
 130:                                              ; preds = %.preheader, %130
-  %indvars.iv85 = phi i64 [ 0, %.preheader ], [ %indvars.iv.next86, %130 ]
+  %indvars.iv85 = phi i64 [ 0, %.preheader ], [ %indvars.iv.next86, %127 ]
   %131 = mul nuw nsw i64 %119, %indvars.iv85
   %132 = getelementptr i32, ptr %123, i64 %131
+  %130 = load i32, ptr %129, align 4, !tbaa !19
+  %131 = sext i32 %130 to i64
+  %132 = getelementptr inbounds i32, ptr %.pre93, i64 %131
   %133 = load i32, ptr %132, align 4, !tbaa !19
   %134 = sext i32 %133 to i64
-  %135 = getelementptr inbounds i32, ptr %.pre93, i64 %134
-  %136 = load i32, ptr %135, align 4, !tbaa !19
+  %135 = getelementptr inbounds i32, ptr %119, i64 %134
+  store i32 %121, ptr %135, align 4, !tbaa !19
+  %136 = load i32, ptr %129, align 4, !tbaa !19
   %137 = sext i32 %136 to i64
-  %138 = getelementptr inbounds i32, ptr %122, i64 %137
-  store i32 %124, ptr %138, align 4, !tbaa !19
-  %139 = load i32, ptr %132, align 4, !tbaa !19
-  %140 = sext i32 %139 to i64
-  %141 = getelementptr inbounds i32, ptr %.pre93, i64 %140
-  %142 = load i32, ptr %141, align 4, !tbaa !19
-  %143 = add nsw i32 %142, 1
-  store i32 %143, ptr %141, align 4, !tbaa !19
+  %138 = getelementptr inbounds i32, ptr %.pre93, i64 %137
+  %139 = load i32, ptr %138, align 4, !tbaa !19
+  %140 = add nsw i32 %139, 1
+  store i32 %140, ptr %138, align 4, !tbaa !19
   %indvars.iv.next86 = add nuw nsw i64 %indvars.iv85, 1
   %exitcond88.not = icmp eq i64 %indvars.iv.next86, 3
   br i1 %exitcond88.not, label %129, label %130, !llvm.loop !121
 
 .body:                                            ; preds = %127, %40, %.body53, %125
-  %.pn41 = phi { ptr, i32 } [ %111, %.body53 ], [ %126, %125 ], [ %128, %127 ], [ %41, %40 ]
+  %.pn41 = phi { ptr, i32 } [ %111, %.body53 ], [ %126, %122 ], [ %128, %124 ], [ %41, %40 ]
   %144 = load ptr, ptr %5, align 8, !tbaa !108
   call void @free(ptr noundef %144) #18
   call void @llvm.lifetime.end.p0(ptr nonnull %5)
@@ -3304,25 +3295,16 @@ _ZN5Eigen8internal23check_size_for_overflowIiEEvm.exit.i.i.i.i.i: ; preds = %32
   %45 = getelementptr inbounds nuw i8, ptr %.sroa.0.0, i64 4
   %46 = load ptr, ptr %3, align 8, !tbaa !108
   %47 = ptrtoint ptr %45 to i64
-  %48 = and i64 %47, 3
-  %.not.i.i.i.i.i.i.i.i.i.i.i.i = icmp eq i64 %48, 0
-  br i1 %.not.i.i.i.i.i.i.i.i.i.i.i.i, label %49, label %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i
-
-49:                                               ; preds = %42
-  %50 = lshr exact i64 %47, 2
-  %51 = sub nsw i64 0, %50
-  %52 = and i64 %51, 3
-  %53 = call i64 @llvm.smin.i64(i64 %52, i64 %44)
-  br label %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i
-
-_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i: ; preds = %49, %42
-  %.0.i.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %53, %49 ], [ %44, %42 ]
-  %54 = sub nsw i64 %44, %.0.i.i.i.i.i.i.i.i.i.i.i.i
-  %55 = sdiv i64 %54, 4
-  %56 = shl nsw i64 %55, 2
-  %57 = add nsw i64 %56, %.0.i.i.i.i.i.i.i.i.i.i.i.i
-  %58 = icmp sgt i64 %.0.i.i.i.i.i.i.i.i.i.i.i.i, 0
-  br i1 %58, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i
+  %47 = lshr exact i64 %47, 2
+  %48 = sub nsw i64 0, %47
+  %49 = and i64 %48, 3
+  %50 = call i64 @llvm.smin.i64(i64 %49, i64 %43)
+  %51 = sub nsw i64 %43, %50
+  %52 = sdiv i64 %51, 4
+  %53 = shl nsw i64 %52, 2
+  %54 = add nsw i64 %53, %50
+  %55 = icmp sgt i64 %50, 0
+  br i1 %55, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i
 
 .lr.ph.i.i.i.i.i.i.i.i.i.i.i.i:                   ; preds = %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i
   %.05.i.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %62, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i ], [ 0, %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i ]
@@ -3331,19 +3313,19 @@ _ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.
   %61 = load i32, ptr %60, align 4, !tbaa !19
   store i32 %61, ptr %59, align 4, !tbaa !19
   %62 = add nuw nsw i64 %.05.i.i.i.i.i.i.i.i.i.i.i.i, 1
-  %exitcond.not.i.i.i.i.i.i.i.i.i.i.i.i = icmp eq i64 %62, %.0.i.i.i.i.i.i.i.i.i.i.i.i
+  %exitcond.not.i.i.i.i.i.i.i.i.i.i.i.i = icmp eq i64 %62, %50
   br i1 %exitcond.not.i.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i, !llvm.loop !114
 
 _ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i: ; preds = %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i, %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i
-  %63 = icmp sgt i64 %54, 3
+  %63 = icmp sgt i64 %51, 3
   br i1 %63, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i, label %._crit_edge.i.i.i.i.i.i.i.i.i.i.i
 
 ._crit_edge.i.i.i.i.i.i.i.i.i.i.i:                ; preds = %.lr.ph.i.i.i.i.i.i.i.i.i.i.i, %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i
-  %64 = icmp slt i64 %57, %44
+  %64 = icmp slt i64 %54, %44
   br i1 %64, label %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit
 
 .lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i:                 ; preds = %._crit_edge.i.i.i.i.i.i.i.i.i.i.i, %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i
-  %.05.i18.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %68, %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i ], [ %57, %._crit_edge.i.i.i.i.i.i.i.i.i.i.i ]
+  %.05.i18.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %68, %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i ], [ %54, %._crit_edge.i.i.i.i.i.i.i.i.i.i.i ]
   %65 = getelementptr inbounds i32, ptr %45, i64 %.05.i18.i.i.i.i.i.i.i.i.i.i.i
   %66 = getelementptr inbounds i32, ptr %46, i64 %.05.i18.i.i.i.i.i.i.i.i.i.i.i
   %67 = load i32, ptr %66, align 4, !tbaa !19
@@ -3353,13 +3335,13 @@ _ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dens
   br i1 %exitcond.not.i19.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit, label %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i, !llvm.loop !114
 
 .lr.ph.i.i.i.i.i.i.i.i.i.i.i:                     ; preds = %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i
-  %.021.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %72, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i ], [ %.0.i.i.i.i.i.i.i.i.i.i.i.i, %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i ]
+  %.021.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %72, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i ], [ %50, %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i ]
   %69 = getelementptr inbounds i32, ptr %45, i64 %.021.i.i.i.i.i.i.i.i.i.i.i
   %70 = getelementptr inbounds i32, ptr %46, i64 %.021.i.i.i.i.i.i.i.i.i.i.i
   %71 = load <2 x i64>, ptr %70, align 1, !tbaa !115
   store <2 x i64> %71, ptr %69, align 16, !tbaa !115
   %72 = add nsw i64 %.021.i.i.i.i.i.i.i.i.i.i.i, 4
-  %73 = icmp slt i64 %72, %57
+  %73 = icmp slt i64 %72, %54
   br i1 %73, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i, label %._crit_edge.i.i.i.i.i.i.i.i.i.i.i, !llvm.loop !116
 
 _ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit: ; preds = %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i, %._crit_edge.i.i.i.i.i.i.i.i.i.i.i
@@ -3398,7 +3380,7 @@ thread-pre-split.i.i.i.i.i.i.i:                   ; preds = %_ZN5Eigen16CommaIni
   br i1 %exitcond.not.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEaSERKS3_.exit, label %.lr.ph.i.i.i.i.i.i.i.i.i, !llvm.loop !117
 
 .lr.ph.i.i.i.i.i.i.i.i:                           ; preds = %74, %.lr.ph.i.i.i.i.i.i.i.i
-  %.011.i.i.i.i.i.i.i.i = phi i64 [ %88, %.lr.ph.i.i.i.i.i.i.i.i ], [ 0, %74 ]
+  %.011.i.i.i.i.i.i.i.i = phi i64 [ %88, %.lr.ph.i.i.i.i.i.i.i.i ], [ 0, %71 ]
   %85 = getelementptr inbounds nuw i32, ptr %75, i64 %.011.i.i.i.i.i.i.i.i
   %86 = getelementptr inbounds nuw i32, ptr %.sroa.0.0, i64 %.011.i.i.i.i.i.i.i.i
   %87 = load <2 x i64>, ptr %86, align 16, !tbaa !115
@@ -3446,7 +3428,7 @@ thread-pre-split.i.i.i.i.i.i.i44:                 ; preds = %_ZN5Eigen15PlainObj
   br i1 %exitcond.not.i.i.i.i.i.i.i.i.i49, label %_ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEaSIS1_EERS1_RKNS_9DenseBaseIT_EE.exit, label %.lr.ph.i.i.i.i.i.i.i.i.i47, !llvm.loop !117
 
 .lr.ph.i.i.i.i.i.i.i.i50:                         ; preds = %93, %.lr.ph.i.i.i.i.i.i.i.i50
-  %.011.i.i.i.i.i.i.i.i51 = phi i64 [ %107, %.lr.ph.i.i.i.i.i.i.i.i50 ], [ 0, %93 ]
+  %.011.i.i.i.i.i.i.i.i51 = phi i64 [ %107, %.lr.ph.i.i.i.i.i.i.i.i50 ], [ 0, %90 ]
   %104 = getelementptr inbounds nuw i32, ptr %95, i64 %.011.i.i.i.i.i.i.i.i51
   %105 = getelementptr inbounds nuw i32, ptr %90, i64 %.011.i.i.i.i.i.i.i.i51
   %106 = load <2 x i64>, ptr %105, align 16, !tbaa !115
@@ -3495,7 +3477,7 @@ _ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEC2IliEERKT_RKT0_.exit: ; preds = %_ZN5E
   br label %.preheader
 
 .preheader:                                       ; preds = %.preheader.lr.ph, %129
-  %indvars.iv89 = phi i64 [ 0, %.preheader.lr.ph ], [ %indvars.iv.next90, %129 ]
+  %indvars.iv89 = phi i64 [ 0, %.preheader.lr.ph ], [ %indvars.iv.next90, %126 ]
   %123 = getelementptr i32, ptr %121, i64 %indvars.iv89
   %124 = trunc nuw nsw i64 %indvars.iv89 to i32
   br label %130
@@ -3522,28 +3504,28 @@ _ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEC2IliEERKT_RKT0_.exit: ; preds = %_ZN5E
   br i1 %exitcond92.not, label %._crit_edge75, label %.preheader, !llvm.loop !127
 
 130:                                              ; preds = %.preheader, %130
-  %indvars.iv85 = phi i64 [ 0, %.preheader ], [ %indvars.iv.next86, %130 ]
+  %indvars.iv85 = phi i64 [ 0, %.preheader ], [ %indvars.iv.next86, %127 ]
   %131 = mul nuw nsw i64 %119, %indvars.iv85
   %132 = getelementptr i32, ptr %123, i64 %131
+  %130 = load i32, ptr %129, align 4, !tbaa !19
+  %131 = sext i32 %130 to i64
+  %132 = getelementptr inbounds i32, ptr %.pre93, i64 %131
   %133 = load i32, ptr %132, align 4, !tbaa !19
   %134 = sext i32 %133 to i64
-  %135 = getelementptr inbounds i32, ptr %.pre93, i64 %134
-  %136 = load i32, ptr %135, align 4, !tbaa !19
+  %135 = getelementptr inbounds i32, ptr %119, i64 %134
+  store i32 %121, ptr %135, align 4, !tbaa !19
+  %136 = load i32, ptr %129, align 4, !tbaa !19
   %137 = sext i32 %136 to i64
-  %138 = getelementptr inbounds i32, ptr %122, i64 %137
-  store i32 %124, ptr %138, align 4, !tbaa !19
-  %139 = load i32, ptr %132, align 4, !tbaa !19
-  %140 = sext i32 %139 to i64
-  %141 = getelementptr inbounds i32, ptr %.pre93, i64 %140
-  %142 = load i32, ptr %141, align 4, !tbaa !19
-  %143 = add nsw i32 %142, 1
-  store i32 %143, ptr %141, align 4, !tbaa !19
+  %138 = getelementptr inbounds i32, ptr %.pre93, i64 %137
+  %139 = load i32, ptr %138, align 4, !tbaa !19
+  %140 = add nsw i32 %139, 1
+  store i32 %140, ptr %138, align 4, !tbaa !19
   %indvars.iv.next86 = add nuw nsw i64 %indvars.iv85, 1
   %exitcond88.not = icmp eq i64 %indvars.iv.next86, 3
   br i1 %exitcond88.not, label %129, label %130, !llvm.loop !128
 
 .body:                                            ; preds = %127, %40, %.body53, %125
-  %.pn41 = phi { ptr, i32 } [ %111, %.body53 ], [ %126, %125 ], [ %128, %127 ], [ %41, %40 ]
+  %.pn41 = phi { ptr, i32 } [ %111, %.body53 ], [ %126, %122 ], [ %128, %124 ], [ %41, %40 ]
   %144 = load ptr, ptr %5, align 8, !tbaa !108
   call void @free(ptr noundef %144) #18
   call void @llvm.lifetime.end.p0(ptr nonnull %5)
diff --git a/bench/linux/optimized/af_netlink.ll b/bench/linux/optimized/af_netlink.ll
index 5f51e263b6f..3c0840d7668 100644
--- a/bench/linux/optimized/af_netlink.ll
+++ b/bench/linux/optimized/af_netlink.ll
@@ -2569,7 +2569,7 @@ define internal fastcc range(i32 -16, -17) i32 @netlink_insert(ptr noundef %0, i
 93:                                               ; preds = %89, %87
   %94 = phi ptr [ %88, %87 ], [ %92, %89 ]
   %95 = icmp eq ptr %94, null
-  br i1 %95, label %211, label %96
+  br i1 %95, label %207, label %96
 
 96:                                               ; preds = %93
   call void @llvm.lifetime.start.p0(ptr nonnull %4)
@@ -2728,57 +2728,53 @@ define internal fastcc range(i32 -16, -17) i32 @netlink_insert(ptr noundef %0, i
   %181 = inttoptr i64 %138 to ptr
   store volatile ptr %181, ptr %34, align 8
   call void asm sideeffect ".pushsection .smp_locks,\22a\22\0A.balign 4\0A.long 671f - .\0A.popsection\0A671:\0A\09lock; incl $0", "=*m,*m,~{memory},~{dirflag},~{fpsr},~{flags}"(ptr nonnull elementtype(i32) %166, ptr nonnull elementtype(i32) %166) #23, !srcloc !29
-  %182 = ptrtoint ptr %34 to i64
-  %183 = and i64 %182, 1
-  %184 = icmp eq i64 %183, 0
-  %185 = select i1 %184, ptr %34, ptr null
   call void asm sideeffect "", "~{memory},~{dirflag},~{fpsr},~{flags}"() #23, !srcloc !76
-  store volatile ptr %185, ptr %94, align 8
+  store volatile ptr %34, ptr %94, align 8
   call void asm sideeffect "", "~{memory},~{dirflag},~{fpsr},~{flags}"() #23, !srcloc !77
   %186 = call i8 asm sideeffect "decl %gs:$0\0A\09/* output condition code e*/\0A", "=*m,={@cce},*m,~{memory},~{dirflag},~{fpsr},~{flags}"(ptr nonnull elementtype(i32) getelementptr inbounds nuw (i8, ptr @pcpu_hot, i64 8), ptr nonnull elementtype(i32) getelementptr inbounds nuw (i8, ptr @pcpu_hot, i64 8)) #23, !srcloc !19
   %187 = icmp ult i8 %186, 2
   call void @llvm.assume(i1 %187)
   %188 = icmp eq i8 %186, 0
-  br i1 %188, label %192, label %189, !prof !9
+  br i1 %188, label %188, label %189, !prof !9
 
 189:                                              ; preds = %180
   %190 = call i64 @llvm.read_register.i64(metadata !0)
   %191 = call i64 asm sideeffect "call __SCT__preempt_schedule", "={rsp},{rsp},~{dirflag},~{fpsr},~{flags}"(i64 %190) #23, !srcloc !78
   call void @llvm.write_register.i64(metadata !0, i64 %191)
-  br label %192
+  br label %188
 
-192:                                              ; preds = %189, %180
-  %193 = and i64 %97, 512
-  %194 = icmp eq i64 %193, 0
-  br i1 %194, label %196, label %195
+188:                                              ; preds = %185, %180
+  %189 = and i64 %97, 512
+  %190 = icmp eq i64 %189, 0
+  br i1 %190, label %192, label %191
 
-195:                                              ; preds = %192
+191:                                              ; preds = %188
   call void asm sideeffect "sti", "~{memory},~{dirflag},~{fpsr},~{flags}"() #23, !srcloc !74
-  br label %196
-
-196:                                              ; preds = %195, %192
-  %197 = load volatile i32, ptr %166, align 4
-  %198 = load i32, ptr %40, align 64
-  %199 = lshr i32 %198, 2
-  %200 = mul nuw i32 %199, 3
-  %201 = icmp ugt i32 %197, %200
-  br i1 %201, label %202, label %211
-
-202:                                              ; preds = %196
-  %203 = getelementptr inbounds nuw i8, ptr %18, i64 24
-  %204 = load i32, ptr %203, align 8
-  %205 = add i32 %204, -1
-  %206 = icmp ult i32 %205, %198
-  br i1 %206, label %211, label %207
-
-207:                                              ; preds = %202
-  %208 = getelementptr inbounds nuw i8, ptr %18, i64 64
-  %209 = load ptr, ptr @system_wq, align 8
-  %210 = call zeroext i1 @queue_work_on(i32 noundef 64, ptr noundef %209, ptr noundef nonnull %208) #23
-  br label %211
-
-211:                                              ; preds = %223, %220, %207, %202, %196, %93
-  %212 = phi ptr [ inttoptr (i64 -12 to ptr), %93 ], [ null, %207 ], [ null, %202 ], [ null, %196 ], [ %213, %220 ], [ %213, %223 ]
+  br label %192
+
+192:                                              ; preds = %191, %188
+  %193 = load volatile i32, ptr %166, align 4
+  %194 = load i32, ptr %40, align 64
+  %195 = lshr i32 %194, 2
+  %196 = mul nuw i32 %195, 3
+  %197 = icmp ugt i32 %193, %196
+  br i1 %197, label %198, label %207
+
+198:                                              ; preds = %192
+  %199 = getelementptr inbounds nuw i8, ptr %18, i64 24
+  %200 = load i32, ptr %199, align 8
+  %201 = add i32 %200, -1
+  %202 = icmp ult i32 %201, %194
+  br i1 %202, label %207, label %203
+
+203:                                              ; preds = %198
+  %204 = getelementptr inbounds nuw i8, ptr %18, i64 64
+  %205 = load ptr, ptr @system_wq, align 8
+  %206 = call zeroext i1 @queue_work_on(i32 noundef 64, ptr noundef %205, ptr noundef nonnull %204) #23
+  br label %207
+
+195:                                              ; preds = %219, %216, %203, %198, %192, %93
+  %208 = phi ptr [ inttoptr (i64 -12 to ptr), %93 ], [ null, %203 ], [ null, %198 ], [ null, %192 ], [ %213, %216 ], [ %209, %219 ]
   call void @__rcu_read_unlock() #23
   br label %224
 
@@ -2802,14 +2798,14 @@ define internal fastcc range(i32 -16, -17) i32 @netlink_insert(ptr noundef %0, i
 220:                                              ; preds = %217, %.loopexit
   %221 = and i64 %97, 512
   %222 = icmp eq i64 %221, 0
-  br i1 %222, label %211, label %223
+  br i1 %222, label %207, label %223
 
 223:                                              ; preds = %220
   call void asm sideeffect "sti", "~{memory},~{dirflag},~{fpsr},~{flags}"() #23, !srcloc !74
-  br label %211
+  br label %207
 
-224:                                              ; preds = %211, %129
-  %225 = phi ptr [ %130, %129 ], [ %212, %211 ]
+224:                                              ; preds = %207, %129
+  %225 = phi ptr [ %130, %129 ], [ %208, %207 ]
   %226 = icmp ugt ptr %225, inttoptr (i64 -4096 to ptr)
   %227 = ptrtoint ptr %225 to i64
   %228 = trunc i64 %227 to i32
@@ -2830,7 +2826,7 @@ define internal fastcc range(i32 -16, -17) i32 @netlink_insert(ptr noundef %0, i
   br label %.thread5
 
 .thread5:                                         ; preds = %224, %.thread5.fold.split, %232
-  %233 = phi i32 [ -98, %232 ], [ -75, %224 ], [ %231, %.thread5.fold.split ]
+  %233 = phi i32 [ -98, %228 ], [ -75, %220 ], [ %231, %.thread5.fold.split ]
   %234 = call i32 asm sideeffect ".pushsection .smp_locks,\22a\22\0A.balign 4\0A.long 671f - .\0A.popsection\0A671:\0A\09lock; xaddl $0, $1\0A", "=r,=*m,0,*m,~{memory},~{cc},~{dirflag},~{fpsr},~{flags}"(ptr nonnull elementtype(i32) %19, i32 -1, ptr nonnull elementtype(i32) %19) #23, !srcloc !30
   %235 = icmp eq i32 %234, 1
   br i1 %235, label %239, label %236
@@ -2856,7 +2852,7 @@ define internal fastcc range(i32 -16, -17) i32 @netlink_insert(ptr noundef %0, i
   br label %.thread7
 
 .thread7:                                         ; preds = %236, %238, %240, %239, %2
-  %243 = phi i32 [ %12, %2 ], [ 0, %240 ], [ %233, %239 ], [ %233, %238 ], [ %233, %236 ]
+  %243 = phi i32 [ %12, %2 ], [ 0, %236 ], [ %233, %235 ], [ %233, %234 ], [ %233, %232 ]
   call void @release_sock(ptr noundef %0) #23
   ret i32 %243
 }
diff --git a/bench/linux/optimized/fatent.ll b/bench/linux/optimized/fatent.ll
index 3b20669159e..e138642d2f9 100644
--- a/bench/linux/optimized/fatent.ll
+++ b/bench/linux/optimized/fatent.ll
@@ -2098,22 +2098,10 @@ define internal range(i32 0, 268435456) i32 @fat32_ent_get(ptr noundef readonly
   %2 = getelementptr inbounds nuw i8, ptr %0, i64 8
   %3 = load ptr, ptr %2, align 8
   %4 = load i32, ptr %3, align 4
-  %5 = ptrtoint ptr %3 to i64
-  %6 = and i64 %5, 3
-  %7 = icmp eq i64 %6, 0
-  br i1 %7, label %9, label %8, !prof !31
-
-8:                                                ; preds = %1
-  tail call void asm sideeffect "427: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 427b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 427) #9, !srcloc !41
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 150, i32 2305, i64 12) #9, !srcloc !42
-  tail call void asm sideeffect "428: nop\0A\09.pushsection .discard.instr_end\0A\09.long 428b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 428) #9, !srcloc !43
-  br label %9
-
-9:                                                ; preds = %8, %1
-  %10 = and i32 %4, 268435455
-  %11 = icmp samesign ugt i32 %10, 268435446
-  %12 = select i1 %11, i32 268435455, i32 %10
-  ret i32 %12
+  %5 = and i32 %4, 268435455
+  %6 = icmp samesign ugt i32 %5, 268435446
+  %7 = select i1 %6, i32 268435455, i32 %5
+  ret i32 %7
 }
 
 ; Function Attrs: fn_ret_thunk_extern nounwind null_pointer_is_valid
@@ -2122,9 +2110,9 @@ define internal void @fat32_ent_put(ptr noundef readonly captures(none) %0, i32
   br i1 %3, label %5, label %4, !prof !31
 
 4:                                                ; preds = %2
-  tail call void asm sideeffect "429: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 429b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 429) #9, !srcloc !44
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 189, i32 2305, i64 12) #9, !srcloc !45
-  tail call void asm sideeffect "430: nop\0A\09.pushsection .discard.instr_end\0A\09.long 430b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 430) #9, !srcloc !46
+  tail call void asm sideeffect "429: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 429b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 429) #9, !srcloc !41
+  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 189, i32 2305, i64 12) #9, !srcloc !42
+  tail call void asm sideeffect "430: nop\0A\09.pushsection .discard.instr_end\0A\09.long 430b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 430) #9, !srcloc !43
   br label %5
 
 5:                                                ; preds = %4, %2
@@ -2184,9 +2172,9 @@ define internal void @fat16_ent_set_ptr(ptr noundef captures(none) initializes((
   br i1 %4, label %6, label %5, !prof !31
 
 5:                                                ; preds = %2
-  tail call void asm sideeffect "417: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 417b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 417) #9, !srcloc !47
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 59, i32 2305, i64 12) #9, !srcloc !48
-  tail call void asm sideeffect "418: nop\0A\09.pushsection .discard.instr_end\0A\09.long 418b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 418) #9, !srcloc !49
+  tail call void asm sideeffect "417: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 417b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 417) #9, !srcloc !44
+  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 59, i32 2305, i64 12) #9, !srcloc !45
+  tail call void asm sideeffect "418: nop\0A\09.pushsection .discard.instr_end\0A\09.long 418b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 418) #9, !srcloc !46
   br label %6
 
 6:                                                ; preds = %5, %2
@@ -2206,22 +2194,10 @@ define internal range(i32 0, 268435456) i32 @fat16_ent_get(ptr noundef readonly
   %2 = getelementptr inbounds nuw i8, ptr %0, i64 8
   %3 = load ptr, ptr %2, align 8
   %4 = load i16, ptr %3, align 2
-  %5 = ptrtoint ptr %3 to i64
-  %6 = and i64 %5, 1
-  %7 = icmp eq i64 %6, 0
-  br i1 %7, label %9, label %8, !prof !31
-
-8:                                                ; preds = %1
-  tail call void asm sideeffect "425: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 425b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 425) #9, !srcloc !50
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 141, i32 2305, i64 12) #9, !srcloc !51
-  tail call void asm sideeffect "426: nop\0A\09.pushsection .discard.instr_end\0A\09.long 426b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 426) #9, !srcloc !52
-  br label %9
-
-9:                                                ; preds = %8, %1
-  %10 = zext i16 %4 to i32
-  %11 = icmp ugt i16 %4, -10
-  %12 = select i1 %11, i32 268435455, i32 %10
-  ret i32 %12
+  %5 = zext i16 %4 to i32
+  %6 = icmp ugt i16 %4, -10
+  %7 = select i1 %6, i32 268435455, i32 %5
+  ret i32 %7
 }
 
 ; Function Attrs: fn_ret_thunk_extern nounwind null_pointer_is_valid
@@ -2276,9 +2252,9 @@ define internal void @fat12_ent_blocknr(ptr noundef readonly captures(none) %0,
   br i1 %12, label %14, label %13, !prof !31
 
 13:                                               ; preds = %8, %4
-  tail call void asm sideeffect "409: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 409b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 409) #9, !srcloc !53
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 28, i32 2305, i64 12) #9, !srcloc !54
-  tail call void asm sideeffect "410: nop\0A\09.pushsection .discard.instr_end\0A\09.long 410b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 410) #9, !srcloc !55
+  tail call void asm sideeffect "409: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 409b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 409) #9, !srcloc !47
+  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 28, i32 2305, i64 12) #9, !srcloc !48
+  tail call void asm sideeffect "410: nop\0A\09.pushsection .discard.instr_end\0A\09.long 410b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 410) #9, !srcloc !49
   br label %14
 
 14:                                               ; preds = %13, %8
@@ -2321,9 +2297,9 @@ define internal void @fat12_ent_set_ptr(ptr noundef captures(none) initializes((
   br i1 %13, label %15, label %14, !prof !31
 
 14:                                               ; preds = %12
-  tail call void asm sideeffect "413: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 413b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 413) #9, !srcloc !56
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 47, i32 2305, i64 12) #9, !srcloc !57
-  tail call void asm sideeffect "414: nop\0A\09.pushsection .discard.instr_end\0A\09.long 414b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 414) #9, !srcloc !58
+  tail call void asm sideeffect "413: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 413b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 413) #9, !srcloc !50
+  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 47, i32 2305, i64 12) #9, !srcloc !51
+  tail call void asm sideeffect "414: nop\0A\09.pushsection .discard.instr_end\0A\09.long 414b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 414) #9, !srcloc !52
   %.pre1 = load ptr, ptr %3, align 8
   br label %15
 
@@ -2345,9 +2321,9 @@ define internal void @fat12_ent_set_ptr(ptr noundef captures(none) initializes((
   br i1 %26, label %28, label %27, !prof !31
 
 27:                                               ; preds = %25
-  tail call void asm sideeffect "415: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 415b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 415) #9, !srcloc !59
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 51, i32 2305, i64 12) #9, !srcloc !60
-  tail call void asm sideeffect "416: nop\0A\09.pushsection .discard.instr_end\0A\09.long 416b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 416) #9, !srcloc !61
+  tail call void asm sideeffect "415: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 415b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 415) #9, !srcloc !53
+  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 51, i32 2305, i64 12) #9, !srcloc !54
+  tail call void asm sideeffect "416: nop\0A\09.pushsection .discard.instr_end\0A\09.long 416b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 416) #9, !srcloc !55
   %.pre = load ptr, ptr %3, align 8
   br label %28
 
@@ -2383,9 +2359,9 @@ define internal noundef range(i32 -5, 1) i32 @fat12_ent_bread(ptr noundef %0, pt
   br i1 %11, label %12, label %13, !prof !13
 
 12:                                               ; preds = %4
-  tail call void asm sideeffect "421: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 421b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 421) #9, !srcloc !62
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 74, i32 2305, i64 12) #9, !srcloc !63
-  tail call void asm sideeffect "422: nop\0A\09.pushsection .discard.instr_end\0A\09.long 422b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 422) #9, !srcloc !64
+  tail call void asm sideeffect "421: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 421b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 421) #9, !srcloc !56
+  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 74, i32 2305, i64 12) #9, !srcloc !57
+  tail call void asm sideeffect "422: nop\0A\09.pushsection .discard.instr_end\0A\09.long 422b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 422) #9, !srcloc !58
   %.pre = load ptr, ptr %6, align 8
   br label %13
 
@@ -2433,9 +2409,9 @@ define internal noundef range(i32 -5, 1) i32 @fat12_ent_bread(ptr noundef %0, pt
   br i1 %43, label %45, label %44, !prof !31
 
 44:                                               ; preds = %37
-  tail call void asm sideeffect "413: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 413b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 413) #9, !srcloc !56
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 47, i32 2305, i64 12) #9, !srcloc !57
-  tail call void asm sideeffect "414: nop\0A\09.pushsection .discard.instr_end\0A\09.long 414b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 414) #9, !srcloc !58
+  tail call void asm sideeffect "413: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 413b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 413) #9, !srcloc !50
+  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 47, i32 2305, i64 12) #9, !srcloc !51
+  tail call void asm sideeffect "414: nop\0A\09.pushsection .discard.instr_end\0A\09.long 414b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 414) #9, !srcloc !52
   %.pre5 = load ptr, ptr %5, align 8
   br label %45
 
@@ -2464,9 +2440,9 @@ define internal noundef range(i32 -5, 1) i32 @fat12_ent_bread(ptr noundef %0, pt
   br i1 %61, label %63, label %62, !prof !31
 
 62:                                               ; preds = %54
-  tail call void asm sideeffect "415: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 415b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 415) #9, !srcloc !59
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 51, i32 2305, i64 12) #9, !srcloc !60
-  tail call void asm sideeffect "416: nop\0A\09.pushsection .discard.instr_end\0A\09.long 416b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 416) #9, !srcloc !61
+  tail call void asm sideeffect "415: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 415b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 415) #9, !srcloc !53
+  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 51, i32 2305, i64 12) #9, !srcloc !54
+  tail call void asm sideeffect "416: nop\0A\09.pushsection .discard.instr_end\0A\09.long 416b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 416) #9, !srcloc !55
   %.pre3 = load ptr, ptr %5, align 8
   %.pre4 = load ptr, ptr %35, align 8
   br label %63
@@ -2648,9 +2624,9 @@ define internal noundef range(i32 0, 2) i32 @fat12_ent_next(ptr noundef captures
   br i1 %24, label %25, label %26, !prof !13
 
 25:                                               ; preds = %22
-  tail call void asm sideeffect "431: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 431b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 431) #9, !srcloc !65
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 204, i32 2305, i64 12) #9, !srcloc !66
-  tail call void asm sideeffect "432: nop\0A\09.pushsection .discard.instr_end\0A\09.long 432b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 432) #9, !srcloc !67
+  tail call void asm sideeffect "431: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 431b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 431) #9, !srcloc !59
+  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 204, i32 2305, i64 12) #9, !srcloc !60
+  tail call void asm sideeffect "432: nop\0A\09.pushsection .discard.instr_end\0A\09.long 432b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 432) #9, !srcloc !61
   %.pre2 = load ptr, ptr %4, align 8
   %.pre3 = load ptr, ptr %3, align 8
   %.phi.trans.insert = getelementptr inbounds nuw i8, ptr %.pre3, i64 40
@@ -2669,9 +2645,9 @@ define internal noundef range(i32 0, 2) i32 @fat12_ent_next(ptr noundef captures
   br i1 %32, label %33, label %34, !prof !13
 
 33:                                               ; preds = %26
-  tail call void asm sideeffect "433: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 433b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 433) #9, !srcloc !68
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 206, i32 2305, i64 12) #9, !srcloc !69
-  tail call void asm sideeffect "434: nop\0A\09.pushsection .discard.instr_end\0A\09.long 434b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 434) #9, !srcloc !70
+  tail call void asm sideeffect "433: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 433b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 433) #9, !srcloc !62
+  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 206, i32 2305, i64 12) #9, !srcloc !63
+  tail call void asm sideeffect "434: nop\0A\09.pushsection .discard.instr_end\0A\09.long 434b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 434) #9, !srcloc !64
   %.pre7 = load ptr, ptr %3, align 8
   %.phi.trans.insert8 = getelementptr inbounds nuw i8, ptr %.pre7, i64 40
   %.pre9 = load ptr, ptr %.phi.trans.insert8, align 8
@@ -2699,9 +2675,9 @@ define internal noundef range(i32 0, 2) i32 @fat12_ent_next(ptr noundef captures
   br i1 %44, label %46, label %45, !prof !31
 
 45:                                               ; preds = %42
-  tail call void asm sideeffect "435: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 435b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 435) #9, !srcloc !71
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 214, i32 2305, i64 12) #9, !srcloc !72
-  tail call void asm sideeffect "436: nop\0A\09.pushsection .discard.instr_end\0A\09.long 436b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 436) #9, !srcloc !73
+  tail call void asm sideeffect "435: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 435b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 435) #9, !srcloc !65
+  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 214, i32 2305, i64 12) #9, !srcloc !66
+  tail call void asm sideeffect "436: nop\0A\09.pushsection .discard.instr_end\0A\09.long 436b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 436) #9, !srcloc !67
   %.pre = load ptr, ptr %4, align 8
   br label %46
 
@@ -2715,9 +2691,9 @@ define internal noundef range(i32 0, 2) i32 @fat12_ent_next(ptr noundef captures
   br i1 %52, label %54, label %53, !prof !31
 
 53:                                               ; preds = %46
-  tail call void asm sideeffect "437: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 437b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 437) #9, !srcloc !74
-  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 215, i32 2305, i64 12) #9, !srcloc !75
-  tail call void asm sideeffect "438: nop\0A\09.pushsection .discard.instr_end\0A\09.long 438b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 438) #9, !srcloc !76
+  tail call void asm sideeffect "437: nop\0A\09.pushsection .discard.instr_begin\0A\09.long 437b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 437) #9, !srcloc !68
+  tail call void asm sideeffect "1:\09.byte 0x0f, 0x0b\0A.pushsection __bug_table,\22aw\22\0A2:\09.long 1b - .\09# bug_entry::bug_addr\0A\09.long ${0:c} - .\09# bug_entry::file\0A\09.word ${1:c}\09# bug_entry::line\0A\09.word ${2:c}\09# bug_entry::flags\0A\09.org 2b+${3:c}\0A.popsection\0A998:\0A\09.pushsection .discard.reachable\0A\09.long 998b\0A\09.popsection\0A\09", "i,i,i,i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @.str.3, i32 215, i32 2305, i64 12) #9, !srcloc !69
+  tail call void asm sideeffect "438: nop\0A\09.pushsection .discard.instr_end\0A\09.long 438b - .\0A\09.popsection\0A\09", "i,~{dirflag},~{fpsr},~{flags}"(i32 438) #9, !srcloc !70
   br label %54
 
 54:                                               ; preds = %53, %46
@@ -2866,39 +2842,33 @@ attributes #11 = { nounwind memory(none) }
 !38 = !{i64 2155580933, i64 2155580742, i64 2155580794, i64 2155580840, i64 2155580868}
 !39 = !{i64 2155581007, i64 2155581036, i64 2155581082, i64 2155581140, i64 2155581194, i64 2155581248, i64 2155581303, i64 2155581334, i64 2155581642, i64 2155581648, i64 2155581695, i64 2155581718, i64 2155581744}
 !40 = !{i64 2155582192, i64 2155582003, i64 2155582053, i64 2155582099, i64 2155582127}
-!41 = !{i64 2155585977, i64 2155585786, i64 2155585838, i64 2155585884, i64 2155585912}
-!42 = !{i64 2155586051, i64 2155586080, i64 2155586126, i64 2155586184, i64 2155586238, i64 2155586292, i64 2155586347, i64 2155586378, i64 2155586686, i64 2155586692, i64 2155586739, i64 2155586762, i64 2155586788}
-!43 = !{i64 2155587236, i64 2155587047, i64 2155587097, i64 2155587143, i64 2155587171}
-!44 = !{i64 2155588196, i64 2155588005, i64 2155588057, i64 2155588103, i64 2155588131}
-!45 = !{i64 2155588270, i64 2155588299, i64 2155588345, i64 2155588403, i64 2155588457, i64 2155588511, i64 2155588566, i64 2155588597, i64 2155588905, i64 2155588911, i64 2155588958, i64 2155588981, i64 2155589007}
-!46 = !{i64 2155589455, i64 2155589266, i64 2155589316, i64 2155589362, i64 2155589390}
-!47 = !{i64 2155574014, i64 2155573823, i64 2155573875, i64 2155573921, i64 2155573949}
-!48 = !{i64 2155574088, i64 2155574117, i64 2155574163, i64 2155574221, i64 2155574275, i64 2155574329, i64 2155574384, i64 2155574415, i64 2155574723, i64 2155574729, i64 2155574776, i64 2155574799, i64 2155574825}
-!49 = !{i64 2155575272, i64 2155575083, i64 2155575133, i64 2155575179, i64 2155575207}
-!50 = !{i64 2155583780, i64 2155583589, i64 2155583641, i64 2155583687, i64 2155583715}
-!51 = !{i64 2155583854, i64 2155583883, i64 2155583929, i64 2155583987, i64 2155584041, i64 2155584095, i64 2155584150, i64 2155584181, i64 2155584489, i64 2155584495, i64 2155584542, i64 2155584565, i64 2155584591}
-!52 = !{i64 2155585039, i64 2155584850, i64 2155584900, i64 2155584946, i64 2155584974}
-!53 = !{i64 2155565662, i64 2155565471, i64 2155565523, i64 2155565569, i64 2155565597}
-!54 = !{i64 2155565736, i64 2155565765, i64 2155565811, i64 2155565869, i64 2155565923, i64 2155565977, i64 2155566032, i64 2155566063, i64 2155566371, i64 2155566377, i64 2155566424, i64 2155566447, i64 2155566473}
-!55 = !{i64 2155566920, i64 2155566731, i64 2155566781, i64 2155566827, i64 2155566855}
-!56 = !{i64 2155569844, i64 2155569653, i64 2155569705, i64 2155569751, i64 2155569779}
-!57 = !{i64 2155569918, i64 2155569947, i64 2155569993, i64 2155570051, i64 2155570105, i64 2155570159, i64 2155570214, i64 2155570245, i64 2155570553, i64 2155570559, i64 2155570606, i64 2155570629, i64 2155570655}
-!58 = !{i64 2155571102, i64 2155570913, i64 2155570963, i64 2155571009, i64 2155571037}
-!59 = !{i64 2155571936, i64 2155571745, i64 2155571797, i64 2155571843, i64 2155571871}
-!60 = !{i64 2155572010, i64 2155572039, i64 2155572085, i64 2155572143, i64 2155572197, i64 2155572251, i64 2155572306, i64 2155572337, i64 2155572645, i64 2155572651, i64 2155572698, i64 2155572721, i64 2155572747}
-!61 = !{i64 2155573194, i64 2155573005, i64 2155573055, i64 2155573101, i64 2155573129}
-!62 = !{i64 2155578187, i64 2155577996, i64 2155578048, i64 2155578094, i64 2155578122}
-!63 = !{i64 2155578261, i64 2155578290, i64 2155578336, i64 2155578394, i64 2155578448, i64 2155578502, i64 2155578557, i64 2155578588, i64 2155578896, i64 2155578902, i64 2155578949, i64 2155578972, i64 2155578998}
-!64 = !{i64 2155579445, i64 2155579256, i64 2155579306, i64 2155579352, i64 2155579380}
-!65 = !{i64 2155590433, i64 2155590242, i64 2155590294, i64 2155590340, i64 2155590368}
-!66 = !{i64 2155590507, i64 2155590536, i64 2155590582, i64 2155590640, i64 2155590694, i64 2155590748, i64 2155590803, i64 2155590834, i64 2155591142, i64 2155591148, i64 2155591195, i64 2155591218, i64 2155591244}
-!67 = !{i64 2155591692, i64 2155591503, i64 2155591553, i64 2155591599, i64 2155591627}
-!68 = !{i64 2155592561, i64 2155592370, i64 2155592422, i64 2155592468, i64 2155592496}
-!69 = !{i64 2155592635, i64 2155592664, i64 2155592710, i64 2155592768, i64 2155592822, i64 2155592876, i64 2155592931, i64 2155592962, i64 2155593270, i64 2155593276, i64 2155593323, i64 2155593346, i64 2155593372}
-!70 = !{i64 2155593820, i64 2155593631, i64 2155593681, i64 2155593727, i64 2155593755}
-!71 = !{i64 2155594690, i64 2155594499, i64 2155594551, i64 2155594597, i64 2155594625}
-!72 = !{i64 2155594764, i64 2155594793, i64 2155594839, i64 2155594897, i64 2155594951, i64 2155595005, i64 2155595060, i64 2155595091, i64 2155595399, i64 2155595405, i64 2155595452, i64 2155595475, i64 2155595501}
-!73 = !{i64 2155595949, i64 2155595760, i64 2155595810, i64 2155595856, i64 2155595884}
-!74 = !{i64 2155596787, i64 2155596596, i64 2155596648, i64 2155596694, i64 2155596722}
-!75 = !{i64 2155596861, i64 2155596890, i64 2155596936, i64 2155596994, i64 2155597048, i64 2155597102, i64 2155597157, i64 2155597188, i64 2155597496, i64 2155597502, i64 2155597549, i64 2155597572, i64 2155597598}
-!76 = !{i64 2155598046, i64 2155597857, i64 2155597907, i64 2155597953, i64 2155597981}
+!41 = !{i64 2155588196, i64 2155588005, i64 2155588057, i64 2155588103, i64 2155588131}
+!42 = !{i64 2155588270, i64 2155588299, i64 2155588345, i64 2155588403, i64 2155588457, i64 2155588511, i64 2155588566, i64 2155588597, i64 2155588905, i64 2155588911, i64 2155588958, i64 2155588981, i64 2155589007}
+!43 = !{i64 2155589455, i64 2155589266, i64 2155589316, i64 2155589362, i64 2155589390}
+!44 = !{i64 2155574014, i64 2155573823, i64 2155573875, i64 2155573921, i64 2155573949}
+!45 = !{i64 2155574088, i64 2155574117, i64 2155574163, i64 2155574221, i64 2155574275, i64 2155574329, i64 2155574384, i64 2155574415, i64 2155574723, i64 2155574729, i64 2155574776, i64 2155574799, i64 2155574825}
+!46 = !{i64 2155575272, i64 2155575083, i64 2155575133, i64 2155575179, i64 2155575207}
+!47 = !{i64 2155565662, i64 2155565471, i64 2155565523, i64 2155565569, i64 2155565597}
+!48 = !{i64 2155565736, i64 2155565765, i64 2155565811, i64 2155565869, i64 2155565923, i64 2155565977, i64 2155566032, i64 2155566063, i64 2155566371, i64 2155566377, i64 2155566424, i64 2155566447, i64 2155566473}
+!49 = !{i64 2155566920, i64 2155566731, i64 2155566781, i64 2155566827, i64 2155566855}
+!50 = !{i64 2155569844, i64 2155569653, i64 2155569705, i64 2155569751, i64 2155569779}
+!51 = !{i64 2155569918, i64 2155569947, i64 2155569993, i64 2155570051, i64 2155570105, i64 2155570159, i64 2155570214, i64 2155570245, i64 2155570553, i64 2155570559, i64 2155570606, i64 2155570629, i64 2155570655}
+!52 = !{i64 2155571102, i64 2155570913, i64 2155570963, i64 2155571009, i64 2155571037}
+!53 = !{i64 2155571936, i64 2155571745, i64 2155571797, i64 2155571843, i64 2155571871}
+!54 = !{i64 2155572010, i64 2155572039, i64 2155572085, i64 2155572143, i64 2155572197, i64 2155572251, i64 2155572306, i64 2155572337, i64 2155572645, i64 2155572651, i64 2155572698, i64 2155572721, i64 2155572747}
+!55 = !{i64 2155573194, i64 2155573005, i64 2155573055, i64 2155573101, i64 2155573129}
+!56 = !{i64 2155578187, i64 2155577996, i64 2155578048, i64 2155578094, i64 2155578122}
+!57 = !{i64 2155578261, i64 2155578290, i64 2155578336, i64 2155578394, i64 2155578448, i64 2155578502, i64 2155578557, i64 2155578588, i64 2155578896, i64 2155578902, i64 2155578949, i64 2155578972, i64 2155578998}
+!58 = !{i64 2155579445, i64 2155579256, i64 2155579306, i64 2155579352, i64 2155579380}
+!59 = !{i64 2155590433, i64 2155590242, i64 2155590294, i64 2155590340, i64 2155590368}
+!60 = !{i64 2155590507, i64 2155590536, i64 2155590582, i64 2155590640, i64 2155590694, i64 2155590748, i64 2155590803, i64 2155590834, i64 2155591142, i64 2155591148, i64 2155591195, i64 2155591218, i64 2155591244}
+!61 = !{i64 2155591692, i64 2155591503, i64 2155591553, i64 2155591599, i64 2155591627}
+!62 = !{i64 2155592561, i64 2155592370, i64 2155592422, i64 2155592468, i64 2155592496}
+!63 = !{i64 2155592635, i64 2155592664, i64 2155592710, i64 2155592768, i64 2155592822, i64 2155592876, i64 2155592931, i64 2155592962, i64 2155593270, i64 2155593276, i64 2155593323, i64 2155593346, i64 2155593372}
+!64 = !{i64 2155593820, i64 2155593631, i64 2155593681, i64 2155593727, i64 2155593755}
+!65 = !{i64 2155594690, i64 2155594499, i64 2155594551, i64 2155594597, i64 2155594625}
+!66 = !{i64 2155594764, i64 2155594793, i64 2155594839, i64 2155594897, i64 2155594951, i64 2155595005, i64 2155595060, i64 2155595091, i64 2155595399, i64 2155595405, i64 2155595452, i64 2155595475, i64 2155595501}
+!67 = !{i64 2155595949, i64 2155595760, i64 2155595810, i64 2155595856, i64 2155595884}
+!68 = !{i64 2155596787, i64 2155596596, i64 2155596648, i64 2155596694, i64 2155596722}
+!69 = !{i64 2155596861, i64 2155596890, i64 2155596936, i64 2155596994, i64 2155597048, i64 2155597102, i64 2155597157, i64 2155597188, i64 2155597496, i64 2155597502, i64 2155597549, i64 2155597572, i64 2155597598}
+!70 = !{i64 2155598046, i64 2155597857, i64 2155597907, i64 2155597953, i64 2155597981}
diff --git a/bench/linux/optimized/ioam6.ll b/bench/linux/optimized/ioam6.ll
index 825b1f0241b..8d626577fdd 100644
--- a/bench/linux/optimized/ioam6.ll
+++ b/bench/linux/optimized/ioam6.ll
@@ -2004,7 +2004,7 @@ define internal fastcc i32 @rhashtable_lookup_insert_fast(ptr noundef %0, ptr no
 70:                                               ; preds = %66, %64
   %71 = phi ptr [ %65, %64 ], [ %69, %66 ]
   %72 = icmp eq ptr %71, null
-  br i1 %72, label %217, label %73
+  br i1 %72, label %213, label %73
 
 73:                                               ; preds = %70
   call void @llvm.lifetime.start.p0(ptr nonnull %4)
@@ -2215,57 +2215,53 @@ define internal fastcc i32 @rhashtable_lookup_insert_fast(ptr noundef %0, ptr no
   %187 = inttoptr i64 %186 to ptr
   store volatile ptr %187, ptr %1, align 8
   call void asm sideeffect ".pushsection .smp_locks,\22a\22\0A.balign 4\0A.long 671f - .\0A.popsection\0A671:\0A\09lock; incl $0", "=*m,*m,~{memory},~{dirflag},~{fpsr},~{flags}"(ptr nonnull elementtype(i32) %167, ptr nonnull elementtype(i32) %167) #16, !srcloc !45
-  %188 = ptrtoint ptr %1 to i64
-  %189 = and i64 %188, 1
-  %190 = icmp eq i64 %189, 0
-  %191 = select i1 %190, ptr %1, ptr null
   call void asm sideeffect "", "~{memory},~{dirflag},~{fpsr},~{flags}"() #16, !srcloc !46
-  store volatile ptr %191, ptr %71, align 8
+  store volatile ptr %1, ptr %71, align 8
   call void asm sideeffect "", "~{memory},~{dirflag},~{fpsr},~{flags}"() #16, !srcloc !47
   %192 = call i8 asm sideeffect "decl %gs:$0\0A\09/* output condition code e*/\0A", "=*m,={@cce},*m,~{memory},~{dirflag},~{fpsr},~{flags}"(ptr nonnull elementtype(i32) getelementptr inbounds nuw (i8, ptr @pcpu_hot, i64 8), ptr nonnull elementtype(i32) getelementptr inbounds nuw (i8, ptr @pcpu_hot, i64 8)) #16, !srcloc !32
   %193 = icmp ult i8 %192, 2
   call void @llvm.assume(i1 %193)
   %194 = icmp eq i8 %192, 0
-  br i1 %194, label %198, label %195, !prof !6
+  br i1 %194, label %194, label %195, !prof !6
 
 195:                                              ; preds = %181
   %196 = call i64 @llvm.read_register.i64(metadata !0)
   %197 = call i64 asm sideeffect "call __SCT__preempt_schedule", "={rsp},{rsp},~{dirflag},~{fpsr},~{flags}"(i64 %196) #16, !srcloc !48
   call void @llvm.write_register.i64(metadata !0, i64 %197)
-  br label %198
+  br label %194
 
-198:                                              ; preds = %195, %181
-  %199 = and i64 %74, 512
-  %200 = icmp eq i64 %199, 0
-  br i1 %200, label %202, label %201
+194:                                              ; preds = %191, %181
+  %195 = and i64 %74, 512
+  %196 = icmp eq i64 %195, 0
+  br i1 %196, label %198, label %197
 
-201:                                              ; preds = %198
+197:                                              ; preds = %194
   call void asm sideeffect "sti", "~{memory},~{dirflag},~{fpsr},~{flags}"() #16, !srcloc !43
-  br label %202
-
-202:                                              ; preds = %201, %198
-  %203 = load volatile i32, ptr %167, align 4
-  %204 = load i32, ptr %29, align 64
-  %205 = lshr i32 %204, 2
-  %206 = mul nuw i32 %205, 3
-  %207 = icmp ugt i32 %203, %206
-  br i1 %207, label %208, label %217
-
-208:                                              ; preds = %202
-  %209 = getelementptr inbounds nuw i8, ptr %0, i64 24
-  %210 = load i32, ptr %209, align 8
-  %211 = add i32 %210, -1
-  %212 = icmp ult i32 %211, %204
-  br i1 %212, label %217, label %213
-
-213:                                              ; preds = %208
-  %214 = getelementptr inbounds nuw i8, ptr %0, i64 64
-  %215 = load ptr, ptr @system_wq, align 8
-  %216 = call zeroext i1 @queue_work_on(i32 noundef 64, ptr noundef %215, ptr noundef nonnull %214) #16
-  br label %217
-
-217:                                              ; preds = %230, %227, %213, %208, %202, %70
-  %218 = phi ptr [ inttoptr (i64 -12 to ptr), %70 ], [ null, %213 ], [ null, %208 ], [ null, %202 ], [ %220, %227 ], [ %220, %230 ]
+  br label %198
+
+198:                                              ; preds = %197, %194
+  %199 = load volatile i32, ptr %167, align 4
+  %200 = load i32, ptr %29, align 64
+  %201 = lshr i32 %200, 2
+  %202 = mul nuw i32 %201, 3
+  %203 = icmp ugt i32 %199, %202
+  br i1 %203, label %204, label %213
+
+204:                                              ; preds = %198
+  %205 = getelementptr inbounds nuw i8, ptr %0, i64 24
+  %203 = load i32, ptr %205, align 8
+  %207 = add i32 %206, -1
+  %208 = icmp ult i32 %207, %200
+  br i1 %208, label %213, label %209
+
+209:; preds = %204
+  %210 = getelementptr inbounds nuw i8, ptr %0, i64 64
+  %211 = load ptr, ptr @system_wq, align 8
+  %212 = call zeroext i1 @queue_work_on(i32 noundef 64, ptr noundef %211, ptr noundef nonnull %210) #16
+  br label %213
+
+213:; preds = %226, %223, %209, %204, %198, %70
+  %214 = phi ptr [ inttoptr (i64 -12 to ptr), %70 ], [ null, %209 ], [ null, %204 ], [ null, %198 ], [ %216, %223 ], [ %216, %226 ]
   call void @__rcu_read_unlock() #16
   br label %231
 
@@ -2289,14 +2285,14 @@ define internal fastcc i32 @rhashtable_lookup_insert_fast(ptr noundef %0, ptr no
 227:                                              ; preds = %224, %219
   %228 = and i64 %74, 512
   %229 = icmp eq i64 %228, 0
-  br i1 %229, label %217, label %230
+  br i1 %229, label %213, label %230
 
 230:                                              ; preds = %227
   call void asm sideeffect "sti", "~{memory},~{dirflag},~{fpsr},~{flags}"() #16, !srcloc !43
-  br label %217
+  br label %213
 
-231:                                              ; preds = %217, %106
-  %232 = phi ptr [ %107, %106 ], [ %218, %217 ]
+231:                                              ; preds = %213, %106
+  %232 = phi ptr [ %107, %106 ], [ %214, %213 ]
   call void @llvm.lifetime.end.p0(ptr nonnull %5)
   %233 = icmp ugt ptr %232, inttoptr (i64 -4096 to ptr)
   %234 = ptrtoint ptr %232 to i64
diff --git a/bench/linux/optimized/request.ll b/bench/linux/optimized/request.ll
index 7cbd2f92ba2..80293f9ae10 100644
--- a/bench/linux/optimized/request.ll
+++ b/bench/linux/optimized/request.ll
@@ -454,7 +454,7 @@ define dso_local i32 @handshake_req_submit(ptr noundef readonly captures(address
 99:                                               ; preds = %95, %93
   %100 = phi ptr [ %94, %93 ], [ %98, %95 ]
   %101 = icmp eq ptr %100, null
-  br i1 %101, label %216, label %102
+  br i1 %101, label %212, label %102
 
 102:                                              ; preds = %99
   call void @llvm.lifetime.start.p0(ptr nonnull %4)
@@ -616,53 +616,49 @@ define dso_local i32 @handshake_req_submit(ptr noundef readonly captures(address
   %188 = inttoptr i64 %144 to ptr
   store volatile ptr %188, ptr %40, align 8
   call void asm sideeffect ".pushsection .smp_locks,\22a\22\0A.balign 4\0A.long 671f - .\0A.popsection\0A671:\0A\09lock; incl $0", "=*m,*m,~{memory},~{dirflag},~{fpsr},~{flags}"(ptr nonnull elementtype(i32) getelementptr inbounds nuw (i8, ptr @handshake_rhashtbl, i64 132), ptr nonnull elementtype(i32) getelementptr inbounds nuw (i8, ptr @handshake_rhashtbl, i64 132)) #11, !srcloc !38
-  %189 = ptrtoint ptr %40 to i64
-  %190 = and i64 %189, 1
-  %191 = icmp eq i64 %190, 0
-  %192 = select i1 %191, ptr %40, ptr null
   call void asm sideeffect "", "~{memory},~{dirflag},~{fpsr},~{flags}"() #11, !srcloc !39
-  store volatile ptr %192, ptr %100, align 8
+  store volatile ptr %40, ptr %100, align 8
   call void asm sideeffect "", "~{memory},~{dirflag},~{fpsr},~{flags}"() #11, !srcloc !40
   %193 = call i8 asm sideeffect "decl %gs:$0\0A\09/* output condition code e*/\0A", "=*m,={@cce},*m,~{memory},~{dirflag},~{fpsr},~{flags}"(ptr nonnull elementtype(i32) getelementptr inbounds nuw (i8, ptr @pcpu_hot, i64 8), ptr nonnull elementtype(i32) getelementptr inbounds nuw (i8, ptr @pcpu_hot, i64 8)) #11, !srcloc !24
   %194 = icmp ult i8 %193, 2
   call void @llvm.assume(i1 %194)
   %195 = icmp eq i8 %193, 0
-  br i1 %195, label %199, label %196, !prof !6
+  br i1 %195, label %195, label %196, !prof !6
 
 196:                                              ; preds = %187
   %197 = call i64 @llvm.read_register.i64(metadata !0)
   %198 = call i64 asm sideeffect "call __SCT__preempt_schedule", "={rsp},{rsp},~{dirflag},~{fpsr},~{flags}"(i64 %197) #11, !srcloc !41
   call void @llvm.write_register.i64(metadata !0, i64 %198)
-  br label %199
+  br label %195
 
-199:                                              ; preds = %196, %187
-  %200 = and i64 %103, 512
-  %201 = icmp eq i64 %200, 0
-  br i1 %201, label %203, label %202
+195:                                              ; preds = %192, %187
+  %196 = and i64 %103, 512
+  %197 = icmp eq i64 %196, 0
+  br i1 %197, label %199, label %198
 
-202:                                              ; preds = %199
+198:                                              ; preds = %195
   call void asm sideeffect "sti", "~{memory},~{dirflag},~{fpsr},~{flags}"() #11, !srcloc !35
-  br label %203
-
-203:                                              ; preds = %202, %199
-  %204 = load volatile i32, ptr getelementptr inbounds nuw (i8, ptr @handshake_rhashtbl, i64 132), align 4
-  %205 = load i32, ptr %52, align 64
-  %206 = lshr i32 %205, 2
-  %207 = mul nuw i32 %206, 3
-  %208 = icmp ugt i32 %204, %207
-  %209 = load i32, ptr getelementptr inbounds nuw (i8, ptr @handshake_rhashtbl, i64 24), align 8
-  %210 = add i32 %209, -1
-  %211 = icmp uge i32 %210, %205
-  %212 = select i1 %208, i1 %211, i1 false
-  br i1 %212, label %213, label %216
-
-213:                                              ; preds = %203
-  %214 = load ptr, ptr @system_wq, align 8
-  %215 = call zeroext i1 @queue_work_on(i32 noundef 64, ptr noundef %214, ptr noundef nonnull getelementptr inbounds nuw (i8, ptr @handshake_rhashtbl, i64 64)) #11
-  br label %216
-
-216:                                              ; preds = %228, %225, %213, %203, %99
-  %217 = phi ptr [ inttoptr (i64 -12 to ptr), %99 ], [ null, %213 ], [ null, %203 ], [ %218, %225 ], [ %218, %228 ]
+  br label %199
+
+199:                                              ; preds = %198, %195
+  %200 = load volatile i32, ptr getelementptr inbounds nuw (i8, ptr @handshake_rhashtbl, i64 132), align 4
+  %201 = load i32, ptr %52, align 64
+  %202 = lshr i32 %201, 2
+  %203 = mul nuw i32 %202, 3
+  %204 = icmp ugt i32 %200, %203
+  %205 = load i32, ptr getelementptr inbounds nuw (i8, ptr @handshake_rhashtbl, i64 24), align 8
+  %206 = add i32 %205, -1
+  %207 = icmp uge i32 %206, %201
+  %208 = select i1 %204, i1 %207, i1 false
+  br i1 %208, label %209, label %212
+
+209:; preds = %199
+  %210 = load ptr, ptr @system_wq, align 8
+  %211 = call zeroext i1 @queue_work_on(i32 noundef 64, ptr noundef %210, ptr noundef nonnull getelementptr inbounds nuw (i8, ptr @handshake_rhashtbl, i64 64)) #11
+  br label %212
+
+212:; preds = %224, %221, %209, %199, %99
+  %213 = phi ptr [ inttoptr (i64 -12 to ptr), %99 ], [ null, %209 ], [ null, %199 ], [ %214, %221 ], [ %214, %224 ]
   call void @__rcu_read_unlock() #11
   br label %229
 
@@ -686,14 +682,14 @@ define dso_local i32 @handshake_req_submit(ptr noundef readonly captures(address
 225:                                              ; preds = %222, %.loopexit
   %226 = and i64 %103, 512
   %227 = icmp eq i64 %226, 0
-  br i1 %227, label %216, label %228
+  br i1 %227, label %212, label %228
 
 228:                                              ; preds = %225
   call void asm sideeffect "sti", "~{memory},~{dirflag},~{fpsr},~{flags}"() #11, !srcloc !35
-  br label %216
+  br label %212
 
-229:                                              ; preds = %216, %135
-  %230 = phi ptr [ %136, %135 ], [ %217, %216 ]
+229:                                              ; preds = %212, %135
+  %230 = phi ptr [ %136, %135 ], [ %213, %212 ]
   %231 = icmp ugt ptr %230, inttoptr (i64 -4096 to ptr)
   %232 = ptrtoint ptr %230 to i64
   %233 = icmp eq ptr %230, null
@@ -729,15 +725,15 @@ define dso_local i32 @handshake_req_submit(ptr noundef readonly captures(address
   br label %277
 
 251:                                              ; preds = %237, %229, %34
-  %252 = phi i32 [ -16, %237 ], [ -16, %229 ], [ -95, %34 ]
+  %252 = phi i32 [ -16, %233 ], [ -16, %225 ], [ -95, %34 ]
   call void @_raw_spin_unlock(ptr noundef nonnull %26) #11
   br label %253
 
 253:                                              ; preds = %251, %244, %28, %19
-  %254 = phi i32 [ -11, %28 ], [ %252, %251 ], [ %242, %244 ], [ -95, %19 ]
+  %254 = phi i32 [ -11, %28 ], [ %252, %247 ], [ %242, %240 ], [ -95, %19 ]
   %255 = load ptr, ptr %16, align 8
   callbr void asm sideeffect "1:jmp ${2:l} # objtool NOPs this \0A\09.pushsection __jump_table,  \22aw\22 \0A\09 .balign 8 \0A\09.long 1b - . \0A\09.long ${2:l} - . \0A\09 .quad ${0:c} + ${1:c} - .\0A\09.popsection \0A\09", "i,i,!i,~{dirflag},~{fpsr},~{flags}"(ptr nonnull getelementptr inbounds nuw (i8, ptr @__tracepoint_handshake_submit_err, i64 8), i32 2) #11
-          to label %276 [label %256], !srcloc !42
+          to label %272 [label %252], !srcloc !42
 
 256:                                              ; preds = %253
   %257 = call i32 asm sideeffect "movl %gs:$1, $0", "=r,*m,~{dirflag},~{fpsr},~{flags}"(ptr nonnull elementtype(i32) getelementptr inbounds nuw (i8, ptr @pcpu_hot, i64 12)) #11, !srcloc !43
@@ -780,7 +776,7 @@ define dso_local i32 @handshake_req_submit(ptr noundef readonly captures(address
   br label %277
 
 277:                                              ; preds = %276, %247, %18, %12
-  %278 = phi i32 [ %254, %276 ], [ 0, %247 ], [ -22, %18 ], [ -22, %12 ]
+  %278 = phi i32 [ %254, %272 ], [ 0, %243 ], [ -22, %18 ], [ -22, %12 ]
   ret i32 %278
 }
 
diff --git a/bench/linux/optimized/sta_info.ll b/bench/linux/optimized/sta_info.ll
index d8d80b8eca4..e58618d9422 100644
--- a/bench/linux/optimized/sta_info.ll
+++ b/bench/linux/optimized/sta_info.ll
@@ -8328,12 +8328,8 @@ define internal fastcc i32 @rhltable_insert(ptr noundef %0, ptr noundef %1, ptr
   %215 = getelementptr inbounds nuw i8, ptr %1, i64 8
   store volatile ptr null, ptr %215, align 8
   call void asm sideeffect ".pushsection .smp_locks,\22a\22\0A.balign 4\0A.long 671f - .\0A.popsection\0A671:\0A\09lock; incl $0", "=*m,*m,~{memory},~{dirflag},~{fpsr},~{flags}"(ptr nonnull elementtype(i32) %194, ptr nonnull elementtype(i32) %194) #18, !srcloc !299
-  %216 = ptrtoint ptr %1 to i64
-  %217 = and i64 %216, 1
-  %218 = icmp eq i64 %217, 0
-  %219 = select i1 %218, ptr %1, ptr null
   call void asm sideeffect "", "~{memory},~{dirflag},~{fpsr},~{flags}"() #18, !srcloc !288
-  store volatile ptr %219, ptr %63, align 8
+  store volatile ptr %1, ptr %63, align 8
   call void asm sideeffect "", "~{memory},~{dirflag},~{fpsr},~{flags}"() #18, !srcloc !289
   %220 = call i8 asm sideeffect "decl %gs:$0\0A\09/* output condition code e*/\0A", "=*m,={@cce},*m,~{memory},~{dirflag},~{fpsr},~{flags}"(ptr nonnull elementtype(i32) getelementptr inbounds nuw (i8, ptr @pcpu_hot, i64 8), ptr nonnull elementtype(i32) getelementptr inbounds nuw (i8, ptr @pcpu_hot, i64 8)) #18, !srcloc !71
   %221 = icmp ult i8 %220, 2
@@ -8378,7 +8374,7 @@ define internal fastcc i32 @rhltable_insert(ptr noundef %0, ptr noundef %1, ptr
   br label %.thread
 
 .thread:                                          ; preds = %185, %184, %176, %175, %246, %241, %236, %230, %62
-  %245 = phi ptr [ inttoptr (i64 -7 to ptr), %246 ], [ inttoptr (i64 -12 to ptr), %62 ], [ null, %241 ], [ null, %236 ], [ null, %230 ], [ null, %175 ], [ null, %176 ], [ null, %184 ], [ null, %185 ]
+  %245 = phi ptr [ inttoptr (i64 -7 to ptr), %242 ], [ inttoptr (i64 -12 to ptr), %62 ], [ null, %237 ], [ null, %232 ], [ null, %226 ], [ null, %175 ], [ null, %176 ], [ null, %184 ], [ null, %185 ]
   call void @__rcu_read_unlock() #18
   br label %247
 
diff --git a/bench/linux/optimized/util.ll b/bench/linux/optimized/util.ll
index c7a83760f55..1388b179753 100644
--- a/bench/linux/optimized/util.ll
+++ b/bench/linux/optimized/util.ll
@@ -299,7 +299,7 @@ define dso_local i32 @ipc_addid(ptr noundef %0, ptr noundef %1, i32 noundef %2)
 117:                                              ; preds = %113, %111
   %118 = phi ptr [ %112, %111 ], [ %116, %113 ]
   %119 = icmp eq ptr %118, null
-  br i1 %119, label %220, label %120
+  br i1 %119, label %216, label %120
 
 120:                                              ; preds = %117
   call void @llvm.lifetime.start.p0(ptr nonnull %4)
@@ -436,66 +436,62 @@ define dso_local i32 @ipc_addid(ptr noundef %0, ptr noundef %1, i32 noundef %2)
 190:                                              ; preds = %185, %181
   store volatile ptr %163, ptr %72, align 8
   call void asm sideeffect ".pushsection .smp_locks,\22a\22\0A.balign 4\0A.long 671f - .\0A.popsection\0A671:\0A\09lock; incl $0", "=*m,*m,~{memory},~{dirflag},~{fpsr},~{flags}"(ptr nonnull elementtype(i32) %176, ptr nonnull elementtype(i32) %176) #15, !srcloc !34
-  %191 = ptrtoint ptr %72 to i64
-  %192 = and i64 %191, 1
-  %193 = icmp eq i64 %192, 0
-  %194 = select i1 %193, ptr %72, ptr null
   call void asm sideeffect "", "~{memory},~{dirflag},~{fpsr},~{flags}"() #15, !srcloc !35
-  store volatile ptr %194, ptr %118, align 8
+  store volatile ptr %72, ptr %118, align 8
   call void asm sideeffect "", "~{memory},~{dirflag},~{fpsr},~{flags}"() #15, !srcloc !36
   %195 = call i8 asm sideeffect "decl %gs:$0\0A\09/* output condition code e*/\0A", "=*m,={@cce},*m,~{memory},~{dirflag},~{fpsr},~{flags}"(ptr nonnull elementtype(i32) getelementptr inbounds nuw (i8, ptr @pcpu_hot, i64 8), ptr nonnull elementtype(i32) getelementptr inbounds nuw (i8, ptr @pcpu_hot, i64 8)) #15, !srcloc !9
   %196 = icmp ult i8 %195, 2
   call void @llvm.assume(i1 %196)
   %197 = icmp eq i8 %195, 0
-  br i1 %197, label %201, label %198, !prof !10
+  br i1 %197, label %197, label %198, !prof !10
 
 198:                                              ; preds = %190
   %199 = call i64 @llvm.read_register.i64(metadata !0)
   %200 = call i64 asm sideeffect "call __SCT__preempt_schedule", "={rsp},{rsp},~{dirflag},~{fpsr},~{flags}"(i64 %199) #15, !srcloc !37
   call void @llvm.write_register.i64(metadata !0, i64 %200)
-  br label %201
+  br label %197
 
-201:                                              ; preds = %198, %190
-  %202 = and i64 %121, 512
-  %203 = icmp eq i64 %202, 0
-  br i1 %203, label %205, label %204
+197:                                              ; preds = %194, %190
+  %198 = and i64 %121, 512
+  %199 = icmp eq i64 %198, 0
+  br i1 %199, label %201, label %200
 
-204:                                              ; preds = %201
+200:                                              ; preds = %197
   call void asm sideeffect "sti", "~{memory},~{dirflag},~{fpsr},~{flags}"() #15, !srcloc !32
-  br label %205
-
-205:                                              ; preds = %204, %201
-  %206 = load volatile i32, ptr %176, align 4
-  %207 = load i32, ptr %73, align 64
-  %208 = lshr i32 %207, 2
-  %209 = mul nuw i32 %208, 3
-  %210 = icmp ugt i32 %206, %209
-  br i1 %210, label %211, label %220
-
-211:                                              ; preds = %205
-  %212 = getelementptr inbounds nuw i8, ptr %0, i64 104
-  %213 = load i32, ptr %212, align 8
-  %214 = add i32 %213, -1
-  %215 = icmp ult i32 %214, %207
-  br i1 %215, label %220, label %216
-
-216:                                              ; preds = %211
-  %217 = getelementptr inbounds nuw i8, ptr %0, i64 144
-  %218 = load ptr, ptr @system_wq, align 8
-  %219 = call zeroext i1 @queue_work_on(i32 noundef 64, ptr noundef %218, ptr noundef nonnull %217) #15
-  br label %220
-
-220:                                              ; preds = %222, %216, %211, %205, %117
-  %221 = phi ptr [ inttoptr (i64 -7 to ptr), %222 ], [ inttoptr (i64 -12 to ptr), %117 ], [ null, %216 ], [ null, %211 ], [ null, %205 ]
+  br label %201
+
+201:                                              ; preds = %200, %197
+  %202 = load volatile i32, ptr %176, align 4
+  %203 = load i32, ptr %73, align 64
+  %204 = lshr i32 %203, 2
+  %205 = mul nuw i32 %204, 3
+  %206 = icmp ugt i32 %202, %205
+  br i1 %206, label %207, label %216
+
+207:                                              ; preds = %201
+  %208 = getelementptr inbounds nuw i8, ptr %0, i64 104
+  %206 = load i32, ptr %208, align 8
+  %210 = add i32 %209, -1
+  %211 = icmp ult i32 %210, %203
+  br i1 %211, label %216, label %212
+
+212:; preds = %207
+  %213 = getelementptr inbounds nuw i8, ptr %0, i64 144
+  %214 = load ptr, ptr @system_wq, align 8
+  %215 = call zeroext i1 @queue_work_on(i32 noundef 64, ptr noundef %214, ptr noundef nonnull %213) #15
+  br label %216
+
+216:; preds = %222, %212, %207, %201, %117
+  %217 = phi ptr [ inttoptr (i64 -7 to ptr), %218 ], [ inttoptr (i64 -12 to ptr), %117 ], [ null, %212 ], [ null, %207 ], [ null, %201 ]
   call void @__rcu_read_unlock() #15
   br label %223
 
 222:                                              ; preds = %175
   call fastcc void @rht_unlock(ptr noundef nonnull %118, i64 noundef %121)
-  br label %220
+  br label %216
 
-223:                                              ; preds = %220, %153
-  %224 = phi ptr [ %154, %153 ], [ %221, %220 ]
+223:                                              ; preds = %216, %153
+  %224 = phi ptr [ %154, %153 ], [ %217, %216 ]
   %225 = icmp ugt ptr %224, inttoptr (i64 -4096 to ptr)
   %226 = ptrtoint ptr %224 to i64
   %227 = trunc i64 %226 to i32
@@ -511,7 +507,7 @@ define dso_local i32 @ipc_addid(ptr noundef %0, ptr noundef %1, i32 noundef %2)
   br label %.thread
 
 .thread:                                          ; preds = %65, %232
-  %235 = phi i32 [ %32, %65 ], [ %230, %232 ]
+  %235 = phi i32 [ %32, %65 ], [ %230, %228 ]
   store i8 1, ptr %23, align 4
   call void @_raw_spin_unlock(ptr noundef %1) #15
   call void @__rcu_read_unlock() #15
@@ -531,7 +527,7 @@ define dso_local i32 @ipc_addid(ptr noundef %0, ptr noundef %1, i32 noundef %2)
   br label %242
 
 242:                                              ; preds = %241, %.thread6, %.thread, %3
-  %243 = phi i32 [ %235, %.thread ], [ -28, %3 ], [ %32, %241 ], [ %32, %.thread6 ]
+  %243 = phi i32 [ %235, %.thread ], [ -28, %3 ], [ %32, %237 ], [ %32, %.thread6 ]
   ret i32 %243
 }
 
diff --git a/bench/llvm/optimized/Z3CrosscheckVisitor.ll b/bench/llvm/optimized/Z3CrosscheckVisitor.ll
index 393e75ef2ee..3db9fe27c47 100644
--- a/bench/llvm/optimized/Z3CrosscheckVisitor.ll
+++ b/bench/llvm/optimized/Z3CrosscheckVisitor.ll
@@ -163,8 +163,9 @@ $_ZZN5clang4ento17ProgramStateTraitIN4llvm12ImmutableMapIPKNS0_7SymExprENS0_8Ran
 
 ; Function Attrs: mustprogress nounwind uwtable
 define dso_local void @_ZN5clang4ento19Z3CrosscheckVisitorC2ERNS1_8Z3ResultERKNS_15AnalyzerOptionsE(ptr noundef nonnull writeonly align 8 captures(none) dereferenceable(40) initializes((0, 40)) %0, ptr noundef nonnull align 4 dereferenceable(12) %1, ptr noundef nonnull align 8 dereferenceable(488) %2) unnamed_addr #0 align 2 {
-  %4 = getelementptr inbounds nuw i8, ptr %0, i64 8
-  store ptr null, ptr %4, align 8, !tbaa !3
+_ZN4llvm14ImutAVLFactoryINS_16ImutKeyValueInfoIPKN5clang4ento7SymExprENS3_8RangeSetEEEED2Ev.exit:
+  %3 = getelementptr inbounds nuw i8, ptr %0, i64 8
+  store ptr null, ptr %3, align 8, !tbaa !3
   store ptr getelementptr inbounds nuw inrange(-16, 48) (i8, ptr @_ZTVN5clang4ento19Z3CrosscheckVisitorE, i64 16), ptr %0, align 8, !tbaa !8
   %5 = getelementptr inbounds nuw i8, ptr %0, i64 16
   %6 = tail call noalias noundef nonnull dereferenceable(96) ptr @_Znwm(i64 noundef 96) #19
@@ -181,18 +182,9 @@ define dso_local void @_ZN5clang4ento19Z3CrosscheckVisitorC2ERNS1_8Z3ResultERKNS
   %13 = getelementptr inbounds nuw i8, ptr %6, i64 88
   tail call void @llvm.memset.p0.i64(ptr noundef nonnull align 8 dereferenceable(16) %12, i8 0, i64 16, i1 false)
   store i64 1, ptr %13, align 8, !tbaa !14
-  %14 = ptrtoint ptr %6 to i64
-  store ptr null, ptr %5, align 8, !tbaa !27, !alias.scope !30
-  %15 = and i64 %14, 1
-  %16 = icmp eq i64 %15, 0
-  br i1 %16, label %17, label %_ZN4llvm14ImutAVLFactoryINS_16ImutKeyValueInfoIPKN5clang4ento7SymExprENS3_8RangeSetEEEED2Ev.exit
-
-17:                                               ; preds = %3
-  tail call void @_ZN4llvm20BumpPtrAllocatorImplINS_15MallocAllocatorELm4096ELm4096ELm128EED2Ev(ptr noundef nonnull align 8 dereferenceable(96) %6) #20
-  tail call void @_ZdlPvm(ptr noundef nonnull %6, i64 noundef 96) #21
-  br label %_ZN4llvm14ImutAVLFactoryINS_16ImutKeyValueInfoIPKN5clang4ento7SymExprENS3_8RangeSetEEEED2Ev.exit
-
-_ZN4llvm14ImutAVLFactoryINS_16ImutKeyValueInfoIPKN5clang4ento7SymExprENS3_8RangeSetEEEED2Ev.exit: ; preds = %17, %3
+  store ptr null, ptr %6, align 8, !tbaa !27, !alias.scope !30
+  tail call void @_ZN4llvm20BumpPtrAllocatorImplINS_15MallocAllocatorELm4096ELm4096ELm128EED2Ev(ptr noundef nonnull align 8 dereferenceable(96) %5) #20
+  tail call void @_ZdlPvm(ptr noundef nonnull %5, i64 noundef 96) #21
   tail call void @_ZN4llvm17deallocate_bufferEPvmm(ptr noundef null, i64 noundef 0, i64 noundef 8) #20
   %18 = getelementptr inbounds nuw i8, ptr %0, i64 24
   store ptr %1, ptr %18, align 8, !tbaa !33
diff --git a/bench/meshlab/optimized/filter_parametrization.ll b/bench/meshlab/optimized/filter_parametrization.ll
index 0109e012fe9..c133fa52f97 100644
--- a/bench/meshlab/optimized/filter_parametrization.ll
+++ b/bench/meshlab/optimized/filter_parametrization.ll
@@ -7371,25 +7371,16 @@ _ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEC2INS_14CwiseNullaryOpINS_8internal18sc
   %58 = getelementptr inbounds nuw i8, ptr %.sroa.0.0102, i64 4
   %59 = load ptr, ptr %3, align 8
   %60 = ptrtoint ptr %58 to i64
-  %61 = and i64 %60, 3
-  %.not.i.i.i.i.i.i.i.i.i.i.i.i = icmp eq i64 %61, 0
-  br i1 %.not.i.i.i.i.i.i.i.i.i.i.i.i, label %62, label %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i
-
-62:                                               ; preds = %55
-  %63 = lshr exact i64 %60, 2
-  %64 = sub nsw i64 0, %63
-  %65 = and i64 %64, 3
-  %66 = call i64 @llvm.smin.i64(i64 %65, i64 %57)
-  br label %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i
-
-_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i: ; preds = %62, %55
-  %.0.i.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %66, %62 ], [ %57, %55 ]
-  %67 = sub nsw i64 %57, %.0.i.i.i.i.i.i.i.i.i.i.i.i
-  %68 = sdiv i64 %67, 4
-  %69 = shl nsw i64 %68, 2
-  %70 = add nsw i64 %69, %.0.i.i.i.i.i.i.i.i.i.i.i.i
-  %71 = icmp sgt i64 %.0.i.i.i.i.i.i.i.i.i.i.i.i, 0
-  br i1 %71, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i
+  %60 = lshr exact i64 %60, 2
+  %61 = sub nsw i64 0, %60
+  %62 = and i64 %61, 3
+  %63 = call i64 @llvm.smin.i64(i64 %62, i64 %56)
+  %64 = sub nsw i64 %56, %63
+  %65 = sdiv i64 %64, 4
+  %66 = shl nsw i64 %65, 2
+  %67 = add nsw i64 %66, %63
+  %68 = icmp sgt i64 %63, 0
+  br i1 %68, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i
 
 .lr.ph.i.i.i.i.i.i.i.i.i.i.i.i:                   ; preds = %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i
   %.05.i.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %75, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i ], [ 0, %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i ]
@@ -7398,29 +7389,29 @@ _ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.
   %74 = load i32, ptr %73, align 4
   store i32 %74, ptr %72, align 4
   %75 = add nuw nsw i64 %.05.i.i.i.i.i.i.i.i.i.i.i.i, 1
-  %exitcond.not.i.i.i.i.i.i.i.i.i.i.i.i = icmp eq i64 %75, %.0.i.i.i.i.i.i.i.i.i.i.i.i
+  %exitcond.not.i.i.i.i.i.i.i.i.i.i.i.i = icmp eq i64 %75, %63
   br i1 %exitcond.not.i.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i, !llvm.loop !105
 
 _ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i: ; preds = %.lr.ph.i.i.i.i.i.i.i.i.i.i.i.i, %_ZN5Eigen8internal13first_alignedILi16EilEET1_PKT0_S2_.exit.i.i.i.i.i.i.i.i.i.i.i
-  %76 = icmp sgt i64 %67, 3
+  %76 = icmp sgt i64 %64, 3
   br i1 %76, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i, label %._crit_edge.i.i.i.i.i.i.i.i.i.i.i
 
 .lr.ph.i.i.i.i.i.i.i.i.i.i.i:                     ; preds = %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i
-  %.021.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %80, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i ], [ %.0.i.i.i.i.i.i.i.i.i.i.i.i, %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i ]
+  %.021.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %80, %.lr.ph.i.i.i.i.i.i.i.i.i.i.i ], [ %63, %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i ]
   %77 = getelementptr inbounds i32, ptr %58, i64 %.021.i.i.i.i.i.i.i.i.i.i.i
   %78 = getelementptr inbounds i32, ptr %59, i64 %.021.i.i.i.i.i.i.i.i.i.i.i
   %79 = load <2 x i64>, ptr %78, align 1
   store <2 x i64> %79, ptr %77, align 16
   %80 = add nsw i64 %.021.i.i.i.i.i.i.i.i.i.i.i, 4
-  %81 = icmp slt i64 %80, %70
+  %81 = icmp slt i64 %80, %67
   br i1 %81, label %.lr.ph.i.i.i.i.i.i.i.i.i.i.i, label %._crit_edge.i.i.i.i.i.i.i.i.i.i.i, !llvm.loop !106
 
 ._crit_edge.i.i.i.i.i.i.i.i.i.i.i:                ; preds = %.lr.ph.i.i.i.i.i.i.i.i.i.i.i, %_ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dense_assignment_kernelINS0_9evaluatorINS_5BlockINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEELin1ELi1ELb0EEEEENS5_IS8_EENS0_9assign_opIiiEELi0EEEEEvRT_ll.exit.i.i.i.i.i.i.i.i.i.i.i
-  %82 = icmp slt i64 %70, %57
+  %82 = icmp slt i64 %67, %57
   br i1 %82, label %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit
 
 .lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i:                 ; preds = %._crit_edge.i.i.i.i.i.i.i.i.i.i.i, %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i
-  %.05.i18.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %86, %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i ], [ %70, %._crit_edge.i.i.i.i.i.i.i.i.i.i.i ]
+  %.05.i18.i.i.i.i.i.i.i.i.i.i.i = phi i64 [ %86, %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i ], [ %67, %._crit_edge.i.i.i.i.i.i.i.i.i.i.i ]
   %83 = getelementptr inbounds i32, ptr %58, i64 %.05.i18.i.i.i.i.i.i.i.i.i.i.i
   %84 = getelementptr inbounds i32, ptr %59, i64 %.05.i18.i.i.i.i.i.i.i.i.i.i.i
   %85 = load i32, ptr %84, align 4
@@ -7431,7 +7422,7 @@ _ZN5Eigen8internal31unaligned_dense_assignment_loopILb0EE3runINS0_31generic_dens
 
 _ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit: ; preds = %.lr.ph.i17.i.i.i.i.i.i.i.i.i.i.i, %._crit_edge.i.i.i.i.i.i.i.i.i.i.i
   %.not.i.i.i.i.i.i.i.i = icmp eq i64 %57, %46
-  br i1 %.not.i.i.i.i.i.i.i.i, label %87, label %_ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEE6resizeEll.exit.i.i.i.i.i.i.i.i
+  br i1 %.not.i.i.i.i.i.i.i.i, label %84, label %_ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEE6resizeEll.exit.i.i.i.i.i.i.i.i
 
 _ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEE6resizeEll.exit.i.i.i.i.i.i.i.i: ; preds = %_ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit
   invoke void @_ZN5Eigen12DenseStorageIiLin1ELin1ELi1ELi0EE6resizeElll(ptr noundef nonnull align 8 dereferenceable(16) %3, i64 noundef %46, i64 noundef %46, i64 noundef 1)
@@ -7440,38 +7431,38 @@ _ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEE6resizeEll.exit.
 .noexc:                                           ; preds = %_ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEE6resizeEll.exit.i.i.i.i.i.i.i.i
   %.pr.i.i.i.i.i.i.i = load i64, ptr %56, align 8
   %.pre = load ptr, ptr %3, align 8
-  br label %87
-
-87:                                               ; preds = %.noexc, %_ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit
-  %88 = phi ptr [ %.pre, %.noexc ], [ %59, %_ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit ]
-  %89 = phi i64 [ %.pr.i.i.i.i.i.i.i, %.noexc ], [ %46, %_ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit ]
-  %90 = sdiv i64 %89, 4
-  %91 = shl nsw i64 %90, 2
-  %92 = icmp sgt i64 %89, 3
-  br i1 %92, label %.lr.ph.i.i.i.i.i.i.i.i, label %._crit_edge.i.i.i.i.i.i.i.i
-
-.lr.ph.i.i.i.i.i.i.i.i:                           ; preds = %87, %.lr.ph.i.i.i.i.i.i.i.i
-  %.011.i.i.i.i.i.i.i.i = phi i64 [ %96, %.lr.ph.i.i.i.i.i.i.i.i ], [ 0, %87 ]
-  %93 = getelementptr inbounds nuw i32, ptr %88, i64 %.011.i.i.i.i.i.i.i.i
-  %94 = getelementptr inbounds nuw i32, ptr %.sroa.0.0102, i64 %.011.i.i.i.i.i.i.i.i
-  %95 = load <2 x i64>, ptr %94, align 16
-  store <2 x i64> %95, ptr %93, align 16
-  %96 = add nuw nsw i64 %.011.i.i.i.i.i.i.i.i, 4
-  %97 = icmp slt i64 %96, %91
-  br i1 %97, label %.lr.ph.i.i.i.i.i.i.i.i, label %._crit_edge.i.i.i.i.i.i.i.i, !llvm.loop !107
+  br label %84
 
-._crit_edge.i.i.i.i.i.i.i.i:                      ; preds = %.lr.ph.i.i.i.i.i.i.i.i, %87
-  %98 = icmp slt i64 %91, %89
-  br i1 %98, label %.lr.ph.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEaSERKS3_.exit
+._crit_edge.i.i.i.i.i.i.i.i:                      ; preds = %.noexc, %_ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit
+  %85 = phi ptr [ %.pre, %.noexc ], [ %58, %_ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit ]
+  %86 = phi i64 [ %.pr.i.i.i.i.i.i.i, %.noexc ], [ %46, %_ZN5Eigen16CommaInitializerINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEcmIS2_EERS3_RKNS_9DenseBaseIT_EE.exit ]
+  %87 = sdiv i64 %86, 4
+  %88 = shl nsw i64 %87, 2
+  %89 = icmp sgt i64 %86, 3
+  br i1 %89, label %.lr.ph.i.i.i.i.i.i.i.i, label %._crit_edge.i.i.i.i.i.i.i.i
+
+.lr.ph.i.i.i.i.i.i.i.i:                           ; preds = %84, %.lr.ph.i.i.i.i.i.i.i.i
+  %.011.i.i.i.i.i.i.i.i = phi i64 [ %93, %.lr.ph.i.i.i.i.i.i.i.i ], [ 0, %84 ]
+  %90 = getelementptr inbounds nuw i32, ptr %85, i64 %.011.i.i.i.i.i.i.i.i
+  %91 = getelementptr inbounds nuw i32, ptr %.sroa.0.0102, i64 %.011.i.i.i.i.i.i.i.i
+  %92 = load <2 x i64>, ptr %91, align 16
+  store <2 x i64> %92, ptr %90, align 16
+  %93 = add nuw nsw i64 %.011.i.i.i.i.i.i.i.i, 4
+  %94 = icmp slt i64 %93, %88
+  br i1 %94, label %.lr.ph.i.i.i.i.i.i.i.i, label %._crit_edge.i.i.i.i.i.i.i.i, !llvm.loop !107
+
+._crit_edge.i.i.i.i.i.i.i.i:                      ; preds = %.lr.ph.i.i.i.i.i.i.i.i, %84
+  %95 = icmp slt i64 %88, %86
+  br i1 %95, label %.lr.ph.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEaSERKS3_.exit
 
 .lr.ph.i.i.i.i.i.i.i.i.i:                         ; preds = %._crit_edge.i.i.i.i.i.i.i.i, %.lr.ph.i.i.i.i.i.i.i.i.i
-  %.05.i.i.i.i.i.i.i.i.i = phi i64 [ %102, %.lr.ph.i.i.i.i.i.i.i.i.i ], [ %91, %._crit_edge.i.i.i.i.i.i.i.i ]
-  %99 = getelementptr inbounds i32, ptr %88, i64 %.05.i.i.i.i.i.i.i.i.i
+  %.05.i.i.i.i.i.i.i.i.i = phi i64 [ %102, %.lr.ph.i.i.i.i.i.i.i.i.i ], [ %88, %._crit_edge.i.i.i.i.i.i.i.i ]
+  %99 = getelementptr inbounds i32, ptr %85, i64 %.05.i.i.i.i.i.i.i.i.i
   %100 = getelementptr inbounds i32, ptr %.sroa.0.0102, i64 %.05.i.i.i.i.i.i.i.i.i
   %101 = load i32, ptr %100, align 4
   store i32 %101, ptr %99, align 4
   %102 = add nsw i64 %.05.i.i.i.i.i.i.i.i.i, 1
-  %exitcond.not.i.i.i.i.i.i.i.i.i = icmp eq i64 %102, %89
+  %exitcond.not.i.i.i.i.i.i.i.i.i = icmp eq i64 %102, %86
   br i1 %exitcond.not.i.i.i.i.i.i.i.i.i, label %_ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEaSERKS3_.exit, label %.lr.ph.i.i.i.i.i.i.i.i.i, !llvm.loop !108
 
 _ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEEaSERKS3_.exit: ; preds = %.lr.ph.i.i.i.i.i.i.i.i.i, %._crit_edge.i.i.i.i.i.i.i.i
@@ -7508,7 +7499,7 @@ _ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEE6resizeEll.exit.
   unreachable
 
 .noexc50:                                         ; preds = %109, %_ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEE6resizeEll.exit.i.i.i.i.i.i.i.i42
-  %.sink.i69 = phi ptr [ %111, %109 ], [ null, %_ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEE6resizeEll.exit.i.i.i.i.i.i.i.i42 ]
+  %.sink.i69 = phi ptr [ %111, %106 ], [ null, %_ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEE6resizeEll.exit.i.i.i.i.i.i.i.i42 ]
   store ptr %.sink.i69, ptr %8, align 8
   store i64 %104, ptr %10, align 8
   br label %114
@@ -7521,7 +7512,7 @@ _ZN5Eigen15PlainObjectBaseINS_6MatrixIiLin1ELi1ELi0ELin1ELi1EEEE6resizeEll.exit.
   br i1 %118, label %.lr.ph.i.i.i.i.i.i.i.i48, label %._crit_edge.i.i.i.i.i.i.i.i44
 
 .lr.ph.i.i.i.i.i.i.i.i48:                         ; preds = %114, %.lr.ph.i.i.i.i.i.i.i.i48
-  %.011.i.i.i.i.i.i.i.i49 = phi i64 [ %122, %.lr.ph.i.i.i.i.i.i.i.i48 ], [ 0, %114 ]
+  %.011.i.i.i.i.i.i.i.i49 = phi i64 [ %122, %.lr.ph.i.i.i.i.i.i.i.i48 ], [ 0, %111 ]
   %119 = getelementptr inbounds nuw i32, ptr %115, i64 %.011.i.i.i.i.i.i.i.i49
   %120 = getelementptr inbounds nuw i32, ptr %103, i64 %.011.i.i.i.i.i.i.i.i49
   %121 = load <2 x i64>, ptr %120, align 16
@@ -7575,7 +7566,7 @@ _ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEaSIS1_EERS1_RKNS_9DenseBaseIT_EE.exit:
   br label %.body
 
 _ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEC2IliEERKT_RKT0_.exit: ; preds = %134, %_ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEaSIS1_EERS1_RKNS_9DenseBaseIT_EE.exit
-  %.sroa.0.0 = phi ptr [ null, %_ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEaSIS1_EERS1_RKNS_9DenseBaseIT_EE.exit ], [ %136, %134 ]
+  %.sroa.0.0 = phi ptr [ null, %_ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEaSIS1_EERS1_RKNS_9DenseBaseIT_EE.exit ], [ %136, %131 ]
   %141 = load ptr, ptr %2, align 8
   store ptr %.sroa.0.0, ptr %2, align 8
   %142 = getelementptr inbounds nuw i8, ptr %2, i64 8
@@ -7586,12 +7577,12 @@ _ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEC2IliEERKT_RKT0_.exit: ; preds = %134,
   br i1 %144, label %.preheader, label %._crit_edge109
 
 .preheader:                                       ; preds = %_ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEC2IliEERKT_RKT0_.exit, %173
-  %indvars.iv122 = phi i64 [ %indvars.iv.next123, %173 ], [ 0, %_ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEC2IliEERKT_RKT0_.exit ]
+  %indvars.iv122 = phi i64 [ %indvars.iv.next123, %170 ], [ 0, %_ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEC2IliEERKT_RKT0_.exit ]
   %145 = trunc nuw nsw i64 %indvars.iv122 to i32
   br label %146
 
 146:                                              ; preds = %.preheader, %146
-  %indvars.iv118 = phi i64 [ 0, %.preheader ], [ %indvars.iv.next119, %146 ]
+  %indvars.iv118 = phi i64 [ 0, %.preheader ], [ %indvars.iv.next119, %143 ]
   %147 = load ptr, ptr %0, align 8
   %148 = load i64, ptr %22, align 8
   %149 = mul nsw i64 %148, %indvars.iv118
@@ -7640,7 +7631,7 @@ _ZN5Eigen6MatrixIiLin1ELi1ELi0ELin1ELi1EEC2IliEERKT_RKT0_.exit: ; preds = %134,
   ret void
 
 .body:                                            ; preds = %53, %139, %37, %171
-  %.pn39 = phi { ptr, i32 } [ %172, %171 ], [ %54, %53 ], [ %38, %37 ], [ %140, %139 ]
+  %.pn39 = phi { ptr, i32 } [ %172, %168 ], [ %54, %53 ], [ %38, %37 ], [ %140, %136 ]
   %177 = load ptr, ptr %8, align 8
   call void @free(ptr noundef %177) #21
   br label %common.resume
diff --git a/bench/openjdk/optimized/methodData.ll b/bench/openjdk/optimized/methodData.ll
index eb582205d0a..23dacf1de4f 100644
--- a/bench/openjdk/optimized/methodData.ll
+++ b/bench/openjdk/optimized/methodData.ll
@@ -247,8 +247,6 @@ $_ZTVN16MetaspaceClosure6MSORefI6MethodEE = comdat any
 @.str.36 = private unnamed_addr constant [20 x i8] c"SpeculativeTrapData\00", align 1
 @.str.37 = private unnamed_addr constant [48 x i8] c"unexpected bytecode for var length profile data\00", align 1
 @UseTypeSpeculation = external local_unnamed_addr global i8, align 1
-@.str.38 = private unnamed_addr constant [61 x i8] c"guarantee(is_aligned(fs, sizeof(FailedSpeculation*))) failed\00", align 1
-@.str.39 = private unnamed_addr constant [50 x i8] c"FailedSpeculation objects must be pointer aligned\00", align 1
 @.str.40 = private unnamed_addr constant [37 x i8] c"guarantee(*cursor != nullptr) failed\00", align 1
 @.str.41 = private unnamed_addr constant [48 x i8] c"cursor must point to non-null FailedSpeculation\00", align 1
 @SpecTrapLimitExtraEntries = external local_unnamed_addr global i32, align 4
@@ -3782,9 +3780,9 @@ _ZL35guarantee_failed_speculations_aliveP7nmethodPP17FailedSpeculation.exit: ; p
   br label %26
 
 26:                                               ; preds = %56, %_ZL35guarantee_failed_speculations_aliveP7nmethodPP17FailedSpeculation.exit
-  %27 = phi ptr [ %.val, %_ZL35guarantee_failed_speculations_aliveP7nmethodPP17FailedSpeculation.exit ], [ %.pre, %56 ]
-  %.024 = phi ptr [ %1, %_ZL35guarantee_failed_speculations_aliveP7nmethodPP17FailedSpeculation.exit ], [ %57, %56 ]
-  %.023 = phi ptr [ null, %_ZL35guarantee_failed_speculations_aliveP7nmethodPP17FailedSpeculation.exit ], [ %.232, %56 ]
+  %27 = phi ptr [ %.val, %_ZL35guarantee_failed_speculations_aliveP7nmethodPP17FailedSpeculation.exit ], [ %.pre, %51 ]
+  %.024 = phi ptr [ %1, %_ZL35guarantee_failed_speculations_aliveP7nmethodPP17FailedSpeculation.exit ], [ %57, %51 ]
+  %.023 = phi ptr [ null, %_ZL35guarantee_failed_speculations_aliveP7nmethodPP17FailedSpeculation.exit ], [ %.232, %51 ]
   %28 = icmp eq ptr %27, null
   br i1 %28, label %29, label %.thread29
 
@@ -3803,16 +3801,7 @@ _ZL35guarantee_failed_speculations_aliveP7nmethodPP17FailedSpeculation.exit: ; p
   store ptr null, ptr %35, align 8
   %36 = getelementptr inbounds nuw i8, ptr %32, i64 16
   tail call void @llvm.memcpy.p0.p0.i64(ptr nonnull align 8 %36, ptr readonly align 1 %2, i64 %6, i1 false)
-  %37 = ptrtoint ptr %32 to i64
-  %38 = and i64 %37, 7
-  %39 = icmp eq i64 %38, 0
-  br i1 %39, label %42, label %40
-
-40:                                               ; preds = %34
-  %41 = load ptr, ptr @g_assert_poison, align 8
-  store i8 88, ptr %41, align 1
-  tail call void (ptr, i32, ptr, ptr, ...) @_Z15report_vm_errorPKciS0_S0_z(ptr noundef nonnull @.str.5, i32 noundef 848, ptr noundef nonnull @.str.38, ptr noundef nonnull @.str.39) #22
-  unreachable
+  br label %37
 
 42:                                               ; preds = %34, %29
   %.1 = phi ptr [ %32, %34 ], [ %.023, %29 ]
@@ -3832,8 +3821,8 @@ _ZL35guarantee_failed_speculations_aliveP7nmethodPP17FailedSpeculation.exit: ; p
   unreachable
 
 .thread29:                                        ; preds = %26, %45
-  %.232 = phi ptr [ %.1, %45 ], [ %.023, %26 ]
-  %48 = phi ptr [ %.pr, %45 ], [ %27, %26 ]
+  %.232 = phi ptr [ %.1, %40 ], [ %.023, %26 ]
+  %48 = phi ptr [ %.pr, %40 ], [ %27, %26 ]
   %49 = load i32, ptr %48, align 8
   %50 = icmp eq i32 %49, %3
   br i1 %50, label %51, label %56
@@ -3858,7 +3847,7 @@ _ZL35guarantee_failed_speculations_aliveP7nmethodPP17FailedSpeculation.exit: ; p
   br label %26, !llvm.loop !24
 
 .thread:                                          ; preds = %31, %42, %54, %55
-  %.0 = phi i1 [ false, %55 ], [ false, %54 ], [ false, %31 ], [ true, %42 ]
+  %.0 = phi i1 [ false, %50 ], [ false, %49 ], [ false, %31 ], [ true, %37 ]
   ret i1 %.0
 }
 
diff --git a/bench/openjdk/optimized/screencast_pipewire.ll b/bench/openjdk/optimized/screencast_pipewire.ll
index 12edb35a81d..e6ef5743c38 100644
--- a/bench/openjdk/optimized/screencast_pipewire.ll
+++ b/bench/openjdk/optimized/screencast_pipewire.ll
@@ -2603,22 +2603,18 @@ define internal void @onStreamParamChanged(ptr noundef %0, i32 noundef %1, ptr n
   %26 = getelementptr inbounds nuw i8, ptr %6, i64 24
   %27 = getelementptr inbounds nuw i8, ptr %6, i64 16
   %.not.i.i.i.i = icmp ult i32 %24, -8
-  %28 = ptrtoint ptr %2 to i64
-  %29 = and i64 %28, 3
-  %30 = icmp eq i64 %29, 0
-  %or.cond.i = and i1 %30, %.not.i.i.i.i
-  br i1 %or.cond.i, label %31, label %spa_format_parse.exit.thread
-
-31:                                               ; preds = %23
-  %32 = zext i32 %25 to i64
-  %33 = zext i32 %24 to i64
-  %34 = add nsw i64 %33, -1
-  %35 = or i64 %34, 7
-  %36 = add nsw i64 %35, 9
-  %.not13.i.i.i.i = icmp ugt i64 %36, %32
+  br i1 %.not.i.i.i.i, label %28, label %spa_format_parse.exit.thread
+
+28:; preds = %23
+  %29 = zext i32 %25 to i64
+  %30 = zext i32 %24 to i64
+  %31 = add nsw i64 %30, -1
+  %32 = or i64 %31, 7
+  %33 = add nsw i64 %32, 9
+  %.not13.i.i.i.i = icmp ugt i64 %33, %29
   br i1 %.not13.i.i.i.i, label %spa_format_parse.exit.thread, label %spa_pod_parser_current.exit.i.i
 
-spa_pod_parser_current.exit.i.i:                  ; preds = %31
+spa_pod_parser_current.exit.i.i:                  ; preds = %28
   %37 = getelementptr inbounds nuw i8, ptr %2, i64 4
   %38 = load i32, ptr %37, align 4
   %39 = icmp ne i32 %38, 15
@@ -2632,7 +2628,7 @@ spa_pod_parser_current.exit.i.i:                  ; preds = %31
   %.not16.i.i = icmp eq i32 %43, 262147
   br i1 %.not16.i.i, label %spa_format_parse.exit, label %spa_format_parse.exit.thread
 
-spa_format_parse.exit.thread:                     ; preds = %spa_pod_parser_current.exit.i.i, %31, %23, %41
+spa_format_parse.exit.thread:                     ; preds = %spa_pod_parser_current.exit.i.i, %28, %23, %41
   call void @llvm.lifetime.end.p0(ptr nonnull %6)
   call void @llvm.lifetime.end.p0(ptr nonnull %7)
   br label %151
@@ -2695,7 +2691,7 @@ spa_pod_is_object.exit.i.i:                       ; preds = %52
   br i1 %.not.i25.i.i.i, label %spa_pod_find_prop.exit.thread.i, label %.lr.ph.i.i.i
 
 .lr.ph.i.i.i:                                     ; preds = %61, %76
-  %.026.i.i.i = phi ptr [ %82, %76 ], [ %64, %61 ]
+  %.026.i.i.i = phi ptr [ %82, %73 ], [ %64, %58 ]
   %68 = getelementptr inbounds nuw i8, ptr %.026.i.i.i, i64 8
   %69 = load i32, ptr %68, align 4
   %70 = zext i32 %69 to i64
@@ -2726,7 +2722,7 @@ spa_pod_find_prop.exit.i:                         ; preds = %73
   br label %spa_pod_find_prop.exit.thread.i
 
 spa_pod_find_prop.exit.thread.i:                  ; preds = %76, %.lr.ph.i.i.i, %spa_pod_find_prop.exit.i, %61, %52
-  %84 = phi i32 [ %.pre27.i, %52 ], [ %.pre27.i, %61 ], [ %.pre.i, %spa_pod_find_prop.exit.i ], [ %.pre27.i, %.lr.ph.i.i.i ], [ %.pre27.i, %76 ]
+  %84 = phi i32 [ %.pre27.i, %49 ], [ %.pre27.i, %58 ], [ %.pre.i, %spa_pod_find_prop.exit.i ], [ %.pre27.i, %.lr.ph.i.i.i ], [ %.pre27.i, %73 ]
   %85 = add i32 %84, 8
   store ptr %2, ptr %4, align 8
   %.sroa.2.0..sroa_idx.i.i.i24 = getelementptr inbounds nuw i8, ptr %4, i64 8
@@ -2827,7 +2823,7 @@ spa_format_video_raw_parse.exit:                  ; preds = %101
   br label %spa_debug_type_find_name.exit
 
 spa_debug_type_find_name.exit:                    ; preds = %128, %142
-  %.0.i33 = phi ptr [ %144, %142 ], [ null, %128 ]
+  %.0.i33 = phi ptr [ %144, %139 ], [ null, %125 ]
   %145 = load i32, ptr %106, align 8
   %146 = getelementptr inbounds nuw i8, ptr %0, i64 76
   %147 = load i32, ptr %146, align 4
@@ -2836,7 +2832,7 @@ spa_debug_type_find_name.exit:                    ; preds = %128, %142
   store i32 1, ptr %148, align 8
   %149 = load ptr, ptr @fp_pw_thread_loop_signal, align 8
   %150 = load ptr, ptr @pw, align 8
-  call void %149(ptr noundef %150, i1 noundef zeroext true) #16
+  call void %146(ptr noundef %150, i1 noundef zeroext true) #16
   br label %151
 
 151:                                              ; preds = %spa_format_video_raw_parse.exit.thread, %spa_format_parse.exit.thread, %spa_format_video_raw_parse.exit, %47, %spa_format_parse.exit, %3, %spa_debug_type_find_name.exit
diff --git a/bench/pingora-rs/optimized/acx590v32pn6eo20biyujqd8g.ll b/bench/pingora-rs/optimized/acx590v32pn6eo20biyujqd8g.ll
index 9117a5d1502..3011fbbe614 100644
--- a/bench/pingora-rs/optimized/acx590v32pn6eo20biyujqd8g.ll
+++ b/bench/pingora-rs/optimized/acx590v32pn6eo20biyujqd8g.ll
@@ -123,9 +123,7 @@ target triple = "x86_64-unknown-linux-gnu"
 @anon.9683d7273f0e2da39fab36790461676f.134 = private unnamed_addr constant [20 x i8] c"SendPingWhilePending", align 1
 @anon.9683d7273f0e2da39fab36790461676f.135 = private unnamed_addr constant [24 x i8] c"SendSettingsWhilePending", align 1
 @anon.9683d7273f0e2da39fab36790461676f.136 = private unnamed_addr constant [22 x i8] c"PeerDisabledServerPush", align 1
-@anon.9683d7273f0e2da39fab36790461676f.140 = private unnamed_addr constant [8 x i8] zeroinitializer, align 8
 @anon.9683d7273f0e2da39fab36790461676f.141 = private unnamed_addr constant [101 x i8] c"/home/dtcxzyw/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/arc-swap-1.7.1/src/debt/helping.rs", align 1
-@anon.9683d7273f0e2da39fab36790461676f.142 = private unnamed_addr constant <{ ptr, [16 x i8] }> <{ ptr @anon.9683d7273f0e2da39fab36790461676f.141, [16 x i8] c"e\00\00\00\00\00\00\00\10\01\00\00\15\00\00\00" }>, align 8
 @anon.9683d7273f0e2da39fab36790461676f.143 = private unnamed_addr constant [64 x i8] c"internal error: entered unreachable code: Invalid control value ", align 1
 @anon.9683d7273f0e2da39fab36790461676f.144 = private unnamed_addr constant <{ ptr, [8 x i8] }> <{ ptr @anon.9683d7273f0e2da39fab36790461676f.143, [8 x i8] c"@\00\00\00\00\00\00\00" }>, align 8
 @anon.9683d7273f0e2da39fab36790461676f.145 = private unnamed_addr constant <{ ptr, [16 x i8] }> <{ ptr @anon.9683d7273f0e2da39fab36790461676f.141, [16 x i8] c"e\00\00\00\00\00\00\00+\01\00\00\16\00\00\00" }>, align 8
@@ -6631,23 +6629,21 @@ define hidden void @_ZN8arc_swap4debt7helping5Slots4help17h2d8e69b46d0bf67aE(ptr
   %6 = alloca [8 x i8], align 8
   %7 = alloca [16 x i8], align 8
   %8 = alloca [48 x i8], align 8
-  %9 = alloca [48 x i8], align 8
+  %9 = alloca [8 x i8], align 8
   %10 = alloca [8 x i8], align 8
-  %11 = alloca [8 x i8], align 8
-  %12 = alloca [8 x i8], align 8
-  call void @llvm.lifetime.start.p0(ptr nonnull %12)
-  %13 = load atomic i64, ptr %1 seq_cst, align 8
-  store i64 %13, ptr %12, align 8
-  %14 = getelementptr inbounds nuw i8, ptr %1, i64 16
-  %15 = getelementptr inbounds nuw i8, ptr %3, i64 8
-  %.val = load ptr, ptr %15, align 8, !nonnull !3, !align !38
-  %16 = getelementptr inbounds nuw i8, ptr %5, i64 8
-  %17 = getelementptr inbounds nuw i8, ptr %1, i64 32
-  %18 = getelementptr inbounds nuw i8, ptr %0, i64 32
+  call void @llvm.lifetime.start.p0(ptr nonnull %10)
+  %11 = load atomic i64, ptr %1 seq_cst, align 8
+  store i64 %11, ptr %10, align 8
+  %12 = getelementptr inbounds nuw i8, ptr %1, i64 16
+  %13 = getelementptr inbounds nuw i8, ptr %3, i64 8
+  %.val = load ptr, ptr %13, align 8, !nonnull !3, !align !38
+  %15 = getelementptr inbounds nuw i8, ptr %5, i64 8
+  %15 = getelementptr inbounds nuw i8, ptr %1, i64 32
+  %16 = getelementptr inbounds nuw i8, ptr %0, i64 32
   br label %.backedge
 
 .backedge:                                        ; preds = %.backedge.backedge, %4
-  %19 = phi i64 [ %13, %4 ], [ %.be, %.backedge.backedge ]
+  %19 = phi i64 [ %11, %4 ], [ %.be, %.backedge.backedge ]
   %20 = and i64 %19, 3
   switch i64 %20, label %default.unreachable31 [
     i64 0, label %25
@@ -6662,7 +6658,7 @@ default.unreachable31:                            ; preds = %.backedge
 .loopexit8:                                       ; preds = %.backedge, %25
   call void @llvm.lifetime.start.p0(ptr nonnull %8)
   call void @llvm.lifetime.start.p0(ptr nonnull %7)
-  store ptr %12, ptr %7, align 8
+  store ptr %10, ptr %7, align 8
   %.sroa.44.0..sroa_idx = getelementptr inbounds nuw i8, ptr %7, i64 8
   store ptr @"_ZN4core3fmt3num55_$LT$impl$u20$core..fmt..UpperHex$u20$for$u20$usize$GT$3fmt17hdee01be989c5eb83E", ptr %.sroa.44.0..sroa_idx, align 8
   store ptr @anon.9683d7273f0e2da39fab36790461676f.144, ptr %8, align 8
@@ -6682,16 +6678,16 @@ default.unreachable31:                            ; preds = %.backedge
   br i1 %26, label %.loopexit, label %.loopexit8, !prof !12
 
 .loopexit:                                        ; preds = %46, %.backedge, %25, %67
-  call void @llvm.lifetime.end.p0(ptr nonnull %12)
+  call void @llvm.lifetime.end.p0(ptr nonnull %10)
   ret void
 
 27:                                               ; preds = %.backedge
-  %28 = load atomic i64, ptr %14 seq_cst, align 8
+  %28 = load atomic i64, ptr %12 seq_cst, align 8
   %.not = icmp eq i64 %28, %2
   br i1 %.not, label %29, label %46
 
 29:                                               ; preds = %27
-  call void @llvm.lifetime.start.p0(ptr nonnull %11)
+  call void @llvm.lifetime.start.p0(ptr nonnull %9)
   call void @llvm.lifetime.start.p0(ptr nonnull %6)
   %30 = load ptr, ptr %.val, align 8, !nonnull !3, !align !38, !noundef !3
   store ptr %30, ptr %6, align 8
@@ -6701,13 +6697,13 @@ default.unreachable31:                            ; preds = %.backedge
   call void @llvm.lifetime.end.p0(ptr nonnull %6)
   call void @llvm.lifetime.start.p0(ptr nonnull %5)
   store ptr %32, ptr %5, align 8
-  store ptr null, ptr %16, align 8
+  store ptr null, ptr %14, align 8
   %.not.i.i = icmp eq ptr %33, null
   br i1 %.not.i.i, label %"_ZN124_$LT$arc_swap..strategy..hybrid..HybridStrategy$LT$Cfg$GT$$u20$as$u20$arc_swap..strategy..sealed..InnerStrategy$LT$T$GT$$GT$16wait_for_readers28_$u7b$$u7b$closure$u7d$$u7d$17hf7bd4b393d731f3aE.exit", label %34
 
 34:                                               ; preds = %29
   %35 = invoke noundef ptr @_ZN8arc_swap7ref_cnt6RefCnt3inc17h6378a83ba566ab00E(ptr noalias noundef nonnull readonly align 8 dereferenceable(8) %5)
-          to label %38 unwind label %36
+          to label %36 unwind label %36
 
 36:                                               ; preds = %41, %34
   %37 = landingpad { ptr, i32 }
@@ -6731,101 +6727,77 @@ default.unreachable31:                            ; preds = %.backedge
   call void @_ZN4core9panicking16panic_in_cleanup17hccd47ddd364deb23E() #22
   unreachable
 
-common.resume:                                    ; preds = %69, %73, %36
-  %common.resume.op = phi { ptr, i32 } [ %37, %36 ], [ %lpad.phi, %73 ], [ %lpad.phi, %69 ]
+common.resume:                                    ; preds = %.loopexit9, %73, %36
+  %common.resume.op = phi { ptr, i32 } [ %37, %34 ], [ %lpad.loopexit, %65 ], [ %lpad.loopexit, %.loopexit9 ]
   resume { ptr, i32 } %common.resume.op
 
 "_ZN124_$LT$arc_swap..strategy..hybrid..HybridStrategy$LT$Cfg$GT$$u20$as$u20$arc_swap..strategy..sealed..InnerStrategy$LT$T$GT$$GT$16wait_for_readers28_$u7b$$u7b$closure$u7d$$u7d$17hf7bd4b393d731f3aE.exit": ; preds = %29, %38, %41
   %44 = load ptr, ptr %5, align 8, !nonnull !3, !noundef !3
   call void @llvm.lifetime.end.p0(ptr nonnull %5)
-  store ptr %44, ptr %11, align 8
-  %45 = invoke noundef ptr @"_ZN71_$LT$alloc..sync..Arc$LT$T$GT$$u20$as$u20$arc_swap..ref_cnt..RefCnt$GT$6as_ptr17h0697427cf70d9e9bE"(ptr noalias noundef nonnull readonly align 8 dereferenceable(8) %11)
-          to label %50 unwind label %.loopexit9
+  store ptr %44, ptr %9, align 8
+  %45 = invoke noundef ptr @"_ZN71_$LT$alloc..sync..Arc$LT$T$GT$$u20$as$u20$arc_swap..ref_cnt..RefCnt$GT$6as_ptr17h0697427cf70d9e9bE"(ptr noalias noundef nonnull readonly align 8 dereferenceable(8) %9)
+          to label %48 unwind label %.loopexit9
 
 46:                                               ; preds = %27
   %47 = load atomic i64, ptr %1 seq_cst, align 8
-  %48 = load i64, ptr %12, align 8, !noundef !3
+  %48 = load i64, ptr %10, align 8, !noundef !3
   %49 = icmp eq i64 %47, %48
   br i1 %49, label %.loopexit, label %76
 
 50:                                               ; preds = %"_ZN124_$LT$arc_swap..strategy..hybrid..HybridStrategy$LT$Cfg$GT$$u20$as$u20$arc_swap..strategy..sealed..InnerStrategy$LT$T$GT$$GT$16wait_for_readers28_$u7b$$u7b$closure$u7d$$u7d$17hf7bd4b393d731f3aE.exit"
   %51 = ptrtoint ptr %45 to i64
-  %52 = load atomic ptr, ptr %17 seq_cst, align 8
-  %53 = load atomic ptr, ptr %18 seq_cst, align 8
+  %52 = load atomic ptr, ptr %15 seq_cst, align 8
+  %53 = load atomic ptr, ptr %16 seq_cst, align 8
   store atomic i64 %51, ptr %53 seq_cst, align 8
   %54 = ptrtoint ptr %53 to i64
-  %55 = and i64 %54, 3
-  store i64 %55, ptr %10, align 8
-  %56 = icmp eq i64 %55, 0
-  br i1 %56, label %58, label %57, !prof !12
-
-57:                                               ; preds = %50
-  call void @llvm.lifetime.start.p0(ptr nonnull %9)
-  store ptr null, ptr %9, align 8
-  invoke void @_ZN4core9panicking13assert_failed17h77d57c61d0ad0ecdE(i8 noundef 0, ptr noalias noundef nonnull readonly align 8 dereferenceable(8) %10, ptr noalias noundef readonly align 8 dereferenceable(8) @anon.9683d7273f0e2da39fab36790461676f.140, ptr noalias noundef nonnull align 8 captures(none) dereferenceable(48) %9, ptr noalias noundef readonly align 8 dereferenceable(24) @anon.9683d7273f0e2da39fab36790461676f.142) #21
-          to label %68 unwind label %.loopexit.split-lp
-
-58:                                               ; preds = %50
-  %59 = or disjoint i64 %54, 1
-  %60 = load i64, ptr %12, align 8, !noundef !3
-  %61 = cmpxchg ptr %1, i64 %60, i64 %59 seq_cst seq_cst, align 8
-  %.sroa.18.0.in.i = extractvalue { i64, i1 } %61, 1
-  br i1 %.sroa.18.0.in.i, label %67, label %62
-
-62:                                               ; preds = %58
-  %.sroa.01.0.i = extractvalue { i64, i1 } %61, 0
-  store i64 %.sroa.01.0.i, ptr %12, align 8
+  %53 = or disjoint i64 %54, 1
+  %54 = load i64, ptr %10, align 8, !noundef !3
+  %55 = cmpxchg ptr %1, i64 %54, i64 %53 seq_cst seq_cst, align 8
+  %.sroa.18.0.in.i = extractvalue { i64, i1 } %55, 1
+  br i1 %.sroa.18.0.in.i, label %61, label %56
+
+56: ; preds = %48
+  %.sroa.01.0.i = extractvalue { i64, i1 } %55, 0
+  store i64 %.sroa.01.0.i, ptr %10, align 8
   call void @llvm.experimental.noalias.scope.decl(metadata !278)
   call void @llvm.experimental.noalias.scope.decl(metadata !281)
-  %63 = load ptr, ptr %11, align 8, !alias.scope !284, !nonnull !3, !noundef !3
+  %63 = load ptr, ptr %9, align 8, !alias.scope !284, !nonnull !3, !noundef !3
   %64 = atomicrmw sub ptr %63, i64 1 release, align 8, !noalias !284
   %65 = icmp eq i64 %64, 1
   br i1 %65, label %66, label %"_ZN4core3ptr124drop_in_place$LT$alloc..sync..Arc$LT$alloc..collections..btree..set..BTreeSet$LT$pingora_load_balancing..Backend$GT$$GT$$GT$17h420a8e38ce4e3b45E.exit"
 
-66:                                               ; preds = %62
+66:                                               ; preds = %56
   fence acquire
-  call void @"_ZN5alloc4sync16Arc$LT$T$C$A$GT$9drop_slow17haef3ab0c4601cc40E"(ptr noalias noundef nonnull align 8 dereferenceable(8) %11)
+  call void @"_ZN5alloc4sync16Arc$LT$T$C$A$GT$9drop_slow17haef3ab0c4601cc40E"(ptr noalias noundef nonnull align 8 dereferenceable(8) %9)
   br label %"_ZN4core3ptr124drop_in_place$LT$alloc..sync..Arc$LT$alloc..collections..btree..set..BTreeSet$LT$pingora_load_balancing..Backend$GT$$GT$$GT$17h420a8e38ce4e3b45E.exit"
 
-"_ZN4core3ptr124drop_in_place$LT$alloc..sync..Arc$LT$alloc..collections..btree..set..BTreeSet$LT$pingora_load_balancing..Backend$GT$$GT$$GT$17h420a8e38ce4e3b45E.exit": ; preds = %62, %66
-  call void @llvm.lifetime.end.p0(ptr nonnull %11)
-  %.pr = load i64, ptr %12, align 8
+"_ZN4core3ptr124drop_in_place$LT$alloc..sync..Arc$LT$alloc..collections..btree..set..BTreeSet$LT$pingora_load_balancing..Backend$GT$$GT$$GT$17h420a8e38ce4e3b45E.exit": ; preds = %56, %66
+  call void @llvm.lifetime.end.p0(ptr nonnull %9)
+  %.pr = load i64, ptr %10, align 8
   br label %.backedge.backedge
 
 .backedge.backedge:                               ; preds = %"_ZN4core3ptr124drop_in_place$LT$alloc..sync..Arc$LT$alloc..collections..btree..set..BTreeSet$LT$pingora_load_balancing..Backend$GT$$GT$$GT$17h420a8e38ce4e3b45E.exit", %76
-  %.be = phi i64 [ %47, %76 ], [ %.pr, %"_ZN4core3ptr124drop_in_place$LT$alloc..sync..Arc$LT$alloc..collections..btree..set..BTreeSet$LT$pingora_load_balancing..Backend$GT$$GT$$GT$17h420a8e38ce4e3b45E.exit" ]
+  %.be = phi i64 [ %47, %68 ], [ %.pr, %"_ZN4core3ptr124drop_in_place$LT$alloc..sync..Arc$LT$alloc..collections..btree..set..BTreeSet$LT$pingora_load_balancing..Backend$GT$$GT$$GT$17h420a8e38ce4e3b45E.exit" ]
   br label %.backedge
 
-67:                                               ; preds = %58
-  store atomic ptr %52, ptr %18 seq_cst, align 8
-  call void @llvm.lifetime.end.p0(ptr nonnull %11)
+67:                                               ; preds = %48
+  store atomic ptr %52, ptr %16 seq_cst, align 8
+  call void @llvm.lifetime.end.p0(ptr nonnull %9)
   br label %.loopexit
 
-68:                                               ; preds = %57
-  unreachable
-
 .loopexit9:                                       ; preds = %"_ZN124_$LT$arc_swap..strategy..hybrid..HybridStrategy$LT$Cfg$GT$$u20$as$u20$arc_swap..strategy..sealed..InnerStrategy$LT$T$GT$$GT$16wait_for_readers28_$u7b$$u7b$closure$u7d$$u7d$17hf7bd4b393d731f3aE.exit"
   %lpad.loopexit = landingpad { ptr, i32 }
           cleanup
-  br label %69
-
-.loopexit.split-lp:                               ; preds = %57
-  %lpad.loopexit.split-lp = landingpad { ptr, i32 }
-          cleanup
-  br label %69
-
-69:                                               ; preds = %.loopexit.split-lp, %.loopexit9
-  %lpad.phi = phi { ptr, i32 } [ %lpad.loopexit, %.loopexit9 ], [ %lpad.loopexit.split-lp, %.loopexit.split-lp ]
   call void @llvm.experimental.noalias.scope.decl(metadata !285)
   call void @llvm.experimental.noalias.scope.decl(metadata !288)
-  %70 = load ptr, ptr %11, align 8, !alias.scope !291, !nonnull !3, !noundef !3
+  %70 = load ptr, ptr %9, align 8, !alias.scope !291, !nonnull !3, !noundef !3
   %71 = atomicrmw sub ptr %70, i64 1 release, align 8, !noalias !291
   %72 = icmp eq i64 %71, 1
   br i1 %72, label %73, label %common.resume
 
-73:                                               ; preds = %69
+73:                                               ; preds = %.loopexit9
   fence acquire
-  invoke void @"_ZN5alloc4sync16Arc$LT$T$C$A$GT$9drop_slow17haef3ab0c4601cc40E"(ptr noalias noundef nonnull align 8 dereferenceable(8) %11)
+  invoke void @"_ZN5alloc4sync16Arc$LT$T$C$A$GT$9drop_slow17haef3ab0c4601cc40E"(ptr noalias noundef nonnull align 8 dereferenceable(8) %9)
           to label %common.resume unwind label %74
 
 74:                                               ; preds = %73
@@ -6835,7 +6807,7 @@ common.resume:                                    ; preds = %69, %73, %36
   unreachable
 
 76:                                               ; preds = %46
-  store i64 %47, ptr %12, align 8
+  store i64 %47, ptr %10, align 8
   br label %.backedge.backedge
 }
 
@@ -6845,23 +6817,21 @@ define hidden void @_ZN8arc_swap4debt7helping5Slots4help17h87e00bf7e6d26d9fE(ptr
   %6 = alloca [8 x i8], align 8
   %7 = alloca [16 x i8], align 8
   %8 = alloca [48 x i8], align 8
-  %9 = alloca [48 x i8], align 8
+  %9 = alloca [8 x i8], align 8
   %10 = alloca [8 x i8], align 8
-  %11 = alloca [8 x i8], align 8
-  %12 = alloca [8 x i8], align 8
-  call void @llvm.lifetime.start.p0(ptr nonnull %12)
-  %13 = load atomic i64, ptr %1 seq_cst, align 8
-  store i64 %13, ptr %12, align 8
-  %14 = getelementptr inbounds nuw i8, ptr %1, i64 16
-  %15 = getelementptr inbounds nuw i8, ptr %3, i64 8
-  %.val = load ptr, ptr %15, align 8, !nonnull !3, !align !38
-  %16 = getelementptr inbounds nuw i8, ptr %5, i64 8
-  %17 = getelementptr inbounds nuw i8, ptr %1, i64 32
-  %18 = getelementptr inbounds nuw i8, ptr %0, i64 32
+  call void @llvm.lifetime.start.p0(ptr nonnull %10)
+  %11 = load atomic i64, ptr %1 seq_cst, align 8
+  store i64 %11, ptr %10, align 8
+  %12 = getelementptr inbounds nuw i8, ptr %1, i64 16
+  %13 = getelementptr inbounds nuw i8, ptr %3, i64 8
+  %.val = load ptr, ptr %13, align 8, !nonnull !3, !align !38
+  %15 = getelementptr inbounds nuw i8, ptr %5, i64 8
+  %15 = getelementptr inbounds nuw i8, ptr %1, i64 32
+  %16 = getelementptr inbounds nuw i8, ptr %0, i64 32
   br label %.backedge
 
 .backedge:                                        ; preds = %.backedge.backedge, %4
-  %19 = phi i64 [ %13, %4 ], [ %.be, %.backedge.backedge ]
+  %19 = phi i64 [ %11, %4 ], [ %.be, %.backedge.backedge ]
   %20 = and i64 %19, 3
   switch i64 %20, label %default.unreachable31 [
     i64 0, label %25
@@ -6876,7 +6846,7 @@ default.unreachable31:                            ; preds = %.backedge
 .loopexit8:                                       ; preds = %.backedge, %25
   call void @llvm.lifetime.start.p0(ptr nonnull %8)
   call void @llvm.lifetime.start.p0(ptr nonnull %7)
-  store ptr %12, ptr %7, align 8
+  store ptr %10, ptr %7, align 8
   %.sroa.44.0..sroa_idx = getelementptr inbounds nuw i8, ptr %7, i64 8
   store ptr @"_ZN4core3fmt3num55_$LT$impl$u20$core..fmt..UpperHex$u20$for$u20$usize$GT$3fmt17hdee01be989c5eb83E", ptr %.sroa.44.0..sroa_idx, align 8
   store ptr @anon.9683d7273f0e2da39fab36790461676f.144, ptr %8, align 8
@@ -6896,16 +6866,16 @@ default.unreachable31:                            ; preds = %.backedge
   br i1 %26, label %.loopexit, label %.loopexit8, !prof !12
 
 .loopexit:                                        ; preds = %46, %.backedge, %25, %67
-  call void @llvm.lifetime.end.p0(ptr nonnull %12)
+  call void @llvm.lifetime.end.p0(ptr nonnull %10)
   ret void
 
 27:                                               ; preds = %.backedge
-  %28 = load atomic i64, ptr %14 seq_cst, align 8
+  %28 = load atomic i64, ptr %12 seq_cst, align 8
   %.not = icmp eq i64 %28, %2
   br i1 %.not, label %29, label %46
 
 29:                                               ; preds = %27
-  call void @llvm.lifetime.start.p0(ptr nonnull %11)
+  call void @llvm.lifetime.start.p0(ptr nonnull %9)
   call void @llvm.lifetime.start.p0(ptr nonnull %6)
   %30 = load ptr, ptr %.val, align 8, !nonnull !3, !align !38, !noundef !3
   store ptr %30, ptr %6, align 8
@@ -6915,13 +6885,13 @@ default.unreachable31:                            ; preds = %.backedge
   call void @llvm.lifetime.end.p0(ptr nonnull %6)
   call void @llvm.lifetime.start.p0(ptr nonnull %5)
   store ptr %32, ptr %5, align 8
-  store ptr null, ptr %16, align 8
+  store ptr null, ptr %14, align 8
   %.not.i.i = icmp eq ptr %33, null
   br i1 %.not.i.i, label %"_ZN124_$LT$arc_swap..strategy..hybrid..HybridStrategy$LT$Cfg$GT$$u20$as$u20$arc_swap..strategy..sealed..InnerStrategy$LT$T$GT$$GT$16wait_for_readers28_$u7b$$u7b$closure$u7d$$u7d$17h720e2579c38d4928E.exit", label %34
 
 34:                                               ; preds = %29
   %35 = invoke noundef ptr @_ZN8arc_swap7ref_cnt6RefCnt3inc17hdf9c7a12f5206704E(ptr noalias noundef nonnull readonly align 8 dereferenceable(8) %5)
-          to label %38 unwind label %36
+          to label %36 unwind label %36
 
 36:                                               ; preds = %41, %34
   %37 = landingpad { ptr, i32 }
@@ -6945,101 +6915,77 @@ default.unreachable31:                            ; preds = %.backedge
   call void @_ZN4core9panicking16panic_in_cleanup17hccd47ddd364deb23E() #22
   unreachable
 
-common.resume:                                    ; preds = %69, %73, %36
-  %common.resume.op = phi { ptr, i32 } [ %37, %36 ], [ %lpad.phi, %73 ], [ %lpad.phi, %69 ]
+common.resume:                                    ; preds = %.loopexit9, %73, %36
+  %common.resume.op = phi { ptr, i32 } [ %37, %34 ], [ %lpad.loopexit, %65 ], [ %lpad.loopexit, %.loopexit9 ]
   resume { ptr, i32 } %common.resume.op
 
 "_ZN124_$LT$arc_swap..strategy..hybrid..HybridStrategy$LT$Cfg$GT$$u20$as$u20$arc_swap..strategy..sealed..InnerStrategy$LT$T$GT$$GT$16wait_for_readers28_$u7b$$u7b$closure$u7d$$u7d$17h720e2579c38d4928E.exit": ; preds = %29, %38, %41
   %44 = load ptr, ptr %5, align 8, !nonnull !3, !noundef !3
   call void @llvm.lifetime.end.p0(ptr nonnull %5)
-  store ptr %44, ptr %11, align 8
-  %45 = invoke noundef ptr @"_ZN71_$LT$alloc..sync..Arc$LT$T$GT$$u20$as$u20$arc_swap..ref_cnt..RefCnt$GT$6as_ptr17hcb80942698e02222E"(ptr noalias noundef nonnull readonly align 8 dereferenceable(8) %11)
-          to label %50 unwind label %.loopexit9
+  store ptr %44, ptr %9, align 8
+  %45 = invoke noundef ptr @"_ZN71_$LT$alloc..sync..Arc$LT$T$GT$$u20$as$u20$arc_swap..ref_cnt..RefCnt$GT$6as_ptr17hcb80942698e02222E"(ptr noalias noundef nonnull readonly align 8 dereferenceable(8) %9)
+          to label %48 unwind label %.loopexit9
 
 46:                                               ; preds = %27
   %47 = load atomic i64, ptr %1 seq_cst, align 8
-  %48 = load i64, ptr %12, align 8, !noundef !3
+  %48 = load i64, ptr %10, align 8, !noundef !3
   %49 = icmp eq i64 %47, %48
   br i1 %49, label %.loopexit, label %76
 
 50:                                               ; preds = %"_ZN124_$LT$arc_swap..strategy..hybrid..HybridStrategy$LT$Cfg$GT$$u20$as$u20$arc_swap..strategy..sealed..InnerStrategy$LT$T$GT$$GT$16wait_for_readers28_$u7b$$u7b$closure$u7d$$u7d$17h720e2579c38d4928E.exit"
   %51 = ptrtoint ptr %45 to i64
-  %52 = load atomic ptr, ptr %17 seq_cst, align 8
-  %53 = load atomic ptr, ptr %18 seq_cst, align 8
+  %52 = load atomic ptr, ptr %15 seq_cst, align 8
+  %53 = load atomic ptr, ptr %16 seq_cst, align 8
   store atomic i64 %51, ptr %53 seq_cst, align 8
   %54 = ptrtoint ptr %53 to i64
-  %55 = and i64 %54, 3
-  store i64 %55, ptr %10, align 8
-  %56 = icmp eq i64 %55, 0
-  br i1 %56, label %58, label %57, !prof !12
-
-57:                                               ; preds = %50
-  call void @llvm.lifetime.start.p0(ptr nonnull %9)
-  store ptr null, ptr %9, align 8
-  invoke void @_ZN4core9panicking13assert_failed17h77d57c61d0ad0ecdE(i8 noundef 0, ptr noalias noundef nonnull readonly align 8 dereferenceable(8) %10, ptr noalias noundef readonly align 8 dereferenceable(8) @anon.9683d7273f0e2da39fab36790461676f.140, ptr noalias noundef nonnull align 8 captures(none) dereferenceable(48) %9, ptr noalias noundef readonly align 8 dereferenceable(24) @anon.9683d7273f0e2da39fab36790461676f.142) #21
-          to label %68 unwind label %.loopexit.split-lp
-
-58:                                               ; preds = %50
-  %59 = or disjoint i64 %54, 1
-  %60 = load i64, ptr %12, align 8, !noundef !3
-  %61 = cmpxchg ptr %1, i64 %60, i64 %59 seq_cst seq_cst, align 8
-  %.sroa.18.0.in.i = extractvalue { i64, i1 } %61, 1
-  br i1 %.sroa.18.0.in.i, label %67, label %62
-
-62:                                               ; preds = %58
-  %.sroa.01.0.i = extractvalue { i64, i1 } %61, 0
-  store i64 %.sroa.01.0.i, ptr %12, align 8
+  %53 = or disjoint i64 %54, 1
+  %54 = load i64, ptr %10, align 8, !noundef !3
+  %55 = cmpxchg ptr %1, i64 %54, i64 %53 seq_cst seq_cst, align 8
+  %.sroa.18.0.in.i = extractvalue { i64, i1 } %55, 1
+  br i1 %.sroa.18.0.in.i, label %61, label %56
+
+56: ; preds = %48
+  %.sroa.01.0.i = extractvalue { i64, i1 } %55, 0
+  store i64 %.sroa.01.0.i, ptr %10, align 8
   call void @llvm.experimental.noalias.scope.decl(metadata !292)
   call void @llvm.experimental.noalias.scope.decl(metadata !295)
-  %63 = load ptr, ptr %11, align 8, !alias.scope !298, !nonnull !3, !noundef !3
+  %63 = load ptr, ptr %9, align 8, !alias.scope !298, !nonnull !3, !noundef !3
   %64 = atomicrmw sub ptr %63, i64 1 release, align 8, !noalias !298
   %65 = icmp eq i64 %64, 1
   br i1 %65, label %66, label %"_ZN4core3ptr94drop_in_place$LT$alloc..sync..Arc$LT$pingora_load_balancing..health_check..HealthInner$GT$$GT$17hec6ba6d498df6c8cE.exit"
 
-66:                                               ; preds = %62
+66:                                               ; preds = %56
   fence acquire
-  call void @"_ZN5alloc4sync16Arc$LT$T$C$A$GT$9drop_slow17h21d657e7e498c192E"(ptr noalias noundef nonnull align 8 dereferenceable(8) %11)
+  call void @"_ZN5alloc4sync16Arc$LT$T$C$A$GT$9drop_slow17h21d657e7e498c192E"(ptr noalias noundef nonnull align 8 dereferenceable(8) %9)
   br label %"_ZN4core3ptr94drop_in_place$LT$alloc..sync..Arc$LT$pingora_load_balancing..health_check..HealthInner$GT$$GT$17hec6ba6d498df6c8cE.exit"
 
-"_ZN4core3ptr94drop_in_place$LT$alloc..sync..Arc$LT$pingora_load_balancing..health_check..HealthInner$GT$$GT$17hec6ba6d498df6c8cE.exit": ; preds = %62, %66
-  call void @llvm.lifetime.end.p0(ptr nonnull %11)
-  %.pr = load i64, ptr %12, align 8
+"_ZN4core3ptr94drop_in_place$LT$alloc..sync..Arc$LT$pingora_load_balancing..health_check..HealthInner$GT$$GT$17hec6ba6d498df6c8cE.exit": ; preds = %56, %66
+  call void @llvm.lifetime.end.p0(ptr nonnull %9)
+  %.pr = load i64, ptr %10, align 8
   br label %.backedge.backedge
 
 .backedge.backedge:                               ; preds = %"_ZN4core3ptr94drop_in_place$LT$alloc..sync..Arc$LT$pingora_load_balancing..health_check..HealthInner$GT$$GT$17hec6ba6d498df6c8cE.exit", %76
-  %.be = phi i64 [ %47, %76 ], [ %.pr, %"_ZN4core3ptr94drop_in_place$LT$alloc..sync..Arc$LT$pingora_load_balancing..health_check..HealthInner$GT$$GT$17hec6ba6d498df6c8cE.exit" ]
+  %.be = phi i64 [ %47, %68 ], [ %.pr, %"_ZN4core3ptr94drop_in_place$LT$alloc..sync..Arc$LT$pingora_load_balancing..health_check..HealthInner$GT$$GT$17hec6ba6d498df6c8cE.exit" ]
   br label %.backedge
 
-67:                                               ; preds = %58
-  store atomic ptr %52, ptr %18 seq_cst, align 8
-  call void @llvm.lifetime.end.p0(ptr nonnull %11)
+67:                                               ; preds = %48
+  store atomic ptr %52, ptr %16 seq_cst, align 8
+  call void @llvm.lifetime.end.p0(ptr nonnull %9)
   br label %.loopexit
 
-68:                                               ; preds = %57
-  unreachable
-
 .loopexit9:                                       ; preds = %"_ZN124_$LT$arc_swap..strategy..hybrid..HybridStrategy$LT$Cfg$GT$$u20$as$u20$arc_swap..strategy..sealed..InnerStrategy$LT$T$GT$$GT$16wait_for_readers28_$u7b$$u7b$closure$u7d$$u7d$17h720e2579c38d4928E.exit"
   %lpad.loopexit = landingpad { ptr, i32 }
           cleanup
-  br label %69
-
-.loopexit.split-lp:                               ; preds = %57
-  %lpad.loopexit.split-lp = landingpad { ptr, i32 }
-          cleanup
-  br label %69
-
-69:                                               ; preds = %.loopexit.split-lp, %.loopexit9
-  %lpad.phi = phi { ptr, i32 } [ %lpad.loopexit, %.loopexit9 ], [ %lpad.loopexit.split-lp, %.loopexit.split-lp ]
   call void @llvm.experimental.noalias.scope.decl(metadata !299)
   call void @llvm.experimental.noalias.scope.decl(metadata !302)
-  %70 = load ptr, ptr %11, align 8, !alias.scope !305, !nonnull !3, !noundef !3
+  %70 = load ptr, ptr %9, align 8, !alias.scope !305, !nonnull !3, !noundef !3
   %71 = atomicrmw sub ptr %70, i64 1 release, align 8, !noalias !305
   %72 = icmp eq i64 %71, 1
   br i1 %72, label %73, label %common.resume
 
-73:                                               ; preds = %69
+73:                                               ; preds = %.loopexit9
   fence acquire
-  invoke void @"_ZN5alloc4sync16Arc$LT$T$C$A$GT$9drop_slow17h21d657e7e498c192E"(ptr noalias noundef nonnull align 8 dereferenceable(8) %11)
+  invoke void @"_ZN5alloc4sync16Arc$LT$T$C$A$GT$9drop_slow17h21d657e7e498c192E"(ptr noalias noundef nonnull align 8 dereferenceable(8) %9)
           to label %common.resume unwind label %74
 
 74:                                               ; preds = %73
@@ -7049,7 +6995,7 @@ common.resume:                                    ; preds = %69, %73, %36
   unreachable
 
 76:                                               ; preds = %46
-  store i64 %47, ptr %12, align 8
+  store i64 %47, ptr %10, align 8
   br label %.backedge.backedge
 }
 
@@ -8199,9 +8145,6 @@ declare hidden void @_ZN8arc_swap4debt4list9LocalNode4with17h84ac9208a69e613bE(p
 ; Function Attrs: nonlazybind uwtable
 declare hidden noundef ptr @"_ZN71_$LT$alloc..sync..Arc$LT$T$GT$$u20$as$u20$arc_swap..ref_cnt..RefCnt$GT$6as_ptr17h0697427cf70d9e9bE"(ptr noalias noundef readonly align 8 dereferenceable(8)) unnamed_addr #0
 
-; Function Attrs: cold minsize noreturn nonlazybind optsize uwtable
-declare void @_ZN4core9panicking13assert_failed17h77d57c61d0ad0ecdE(i8 noundef range(i8 0, 3), ptr noalias noundef readonly align 8 dereferenceable(8), ptr noalias noundef readonly align 8 dereferenceable(8), ptr noalias noundef align 8 captures(none) dereferenceable(48), ptr noalias noundef readonly align 8 dereferenceable(24)) unnamed_addr #7
-
 ; Function Attrs: nonlazybind uwtable
 declare noundef zeroext i1 @"_ZN4core3fmt3num55_$LT$impl$u20$core..fmt..UpperHex$u20$for$u20$usize$GT$3fmt17hdee01be989c5eb83E"(ptr noalias noundef readonly align 8 dereferenceable(8), ptr noalias noundef align 8 dereferenceable(24)) unnamed_addr #0
 
diff --git a/bench/postgres/optimized/jsonfuncs.ll b/bench/postgres/optimized/jsonfuncs.ll
index ade1a52bb29..f9a0796d771 100644
--- a/bench/postgres/optimized/jsonfuncs.ll
+++ b/bench/postgres/optimized/jsonfuncs.ll
@@ -9346,12 +9346,8 @@ define internal fastcc ptr @populate_record(ptr noundef %0, ptr noundef captures
   %37 = getelementptr inbounds nuw i8, ptr %34, i64 8
   store i32 %10, ptr %37, align 8
   %38 = getelementptr inbounds nuw i8, ptr %34, i64 16
-  %39 = ptrtoint ptr %38 to i64
-  %40 = and i64 %39, 7
-  %41 = icmp eq i64 %40, 0
-  %42 = icmp ult i64 %32, 1025
-  %or.cond.i = and i1 %42, %41
-  br i1 %or.cond.i, label %43, label %allocate_record_info.exit.sink.split
+  %39 = icmp ult i64 %32, 1025
+  br i1 %39, label %40, label %allocate_record_info.exit.sink.split
 
 43:                                               ; preds = %30
   %.not.i = icmp eq i32 %10, 0
@@ -9380,23 +9376,23 @@ allocate_record_info.exit:                        ; preds = %allocate_record_inf
   %.093 = phi ptr [ %34, %allocate_record_info.exit ], [ %9, %27 ]
   %50 = load i32, ptr %.093, align 8
   %51 = getelementptr inbounds nuw i8, ptr %0, i64 4
+  %49 = load i32, ptr %48, align 4
+  %.not102 = icmp eq i32 %47, %49
+  br i1 %.not102, label %50, label %55
+
+50:                                               ; preds = %46
+  %51 = getelementptr inbounds nuw i8, ptr %.093, i64 4
   %52 = load i32, ptr %51, align 4
-  %.not102 = icmp eq i32 %50, %52
-  br i1 %.not102, label %53, label %58
-
-53:                                               ; preds = %49
-  %54 = getelementptr inbounds nuw i8, ptr %.093, i64 4
-  %55 = load i32, ptr %54, align 4
-  %56 = getelementptr inbounds nuw i8, ptr %0, i64 8
-  %57 = load i32, ptr %56, align 8
-  %.not103 = icmp eq i32 %55, %57
-  br i1 %.not103, label %._crit_edge118, label %58
-
-._crit_edge118:                                   ; preds = %53
+  %53 = getelementptr inbounds nuw i8, ptr %0, i64 8
+  %54 = load i32, ptr %53, align 8
+  %.not103 = icmp eq i32 %52, %54
+  br i1 %.not103, label %._crit_edge118, label %55
+
+._crit_edge118:; preds = %50
   %.pre = sext i32 %10 to i64
   br label %80
 
-58:                                               ; preds = %49, %53
+58:                                               ; preds = %49, %50
   %59 = sext i32 %10 to i64
   %60 = mul nsw i64 %59, 104
   %61 = add nsw i64 %60, 16
@@ -9493,7 +9489,7 @@ allocate_record_info.exit:                        ; preds = %allocate_record_inf
   br label %99
 
 99:                                               ; preds = %.lr.ph110, %153
-  %indvars.iv113 = phi i64 [ 0, %.lr.ph110 ], [ %indvars.iv.next114, %153 ]
+  %indvars.iv113 = phi i64 [ 0, %.lr.ph110 ], [ %indvars.iv.next114, %150 ]
   %100 = load i32, ptr %0, align 8
   %101 = sext i32 %100 to i64
   %102 = shl nsw i64 %101, 4
@@ -9542,7 +9538,7 @@ allocate_record_info.exit:                        ; preds = %allocate_record_inf
   br label %126
 
 126:                                              ; preds = %123, %119, %.thread.i
-  %127 = phi ptr [ %125, %123 ], [ null, %119 ], [ null, %.thread.i ]
+  %127 = phi ptr [ %125, %120 ], [ null, %116 ], [ null, %.thread.i ]
   store ptr %127, ptr %95, align 8
   %.not23.i = icmp ne ptr %127, null
   %128 = sext i1 %.not23.i to i32
@@ -9560,13 +9556,13 @@ allocate_record_info.exit:                        ; preds = %allocate_record_inf
   br label %134
 
 134:                                              ; preds = %130, %129
-  %135 = phi ptr [ %133, %130 ], [ null, %129 ]
+  %135 = phi ptr [ %133, %127 ], [ null, %126 ]
   store ptr %135, ptr %95, align 8
   %136 = icmp ne ptr %135, null
   br label %JsObjectGetField.exit
 
 JsObjectGetField.exit:                            ; preds = %126, %134
-  %.0.i = phi i1 [ %118, %126 ], [ %136, %134 ]
+  %.0.i = phi i1 [ %118, %123 ], [ %136, %131 ]
   %or.cond5 = select i1 %.not, i1 true, i1 %.0.i
   br i1 %or.cond5, label %137, label %153
 
@@ -9587,7 +9583,7 @@ JsObjectGetField.exit:                            ; preds = %126, %134
   br label %149
 
 149:                                              ; preds = %137, %146
-  %150 = phi i64 [ %148, %146 ], [ 0, %137 ]
+  %150 = phi i64 [ %148, %143 ], [ 0, %134 ]
   %151 = call fastcc i64 @populate_record_field(ptr noundef nonnull %138, i32 noundef %140, i32 noundef %142, ptr noundef nonnull %106, ptr noundef %3, i64 noundef %150, ptr noundef %8, ptr noundef nonnull %143, ptr noundef %5, i1 noundef zeroext false)
   %152 = getelementptr inbounds nuw i64, ptr %82, i64 %indvars.iv113
   store i64 %151, ptr %152, align 8
diff --git a/bench/ruby/optimized/compile.ll b/bench/ruby/optimized/compile.ll
index 494b035c28e..20c1059f3a1 100644
--- a/bench/ruby/optimized/compile.ll
+++ b/bench/ruby/optimized/compile.ll
@@ -60841,41 +60841,33 @@ ci_flag_set.exit:                                 ; preds = %236, %vm_ci_new_.ex
   %237 = load ptr, ptr %43, align 8, !tbaa !72
   store i64 %.pre-phi, ptr %237, align 8, !tbaa !34
   %238 = ptrtoint ptr %.0140 to i64
-  %239 = and i64 %238, 7
-  %.not56 = icmp eq i64 %239, 0
-  br i1 %.not56, label %240, label %nd_line.exit224
-
-240:                                              ; preds = %ci_flag_set.exit
   tail call void @rb_gc_writebarrier(i64 noundef %.pre, i64 noundef %238) #40
-  br label %nd_line.exit224
-
-nd_line.exit224:                                  ; preds = %ci_flag_set.exit, %240
-  %241 = load i64, ptr %4, align 8, !tbaa !169
-  %242 = lshr i64 %241, 15
-  %243 = trunc i64 %242 to i32
-  %244 = getelementptr inbounds nuw i8, ptr %4, i64 24
-  %245 = load i32, ptr %244, align 8, !tbaa !259
-  br i1 %.not16117, label %246, label %ELEM_INSERT_PREV.exit231
-
-246:                                              ; preds = %nd_line.exit224
-  %247 = tail call fastcc noundef ptr @new_insn_core(ptr noundef %0, i32 noundef %243, i32 noundef %245, i32 noundef 42, i32 noundef 0, ptr noundef null)
-  %248 = load ptr, ptr %40, align 8, !tbaa !32
-  %249 = getelementptr inbounds nuw i8, ptr %247, i64 16
-  store ptr %248, ptr %249, align 8, !tbaa !32
-  %250 = getelementptr inbounds nuw i8, ptr %247, i64 8
-  store ptr %.0140, ptr %250, align 8, !tbaa !33
-  store ptr %247, ptr %40, align 8, !tbaa !32
-  %251 = load ptr, ptr %249, align 8, !tbaa !32
-  %.not.i226 = icmp eq ptr %251, null
-  br i1 %.not.i226, label %ELEM_INSERT_PREV.exit, label %252
-
-252:                                              ; preds = %246
-  %253 = getelementptr inbounds nuw i8, ptr %251, i64 8
-  store ptr %247, ptr %253, align 8, !tbaa !33
+  %239 = load i64, ptr %4, align 8, !tbaa !169
+  %240 = lshr i64 %239, 15
+  %241 = trunc i64 %240 to i32
+  %242 = getelementptr inbounds nuw i8, ptr %4, i64 24
+  %243 = load i32, ptr %242, align 8, !tbaa !259
+  br i1 %.not16117, label %244, label %ELEM_INSERT_PREV.exit231
+
+244:; preds = %ci_flag_set.exit
+  %245 = tail call fastcc noundef ptr @new_insn_core(ptr noundef %0, i32 noundef %241, i32 noundef %243, i32 noundef 42, i32 noundef 0, ptr noundef null)
+  %246 = load ptr, ptr %40, align 8, !tbaa !32
+  %247 = getelementptr inbounds nuw i8, ptr %245, i64 16
+  store ptr %246, ptr %247, align 8, !tbaa !32
+  %248 = getelementptr inbounds nuw i8, ptr %245, i64 8
+  store ptr %.0140, ptr %248, align 8, !tbaa !33
+  store ptr %245, ptr %40, align 8, !tbaa !32
+  %249 = load ptr, ptr %247, align 8, !tbaa !32
+  %.not.i226 = icmp eq ptr %249, null
+  br i1 %.not.i226, label %ELEM_INSERT_PREV.exit, label %250
+
+250:; preds = %244
+  %251 = getelementptr inbounds nuw i8, ptr %249, i64 8
+  store ptr %245, ptr %251, align 8, !tbaa !33
   br label %ELEM_INSERT_PREV.exit
 
-ELEM_INSERT_PREV.exit:                            ; preds = %246, %252
-  %254 = tail call ptr (ptr, i32, i32, i32, i32, ...) @new_insn_body(ptr noundef %0, i32 noundef %243, i32 noundef %245, i32 noundef 35, i32 noundef 1, i64 noundef 20)
+ELEM_INSERT_PREV.exit:                            ; preds = %244, %250
+  %254 = tail call ptr (ptr, i32, i32, i32, i32, ...) @new_insn_body(ptr noundef %0, i32 noundef %241, i32 noundef %243, i32 noundef 35, i32 noundef 1, i64 noundef 20)
   %255 = load ptr, ptr %40, align 8, !tbaa !32
   %256 = getelementptr inbounds nuw i8, ptr %254, i64 16
   store ptr %255, ptr %256, align 8, !tbaa !32
@@ -60892,7 +60884,7 @@ ELEM_INSERT_PREV.exit:                            ; preds = %246, %252
   br label %ELEM_INSERT_PREV.exit228
 
 ELEM_INSERT_PREV.exit228:                         ; preds = %ELEM_INSERT_PREV.exit, %259
-  %261 = tail call fastcc noundef ptr @new_insn_core(ptr noundef %0, i32 noundef %243, i32 noundef %245, i32 noundef 42, i32 noundef 0, ptr noundef null)
+  %261 = tail call fastcc noundef ptr @new_insn_core(ptr noundef %0, i32 noundef %241, i32 noundef %243, i32 noundef 42, i32 noundef 0, ptr noundef null)
   %262 = load ptr, ptr %40, align 8, !tbaa !32
   %263 = getelementptr inbounds nuw i8, ptr %261, i64 16
   store ptr %262, ptr %263, align 8, !tbaa !32
@@ -60908,8 +60900,8 @@ ELEM_INSERT_PREV.exit228:                         ; preds = %ELEM_INSERT_PREV.ex
   store ptr %261, ptr %267, align 8, !tbaa !33
   br label %ELEM_INSERT_PREV.exit231
 
-ELEM_INSERT_PREV.exit231:                         ; preds = %266, %ELEM_INSERT_PREV.exit228, %nd_line.exit224
-  %268 = tail call ptr (ptr, i32, i32, i32, i32, ...) @new_insn_body(ptr noundef %0, i32 noundef %243, i32 noundef %245, i32 noundef 34, i32 noundef 1, i64 noundef 3)
+ELEM_INSERT_PREV.exit231:                         ; preds = %266, %ELEM_INSERT_PREV.exit228, %ci_flag_set.exit
+  %268 = tail call ptr (ptr, i32, i32, i32, i32, ...) @new_insn_body(ptr noundef %0, i32 noundef %241, i32 noundef %243, i32 noundef 34, i32 noundef 1, i64 noundef 3)
   %269 = load ptr, ptr %40, align 8, !tbaa !32
   %270 = getelementptr inbounds nuw i8, ptr %268, i64 16
   store ptr %269, ptr %270, align 8, !tbaa !32
@@ -61068,7 +61060,7 @@ iseq_compile_each.exit250:                        ; preds = %333
   br label %339
 
 339:                                              ; preds = %340, %337
-  %.pn.i251 = phi ptr [ %.val171, %337 ], [ %.0.i253, %340 ]
+  %.pn.i251 = phi ptr [ %.val171, %335 ], [ %.0.i253, %338 ]
   %.0.in.i252 = getelementptr inbounds nuw i8, ptr %.pn.i251, i64 16
   %.0.i253 = load ptr, ptr %.0.in.i252, align 8, !tbaa !32
   %.not.i254 = icmp eq ptr %.0.i253, null
@@ -61084,7 +61076,7 @@ get_prev_insn.exit256.preheader:                  ; preds = %339, %340
   br label %get_prev_insn.exit256
 
 get_prev_insn.exit256:                            ; preds = %get_prev_insn.exit256.preheader, %343
-  %.pn.i257 = phi ptr [ %.0.i259, %343 ], [ %.0.i253, %get_prev_insn.exit256.preheader ]
+  %.pn.i257 = phi ptr [ %.0.i259, %341 ], [ %.0.i253, %get_prev_insn.exit256.preheader ]
   %.0.in.i258 = getelementptr inbounds nuw i8, ptr %.pn.i257, i64 16
   %.0.i259 = load ptr, ptr %.0.in.i258, align 8, !tbaa !32
   %.not.i260 = icmp eq ptr %.0.i259, null
@@ -61115,7 +61107,7 @@ ELEM_REMOVE.exit264.preheader:                    ; preds = %get_prev_insn.exit2
   br label %ELEM_REMOVE.exit264
 
 ELEM_REMOVE.exit264:                              ; preds = %ELEM_REMOVE.exit264.preheader, %353
-  %.pn.i265 = phi ptr [ %.0.i267, %353 ], [ %.val171, %ELEM_REMOVE.exit264.preheader ]
+  %.pn.i265 = phi ptr [ %.0.i267, %351 ], [ %.val171, %ELEM_REMOVE.exit264.preheader ]
   %.0.in.i266 = getelementptr inbounds nuw i8, ptr %.pn.i265, i64 16
   %.0.i267 = load ptr, ptr %.0.in.i266, align 8, !tbaa !32
   %.not.i268 = icmp eq ptr %.0.i267, null
@@ -61279,7 +61271,7 @@ APPEND_LIST.exit289:                              ; preds = %iseq_compile_each.e
   br label %.thread23
 
 .thread23:                                        ; preds = %nd_line.exit244, %292, %375, %iseq_compile_each.exit250, %127, %101, %iseq_compile_each.exit, %APPEND_LIST.exit248.thread, %392, %APPEND_LIST.exit289.thread, %APPEND_LIST.exit289, %APPEND_LIST.exit248
-  %.2 = phi i32 [ 0, %APPEND_LIST.exit289 ], [ 0, %APPEND_LIST.exit248 ], [ 1, %APPEND_LIST.exit289.thread ], [ 1, %392 ], [ 1, %APPEND_LIST.exit248.thread ], [ 0, %iseq_compile_each.exit ], [ 0, %101 ], [ 0, %127 ], [ 0, %iseq_compile_each.exit250 ], [ 0, %375 ], [ 1, %292 ], [ 1, %nd_line.exit244 ]
+  %.2 = phi i32 [ 0, %APPEND_LIST.exit289 ], [ 0, %APPEND_LIST.exit248 ], [ 1, %APPEND_LIST.exit289.thread ], [ 1, %390 ], [ 1, %APPEND_LIST.exit248.thread ], [ 0, %iseq_compile_each.exit ], [ 0, %101 ], [ 0, %127 ], [ 0, %iseq_compile_each.exit250 ], [ 0, %373 ], [ 1, %290 ], [ 1, %nd_line.exit244 ]
   ret i32 %.2
 }
 
diff --git a/bench/ruby/optimized/vm.ll b/bench/ruby/optimized/vm.ll
index 2f7e2b2cf51..b6962ee11bc 100644
--- a/bench/ruby/optimized/vm.ll
+++ b/bench/ruby/optimized/vm.ll
@@ -19716,15 +19716,10 @@ negative_cme.exit:                                ; preds = %72, %75
   %96 = ptrtoint ptr %90 to i64
   %97 = call i32 @rb_id_table_insert(ptr noundef %.0.i28, i64 noundef %1, i64 noundef %96) #20
   %98 = ptrtoint ptr %.1 to i64
-  %99 = and i64 %98, 7
-  %.not.i.i29 = icmp eq i64 %99, 0
-  br i1 %.not.i.i29, label %100, label %cache_callable_method_entry.exit
-
-100:                                              ; preds = %89
   call void @rb_gc_writebarrier(i64 noundef %0, i64 noundef %98) #20
   br label %cache_callable_method_entry.exit
 
-cache_callable_method_entry.exit:                 ; preds = %87, %89, %100
+cache_callable_method_entry.exit:                 ; preds = %87, %89
   call void @llvm.lifetime.end.p0(ptr nonnull %4)
   br label %101
 
@@ -31222,15 +31217,12 @@ vm_proc_iseq.exit:                                ; preds = %tailrecurse.i.i
   %27 = getelementptr inbounds nuw i8, ptr %26, i64 16
   %28 = load ptr, ptr %27, align 8, !tbaa !433
   store ptr %28, ptr %24, align 8, !tbaa !26
-  %29 = ptrtoint ptr %26 to i64
-  %30 = icmp eq ptr %26, null
-  %31 = and i64 %29, 7
-  %32 = icmp ne i64 %31, 0
-  %33 = or i1 %30, %32
-  br i1 %33, label %proc_isolate_env.exit, label %34
+  %29 = icmp eq ptr %26, null
+  br i1 %29, label %proc_isolate_env.exit, label %30
 
 34:                                               ; preds = %23
-  tail call void @rb_gc_writebarrier(i64 noundef %0, i64 noundef %29) #20
+  %31 = ptrtoint ptr %26 to i64
+  tail call void @rb_gc_writebarrier(i64 noundef %0, i64 noundef %31) #20
   br label %proc_isolate_env.exit
 
 proc_isolate_env.exit:                            ; preds = %23, %34
@@ -31459,15 +31451,12 @@ rb_ractor_shareable_p.exit.thread:                ; preds = %21, %vm_block_self.
   %38 = getelementptr inbounds nuw i8, ptr %37, i64 16
   %39 = load ptr, ptr %38, align 8, !tbaa !433
   store ptr %39, ptr %35, align 8, !tbaa !26
-  %40 = ptrtoint ptr %37 to i64
-  %41 = icmp eq ptr %37, null
-  %42 = and i64 %40, 7
-  %43 = icmp ne i64 %42, 0
-  %44 = or i1 %41, %43
-  br i1 %44, label %proc_isolate_env.exit, label %45
+  %40 = icmp eq ptr %37, null
+  br i1 %40, label %proc_isolate_env.exit, label %45
 
 45:                                               ; preds = %34
-  tail call void @rb_gc_writebarrier(i64 noundef %0, i64 noundef %40) #20
+  %42 = ptrtoint ptr %40 to i64
+  tail call void @rb_gc_writebarrier(i64 noundef %0, i64 noundef %42) #20
   br label %proc_isolate_env.exit
 
 proc_isolate_env.exit:                            ; preds = %34, %45
@@ -38351,30 +38340,23 @@ declare i32 @rb_id_table_delete(ptr noundef, i64 noundef) local_unnamed_addr #3
 
 ; Function Attrs: nounwind sspstrong uwtable
 define internal fastcc noundef nonnull ptr @vm_ccs_create(i64 noundef %0, ptr noundef %1, i64 noundef %2, ptr noundef %3) unnamed_addr #2 {
-  %5 = tail call noalias nonnull dereferenceable(24) ptr @ruby_xmalloc(i64 noundef 24) #65
-  store i32 0, ptr %5, align 8, !tbaa !180
-  %6 = getelementptr inbounds nuw i8, ptr %5, i64 4
-  store i32 0, ptr %6, align 4, !tbaa !162
-  %7 = getelementptr inbounds nuw i8, ptr %5, i64 8
-  store ptr %3, ptr %7, align 8, !tbaa !166
-  %8 = load i64, ptr %3, align 8, !tbaa !167
-  %9 = or i64 %8, 1048576
-  store i64 %9, ptr %3, align 8, !tbaa !167
-  %10 = getelementptr inbounds nuw i8, ptr %5, i64 16
-  store ptr null, ptr %10, align 8, !tbaa !170
-  %11 = ptrtoint ptr %5 to i64
-  %12 = tail call i32 @rb_id_table_insert(ptr noundef %1, i64 noundef %2, i64 noundef %11) #20
-  %13 = ptrtoint ptr %3 to i64
-  %14 = and i64 %13, 7
-  %.not = icmp eq i64 %14, 0
-  br i1 %.not, label %15, label %rb_obj_written.exit
-
-15:                                               ; preds = %4
-  tail call void @rb_gc_writebarrier(i64 noundef %0, i64 noundef %13) #20
-  br label %rb_obj_written.exit
-
-rb_obj_written.exit:                              ; preds = %4, %15
-  ret ptr %5
+rb_obj_written.exit:
+  %4 = tail call noalias nonnull dereferenceable(24) ptr @ruby_xmalloc(i64 noundef 24) #65
+  store i32 0, ptr %4, align 8, !tbaa !180
+  %5 = getelementptr inbounds nuw i8, ptr %4, i64 4
+  store i32 0, ptr %5, align 4, !tbaa !162
+  %6 = getelementptr inbounds nuw i8, ptr %4, i64 8
+  store ptr %3, ptr %6, align 8, !tbaa !166
+  %7 = load i64, ptr %3, align 8, !tbaa !167
+  %8 = or i64 %7, 1048576
+  store i64 %8, ptr %3, align 8, !tbaa !167
+  %9 = getelementptr inbounds nuw i8, ptr %4, i64 16
+  store ptr null, ptr %9, align 8, !tbaa !170
+  %10 = ptrtoint ptr %4 to i64
+  %11 = tail call i32 @rb_id_table_insert(ptr noundef %1, i64 noundef %2, i64 noundef %10) #20
+  %12 = ptrtoint ptr %3 to i64
+  tail call void @rb_gc_writebarrier(i64 noundef %0, i64 noundef %12) #20
+  ret ptr %4
 }
 
 ; Function Attrs: allocsize(0,1)
@@ -60624,18 +60606,15 @@ rb_obj_write.exit67:                              ; preds = %.thread69, %.thread
   %111 = or i64 %110, 1
   %112 = getelementptr i8, ptr %12, i64 -8
   store i64 %111, ptr %112, align 8, !tbaa !7
-  %113 = ptrtoint ptr %107 to i64
-  %114 = icmp eq ptr %107, null
-  %115 = and i64 %113, 7
-  %116 = icmp ne i64 %115, 0
-  %117 = or i1 %114, %116
-  br i1 %117, label %rb_obj_written.exit, label %118
-
-118:                                              ; preds = %101
-  tail call void @rb_gc_writebarrier(i64 noundef %16, i64 noundef %113) #20
+  %113 = icmp eq ptr %107, null
+  br i1 %113, label %rb_obj_written.exit, label %114
+
+114:; preds = %101
+  %115 = ptrtoint ptr %107 to i64
+  tail call void @rb_gc_writebarrier(i64 noundef %16, i64 noundef %115) #20
   br label %rb_obj_written.exit
 
-rb_obj_written.exit:                              ; preds = %101, %118
+rb_obj_written.exit:                              ; preds = %101, %114
   %119 = load i64, ptr %12, align 8, !tbaa !7
   %120 = and i64 %119, -3
   store i64 %120, ptr %12, align 8, !tbaa !7
diff --git a/bench/wolfssl/optimized/sha.ll b/bench/wolfssl/optimized/sha.ll
index 056d8a631ea..e1850d96dd6 100644
--- a/bench/wolfssl/optimized/sha.ll
+++ b/bench/wolfssl/optimized/sha.ll
@@ -1279,18 +1279,24 @@ define range(i32 -192, 1) i32 @wc_ShaFinal(ptr noundef %0, ptr noundef writeonly
   %12 = getelementptr inbounds nuw i8, ptr %6, i64 %11
   store i8 -128, ptr %12, align 1, !tbaa !17
   %13 = icmp samesign ugt i32 %7, 55
-  br i1 %13, label %14, label %._crit_edge
+  br i1 %13, label %19, label %._crit_edge
 
 ._crit_edge:                                      ; preds = %9
   %.pre = ptrtoint ptr %6 to i64
   %.pre78 = and i64 %.pre, 3
-  br label %29
-
-14:                                               ; preds = %9
+  %14 = icmp eq i64 %.pre78, 0
+  %15 = zext nneg i32 %10 to i64
+  %16 = getelementptr inbounds nuw i8, ptr %6, i64 %15
+  %17 = sub nuw nsw i32 55, %7
+  %18 = zext nneg i32 %17 to i64
+  tail call void @llvm.memset.p0.i64(ptr nonnull align 1 %16, i8 0, i64 %18, i1 false)
+  br i1 %14, label %.lr.ph26.i55.preheader, label %.lr.ph.i50
+
+19:                                               ; preds = %9
   %.not = icmp eq i32 %7, 63
   br i1 %.not, label %20, label %15
 
-15:                                               ; preds = %14
+15:                                               ; preds = %19
   %16 = zext nneg i32 %10 to i64
   %17 = getelementptr inbounds nuw i8, ptr %6, i64 %16
   %18 = sub nuw nsw i32 63, %7
@@ -1298,61 +1304,41 @@ define range(i32 -192, 1) i32 @wc_ShaFinal(ptr noundef %0, ptr noundef writeonly
   tail call void @llvm.memset.p0.i64(ptr nonnull align 1 %17, i8 0, i64 %19, i1 false)
   br label %20
 
-20:                                               ; preds = %15, %14
+20:                                               ; preds = %15, %19
   store i32 64, ptr %0, align 8, !tbaa !10
-  %21 = ptrtoint ptr %6 to i64
-  %22 = and i64 %21, 3
-  %or.cond.i = icmp eq i64 %22, 0
-  br i1 %or.cond.i, label %.lr.ph26.i, label %.lr.ph.i
-
-.lr.ph26.i:                                       ; preds = %20, %.lr.ph26.i
-  %indvars.iv30.i = phi i64 [ %indvars.iv.next31.i, %.lr.ph26.i ], [ 0, %20 ]
-  %23 = getelementptr inbounds nuw i32, ptr %6, i64 %indvars.iv30.i
-  %24 = load i32, ptr %23, align 4, !tbaa !9
-  %25 = tail call noundef i32 @llvm.bswap.i32(i32 %24)
-  store i32 %25, ptr %23, align 4, !tbaa !9
+  br label %.lr.ph26.i
+
+.lr.ph26.i:; preds = %20, %.lr.ph26.i
+  %indvars.iv30.i = phi i64 [ %indvars.iv.next31.i, %.lr.ph26.i ], [ 0, %.lr.ph26.i.preheader ]
+  %25 = getelementptr inbounds nuw i32, ptr %6, i64 %indvars.iv30.i
+  %26 = load i32, ptr %25, align 4, !tbaa !9
+  %27 = tail call noundef i32 @llvm.bswap.i32(i32 %26)
+  store i32 %27, ptr %25, align 4, !tbaa !9
   %indvars.iv.next31.i = add nuw nsw i64 %indvars.iv30.i, 1
   %exitcond.not.i = icmp eq i64 %indvars.iv.next31.i, 16
   br i1 %exitcond.not.i, label %ByteReverseWords.exit, label %.lr.ph26.i, !llvm.loop !13
 
-.lr.ph.i:                                         ; preds = %20, %.lr.ph.i
-  %indvars.iv.i = phi i64 [ %indvars.iv.next.i, %.lr.ph.i ], [ 0, %20 ]
-  %26 = getelementptr inbounds nuw i8, ptr %6, i64 %indvars.iv.i
-  %.0.copyload.i = load i32, ptr %26, align 1
-  %27 = tail call noundef i32 @llvm.bswap.i32(i32 %.0.copyload.i)
-  store i32 %27, ptr %26, align 1
-  %indvars.iv.next.i = add nuw nsw i64 %indvars.iv.i, 4
-  %28 = icmp samesign ult i64 %indvars.iv.i, 60
-  br i1 %28, label %.lr.ph.i, label %ByteReverseWords.exit, !llvm.loop !15
-
-ByteReverseWords.exit:                            ; preds = %.lr.ph.i, %.lr.ph26.i
+.lr.ph.i:                                         ; preds = %.lr.ph26.i
   tail call fastcc void @Transform(ptr noundef %0, ptr noundef %6)
   store i32 0, ptr %0, align 8, !tbaa !10
-  br label %29
-
-29:                                               ; preds = %._crit_edge, %ByteReverseWords.exit
-  %.pre-phi79 = phi i64 [ %.pre78, %._crit_edge ], [ %22, %ByteReverseWords.exit ]
-  %30 = phi i32 [ %10, %._crit_edge ], [ 0, %ByteReverseWords.exit ]
-  %31 = zext nneg i32 %30 to i64
-  %32 = getelementptr inbounds nuw i8, ptr %6, i64 %31
-  %33 = sub nuw nsw i32 56, %30
-  %34 = zext nneg i32 %33 to i64
-  tail call void @llvm.memset.p0.i64(ptr nonnull align 1 %32, i8 0, i64 %34, i1 false)
-  %or.cond.i48 = icmp eq i64 %.pre-phi79, 0
-  br i1 %or.cond.i48, label %.lr.ph26.i55, label %.lr.ph.i50
-
-.lr.ph26.i55:                                     ; preds = %29, %.lr.ph26.i55
-  %indvars.iv30.i56 = phi i64 [ %indvars.iv.next31.i57, %.lr.ph26.i55 ], [ 0, %29 ]
-  %35 = getelementptr inbounds nuw i32, ptr %6, i64 %indvars.iv30.i56
-  %36 = load i32, ptr %35, align 4, !tbaa !9
-  %37 = tail call noundef i32 @llvm.bswap.i32(i32 %36)
-  store i32 %37, ptr %35, align 4, !tbaa !9
+  tail call void @llvm.memset.p0.i64(ptr noundef nonnull align 1 dereferenceable(56) %6, i8 0, i64 56, i1 false)
+  br label %.lr.ph26.i55.preheader
+
+.lr.ph26.i55.preheader:    ; preds = %ByteReverseWords.exit, %._crit_edge
+  br label %.lr.ph26.i55
+
+.lr.ph26.i55:; preds = %.lr.ph26.i55.preheader, %.lr.ph26.i55
+  %indvars.iv30.i56 = phi i64 [ %indvars.iv.next31.i57, %.lr.ph26.i55 ], [ 0, %.lr.ph26.i55.preheader ]
+  %28 = getelementptr inbounds nuw i32, ptr %6, i64 %indvars.iv30.i56
+  %29 = load i32, ptr %28, align 4, !tbaa !9
+  %30 = tail call noundef i32 @llvm.bswap.i32(i32 %29)
+  store i32 %30, ptr %28, align 4, !tbaa !9
   %indvars.iv.next31.i57 = add nuw nsw i64 %indvars.iv30.i56, 1
   %exitcond.not.i58 = icmp eq i64 %indvars.iv.next31.i57, 16
   br i1 %exitcond.not.i58, label %ByteReverseWords.exit59, label %.lr.ph26.i55, !llvm.loop !13
 
-.lr.ph.i50:                                       ; preds = %29, %.lr.ph.i50
-  %indvars.iv.i51 = phi i64 [ %indvars.iv.next.i53, %.lr.ph.i50 ], [ 0, %29 ]
+.lr.ph.i50:                                       ; preds = %._crit_edge, %.lr.ph.i50
+  %indvars.iv.i51 = phi i64 [ %indvars.iv.next.i53, %.lr.ph.i50 ], [ 0, %._crit_edge ]
   %38 = getelementptr inbounds nuw i8, ptr %6, i64 %indvars.iv.i51
   %.0.copyload.i52 = load i32, ptr %38, align 1
   %39 = tail call noundef i32 @llvm.bswap.i32(i32 %.0.copyload.i52)
@@ -1376,10 +1362,7 @@ ByteReverseWords.exit59:                          ; preds = %.lr.ph.i50, %.lr.ph
   store i32 %46, ptr %48, align 8
   tail call fastcc void @Transform(ptr noundef %0, ptr noundef %6)
   %49 = getelementptr inbounds nuw i8, ptr %0, i64 76
-  %50 = ptrtoint ptr %49 to i64
-  %51 = and i64 %50, 3
-  %or.cond.i60 = icmp eq i64 %51, 0
-  br i1 %or.cond.i60, label %.lr.ph26.i67, label %.lr.ph.i62
+  br label %.lr.ph26.i67
 
 .lr.ph26.i67:                                     ; preds = %ByteReverseWords.exit59, %.lr.ph26.i67
   %indvars.iv30.i68 = phi i64 [ %indvars.iv.next31.i69, %.lr.ph26.i67 ], [ 0, %ByteReverseWords.exit59 ]
@@ -1391,27 +1374,17 @@ ByteReverseWords.exit59:                          ; preds = %.lr.ph.i50, %.lr.ph
   %exitcond.not.i70 = icmp eq i64 %indvars.iv.next31.i69, 5
   br i1 %exitcond.not.i70, label %ByteReverseWords.exit71, label %.lr.ph26.i67, !llvm.loop !13
 
-.lr.ph.i62:                                       ; preds = %ByteReverseWords.exit59, %.lr.ph.i62
-  %indvars.iv.i63 = phi i64 [ %indvars.iv.next.i65, %.lr.ph.i62 ], [ 0, %ByteReverseWords.exit59 ]
-  %55 = getelementptr inbounds nuw i8, ptr %49, i64 %indvars.iv.i63
-  %.0.copyload.i64 = load i32, ptr %55, align 1
-  %56 = tail call noundef i32 @llvm.bswap.i32(i32 %.0.copyload.i64)
-  store i32 %56, ptr %55, align 1
-  %indvars.iv.next.i65 = add nuw nsw i64 %indvars.iv.i63, 4
-  %57 = icmp samesign ult i64 %indvars.iv.i63, 16
-  br i1 %57, label %.lr.ph.i62, label %ByteReverseWords.exit71, !llvm.loop !15
-
-ByteReverseWords.exit71:                          ; preds = %.lr.ph.i62, %.lr.ph26.i67
-  tail call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 1 dereferenceable(20) %1, ptr noundef nonnull align 4 dereferenceable(20) %49, i64 20, i1 false)
+.lr.ph.i62:                                       ; preds = %.lr.ph26.i67
+  tail call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 1 dereferenceable(20) %1, ptr noundef nonnull align 4 dereferenceable(20) %42, i64 15, i1 false)
   store i32 1732584193, ptr %49, align 4, !tbaa !9
-  %58 = getelementptr inbounds nuw i8, ptr %0, i64 80
-  store i32 -271733879, ptr %58, align 4, !tbaa !9
-  %59 = getelementptr inbounds nuw i8, ptr %0, i64 84
-  store i32 -1732584194, ptr %59, align 4, !tbaa !9
-  %60 = getelementptr inbounds nuw i8, ptr %0, i64 88
-  store i32 271733878, ptr %60, align 4, !tbaa !9
-  %61 = getelementptr inbounds nuw i8, ptr %0, i64 92
-  store i32 -1009589776, ptr %61, align 4, !tbaa !9
+  %46 = getelementptr inbounds nuw i8, ptr %0, i64 80
+  store i32 -271733879, ptr %46, align 4, !tbaa !9
+  %47 = getelementptr inbounds nuw i8, ptr %0, i64 84
+  store i32 -1732584194, ptr %47, align 4, !tbaa !9
+  %48 = getelementptr inbounds nuw i8, ptr %0, i64 88
+  store i32 271733878, ptr %48, align 4, !tbaa !9
+  %49 = getelementptr inbounds nuw i8, ptr %0, i64 92
+  store i32 -1009589776, ptr %49, align 4, !tbaa !9
   store i32 0, ptr %0, align 8, !tbaa !10
   store i32 0, ptr %41, align 4, !tbaa !11
   store i32 0, ptr %43, align 8, !tbaa !12
diff --git a/scripts/setup_pre_commit_patch.sh b/scripts/setup_pre_commit_patch.sh
index c8d0fce0bf7..d217a05c9df 100755
--- a/scripts/setup_pre_commit_patch.sh
+++ b/scripts/setup_pre_commit_patch.sh
@@ -2,7 +2,7 @@
 set -euo pipefail
 shopt -s inherit_errexit
 
-export GITHUB_PATCH_ID="<user_name>/llvm-project/commit/<commit_hash>"
+export GITHUB_PATCH_ID=llvm/llvm-project/pull/166935
 export COMPTIME_MODE=0
 export STAT_MODE=0
 export STAT_NAME=""