diff --git a/bench/abc/optimized/giaSif.ll b/bench/abc/optimized/giaSif.ll
index d72d00db7a1..7ff972a5048 100644
--- a/bench/abc/optimized/giaSif.ll
+++ b/bench/abc/optimized/giaSif.ll
@@ -5412,8 +5412,8 @@ define void @Gia_ManSifPrintTimes(ptr noundef readonly captures(none) %0, ptr no
   %21 = zext i1 %20 to i32
   %22 = add i32 %18, -65536
   %23 = add i32 %22, %21
-  %24 = tail call range(i32 -2147483648, 536870913) i32 @llvm.smin.i32(i32 %23, i32 255)
-  %25 = tail call noundef i32 @llvm.smax.i32(i32 %24, i32 -255)
+  %24 = tail call i32 @llvm.smax.i32(i32 %23, i32 -255)
+  %25 = tail call i32 @llvm.smin.i32(i32 %24, i32 255)
   %26 = icmp sgt i32 %23, -1
   %27 = sub nsw i32 0, %25
   %.sink = select i1 %26, i32 %25, i32 %27
diff --git a/bench/abc/optimized/giaUtil.ll b/bench/abc/optimized/giaUtil.ll
index f3dc18e1c2b..25a423c7d0d 100644
--- a/bench/abc/optimized/giaUtil.ll
+++ b/bench/abc/optimized/giaUtil.ll
@@ -19322,8 +19322,8 @@ Vec_WrdStartTruthTables.exit:                     ; preds = %..loopexit28_crit_e
   br i1 %58, label %.lr.ph.preheader.i.us, label %Abc_TtFill.exit
 
 .lr.ph.preheader.i.us:                            ; preds = %.lr.ph95, %Abc_TtPrintBinary1.exit78.us
-  %.val54.val94.us = phi i32 [ %.val54.val.us, %Abc_TtPrintBinary1.exit78.us ], [ %.val54.val91, %.lr.ph95 ]
-  %.192.us = phi i32 [ %85, %Abc_TtPrintBinary1.exit78.us ], [ 0, %.lr.ph95 ]
+  %.val54.val94.us = phi i32 [ %.val54.val.us, %Abc_TtPrintBinary1.exit77.us ], [ %.val54.val91, %.lr.ph94 ]
+  %.192.us = phi i32 [ %85, %Abc_TtPrintBinary1.exit77.us ], [ 0, %.lr.ph94 ]
   tail call void @llvm.memset.p0.i64(ptr align 8 %50, i8 -1, i64 %60, i1 false), !tbaa !205
   %62 = icmp sgt i32 %.val54.val94.us, 0
   br i1 %62, label %.lr.ph89.us, label %._crit_edge.split.us.us
@@ -19349,11 +19349,11 @@ Vec_WrdStartTruthTables.exit:                     ; preds = %..loopexit28_crit_e
   br label %.preheader.us.i73.us
 
 .preheader.us.i73.us:                             ; preds = %._crit_edge.us.i77.us, %.preheader.us.preheader.i71.us
-  %.014.us.i74.us = phi ptr [ %83, %._crit_edge.us.i77.us ], [ %50, %.preheader.us.preheader.i71.us ]
+  %.014.us.i74.us = phi ptr [ %83, %._crit_edge.us.i76.us ], [ %50, %.preheader.us.preheader.i71.us ]
   br label %73
 
 73:                                               ; preds = %73, %.preheader.us.i73.us
-  %.01213.us.i75.us = phi i32 [ 0, %.preheader.us.i73.us ], [ %82, %73 ]
+  %.01213.us.i75.us = phi i32 [ 0, %.preheader.us.i72.us ], [ %82, %73 ]
   %74 = lshr i32 %.01213.us.i75.us, 5
   %75 = zext nneg i32 %74 to i64
   %76 = getelementptr inbounds nuw i32, ptr %.014.us.i74.us, i64 %75
@@ -19387,7 +19387,7 @@ Abc_TtPrintBinary1.exit78.us:                     ; preds = %._crit_edge.us.i77.
   br label %89
 
 89:                                               ; preds = %Abc_TtAndCompl.exit.us.us, %.lr.ph89.us
-  %indvars.iv103 = phi i64 [ %indvars.iv.next104, %Abc_TtAndCompl.exit.us.us ], [ 0, %.lr.ph89.us ]
+  %indvars.iv103 = phi i64 [ %indvars.iv.next104, %Abc_TtAndCompl.exit.us.us ], [ 0, %.lr.ph88.us ]
   %90 = mul nuw nsw i64 %indvars.iv103, %59
   %91 = getelementptr inbounds nuw i64, ptr %.val63.us, i64 %90
   %92 = trunc nuw nsw i64 %indvars.iv103 to i32
@@ -19482,7 +19482,7 @@ Abc_TtPrintBinary1.exit:                          ; preds = %._crit_edge.us.i, %
   br i1 %131, label %.lr.ph, label %.preheader, !llvm.loop !268
 
 Abc_TtFill.exit:                                  ; preds = %.lr.ph95, %Abc_TtPrintBinary1.exit78
-  %.192 = phi i32 [ %154, %Abc_TtPrintBinary1.exit78 ], [ 0, %.lr.ph95 ]
+  %.192 = phi i32 [ %154, %Abc_TtPrintBinary1.exit77 ], [ 0, %.lr.ph94 ]
   %.val56 = load ptr, ptr %8, align 8, !tbaa !57
   %132 = getelementptr i8, ptr %.val56, i64 4
   %.val56.val = load i32, ptr %132, align 4, !tbaa !58
@@ -19503,11 +19503,11 @@ Abc_TtFill.exit:                                  ; preds = %.lr.ph95, %Abc_TtPr
   br label %.preheader.us.i73
 
 .preheader.us.i73:                                ; preds = %._crit_edge.us.i77, %.preheader.us.preheader.i71
-  %.014.us.i74 = phi ptr [ %152, %._crit_edge.us.i77 ], [ %50, %.preheader.us.preheader.i71 ]
+  %.014.us.i74 = phi ptr [ %152, %._crit_edge.us.i76 ], [ %50, %.preheader.us.preheader.i71 ]
   br label %142
 
 142:                                              ; preds = %142, %.preheader.us.i73
-  %.01213.us.i75 = phi i32 [ 0, %.preheader.us.i73 ], [ %151, %142 ]
+  %.01213.us.i75 = phi i32 [ 0, %.preheader.us.i72 ], [ %151, %142 ]
   %143 = lshr i32 %.01213.us.i75, 5
   %144 = zext nneg i32 %143 to i64
   %145 = getelementptr inbounds nuw i32, ptr %.014.us.i74, i64 %144
diff --git a/bench/abc/optimized/rsbDec6.ll b/bench/abc/optimized/rsbDec6.ll
index 8553333d1be..032cf0a12e3 100644
--- a/bench/abc/optimized/rsbDec6.ll
+++ b/bench/abc/optimized/rsbDec6.ll
@@ -3687,8 +3687,8 @@ define void @Rsb_DecPrintFunc(ptr noundef readonly captures(none) %0, i32 nounde
   br i1 %or.cond.i, label %Abc_TtPrintBinary.exit51.critedge, label %.preheader.us.preheader.i
 
 .preheader.us.preheader.i:                        ; preds = %6
-  %57 = tail call noundef range(i32 -2147483648, 65) i32 @llvm.smin.i32(i32 %13, i32 64)
-  %smax.i = tail call i32 @llvm.smax.i32(i32 %57, i32 1)
+  %57 = tail call i32 @llvm.smax.i32(i32 %13, i32 1)
+  %smax.i = tail call i32 @llvm.umin.i32(i32 %57, i32 64)
   br label %.preheader.us.i
 
 .preheader.us.i:                                  ; preds = %._crit_edge.us.i, %.preheader.us.preheader.i
@@ -3696,7 +3696,7 @@ define void @Rsb_DecPrintFunc(ptr noundef readonly captures(none) %0, i32 nounde
   br label %58
 
 58:                                               ; preds = %58, %.preheader.us.i
-  %.01112.us.i = phi i32 [ 0, %.preheader.us.i ], [ %67, %58 ]
+  %.01112.us.i = phi i32 [ 0, %.preheader.us.i ], [ %67, %59 ]
   %59 = lshr i32 %.01112.us.i, 5
   %60 = zext nneg i32 %59 to i64
   %61 = getelementptr inbounds nuw i32, ptr %.013.us.i, i64 %60
@@ -3720,11 +3720,11 @@ Abc_TtPrintBinary.exit:                           ; preds = %._crit_edge.us.i
   br label %.preheader.us.i45
 
 .preheader.us.i45:                                ; preds = %._crit_edge.us.i49, %Abc_TtPrintBinary.exit
-  %.013.us.i46 = phi ptr [ %81, %._crit_edge.us.i49 ], [ %7, %Abc_TtPrintBinary.exit ]
+  %.013.us.i46 = phi ptr [ %81, %._crit_edge.us.i48 ], [ %7, %Abc_TtPrintBinary.exit ]
   br label %71
 
 71:                                               ; preds = %71, %.preheader.us.i45
-  %.01112.us.i47 = phi i32 [ 0, %.preheader.us.i45 ], [ %80, %71 ]
+  %.01112.us.i47 = phi i32 [ 0, %.preheader.us.i44 ], [ %80, %72 ]
   %72 = lshr i32 %.01112.us.i47, 5
   %73 = zext nneg i32 %72 to i64
   %74 = getelementptr inbounds nuw i32, ptr %.013.us.i46, i64 %73
@@ -3754,16 +3754,16 @@ Abc_TtPrintBinary.exit51:                         ; preds = %._crit_edge.us.i49,
   br i1 %or.cond.i, label %Abc_TtPrintBinary.exit62, label %.preheader.us.preheader.i54
 
 .preheader.us.preheader.i54:                      ; preds = %Abc_TtPrintBinary.exit51
-  %86 = call noundef range(i32 -2147483648, 65) i32 @llvm.smin.i32(i32 %13, i32 64)
-  %smax.i55 = call i32 @llvm.smax.i32(i32 %86, i32 1)
+  %86 = call i32 @llvm.smax.i32(i32 %13, i32 1)
+  %smax.i55 = call i32 @llvm.umin.i32(i32 %86, i32 64)
   br label %.preheader.us.i56
 
 .preheader.us.i56:                                ; preds = %._crit_edge.us.i60, %.preheader.us.preheader.i54
-  %.013.us.i57 = phi ptr [ %97, %._crit_edge.us.i60 ], [ %8, %.preheader.us.preheader.i54 ]
+  %.013.us.i57 = phi ptr [ %97, %._crit_edge.us.i58 ], [ %8, %.preheader.us.preheader.i53 ]
   br label %87
 
 87:                                               ; preds = %87, %.preheader.us.i56
-  %.01112.us.i58 = phi i32 [ 0, %.preheader.us.i56 ], [ %96, %87 ]
+  %.01112.us.i58 = phi i32 [ 0, %.preheader.us.i54 ], [ %96, %89 ]
   %88 = lshr i32 %.01112.us.i58, 5
   %89 = zext nneg i32 %88 to i64
   %90 = getelementptr inbounds nuw i32, ptr %.013.us.i57, i64 %89
@@ -3803,11 +3803,11 @@ Abc_TtPrintBinary.exit62:                         ; preds = %._crit_edge.us.i60,
   br label %.preheader.us.i67
 
 .preheader.us.i67:                                ; preds = %._crit_edge.us.i71, %.preheader.us.preheader.i65
-  %.013.us.i68 = phi ptr [ %119, %._crit_edge.us.i71 ], [ %2, %.preheader.us.preheader.i65 ]
+  %.013.us.i68 = phi ptr [ %119, %._crit_edge.us.i68 ], [ %2, %.preheader.us.preheader.i63 ]
   br label %109
 
 109:                                              ; preds = %109, %.preheader.us.i67
-  %.01112.us.i69 = phi i32 [ 0, %.preheader.us.i67 ], [ %118, %109 ]
+  %.01112.us.i69 = phi i32 [ 0, %.preheader.us.i64 ], [ %118, %111 ]
   %110 = lshr i32 %.01112.us.i69, 5
   %111 = zext nneg i32 %110 to i64
   %112 = getelementptr inbounds nuw i32, ptr %.013.us.i68, i64 %111
@@ -3857,8 +3857,8 @@ Abc_TtPrintBinary.exit73:                         ; preds = %._crit_edge.us.i71,
   br i1 %129, label %.lr.ph78, label %._crit_edge79
 
 .lr.ph78:                                         ; preds = %._crit_edge, %.lr.ph78
-  %indvars.iv81 = phi i64 [ %indvars.iv.next82, %.lr.ph78 ], [ 0, %._crit_edge ]
-  %130 = phi ptr [ %135, %.lr.ph78 ], [ %127, %._crit_edge ]
+  %indvars.iv81 = phi i64 [ %indvars.iv.next82, %.lr.ph75 ], [ 0, %._crit_edge ]
+  %130 = phi ptr [ %135, %.lr.ph75 ], [ %127, %._crit_edge ]
   %131 = getelementptr i8, ptr %130, i64 8
   %.val32 = load ptr, ptr %131, align 8, !tbaa !25
   %132 = getelementptr inbounds nuw i32, ptr %.val32, i64 %indvars.iv81
diff --git a/bench/box2d/optimized/imgui_draw.ll b/bench/box2d/optimized/imgui_draw.ll
index 1e49b7f4a9b..adeecfbbc28 100644
--- a/bench/box2d/optimized/imgui_draw.ll
+++ b/bench/box2d/optimized/imgui_draw.ll
@@ -1378,8 +1378,8 @@ define dso_local void @_ZN20ImDrawListSharedData29SetCircleTessellationMaxErrorE
   %23 = add nsw i32 %22, 1
   %24 = sdiv i32 %23, 2
   %25 = shl nsw i32 %24, 1
-  %26 = tail call i32 @llvm.smin.i32(i32 range(i32 -2147483646, -2147483648) %25, i32 512)
-  %27 = tail call i32 @llvm.smax.i32(i32 %26, i32 4)
+  %26 = tail call i32 @llvm.smax.i32(i32 %25, i32 4)
+  %27 = tail call i32 @llvm.umin.i32(i32 %26, i32 512)
   %28 = trunc i32 %27 to i8
   br label %29
 
@@ -3336,8 +3336,8 @@ define dso_local noundef range(i32 0, 513) i32 @_ZNK10ImDrawList27_CalcCircleAut
   %24 = add nsw i32 %23, 1
   %25 = sdiv i32 %24, 2
   %26 = shl nsw i32 %25, 1
-  %27 = tail call i32 @llvm.smin.i32(i32 range(i32 -2147483646, -2147483648) %26, i32 512)
-  %28 = tail call i32 @llvm.smax.i32(i32 %27, i32 4)
+  %27 = tail call i32 @llvm.smax.i32(i32 %26, i32 4)
+  %28 = tail call i32 @llvm.umin.i32(i32 %27, i32 512)
   br label %29
 
 29:                                               ; preds = %13, %7
@@ -5306,8 +5306,8 @@ _ZN8ImVectorI6ImVec2E9push_backERKS0_.exit:       ; preds = %._ZN8ImVectorI6ImVe
   %63 = add nsw i32 %62, 1
   %64 = sdiv i32 %63, 2
   %65 = shl nsw i32 %64, 1
-  %66 = tail call i32 @llvm.smin.i32(i32 range(i32 -2147483646, -2147483648) %65, i32 512)
-  %67 = tail call i32 @llvm.smax.i32(i32 %66, i32 4)
+  %66 = tail call i32 @llvm.smax.i32(i32 %65, i32 4)
+  %67 = tail call i32 @llvm.umin.i32(i32 %66, i32 512)
   br label %_ZNK10ImDrawList27_CalcCircleAutoSegmentCountEf.exit
 
 _ZNK10ImDrawList27_CalcCircleAutoSegmentCountEf.exit: ; preds = %46, %52
@@ -6124,8 +6124,8 @@ _ZN8ImVectorI6ImVec2E9push_backERKS0_.exit81:     ; preds = %._ZN8ImVectorI6ImVe
   %217 = add nsw i32 %216, 1
   %218 = sdiv i32 %217, 2
   %219 = shl nsw i32 %218, 1
-  %220 = tail call i32 @llvm.smin.i32(i32 range(i32 -2147483646, -2147483648) %219, i32 512)
-  %221 = tail call i32 @llvm.smax.i32(i32 %220, i32 4)
+  %220 = tail call i32 @llvm.smax.i32(i32 %219, i32 4)
+  %221 = tail call i32 @llvm.umin.i32(i32 %220, i32 512)
   br label %_ZNK10ImDrawList27_CalcCircleAutoSegmentCountEf.exit
 
 _ZNK10ImDrawList27_CalcCircleAutoSegmentCountEf.exit: ; preds = %200, %206
@@ -6185,8 +6185,8 @@ define dso_local void @_ZN10ImDrawList19PathEllipticalArcToERK6ImVec2S2_fffi(ptr
   %36 = add nsw i32 %35, 1
   %37 = sdiv i32 %36, 2
   %38 = shl nsw i32 %37, 1
-  %39 = tail call i32 @llvm.smin.i32(i32 range(i32 -2147483646, -2147483648) %38, i32 512)
-  %40 = tail call i32 @llvm.smax.i32(i32 %39, i32 4)
+  %39 = tail call i32 @llvm.smax.i32(i32 %38, i32 4)
+  %40 = tail call i32 @llvm.umin.i32(i32 %39, i32 512)
   br label %_ZNK10ImDrawList27_CalcCircleAutoSegmentCountEf.exit
 
 _ZNK10ImDrawList27_CalcCircleAutoSegmentCountEf.exit: ; preds = %25, %19, %7
@@ -9004,8 +9004,8 @@ define dso_local void @_ZN10ImDrawList10AddEllipseERK6ImVec2S2_jfif(ptr noundef
   %38 = add nsw i32 %37, 1
   %39 = sdiv i32 %38, 2
   %40 = shl nsw i32 %39, 1
-  %41 = tail call i32 @llvm.smin.i32(i32 range(i32 -2147483646, -2147483648) %40, i32 512)
-  %42 = tail call i32 @llvm.smax.i32(i32 %41, i32 4)
+  %41 = tail call i32 @llvm.smax.i32(i32 %40, i32 4)
+  %42 = tail call i32 @llvm.umin.i32(i32 %41, i32 512)
   br label %_ZNK10ImDrawList27_CalcCircleAutoSegmentCountEf.exit
 
 _ZNK10ImDrawList27_CalcCircleAutoSegmentCountEf.exit: ; preds = %27, %21, %9
@@ -9072,8 +9072,8 @@ define dso_local void @_ZN10ImDrawList16AddEllipseFilledERK6ImVec2S2_jfi(ptr nou
   %37 = add nsw i32 %36, 1
   %38 = sdiv i32 %37, 2
   %39 = shl nsw i32 %38, 1
-  %40 = tail call i32 @llvm.smin.i32(i32 range(i32 -2147483646, -2147483648) %39, i32 512)
-  %41 = tail call i32 @llvm.smax.i32(i32 %40, i32 4)
+  %40 = tail call i32 @llvm.smax.i32(i32 %39, i32 4)
+  %41 = tail call i32 @llvm.umin.i32(i32 %40, i32 512)
   br label %_ZNK10ImDrawList27_CalcCircleAutoSegmentCountEf.exit
 
 _ZNK10ImDrawList27_CalcCircleAutoSegmentCountEf.exit: ; preds = %26, %20, %8
diff --git a/bench/bullet3/optimized/btMultiSphereShape.ll b/bench/bullet3/optimized/btMultiSphereShape.ll
index c54379b69f4..1b405fef106 100644
--- a/bench/bullet3/optimized/btMultiSphereShape.ll
+++ b/bench/bullet3/optimized/btMultiSphereShape.ll
@@ -377,8 +377,8 @@ define dso_local { <2 x float>, <2 x float> } @_ZNK18btMultiSphereShape37localGe
   %.026100 = phi i32 [ 0, %.lr.ph105 ], [ %82, %81 ]
   %.sroa.4.099 = phi <2 x float> [ zeroinitializer, %.lr.ph105 ], [ %.sroa.4.1, %81 ]
   %.sroa.089.098 = phi <2 x float> [ zeroinitializer, %.lr.ph105 ], [ %.sroa.089.1, %81 ]
-  %25 = tail call i32 @llvm.smin.i32(i32 %indvars.iv, i32 128)
-  %26 = tail call i32 @llvm.smax.i32(i32 %25, i32 1)
+  %25 = tail call i32 @llvm.smax.i32(i32 %indvars.iv, i32 1)
+  %26 = tail call i32 @llvm.umin.i32(i32 %25, i32 128)
   %smax = zext nneg i32 %26 to i64
   call void @llvm.lifetime.start.p0(i64 2048, ptr nonnull %3) #14
   %27 = sub nsw i32 %15, %.026100
@@ -538,8 +538,8 @@ define dso_local void @_ZNK18btMultiSphereShape49batchedUnitVectorGetSupportingV
   %.03081 = phi ptr [ %19, %.lr.ph84 ], [ %.131.lcssa, %90 ]
   %.03280 = phi i32 [ 0, %.lr.ph84 ], [ %91, %90 ]
   %.03379 = phi ptr [ %18, %.lr.ph84 ], [ %.134.lcssa, %90 ]
-  %24 = tail call i32 @llvm.smin.i32(i32 %indvars.iv, i32 128)
-  %25 = tail call i32 @llvm.smax.i32(i32 %24, i32 1)
+  %24 = tail call i32 @llvm.smax.i32(i32 %indvars.iv, i32 1)
+  %25 = tail call i32 @llvm.umin.i32(i32 %24, i32 128)
   %smax = zext nneg i32 %25 to i64
   call void @llvm.lifetime.start.p0(i64 2048, ptr nonnull %5) #14
   %26 = sub nsw i32 %15, %.03280
@@ -1038,6 +1038,9 @@ declare i32 @llvm.smin.i32(i32, i32) #13
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smax.i32(i32, i32) #13
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #13
+
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare float @llvm.sqrt.f32(float) #13
 
diff --git a/bench/ceres/optimized/gradient_checker.ll b/bench/ceres/optimized/gradient_checker.ll
index 3f3bc9327cd..3a150444eed 100644
--- a/bench/ceres/optimized/gradient_checker.ll
+++ b/bench/ceres/optimized/gradient_checker.ll
@@ -4846,55 +4846,55 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
 17:                                               ; preds = %_ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit
   %18 = add nsw i64 %13, -128
   %19 = sdiv i64 %18, 64
-  %.sroa.speculated157 = tail call i64 @llvm.smin.i64(i64 %19, i64 320)
-  %.sroa.speculated163 = tail call i64 @llvm.smax.i64(i64 %.sroa.speculated157, i64 8)
+  %.sroa.speculated157 = tail call i64 @llvm.smax.i64(i64 %19, i64 8)
+  %.sroa.speculated163 = tail call i64 @llvm.umin.i64(i64 %.sroa.speculated157, i64 320)
   %20 = load i64, ptr %0, align 8, !tbaa !119
   %21 = icmp slt i64 %.sroa.speculated163, %20
-  br i1 %21, label %22, label %24
+  br i1 %21, label %22, label %26
 
 22:                                               ; preds = %17
   %23 = and i64 %.sroa.speculated163, 504
   store i64 %23, ptr %0, align 8, !tbaa !119
-  br label %24
+  br label %26
 
-24:                                               ; preds = %22, %17
-  %25 = phi i64 [ %23, %22 ], [ %20, %17 ]
-  %26 = sub nsw i64 %14, %13
-  %27 = shl i64 %25, 5
-  %28 = udiv i64 %26, %27
-  %29 = load i64, ptr %2, align 8, !tbaa !119
-  %30 = add nsw i64 %3, -1
-  %31 = add i64 %30, %29
-  %32 = sdiv i64 %31, %3
-  %.not114 = icmp sgt i64 %28, %32
-  br i1 %.not114, label %36, label %33
-
-33:                                               ; preds = %24
-  %34 = srem i64 %28, 4
-  %35 = sub nsw i64 %28, %34
-  br label %40
-
-36:                                               ; preds = %24
-  %37 = add nsw i64 %32, 3
-  %38 = srem i64 %37, 4
-  %39 = sub nsw i64 %37, %38
-  %.sroa.speculated153 = tail call i64 @llvm.smin.i64(i64 %39, i64 %29)
-  br label %40
-
-40:                                               ; preds = %36, %33
-  %storemerge184 = phi i64 [ %35, %33 ], [ %.sroa.speculated153, %36 ]
+26:                                               ; preds = %24, %17
+  %27 = phi i64 [ %25, %24 ], [ %22, %17 ]
+  %28 = sub nsw i64 %14, %13
+  %29 = shl i64 %27, 5
+  %30 = udiv i64 %28, %29
+  %31 = load i64, ptr %2, align 8, !tbaa !119
+  %32 = add nsw i64 %3, -1
+  %33 = add i64 %32, %31
+  %34 = sdiv i64 %33, %3
+  %.not114 = icmp sgt i64 %30, %34
+  br i1 %.not114, label %38, label %35
+
+35:                                               ; preds = %26
+  %36 = srem i64 %30, 4
+  %37 = sub nsw i64 %30, %36
+  br label %42
+
+38:                                               ; preds = %26
+  %39 = add nsw i64 %34, 3
+  %40 = srem i64 %39, 4
+  %41 = sub nsw i64 %39, %40
+  %.sroa.speculated153 = tail call i64 @llvm.smin.i64(i64 %41, i64 %31)
+  br label %42
+
+24:                                               ; preds = %38, %35
+  %25 = phi i64 [ %37, %35 ], [ %.sroa.speculated153, %38 ]
   store i64 %storemerge184, ptr %2, align 8, !tbaa !119
   %41 = icmp sgt i64 %15, %14
   br i1 %41, label %42, label %.critedge116
 
-42:                                               ; preds = %40
+42:                                               ; preds = %42
   %43 = sub nsw i64 %15, %14
   %44 = load i64, ptr %0, align 8, !tbaa !119
   %45 = shl i64 %3, 3
   %46 = mul i64 %45, %44
   %47 = udiv i64 %43, %46
   %48 = load i64, ptr %1, align 8, !tbaa !119
-  %49 = add i64 %30, %48
+  %49 = add i64 %32, %48
   %50 = sdiv i64 %49, %3
   %51 = icmp slt i64 %47, %50
   %52 = icmp sgt i64 %47, 3
@@ -4921,147 +4921,147 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   %63 = load i64, ptr %0, align 8, !tbaa !119
   %.sroa.speculated145 = tail call i64 @llvm.smax.i64(i64 %63, i64 %62)
   %64 = icmp slt i64 %.sroa.speculated145, 48
-  br i1 %64, label %.critedge116, label %65
-
-65:                                               ; preds = %59
-  %66 = add nsw i64 %13, -128
-  %67 = sdiv i64 %66, 64
-  %68 = and i64 %67, -8
-  %.sroa.speculated140 = tail call i64 @llvm.smax.i64(i64 %68, i64 1)
-  %69 = icmp sgt i64 %63, %.sroa.speculated140
-  br i1 %69, label %70, label %84
-
-70:                                               ; preds = %65
-  %71 = urem i64 %63, %.sroa.speculated140
-  %72 = udiv i64 %63, %.sroa.speculated140
-  %73 = icmp eq i64 %71, 0
-  br i1 %73, label %82, label %74
-
-74:                                               ; preds = %70
-  %75 = xor i64 %71, -1
-  %76 = add nsw i64 %.sroa.speculated140, %75
-  %77 = shl i64 %72, 3
-  %78 = add i64 %77, 8
-  %79 = sdiv i64 %76, %78
-  %80 = shl nsw i64 %79, 3
-  %81 = sub nsw i64 %.sroa.speculated140, %80
-  br label %82
-
-82:                                               ; preds = %70, %74
-  %83 = phi i64 [ %81, %74 ], [ %.sroa.speculated140, %70 ]
-  store i64 %83, ptr %0, align 8, !tbaa !119
-  %.pre = load i64, ptr %1, align 8, !tbaa !119
+  br i1 %64, label %.critedge116, label %67
+
+82:                                               ; preds = %61
+  %68 = add nsw i64 %13, -128
+  %69 = sdiv i64 %68, 64
+  %70 = and i64 %69, -8
+  %.sroa.speculated140 = tail call i64 @llvm.smax.i64(i64 %70, i64 1)
+  %71 = icmp sgt i64 %65, %.sroa.speculated140
+  br i1 %71, label %72, label %86
+
+72:                                               ; preds = %67
+  %73 = urem i64 %65, %.sroa.speculated140
+  %74 = udiv i64 %65, %.sroa.speculated140
+  %75 = icmp eq i64 %73, 0
+  br i1 %75, label %84, label %76
+
+76:                                               ; preds = %72
+  %77 = xor i64 %73, -1
+  %78 = add nsw i64 %.sroa.speculated140, %77
+  %79 = shl i64 %74, 3
+  %80 = add i64 %79, 8
+  %81 = sdiv i64 %78, %80
+  %82 = shl nsw i64 %81, 3
+  %83 = sub nsw i64 %.sroa.speculated140, %82
   br label %84
 
-84:                                               ; preds = %82, %65
-  %85 = phi i64 [ %83, %82 ], [ %63, %65 ]
-  %86 = phi i64 [ %.pre, %82 ], [ %60, %65 ]
-  %87 = shl i64 %86, 3
-  %88 = mul i64 %87, %85
-  %89 = sub nsw i64 %66, %88
-  %90 = shl nsw i64 %85, 5
-  %.not = icmp slt i64 %89, %90
-  br i1 %.not, label %94, label %91
-
-91:                                               ; preds = %84
-  %92 = shl i64 %85, 3
-  %93 = udiv i64 %89, %92
-  br label %97
-
-94:                                               ; preds = %84
-  %95 = shl i64 %.sroa.speculated140, 5
-  %96 = udiv i64 4718592, %95
-  br label %97
-
-97:                                               ; preds = %94, %91
-  %storemerge = phi i64 [ %96, %94 ], [ %93, %91 ]
-  %98 = shl i64 %85, 4
-  %99 = udiv i64 1572864, %98
-  %.sroa.speculated135 = tail call i64 @llvm.smin.i64(i64 %storemerge, i64 %99)
-  %100 = and i64 %.sroa.speculated135, -4
-  %101 = load i64, ptr %2, align 8, !tbaa !119
-  %102 = icmp sgt i64 %101, %100
-  br i1 %102, label %103, label %116
-
-103:                                              ; preds = %97
-  %104 = srem i64 %101, %100
-  %105 = sdiv i64 %101, %100
-  %106 = icmp eq i64 %104, 0
-  br i1 %106, label %114, label %107
-
-107:                                              ; preds = %103
-  %108 = sub nsw i64 %100, %104
-  %109 = shl i64 %105, 2
-  %110 = add i64 %109, 4
-  %111 = sdiv i64 %108, %110
-  %112 = shl nsw i64 %111, 2
-  %113 = sub nsw i64 %100, %112
-  br label %114
-
-114:                                              ; preds = %103, %107
-  %115 = phi i64 [ %113, %107 ], [ %100, %103 ]
-  store i64 %115, ptr %2, align 8, !tbaa !119
-  br label %.critedge116
-
-116:                                              ; preds = %97
-  %117 = icmp eq i64 %63, %85
-  br i1 %117, label %118, label %.critedge116
-
-118:                                              ; preds = %116
-  %119 = shl i64 %63, 3
-  %120 = mul i64 %119, %101
-  %121 = icmp slt i64 %120, 1025
-  br i1 %121, label %126, label %122
+84:                                               ; preds = %72, %76
+  %85 = phi i64 [ %83, %76 ], [ %.sroa.speculated140, %72 ]
+  store i64 %85, ptr %0, align 8, !tbaa !119
+  %.pre = load i64, ptr %1, align 8, !tbaa !119
+  br label %86
 
-122:                                              ; preds = %118
-  %123 = icmp ne i64 %15, 0
-  %124 = icmp samesign ult i64 %120, 32769
-  %or.cond4 = and i1 %123, %124
-  br i1 %or.cond4, label %125, label %126
+122:                                              ; preds = %84, %67
+  %87 = phi i64 [ %85, %84 ], [ %65, %67 ]
+  %88 = phi i64 [ %.pre, %84 ], [ %60, %67 ]
+  %89 = shl i64 %88, 3
+  %90 = mul i64 %89, %87
+  %91 = sub nsw i64 %68, %90
+  %92 = shl nsw i64 %87, 5
+  %.not = icmp slt i64 %91, %92
+  br i1 %.not, label %96, label %93
+
+93:                                               ; preds = %86
+  %94 = shl i64 %87, 3
+  %95 = udiv i64 %91, %94
+  br label %99
 
 125:                                              ; preds = %122
-  %.sroa.speculated128 = tail call i64 @llvm.smin.i64(i64 %86, i64 576)
-  br label %126
+  %97 = shl i64 %.sroa.speculated140, 5
+  %98 = udiv i64 4718592, %97
+  br label %99
+
+99:                                               ; preds = %96, %93
+  %storemerge = phi i64 [ %98, %96 ], [ %95, %93 ]
+  %100 = shl i64 %87, 4
+  %101 = udiv i64 1572864, %100
+  %.sroa.speculated135 = tail call i64 @llvm.smin.i64(i64 %storemerge, i64 %101)
+  %102 = and i64 %.sroa.speculated135, -4
+  %103 = load i64, ptr %2, align 8, !tbaa !119
+  %104 = icmp sgt i64 %103, %102
+  br i1 %104, label %105, label %118
+
+105:                                              ; preds = %99
+  %106 = srem i64 %103, %102
+  %107 = sdiv i64 %103, %102
+  %108 = icmp eq i64 %106, 0
+  br i1 %108, label %116, label %109
+
+109:                                              ; preds = %105
+  %110 = sub nsw i64 %102, %106
+  %111 = shl i64 %107, 2
+  %112 = add i64 %111, 4
+  %113 = sdiv i64 %110, %112
+  %114 = shl nsw i64 %113, 2
+  %115 = sub nsw i64 %102, %114
+  br label %116
+
+116:                                              ; preds = %105, %109
+  %117 = phi i64 [ %115, %109 ], [ %102, %105 ]
+  store i64 %117, ptr %2, align 8, !tbaa !119
+  br label %.critedge116
 
-126:                                              ; preds = %118, %122, %125
-  %.0182 = phi i64 [ %.sroa.speculated128, %125 ], [ %86, %122 ], [ %86, %118 ]
-  %.093 = phi i64 [ %14, %125 ], [ 1572864, %122 ], [ %13, %118 ]
-  %127 = mul i64 %63, 24
-  %128 = udiv i64 %.093, %127
-  %.sroa.speculated = tail call i64 @llvm.smin.i64(i64 %.0182, i64 %128)
-  %129 = icmp sgt i64 %.sroa.speculated, 4
-  br i1 %129, label %130, label %132
-
-130:                                              ; preds = %126
-  %131 = and i64 %.sroa.speculated, 9223372036854775804
-  br label %134
-
-132:                                              ; preds = %126
-  %133 = icmp eq i64 %.sroa.speculated, 0
-  br i1 %133, label %.critedge116, label %134
-
-134:                                              ; preds = %132, %130
-  %.0 = phi i64 [ %131, %130 ], [ %.sroa.speculated, %132 ]
-  %135 = srem i64 %86, %.0
-  %136 = sdiv i64 %86, %.0
-  %137 = icmp eq i64 %135, 0
-  br i1 %137, label %.critedge, label %138
-
-138:                                              ; preds = %134
-  %139 = sub nsw i64 %.0, %135
-  %140 = shl i64 %136, 2
-  %141 = add i64 %140, 4
-  %142 = sdiv i64 %139, %141
-  %143 = shl nsw i64 %142, 2
-  %144 = sub nsw i64 %.0, %143
+126:                                              ; preds = %99
+  %119 = icmp eq i64 %65, %87
+  br i1 %119, label %120, label %.critedge116
+
+120:; preds = %126
+  %121 = shl i64 %65, 3
+  %122 = mul i64 %121, %103
+  %123 = icmp slt i64 %122, 1025
+  br i1 %123, label %128, label %124
+
+124:; preds = %120
+  %125 = icmp ne i64 %15, 0
+  %126 = icmp samesign ult i64 %122, 32769
+  %or.cond4 = and i1 %125, %126
+  br i1 %or.cond4, label %127, label %128
+
+127:                                              ; preds = %124
+  %.sroa.speculated128 = tail call i64 @llvm.smin.i64(i64 %88, i64 576)
+  br label %128
+
+128:; preds = %120, %124, %127
+  %.0182 = phi i64 [ %.sroa.speculated128, %127 ], [ %88, %124 ], [ %88, %120 ]
+  %.093 = phi i64 [ %14, %127 ], [ 1572864, %124 ], [ %13, %120 ]
+  %129 = mul i64 %65, 24
+  %130 = udiv i64 %.093, %129
+  %.sroa.speculated = tail call i64 @llvm.smin.i64(i64 %.0182, i64 %130)
+  %131 = icmp sgt i64 %.sroa.speculated, 4
+  br i1 %131, label %132, label %134
+
+132:; preds = %128
+  %133 = and i64 %.sroa.speculated, 9223372036854775804
+  br label %136
+
+134:                                              ; preds = %128
+  %135 = icmp eq i64 %.sroa.speculated, 0
+  br i1 %135, label %.critedge116, label %136
+
+136:                                              ; preds = %134, %132
+  %.0 = phi i64 [ %133, %132 ], [ %.sroa.speculated, %134 ]
+  %137 = srem i64 %88, %.0
+  %138 = sdiv i64 %88, %.0
+  %139 = icmp eq i64 %137, 0
+  br i1 %139, label %.critedge, label %140
+
+140:                                              ; preds = %136
+  %141 = sub nsw i64 %.0, %137
+  %142 = shl i64 %138, 2
+  %143 = add i64 %142, 4
+  %144 = sdiv i64 %141, %143
+  %145 = shl nsw i64 %144, 2
+  %146 = sub nsw i64 %.0, %145
   br label %.critedge
 
-.critedge:                                        ; preds = %134, %138
-  %145 = phi i64 [ %144, %138 ], [ %.0, %134 ]
+.critedge:                                        ; preds = %136, %140
+  %145 = phi i64 [ %146, %140 ], [ %.0, %136 ]
   store i64 %145, ptr %1, align 8, !tbaa !119
   br label %.critedge116
 
-.critedge116:                                     ; preds = %.critedge, %116, %114, %132, %53, %55, %40, %59
+.critedge116:                                     ; preds = %.critedge, %118, %116, %134, %53, %55, %42, %59
   ret void
 }
 
diff --git a/bench/ceres/optimized/manifold.ll b/bench/ceres/optimized/manifold.ll
index 83b1973cbe6..98d959548ec 100644
--- a/bench/ceres/optimized/manifold.ll
+++ b/bench/ceres/optimized/manifold.ll
@@ -3718,55 +3718,55 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
 17:                                               ; preds = %_ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit
   %18 = add nsw i64 %13, -128
   %19 = sdiv i64 %18, 64
-  %.sroa.speculated157 = tail call i64 @llvm.smin.i64(i64 %19, i64 320)
-  %.sroa.speculated163 = tail call i64 @llvm.smax.i64(i64 %.sroa.speculated157, i64 8)
+  %.sroa.speculated157 = tail call i64 @llvm.smax.i64(i64 %19, i64 8)
+  %.sroa.speculated163 = tail call i64 @llvm.umin.i64(i64 %.sroa.speculated157, i64 320)
   %20 = load i64, ptr %0, align 8, !tbaa !59
   %21 = icmp slt i64 %.sroa.speculated163, %20
-  br i1 %21, label %22, label %24
+  br i1 %21, label %22, label %26
 
 22:                                               ; preds = %17
   %23 = and i64 %.sroa.speculated163, 504
   store i64 %23, ptr %0, align 8, !tbaa !59
-  br label %24
-
-24:                                               ; preds = %22, %17
-  %25 = phi i64 [ %23, %22 ], [ %20, %17 ]
-  %26 = sub nsw i64 %14, %13
-  %27 = shl i64 %25, 5
-  %28 = udiv i64 %26, %27
-  %29 = load i64, ptr %2, align 8, !tbaa !59
-  %30 = add nsw i64 %3, -1
-  %31 = add i64 %30, %29
-  %32 = sdiv i64 %31, %3
-  %.not114 = icmp sgt i64 %28, %32
-  br i1 %.not114, label %36, label %33
-
-33:                                               ; preds = %24
-  %34 = srem i64 %28, 4
-  %35 = sub nsw i64 %28, %34
-  br label %40
-
-36:                                               ; preds = %24
-  %37 = add nsw i64 %32, 3
-  %38 = srem i64 %37, 4
-  %39 = sub nsw i64 %37, %38
-  %.sroa.speculated153 = tail call i64 @llvm.smin.i64(i64 %39, i64 %29)
-  br label %40
+  br label %26
 
-40:                                               ; preds = %36, %33
-  %storemerge184 = phi i64 [ %35, %33 ], [ %.sroa.speculated153, %36 ]
+36:                                               ; preds = %24, %17
+  %27 = phi i64 [ %23, %24 ], [ %22, %17 ]
+  %28 = sub nsw i64 %14, %13
+  %29 = shl i64 %27, 5
+  %30 = udiv i64 %28, %29
+  %31 = load i64, ptr %2, align 8, !tbaa !59
+  %32 = add nsw i64 %3, -1
+  %33 = add i64 %32, %31
+  %34 = sdiv i64 %33, %3
+  %.not114 = icmp sgt i64 %30, %34
+  br i1 %.not114, label %38, label %35
+
+35:                                               ; preds = %26
+  %36 = srem i64 %30, 4
+  %37 = sub nsw i64 %30, %36
+  br label %42
+
+40:                                               ; preds = %36
+  %39 = add nsw i64 %34, 3
+  %40 = srem i64 %39, 4
+  %41 = sub nsw i64 %39, %40
+  %.sroa.speculated153 = tail call i64 @llvm.smin.i64(i64 %41, i64 %31)
+  br label %42
+
+42:                                               ; preds = %38, %35
+  %storemerge184 = phi i64 [ %37, %35 ], [ %.sroa.speculated153, %38 ]
   store i64 %storemerge184, ptr %2, align 8, !tbaa !59
   %41 = icmp sgt i64 %15, %14
   br i1 %41, label %42, label %.critedge116
 
-42:                                               ; preds = %40
+42:                                               ; preds = %42
   %43 = sub nsw i64 %15, %14
   %44 = load i64, ptr %0, align 8, !tbaa !59
   %45 = shl i64 %3, 3
   %46 = mul i64 %45, %44
   %47 = udiv i64 %43, %46
   %48 = load i64, ptr %1, align 8, !tbaa !59
-  %49 = add i64 %30, %48
+  %49 = add i64 %32, %48
   %50 = sdiv i64 %49, %3
   %51 = icmp slt i64 %47, %50
   %52 = icmp sgt i64 %47, 3
@@ -3801,13 +3801,13 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   %68 = and i64 %67, -8
   %.sroa.speculated140 = tail call i64 @llvm.smax.i64(i64 %68, i64 1)
   %69 = icmp sgt i64 %63, %.sroa.speculated140
-  br i1 %69, label %70, label %84
+  br i1 %69, label %70, label %86
 
 70:                                               ; preds = %65
   %71 = urem i64 %63, %.sroa.speculated140
   %72 = udiv i64 %63, %.sroa.speculated140
   %73 = icmp eq i64 %71, 0
-  br i1 %73, label %82, label %74
+  br i1 %73, label %84, label %74
 
 74:                                               ; preds = %70
   %75 = xor i64 %71, -1
@@ -3817,71 +3817,71 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   %79 = sdiv i64 %76, %78
   %80 = shl nsw i64 %79, 3
   %81 = sub nsw i64 %.sroa.speculated140, %80
-  br label %82
-
-82:                                               ; preds = %70, %74
-  %83 = phi i64 [ %81, %74 ], [ %.sroa.speculated140, %70 ]
-  store i64 %83, ptr %0, align 8, !tbaa !59
-  %.pre = load i64, ptr %1, align 8, !tbaa !59
   br label %84
 
-84:                                               ; preds = %82, %65
-  %85 = phi i64 [ %83, %82 ], [ %63, %65 ]
-  %86 = phi i64 [ %.pre, %82 ], [ %60, %65 ]
-  %87 = shl i64 %86, 3
-  %88 = mul i64 %87, %85
-  %89 = sub nsw i64 %66, %88
-  %90 = shl nsw i64 %85, 5
-  %.not = icmp slt i64 %89, %90
-  br i1 %.not, label %94, label %91
-
-91:                                               ; preds = %84
-  %92 = shl i64 %85, 3
-  %93 = udiv i64 %89, %92
-  br label %97
-
-94:                                               ; preds = %84
-  %95 = shl i64 %.sroa.speculated140, 5
-  %96 = udiv i64 4718592, %95
-  br label %97
-
-97:                                               ; preds = %94, %91
-  %storemerge = phi i64 [ %96, %94 ], [ %93, %91 ]
-  %98 = shl i64 %85, 4
-  %99 = udiv i64 1572864, %98
-  %.sroa.speculated135 = tail call i64 @llvm.smin.i64(i64 %storemerge, i64 %99)
-  %100 = and i64 %.sroa.speculated135, -4
-  %101 = load i64, ptr %2, align 8, !tbaa !59
-  %102 = icmp sgt i64 %101, %100
-  br i1 %102, label %103, label %116
-
-103:                                              ; preds = %97
-  %104 = srem i64 %101, %100
-  %105 = sdiv i64 %101, %100
-  %106 = icmp eq i64 %104, 0
-  br i1 %106, label %114, label %107
-
-107:                                              ; preds = %103
-  %108 = sub nsw i64 %100, %104
-  %109 = shl i64 %105, 2
-  %110 = add i64 %109, 4
-  %111 = sdiv i64 %108, %110
-  %112 = shl nsw i64 %111, 2
-  %113 = sub nsw i64 %100, %112
-  br label %114
-
-114:                                              ; preds = %103, %107
-  %115 = phi i64 [ %113, %107 ], [ %100, %103 ]
-  store i64 %115, ptr %2, align 8, !tbaa !59
+84:                                               ; preds = %72, %76
+  %85 = phi i64 [ %83, %76 ], [ %.sroa.speculated138, %72 ]
+  store i64 %85, ptr %0, align 8, !tbaa !59
+  %.pre = load i64, ptr %1, align 8, !tbaa !59
+  br label %86
+
+86:; preds = %84, %65
+  %87 = phi i64 [ %85, %84 ], [ %65, %67 ]
+  %88 = phi i64 [ %.pre, %84 ], [ %62, %67 ]
+  %89 = shl i64 %88, 3
+  %90 = mul i64 %89, %87
+  %91 = sub nsw i64 %68, %90
+  %92 = shl nsw i64 %87, 5
+  %.not = icmp slt i64 %91, %92
+  br i1 %.not, label %96, label %93
+
+93:; preds = %86
+  %94 = shl i64 %87, 3
+  %95 = udiv i64 %91, %94
+  br label %99
+
+96:; preds = %86
+  %98 = shl i64 %.sroa.speculated138, 5
+  %99 = udiv i64 4718592, %98
+  br label %99
+
+99:; preds = %96, %93
+  %storemerge = phi i64 [ %98, %96 ], [ %95, %93 ]
+  %100 = shl i64 %87, 4
+  %101 = udiv i64 1572864, %100
+  %.sroa.speculated135 = tail call i64 @llvm.smin.i64(i64 %storemerge, i64 %101)
+  %102 = and i64 %.sroa.speculated135, -4
+  %103 = load i64, ptr %2, align 8, !tbaa !59
+  %106 = icmp sgt i64 %103, %102
+  br i1 %106, label %105, label %118
+
+107:                                              ; preds = %99
+  %106 = srem i64 %103, %102
+  %107 = sdiv i64 %103, %102
+  %108 = icmp eq i64 %106, 0
+  br i1 %108, label %116, label %109
+
+109:; preds = %105
+  %110 = sub nsw i64 %102, %106
+  %111 = shl i64 %107, 2
+  %112 = add i64 %111, 4
+  %113 = sdiv i64 %110, %112
+  %114 = shl nsw i64 %113, 2
+  %115 = sub nsw i64 %102, %114
+  br label %116
+
+116:                                              ; preds = %105, %109
+  %117 = phi i64 [ %115, %109 ], [ %102, %105 ]
+  store i64 %117, ptr %2, align 8, !tbaa !59
   br label %.critedge116
 
-116:                                              ; preds = %97
-  %117 = icmp eq i64 %63, %85
+116:                                              ; preds = %99
+  %117 = icmp eq i64 %63, %87
   br i1 %117, label %118, label %.critedge116
 
 118:                                              ; preds = %116
   %119 = shl i64 %63, 3
-  %120 = mul i64 %119, %101
+  %120 = mul i64 %119, %103
   %121 = icmp slt i64 %120, 1025
   br i1 %121, label %126, label %122
 
@@ -3892,12 +3892,12 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br i1 %or.cond4, label %125, label %126
 
 125:                                              ; preds = %122
-  %.sroa.speculated128 = tail call i64 @llvm.smin.i64(i64 %86, i64 576)
+  %.sroa.speculated128 = tail call i64 @llvm.smin.i64(i64 %88, i64 576)
   br label %126
 
 126:                                              ; preds = %118, %122, %125
-  %.0182 = phi i64 [ %.sroa.speculated128, %125 ], [ %86, %122 ], [ %86, %118 ]
-  %.093 = phi i64 [ %14, %125 ], [ 1572864, %122 ], [ %13, %118 ]
+  %.0182 = phi i64 [ %.sroa.speculated128, %127 ], [ %88, %124 ], [ %88, %120 ]
+  %.093 = phi i64 [ %14, %127 ], [ 1572864, %124 ], [ %13, %120 ]
   %127 = mul i64 %63, 24
   %128 = udiv i64 %.093, %127
   %.sroa.speculated = tail call i64 @llvm.smin.i64(i64 %.0182, i64 %128)
@@ -3913,9 +3913,9 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br i1 %133, label %.critedge116, label %134
 
 134:                                              ; preds = %132, %130
-  %.0 = phi i64 [ %131, %130 ], [ %.sroa.speculated, %132 ]
-  %135 = srem i64 %86, %.0
-  %136 = sdiv i64 %86, %.0
+  %.0 = phi i64 [ %131, %132 ], [ %.sroa.speculated, %134 ]
+  %135 = srem i64 %88, %.0
+  %136 = sdiv i64 %88, %.0
   %137 = icmp eq i64 %135, 0
   br i1 %137, label %.critedge, label %138
 
@@ -3929,11 +3929,11 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br label %.critedge
 
 .critedge:                                        ; preds = %134, %138
-  %145 = phi i64 [ %144, %138 ], [ %.0, %134 ]
+  %145 = phi i64 [ %144, %140 ], [ %.0, %136 ]
   store i64 %145, ptr %1, align 8, !tbaa !59
   br label %.critedge116
 
-.critedge116:                                     ; preds = %.critedge, %116, %114, %132, %53, %55, %40, %59
+.critedge116:                                     ; preds = %.critedge, %116, %116, %132, %53, %55, %42, %59
   ret void
 }
 
@@ -7157,6 +7157,9 @@ declare void @llvm.experimental.noalias.scope.decl(metadata) #34
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.smax.i64(i64, i64) #33
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i64 @llvm.umin.i64(i64, i64) #33
+
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare double @llvm.sqrt.f64(double) #33
 
diff --git a/bench/ceres/optimized/schur_complement_solver.ll b/bench/ceres/optimized/schur_complement_solver.ll
index cb611263ed8..c804b279cd2 100644
--- a/bench/ceres/optimized/schur_complement_solver.ll
+++ b/bench/ceres/optimized/schur_complement_solver.ll
@@ -8194,55 +8194,55 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
 17:                                               ; preds = %_ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit
   %18 = add nsw i64 %13, -128
   %19 = sdiv i64 %18, 64
-  %.sroa.speculated157 = tail call i64 @llvm.smin.i64(i64 %19, i64 320)
-  %.sroa.speculated163 = tail call i64 @llvm.smax.i64(i64 %.sroa.speculated157, i64 8)
+  %.sroa.speculated157 = tail call i64 @llvm.smax.i64(i64 %19, i64 8)
+  %.sroa.speculated163 = tail call i64 @llvm.umin.i64(i64 %.sroa.speculated157, i64 320)
   %20 = load i64, ptr %0, align 8, !tbaa !382
   %21 = icmp slt i64 %.sroa.speculated163, %20
-  br i1 %21, label %22, label %24
+  br i1 %21, label %22, label %26
 
 22:                                               ; preds = %17
   %23 = and i64 %.sroa.speculated163, 504
   store i64 %23, ptr %0, align 8, !tbaa !382
-  br label %24
-
-24:                                               ; preds = %22, %17
-  %25 = phi i64 [ %23, %22 ], [ %20, %17 ]
-  %26 = sub nsw i64 %14, %13
-  %27 = shl i64 %25, 5
-  %28 = udiv i64 %26, %27
-  %29 = load i64, ptr %2, align 8, !tbaa !382
-  %30 = add nsw i64 %3, -1
-  %31 = add i64 %30, %29
-  %32 = sdiv i64 %31, %3
-  %.not114 = icmp sgt i64 %28, %32
-  br i1 %.not114, label %36, label %33
-
-33:                                               ; preds = %24
-  %34 = srem i64 %28, 4
-  %35 = sub nsw i64 %28, %34
-  br label %40
-
-36:                                               ; preds = %24
-  %37 = add nsw i64 %32, 3
-  %38 = srem i64 %37, 4
-  %39 = sub nsw i64 %37, %38
-  %.sroa.speculated153 = tail call i64 @llvm.smin.i64(i64 %39, i64 %29)
-  br label %40
+  br label %26
 
-40:                                               ; preds = %36, %33
-  %storemerge184 = phi i64 [ %35, %33 ], [ %.sroa.speculated153, %36 ]
+40:                                               ; preds = %24, %17
+  %storemerge184 = phi i64 [ %25, %24 ], [ %22, %17 ]
+  %28 = sub nsw i64 %14, %13
+  %29 = shl i64 %27, 5
+  %30 = udiv i64 %28, %29
+  %31 = load i64, ptr %2, align 8, !tbaa !382
+  %32 = add nsw i64 %3, -1
+  %33 = add i64 %32, %31
+  %34 = sdiv i64 %33, %3
+  %.not114 = icmp sgt i64 %30, %34
+  br i1 %.not114, label %38, label %35
+
+35:                                               ; preds = %26
+  %36 = srem i64 %30, 4
+  %37 = sub nsw i64 %30, %36
+  br label %42
+
+38:                                               ; preds = %26
+  %39 = add nsw i64 %34, 3
+  %40 = srem i64 %39, 4
+  %41 = sub nsw i64 %39, %40
+  %.sroa.speculated153 = tail call i64 @llvm.smin.i64(i64 %41, i64 %31)
+  br label %42
+
+42:                                               ; preds = %38, %35
+  %storemerge184 = phi i64 [ %37, %35 ], [ %.sroa.speculated153, %38 ]
   store i64 %storemerge184, ptr %2, align 8, !tbaa !382
   %41 = icmp sgt i64 %15, %14
   br i1 %41, label %42, label %.critedge116
 
-42:                                               ; preds = %40
+42:                                               ; preds = %42
   %43 = sub nsw i64 %15, %14
   %44 = load i64, ptr %0, align 8, !tbaa !382
   %45 = shl i64 %3, 3
   %46 = mul i64 %45, %44
   %47 = udiv i64 %43, %46
   %48 = load i64, ptr %1, align 8, !tbaa !382
-  %49 = add i64 %30, %48
+  %49 = add i64 %32, %48
   %50 = sdiv i64 %49, %3
   %51 = icmp slt i64 %47, %50
   %52 = icmp sgt i64 %47, 3
@@ -8277,87 +8277,87 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   %68 = and i64 %67, -8
   %.sroa.speculated140 = tail call i64 @llvm.smax.i64(i64 %68, i64 1)
   %69 = icmp sgt i64 %63, %.sroa.speculated140
-  br i1 %69, label %70, label %84
-
-70:                                               ; preds = %65
-  %71 = urem i64 %63, %.sroa.speculated140
-  %72 = udiv i64 %63, %.sroa.speculated140
-  %73 = icmp eq i64 %71, 0
-  br i1 %73, label %82, label %74
-
-74:                                               ; preds = %70
-  %75 = xor i64 %71, -1
-  %76 = add nsw i64 %.sroa.speculated140, %75
-  %77 = shl i64 %72, 3
-  %78 = add i64 %77, 8
-  %79 = sdiv i64 %76, %78
-  %80 = shl nsw i64 %79, 3
-  %81 = sub nsw i64 %.sroa.speculated140, %80
-  br label %82
-
-82:                                               ; preds = %70, %74
-  %83 = phi i64 [ %81, %74 ], [ %.sroa.speculated140, %70 ]
-  store i64 %83, ptr %0, align 8, !tbaa !382
-  %.pre = load i64, ptr %1, align 8, !tbaa !382
+  br i1 %69, label %72, label %86
+
+82:                                               ; preds = %67
+  %73 = urem i64 %65, %.sroa.speculated140
+  %74 = udiv i64 %63, %.sroa.speculated138
+  %75 = icmp eq i64 %73, 0
+  br i1 %75, label %84, label %76
+
+76:                                               ; preds = %72
+  %77 = xor i64 %73, -1
+  %78 = add nsw i64 %.sroa.speculated140, %77
+  %79 = shl i64 %74, 3
+  %80 = add i64 %79, 8
+  %81 = sdiv i64 %78, %80
+  %82 = shl nsw i64 %81, 3
+  %83 = sub nsw i64 %.sroa.speculated140, %82
   br label %84
 
-84:                                               ; preds = %82, %65
-  %85 = phi i64 [ %83, %82 ], [ %63, %65 ]
-  %86 = phi i64 [ %.pre, %82 ], [ %60, %65 ]
-  %87 = shl i64 %86, 3
-  %88 = mul i64 %87, %85
-  %89 = sub nsw i64 %66, %88
-  %90 = shl nsw i64 %85, 5
-  %.not = icmp slt i64 %89, %90
-  br i1 %.not, label %94, label %91
-
-91:                                               ; preds = %84
-  %92 = shl i64 %85, 3
-  %93 = udiv i64 %89, %92
-  br label %97
+84:                                               ; preds = %82, %76
+  %85 = phi i64 [ %83, %76 ], [ %.sroa.speculated138, %72 ]
+  store i64 %85, ptr %0, align 8, !tbaa !382
+  %.pre = load i64, ptr %1, align 8, !tbaa !382
+  br label %86
 
-94:                                               ; preds = %84
-  %95 = shl i64 %.sroa.speculated140, 5
-  %96 = udiv i64 4718592, %95
-  br label %97
+94:                                               ; preds = %84, %67
+  %87 = phi i64 [ %85, %84 ], [ %65, %67 ]
+  %88 = phi i64 [ %.pre, %84 ], [ %62, %67 ]
+  %89 = shl i64 %88, 3
+  %90 = mul i64 %89, %87
+  %91 = sub nsw i64 %68, %90
+  %92 = shl nsw i64 %87, 5
+  %.not = icmp slt i64 %91, %92
+  br i1 %.not, label %96, label %93
+
+93:                                               ; preds = %86
+  %94 = shl i64 %87, 3
+  %95 = udiv i64 %91, %94
+  br label %99
+
+96:                                               ; preds = %86
+  %97 = shl i64 %.sroa.speculated140, 5
+  %98 = udiv i64 4718592, %97
+  br label %99
+
+99:                                               ; preds = %96, %93
+  %storemerge = phi i64 [ %98, %96 ], [ %95, %93 ]
+  %100 = shl i64 %87, 4
+  %101 = udiv i64 1572864, %100
+  %.sroa.speculated135 = tail call i64 @llvm.smin.i64(i64 %storemerge, i64 %101)
+  %102 = and i64 %.sroa.speculated135, -4
+  %103 = load i64, ptr %2, align 8, !tbaa !382
+  %104 = icmp sgt i64 %103, %102
+  br i1 %104, label %105, label %118
+
+105:                                              ; preds = %99
+  %106 = srem i64 %103, %102
+  %107 = sdiv i64 %103, %102
+  %108 = icmp eq i64 %106, 0
+  br i1 %108, label %116, label %109
+
+109:                                              ; preds = %105
+  %110 = sub nsw i64 %102, %106
+  %111 = shl i64 %107, 2
+  %112 = add i64 %111, 4
+  %113 = sdiv i64 %110, %112
+  %114 = shl nsw i64 %113, 2
+  %115 = sub nsw i64 %102, %114
+  br label %116
 
-97:                                               ; preds = %94, %91
-  %storemerge = phi i64 [ %96, %94 ], [ %93, %91 ]
-  %98 = shl i64 %85, 4
-  %99 = udiv i64 1572864, %98
-  %.sroa.speculated135 = tail call i64 @llvm.smin.i64(i64 %storemerge, i64 %99)
-  %100 = and i64 %.sroa.speculated135, -4
-  %101 = load i64, ptr %2, align 8, !tbaa !382
-  %102 = icmp sgt i64 %101, %100
-  br i1 %102, label %103, label %116
-
-103:                                              ; preds = %97
-  %104 = srem i64 %101, %100
-  %105 = sdiv i64 %101, %100
-  %106 = icmp eq i64 %104, 0
-  br i1 %106, label %114, label %107
-
-107:                                              ; preds = %103
-  %108 = sub nsw i64 %100, %104
-  %109 = shl i64 %105, 2
-  %110 = add i64 %109, 4
-  %111 = sdiv i64 %108, %110
-  %112 = shl nsw i64 %111, 2
-  %113 = sub nsw i64 %100, %112
-  br label %114
-
-114:                                              ; preds = %103, %107
-  %115 = phi i64 [ %113, %107 ], [ %100, %103 ]
-  store i64 %115, ptr %2, align 8, !tbaa !382
+97:                                               ; preds = %105, %109
+  %storemerge = phi i64 [ %115, %109 ], [ %102, %105 ]
+  store i64 %117, ptr %2, align 8, !tbaa !382
   br label %.critedge116
 
-116:                                              ; preds = %97
-  %117 = icmp eq i64 %63, %85
+116:                                              ; preds = %99
+  %117 = icmp eq i64 %63, %87
   br i1 %117, label %118, label %.critedge116
 
 118:                                              ; preds = %116
   %119 = shl i64 %63, 3
-  %120 = mul i64 %119, %101
+  %120 = mul i64 %119, %103
   %121 = icmp slt i64 %120, 1025
   br i1 %121, label %126, label %122
 
@@ -8368,12 +8368,12 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br i1 %or.cond4, label %125, label %126
 
 125:                                              ; preds = %122
-  %.sroa.speculated128 = tail call i64 @llvm.smin.i64(i64 %86, i64 576)
+  %.sroa.speculated128 = tail call i64 @llvm.smin.i64(i64 %88, i64 576)
   br label %126
 
 126:                                              ; preds = %118, %122, %125
-  %.0182 = phi i64 [ %.sroa.speculated128, %125 ], [ %86, %122 ], [ %86, %118 ]
-  %.093 = phi i64 [ %14, %125 ], [ 1572864, %122 ], [ %13, %118 ]
+  %.0182 = phi i64 [ %.sroa.speculated128, %127 ], [ %88, %124 ], [ %88, %120 ]
+  %.093 = phi i64 [ %14, %127 ], [ 1572864, %124 ], [ %13, %120 ]
   %127 = mul i64 %63, 24
   %128 = udiv i64 %.093, %127
   %.sroa.speculated = tail call i64 @llvm.smin.i64(i64 %.0182, i64 %128)
@@ -8389,9 +8389,9 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br i1 %133, label %.critedge116, label %134
 
 134:                                              ; preds = %132, %130
-  %.0 = phi i64 [ %131, %130 ], [ %.sroa.speculated, %132 ]
-  %135 = srem i64 %86, %.0
-  %136 = sdiv i64 %86, %.0
+  %.0 = phi i64 [ %131, %132 ], [ %.sroa.speculated, %134 ]
+  %135 = srem i64 %88, %.0
+  %136 = sdiv i64 %88, %.0
   %137 = icmp eq i64 %135, 0
   br i1 %137, label %.critedge, label %138
 
@@ -8405,11 +8405,11 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br label %.critedge
 
 .critedge:                                        ; preds = %134, %138
-  %145 = phi i64 [ %144, %138 ], [ %.0, %134 ]
+  %145 = phi i64 [ %144, %140 ], [ %.0, %136 ]
   store i64 %145, ptr %1, align 8, !tbaa !382
   br label %.critedge116
 
-.critedge116:                                     ; preds = %.critedge, %116, %114, %132, %53, %55, %40, %59
+.critedge116:                                     ; preds = %.critedge, %116, %116, %132, %53, %55, %42, %59
   ret void
 }
 
diff --git a/bench/cpython/optimized/socketmodule.ll b/bench/cpython/optimized/socketmodule.ll
index 2f272ca8e75..b3978923204 100644
--- a/bench/cpython/optimized/socketmodule.ll
+++ b/bench/cpython/optimized/socketmodule.ll
@@ -9704,9 +9704,9 @@ define internal fastcc range(i32 -1, 1) i32 @sock_call_ex(ptr noundef %0, i32 no
   %17 = getelementptr inbounds nuw i8, ptr %0, i64 40
   br i1 %or.cond, label %.split92.us, label %.split92
 
-.split92.us:                                      ; preds = %7, %.split92.us.backedge
-  %.045.us = phi i64 [ %.348.us, %.split92.us.backedge ], [ 0, %7 ]
-  %.043.us = phi i32 [ %.3.us, %.split92.us.backedge ], [ 0, %7 ]
+.split92.us:                                      ; preds = %7, %.split91.us.backedge
+  %.045.us = phi i64 [ %.348.us, %.split91.us.backedge ], [ 0, %7 ]
+  %.043.us = phi i32 [ %.3.us, %.split91.us.backedge ], [ 0, %7 ]
   br i1 %10, label %28, label %18
 
 18:                                               ; preds = %.split92.us
@@ -9719,8 +9719,8 @@ define internal fastcc range(i32 -1, 1) i32 @sock_call_ex(ptr noundef %0, i32 no
   store i32 %.val64.us, ptr %8, align 4, !tbaa !196
   store i16 %spec.select.i67, ptr %14, align 4, !tbaa !198
   %21 = call i64 @_PyTime_AsMilliseconds(i64 noundef %6, i32 noundef 1) #13
-  %spec.store.select.i68.us = call i64 @llvm.smin.i64(i64 %21, i64 2147483647)
-  %spec.store.select1.i69.us = call i64 @llvm.smax.i64(i64 %spec.store.select.i68.us, i64 -1)
+  %spec.store.select.i68.us = call i64 @llvm.smax.i64(i64 %21, i64 -1)
+  %spec.store.select1.i69.us = call i64 @llvm.smin.i64(i64 %spec.store.select.i68.us, i64 2147483647)
   %22 = call ptr @PyEval_SaveThread() #13
   %23 = trunc nsw i64 %spec.store.select1.i69.us to i32
   %24 = call i32 @poll(ptr noundef nonnull %8, i64 noundef 1, i32 noundef %23) #13
@@ -9734,7 +9734,7 @@ define internal fastcc range(i32 -1, 1) i32 @sock_call_ex(ptr noundef %0, i32 no
   br label %internal_select.exit72.us
 
 internal_select.exit72.us:                        ; preds = %26, %20, %18
-  %.0.i71.us = phi i32 [ 0, %18 ], [ -1, %20 ], [ %..i70.us, %26 ]
+  %.0.i71.us = phi i32 [ 0, %18 ], [ -1, %20 ], [ %..i70.us, %27 ]
   call void @llvm.lifetime.end.p0(i64 8, ptr nonnull %8) #13
   br label %43
 
@@ -9752,8 +9752,8 @@ internal_select.exit72.us:                        ; preds = %26, %20, %18
   br label %33
 
 33:                                               ; preds = %.thread.us, %29
-  %.04176.us = phi i64 [ %6, %.thread.us ], [ %30, %29 ]
-  %.24775.us = phi i64 [ %32, %.thread.us ], [ %.045.us, %29 ]
+  %.04176.us = phi i64 [ %6, %.thread.us ], [ %30, %30 ]
+  %.24775.us = phi i64 [ %32, %.thread.us ], [ %.045.us, %30 ]
   %.val.us = load i32, ptr %12, align 8, !tbaa !83
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %9) #13
   %34 = icmp eq i32 %.val.us, -1
@@ -9763,8 +9763,8 @@ internal_select.exit72.us:                        ; preds = %26, %20, %18
   store i32 %.val.us, ptr %9, align 4, !tbaa !196
   store i16 %spec.select.i67, ptr %16, align 4, !tbaa !198
   %36 = call i64 @_PyTime_AsMilliseconds(i64 noundef %.04176.us, i32 noundef 1) #13
-  %spec.store.select.i.us = call i64 @llvm.smin.i64(i64 %36, i64 2147483647)
-  %spec.store.select1.i.us = call i64 @llvm.smax.i64(i64 %spec.store.select.i.us, i64 -1)
+  %spec.store.select.i.us = call i64 @llvm.smax.i64(i64 %36, i64 -1)
+  %spec.store.select1.i.us = call i64 @llvm.smin.i64(i64 %spec.store.select.i.us, i64 2147483647)
   %37 = call ptr @PyEval_SaveThread() #13
   %38 = trunc nsw i64 %spec.store.select1.i.us to i32
   %39 = call i32 @poll(ptr noundef nonnull %9, i64 noundef 1, i32 noundef %38) #13
@@ -9778,15 +9778,15 @@ internal_select.exit72.us:                        ; preds = %26, %20, %18
   br label %internal_select.exit.us
 
 internal_select.exit.us:                          ; preds = %41, %35, %33
-  %.0.i.us = phi i32 [ 0, %33 ], [ -1, %35 ], [ %..i.us, %41 ]
+  %.0.i.us = phi i32 [ 0, %34 ], [ -1, %36 ], [ %..i.us, %43 ]
   call void @llvm.lifetime.end.p0(i64 8, ptr nonnull %9) #13
   br label %43
 
 43:                                               ; preds = %internal_select.exit.us, %internal_select.exit72.us
-  %.348.us = phi i64 [ %.045.us, %internal_select.exit72.us ], [ %.24775.us, %internal_select.exit.us ]
-  %.3.us = phi i32 [ %.043.us, %internal_select.exit72.us ], [ 1, %internal_select.exit.us ]
-  %.1.us = phi i32 [ %.0.i71.us, %internal_select.exit72.us ], [ %.0.i.us, %internal_select.exit.us ]
-  switch i32 %.1.us, label %51 [
+  %.348.us = phi i64 [ %.045.us, %internal_select.exit71.us ], [ %.24775.us, %internal_select.exit.us ]
+  %.3.us = phi i32 [ %.043.us, %internal_select.exit71.us ], [ 1, %internal_select.exit.us ]
+  %.1.us = phi i32 [ %.0.i71.us, %internal_select.exit71.us ], [ %.0.i.us, %internal_select.exit.us ]
+  switch i32 %.1.us, label %53 [
     i32 -1, label %44
     i32 1, label %.thread77
   ]
@@ -9808,12 +9808,12 @@ internal_select.exit.us:                          ; preds = %41, %35, %33
 49:                                               ; preds = %._crit_edge
   %50 = call i32 @PyErr_CheckSignals() #13
   %.not62.us = icmp eq i32 %50, 0
-  br i1 %.not62.us, label %.split92.us.backedge, label %.split101.us
+  br i1 %.not62.us, label %.split91.us.backedge, label %.split101.us
 
 51:                                               ; preds = %43
   br i1 %.not61, label %.split.us.us, label %.split.us97
 
-.split.us97:                                      ; preds = %51, %58
+.split.us97:                                      ; preds = %51, %60
   %52 = call ptr @PyEval_SaveThread() #13
   %53 = call i32 %2(ptr noundef %0, ptr noundef %3) #13
   call void @PyEval_RestoreThread(ptr noundef %52) #13
@@ -9825,40 +9825,40 @@ internal_select.exit.us:                          ; preds = %41, %35, %33
   %56 = load i32, ptr %55, align 4, !tbaa !35
   store i32 %56, ptr %5, align 4, !tbaa !35
   %57 = icmp eq i32 %56, 4
-  br i1 %57, label %58, label %.split87.us94
-
-58:                                               ; preds = %54
-  %59 = call i32 @PyErr_CheckSignals() #13
-  %.not58.us96 = icmp eq i32 %59, 0
-  br i1 %.not58.us96, label %.split.us97, label %.split90.us
-
-.split87.us94:                                    ; preds = %54, %65
-  %.us-phi.us = phi i32 [ %67, %65 ], [ %56, %54 ]
-  %60 = load i64, ptr %17, align 8, !tbaa !90
-  %61 = icmp sgt i64 %60, 0
-  %62 = icmp eq i32 %.us-phi.us, 11
-  %or.cond63.us = and i1 %62, %61
-  br i1 %or.cond63.us, label %.split92.us.backedge, label %.split106.us
-
-.split92.us.backedge:                             ; preds = %.split87.us94, %49
-  br label %.split92.us
-
-.split.us.us:                                     ; preds = %51, %69
-  %63 = call ptr @PyEval_SaveThread() #13
-  %64 = call i32 %2(ptr noundef %0, ptr noundef %3) #13
-  call void @PyEval_RestoreThread(ptr noundef %63) #13
-  %.not56.us.us = icmp eq i32 %64, 0
-  br i1 %.not56.us.us, label %65, label %.split85.us
-
-65:                                               ; preds = %.split.us.us
-  %66 = tail call ptr @__errno_location() #14
-  %67 = load i32, ptr %66, align 4, !tbaa !35
-  %68 = icmp eq i32 %67, 4
-  br i1 %68, label %69, label %.split87.us94
+  br i1 %57, label %60, label %.split86.us93
 
-69:                                               ; preds = %65
+69:                                               ; preds = %56
   %70 = call i32 @PyErr_CheckSignals() #13
   %.not58.us.us = icmp eq i32 %70, 0
+  br i1 %.not58.us95, label %.split.us96, label %.split89.us
+
+.split86.us93:                                    ; preds = %56, %67
+  %.us-phi.us = phi i32 [ %69, %67 ], [ %58, %56 ]
+  %62 = load i64, ptr %17, align 8, !tbaa !90
+  %63 = icmp sgt i64 %62, 0
+  %64 = icmp eq i32 %.us-phi.us, 11
+  %or.cond63.us = and i1 %64, %63
+  br i1 %or.cond63.us, label %.split91.us.backedge, label %.split105.us
+
+.split91.us.backedge:                             ; preds = %.split86.us93, %51
+  br label %.split91.us
+
+.split.us.us:                                     ; preds = %53, %71
+  %65 = call ptr @PyEval_SaveThread() #13
+  %66 = call i32 %2(ptr noundef %0, ptr noundef %3) #13
+  call void @PyEval_RestoreThread(ptr noundef %65) #13
+  %.not56.us.us = icmp eq i32 %66, 0
+  br i1 %.not56.us.us, label %67, label %.split84.us
+
+67:                                               ; preds = %.split.us.us
+  %68 = tail call ptr @__errno_location() #14
+  %69 = load i32, ptr %68, align 4, !tbaa !35
+  %70 = icmp eq i32 %69, 4
+  br i1 %70, label %71, label %.split86.us93
+
+71:                                               ; preds = %67
+  %72 = call i32 @PyErr_CheckSignals() #13
+  %.not58.us.us = icmp eq i32 %72, 0
   br i1 %.not58.us.us, label %.split.us.us, label %.critedge
 
 .split92:                                         ; preds = %7
@@ -9902,7 +9902,7 @@ internal_select.exit.us:                          ; preds = %41, %35, %33
 .split99.us:                                      ; preds = %._crit_edge
   %83 = getelementptr inbounds nuw i8, ptr %0, i64 32
   %84 = load ptr, ptr %83, align 8, !tbaa !89
-  %85 = call ptr %84() #13
+  %85 = call ptr %86() #13
   br label %.critedge
 
 .thread77:                                        ; preds = %29, %43
@@ -9946,7 +9946,7 @@ internal_select.exit.us:                          ; preds = %41, %35, %33
 .split.backedge:                                  ; preds = %95, %.split87
   br label %.split
 
-.split90.us:                                      ; preds = %95, %58
+.split90.us:                                      ; preds = %95, %60
   store i32 -1, ptr %5, align 4, !tbaa !35
   br label %.critedge
 
@@ -9957,17 +9957,17 @@ internal_select.exit.us:                          ; preds = %41, %35, %33
   %or.cond63 = and i1 %99, %98
   br i1 %or.cond63, label %.split.backedge, label %.critedge
 
-.split106.us:                                     ; preds = %.split87.us94
+.split106.us:                                     ; preds = %.split86.us93
   br i1 %.not61, label %.critedge122, label %.critedge
 
 .critedge122:                                     ; preds = %.split87.us.us116, %.split106.us
   %100 = getelementptr inbounds nuw i8, ptr %0, i64 32
   %101 = load ptr, ptr %100, align 8, !tbaa !89
-  %102 = call ptr %101() #13
+  %102 = call ptr %103() #13
   br label %.critedge
 
-.critedge:                                        ; preds = %.split87, %77, %69, %.split106.us, %.critedge122, %.split90.us, %.split85.us, %.split85.us.thread, %86, %87, %.split101.us, %82, %.split99.us
-  %.0 = phi i32 [ -1, %.split99.us ], [ -1, %82 ], [ -1, %.split101.us ], [ -1, %87 ], [ -1, %86 ], [ 0, %.split85.us.thread ], [ 0, %.split85.us ], [ -1, %.split90.us ], [ -1, %.critedge122 ], [ -1, %.split106.us ], [ -1, %69 ], [ -1, %77 ], [ -1, %.split87 ]
+.critedge:                                        ; preds = %.split87, %77, %71, %.split106.us, %.critedge122, %.split90.us, %.split85.us, %.split85.us.thread, %86, %87, %.split101.us, %82, %.split99.us
+  %.0 = phi i32 [ -1, %.split98.us ], [ -1, %84 ], [ -1, %.split100.us ], [ -1, %89 ], [ -1, %88 ], [ 0, %.split84.us.thread ], [ 0, %.split84.us ], [ -1, %.split89.us ], [ -1, %.critedge121 ], [ -1, %.split105.us ], [ -1, %71 ], [ -1, %79 ], [ -1, %.split86 ]
   ret i32 %.0
 }
 
diff --git a/bench/darktable/optimized/decoders_libraw.ll b/bench/darktable/optimized/decoders_libraw.ll
index 298dd596056..d3a22781bf6 100644
--- a/bench/darktable/optimized/decoders_libraw.ll
+++ b/bench/darktable/optimized/decoders_libraw.ll
@@ -3535,8 +3535,8 @@ _ZN15iiq_bitstream_t3getEc.exit94:                ; preds = %_ZN15iiq_bitstream_
   %.sroa.0.5 = phi i64 [ %.pre.i.i.i85155, %_ZN15iiq_bitstream_t3getEc.exit87 ], [ %.sroa.0.20, %_ZN15iiq_bitstream_t3getEc.exit94 ]
   %.045 = phi i32 [ %165, %_ZN15iiq_bitstream_t3getEc.exit87 ], [ %200, %_ZN15iiq_bitstream_t3getEc.exit94 ]
   %202 = shl i32 %.045, 2
-  %203 = tail call i32 @llvm.smin.i32(i32 %202, i32 65535)
-  %204 = tail call i32 @llvm.smax.i32(i32 %203, i32 0)
+  %203 = tail call i32 @llvm.smax.i32(i32 %202, i32 0)
+  %204 = tail call i32 @llvm.umin.i32(i32 %203, i32 65535)
   %205 = trunc nuw i32 %204 to i16
   %206 = getelementptr inbounds nuw i16, ptr %.1177, i64 %indvars.iv
   store i16 %205, ptr %206, align 2, !tbaa !80
@@ -4355,9 +4355,6 @@ declare i32 @llvm.bswap.i32(i32) #12
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i8 @llvm.fshl.i8(i8, i8, i8) #12
 
-; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smin.i32(i32, i32) #12
-
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i16 @llvm.umax.i16(i16, i16) #12
 
diff --git a/bench/darktable/optimized/identify.ll b/bench/darktable/optimized/identify.ll
index 6193f4952dc..b51f8d257bc 100644
--- a/bench/darktable/optimized/identify.ll
+++ b/bench/darktable/optimized/identify.ll
@@ -8385,18 +8385,19 @@ define void @_ZN6LibRaw27identify_process_dng_fieldsEv(ptr noundef nonnull align
 
 ._crit_edge664:                                   ; preds = %.lr.ph663, %.preheader552
   %.0323.lcssa = phi i32 [ 0, %.preheader552 ], [ %694, %.lr.ph663 ]
-  %685 = call i32 @llvm.smax.i32(i32 %invariant.smin660, i32 1)
-  %686 = sdiv i32 %.0323.lcssa, %685
-  %687 = getelementptr inbounds nuw i8, ptr %0, i64 136592
-  %688 = load i32, ptr %687, align 8, !tbaa !92
-  %689 = getelementptr inbounds nuw i8, ptr %0, i64 136596
-  %690 = load i32, ptr %689, align 4, !tbaa !92
-  %691 = mul i32 %690, %688
-  %.not489 = icmp eq i32 %691, 0
-  br i1 %.not489, label %701, label %.preheader
+  %685 = call i32 @llvm.smax.i32(i32 %683, i32 1)
+  %686 = call i32 @llvm.umin.i32(i32 %685, i32 4)
+  %687 = sdiv i32 %.0323.lcssa, %686
+  %688 = getelementptr inbounds nuw i8, ptr %0, i64 136592
+  %689 = load i32, ptr %688, align 8, !tbaa !92
+  %690 = getelementptr inbounds nuw i8, ptr %0, i64 136596
+  %691 = load i32, ptr %690, align 4, !tbaa !92
+  %692 = mul i32 %691, %689
+  %.not489 = icmp eq i32 %692, 0
+  br i1 %.not489, label %702, label %.preheader
 
 .preheader:                                       ; preds = %._crit_edge664
-  %invariant.umin667 = call i32 @llvm.umin.i32(i32 %691, i32 4096)
+  %invariant.umin667 = call i32 @llvm.umin.i32(i32 %692, i32 4096)
   %wide.trip.count800 = zext nneg i32 %invariant.umin667 to i64
   br label %696
 
@@ -8415,8 +8416,8 @@ define void @_ZN6LibRaw27identify_process_dng_fieldsEv(ptr noundef nonnull align
   br label %701
 
 696:                                              ; preds = %.preheader, %696
-  %indvars.iv795 = phi i64 [ 0, %.preheader ], [ %indvars.iv.next796, %696 ]
-  %.1322668 = phi i32 [ 0, %.preheader ], [ %700, %696 ]
+  %indvars.iv795 = phi i64 [ 0, %.preheader ], [ %indvars.iv.next796, %697 ]
+  %.1322668 = phi i32 [ 0, %.preheader ], [ %700, %697 ]
   %697 = add nuw nsw i64 %indvars.iv795, 6
   %698 = getelementptr inbounds nuw [4104 x i32], ptr %656, i64 0, i64 %697
   %699 = load i32, ptr %698, align 4, !tbaa !92
@@ -8430,7 +8431,7 @@ define void @_ZN6LibRaw27identify_process_dng_fieldsEv(ptr noundef nonnull align
   br i1 %684, label %.lr.ph675, label %._crit_edge676
 
 .lr.ph675:                                        ; preds = %701
-  %702 = add i32 %.0321, %686
+  %702 = add i32 %.0321, %687
   %703 = add i32 %702, %655
   %704 = sub i32 %522, %703
   %705 = uitofp i32 %704 to float
@@ -8449,7 +8450,7 @@ define void @_ZN6LibRaw27identify_process_dng_fieldsEv(ptr noundef nonnull align
   br i1 %.not490, label %719, label %715
 
 713:                                              ; preds = %.lr.ph675, %713
-  %indvars.iv802 = phi i64 [ 0, %.lr.ph675 ], [ %indvars.iv.next803, %713 ]
+  %indvars.iv802 = phi i64 [ 0, %.lr.ph675 ], [ %indvars.iv.next803, %714 ]
   %714 = getelementptr inbounds nuw [4 x i64], ptr %710, i64 0, i64 %indvars.iv802
   store i64 %709, ptr %714, align 8, !tbaa !259
   %indvars.iv.next803 = add nuw nsw i64 %indvars.iv802, 1
diff --git a/bench/darktable/optimized/kodak_decoders.ll b/bench/darktable/optimized/kodak_decoders.ll
index a082a94489d..7d20efa56f5 100644
--- a/bench/darktable/optimized/kodak_decoders.ll
+++ b/bench/darktable/optimized/kodak_decoders.ll
@@ -1572,8 +1572,8 @@ _ZNSt6vectorIhSaIhEED2Ev.exit:                    ; preds = %41, %35, %25, %24
   %indvars.iv = phi i64 [ 0, %53 ], [ %indvars.iv.next, %71 ]
   %72 = getelementptr inbounds nuw [3 x i32], ptr %2, i64 0, i64 %indvars.iv
   %73 = load i32, ptr %72, align 4, !tbaa !142
-  %narrow = tail call i32 @llvm.smin.i32(i32 %73, i32 255)
-  %narrow43 = tail call i32 @llvm.smax.i32(i32 %narrow, i32 0)
+  %narrow = tail call i32 @llvm.smax.i32(i32 %73, i32 0)
+  %narrow43 = tail call i32 @llvm.umin.i32(i32 %narrow, i32 255)
   %74 = zext nneg i32 %narrow43 to i64
   %75 = getelementptr inbounds nuw [65536 x i16], ptr %23, i64 0, i64 %74
   %76 = load i16, ptr %75, align 2, !tbaa !75
@@ -1756,8 +1756,8 @@ _ZNSt6vectorIhSaIhEEC2EmRKS0_.exit:               ; preds = %.noexc, %7
   %indvars.iv61 = phi i64 [ 0, %.lr.ph.split.us ], [ %indvars.iv.next62, %75 ]
   %76 = getelementptr inbounds nuw [3 x i32], ptr %2, i64 0, i64 %indvars.iv61
   %77 = load i32, ptr %76, align 4, !tbaa !142
-  %narrow.us = tail call i32 @llvm.smin.i32(i32 %77, i32 255)
-  %narrow47.us = tail call i32 @llvm.smax.i32(i32 %narrow.us, i32 0)
+  %narrow.us = tail call i32 @llvm.smax.i32(i32 %77, i32 0)
+  %narrow47.us = tail call i32 @llvm.umin.i32(i32 %narrow.us, i32 255)
   %78 = zext nneg i32 %narrow47.us to i64
   %79 = getelementptr inbounds nuw [65536 x i16], ptr %23, i64 0, i64 %78
   %80 = load i16, ptr %79, align 2, !tbaa !75
@@ -1773,8 +1773,8 @@ _ZNSt6vectorIhSaIhEEC2EmRKS0_.exit:               ; preds = %.noexc, %7
   br i1 %exitcond64.not, label %70, label %75, !llvm.loop !151
 
 .lr.ph.split:                                     ; preds = %.lr.ph, %119
-  %indvars.iv58 = phi i64 [ %indvars.iv.next59, %119 ], [ 0, %.lr.ph ]
-  %87 = phi i32 [ %121, %119 ], [ %45, %.lr.ph ]
+  %indvars.iv58 = phi i64 [ %indvars.iv.next59, %123 ], [ 0, %.lr.ph ]
+  %87 = phi i32 [ %121, %123 ], [ %45, %.lr.ph ]
   %88 = getelementptr inbounds nuw i8, ptr %.sroa.037.0, i64 %indvars.iv58
   %89 = load i8, ptr %88, align 1, !tbaa !79
   %90 = zext i8 %89 to i32
@@ -1802,11 +1802,11 @@ _ZNSt6vectorIhSaIhEEC2EmRKS0_.exit:               ; preds = %.noexc, %7
   br label %107
 
 107:                                              ; preds = %.lr.ph.split, %107
-  %indvars.iv = phi i64 [ 0, %.lr.ph.split ], [ %indvars.iv.next, %107 ]
+  %indvars.iv = phi i64 [ 0, %.lr.ph.split ], [ %indvars.iv.next, %109 ]
   %108 = getelementptr inbounds nuw [3 x i32], ptr %2, i64 0, i64 %indvars.iv
   %109 = load i32, ptr %108, align 4, !tbaa !142
-  %narrow = tail call i32 @llvm.smin.i32(i32 %109, i32 255)
-  %narrow47 = tail call i32 @llvm.smax.i32(i32 %narrow, i32 0)
+  %narrow = tail call i32 @llvm.smax.i32(i32 %109, i32 0)
+  %narrow47 = tail call i32 @llvm.umin.i32(i32 %narrow, i32 255)
   %110 = zext nneg i32 %narrow47 to i64
   %111 = getelementptr inbounds nuw [65536 x i16], ptr %23, i64 0, i64 %110
   %112 = load i16, ptr %111, align 2, !tbaa !75
@@ -2652,8 +2652,8 @@ define void @_ZN6LibRaw20kodak_ycbcr_load_rawEv(ptr noundef nonnull align 8 dere
   %74 = getelementptr inbounds nuw [3 x i32], ptr %4, i64 0, i64 %indvars.iv
   %75 = load i32, ptr %74, align 4, !tbaa !142
   %76 = add nsw i32 %75, %60
-  %narrow = tail call i32 @llvm.smin.i32(i32 %76, i32 4095)
-  %narrow73 = tail call i32 @llvm.smax.i32(i32 %narrow, i32 0)
+  %narrow = tail call i32 @llvm.smax.i32(i32 %76, i32 0)
+  %narrow73 = tail call i32 @llvm.umin.i32(i32 %narrow, i32 4095)
   %77 = zext nneg i32 %narrow73 to i64
   %78 = getelementptr inbounds nuw [65536 x i16], ptr %25, i64 0, i64 %77
   %79 = load i16, ptr %78, align 2, !tbaa !75
@@ -3029,9 +3029,6 @@ declare i64 @llvm.smin.i64(i64, i64) #11
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.umax.i32(i32, i32) #11
 
-; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smin.i32(i32, i32) #11
-
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.umin.i32(i32, i32) #11
 
diff --git a/bench/darktable/optimized/load_mfbacks.ll b/bench/darktable/optimized/load_mfbacks.ll
index 51015de0280..85f766e5275 100644
--- a/bench/darktable/optimized/load_mfbacks.ll
+++ b/bench/darktable/optimized/load_mfbacks.ll
@@ -1944,8 +1944,8 @@ _ZNSt6vectorIjSaIjEE9push_backERKj.exit:          ; preds = %129, %199, %200, %_
   %489 = uitofp i16 %488 to float
   %490 = fmul reassoc nsz arcp contract afn float %483, %489
   %491 = fptosi float %490 to i32
-  %492 = call i32 @llvm.smin.i32(i32 %491, i32 65535)
-  %493 = call i32 @llvm.smax.i32(i32 %492, i32 0)
+  %492 = call i32 @llvm.smax.i32(i32 %491, i32 0)
+  %493 = call i32 @llvm.umin.i32(i32 %492, i32 65535)
   %494 = trunc nuw i32 %493 to i16
   store i16 %494, ptr %487, align 2, !tbaa !74
   %495 = add nuw nsw i32 %storemerge316692, 1
@@ -2658,8 +2658,8 @@ _ZN6LibRaw27phase_one_fix_col_pixel_avgEjj.exit:  ; preds = %634, %656
   %801 = fadd reassoc nsz arcp contract afn float %800, %751
   %802 = fmul reassoc nsz arcp contract afn float %801, 2.000000e+00
   %803 = fptosi float %802 to i32
-  %804 = call i32 @llvm.smin.i32(i32 %803, i32 65535)
-  %805 = call i32 @llvm.smax.i32(i32 %804, i32 0)
+  %804 = call i32 @llvm.smax.i32(i32 %803, i32 0)
+  %805 = call i32 @llvm.umin.i32(i32 %804, i32 65535)
   %806 = trunc nuw i32 %805 to i16
   store i16 %806, ptr %747, align 2, !tbaa !74
   %807 = add nuw nsw i32 %storemerge745.us.us, 1
@@ -2748,8 +2748,8 @@ _ZN6LibRaw27phase_one_fix_col_pixel_avgEjj.exit:  ; preds = %634, %656
   %851 = fadd reassoc nsz arcp contract afn float %850, %844
   %852 = fmul reassoc nsz arcp contract afn float %851, 2.000000e+00
   %853 = fptosi float %852 to i32
-  %854 = call i32 @llvm.smin.i32(i32 %853, i32 65535)
-  %855 = call i32 @llvm.smax.i32(i32 %854, i32 0)
+  %854 = call i32 @llvm.smax.i32(i32 %853, i32 0)
+  %855 = call i32 @llvm.umin.i32(i32 %854, i32 65535)
   %856 = trunc nuw i32 %855 to i16
   store i16 %856, ptr %835, align 2, !tbaa !74
   %857 = add nuw nsw i32 %storemerge745, 1
@@ -4936,9 +4936,6 @@ declare i32 @llvm.umin.i32(i32, i32) #17
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.umax.i32(i32, i32) #17
 
-; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smin.i32(i32, i32) #17
-
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smax.i32(i32, i32) #17
 
diff --git a/bench/faiss/optimized/AdditiveQuantizer.ll b/bench/faiss/optimized/AdditiveQuantizer.ll
index ba03de31c6f..9f8b7b1974d 100644
--- a/bench/faiss/optimized/AdditiveQuantizer.ll
+++ b/bench/faiss/optimized/AdditiveQuantizer.ll
@@ -1858,8 +1858,8 @@ define noundef range(i64 0, 4294967296) i64 @_ZNK5faiss17AdditiveQuantizer11enco
   %20 = fmul float %19, 2.560000e+02
   %21 = tail call float @llvm.floor.f32(float %20)
   %22 = fptosi float %21 to i32
-  %23 = tail call i32 @llvm.smin.i32(i32 %22, i32 255)
-  %24 = tail call i32 @llvm.smax.i32(i32 %23, i32 0)
+  %23 = tail call i32 @llvm.smax.i32(i32 %22, i32 0)
+  %24 = tail call i32 @llvm.umin.i32(i32 %23, i32 255)
   %25 = zext nneg i32 %24 to i64
   br label %48
 
@@ -1874,8 +1874,8 @@ define noundef range(i64 0, 4294967296) i64 @_ZNK5faiss17AdditiveQuantizer11enco
   %34 = fmul float %33, 1.600000e+01
   %35 = tail call float @llvm.floor.f32(float %34)
   %36 = fptosi float %35 to i32
-  %37 = tail call i32 @llvm.smin.i32(i32 %36, i32 15)
-  %38 = tail call i32 @llvm.smax.i32(i32 %37, i32 0)
+  %37 = tail call i32 @llvm.smax.i32(i32 %36, i32 0)
+  %38 = tail call i32 @llvm.umin.i32(i32 %37, i32 15)
   %39 = zext nneg i32 %38 to i64
   br label %48
 
@@ -5477,9 +5477,6 @@ declare void @llvm.assume(i1 noundef) #23
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare float @llvm.floor.f32(float) #24
 
-; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smin.i32(i32, i32) #24
-
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smax.i32(i32, i32) #24
 
@@ -5492,6 +5489,9 @@ declare i64 @llvm.smin.i64(i64, i64) #24
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.umax.i64(i64, i64) #24
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #24
+
 attributes #0 = { mustprogress uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { cold mustprogress noreturn nounwind memory(inaccessiblemem: write) uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #2 = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
diff --git a/bench/freetype/optimized/sdf.ll b/bench/freetype/optimized/sdf.ll
index ffab7c492d2..2791c0706bf 100644
--- a/bench/freetype/optimized/sdf.ll
+++ b/bench/freetype/optimized/sdf.ll
@@ -3086,7 +3086,7 @@ get_control_box.exit.i:                           ; preds = %.critedge4.i.i, %.l
   br i1 %221, label %.lr.ph.us.i, label %._crit_edge229.i, !llvm.loop !162
 
 .lr.ph.split.us.us.i:                             ; preds = %756, %.lr.ph.split.us.us.preheader.i
-  %indvars.iv.i = phi i64 [ %212, %.lr.ph.split.us.us.preheader.i ], [ %indvars.iv.next.i, %756 ]
+  %indvars.iv.i = phi i64 [ %212, %.lr.ph.split.us.us.preheader.i ], [ %indvars.iv.next.i, %758 ]
   %222 = icmp sgt i64 %indvars.iv.i, -1
   %.not166.us.us.i = icmp slt i64 %indvars.iv.i, %117
   %or.cond170.us.us.i = and i1 %222, %.not166.us.us.i
@@ -3704,8 +3704,8 @@ get_min_distance_conic.exit.i.us.us.i:            ; preds = %623, %585
   %663 = ashr exact i64 %sext39.i.i.us.us.i, 32
   %664 = call i64 @FT_DivFix(i64 noundef %662, i64 noundef %663) #11
   %665 = trunc i64 %664 to i32
-  %spec.store.select.i.i.us.us.i = call i32 @llvm.smin.i32(i32 %665, i32 65536)
-  %spec.store.select4.i.i.us.us.i = call i32 @llvm.smax.i32(i32 %spec.store.select.i.i.us.us.i, i32 0)
+  %spec.store.select.i.i.us.us.i = call i32 @llvm.smax.i32(i32 %665, i32 0)
+  %spec.store.select4.i.i.us.us.i = call i32 @llvm.umin.i32(i32 %spec.store.select.i.i.us.us.i, i32 65536)
   %.tr.i.i.us.us.i = trunc i64 %648 to i32
   %666 = shl i32 %.tr.i.i.us.us.i, 10
   %667 = sext i32 %666 to i64
@@ -3790,7 +3790,7 @@ get_min_distance_conic.exit.i.us.us.i:            ; preds = %623, %585
   br label %get_min_distance_line.exit.i.us.us.i
 
 get_min_distance_line.exit.i.us.us.i:             ; preds = %708, %647
-  %.sink.i.i.us.us.i = phi i32 [ %731, %708 ], [ 65536, %647 ]
+  %.sink.i.i.us.us.i = phi i32 [ %731, %710 ], [ 65536, %647 ]
   call void @llvm.lifetime.end.p0(i64 16, ptr nonnull %13) #11
   call void @llvm.lifetime.end.p0(i64 16, ptr nonnull %12) #11
   br label %732
diff --git a/bench/g2o/optimized/edge_se2_lotsofxy.ll b/bench/g2o/optimized/edge_se2_lotsofxy.ll
index f133f1b8d3c..6bdb1ca5acc 100644
--- a/bench/g2o/optimized/edge_se2_lotsofxy.ll
+++ b/bench/g2o/optimized/edge_se2_lotsofxy.ll
@@ -6644,8 +6644,8 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
 17:                                               ; preds = %_ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit
   %18 = add nsw i64 %13, -128
   %19 = sdiv i64 %18, 64
-  %.sroa.speculated157 = tail call i64 @llvm.smin.i64(i64 %19, i64 320)
-  %.sroa.speculated163 = tail call i64 @llvm.smax.i64(i64 %.sroa.speculated157, i64 8)
+  %.sroa.speculated157 = tail call i64 @llvm.smax.i64(i64 %19, i64 8)
+  %.sroa.speculated163 = tail call i64 @llvm.umin.i64(i64 %.sroa.speculated157, i64 320)
   %20 = load i64, ptr %0, align 8, !tbaa !131
   %21 = icmp slt i64 %.sroa.speculated163, %20
   br i1 %21, label %22, label %24
@@ -6656,7 +6656,7 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br label %24
 
 24:                                               ; preds = %22, %17
-  %25 = phi i64 [ %23, %22 ], [ %20, %17 ]
+  %25 = phi i64 [ %23, %24 ], [ %20, %17 ]
   %26 = sub nsw i64 %14, %13
   %27 = shl i64 %25, 5
   %28 = udiv i64 %26, %27
@@ -6680,7 +6680,7 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br label %40
 
 40:                                               ; preds = %36, %33
-  %storemerge184 = phi i64 [ %35, %33 ], [ %.sroa.speculated153, %36 ]
+  %storemerge184 = phi i64 [ %35, %35 ], [ %.sroa.speculated153, %38 ]
   store i64 %storemerge184, ptr %2, align 8, !tbaa !131
   %41 = icmp sgt i64 %15, %14
   br i1 %41, label %42, label %.critedge116
@@ -6727,7 +6727,7 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   %68 = and i64 %67, -8
   %.sroa.speculated140 = tail call i64 @llvm.smax.i64(i64 %68, i64 1)
   %69 = icmp sgt i64 %63, %.sroa.speculated140
-  br i1 %69, label %70, label %84
+  br i1 %69, label %70, label %86
 
 70:                                               ; preds = %65
   %71 = urem i64 %63, %.sroa.speculated140
@@ -6746,68 +6746,68 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br label %82
 
 82:                                               ; preds = %70, %74
-  %83 = phi i64 [ %81, %74 ], [ %.sroa.speculated140, %70 ]
+  %83 = phi i64 [ %81, %76 ], [ %.sroa.speculated140, %72 ]
   store i64 %83, ptr %0, align 8, !tbaa !131
   %.pre = load i64, ptr %1, align 8, !tbaa !131
-  br label %84
-
-84:                                               ; preds = %82, %65
-  %85 = phi i64 [ %83, %82 ], [ %63, %65 ]
-  %86 = phi i64 [ %.pre, %82 ], [ %60, %65 ]
-  %87 = shl i64 %86, 3
-  %88 = mul i64 %87, %85
-  %89 = sub nsw i64 %66, %88
-  %90 = shl nsw i64 %85, 5
-  %.not = icmp slt i64 %89, %90
-  br i1 %.not, label %94, label %91
-
-91:                                               ; preds = %84
-  %92 = shl i64 %85, 3
-  %93 = udiv i64 %89, %92
-  br label %97
-
-94:                                               ; preds = %84
-  %95 = shl i64 %.sroa.speculated140, 5
-  %96 = udiv i64 4718592, %95
-  br label %97
-
-97:                                               ; preds = %94, %91
-  %storemerge = phi i64 [ %96, %94 ], [ %93, %91 ]
-  %98 = shl i64 %85, 4
-  %99 = udiv i64 1572864, %98
-  %.sroa.speculated135 = tail call i64 @llvm.smin.i64(i64 %storemerge, i64 %99)
-  %100 = and i64 %.sroa.speculated135, -4
-  %101 = load i64, ptr %2, align 8, !tbaa !131
-  %102 = icmp sgt i64 %101, %100
-  br i1 %102, label %103, label %116
-
-103:                                              ; preds = %97
-  %104 = srem i64 %101, %100
-  %105 = sdiv i64 %101, %100
-  %106 = icmp eq i64 %104, 0
-  br i1 %106, label %114, label %107
-
-107:                                              ; preds = %103
-  %108 = sub nsw i64 %100, %104
-  %109 = shl i64 %105, 2
-  %110 = add i64 %109, 4
-  %111 = sdiv i64 %108, %110
-  %112 = shl nsw i64 %111, 2
-  %113 = sub nsw i64 %100, %112
-  br label %114
-
-114:                                              ; preds = %103, %107
-  %115 = phi i64 [ %113, %107 ], [ %100, %103 ]
-  store i64 %115, ptr %2, align 8, !tbaa !131
+  br label %86
+
+86:                                               ; preds = %84, %67
+  %87 = phi i64 [ %85, %84 ], [ %65, %67 ]
+  %88 = phi i64 [ %.pre, %84 ], [ %62, %67 ]
+  %89 = shl i64 %88, 3
+  %90 = mul i64 %89, %87
+  %91 = sub nsw i64 %68, %90
+  %92 = shl nsw i64 %87, 5
+  %.not = icmp slt i64 %91, %92
+  br i1 %.not, label %96, label %93
+
+93:                                               ; preds = %86
+  %94 = shl i64 %87, 3
+  %95 = udiv i64 %91, %94
+  br label %99
+
+96:                                               ; preds = %86
+  %97 = shl i64 %.sroa.speculated140, 5
+  %98 = udiv i64 4718592, %97
+  br label %99
+
+99:                                               ; preds = %96, %93
+  %storemerge = phi i64 [ %98, %96 ], [ %95, %93 ]
+  %100 = shl i64 %87, 4
+  %101 = udiv i64 1572864, %100
+  %.sroa.speculated135 = tail call i64 @llvm.smin.i64(i64 %storemerge, i64 %101)
+  %102 = and i64 %.sroa.speculated135, -4
+  %103 = load i64, ptr %2, align 8, !tbaa !131
+  %104 = icmp sgt i64 %103, %102
+  br i1 %104, label %105, label %118
+
+105:                                              ; preds = %99
+  %106 = srem i64 %103, %102
+  %107 = sdiv i64 %103, %102
+  %108 = icmp eq i64 %106, 0
+  br i1 %108, label %116, label %109
+
+84:                                               ; preds = %105
+  %110 = sub nsw i64 %102, %106
+  %111 = shl i64 %107, 2
+  %112 = add i64 %111, 4
+  %113 = sdiv i64 %110, %112
+  %114 = shl nsw i64 %113, 2
+  %115 = sub nsw i64 %102, %114
+  br label %116
+
+116:                                              ; preds = %105, %109
+  %117 = phi i64 [ %115, %109 ], [ %102, %105 ]
+  store i64 %117, ptr %2, align 8, !tbaa !131
   br label %.critedge116
 
-116:                                              ; preds = %97
-  %117 = icmp eq i64 %63, %85
+116:                                              ; preds = %99
+  %117 = icmp eq i64 %63, %87
   br i1 %117, label %118, label %.critedge116
 
 118:                                              ; preds = %116
   %119 = shl i64 %63, 3
-  %120 = mul i64 %119, %101
+  %120 = mul i64 %119, %103
   %121 = icmp slt i64 %120, 1025
   br i1 %121, label %126, label %122
 
@@ -6818,12 +6818,12 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br i1 %or.cond4, label %125, label %126
 
 125:                                              ; preds = %122
-  %.sroa.speculated128 = tail call i64 @llvm.smin.i64(i64 %86, i64 576)
+  %.sroa.speculated128 = tail call i64 @llvm.smin.i64(i64 %88, i64 576)
   br label %126
 
 126:                                              ; preds = %118, %122, %125
-  %.0182 = phi i64 [ %.sroa.speculated128, %125 ], [ %86, %122 ], [ %86, %118 ]
-  %.093 = phi i64 [ %14, %125 ], [ 1572864, %122 ], [ %13, %118 ]
+  %.0182 = phi i64 [ %.sroa.speculated128, %127 ], [ %88, %124 ], [ %88, %120 ]
+  %.093 = phi i64 [ %14, %127 ], [ 1572864, %124 ], [ %13, %120 ]
   %127 = mul i64 %63, 24
   %128 = udiv i64 %.093, %127
   %.sroa.speculated = tail call i64 @llvm.smin.i64(i64 %.0182, i64 %128)
@@ -6839,9 +6839,9 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br i1 %133, label %.critedge116, label %134
 
 134:                                              ; preds = %132, %130
-  %.0 = phi i64 [ %131, %130 ], [ %.sroa.speculated, %132 ]
-  %135 = srem i64 %86, %.0
-  %136 = sdiv i64 %86, %.0
+  %.0 = phi i64 [ %131, %132 ], [ %.sroa.speculated, %134 ]
+  %135 = srem i64 %88, %.0
+  %136 = sdiv i64 %88, %.0
   %137 = icmp eq i64 %135, 0
   br i1 %137, label %.critedge, label %138
 
@@ -6855,11 +6855,11 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br label %.critedge
 
 .critedge:                                        ; preds = %134, %138
-  %145 = phi i64 [ %144, %138 ], [ %.0, %134 ]
+  %145 = phi i64 [ %144, %140 ], [ %.0, %136 ]
   store i64 %145, ptr %1, align 8, !tbaa !131
   br label %.critedge116
 
-.critedge116:                                     ; preds = %.critedge, %116, %114, %132, %53, %55, %40, %59
+.critedge116:                                     ; preds = %.critedge, %116, %116, %132, %53, %55, %40, %59
   ret void
 }
 
@@ -12118,12 +12118,12 @@ declare i64 @llvm.smax.i64(i64, i64) #31
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.umax.i64(i64, i64) #31
 
-; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)
-declare void @llvm.memmove.p0.p0.i64(ptr writeonly captures(none), ptr readonly captures(none), i64, i1 immarg) #33
-
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.umin.i64(i64, i64) #31
 
+; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.memmove.p0.p0.i64(ptr writeonly captures(none), ptr readonly captures(none), i64, i1 immarg) #33
+
 ; Function Attrs: nofree nounwind willreturn allockind("alloc,zeroed") allocsize(0,1) memory(inaccessiblemem: readwrite)
 declare noalias noundef ptr @calloc(i64 noundef, i64 noundef) local_unnamed_addr #34
 
diff --git a/bench/g2o/optimized/edge_se3_calib.ll b/bench/g2o/optimized/edge_se3_calib.ll
index 2d14d3ed8d7..ae7e82d8360 100644
--- a/bench/g2o/optimized/edge_se3_calib.ll
+++ b/bench/g2o/optimized/edge_se3_calib.ll
@@ -5210,8 +5210,8 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
 17:                                               ; preds = %_ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit
   %18 = add nsw i64 %13, -128
   %19 = sdiv i64 %18, 64
-  %.sroa.speculated157 = tail call i64 @llvm.smin.i64(i64 %19, i64 320)
-  %.sroa.speculated163 = tail call i64 @llvm.smax.i64(i64 %.sroa.speculated157, i64 8)
+  %.sroa.speculated157 = tail call i64 @llvm.smax.i64(i64 %19, i64 8)
+  %.sroa.speculated163 = tail call i64 @llvm.umin.i64(i64 %.sroa.speculated157, i64 320)
   %20 = load i64, ptr %0, align 8, !tbaa !176
   %21 = icmp slt i64 %.sroa.speculated163, %20
   br i1 %21, label %22, label %24
@@ -5222,7 +5222,7 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br label %24
 
 24:                                               ; preds = %22, %17
-  %25 = phi i64 [ %23, %22 ], [ %20, %17 ]
+  %25 = phi i64 [ %23, %24 ], [ %20, %17 ]
   %26 = sub nsw i64 %14, %13
   %27 = shl i64 %25, 5
   %28 = udiv i64 %26, %27
@@ -5246,7 +5246,7 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br label %40
 
 40:                                               ; preds = %36, %33
-  %storemerge184 = phi i64 [ %35, %33 ], [ %.sroa.speculated153, %36 ]
+  %storemerge184 = phi i64 [ %35, %35 ], [ %.sroa.speculated153, %38 ]
   store i64 %storemerge184, ptr %2, align 8, !tbaa !176
   %41 = icmp sgt i64 %15, %14
   br i1 %41, label %42, label %.critedge116
@@ -5312,14 +5312,14 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br label %82
 
 82:                                               ; preds = %70, %74
-  %83 = phi i64 [ %81, %74 ], [ %.sroa.speculated140, %70 ]
+  %83 = phi i64 [ %81, %76 ], [ %.sroa.speculated140, %72 ]
   store i64 %83, ptr %0, align 8, !tbaa !176
   %.pre = load i64, ptr %1, align 8, !tbaa !176
   br label %84
 
 84:                                               ; preds = %82, %65
-  %85 = phi i64 [ %83, %82 ], [ %63, %65 ]
-  %86 = phi i64 [ %.pre, %82 ], [ %60, %65 ]
+  %85 = phi i64 [ %83, %84 ], [ %63, %67 ]
+  %86 = phi i64 [ %.pre, %84 ], [ %60, %67 ]
   %87 = shl i64 %86, 3
   %88 = mul i64 %87, %85
   %89 = sub nsw i64 %66, %88
@@ -5338,7 +5338,7 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br label %97
 
 97:                                               ; preds = %94, %91
-  %storemerge = phi i64 [ %96, %94 ], [ %93, %91 ]
+  %storemerge = phi i64 [ %96, %96 ], [ %93, %93 ]
   %98 = shl i64 %85, 4
   %99 = udiv i64 1572864, %98
   %.sroa.speculated135 = tail call i64 @llvm.smin.i64(i64 %storemerge, i64 %99)
@@ -5363,7 +5363,7 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   br label %114
 
 114:                                              ; preds = %103, %107
-  %115 = phi i64 [ %113, %107 ], [ %100, %103 ]
+  %115 = phi i64 [ %113, %109 ], [ %100, %105 ]
   store i64 %115, ptr %2, align 8, !tbaa !176
   br label %.critedge116
 
@@ -5375,57 +5375,57 @@ _ZN5Eigen8internal20manage_caching_sizesENS_6ActionEPlS2_S2_.exit: ; preds = %4,
   %119 = shl i64 %63, 3
   %120 = mul i64 %119, %101
   %121 = icmp slt i64 %120, 1025
-  br i1 %121, label %126, label %122
+  br i1 %121, label %128, label %122
 
 122:                                              ; preds = %118
   %123 = icmp ne i64 %15, 0
   %124 = icmp samesign ult i64 %120, 32769
   %or.cond4 = and i1 %123, %124
-  br i1 %or.cond4, label %125, label %126
+  br i1 %or.cond4, label %125, label %128
 
 125:                                              ; preds = %122
   %.sroa.speculated128 = tail call i64 @llvm.smin.i64(i64 %86, i64 576)
-  br label %126
-
-126:                                              ; preds = %118, %122, %125
-  %.0182 = phi i64 [ %.sroa.speculated128, %125 ], [ %86, %122 ], [ %86, %118 ]
-  %.093 = phi i64 [ %14, %125 ], [ 1572864, %122 ], [ %13, %118 ]
-  %127 = mul i64 %63, 24
-  %128 = udiv i64 %.093, %127
-  %.sroa.speculated = tail call i64 @llvm.smin.i64(i64 %.0182, i64 %128)
-  %129 = icmp sgt i64 %.sroa.speculated, 4
-  br i1 %129, label %130, label %132
-
-130:                                              ; preds = %126
-  %131 = and i64 %.sroa.speculated, 9223372036854775804
-  br label %134
+  br label %128
 
-132:                                              ; preds = %126
-  %133 = icmp eq i64 %.sroa.speculated, 0
-  br i1 %133, label %.critedge116, label %134
-
-134:                                              ; preds = %132, %130
-  %.0 = phi i64 [ %131, %130 ], [ %.sroa.speculated, %132 ]
-  %135 = srem i64 %86, %.0
-  %136 = sdiv i64 %86, %.0
-  %137 = icmp eq i64 %135, 0
-  br i1 %137, label %.critedge, label %138
-
-138:                                              ; preds = %134
-  %139 = sub nsw i64 %.0, %135
-  %140 = shl i64 %136, 2
-  %141 = add i64 %140, 4
-  %142 = sdiv i64 %139, %141
-  %143 = shl nsw i64 %142, 2
-  %144 = sub nsw i64 %.0, %143
+132:                                              ; preds = %120, %124, %127
+  %.0182 = phi i64 [ %.sroa.speculated128, %127 ], [ %88, %124 ], [ %88, %120 ]
+  %.093 = phi i64 [ %14, %127 ], [ 1572864, %124 ], [ %13, %120 ]
+  %129 = mul i64 %65, 24
+  %130 = udiv i64 %.093, %129
+  %.sroa.speculated = tail call i64 @llvm.smin.i64(i64 %.0182, i64 %130)
+  %131 = icmp sgt i64 %.sroa.speculated, 4
+  br i1 %131, label %132, label %134
+
+132:; preds = %128
+  %133 = and i64 %.sroa.speculated, 9223372036854775804
+  br label %136
+
+134:; preds = %132
+  %135 = icmp eq i64 %.sroa.speculated, 0
+  br i1 %135, label %.critedge116, label %136
+
+136:; preds = %134, %132
+  %.0 = phi i64 [ %133, %132 ], [ %.sroa.speculated, %134 ]
+  %137 = srem i64 %88, %.0
+  %138 = sdiv i64 %88, %.0
+  %139 = icmp eq i64 %137, 0
+  br i1 %139, label %.critedge, label %140
+
+140:                                              ; preds = %136
+  %141 = sub nsw i64 %.0, %137
+  %142 = shl i64 %138, 2
+  %143 = add i64 %142, 4
+  %144 = sdiv i64 %141, %143
+  %145 = shl nsw i64 %144, 2
+  %146 = sub nsw i64 %.0, %145
   br label %.critedge
 
-.critedge:                                        ; preds = %134, %138
-  %145 = phi i64 [ %144, %138 ], [ %.0, %134 ]
+.critedge:                                        ; preds = %136, %140
+  %145 = phi i64 [ %146, %140 ], [ %.0, %136 ]
   store i64 %145, ptr %1, align 8, !tbaa !176
   br label %.critedge116
 
-.critedge116:                                     ; preds = %.critedge, %116, %114, %132, %53, %55, %40, %59
+.critedge116:                                     ; preds = %.critedge, %116, %114, %134, %53, %55, %40, %59
   ret void
 }
 
@@ -9429,12 +9429,12 @@ declare i64 @llvm.umax.i64(i64, i64) #25
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.smax.i64(i64, i64) #25
 
-; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)
-declare void @llvm.memmove.p0.p0.i64(ptr writeonly captures(none), ptr readonly captures(none), i64, i1 immarg) #28
-
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.umin.i64(i64, i64) #25
 
+; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.memmove.p0.p0.i64(ptr writeonly captures(none), ptr readonly captures(none), i64, i1 immarg) #28
+
 attributes #0 = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+crc32,+cx8,+fxsr,+mmx,+popcnt,+sse,+sse2,+sse3,+sse4.1,+sse4.2,+ssse3,+x87" "tune-cpu"="generic" }
 attributes #1 = { nounwind "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+crc32,+cx8,+fxsr,+mmx,+popcnt,+sse,+sse2,+sse3,+sse4.1,+sse4.2,+ssse3,+x87" "tune-cpu"="generic" }
 attributes #2 = { nofree nounwind }
diff --git a/bench/glslang/optimized/ParseContextBase.ll b/bench/glslang/optimized/ParseContextBase.ll
index 81fb8a05697..fa4c9cfbcb6 100644
--- a/bench/glslang/optimized/ParseContextBase.ll
+++ b/bench/glslang/optimized/ParseContextBase.ll
@@ -2776,10 +2776,10 @@ define void @_ZN7glslang17TParseContextBase20parseSwizzleSelectorERKNS_10TSource
   br i1 %19, label %.lr.ph, label %.preheader
 
 .lr.ph:                                           ; preds = %16
-  %20 = tail call i32 @llvm.umin.i32(i32 %18, i32 4)
-  %21 = getelementptr inbounds nuw i8, ptr %2, i64 8
-  %22 = getelementptr inbounds nuw i8, ptr %4, i64 4
-  %wide.trip.count = zext nneg i32 %20 to i64
+  %20 = getelementptr inbounds nuw i8, ptr %2, i64 8
+  %21 = getelementptr inbounds nuw i8, ptr %4, i64 4
+  %22 = tail call i32 @llvm.umin.i32(i32 %18, i32 4)
+  %wide.trip.count = zext nneg i32 %22 to i64
   br label %26
 
 .preheader:                                       ; preds = %130, %16
@@ -2794,7 +2794,7 @@ define void @_ZN7glslang17TParseContextBase20parseSwizzleSelectorERKNS_10TSource
 
 26:                                               ; preds = %.lr.ph, %130
   %indvars.iv = phi i64 [ 0, %.lr.ph ], [ %indvars.iv.next, %130 ]
-  %27 = load ptr, ptr %21, align 8
+  %27 = load ptr, ptr %20, align 8
   %28 = getelementptr inbounds nuw i8, ptr %27, i64 %indvars.iv
   %29 = load i8, ptr %28, align 1
   switch i8 %29, label %126 [
@@ -2821,7 +2821,7 @@ define void @_ZN7glslang17TParseContextBase20parseSwizzleSelectorERKNS_10TSource
   %34 = add nsw i32 %31, 1
   store i32 %34, ptr %4, align 4
   %35 = sext i32 %31 to i64
-  %36 = getelementptr inbounds [4 x i32], ptr %22, i64 0, i64 %35
+  %36 = getelementptr inbounds [4 x i32], ptr %21, i64 0, i64 %35
   store i32 0, ptr %36, align 4
   br label %_ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit
 
@@ -2839,7 +2839,7 @@ _ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit: ; preds = %30, %33
   %42 = add nsw i32 %39, 1
   store i32 %42, ptr %4, align 4
   %43 = sext i32 %39 to i64
-  %44 = getelementptr inbounds [4 x i32], ptr %22, i64 0, i64 %43
+  %44 = getelementptr inbounds [4 x i32], ptr %21, i64 0, i64 %43
   store i32 0, ptr %44, align 4
   br label %_ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit57
 
@@ -2857,7 +2857,7 @@ _ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit57: ; preds = %38, %41
   %50 = add nsw i32 %47, 1
   store i32 %50, ptr %4, align 4
   %51 = sext i32 %47 to i64
-  %52 = getelementptr inbounds [4 x i32], ptr %22, i64 0, i64 %51
+  %52 = getelementptr inbounds [4 x i32], ptr %21, i64 0, i64 %51
   store i32 0, ptr %52, align 4
   br label %_ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit58
 
@@ -2875,7 +2875,7 @@ _ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit58: ; preds = %46, %49
   %58 = add nsw i32 %55, 1
   store i32 %58, ptr %4, align 4
   %59 = sext i32 %55 to i64
-  %60 = getelementptr inbounds [4 x i32], ptr %22, i64 0, i64 %59
+  %60 = getelementptr inbounds [4 x i32], ptr %21, i64 0, i64 %59
   store i32 1, ptr %60, align 4
   br label %_ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit59
 
@@ -2893,7 +2893,7 @@ _ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit59: ; preds = %54, %57
   %66 = add nsw i32 %63, 1
   store i32 %66, ptr %4, align 4
   %67 = sext i32 %63 to i64
-  %68 = getelementptr inbounds [4 x i32], ptr %22, i64 0, i64 %67
+  %68 = getelementptr inbounds [4 x i32], ptr %21, i64 0, i64 %67
   store i32 1, ptr %68, align 4
   br label %_ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit60
 
@@ -2911,7 +2911,7 @@ _ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit60: ; preds = %62, %65
   %74 = add nsw i32 %71, 1
   store i32 %74, ptr %4, align 4
   %75 = sext i32 %71 to i64
-  %76 = getelementptr inbounds [4 x i32], ptr %22, i64 0, i64 %75
+  %76 = getelementptr inbounds [4 x i32], ptr %21, i64 0, i64 %75
   store i32 1, ptr %76, align 4
   br label %_ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit61
 
@@ -2929,7 +2929,7 @@ _ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit61: ; preds = %70, %73
   %82 = add nsw i32 %79, 1
   store i32 %82, ptr %4, align 4
   %83 = sext i32 %79 to i64
-  %84 = getelementptr inbounds [4 x i32], ptr %22, i64 0, i64 %83
+  %84 = getelementptr inbounds [4 x i32], ptr %21, i64 0, i64 %83
   store i32 2, ptr %84, align 4
   br label %_ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit62
 
@@ -2947,7 +2947,7 @@ _ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit62: ; preds = %78, %81
   %90 = add nsw i32 %87, 1
   store i32 %90, ptr %4, align 4
   %91 = sext i32 %87 to i64
-  %92 = getelementptr inbounds [4 x i32], ptr %22, i64 0, i64 %91
+  %92 = getelementptr inbounds [4 x i32], ptr %21, i64 0, i64 %91
   store i32 2, ptr %92, align 4
   br label %_ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit63
 
@@ -2965,7 +2965,7 @@ _ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit63: ; preds = %86, %89
   %98 = add nsw i32 %95, 1
   store i32 %98, ptr %4, align 4
   %99 = sext i32 %95 to i64
-  %100 = getelementptr inbounds [4 x i32], ptr %22, i64 0, i64 %99
+  %100 = getelementptr inbounds [4 x i32], ptr %21, i64 0, i64 %99
   store i32 2, ptr %100, align 4
   br label %_ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit64
 
@@ -2983,7 +2983,7 @@ _ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit64: ; preds = %94, %97
   %106 = add nsw i32 %103, 1
   store i32 %106, ptr %4, align 4
   %107 = sext i32 %103 to i64
-  %108 = getelementptr inbounds [4 x i32], ptr %22, i64 0, i64 %107
+  %108 = getelementptr inbounds [4 x i32], ptr %21, i64 0, i64 %107
   store i32 3, ptr %108, align 4
   br label %_ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit65
 
@@ -3001,7 +3001,7 @@ _ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit65: ; preds = %102, %105
   %114 = add nsw i32 %111, 1
   store i32 %114, ptr %4, align 4
   %115 = sext i32 %111 to i64
-  %116 = getelementptr inbounds [4 x i32], ptr %22, i64 0, i64 %115
+  %116 = getelementptr inbounds [4 x i32], ptr %21, i64 0, i64 %115
   store i32 3, ptr %116, align 4
   br label %_ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit66
 
@@ -3019,7 +3019,7 @@ _ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit66: ; preds = %110, %113
   %122 = add nsw i32 %119, 1
   store i32 %122, ptr %4, align 4
   %123 = sext i32 %119 to i64
-  %124 = getelementptr inbounds [4 x i32], ptr %22, i64 0, i64 %123
+  %124 = getelementptr inbounds [4 x i32], ptr %21, i64 0, i64 %123
   store i32 3, ptr %124, align 4
   br label %_ZN7glslang17TSwizzleSelectorsIiE9push_backEi.exit67
 
diff --git a/bench/gromacs/optimized/xtc2.ll b/bench/gromacs/optimized/xtc2.ll
index 839d78b711e..4f2850d6aa6 100644
--- a/bench/gromacs/optimized/xtc2.ll
+++ b/bench/gromacs/optimized/xtc2.ll
@@ -448,12 +448,12 @@ positive_int.exit480:                             ; preds = %142, %145, %147
   br label %167
 
 167:                                              ; preds = %.lr.ph747, %627
-  %.0378745 = phi i32 [ %119, %.lr.ph747 ], [ %.1379, %627 ]
-  %.0384744 = phi i32 [ 0, %.lr.ph747 ], [ %.1385, %627 ]
-  %.0388743 = phi i32 [ %23, %.lr.ph747 ], [ %.2390, %627 ]
-  %.0393742 = phi ptr [ %1, %.lr.ph747 ], [ %.2395, %627 ]
-  %.0400741 = phi i32 [ 0, %.lr.ph747 ], [ %.1401, %627 ]
-  %.0740 = phi i32 [ 0, %.lr.ph747 ], [ %.1628, %627 ]
+  %.0378745 = phi i32 [ %119, %.lr.ph747 ], [ %.1379, %628 ]
+  %.0384744 = phi i32 [ 0, %.lr.ph747 ], [ %.1385, %628 ]
+  %.0388743 = phi i32 [ %23, %.lr.ph747 ], [ %.2390, %628 ]
+  %.0393742 = phi ptr [ %1, %.lr.ph747 ], [ %.2395, %628 ]
+  %.0400741 = phi i32 [ 0, %.lr.ph747 ], [ %.1401, %628 ]
+  %.0740 = phi i32 [ 0, %.lr.ph747 ], [ %.1628, %628 ]
   %168 = icmp slt i32 %.0388743, 0
   br i1 %168, label %169, label %172
 
@@ -1507,8 +1507,8 @@ Ptngc_find_magic_index.exit558:                   ; preds = %507
 .critedge:                                        ; preds = %553, %.critedge
   %.6956 = phi i32 [ %560, %.critedge ], [ %.5, %553 ]
   %.4382955 = phi i32 [ %561, %.critedge ], [ %.0378745, %553 ]
-  %spec.store.select10 = call i32 @llvm.smin.i32(i32 %.6956, i32 2)
-  %spec.store.select15 = call i32 @llvm.smax.i32(i32 %spec.store.select10, i32 -2)
+  %spec.store.select10 = call i32 @llvm.smax.i32(i32 %.6956, i32 -2)
+  %spec.store.select15 = call i32 @llvm.smin.i32(i32 %spec.store.select10, i32 2)
   %560 = sub nsw i32 %.6956, %spec.store.select15
   %561 = add nsw i32 %spec.store.select15, %.4382955
   %562 = icmp slt i32 %.6956, 0
@@ -1545,8 +1545,8 @@ Ptngc_find_magic_index.exit558:                   ; preds = %507
   br label %575
 
 575:                                              ; preds = %._crit_edge957, %569, %._crit_edge717, %523
-  %.2402 = phi i32 [ %.0400741, %523 ], [ %.0357, %569 ], [ %.0400741, %._crit_edge957 ], [ %.0400741, %._crit_edge717 ]
-  %.2380 = phi i32 [ %.0378745, %523 ], [ %574, %569 ], [ %.4382.lcssa, %._crit_edge957 ], [ %.0378745, %._crit_edge717 ]
+  %.2402 = phi i32 [ %.0400741, %523 ], [ %.0357, %570 ], [ %.0400741, %._crit_edge957 ], [ %.0400741, %._crit_edge717 ]
+  %.2380 = phi i32 [ %.0378745, %523 ], [ %574, %570 ], [ %.4382.lcssa, %._crit_edge957 ], [ %.0378745, %._crit_edge717 ]
   %576 = load i32, ptr %17, align 4, !tbaa !3
   %.not429 = icmp eq i32 %576, 0
   br i1 %.not429, label %585, label %577
@@ -1640,9 +1640,9 @@ Ptngc_find_magic_index.exit558:                   ; preds = %507
   br i1 %exitcond834.not, label %._crit_edge727, label %.lr.ph726, !llvm.loop !49
 
 ._crit_edge727:                                   ; preds = %.lr.ph726, %586
-  %.lcssa723 = phi i32 [ %.promoted722888, %586 ], [ %623, %.lr.ph726 ]
-  %.lcssa721 = phi i32 [ %.promoted720890, %586 ], [ %614, %.lr.ph726 ]
-  %.lcssa719 = phi i32 [ %.promoted892, %586 ], [ %605, %.lr.ph726 ]
+  %.lcssa723 = phi i32 [ %.promoted722888, %587 ], [ %623, %.lr.ph726 ]
+  %.lcssa721 = phi i32 [ %.promoted720890, %587 ], [ %614, %.lr.ph726 ]
+  %.lcssa719 = phi i32 [ %.promoted892, %587 ], [ %605, %.lr.ph726 ]
   store i32 %.lcssa719, ptr %19, align 4
   store i32 %.lcssa721, ptr %151, align 4
   store i32 %.lcssa723, ptr %152, align 4
diff --git a/bench/grpc/optimized/compression_internal.ll b/bench/grpc/optimized/compression_internal.ll
index 157f4bd96e1..53665691460 100644
--- a/bench/grpc/optimized/compression_internal.ll
+++ b/bench/grpc/optimized/compression_internal.ll
@@ -996,8 +996,8 @@ define void @_ZN9grpc_core33CompressionOptionsFromChannelArgsERKNS_11ChannelArgs
   %.sroa.015.0.extract.trunc = trunc i64 %3 to i32
   %6 = getelementptr inbounds nuw i8, ptr %0, i64 4
   store i32 1, ptr %6, align 4, !tbaa !79
-  %..i = tail call i32 @llvm.smin.i32(i32 %.sroa.015.0.extract.trunc, i32 3)
-  %.0.i = tail call i32 @llvm.smax.i32(i32 %..i, i32 0)
+  %..i = tail call i32 @llvm.smax.i32(i32 %.sroa.015.0.extract.trunc, i32 0)
+  %.0.i = tail call i32 @llvm.umin.i32(i32 %..i, i32 3)
   %7 = getelementptr inbounds nuw i8, ptr %0, i64 8
   store i32 %.0.i, ptr %7, align 4, !tbaa !85
   br label %8
@@ -1012,8 +1012,8 @@ define void @_ZN9grpc_core33CompressionOptionsFromChannelArgsERKNS_11ChannelArgs
   %.sroa.08.0.extract.trunc = trunc i64 %9 to i32
   %12 = getelementptr inbounds nuw i8, ptr %0, i64 12
   store i32 1, ptr %12, align 4, !tbaa !86
-  %..i3 = tail call i32 @llvm.smin.i32(i32 %.sroa.08.0.extract.trunc, i32 2)
-  %.0.i4 = tail call i32 @llvm.smax.i32(i32 %..i3, i32 0)
+  %..i3 = tail call i32 @llvm.smax.i32(i32 %.sroa.08.0.extract.trunc, i32 0)
+  %.0.i4 = tail call i32 @llvm.umin.i32(i32 %..i3, i32 2)
   %13 = getelementptr inbounds nuw i8, ptr %0, i64 16
   store i32 %.0.i4, ptr %13, align 4, !tbaa !87
   br label %14
@@ -1419,10 +1419,10 @@ declare i64 @llvm.umin.i64(i64, i64) #26
 declare i32 @bcmp(ptr captures(none), ptr captures(none), i64) local_unnamed_addr #27
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smin.i32(i32, i32) #26
+declare i32 @llvm.smax.i32(i32, i32) #26
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smax.i32(i32, i32) #26
+declare i32 @llvm.umin.i32(i32, i32) #26
 
 attributes #0 = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { nounwind "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
diff --git a/bench/grpc/optimized/flow_control.ll b/bench/grpc/optimized/flow_control.ll
index 4ee00b7ba85..a5aa735f90d 100644
--- a/bench/grpc/optimized/flow_control.ll
+++ b/bench/grpc/optimized/flow_control.ll
@@ -1139,8 +1139,8 @@ define noundef range(i32 0, -2147483648) i32 @_ZNK9grpc_core6chttp220TransportFl
 
 10:                                               ; preds = %2
   %11 = sub nsw i64 %8, %.pre
-  %..i = tail call i64 @llvm.smin.i64(i64 %11, i64 2147483647)
-  %.0.i = tail call i64 @llvm.smax.i64(i64 %..i, i64 0)
+  %..i = tail call i64 @llvm.smax.i64(i64 %11, i64 0)
+  %.0.i = tail call i64 @llvm.umin.i64(i64 %..i, i64 2147483647)
   %12 = trunc nuw nsw i64 %.0.i to i32
   br label %13
 
@@ -2182,8 +2182,8 @@ define noundef range(i32 0, -2147483648) i32 @_ZNK9grpc_core6chttp217StreamFlowC
   %17 = phi i64 [ %.pre, %15 ], [ %.pre.i, %._crit_edge.i ], [ %12, %9 ]
   %.0.i = phi i64 [ %16, %15 ], [ %.pre.i, %._crit_edge.i ], [ %spec.select.i, %9 ]
   %18 = sub nsw i64 %.0.i, %17
-  %..i = tail call i64 @llvm.smin.i64(i64 %18, i64 2147483647)
-  %.0.i1 = tail call i64 @llvm.smax.i64(i64 %..i, i64 0)
+  %..i = tail call i64 @llvm.smax.i64(i64 %18, i64 0)
+  %.0.i1 = tail call i64 @llvm.umin.i64(i64 %..i, i64 2147483647)
   %19 = trunc nuw nsw i64 %.0.i1 to i32
   ret i32 %19
 }
diff --git a/bench/icu/optimized/csrucode.ll b/bench/icu/optimized/csrucode.ll
index 70ba8b6778f..c5ce25094a2 100644
--- a/bench/icu/optimized/csrucode.ll
+++ b/bench/icu/optimized/csrucode.ll
@@ -85,8 +85,8 @@ define noundef signext range(i8 0, 2) i8 @_ZNK6icu_7722CharsetRecog_UTF_16_BE5ma
   br i1 %10, label %.lr.ph, label %.thread
 
 .lr.ph:                                           ; preds = %3, %29
-  %indvars.iv = phi i64 [ %indvars.iv.next, %29 ], [ 0, %3 ]
-  %.036 = phi i32 [ %.1.i, %29 ], [ 10, %3 ]
+  %indvars.iv = phi i64 [ %indvars.iv.next, %31 ], [ 0, %3 ]
+  %.036 = phi i32 [ %.1.i, %31 ], [ 10, %3 ]
   %11 = getelementptr inbounds nuw i8, ptr %5, i64 %indvars.iv
   %12 = load i8, ptr %11, align 1, !tbaa !12
   %13 = zext i8 %12 to i16
@@ -119,9 +119,9 @@ define noundef signext range(i8 0, 2) i8 @_ZNK6icu_7722CharsetRecog_UTF_16_BE5ma
 
 _ZN6icu_77L16adjustConfidenceEDsi.exit:           ; preds = %23, %25
   %.0.i = phi i32 [ %24, %23 ], [ %spec.select.i, %25 ]
-  %spec.store.select.i = tail call i32 @llvm.smin.i32(i32 %.0.i, i32 100)
-  %.1.i = tail call noundef range(i32 0, 101) i32 @llvm.smax.i32(i32 %spec.store.select.i, i32 0)
-  switch i32 %.1.i, label %29 [
+  %spec.store.select.i = tail call i32 @llvm.smax.i32(i32 %.0.i, i32 0)
+  %.1.i = tail call noundef range(i32 0, 101) i32 @llvm.umin.i32(i32 %spec.store.select.i, i32 100)
+  switch i32 %.1.i, label %31 [
     i32 100, label %.thread
     i32 0, label %.thread
   ]
@@ -133,7 +133,7 @@ _ZN6icu_77L16adjustConfidenceEDsi.exit:           ; preds = %23, %25
   br i1 %31, label %.lr.ph, label %.thread, !llvm.loop !13
 
 .thread:                                          ; preds = %29, %.lr.ph, %_ZN6icu_77L16adjustConfidenceEDsi.exit, %_ZN6icu_77L16adjustConfidenceEDsi.exit, %3
-  %.1 = phi i32 [ 10, %3 ], [ %.1.i, %_ZN6icu_77L16adjustConfidenceEDsi.exit ], [ %.1.i, %_ZN6icu_77L16adjustConfidenceEDsi.exit ], [ 100, %.lr.ph ], [ %.1.i, %29 ]
+  %.1 = phi i32 [ 10, %3 ], [ %.1.i, %_ZN6icu_77L16adjustConfidenceEDsi.exit ], [ %.1.i, %_ZN6icu_77L16adjustConfidenceEDsi.exit ], [ 100, %.lr.ph ], [ %.1.i, %31 ]
   %32 = icmp slt i32 %7, 4
   %33 = icmp samesign ult i32 %.1, 100
   %or.cond6 = and i1 %32, %33
@@ -170,8 +170,8 @@ define noundef signext range(i8 0, 2) i8 @_ZNK6icu_7722CharsetRecog_UTF_16_LE5ma
   br i1 %10, label %.lr.ph, label %.thread
 
 .lr.ph:                                           ; preds = %3, %39
-  %indvars.iv = phi i64 [ %indvars.iv.next, %39 ], [ 0, %3 ]
-  %.039 = phi i32 [ %.1.i, %39 ], [ 10, %3 ]
+  %indvars.iv = phi i64 [ %indvars.iv.next, %41 ], [ 0, %3 ]
+  %.039 = phi i32 [ %.1.i, %41 ], [ 10, %3 ]
   %11 = getelementptr inbounds nuw i8, ptr %5, i64 %indvars.iv
   %12 = load i8, ptr %11, align 1, !tbaa !12
   %13 = zext i8 %12 to i16
@@ -221,9 +221,9 @@ define noundef signext range(i8 0, 2) i8 @_ZNK6icu_7722CharsetRecog_UTF_16_LE5ma
 
 _ZN6icu_77L16adjustConfidenceEDsi.exit:           ; preds = %33, %35
   %.0.i = phi i32 [ %34, %33 ], [ %spec.select.i, %35 ]
-  %spec.store.select.i = tail call i32 @llvm.smin.i32(i32 %.0.i, i32 100)
-  %.1.i = tail call noundef range(i32 0, 101) i32 @llvm.smax.i32(i32 %spec.store.select.i, i32 0)
-  switch i32 %.1.i, label %39 [
+  %spec.store.select.i = tail call i32 @llvm.smax.i32(i32 %.0.i, i32 0)
+  %.1.i = tail call noundef range(i32 0, 101) i32 @llvm.umin.i32(i32 %spec.store.select.i, i32 100)
+  switch i32 %.1.i, label %41 [
     i32 100, label %.thread
     i32 0, label %.thread
   ]
@@ -235,7 +235,7 @@ _ZN6icu_77L16adjustConfidenceEDsi.exit:           ; preds = %33, %35
   br i1 %41, label %.lr.ph, label %.thread, !llvm.loop !15
 
 .thread:                                          ; preds = %39, %_ZN6icu_77L16adjustConfidenceEDsi.exit, %_ZN6icu_77L16adjustConfidenceEDsi.exit, %3, %21, %23, %27
-  %.1 = phi i32 [ %spec.select, %27 ], [ 100, %21 ], [ 100, %23 ], [ 10, %3 ], [ %.1.i, %_ZN6icu_77L16adjustConfidenceEDsi.exit ], [ %.1.i, %_ZN6icu_77L16adjustConfidenceEDsi.exit ], [ %.1.i, %39 ]
+  %.1 = phi i32 [ %spec.select, %27 ], [ 100, %21 ], [ 100, %23 ], [ 10, %3 ], [ %.1.i, %_ZN6icu_77L16adjustConfidenceEDsi.exit ], [ %.1.i, %_ZN6icu_77L16adjustConfidenceEDsi.exit ], [ %.1.i, %41 ]
   %42 = icmp slt i32 %7, 4
   %43 = icmp samesign ult i32 %.1, 100
   %or.cond6 = and i1 %42, %43
@@ -392,6 +392,9 @@ declare i32 @llvm.smin.i32(i32, i32) #8
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smax.i32(i32, i32) #8
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #8
+
 attributes #0 = { nounwind "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { mustprogress nounwind uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #2 = { cold mustprogress noreturn nounwind memory(inaccessiblemem: write) uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
diff --git a/bench/image-rs/optimized/254ue5dpb10tdnze.ll b/bench/image-rs/optimized/254ue5dpb10tdnze.ll
index 56c900a74c6..47a49b552d0 100644
--- a/bench/image-rs/optimized/254ue5dpb10tdnze.ll
+++ b/bench/image-rs/optimized/254ue5dpb10tdnze.ll
@@ -15575,8 +15575,8 @@ define hidden void @_ZN5image8imageops8colorops8brighten17h35be5e44376220d5E(ptr
   %26 = add nuw i32 %23, 1
   %27 = zext i8 %.sroa.0.0.copyload.i.i to i32
   %28 = add i32 %2, %27
-  %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i = tail call i32 @llvm.smin.i32(i32 %28, i32 255)
-  %.0.in.sroa.speculated.i.i.i.i = tail call i32 @llvm.smax.i32(i32 %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i, i32 0)
+  %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i = tail call i32 @llvm.smax.i32(i32 %28, i32 0)
+  %.0.in.sroa.speculated.i.i.i.i = tail call i32 @llvm.umin.i32(i32 %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i, i32 255)
   %29 = trunc nuw i32 %.0.in.sroa.speculated.i.i.i.i to i8
   %30 = icmp uge i32 %23, %10
   %31 = icmp uge i32 %.sroa.8.138, %12
@@ -15596,11 +15596,11 @@ define hidden void @_ZN5image8imageops8colorops8brighten17h35be5e44376220d5E(ptr
   store i32 %12, ptr %34, align 4, !noalias !1668
   store ptr %5, ptr %6, align 8, !noalias !1668
   %35 = getelementptr inbounds nuw i8, ptr %6, i64 8
-  store ptr @"_ZN50_$LT$$LP$U$C$T$RP$$u20$as$u20$core..fmt..Debug$GT$3fmt17h297b96a1d93efb91E", ptr %35, align 8, !noalias !1668
-  %36 = getelementptr inbounds nuw i8, ptr %6, i64 16
-  store ptr %4, ptr %36, align 8, !noalias !1668
-  %37 = getelementptr inbounds nuw i8, ptr %6, i64 24
   store ptr @"_ZN50_$LT$$LP$U$C$T$RP$$u20$as$u20$core..fmt..Debug$GT$3fmt17h297b96a1d93efb91E", ptr %37, align 8, !noalias !1668
+  %38 = getelementptr inbounds nuw i8, ptr %6, i64 16
+  store ptr %4, ptr %38, align 8, !noalias !1668
+  %39 = getelementptr inbounds nuw i8, ptr %6, i64 24
+  store ptr @"_ZN50_$LT$$LP$U$C$T$RP$$u20$as$u20$core..fmt..Debug$GT$3fmt17h297b96a1d93efb91E", ptr %39, align 8, !noalias !1668
   store ptr @anon.f0c04611e5e2b4be9a0dda14dd5d7c38.45, ptr %7, align 8, !alias.scope !1672, !noalias !1675
   %38 = getelementptr inbounds nuw i8, ptr %7, i64 8
   store i64 2, ptr %38, align 8, !alias.scope !1672, !noalias !1675
@@ -15723,8 +15723,8 @@ define hidden void @_ZN5image8imageops8colorops8brighten17h48e73d165d4f9ba6E(ptr
   %27 = add nuw i32 %24, 1
   %28 = zext i16 %.sroa.0.0.copyload.i.i to i32
   %29 = add i32 %2, %28
-  %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i = tail call i32 @llvm.smin.i32(i32 %29, i32 65535)
-  %.0.in.sroa.speculated.i.i.i.i = tail call i32 @llvm.smax.i32(i32 %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i, i32 0)
+  %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i = tail call i32 @llvm.smax.i32(i32 %29, i32 0)
+  %.0.in.sroa.speculated.i.i.i.i = tail call i32 @llvm.umin.i32(i32 %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i, i32 65535)
   %30 = trunc nuw i32 %.0.in.sroa.speculated.i.i.i.i to i16
   tail call void @llvm.experimental.noalias.scope.decl(metadata !1687)
   %31 = icmp uge i32 %24, %10
@@ -15745,11 +15745,11 @@ define hidden void @_ZN5image8imageops8colorops8brighten17h48e73d165d4f9ba6E(ptr
   store i32 %12, ptr %35, align 4, !noalias !1690
   store ptr %5, ptr %6, align 8, !noalias !1690
   %36 = getelementptr inbounds nuw i8, ptr %6, i64 8
-  store ptr @"_ZN50_$LT$$LP$U$C$T$RP$$u20$as$u20$core..fmt..Debug$GT$3fmt17h297b96a1d93efb91E", ptr %36, align 8, !noalias !1690
-  %37 = getelementptr inbounds nuw i8, ptr %6, i64 16
-  store ptr %4, ptr %37, align 8, !noalias !1690
-  %38 = getelementptr inbounds nuw i8, ptr %6, i64 24
   store ptr @"_ZN50_$LT$$LP$U$C$T$RP$$u20$as$u20$core..fmt..Debug$GT$3fmt17h297b96a1d93efb91E", ptr %38, align 8, !noalias !1690
+  %39 = getelementptr inbounds nuw i8, ptr %6, i64 16
+  store ptr %4, ptr %39, align 8, !noalias !1690
+  %40 = getelementptr inbounds nuw i8, ptr %6, i64 24
+  store ptr @"_ZN50_$LT$$LP$U$C$T$RP$$u20$as$u20$core..fmt..Debug$GT$3fmt17h297b96a1d93efb91E", ptr %40, align 8, !noalias !1690
   store ptr @anon.f0c04611e5e2b4be9a0dda14dd5d7c38.45, ptr %7, align 8, !alias.scope !1692, !noalias !1695
   %39 = getelementptr inbounds nuw i8, ptr %7, i64 8
   store i64 2, ptr %39, align 8, !alias.scope !1692, !noalias !1695
@@ -16383,7 +16383,7 @@ define hidden void @_ZN5image8imageops8colorops8brighten17h75f48f95ff6b6f5bE(ptr
 15:                                               ; preds = %.loopexit.split-lp, %.loopexit
   %lpad.phi = phi { ptr, i32 } [ %lpad.loopexit, %.loopexit ], [ %lpad.loopexit.split-lp, %.loopexit.split-lp ]
   invoke void @"_ZN4core3ptr113drop_in_place$LT$image..buffer_..ImageBuffer$LT$image..color..LumaA$LT$u8$GT$$C$alloc..vec..Vec$LT$u8$GT$$GT$$GT$17ha1a53ec1e6e57852E"(ptr noalias noundef nonnull align 8 dereferenceable(32) %8) #19
-          to label %59 unwind label %57
+          to label %61 unwind label %57
 
 16:                                               ; preds = %3
   %.fca.1.extract.i = extractvalue { i64, i64 } %13, 1
@@ -16425,8 +16425,8 @@ define hidden void @_ZN5image8imageops8colorops8brighten17h75f48f95ff6b6f5bE(ptr
   %27 = and i16 %.sroa.0.0.copyload.i.i, 255
   %28 = zext nneg i16 %27 to i32
   %29 = add i32 %2, %28
-  %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i = tail call i32 @llvm.smin.i32(i32 %29, i32 255)
-  %.0.in.sroa.speculated.i.i.i.i = tail call i32 @llvm.smax.i32(i32 %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i, i32 0)
+  %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i = tail call i32 @llvm.smax.i32(i32 %29, i32 0)
+  %.0.in.sroa.speculated.i.i.i.i = tail call i32 @llvm.umin.i32(i32 %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i, i32 255)
   %.sroa.0.0.insert.ext.i = trunc nuw nsw i32 %.0.in.sroa.speculated.i.i.i.i to i16
   %.sroa.0.0.insert.mask.i = and i16 %.sroa.0.0.copyload.i.i, -256
   %.sroa.0.0.insert.insert.i = or disjoint i16 %.sroa.0.0.insert.mask.i, %.sroa.0.0.insert.ext.i
@@ -16448,11 +16448,11 @@ define hidden void @_ZN5image8imageops8colorops8brighten17h75f48f95ff6b6f5bE(ptr
   store i32 %12, ptr %34, align 4, !noalias !1810
   store ptr %5, ptr %6, align 8, !noalias !1810
   %35 = getelementptr inbounds nuw i8, ptr %6, i64 8
-  store ptr @"_ZN50_$LT$$LP$U$C$T$RP$$u20$as$u20$core..fmt..Debug$GT$3fmt17h297b96a1d93efb91E", ptr %35, align 8, !noalias !1810
-  %36 = getelementptr inbounds nuw i8, ptr %6, i64 16
-  store ptr %4, ptr %36, align 8, !noalias !1810
-  %37 = getelementptr inbounds nuw i8, ptr %6, i64 24
   store ptr @"_ZN50_$LT$$LP$U$C$T$RP$$u20$as$u20$core..fmt..Debug$GT$3fmt17h297b96a1d93efb91E", ptr %37, align 8, !noalias !1810
+  %38 = getelementptr inbounds nuw i8, ptr %6, i64 16
+  store ptr %4, ptr %38, align 8, !noalias !1810
+  %39 = getelementptr inbounds nuw i8, ptr %6, i64 24
+  store ptr @"_ZN50_$LT$$LP$U$C$T$RP$$u20$as$u20$core..fmt..Debug$GT$3fmt17h297b96a1d93efb91E", ptr %39, align 8, !noalias !1810
   store ptr @anon.f0c04611e5e2b4be9a0dda14dd5d7c38.45, ptr %7, align 8, !alias.scope !1814, !noalias !1817
   %38 = getelementptr inbounds nuw i8, ptr %7, i64 8
   store i64 2, ptr %38, align 8, !alias.scope !1814, !noalias !1817
@@ -16551,7 +16551,7 @@ define hidden void @_ZN5image8imageops8colorops8brighten17hdbae2855b93cb8bbE(ptr
 15:                                               ; preds = %.loopexit.split-lp, %.loopexit
   %lpad.phi = phi { ptr, i32 } [ %lpad.loopexit, %.loopexit ], [ %lpad.loopexit.split-lp, %.loopexit.split-lp ]
   invoke void @"_ZN4core3ptr115drop_in_place$LT$image..buffer_..ImageBuffer$LT$image..color..LumaA$LT$u16$GT$$C$alloc..vec..Vec$LT$u16$GT$$GT$$GT$17h11586b09306c37f2E"(ptr noalias noundef nonnull align 8 dereferenceable(32) %8) #19
-          to label %59 unwind label %57
+          to label %61 unwind label %57
 
 16:                                               ; preds = %3
   %.fca.1.extract.i = extractvalue { i64, i64 } %13, 1
@@ -16587,8 +16587,8 @@ define hidden void @_ZN5image8imageops8colorops8brighten17hdbae2855b93cb8bbE(ptr
   %27 = add nuw i32 %24, 1
   %28 = and i32 %.sroa.0.0.copyload.i.i, 65535
   %29 = add i32 %28, %2
-  %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i = tail call i32 @llvm.smin.i32(i32 %29, i32 65535)
-  %.0.in.sroa.speculated.i.i.i.i = tail call i32 @llvm.smax.i32(i32 %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i, i32 0)
+  %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i = tail call i32 @llvm.smax.i32(i32 %29, i32 0)
+  %.0.in.sroa.speculated.i.i.i.i = tail call i32 @llvm.umin.i32(i32 %.0.in.sroa.speculate.load.3.sroa.speculated.i.i.i.i, i32 65535)
   %.sroa.0.0.insert.mask.i = and i32 %.sroa.0.0.copyload.i.i, -65536
   %.sroa.0.0.insert.insert.i = or disjoint i32 %.0.in.sroa.speculated.i.i.i.i, %.sroa.0.0.insert.mask.i
   tail call void @llvm.experimental.noalias.scope.decl(metadata !1829)
@@ -16610,11 +16610,11 @@ define hidden void @_ZN5image8imageops8colorops8brighten17hdbae2855b93cb8bbE(ptr
   store i32 %12, ptr %34, align 4, !noalias !1832
   store ptr %5, ptr %6, align 8, !noalias !1832
   %35 = getelementptr inbounds nuw i8, ptr %6, i64 8
-  store ptr @"_ZN50_$LT$$LP$U$C$T$RP$$u20$as$u20$core..fmt..Debug$GT$3fmt17h297b96a1d93efb91E", ptr %35, align 8, !noalias !1832
-  %36 = getelementptr inbounds nuw i8, ptr %6, i64 16
-  store ptr %4, ptr %36, align 8, !noalias !1832
-  %37 = getelementptr inbounds nuw i8, ptr %6, i64 24
   store ptr @"_ZN50_$LT$$LP$U$C$T$RP$$u20$as$u20$core..fmt..Debug$GT$3fmt17h297b96a1d93efb91E", ptr %37, align 8, !noalias !1832
+  %38 = getelementptr inbounds nuw i8, ptr %6, i64 16
+  store ptr %4, ptr %38, align 8, !noalias !1832
+  %39 = getelementptr inbounds nuw i8, ptr %6, i64 24
+  store ptr @"_ZN50_$LT$$LP$U$C$T$RP$$u20$as$u20$core..fmt..Debug$GT$3fmt17h297b96a1d93efb91E", ptr %39, align 8, !noalias !1832
   store ptr @anon.f0c04611e5e2b4be9a0dda14dd5d7c38.45, ptr %7, align 8, !alias.scope !1834, !noalias !1837
   %38 = getelementptr inbounds nuw i8, ptr %7, i64 8
   store i64 2, ptr %38, align 8, !alias.scope !1834, !noalias !1837
diff --git a/bench/libquic/optimized/hybrid_slow_start.ll b/bench/libquic/optimized/hybrid_slow_start.ll
index d9cf14b2982..ecc89d94b02 100644
--- a/bench/libquic/optimized/hybrid_slow_start.ll
+++ b/bench/libquic/optimized/hybrid_slow_start.ll
@@ -118,8 +118,8 @@ define noundef zeroext i1 @_ZN3net15HybridSlowStart19ShouldExitSlowStartENS_8Qui
 
 32:                                               ; preds = %30
   %33 = ashr i64 %4, 3
-  %.sroa.speculated18 = tail call i64 @llvm.smin.i64(i64 %33, i64 16000)
-  %.sroa.speculated = tail call i64 @llvm.smax.i64(i64 %.sroa.speculated18, i64 4000)
+  %.sroa.speculated18 = tail call i64 @llvm.smax.i64(i64 %33, i64 4000)
+  %.sroa.speculated = tail call i64 @llvm.umin.i64(i64 %.sroa.speculated18, i64 16000)
   %34 = add nsw i64 %.sroa.speculated, %4
   %35 = icmp slt i64 %34, %.sroa.26.0.copyload
   br i1 %35, label %36, label %.thread
@@ -129,7 +129,7 @@ define noundef zeroext i1 @_ZN3net15HybridSlowStart19ShouldExitSlowStartENS_8Qui
   br label %.thread
 
 .thread:                                          ; preds = %18, %32, %36, %30
-  %37 = phi i1 [ false, %18 ], [ false, %32 ], [ true, %36 ], [ false, %30 ]
+  %37 = phi i1 [ false, %18 ], [ false, %32 ], [ true, %38 ], [ false, %30 ]
   %38 = icmp ugt i64 %5, 15
   %39 = and i1 %38, %37
   br label %40
@@ -143,10 +143,10 @@ define noundef zeroext i1 @_ZN3net15HybridSlowStart19ShouldExitSlowStartENS_8Qui
 declare void @llvm.memset.p0.i64(ptr writeonly captures(none), i8, i64, i1 immarg) #3
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i64 @llvm.smin.i64(i64, i64) #4
+declare i64 @llvm.smax.i64(i64, i64) #4
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i64 @llvm.smax.i64(i64, i64) #4
+declare i64 @llvm.umin.i64(i64, i64) #4
 
 attributes #0 = { mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: write) uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: readwrite) uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
diff --git a/bench/libwebp/optimized/analysis_enc.ll b/bench/libwebp/optimized/analysis_enc.ll
index 554dad57438..c34ea798001 100644
--- a/bench/libwebp/optimized/analysis_enc.ll
+++ b/bench/libwebp/optimized/analysis_enc.ll
@@ -715,13 +715,13 @@ SmoothSegmentMap.exit.i:                          ; preds = %._crit_edge70.i.i,
   %367 = sub nsw i32 %363, %.060.i.i
   %368 = mul nsw i32 %367, 255
   %369 = sdiv i32 %368, %359
-  %370 = call i32 @llvm.smin.i32(i32 %366, i32 127)
-  %371 = call i32 @llvm.smax.i32(i32 %370, i32 -127)
+  %370 = call i32 @llvm.smax.i32(i32 %366, i32 -127)
+  %371 = call i32 @llvm.smin.i32(i32 %370, i32 127)
   %372 = getelementptr inbounds nuw [4 x %struct.VP8SegmentInfo], ptr %360, i64 0, i64 %indvars.iv52.i.i
   %373 = getelementptr inbounds nuw i8, ptr %372, i64 672
   store i32 %371, ptr %373, align 8, !tbaa !74
-  %374 = call i32 @llvm.smin.i32(i32 %369, i32 255)
-  %375 = call i32 @llvm.smax.i32(i32 %374, i32 0)
+  %374 = call i32 @llvm.smax.i32(i32 %369, i32 0)
+  %375 = call i32 @llvm.umin.i32(i32 %374, i32 255)
   %376 = getelementptr inbounds nuw i8, ptr %372, i64 676
   store i32 %375, ptr %376, align 4, !tbaa !77
   %indvars.iv.next53.i.i = add nuw nsw i64 %indvars.iv52.i.i, 1
diff --git a/bench/libwebp/optimized/frame_dec.ll b/bench/libwebp/optimized/frame_dec.ll
index b4786f3f998..f058cd93a45 100644
--- a/bench/libwebp/optimized/frame_dec.ll
+++ b/bench/libwebp/optimized/frame_dec.ll
@@ -1357,8 +1357,8 @@ define hidden i32 @VP8EnterCritical(ptr noundef %0, ptr noundef %1) local_unname
 
 .split.us.us.i:                                   ; preds = %80, %77, %72
   %.048.us.i = phi i32 [ %75, %72 ], [ %79, %77 ], [ %81, %80 ]
-  %82 = tail call i32 @llvm.smin.i32(i32 %.048.us.i, i32 63)
-  %83 = tail call i32 @llvm.smax.i32(i32 %82, i32 0)
+  %82 = tail call i32 @llvm.smax.i32(i32 %.048.us.i, i32 0)
+  %83 = tail call i32 @llvm.umin.i32(i32 %82, i32 63)
   %.not56.us.us.i = icmp slt i32 %.048.us.i, 1
   %84 = shl nuw nsw i32 %83, 1
   %85 = icmp samesign ugt i32 %.048.us.i, 39
@@ -1392,9 +1392,9 @@ define hidden i32 @VP8EnterCritical(ptr noundef %0, ptr noundef %1) local_unname
   br label %.split59.us.us.i
 
 .split.us.split.split.us65.i:                     ; preds = %.split.us.split.us64.i
-  %spec.store.select.us.us66.i = tail call i32 @llvm.smax.i32(i32 %82, i32 1)
+  %spec.store.select.us.us66.i = tail call i32 @llvm.umax.i32(i32 %83, i32 1)
   %96 = trunc nuw nsw i32 %spec.store.select.us.us66.i to i8
-  %97 = add nuw nsw i32 %84, %spec.store.select.us.us66.i
+  %97 = add nuw nsw i32 %spec.store.select.us.us66.i, %84
   %98 = trunc nuw i32 %97 to i8
   %99 = getelementptr inbounds nuw [4 x [2 x %struct.VP8FInfo]], ptr %64, i64 0, i64 %indvars.iv84.i, i64 0
   %100 = getelementptr inbounds nuw i8, ptr %99, i64 1
@@ -1457,8 +1457,8 @@ define hidden i32 @VP8EnterCritical(ptr noundef %0, ptr noundef %1) local_unname
   br label %124
 
 124:                                              ; preds = %150, %.split.i.us
-  %.not55.i.us = phi i1 [ true, %.split.i.us ], [ false, %150 ]
-  %indvars.iv.i.us = phi i64 [ 0, %.split.i.us ], [ 1, %150 ]
+  %.not55.i.us = phi i1 [ true, %.split.i.us ], [ false, %151 ]
+  %indvars.iv.i.us = phi i64 [ 0, %.split.i.us ], [ 1, %151 ]
   %125 = getelementptr inbounds nuw [4 x [2 x %struct.VP8FInfo]], ptr %64, i64 0, i64 %indvars.iv72.i.us, i64 %indvars.iv.i.us
   br i1 %.not55.i.us, label %129, label %126
 
@@ -1468,9 +1468,9 @@ define hidden i32 @VP8EnterCritical(ptr noundef %0, ptr noundef %1) local_unname
   br label %129
 
 129:                                              ; preds = %126, %124
-  %.046.i.us = phi i32 [ %128, %126 ], [ %123, %124 ]
-  %130 = tail call i32 @llvm.smin.i32(i32 %.046.i.us, i32 63)
-  %131 = tail call i32 @llvm.smax.i32(i32 %130, i32 0)
+  %.046.i.us = phi i32 [ %128, %127 ], [ %123, %125 ]
+  %130 = tail call i32 @llvm.smax.i32(i32 %.046.i.us, i32 0)
+  %131 = tail call i32 @llvm.umin.i32(i32 %130, i32 63)
   %.not56.i.us = icmp slt i32 %.046.i.us, 1
   br i1 %.not56.i.us, label %149, label %132
 
@@ -1488,7 +1488,7 @@ define hidden i32 @VP8EnterCritical(ptr noundef %0, ptr noundef %1) local_unname
   br label %138
 
 138:                                              ; preds = %135, %132
-  %.0.i.us = phi i32 [ %131, %132 ], [ %spec.select.i.us, %135 ]
+  %.0.i.us = phi i32 [ %131, %133 ], [ %spec.select.i.us, %136 ]
   %spec.store.select.i.us = tail call i32 @llvm.smax.i32(i32 %.0.i.us, i32 1)
   %139 = trunc nuw nsw i32 %spec.store.select.i.us to i8
   %140 = getelementptr inbounds nuw i8, ptr %125, i64 1
@@ -1538,13 +1538,13 @@ define hidden i32 @VP8EnterCritical(ptr noundef %0, ptr noundef %1) local_unname
   br label %.split.i
 
 .split.i:                                         ; preds = %158, %154
-  %.048.i = phi i32 [ %157, %154 ], [ %160, %158 ]
+  %.048.i = phi i32 [ %157, %155 ], [ %160, %159 ]
   %161 = add nsw i32 %.048.i, %121
   br label %162
 
 162:                                              ; preds = %188, %.split.i
-  %.not55.i = phi i1 [ true, %.split.i ], [ false, %188 ]
-  %indvars.iv.i = phi i64 [ 0, %.split.i ], [ 1, %188 ]
+  %.not55.i = phi i1 [ true, %.split.i ], [ false, %189 ]
+  %indvars.iv.i = phi i64 [ 0, %.split.i ], [ 1, %189 ]
   %163 = getelementptr inbounds nuw [4 x [2 x %struct.VP8FInfo]], ptr %64, i64 0, i64 %indvars.iv72.i, i64 %indvars.iv.i
   br i1 %.not55.i, label %167, label %164
 
@@ -1554,9 +1554,9 @@ define hidden i32 @VP8EnterCritical(ptr noundef %0, ptr noundef %1) local_unname
   br label %167
 
 167:                                              ; preds = %164, %162
-  %.046.i = phi i32 [ %166, %164 ], [ %161, %162 ]
-  %168 = tail call i32 @llvm.smin.i32(i32 %.046.i, i32 63)
-  %169 = tail call i32 @llvm.smax.i32(i32 %168, i32 0)
+  %.046.i = phi i32 [ %166, %165 ], [ %161, %163 ]
+  %168 = tail call i32 @llvm.smax.i32(i32 %.046.i, i32 0)
+  %169 = tail call i32 @llvm.umin.i32(i32 %168, i32 63)
   %.not56.i = icmp slt i32 %.046.i, 1
   br i1 %.not56.i, label %187, label %170
 
@@ -1574,7 +1574,7 @@ define hidden i32 @VP8EnterCritical(ptr noundef %0, ptr noundef %1) local_unname
   br label %176
 
 176:                                              ; preds = %173, %170
-  %.0.i = phi i32 [ %169, %170 ], [ %spec.select.i, %173 ]
+  %.0.i = phi i32 [ %169, %171 ], [ %spec.select.i, %174 ]
   %spec.store.select.i = tail call i32 @llvm.smax.i32(i32 %.0.i, i32 1)
   %177 = trunc nuw nsw i32 %spec.store.select.i to i8
   %178 = getelementptr inbounds nuw i8, ptr %163, i64 1
@@ -1969,6 +1969,12 @@ declare i32 @llvm.smin.i32(i32, i32) #6
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smax.i32(i32, i32) #6
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #6
+
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umax.i32(i32, i32) #6
+
 attributes #0 = { nounwind uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
 attributes #2 = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
diff --git a/bench/libwebp/optimized/quant_enc.ll b/bench/libwebp/optimized/quant_enc.ll
index f789dce98a7..48fec55c8d8 100644
--- a/bench/libwebp/optimized/quant_enc.ll
+++ b/bench/libwebp/optimized/quant_enc.ll
@@ -124,8 +124,8 @@ QualityToJPEGCompression.exit:                    ; preds = %16, %22
   %50 = fsub double 1.000000e+00, %49
   %51 = fmul double %50, 1.270000e+02
   %52 = fptosi double %51 to i32
-  %53 = tail call i32 @llvm.smin.i32(i32 %52, i32 127)
-  %54 = tail call i32 @llvm.smax.i32(i32 %53, i32 0)
+  %53 = tail call i32 @llvm.smax.i32(i32 %52, i32 0)
+  %54 = tail call i32 @llvm.umin.i32(i32 %53, i32 127)
   %55 = getelementptr inbounds nuw i8, ptr %44, i64 680
   store i32 %54, ptr %55, align 8, !tbaa !31
   %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
@@ -165,11 +165,11 @@ QualityToJPEGCompression.exit:                    ; preds = %16, %22
   %70 = load i32, ptr %7, align 4, !tbaa !25
   %71 = mul nsw i32 %69, %70
   %72 = sdiv i32 %71, 100
-  %73 = tail call i32 @llvm.smin.i32(i32 %72, i32 6)
-  %74 = tail call i32 @llvm.smax.i32(i32 %73, i32 -4)
+  %73 = tail call i32 @llvm.smax.i32(i32 %72, i32 -4)
+  %74 = tail call i32 @llvm.smin.i32(i32 %73, i32 6)
   %75 = sdiv i32 %70, -25
-  %76 = tail call i32 @llvm.smin.i32(i32 %75, i32 15)
-  %77 = tail call i32 @llvm.smax.i32(i32 %76, i32 -15)
+  %76 = tail call i32 @llvm.smax.i32(i32 %75, i32 -15)
+  %77 = tail call i32 @llvm.smin.i32(i32 %76, i32 15)
   %78 = getelementptr inbounds nuw i8, ptr %0, i64 3596
   store i32 0, ptr %78, align 4, !tbaa !40
   %79 = getelementptr inbounds nuw i8, ptr %0, i64 3600
@@ -192,8 +192,8 @@ QualityToJPEGCompression.exit:                    ; preds = %16, %22
   %89 = getelementptr inbounds nuw [4 x %struct.VP8SegmentInfo], ptr %86, i64 0, i64 %indvars.iv.i
   %90 = getelementptr inbounds nuw i8, ptr %89, i64 680
   %91 = load i32, ptr %90, align 8, !tbaa !31
-  %92 = tail call i32 @llvm.smin.i32(i32 %91, i32 127)
-  %93 = tail call i32 @llvm.smax.i32(i32 %92, i32 0)
+  %92 = tail call i32 @llvm.smax.i32(i32 %91, i32 0)
+  %93 = tail call i32 @llvm.umin.i32(i32 %92, i32 127)
   %94 = zext nneg i32 %93 to i64
   %95 = getelementptr inbounds nuw [128 x i16], ptr @kAcTable, i64 0, i64 %94
   %96 = load i16, ptr %95, align 2, !tbaa !46
@@ -402,15 +402,15 @@ SimplifySegments.exit:                            ; preds = %177, %._crit_edge.i
   %194 = load i32, ptr %193, align 8, !tbaa !31
   %195 = load i32, ptr %78, align 4, !tbaa !40
   %196 = add nsw i32 %195, %194
-  %197 = tail call i32 @llvm.smin.i32(i32 %196, i32 127)
-  %198 = tail call i32 @llvm.smax.i32(i32 %197, i32 0)
+  %197 = tail call i32 @llvm.smax.i32(i32 %196, i32 0)
+  %198 = tail call i32 @llvm.umin.i32(i32 %197, i32 127)
   %199 = zext nneg i32 %198 to i64
   %200 = getelementptr inbounds nuw [128 x i8], ptr @kDcTable, i64 0, i64 %199
   %201 = load i8, ptr %200, align 1, !tbaa !58
   %202 = zext i8 %201 to i16
   store i16 %202, ptr %192, align 8, !tbaa !46
-  %203 = tail call i32 @llvm.smin.i32(i32 %194, i32 127)
-  %204 = tail call i32 @llvm.smax.i32(i32 %203, i32 0)
+  %203 = tail call i32 @llvm.smax.i32(i32 %194, i32 0)
+  %204 = tail call i32 @llvm.umin.i32(i32 %203, i32 127)
   %205 = zext nneg i32 %204 to i64
   %206 = getelementptr inbounds nuw [128 x i16], ptr @kAcTable, i64 0, i64 %205
   %207 = load i16, ptr %206, align 2, !tbaa !46
@@ -418,8 +418,8 @@ SimplifySegments.exit:                            ; preds = %177, %._crit_edge.i
   store i16 %207, ptr %208, align 2, !tbaa !46
   %209 = load i32, ptr %79, align 8, !tbaa !41
   %210 = add nsw i32 %209, %194
-  %211 = tail call i32 @llvm.smin.i32(i32 %210, i32 127)
-  %212 = tail call i32 @llvm.smax.i32(i32 %211, i32 0)
+  %211 = tail call i32 @llvm.smax.i32(i32 %210, i32 0)
+  %212 = tail call i32 @llvm.umin.i32(i32 %211, i32 127)
   %213 = zext nneg i32 %212 to i64
   %214 = getelementptr inbounds nuw [128 x i8], ptr @kDcTable, i64 0, i64 %213
   %215 = load i8, ptr %214, align 1, !tbaa !58
@@ -429,8 +429,8 @@ SimplifySegments.exit:                            ; preds = %177, %._crit_edge.i
   store i16 %217, ptr %218, align 8, !tbaa !46
   %219 = load i32, ptr %80, align 4, !tbaa !42
   %220 = add nsw i32 %219, %194
-  %221 = tail call i32 @llvm.smin.i32(i32 %220, i32 127)
-  %222 = tail call i32 @llvm.smax.i32(i32 %221, i32 0)
+  %221 = tail call i32 @llvm.smax.i32(i32 %220, i32 0)
+  %222 = tail call i32 @llvm.umin.i32(i32 %221, i32 127)
   %223 = zext nneg i32 %222 to i64
   %224 = getelementptr inbounds nuw [128 x i16], ptr @kAcTable2, i64 0, i64 %223
   %225 = load i16, ptr %224, align 2, !tbaa !46
@@ -438,8 +438,8 @@ SimplifySegments.exit:                            ; preds = %177, %._crit_edge.i
   store i16 %225, ptr %226, align 2, !tbaa !46
   %227 = load i32, ptr %81, align 8, !tbaa !43
   %228 = add nsw i32 %227, %194
-  %229 = tail call i32 @llvm.smin.i32(i32 %228, i32 117)
-  %230 = tail call i32 @llvm.smax.i32(i32 %229, i32 0)
+  %229 = tail call i32 @llvm.smax.i32(i32 %228, i32 0)
+  %230 = tail call i32 @llvm.umin.i32(i32 %229, i32 117)
   %231 = zext nneg i32 %230 to i64
   %232 = getelementptr inbounds nuw [128 x i8], ptr @kDcTable, i64 0, i64 %231
   %233 = load i8, ptr %232, align 1, !tbaa !58
@@ -448,8 +448,8 @@ SimplifySegments.exit:                            ; preds = %177, %._crit_edge.i
   store i16 %234, ptr %235, align 8, !tbaa !46
   %236 = load i32, ptr %82, align 4, !tbaa !44
   %237 = add nsw i32 %236, %194
-  %238 = tail call i32 @llvm.smin.i32(i32 %237, i32 127)
-  %239 = tail call i32 @llvm.smax.i32(i32 %238, i32 0)
+  %238 = tail call i32 @llvm.smax.i32(i32 %237, i32 0)
+  %239 = tail call i32 @llvm.umin.i32(i32 %238, i32 127)
   %240 = zext nneg i32 %239 to i64
   %241 = getelementptr inbounds nuw [128 x i16], ptr @kAcTable, i64 0, i64 %240
   %242 = load i16, ptr %241, align 2, !tbaa !46
diff --git a/bench/linux/optimized/ntp.ll b/bench/linux/optimized/ntp.ll
index 25983ce4f22..9e61c78f75a 100644
--- a/bench/linux/optimized/ntp.ll
+++ b/bench/linux/optimized/ntp.ll
@@ -444,8 +444,8 @@ define dso_local range(i32 0, 6) i32 @__do_adjtimex(ptr noundef captures(none) i
   %86 = getelementptr inbounds nuw i8, ptr %0, i64 16
   %87 = load i64, ptr %86, align 8
   %88 = mul i64 %87, 65536000
-  %89 = tail call i64 @llvm.smin.i64(i64 %88, i64 2147483648000000)
-  %90 = tail call i64 @llvm.smax.i64(i64 %89, i64 -2147483648000000)
+  %89 = tail call i64 @llvm.smax.i64(i64 %88, i64 -2147483648000000)
+  %90 = tail call i64 @llvm.smin.i64(i64 %89, i64 2147483648000000)
   store i64 %90, ptr @time_freq, align 8
   br label %91
 
@@ -483,8 +483,8 @@ define dso_local range(i32 0, 6) i32 @__do_adjtimex(ptr noundef captures(none) i
   %110 = icmp eq i32 %109, 0
   %111 = add i64 %108, 4
   %spec.select = select i1 %110, i64 %111, i64 %108
-  %112 = tail call i64 @llvm.smin.i64(i64 %spec.select, i64 10)
-  %113 = tail call i64 @llvm.smax.i64(i64 %112, i64 0)
+  %112 = tail call i64 @llvm.smax.i64(i64 %spec.select, i64 0)
+  %113 = tail call i64 @llvm.umin.i64(i64 %112, i64 10)
   store i64 %113, ptr @time_constant, align 8
   br label %114
 
@@ -594,8 +594,8 @@ define dso_local range(i32 0, 6) i32 @__do_adjtimex(ptr noundef captures(none) i
   %182 = load i64, ptr @time_freq, align 8
   %183 = add nsw i64 %182, %168
   %184 = add i64 %183, %181
-  %185 = tail call i64 @llvm.smin.i64(i64 %184, i64 2147483648000000)
-  %186 = tail call i64 @llvm.smax.i64(i64 %185, i64 -2147483648000000)
+  %185 = tail call i64 @llvm.smax.i64(i64 %184, i64 -2147483648000000)
+  %186 = tail call i64 @llvm.smin.i64(i64 %185, i64 2147483648000000)
   store i64 %186, ptr @time_freq, align 8
   %187 = shl nsw i64 %143, 32
   %188 = sdiv i64 %187, 1000
@@ -1003,6 +1003,9 @@ declare i64 @llvm.smin.i64(i64, i64) #8
 ; Function Attrs: mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.smax.i64(i64, i64) #8
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i64 @llvm.umin.i64(i64, i64) #9
+
 attributes #0 = { fn_ret_thunk_extern mustprogress nofree norecurse nosync nounwind null_pointer_is_valid willreturn memory(readwrite, argmem: none, inaccessiblemem: none) "min-legal-vector-width"="0" "no-jump-tables"="true" "no-trapping-math"="true" "patchable-function-entry"="0" "patchable-function-prefix"="16" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+retpoline-external-thunk,+retpoline-indirect-branches,+retpoline-indirect-calls,-3dnow,-3dnowa,-aes,-avx,-avx10.1-256,-avx10.1-512,-avx2,-avx512bf16,-avx512bitalg,-avx512bw,-avx512cd,-avx512dq,-avx512er,-avx512f,-avx512fp16,-avx512ifma,-avx512pf,-avx512vbmi,-avx512vbmi2,-avx512vl,-avx512vnni,-avx512vp2intersect,-avx512vpopcntdq,-avxifma,-avxneconvert,-avxvnni,-avxvnniint16,-avxvnniint8,-f16c,-fma,-fma4,-gfni,-kl,-mmx,-pclmul,-sha,-sha512,-sm3,-sm4,-sse,-sse2,-sse3,-sse4.1,-sse4.2,-sse4a,-ssse3,-vaes,-vpclmulqdq,-widekl,-x87,-xop" "tune-cpu"="generic" }
 attributes #1 = { fn_ret_thunk_extern mustprogress nofree norecurse nosync nounwind null_pointer_is_valid willreturn memory(read, argmem: none, inaccessiblemem: none) "min-legal-vector-width"="0" "no-jump-tables"="true" "no-trapping-math"="true" "patchable-function-entry"="0" "patchable-function-prefix"="16" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+retpoline-external-thunk,+retpoline-indirect-branches,+retpoline-indirect-calls,-3dnow,-3dnowa,-aes,-avx,-avx10.1-256,-avx10.1-512,-avx2,-avx512bf16,-avx512bitalg,-avx512bw,-avx512cd,-avx512dq,-avx512er,-avx512f,-avx512fp16,-avx512ifma,-avx512pf,-avx512vbmi,-avx512vbmi2,-avx512vl,-avx512vnni,-avx512vp2intersect,-avx512vpopcntdq,-avxifma,-avxneconvert,-avxvnni,-avxvnniint16,-avxvnniint8,-f16c,-fma,-fma4,-gfni,-kl,-mmx,-pclmul,-sha,-sha512,-sm3,-sm4,-sse,-sse2,-sse3,-sse4.1,-sse4.2,-sse4a,-ssse3,-vaes,-vpclmulqdq,-widekl,-x87,-xop" "tune-cpu"="generic" }
 attributes #2 = { mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
@@ -1012,8 +1015,9 @@ attributes #5 = { cold fn_ret_thunk_extern nounwind null_pointer_is_valid optsiz
 attributes #6 = { null_pointer_is_valid "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+retpoline-external-thunk,+retpoline-indirect-branches,+retpoline-indirect-calls,-3dnow,-3dnowa,-aes,-avx,-avx10.1-256,-avx10.1-512,-avx2,-avx512bf16,-avx512bitalg,-avx512bw,-avx512cd,-avx512dq,-avx512er,-avx512f,-avx512fp16,-avx512ifma,-avx512pf,-avx512vbmi,-avx512vbmi2,-avx512vl,-avx512vnni,-avx512vp2intersect,-avx512vpopcntdq,-avxifma,-avxneconvert,-avxvnni,-avxvnniint16,-avxvnniint8,-f16c,-fma,-fma4,-gfni,-kl,-mmx,-pclmul,-sha,-sha512,-sm3,-sm4,-sse,-sse2,-sse3,-sse4.1,-sse4.2,-sse4a,-ssse3,-vaes,-vpclmulqdq,-widekl,-x87,-xop" "tune-cpu"="generic" }
 attributes #7 = { mustprogress nocallback nofree nounwind willreturn memory(argmem: write) }
 attributes #8 = { mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none) }
-attributes #9 = { cold nounwind }
-attributes #10 = { nounwind }
+attributes #9 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
+attributes #10 = { cold nounwind }
+attributes #11 = { nounwind }
 
 !llvm.module.flags = !{!0, !1, !2, !3, !4}
 
diff --git a/bench/llama.cpp/optimized/ggml-quants.ll b/bench/llama.cpp/optimized/ggml-quants.ll
index 338c7aa2fbf..b9c1ede5968 100644
--- a/bench/llama.cpp/optimized/ggml-quants.ll
+++ b/bench/llama.cpp/optimized/ggml-quants.ll
@@ -1462,9 +1462,9 @@ make_qkx2_quants.exit:                            ; preds = %.loopexit.i, %.preh
   %218 = fadd float %217, 0x4168000000000000
   %219 = bitcast float %218 to i32
   %220 = and i32 %219, 8388607
-  %221 = tail call i32 @llvm.umin.i32(i32 %220, i32 4194307)
-  %222 = tail call i32 @llvm.umax.i32(i32 %221, i32 4194304)
-  %223 = trunc i32 %222 to i8
+  %221 = tail call i32 @llvm.usub.sat.i32(i32 %220, i32 4194304)
+  %222 = tail call i32 @llvm.umin.i32(i32 %221, i32 3)
+  %223 = trunc nuw nsw i32 %222 to i8
   %224 = getelementptr inbounds nuw [256 x i8], ptr %4, i64 0, i64 %213
   store i8 %223, ptr %224, align 1, !tbaa !13
   %indvars.iv.next168 = add nuw nsw i64 %indvars.iv167, 1
@@ -2459,9 +2459,9 @@ make_qp_quants.exit.us:                           ; preds = %370, %.preheader191
   %449 = fadd float %448, 0x4168000000000000
   %450 = bitcast float %449 to i32
   %451 = and i32 %450, 8388607
-  %452 = tail call i32 @llvm.umin.i32(i32 %451, i32 4194307)
-  %453 = tail call i32 @llvm.umax.i32(i32 %452, i32 4194304)
-  %454 = trunc i32 %453 to i8
+  %452 = tail call i32 @llvm.usub.sat.i32(i32 %451, i32 4194304)
+  %453 = tail call i32 @llvm.umin.i32(i32 %452, i32 3)
+  %454 = trunc nuw nsw i32 %453 to i8
   %455 = getelementptr inbounds nuw [256 x i8], ptr %6, i64 0, i64 %444
   store i8 %454, ptr %455, align 1, !tbaa !13, !noalias !86
   %indvars.iv.next159.i.us = add nuw nsw i64 %indvars.iv158.i.us, 1
@@ -2625,8 +2625,8 @@ define void @quantize_row_q3_K_ref(ptr noalias noundef readonly captures(none) %
   %30 = fsub float 0x4168000000000000, %29
   %31 = bitcast float %30 to i32
   %32 = and i32 %31, 8388607
-  %33 = tail call i32 @llvm.umin.i32(i32 %32, i32 4194307)
-  %34 = tail call i32 @llvm.umax.i32(i32 %33, i32 4194300)
+  %33 = tail call i32 @llvm.umax.i32(i32 %32, i32 4194300)
+  %34 = tail call i32 @llvm.umin.i32(i32 %33, i32 4194307)
   %35 = add nsw i32 %34, -4194304
   %36 = trunc i32 %34 to i8
   %37 = getelementptr inbounds nuw i8, ptr %17, i64 %indvars.iv196.i
@@ -2680,8 +2680,8 @@ define void @quantize_row_q3_K_ref(ptr noalias noundef readonly captures(none) %
   %63 = fadd float %62, 0x4168000000000000
   %64 = bitcast float %63 to i32
   %65 = and i32 %64, 8388607
-  %66 = tail call i32 @llvm.umin.i32(i32 %65, i32 4194307)
-  %67 = tail call i32 @llvm.umax.i32(i32 %66, i32 4194300)
+  %66 = tail call i32 @llvm.umax.i32(i32 %65, i32 4194300)
+  %67 = tail call i32 @llvm.umin.i32(i32 %66, i32 4194307)
   %68 = add nsw i32 %67, -4194304
   %69 = sext i8 %51 to i32
   %.not.i = icmp eq i32 %68, %69
@@ -2786,8 +2786,8 @@ make_q3_quants.exit:                              ; preds = %.preheader.preheade
   %127 = bitcast float %126 to i32
   %sext = shl i32 %127, 24
   %128 = ashr exact i32 %sext, 24
-  %129 = tail call i32 @llvm.smin.i32(i32 %128, i32 31)
-  %130 = tail call i32 @llvm.smax.i32(i32 %129, i32 -32)
+  %129 = tail call i32 @llvm.smax.i32(i32 %128, i32 -32)
+  %130 = tail call i32 @llvm.smin.i32(i32 %129, i32 31)
   %131 = icmp samesign ult i64 %indvars.iv147, 8
   br i1 %131, label %132, label %136
 
@@ -2894,8 +2894,8 @@ make_q3_quants.exit:                              ; preds = %.preheader.preheade
   %196 = fadd float %195, 0x4168000000000000
   %197 = bitcast float %196 to i32
   %198 = and i32 %197, 8388607
-  %199 = tail call i32 @llvm.umin.i32(i32 %198, i32 4194307)
-  %200 = tail call i32 @llvm.umax.i32(i32 %199, i32 4194300)
+  %199 = tail call i32 @llvm.umax.i32(i32 %198, i32 4194300)
+  %200 = tail call i32 @llvm.umin.i32(i32 %199, i32 4194307)
   %201 = trunc i32 %200 to i8
   %202 = add nsw i8 %201, 4
   %203 = getelementptr inbounds nuw [256 x i8], ptr %4, i64 0, i64 %192
@@ -3272,8 +3272,8 @@ define i64 @quantize_q3_K(ptr noalias noundef readonly captures(none) %0, ptr no
   %54 = fsub float 0x4168000000000000, %53
   %55 = bitcast float %54 to i32
   %56 = and i32 %55, 8388607
-  %57 = tail call i32 @llvm.umin.i32(i32 %56, i32 4194307)
-  %58 = tail call i32 @llvm.umax.i32(i32 %57, i32 4194300)
+  %57 = tail call i32 @llvm.umax.i32(i32 %56, i32 4194300)
+  %58 = tail call i32 @llvm.umin.i32(i32 %57, i32 4194307)
   %..i.i.us = add nsw i32 %58, -4194304
   %59 = trunc i32 %58 to i8
   %60 = add nsw i8 %59, 4
@@ -3321,8 +3321,8 @@ define i64 @quantize_q3_K(ptr noalias noundef readonly captures(none) %0, ptr no
   %84 = fadd float %83, 0x4168000000000000
   %85 = bitcast float %84 to i32
   %86 = and i32 %85, 8388607
-  %87 = tail call i32 @llvm.umin.i32(i32 %86, i32 4194307)
-  %88 = tail call i32 @llvm.umax.i32(i32 %87, i32 4194300)
+  %87 = tail call i32 @llvm.umax.i32(i32 %86, i32 4194300)
+  %88 = tail call i32 @llvm.umin.i32(i32 %87, i32 4194307)
   %.202.i.i.us = add nsw i32 %88, -4194304
   %89 = getelementptr inbounds nuw float, ptr %8, i64 %indvars.iv236.i.i.us
   %90 = load float, ptr %89, align 4, !tbaa !8, !alias.scope !194, !noalias !200
@@ -3353,8 +3353,8 @@ define i64 @quantize_q3_K(ptr noalias noundef readonly captures(none) %0, ptr no
   %104 = fadd float %103, 0x4168000000000000
   %105 = bitcast float %104 to i32
   %106 = and i32 %105, 8388607
-  %107 = tail call i32 @llvm.umin.i32(i32 %106, i32 4194307)
-  %108 = tail call i32 @llvm.umax.i32(i32 %107, i32 4194300)
+  %107 = tail call i32 @llvm.umax.i32(i32 %106, i32 4194300)
+  %108 = tail call i32 @llvm.umin.i32(i32 %107, i32 4194307)
   %109 = trunc i32 %108 to i8
   %110 = add nsw i8 %109, 4
   %111 = getelementptr inbounds nuw i8, ptr %46, i64 %indvars.iv246.i.i.us
@@ -3428,8 +3428,8 @@ make_qx_quants.exit.i.us:                         ; preds = %115, %.preheader.pr
   %134 = fsub float 0x4168000000000000, %133
   %135 = bitcast float %134 to i32
   %136 = and i32 %135, 8388607
-  %137 = tail call i32 @llvm.umin.i32(i32 %136, i32 4194335)
-  %138 = tail call i32 @llvm.umax.i32(i32 %137, i32 4194272)
+  %137 = tail call i32 @llvm.umax.i32(i32 %136, i32 4194272)
+  %138 = tail call i32 @llvm.umin.i32(i32 %137, i32 4194335)
   %..i.us = add nsw i32 %138, -4194304
   %139 = trunc i32 %138 to i8
   %140 = add nsw i8 %139, 32
@@ -3477,8 +3477,8 @@ make_qx_quants.exit.i.us:                         ; preds = %115, %.preheader.pr
   %164 = fadd float %163, 0x4168000000000000
   %165 = bitcast float %164 to i32
   %166 = and i32 %165, 8388607
-  %167 = tail call i32 @llvm.umin.i32(i32 %166, i32 4194335)
-  %168 = tail call i32 @llvm.umax.i32(i32 %167, i32 4194272)
+  %167 = tail call i32 @llvm.umax.i32(i32 %166, i32 4194272)
+  %168 = tail call i32 @llvm.umin.i32(i32 %167, i32 4194335)
   %.202.i.us = add nsw i32 %168, -4194304
   %169 = getelementptr inbounds nuw float, ptr %9, i64 %indvars.iv236.i.us
   %170 = load float, ptr %169, align 4, !tbaa !8, !alias.scope !212, !noalias !216
@@ -3509,8 +3509,8 @@ make_qx_quants.exit.i.us:                         ; preds = %115, %.preheader.pr
   %184 = fadd float %183, 0x4168000000000000
   %185 = bitcast float %184 to i32
   %186 = and i32 %185, 8388607
-  %187 = tail call i32 @llvm.umin.i32(i32 %186, i32 4194335)
-  %188 = tail call i32 @llvm.umax.i32(i32 %187, i32 4194272)
+  %187 = tail call i32 @llvm.umax.i32(i32 %186, i32 4194272)
+  %188 = tail call i32 @llvm.umin.i32(i32 %187, i32 4194335)
   %189 = trunc i32 %188 to i8
   %190 = add nsw i8 %189, 32
   %191 = getelementptr inbounds nuw i8, ptr %10, i64 %indvars.iv246.i.us
@@ -3661,8 +3661,8 @@ make_qx_quants.exit.us:                           ; preds = %195, %.preheader.pr
   %283 = fadd float %282, 0x4168000000000000
   %284 = bitcast float %283 to i32
   %285 = and i32 %284, 8388607
-  %286 = tail call i32 @llvm.umin.i32(i32 %285, i32 4194307)
-  %287 = tail call i32 @llvm.umax.i32(i32 %286, i32 4194300)
+  %286 = tail call i32 @llvm.umax.i32(i32 %285, i32 4194300)
+  %287 = tail call i32 @llvm.umin.i32(i32 %286, i32 4194307)
   %288 = trunc i32 %287 to i8
   %289 = add nsw i8 %288, 4
   %290 = getelementptr inbounds nuw [256 x i8], ptr %6, i64 0, i64 %279
@@ -4221,9 +4221,9 @@ get_scale_min_k4.exit:                            ; preds = %232, %238
   %266 = fadd float %265, 0x4168000000000000
   %267 = bitcast float %266 to i32
   %268 = and i32 %267, 8388607
-  %269 = tail call i32 @llvm.umin.i32(i32 %268, i32 4194319)
-  %270 = tail call i32 @llvm.umax.i32(i32 %269, i32 4194304)
-  %271 = trunc i32 %270 to i8
+  %269 = tail call i32 @llvm.usub.sat.i32(i32 %268, i32 4194304)
+  %270 = tail call i32 @llvm.umin.i32(i32 %269, i32 15)
+  %271 = trunc nuw nsw i32 %270 to i8
   %272 = getelementptr inbounds nuw [256 x i8], ptr %4, i64 0, i64 %261
   store i8 %271, ptr %272, align 1, !tbaa !13
   %indvars.iv.next178 = add nuw nsw i64 %indvars.iv177, 1
@@ -5300,9 +5300,9 @@ get_scale_min_k4.exit.i.us:                       ; preds = %468, %454
   %488 = fadd float %487, 0x4168000000000000
   %489 = bitcast float %488 to i32
   %490 = and i32 %489, 8388607
-  %491 = tail call i32 @llvm.umin.i32(i32 %490, i32 4194319)
-  %492 = tail call i32 @llvm.umax.i32(i32 %491, i32 4194304)
-  %493 = trunc i32 %492 to i8
+  %491 = tail call i32 @llvm.usub.sat.i32(i32 %490, i32 4194304)
+  %492 = tail call i32 @llvm.umin.i32(i32 %491, i32 15)
+  %493 = trunc nuw nsw i32 %492 to i8
   %494 = getelementptr inbounds nuw [256 x i8], ptr %6, i64 0, i64 %483
   store i8 %493, ptr %494, align 1, !tbaa !13, !noalias !279
   %indvars.iv.next185.i.us = add nuw nsw i64 %indvars.iv184.i.us, 1
@@ -5818,9 +5818,9 @@ get_scale_min_k4.exit:                            ; preds = %232, %238
   %266 = fadd float %265, 0x4168000000000000
   %267 = bitcast float %266 to i32
   %268 = and i32 %267, 8388607
-  %269 = tail call i32 @llvm.umin.i32(i32 %268, i32 4194335)
-  %270 = tail call i32 @llvm.umax.i32(i32 %269, i32 4194304)
-  %271 = trunc i32 %270 to i8
+  %269 = tail call i32 @llvm.usub.sat.i32(i32 %268, i32 4194304)
+  %270 = tail call i32 @llvm.umin.i32(i32 %269, i32 31)
+  %271 = trunc nuw nsw i32 %270 to i8
   %272 = getelementptr inbounds nuw [256 x i8], ptr %4, i64 0, i64 %261
   store i8 %271, ptr %272, align 1, !tbaa !13
   %indvars.iv.next206 = add nuw nsw i64 %indvars.iv205, 1
@@ -6940,9 +6940,9 @@ get_scale_min_k4.exit.i.us:                       ; preds = %470, %456
   %490 = fadd float %489, 0x4168000000000000
   %491 = bitcast float %490 to i32
   %492 = and i32 %491, 8388607
-  %493 = tail call i32 @llvm.umin.i32(i32 %492, i32 4194335)
-  %494 = tail call i32 @llvm.umax.i32(i32 %493, i32 4194304)
-  %495 = trunc i32 %494 to i8
+  %493 = tail call i32 @llvm.usub.sat.i32(i32 %492, i32 4194304)
+  %494 = tail call i32 @llvm.umin.i32(i32 %493, i32 31)
+  %495 = trunc nuw nsw i32 %494 to i8
   %496 = getelementptr inbounds nuw [256 x i8], ptr %6, i64 0, i64 %485
   store i8 %495, ptr %496, align 1, !tbaa !13, !noalias !387
   %indvars.iv.next215.i.us = add nuw nsw i64 %indvars.iv214.i.us, 1
@@ -7120,8 +7120,8 @@ define void @quantize_row_q6_K_ref(ptr noalias noundef readonly captures(none) %
   %27 = fsub float 0x4168000000000000, %26
   %28 = bitcast float %27 to i32
   %29 = and i32 %28, 8388607
-  %30 = tail call i32 @llvm.umin.i32(i32 %29, i32 4194335)
-  %31 = tail call i32 @llvm.umax.i32(i32 %30, i32 4194272)
+  %30 = tail call i32 @llvm.umax.i32(i32 %29, i32 4194272)
+  %31 = tail call i32 @llvm.umin.i32(i32 %30, i32 4194335)
   %..i = add nsw i32 %31, -4194304
   %32 = trunc i32 %31 to i8
   %33 = add nsw i8 %32, 32
@@ -7168,8 +7168,8 @@ define void @quantize_row_q6_K_ref(ptr noalias noundef readonly captures(none) %
   %56 = fadd float %55, 0x4168000000000000
   %57 = bitcast float %56 to i32
   %58 = and i32 %57, 8388607
-  %59 = tail call i32 @llvm.umin.i32(i32 %58, i32 4194335)
-  %60 = tail call i32 @llvm.umax.i32(i32 %59, i32 4194272)
+  %59 = tail call i32 @llvm.umax.i32(i32 %58, i32 4194272)
+  %60 = tail call i32 @llvm.umin.i32(i32 %59, i32 4194335)
   %.202.us.i = add nsw i32 %60, -4194304
   %61 = fmul float %54, %54
   %62 = fmul float %54, %61
@@ -7204,8 +7204,8 @@ define void @quantize_row_q6_K_ref(ptr noalias noundef readonly captures(none) %
   %78 = fadd float %77, 0x4168000000000000
   %79 = bitcast float %78 to i32
   %80 = and i32 %79, 8388607
-  %81 = tail call i32 @llvm.umin.i32(i32 %80, i32 4194335)
-  %82 = tail call i32 @llvm.umax.i32(i32 %81, i32 4194272)
+  %81 = tail call i32 @llvm.umax.i32(i32 %80, i32 4194272)
+  %82 = tail call i32 @llvm.umin.i32(i32 %81, i32 4194335)
   %83 = trunc i32 %82 to i8
   %84 = add nsw i8 %83, 32
   %85 = getelementptr inbounds nuw i8, ptr %14, i64 %indvars.iv246.i
@@ -7319,8 +7319,8 @@ make_qx_quants.exit:                              ; preds = %86, %.preheader.pre
   %148 = fadd float %147, 0x4168000000000000
   %149 = bitcast float %148 to i32
   %150 = and i32 %149, 8388607
-  %151 = tail call i32 @llvm.umin.i32(i32 %150, i32 4194335)
-  %152 = tail call i32 @llvm.umax.i32(i32 %151, i32 4194272)
+  %151 = tail call i32 @llvm.umax.i32(i32 %150, i32 4194272)
+  %152 = tail call i32 @llvm.umin.i32(i32 %151, i32 4194335)
   %153 = trunc i32 %152 to i8
   %154 = add nsw i8 %153, 32
   %155 = getelementptr inbounds nuw [256 x i8], ptr %4, i64 0, i64 %144
@@ -7580,8 +7580,8 @@ define i64 @quantize_q6_K(ptr noalias noundef readonly captures(none) %0, ptr no
   %31 = fsub float 0x4168000000000000, %30
   %32 = bitcast float %31 to i32
   %33 = and i32 %32, 8388607
-  %34 = tail call i32 @llvm.umin.i32(i32 %33, i32 4194335)
-  %35 = tail call i32 @llvm.umax.i32(i32 %34, i32 4194272)
+  %34 = tail call i32 @llvm.umax.i32(i32 %33, i32 4194272)
+  %35 = tail call i32 @llvm.umin.i32(i32 %34, i32 4194335)
   %..i.i.us = add nsw i32 %35, -4194304
   %36 = trunc i32 %35 to i8
   %37 = add nsw i8 %36, 32
@@ -7629,8 +7629,8 @@ define i64 @quantize_q6_K(ptr noalias noundef readonly captures(none) %0, ptr no
   %61 = fadd float %60, 0x4168000000000000
   %62 = bitcast float %61 to i32
   %63 = and i32 %62, 8388607
-  %64 = tail call i32 @llvm.umin.i32(i32 %63, i32 4194335)
-  %65 = tail call i32 @llvm.umax.i32(i32 %64, i32 4194272)
+  %64 = tail call i32 @llvm.umax.i32(i32 %63, i32 4194272)
+  %65 = tail call i32 @llvm.umin.i32(i32 %64, i32 4194335)
   %.202.i.i.us = add nsw i32 %65, -4194304
   %66 = getelementptr inbounds nuw float, ptr %22, i64 %indvars.iv236.i.i.us
   %67 = load float, ptr %66, align 4, !tbaa !8, !alias.scope !471, !noalias !476
@@ -7661,8 +7661,8 @@ define i64 @quantize_q6_K(ptr noalias noundef readonly captures(none) %0, ptr no
   %81 = fadd float %80, 0x4168000000000000
   %82 = bitcast float %81 to i32
   %83 = and i32 %82, 8388607
-  %84 = tail call i32 @llvm.umin.i32(i32 %83, i32 4194335)
-  %85 = tail call i32 @llvm.umax.i32(i32 %84, i32 4194272)
+  %84 = tail call i32 @llvm.umax.i32(i32 %83, i32 4194272)
+  %85 = tail call i32 @llvm.umin.i32(i32 %84, i32 4194335)
   %86 = trunc i32 %85 to i8
   %87 = add nsw i8 %86, 32
   %88 = getelementptr inbounds nuw i8, ptr %23, i64 %indvars.iv246.i.i.us
@@ -7780,8 +7780,8 @@ make_qx_quants.exit.i.us:                         ; preds = %92, %.preheader.pre
   %152 = fadd float %151, 0x4168000000000000
   %153 = bitcast float %152 to i32
   %154 = and i32 %153, 8388607
-  %155 = tail call i32 @llvm.umin.i32(i32 %154, i32 4194335)
-  %156 = tail call i32 @llvm.umax.i32(i32 %155, i32 4194272)
+  %155 = tail call i32 @llvm.umax.i32(i32 %154, i32 4194272)
+  %156 = tail call i32 @llvm.umin.i32(i32 %155, i32 4194335)
   %157 = trunc i32 %156 to i8
   %158 = add nsw i8 %157, 32
   %159 = getelementptr inbounds nuw [256 x i8], ptr %6, i64 0, i64 %148
@@ -8078,8 +8078,8 @@ quantize_row_q4_0_ref.exit:                       ; preds = %45, %8
   %97 = fsub float 0x4168000000000000, %96
   %98 = bitcast float %97 to i32
   %99 = and i32 %98, 8388607
-  %100 = tail call i32 @llvm.umin.i32(i32 %99, i32 4194311)
-  %101 = tail call i32 @llvm.umax.i32(i32 %100, i32 4194296)
+  %100 = tail call i32 @llvm.umax.i32(i32 %99, i32 4194296)
+  %101 = tail call i32 @llvm.umin.i32(i32 %100, i32 4194311)
   %..i.us.us = add nsw i32 %101, -4194304
   %102 = trunc i32 %101 to i8
   %103 = add nsw i8 %102, 8
@@ -8127,8 +8127,8 @@ quantize_row_q4_0_ref.exit:                       ; preds = %45, %8
   %127 = fadd float %126, 0x4168000000000000
   %128 = bitcast float %127 to i32
   %129 = and i32 %128, 8388607
-  %130 = tail call i32 @llvm.umin.i32(i32 %129, i32 4194311)
-  %131 = tail call i32 @llvm.umax.i32(i32 %130, i32 4194296)
+  %130 = tail call i32 @llvm.umax.i32(i32 %129, i32 4194296)
+  %131 = tail call i32 @llvm.umin.i32(i32 %130, i32 4194311)
   %.202.i.us.us = add nsw i32 %131, -4194304
   %132 = getelementptr inbounds nuw float, ptr %6, i64 %indvars.iv236.i.us.us
   %133 = load float, ptr %132, align 4, !tbaa !8, !alias.scope !503, !noalias !507
@@ -8159,8 +8159,8 @@ quantize_row_q4_0_ref.exit:                       ; preds = %45, %8
   %147 = fadd float %146, 0x4168000000000000
   %148 = bitcast float %147 to i32
   %149 = and i32 %148, 8388607
-  %150 = tail call i32 @llvm.umin.i32(i32 %149, i32 4194311)
-  %151 = tail call i32 @llvm.umax.i32(i32 %150, i32 4194296)
+  %150 = tail call i32 @llvm.umax.i32(i32 %149, i32 4194296)
+  %151 = tail call i32 @llvm.umin.i32(i32 %150, i32 4194311)
   %152 = trunc i32 %151 to i8
   %153 = add nsw i8 %152, 8
   %154 = getelementptr inbounds nuw i8, ptr %7, i64 %indvars.iv246.i.us.us
@@ -8688,8 +8688,8 @@ define i64 @quantize_q5_0(ptr noalias noundef readonly captures(none) %0, ptr no
   %46 = fsub float 0x4168000000000000, %45
   %47 = bitcast float %46 to i32
   %48 = and i32 %47, 8388607
-  %49 = tail call i32 @llvm.umin.i32(i32 %48, i32 4194319)
-  %50 = tail call i32 @llvm.umax.i32(i32 %49, i32 4194288)
+  %49 = tail call i32 @llvm.umax.i32(i32 %48, i32 4194288)
+  %50 = tail call i32 @llvm.umin.i32(i32 %49, i32 4194319)
   %..i.i.us.us = add nsw i32 %50, -4194304
   %51 = trunc i32 %50 to i8
   %52 = add nsw i8 %51, 16
@@ -8737,8 +8737,8 @@ define i64 @quantize_q5_0(ptr noalias noundef readonly captures(none) %0, ptr no
   %76 = fadd float %75, 0x4168000000000000
   %77 = bitcast float %76 to i32
   %78 = and i32 %77, 8388607
-  %79 = tail call i32 @llvm.umin.i32(i32 %78, i32 4194319)
-  %80 = tail call i32 @llvm.umax.i32(i32 %79, i32 4194288)
+  %79 = tail call i32 @llvm.umax.i32(i32 %78, i32 4194288)
+  %80 = tail call i32 @llvm.umin.i32(i32 %79, i32 4194319)
   %.202.i.i.us.us = add nsw i32 %80, -4194304
   %81 = getelementptr inbounds nuw float, ptr %6, i64 %indvars.iv236.i.i.us.us
   %82 = load float, ptr %81, align 4, !tbaa !8, !alias.scope !551, !noalias !556
@@ -8769,8 +8769,8 @@ define i64 @quantize_q5_0(ptr noalias noundef readonly captures(none) %0, ptr no
   %96 = fadd float %95, 0x4168000000000000
   %97 = bitcast float %96 to i32
   %98 = and i32 %97, 8388607
-  %99 = tail call i32 @llvm.umin.i32(i32 %98, i32 4194319)
-  %100 = tail call i32 @llvm.umax.i32(i32 %99, i32 4194288)
+  %99 = tail call i32 @llvm.umax.i32(i32 %98, i32 4194288)
+  %100 = tail call i32 @llvm.umin.i32(i32 %99, i32 4194319)
   %101 = trunc i32 %100 to i8
   %102 = add nsw i8 %101, 16
   %103 = getelementptr inbounds nuw i8, ptr %7, i64 %indvars.iv246.i.i.us.us
@@ -12021,9 +12021,9 @@ make_qp_quants.exit.i.us:                         ; preds = %219, %.preheader191
   %234 = fadd float %233, 0x4168000000000000
   %235 = bitcast float %234 to i32
   %236 = and i32 %235, 8388607
-  %237 = tail call i32 @llvm.umin.i32(i32 %236, i32 4194306)
-  %238 = tail call i32 @llvm.umax.i32(i32 %237, i32 4194304)
-  %239 = trunc i32 %238 to i8
+  %237 = tail call i32 @llvm.usub.sat.i32(i32 %236, i32 4194304)
+  %238 = tail call i32 @llvm.umin.i32(i32 %237, i32 2)
+  %239 = trunc nuw nsw i32 %238 to i8
   %240 = getelementptr inbounds nuw [32 x i8], ptr %10, i64 0, i64 %229
   store i8 %239, ptr %240, align 1, !tbaa !13, !noalias !709
   %indvars.iv.next459.i.us = add nuw nsw i64 %indvars.iv458.i.us, 1
@@ -12207,12 +12207,12 @@ iq2_find_best_neighbour.exit.i.us:                ; preds = %286, %249
   %329 = fadd float %328, 0x4168000000000000
   %330 = bitcast float %329 to i32
   %331 = and i32 %330, 8388607
-  %332 = tail call i32 @llvm.umin.i32(i32 %331, i32 4194306)
-  %333 = tail call i32 @llvm.umax.i32(i32 %332, i32 4194304)
+  %332 = tail call i32 @llvm.usub.sat.i32(i32 %331, i32 4194304)
+  %333 = tail call i32 @llvm.umin.i32(i32 %332, i32 2)
   %indvars.iv474.tr.i.us = trunc i64 %indvars.iv474.i.us to i32
   %334 = shl i32 %indvars.iv474.tr.i.us, 1
-  %335 = shl i32 %333, %334
-  %336 = trunc i32 %335 to i16
+  %335 = shl nuw nsw i32 %333, %334
+  %336 = trunc nuw i32 %335 to i16
   %337 = or i16 %.0287401.i.us, %336
   %indvars.iv.next475.i.us = add nuw nsw i64 %indvars.iv474.i.us, 1
   %exitcond476.not.i.us = icmp eq i64 %indvars.iv.next475.i.us, 8
@@ -12506,9 +12506,9 @@ iq2_find_best_neighbour.exit355.i.us:             ; preds = %375, %.iq2_find_bes
   %490 = fadd float %489, 0x4168000000000000
   %491 = bitcast float %490 to i32
   %492 = and i32 %491, 8388607
-  %493 = tail call i32 @llvm.umin.i32(i32 %492, i32 4194319)
-  %494 = tail call i32 @llvm.umax.i32(i32 %493, i32 4194304)
-  %495 = shl i32 %494, 28
+  %493 = tail call i32 @llvm.usub.sat.i32(i32 %492, i32 4194304)
+  %494 = tail call i32 @llvm.umin.i32(i32 %493, i32 15)
+  %495 = shl nuw i32 %494, 28
   %496 = shl nuw nsw i64 %indvars.iv504.i.us, 1
   %497 = or disjoint i64 %496, 1
   %498 = getelementptr inbounds nuw [16 x i32], ptr %13, i64 0, i64 %497
@@ -12890,9 +12890,9 @@ define i64 @quantize_iq2_xs(ptr noalias noundef readonly captures(none) %0, ptr
   %119 = fadd float %118, 0x4168000000000000
   %120 = bitcast float %119 to i32
   %121 = and i32 %120, 8388607
-  %122 = tail call i32 @llvm.umin.i32(i32 %121, i32 4194306)
-  %123 = tail call i32 @llvm.umax.i32(i32 %122, i32 4194304)
-  %124 = trunc i32 %123 to i8
+  %122 = tail call i32 @llvm.usub.sat.i32(i32 %121, i32 4194304)
+  %123 = tail call i32 @llvm.umin.i32(i32 %122, i32 2)
+  %124 = trunc nuw nsw i32 %123 to i8
   %125 = getelementptr inbounds nuw [16 x i8], ptr %10, i64 0, i64 %114
   store i8 %124, ptr %125, align 1, !tbaa !13, !noalias !789
   %indvars.iv.next470.i.us = add nuw nsw i64 %indvars.iv469.i.us, 1
@@ -13102,14 +13102,14 @@ iq2_find_best_neighbour.exit.i.us:                ; preds = %168, %134
   %216 = fadd float %215, 0x4168000000000000
   %217 = bitcast float %216 to i32
   %218 = and i32 %217, 8388607
-  %219 = tail call i32 @llvm.umin.i32(i32 %218, i32 4194306)
-  %220 = tail call i32 @llvm.umax.i32(i32 %219, i32 4194304)
+  %219 = tail call i32 @llvm.usub.sat.i32(i32 %218, i32 4194304)
+  %220 = tail call i32 @llvm.umin.i32(i32 %219, i32 2)
   %indvars.iv493.tr.i.us = trunc i64 %indvars.iv493.i.us to i32
   %221 = shl i32 %indvars.iv493.tr.i.us, 1
-  %222 = shl i32 %220, %221
-  %223 = trunc i32 %222 to i16
+  %222 = shl nuw nsw i32 %220, %221
+  %223 = trunc nuw i32 %222 to i16
   %224 = or i16 %.0302418.i.us, %223
-  %225 = trunc i32 %220 to i8
+  %225 = trunc nuw nsw i32 %220 to i8
   %226 = getelementptr inbounds nuw [16 x i8], ptr %9, i64 0, i64 %211
   store i8 %225, ptr %226, align 1, !tbaa !13, !noalias !789
   %indvars.iv.next494.i.us = add nuw nsw i64 %indvars.iv493.i.us, 1
@@ -13363,8 +13363,8 @@ iq2_find_best_neighbour.exit369.i.us:             ; preds = %261, %227, %208
   %359 = fadd float %358, 0x4168000000000000
   %360 = bitcast float %359 to i32
   %361 = and i32 %360, 8388607
-  %362 = tail call i32 @llvm.umin.i32(i32 %361, i32 4194319)
-  %363 = tail call i32 @llvm.umax.i32(i32 %362, i32 4194304)
+  %362 = tail call i32 @llvm.usub.sat.i32(i32 %361, i32 4194304)
+  %363 = tail call i32 @llvm.umin.i32(i32 %362, i32 15)
   %364 = and i64 %indvars.iv516.i.us, 1
   %365 = icmp eq i64 %364, 0
   br i1 %365, label %373, label %366
@@ -13374,14 +13374,14 @@ iq2_find_best_neighbour.exit369.i.us:             ; preds = %261, %227, %208
   %368 = and i64 %367, 2147483647
   %369 = getelementptr inbounds nuw [8 x i8], ptr %26, i64 0, i64 %368
   %370 = load i8, ptr %369, align 1, !tbaa !13, !alias.scope !795, !noalias !798
-  %.tr.i.us = trunc i32 %363 to i8
+  %.tr.i.us = trunc nuw nsw i32 %363 to i8
   %371 = shl nuw i8 %.tr.i.us, 4
   %372 = or i8 %370, %371
   store i8 %372, ptr %369, align 1, !tbaa !13, !alias.scope !795, !noalias !798
   br label %378
 
 373:                                              ; preds = %354
-  %374 = trunc i32 %363 to i8
+  %374 = trunc nuw nsw i32 %363 to i8
   %375 = lshr exact i64 %indvars.iv516.i.us, 1
   %376 = and i64 %375, 2147483647
   %377 = getelementptr inbounds nuw [8 x i8], ptr %26, i64 0, i64 %376
@@ -14241,9 +14241,9 @@ define internal fastcc void @quantize_row_iq3_xxs_impl(ptr noalias noundef reado
   %140 = fadd float %139, 0x4168000000000000
   %141 = bitcast float %140 to i32
   %142 = and i32 %141, 8388607
-  %143 = tail call i32 @llvm.umin.i32(i32 %142, i32 4194311)
-  %144 = tail call i32 @llvm.umax.i32(i32 %143, i32 4194304)
-  %145 = trunc i32 %144 to i8
+  %143 = tail call i32 @llvm.usub.sat.i32(i32 %142, i32 4194304)
+  %144 = tail call i32 @llvm.umin.i32(i32 %143, i32 7)
+  %145 = trunc nuw nsw i32 %144 to i8
   %146 = getelementptr inbounds nuw [32 x i8], ptr %9, i64 0, i64 %135
   store i8 %145, ptr %146, align 1, !tbaa !13
   %indvars.iv.next520 = add nuw nsw i64 %indvars.iv519, 1
@@ -14473,12 +14473,12 @@ iq3_find_best_neighbour.exit:                     ; preds = %196, %147
   %254 = fadd float %253, 0x4168000000000000
   %255 = bitcast float %254 to i32
   %256 = and i32 %255, 8388607
-  %257 = tail call i32 @llvm.umin.i32(i32 %256, i32 4194311)
-  %258 = tail call i32 @llvm.umax.i32(i32 %257, i32 4194304)
+  %257 = tail call i32 @llvm.usub.sat.i32(i32 %256, i32 4194304)
+  %258 = tail call i32 @llvm.umin.i32(i32 %257, i32 7)
   %259 = trunc i64 %indvars.iv545 to i32
   %260 = mul i32 %259, 3
-  %261 = shl i32 %258, %260
-  %262 = trunc i32 %261 to i16
+  %261 = shl nuw nsw i32 %258, %260
+  %262 = trunc nuw nsw i32 %261 to i16
   %263 = or i16 %.0336459, %262
   %indvars.iv.next546 = add nuw nsw i64 %indvars.iv545, 1
   %exitcond547.not = icmp eq i64 %indvars.iv.next546, 4
@@ -14795,9 +14795,9 @@ iq3_find_best_neighbour.exit406:                  ; preds = %298, %.iq3_find_bes
   %426 = fadd float %425, 0x4168000000000000
   %427 = bitcast float %426 to i32
   %428 = and i32 %427, 8388607
-  %429 = tail call i32 @llvm.umin.i32(i32 %428, i32 4194319)
-  %430 = tail call i32 @llvm.umax.i32(i32 %429, i32 4194304)
-  %431 = shl i32 %430, 28
+  %429 = tail call i32 @llvm.usub.sat.i32(i32 %428, i32 4194304)
+  %430 = tail call i32 @llvm.umin.i32(i32 %429, i32 15)
+  %431 = shl nuw i32 %430, 28
   %432 = getelementptr inbounds nuw i32, ptr %29, i64 %indvars.iv575
   %433 = load i32, ptr %432, align 4, !tbaa !90
   %434 = or i32 %431, %433
@@ -15075,9 +15075,9 @@ define i64 @quantize_iq3_s(ptr noalias noundef readonly captures(none) %0, ptr n
   %102 = fadd float %101, 0x4168000000000000
   %103 = bitcast float %102 to i32
   %104 = and i32 %103, 8388607
-  %105 = tail call i32 @llvm.umin.i32(i32 %104, i32 4194311)
-  %106 = tail call i32 @llvm.umax.i32(i32 %105, i32 4194304)
-  %107 = trunc i32 %106 to i8
+  %105 = tail call i32 @llvm.usub.sat.i32(i32 %104, i32 4194304)
+  %106 = tail call i32 @llvm.umin.i32(i32 %105, i32 7)
+  %107 = trunc nuw nsw i32 %106 to i8
   %108 = getelementptr inbounds nuw i8, ptr %10, i64 %97
   store i8 %107, ptr %108, align 1, !tbaa !13, !noalias !942
   %indvars.iv.next555.i.us = add nuw nsw i64 %indvars.iv554.i.us, 1
@@ -15282,12 +15282,12 @@ iq3_find_best_neighbour.exit.i.us:                ; preds = %154, %117
   %200 = fadd float %199, 0x4168000000000000
   %201 = bitcast float %200 to i32
   %202 = and i32 %201, 8388607
-  %203 = tail call i32 @llvm.umin.i32(i32 %202, i32 4194311)
-  %204 = tail call i32 @llvm.umax.i32(i32 %203, i32 4194304)
+  %203 = tail call i32 @llvm.usub.sat.i32(i32 %202, i32 4194304)
+  %204 = tail call i32 @llvm.umin.i32(i32 %203, i32 7)
   %205 = trunc i64 %indvars.iv582.i.us to i32
   %206 = mul i32 %205, 3
-  %207 = shl i32 %204, %206
-  %208 = trunc i32 %207 to i16
+  %207 = shl nuw nsw i32 %204, %206
+  %208 = trunc nuw nsw i32 %207 to i16
   %209 = or i16 %.0380494.i.us, %208
   %indvars.iv.next583.i.us = add nuw nsw i64 %indvars.iv582.i.us, 1
   %exitcond584.not.i.us = icmp eq i64 %indvars.iv.next583.i.us, 4
@@ -15577,8 +15577,8 @@ iq3_find_best_neighbour.exit443.i.us:             ; preds = %246, %.iq3_find_bes
   %353 = fadd float %352, 0x4168000000000000
   %354 = bitcast float %353 to i32
   %355 = and i32 %354, 8388607
-  %356 = tail call i32 @llvm.umin.i32(i32 %355, i32 4194319)
-  %357 = tail call i32 @llvm.umax.i32(i32 %356, i32 4194304)
+  %356 = tail call i32 @llvm.usub.sat.i32(i32 %355, i32 4194304)
+  %357 = tail call i32 @llvm.umin.i32(i32 %356, i32 15)
   %358 = getelementptr inbounds nuw i8, ptr %349, i64 4
   %359 = load float, ptr %358, align 4, !tbaa !8, !noalias !942
   %360 = tail call float @llvm.fmuladd.f32(float %346, float %359, float -1.000000e+00)
@@ -15586,11 +15586,11 @@ iq3_find_best_neighbour.exit443.i.us:             ; preds = %246, %.iq3_find_bes
   %362 = fadd float %361, 0x4168000000000000
   %363 = bitcast float %362 to i32
   %364 = and i32 %363, 8388607
-  %365 = tail call i32 @llvm.umin.i32(i32 %364, i32 4194319)
-  %366 = tail call i32 @llvm.umax.i32(i32 %365, i32 4194304)
+  %365 = tail call i32 @llvm.usub.sat.i32(i32 %364, i32 4194304)
+  %366 = tail call i32 @llvm.umin.i32(i32 %365, i32 15)
   %367 = shl nuw nsw i32 %366, 4
-  %368 = add nuw nsw i32 %367, %357
-  %369 = trunc i32 %368 to i8
+  %368 = or disjoint i32 %367, %357
+  %369 = trunc nuw i32 %368 to i8
   %370 = lshr exact i64 %indvars.iv616.i.us, 1
   %371 = getelementptr inbounds nuw [4 x i8], ptr %347, i64 0, i64 %370
   store i8 %369, ptr %371, align 1, !tbaa !13, !alias.scope !948, !noalias !951
@@ -16269,8 +16269,8 @@ define i64 @quantize_iq1_s(ptr noalias noundef readonly captures(none) %0, ptr n
   %286 = fadd float %285, 0x4168000000000000
   %287 = bitcast float %286 to i32
   %288 = and i32 %287, 8388607
-  %289 = call i32 @llvm.umin.i32(i32 %288, i32 4194311)
-  %290 = call i32 @llvm.umax.i32(i32 %289, i32 4194304)
+  %289 = call i32 @llvm.usub.sat.i32(i32 %288, i32 4194304)
+  %290 = call i32 @llvm.umin.i32(i32 %289, i32 7)
   %291 = getelementptr inbounds nuw i8, ptr %13, i64 %indvars.iv486.i.us
   %292 = load i8, ptr %291, align 1, !tbaa !13, !noalias !1021
   %293 = icmp eq i8 %292, -1
@@ -16278,7 +16278,7 @@ define i64 @quantize_iq1_s(ptr noalias noundef readonly captures(none) %0, ptr n
   %spec.select.i.us = select i1 %293, i32 %294, i32 %290
   %295 = getelementptr inbounds nuw [8 x i16], ptr %40, i64 0, i64 %indvars.iv486.i.us
   %296 = load i16, ptr %295, align 2, !tbaa !635, !alias.scope !1016, !noalias !1022
-  %.0.tr.i.us = trunc i32 %spec.select.i.us to i16
+  %.0.tr.i.us = trunc nuw nsw i32 %spec.select.i.us to i16
   %297 = shl nuw i16 %.0.tr.i.us, 12
   %298 = or i16 %297, %296
   store i16 %298, ptr %295, align 2, !tbaa !635, !alias.scope !1016, !noalias !1022
@@ -17130,17 +17130,17 @@ define i64 @quantize_iq1_m(ptr noalias noundef readonly captures(none) %0, ptr n
   %382 = fadd float %381, 0x4168000000000000
   %383 = bitcast float %382 to i32
   %384 = and i32 %383, 8388607
-  %385 = call i32 @llvm.umin.i32(i32 %384, i32 4194311)
-  %386 = call i32 @llvm.umax.i32(i32 %385, i32 4194304)
+  %385 = call i32 @llvm.usub.sat.i32(i32 %384, i32 4194304)
+  %386 = call i32 @llvm.umin.i32(i32 %385, i32 7)
   %387 = trunc nuw nsw i64 %indvars.iv823.i.us to i32
   %388 = and i32 %387, 3
   %389 = mul nuw nsw i32 %388, 3
-  %390 = shl nuw i32 %386, %389
+  %390 = shl nuw nsw i32 %386, %389
   %391 = lshr i64 %indvars.iv823.i.us, 2
   %392 = and i64 %391, 1073741823
   %393 = getelementptr inbounds nuw i16, ptr %44, i64 %392
   %394 = load i16, ptr %393, align 2, !tbaa !635, !alias.scope !1047, !noalias !1052
-  %395 = trunc i32 %390 to i16
+  %395 = trunc nuw nsw i32 %390 to i16
   %396 = or i16 %394, %395
   store i16 %396, ptr %393, align 2, !tbaa !635, !alias.scope !1047, !noalias !1052
   %397 = getelementptr inbounds nuw i8, ptr %13, i64 %indvars.iv823.i.us
@@ -17190,7 +17190,7 @@ define i64 @quantize_iq1_m(ptr noalias noundef readonly captures(none) %0, ptr n
   %422 = shl nuw nsw i64 %indvars.iv823.i.us, 1
   %423 = zext i8 %404 to i32
   %424 = shl nuw nsw i32 %386, 1
-  %425 = add nsw i32 %424, -8388607
+  %425 = or disjoint i32 %424, 1
   %426 = uitofp nneg i32 %425 to float
   %427 = and i8 %398, 1
   %428 = icmp eq i8 %427, 0
@@ -17798,8 +17798,8 @@ best_index_int8.exit265:                          ; preds = %107, %112, %123
   %184 = fadd float %183, 0x4168000000000000
   %185 = bitcast float %184 to i32
   %186 = and i32 %185, 8388607
-  %187 = tail call i32 @llvm.umin.i32(i32 %186, i32 4194335)
-  %188 = tail call i32 @llvm.umax.i32(i32 %187, i32 4194272)
+  %187 = tail call i32 @llvm.umax.i32(i32 %186, i32 4194272)
+  %188 = tail call i32 @llvm.umin.i32(i32 %187, i32 4194335)
   %189 = add nsw i32 %188, -4194304
   %190 = sitofp i32 %189 to float
   %191 = fmul float %156, %190
@@ -18456,9 +18456,9 @@ define i64 @quantize_iq2_s(ptr noalias noundef readonly captures(none) %0, ptr n
   %96 = fadd float %95, 0x4168000000000000
   %97 = bitcast float %96 to i32
   %98 = and i32 %97, 8388607
-  %99 = tail call i32 @llvm.umin.i32(i32 %98, i32 4194306)
-  %100 = tail call i32 @llvm.umax.i32(i32 %99, i32 4194304)
-  %101 = trunc i32 %100 to i8
+  %99 = tail call i32 @llvm.usub.sat.i32(i32 %98, i32 4194304)
+  %100 = tail call i32 @llvm.umin.i32(i32 %99, i32 2)
+  %101 = trunc nuw nsw i32 %100 to i8
   %102 = getelementptr inbounds nuw [16 x i8], ptr %10, i64 0, i64 %91
   store i8 %101, ptr %102, align 1, !tbaa !13, !noalias !1105
   %indvars.iv.next447.i.us = add nuw nsw i64 %indvars.iv446.i.us, 1
@@ -18668,14 +18668,14 @@ iq2_find_best_neighbour.exit.i.us:                ; preds = %145, %111
   %193 = fadd float %192, 0x4168000000000000
   %194 = bitcast float %193 to i32
   %195 = and i32 %194, 8388607
-  %196 = tail call i32 @llvm.umin.i32(i32 %195, i32 4194306)
-  %197 = tail call i32 @llvm.umax.i32(i32 %196, i32 4194304)
+  %196 = tail call i32 @llvm.usub.sat.i32(i32 %195, i32 4194304)
+  %197 = tail call i32 @llvm.umin.i32(i32 %196, i32 2)
   %indvars.iv470.tr.i.us = trunc i64 %indvars.iv470.i.us to i32
   %198 = shl i32 %indvars.iv470.tr.i.us, 1
-  %199 = shl i32 %197, %198
-  %200 = trunc i32 %199 to i16
+  %199 = shl nuw nsw i32 %197, %198
+  %200 = trunc nuw i32 %199 to i16
   %201 = or i16 %.0288394.i.us, %200
-  %202 = trunc i32 %197 to i8
+  %202 = trunc nuw nsw i32 %197 to i8
   %203 = getelementptr inbounds nuw [16 x i8], ptr %9, i64 0, i64 %188
   store i8 %202, ptr %203, align 1, !tbaa !13, !noalias !1105
   %indvars.iv.next471.i.us = add nuw nsw i64 %indvars.iv470.i.us, 1
@@ -18935,8 +18935,8 @@ iq2_find_best_neighbour.exit345.i.us:             ; preds = %238, %204, %185
   %343 = fadd float %342, 0x4168000000000000
   %344 = bitcast float %343 to i32
   %345 = and i32 %344, 8388607
-  %346 = tail call i32 @llvm.umin.i32(i32 %345, i32 4194319)
-  %347 = tail call i32 @llvm.umax.i32(i32 %346, i32 4194304)
+  %346 = tail call i32 @llvm.usub.sat.i32(i32 %345, i32 4194304)
+  %347 = tail call i32 @llvm.umin.i32(i32 %346, i32 15)
   %348 = and i64 %indvars.iv493.i.us, 1
   %349 = icmp eq i64 %348, 0
   br i1 %349, label %357, label %350
@@ -18946,14 +18946,14 @@ iq2_find_best_neighbour.exit345.i.us:             ; preds = %238, %204, %185
   %352 = and i64 %351, 2147483647
   %353 = getelementptr inbounds nuw [8 x i8], ptr %337, i64 0, i64 %352
   %354 = load i8, ptr %353, align 1, !tbaa !13, !alias.scope !1111, !noalias !1114
-  %.tr.i.us = trunc i32 %347 to i8
+  %.tr.i.us = trunc nuw nsw i32 %347 to i8
   %355 = shl nuw i8 %.tr.i.us, 4
   %356 = or i8 %354, %355
   store i8 %356, ptr %353, align 1, !tbaa !13, !alias.scope !1111, !noalias !1114
   br label %362
 
 357:                                              ; preds = %338
-  %358 = trunc i32 %347 to i8
+  %358 = trunc nuw nsw i32 %347 to i8
   %359 = lshr exact i64 %indvars.iv493.i.us, 1
   %360 = and i64 %359, 2147483647
   %361 = getelementptr inbounds nuw [8 x i8], ptr %337, i64 0, i64 %360
@@ -20285,6 +20285,9 @@ declare noundef i32 @puts(ptr noundef readonly captures(none)) local_unnamed_add
 ; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: readwrite)
 declare void @llvm.experimental.noalias.scope.decl(metadata) #22
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.usub.sat.i32(i32, i32) #20
+
 attributes #0 = { nofree norecurse nosync nounwind memory(argmem: readwrite) uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
 attributes #2 = { mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none) }
diff --git a/bench/lodepng/optimized/pngdetail.ll b/bench/lodepng/optimized/pngdetail.ll
index 19556f050e5..c501026c925 100644
--- a/bench/lodepng/optimized/pngdetail.ll
+++ b/bench/lodepng/optimized/pngdetail.ll
@@ -2649,7 +2649,7 @@ _Z11HueToLetteri.exit47:                          ; preds = %_Z8RGBtoHSLhhhPhS_S
   br i1 %or.cond5, label %186, label %202
 
 186:                                              ; preds = %_Z11HueToLetteri.exit47.thread, %_Z11HueToLetteri.exit47
-  %spec.select110117123 = phi i8 [ %spec.select110114, %_Z11HueToLetteri.exit47.thread ], [ %spec.select110, %_Z11HueToLetteri.exit47 ]
+  %spec.select110117123 = phi i8 [ %spec.select110114, %_Z11HueToLetteri.exit44.thread ], [ %spec.select110, %_Z11HueToLetteri.exit44 ]
   %187 = icmp ult i8 %spec.select110117123, 16
   br i1 %187, label %_Z17lightnessToLetteri.exit, label %188
 
diff --git a/bench/luau/optimized/isocline.ll b/bench/luau/optimized/isocline.ll
index 063cfc3a832..46c6e91410c 100644
--- a/bench/luau/optimized/isocline.ll
+++ b/bench/luau/optimized/isocline.ll
@@ -5390,8 +5390,8 @@ ic_get_env.exit.thread8:                          ; preds = %1, %ic_get_env.exit
   %8 = phi ptr [ %.pre.i, %ic_get_env.exit ], [ %2, %1 ]
   %9 = getelementptr inbounds nuw i8, ptr %8, i64 120
   %10 = load i64, ptr %9, align 8, !tbaa !163
-  %11 = tail call i64 @llvm.smin.i64(i64 %0, i64 5000)
-  %12 = tail call i64 @llvm.smax.i64(i64 %11, i64 0)
+  %11 = tail call i64 @llvm.smax.i64(i64 %0, i64 0)
+  %12 = tail call i64 @llvm.umin.i64(i64 %11, i64 5000)
   store i64 %12, ptr %9, align 8, !tbaa !163
   br label %ic_get_env.exit.thread
 
@@ -5426,12 +5426,12 @@ ic_get_env.exit.thread5:                          ; preds = %2, %ic_get_env.exit
   br i1 %12, label %ic_get_env.exit.thread, label %13
 
 13:                                               ; preds = %ic_get_env.exit.thread5
-  %14 = tail call i64 @llvm.smin.i64(i64 %0, i64 1000)
-  %15 = tail call i64 @llvm.smax.i64(i64 %14, i64 0)
+  %14 = tail call i64 @llvm.smax.i64(i64 %0, i64 0)
+  %15 = tail call i64 @llvm.umin.i64(i64 %14, i64 1000)
   %16 = getelementptr inbounds nuw i8, ptr %11, i64 192
   store i64 %15, ptr %16, align 8, !tbaa !164
-  %17 = tail call i64 @llvm.smin.i64(i64 %1, i64 1000)
-  %18 = tail call i64 @llvm.smax.i64(i64 %17, i64 0)
+  %17 = tail call i64 @llvm.smax.i64(i64 %1, i64 0)
+  %18 = tail call i64 @llvm.umin.i64(i64 %17, i64 1000)
   %19 = getelementptr inbounds nuw i8, ptr %11, i64 200
   store i64 %18, ptr %19, align 8, !tbaa !165
   br label %ic_get_env.exit.thread
@@ -9430,15 +9430,15 @@ color_from_ansi256.exit:                          ; preds = %100, %105, %108
   %133 = load i64, ptr %6, align 8, !tbaa !57
   %134 = load i64, ptr %7, align 8, !tbaa !57
   %135 = load i64, ptr %8, align 8, !tbaa !57
-  %136 = call i64 @llvm.smin.i64(i64 %133, i64 255)
-  %137 = call i64 @llvm.smax.i64(i64 %136, i64 0)
+  %136 = call i64 @llvm.smax.i64(i64 %133, i64 0)
+  %137 = call i64 @llvm.umin.i64(i64 %136, i64 255)
   %138 = shl nuw nsw i64 %137, 16
-  %139 = call i64 @llvm.smin.i64(i64 %134, i64 255)
-  %140 = call i64 @llvm.smax.i64(i64 %139, i64 0)
+  %139 = call i64 @llvm.smax.i64(i64 %134, i64 0)
+  %140 = call i64 @llvm.umin.i64(i64 %139, i64 255)
   %141 = shl nuw nsw i64 %140, 8
   %142 = or disjoint i64 %141, %138
-  %143 = call i64 @llvm.smin.i64(i64 %135, i64 255)
-  %144 = call i64 @llvm.smax.i64(i64 %143, i64 0)
+  %143 = call i64 @llvm.smax.i64(i64 %135, i64 0)
+  %144 = call i64 @llvm.umin.i64(i64 %143, i64 255)
   %145 = or disjoint i64 %142, %144
   %146 = or disjoint i64 %145, 16777216
   %147 = load i64, ptr %4, align 8, !tbaa !57
diff --git a/bench/lvgl/optimized/lv_anim.ll b/bench/lvgl/optimized/lv_anim.ll
index 92b199605d3..02e1d2cf8b0 100644
--- a/bench/lvgl/optimized/lv_anim.ll
+++ b/bench/lvgl/optimized/lv_anim.ll
@@ -1289,8 +1289,8 @@ define i32 @lv_anim_path_bounce(ptr noundef readonly captures(none) %0) local_un
 38:                                               ; preds = %18, %29, %35, %33, %24, %13
   %.046 = phi i32 [ %11, %13 ], [ %21, %18 ], [ %26, %24 ], [ %32, %29 ], [ %37, %35 ], [ %11, %33 ]
   %.0 = phi i32 [ %narrow, %13 ], [ %20, %18 ], [ %25, %24 ], [ %31, %29 ], [ %36, %35 ], [ %6, %33 ]
-  %spec.store.select = tail call i32 @llvm.smin.i32(i32 %.0, i32 1024)
-  %spec.store.select9 = tail call i32 @llvm.smax.i32(i32 %spec.store.select, i32 0)
+  %spec.store.select = tail call i32 @llvm.smax.i32(i32 %.0, i32 0)
+  %spec.store.select9 = tail call i32 @llvm.umin.i32(i32 %spec.store.select, i32 1024)
   %39 = tail call i32 @lv_bezier3(i32 noundef %spec.store.select9, i32 noundef 0, i32 noundef 500, i32 noundef 800, i32 noundef 1024) #10
   %40 = mul nsw i32 %39, %.046
   %41 = ashr i32 %40, 10
@@ -1707,9 +1707,6 @@ declare i32 @lv_cubic_bezier(i32 noundef, i32 noundef, i32 noundef, i32 noundef,
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.umax.i32(i32, i32) #9
 
-; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smin.i32(i32, i32) #9
-
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smax.i32(i32, i32) #9
 
diff --git a/bench/lvgl/optimized/lv_roller.ll b/bench/lvgl/optimized/lv_roller.ll
index 3177a035610..1ff3764a620 100644
--- a/bench/lvgl/optimized/lv_roller.ll
+++ b/bench/lvgl/optimized/lv_roller.ll
@@ -831,8 +831,8 @@ define void @lv_roller_set_options(ptr noundef %0, ptr noundef %1, i32 noundef %
   %34 = add nsw i32 %31, %.sroa.0.0.extract.trunc.i
   %35 = mul i32 %34, %30
   %36 = sdiv i32 1000, %35
-  %spec.select67 = tail call i32 @llvm.smin.i32(i32 %36, i32 15)
-  %37 = tail call i32 @llvm.smax.i32(i32 %spec.select67, i32 3)
+  %spec.select67 = tail call i32 @llvm.smax.i32(i32 %36, i32 3)
+  %37 = tail call i32 @llvm.umin.i32(i32 %spec.select67, i32 15)
   %38 = getelementptr inbounds nuw i8, ptr %0, i64 76
   %spec.select68 = or i32 %37, 1
   store i32 %spec.select68, ptr %38, align 4, !tbaa !42
@@ -1827,6 +1827,9 @@ declare i32 @llvm.abs.i32(i32, i1 immarg) #5
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smin.i32(i32, i32) #5
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #5
+
 attributes #0 = { nounwind uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
 attributes #2 = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
diff --git a/bench/meshlab/optimized/gltf_loader.ll b/bench/meshlab/optimized/gltf_loader.ll
index bf12aad5443..6242ded1316 100644
--- a/bench/meshlab/optimized/gltf_loader.ll
+++ b/bench/meshlab/optimized/gltf_loader.ll
@@ -40006,8 +40006,8 @@ define internal fastcc noundef range(i32 0, 2) i32 @_ZL19stbi_write_jpg_coreP19s
 
 30:                                               ; preds = %6
   %.not = icmp eq i32 %5, 0
-  %31 = tail call i32 @llvm.smin.i32(i32 %5, i32 100)
-  %32 = tail call i32 @llvm.smax.i32(i32 %31, i32 1)
+  %31 = tail call i32 @llvm.smax.i32(i32 %5, i32 1)
+  %32 = tail call i32 @llvm.umin.i32(i32 %31, i32 100)
   %33 = select i1 %.not, i32 90, i32 %32
   %34 = icmp samesign ult i32 %33, 50
   br i1 %34, label %35, label %37
@@ -40034,8 +40034,8 @@ define internal fastcc noundef range(i32 0, 2) i32 @_ZL19stbi_write_jpg_coreP19s
   %45 = mul nsw i32 %44, %41
   %46 = add nsw i32 %45, 50
   %47 = sdiv i32 %46, 100
-  %48 = tail call i32 @llvm.smin.i32(i32 %47, i32 255)
-  %49 = tail call i32 @llvm.smax.i32(i32 %48, i32 1)
+  %48 = tail call i32 @llvm.smax.i32(i32 %47, i32 1)
+  %49 = tail call i32 @llvm.umin.i32(i32 %48, i32 255)
   %50 = trunc nuw i32 %49 to i8
   %51 = getelementptr inbounds nuw [64 x i8], ptr @_ZL17stbiw__jpg_ZigZag, i64 0, i64 %indvars.iv
   %52 = load i8, ptr %51, align 1
@@ -40047,8 +40047,8 @@ define internal fastcc noundef range(i32 0, 2) i32 @_ZL19stbi_write_jpg_coreP19s
   %57 = mul nsw i32 %56, %41
   %58 = add nsw i32 %57, 50
   %59 = sdiv i32 %58, 100
-  %60 = tail call i32 @llvm.smin.i32(i32 %59, i32 255)
-  %61 = tail call i32 @llvm.smax.i32(i32 %60, i32 1)
+  %60 = tail call i32 @llvm.smax.i32(i32 %59, i32 1)
+  %61 = tail call i32 @llvm.umin.i32(i32 %60, i32 255)
   %62 = trunc nuw i32 %61 to i8
   %63 = getelementptr inbounds nuw [64 x i8], ptr %18, i64 0, i64 %53
   store i8 %62, ptr %63, align 1
diff --git a/bench/meshlab/optimized/mainwindow_RunTime.ll b/bench/meshlab/optimized/mainwindow_RunTime.ll
index cd30ab89215..2af8edfa968 100644
--- a/bench/meshlab/optimized/mainwindow_RunTime.ll
+++ b/bench/meshlab/optimized/mainwindow_RunTime.ll
@@ -1264,9 +1264,9 @@ _ZN7QStringD2Ev.exit:                             ; preds = %40, %_ZN9QtPrivate8
   br i1 %51, label %.lr.ph, label %.lr.ph78
 
 .lr.ph:                                           ; preds = %_ZN7QStringD2Ev.exit
-  %52 = call i32 @llvm.umin.i32(i32 %50, i32 9)
-  %53 = getelementptr inbounds nuw i8, ptr %0, i64 704
-  %wide.trip.count = zext nneg i32 %52 to i64
+  %52 = getelementptr inbounds nuw i8, ptr %0, i64 704
+  %53 = call i32 @llvm.umin.i32(i32 %50, i32 9)
+  %wide.trip.count = zext nneg i32 %53 to i64
   br label %57
 
 .preheader:                                       ; preds = %_ZN7QStringD2Ev.exit69
@@ -1391,7 +1391,7 @@ _ZN9QtPrivate8RefCount5derefEv.exit.thread2.i55:  ; preds = %_ZN9QtPrivate8RefCo
   br label %_ZN7QStringD2Ev.exit60
 
 _ZN7QStringD2Ev.exit60:                           ; preds = %_ZN7QStringD2Ev.exit54, %_ZN9QtPrivate8RefCount5derefEv.exit.i56, %_ZN9QtPrivate8RefCount5derefEv.exit.thread2.i55
-  %87 = getelementptr inbounds nuw [9 x ptr], ptr %53, i64 0, i64 %indvars.iv
+  %87 = getelementptr inbounds nuw [9 x ptr], ptr %52, i64 0, i64 %indvars.iv
   %88 = load ptr, ptr %87, align 8
   invoke void @_ZN7QAction7setTextERK7QString(ptr noundef nonnull align 8 dereferenceable(16) %88, ptr noundef nonnull align 8 dereferenceable(8) %8)
           to label %89 unwind label %133
@@ -2035,9 +2035,9 @@ _ZN7QStringD2Ev.exit:                             ; preds = %17, %_ZN9QtPrivate8
   br i1 %28, label %.lr.ph, label %.lr.ph70
 
 .lr.ph:                                           ; preds = %_ZN7QStringD2Ev.exit
-  %29 = call i32 @llvm.umin.i32(i32 %27, i32 9)
-  %30 = getelementptr inbounds nuw i8, ptr %0, i64 776
-  %wide.trip.count = zext nneg i32 %29 to i64
+  %29 = getelementptr inbounds nuw i8, ptr %0, i64 776
+  %30 = call i32 @llvm.umin.i32(i32 %27, i32 9)
+  %wide.trip.count = zext nneg i32 %30 to i64
   br label %34
 
 .preheader:                                       ; preds = %_ZN7QStringD2Ev.exit61
@@ -2162,7 +2162,7 @@ _ZN9QtPrivate8RefCount5derefEv.exit.thread2.i47:  ; preds = %_ZN9QtPrivate8RefCo
   br label %_ZN7QStringD2Ev.exit52
 
 _ZN7QStringD2Ev.exit52:                           ; preds = %_ZN7QStringD2Ev.exit46, %_ZN9QtPrivate8RefCount5derefEv.exit.i48, %_ZN9QtPrivate8RefCount5derefEv.exit.thread2.i47
-  %64 = getelementptr inbounds nuw [9 x ptr], ptr %30, i64 0, i64 %indvars.iv
+  %64 = getelementptr inbounds nuw [9 x ptr], ptr %29, i64 0, i64 %indvars.iv
   %65 = load ptr, ptr %64, align 8
   invoke void @_ZN7QAction7setTextERK7QString(ptr noundef nonnull align 8 dereferenceable(16) %65, ptr noundef nonnull align 8 dereferenceable(8) %7)
           to label %66 unwind label %108
diff --git a/bench/minetest/optimized/l_env.ll b/bench/minetest/optimized/l_env.ll
index 28f6a2c0489..f7f640ea451 100644
--- a/bench/minetest/optimized/l_env.ll
+++ b/bench/minetest/optimized/l_env.ll
@@ -5865,9 +5865,9 @@ for.cond9.preheader.i.i.i:                        ; preds = %for.cond9.for.inc23
   %cond98.i.i.i = call i32 @llvm.smax.i32(i32 %cond96.i.i.i, i32 0)
   %conv99.i.i.i = trunc nuw nsw i32 %cond98.i.i.i to i16
   %sub156.i.i.i = sub nsw i32 %conv.i13.i.i320.i.i.i, %conv77.i.i.i
-  %cond174.i.i.i = call i32 @llvm.smin.i32(i32 %sub156.i.i.i, i32 15)
-  %cond176.i.i.i = call i32 @llvm.smax.i32(i32 %cond174.i.i.i, i32 0)
-  %cmp181.not372.i.i.i = icmp sgt i32 %cond96.i.i.i, %cond176.i.i.i
+  %cond174.i.i.i = call i32 @llvm.smax.i32(i32 %sub156.i.i.i, i32 0)
+  %cond176.i.i.i = call i32 @llvm.umin.i32(i32 %cond174.i.i.i, i32 15)
+  %cmp181.not372.i.i.i = icmp slt i32 %47, %cond96.i.i.i
   %cmp181.not372.fr.i.i.i = freeze i1 %cmp181.not372.i.i.i
   br i1 %cmp181.not372.fr.i.i.i, label %for.cond16.preheader.us385.i.i.i, label %for.cond16.preheader.i.i.i
 
@@ -5907,9 +5907,9 @@ for.cond16.preheader.i.i.i:                       ; preds = %for.cond9.preheader
   %cond46.i.i.i = call i32 @llvm.smax.i32(i32 %cond.i.i.i, i32 0)
   %conv47.i.i.i = trunc nuw nsw i32 %cond46.i.i.i to i16
   %sub104.i.i.i = sub nsw i32 %conv.i.i.i309.i.i.i, %conv30.i.i.i
-  %cond122.i.i.i = call i32 @llvm.smin.i32(i32 %sub104.i.i.i, i32 15)
-  %cond124.i.i.i = call i32 @llvm.smax.i32(i32 %cond122.i.i.i, i32 0)
-  %cmp193.not366.i.i.i = icmp sgt i32 %cond.i.i.i, %cond124.i.i.i
+  %cond122.i.i.i = call i32 @llvm.smax.i32(i32 %sub104.i.i.i, i32 0)
+  %cond124.i.i.i = call i32 @llvm.umin.i32(i32 %cond122.i.i.i, i32 15)
+  %cmp193.not366.i.i.i = icmp slt i32 %cond122.i.i.i, %cond.i.i.i
   %cmp193.not366.i.fr.i.i = freeze i1 %cmp193.not366.i.i.i
   %invariant.op133 = or disjoint i48 %bp.sroa.7.0.insert.shift.i.i.i, %bp.sroa.0.0.insert.ext.i.i.i
   br i1 %cmp193.not366.i.fr.i.i, label %for.body22.i.us.i.i, label %for.body22.i.i.i
@@ -5944,9 +5944,9 @@ call24.i.i.noexc.i:                               ; preds = %for.body22.i.i.i
   %cond72.i.i.i = call i32 @llvm.smax.i32(i32 %cond70.i.i.i, i32 0)
   %conv73.i.i.i = trunc nuw nsw i32 %cond72.i.i.i to i16
   %sub130.i.i.i = sub nsw i32 %conv127.i.i.i, %conv51.i.i.i
-  %cond148.i.i.i = call i32 @llvm.smin.i32(i32 %sub130.i.i.i, i32 15)
-  %cond150.i.i.i = call i32 @llvm.smax.i32(i32 %cond148.i.i.i, i32 0)
-  %cmp187.not369.i.i.i = icmp sgt i32 %cond70.i.i.i, %cond150.i.i.i
+  %cond148.i.i.i = call i32 @llvm.smax.i32(i32 %sub130.i.i.i, i32 0)
+  %cond150.i.i.i = call i32 @llvm.umin.i32(i32 %cond148.i.i.i, i32 15)
+  %cmp187.not369.i.i.i = icmp slt i32 %51, %cond70.i.i.i
   %tobool.not.i.i99.i = icmp eq ptr %call24.i.i102.i, null
   %data.i.i.i.i = getelementptr inbounds nuw i8, ptr %call24.i.i102.i, i64 24
   br i1 %cmp187.not369.i.i.i, label %for.cond178.cleanup215_crit_edge.i.i.i, label %for.cond184.preheader.i.i.i
@@ -6384,9 +6384,9 @@ for.cond9.preheader.i.i205.i:                     ; preds = %for.cond9.for.inc23
   %cond98.i.i212.i = call i32 @llvm.smax.i32(i32 %cond96.i.i211.i, i32 0)
   %conv99.i.i213.i = trunc nuw nsw i32 %cond98.i.i212.i to i16
   %sub156.i.i214.i = sub nsw i32 %conv.i13.i.i320.i.i187.i, %conv77.i.i209.i
-  %cond174.i.i215.i = call i32 @llvm.smin.i32(i32 %sub156.i.i214.i, i32 15)
-  %cond176.i.i216.i = call i32 @llvm.smax.i32(i32 %cond174.i.i215.i, i32 0)
-  %cmp181.not370.i.i.i = icmp sgt i32 %cond96.i.i211.i, %cond176.i.i216.i
+  %cond174.i.i215.i = call i32 @llvm.smax.i32(i32 %sub156.i.i214.i, i32 0)
+  %cond176.i.i216.i = call i32 @llvm.umin.i32(i32 %cond174.i.i215.i, i32 15)
+  %cmp181.not370.i.i.i = icmp slt i32 %87, %cond96.i.i211.i
   %cmp181.not370.fr.i.i.i = freeze i1 %cmp181.not370.i.i.i
   br i1 %cmp181.not370.fr.i.i.i, label %for.cond16.preheader.us383.i.i.i, label %for.cond16.preheader.i.i217.i
 
@@ -6427,9 +6427,9 @@ for.cond16.preheader.i.i217.i:                    ; preds = %for.cond9.preheader
   %cond46.i.i223.i = call i32 @llvm.smax.i32(i32 %cond.i.i222.i, i32 0)
   %conv47.i.i224.i = trunc nuw nsw i32 %cond46.i.i223.i to i16
   %sub104.i.i225.i = sub nsw i32 %conv.i.i.i309.i.i176.i, %conv30.i.i220.i
-  %cond122.i.i226.i = call i32 @llvm.smin.i32(i32 %sub104.i.i225.i, i32 15)
-  %cond124.i.i227.i = call i32 @llvm.smax.i32(i32 %cond122.i.i226.i, i32 0)
-  %cmp193.not364.i.i.i = icmp sgt i32 %cond.i.i222.i, %cond124.i.i227.i
+  %cond122.i.i226.i = call i32 @llvm.smax.i32(i32 %sub104.i.i225.i, i32 0)
+  %cond124.i.i227.i = call i32 @llvm.umin.i32(i32 %cond122.i.i226.i, i32 15)
+  %cmp193.not364.i.i.i = icmp slt i32 %89, %cond.i.i222.i
   %cmp193.not364.i.fr.i.i = freeze i1 %cmp193.not364.i.i.i
   %invariant.op122 = or disjoint i48 %bp.sroa.7.0.insert.shift.i.i207.i, %bp.sroa.0.0.insert.ext.i.i218.i
   br i1 %cmp193.not364.i.fr.i.i, label %for.body22.i.us.i352.i, label %for.body22.i.i228.i
@@ -6465,9 +6465,9 @@ call24.i.i.noexc372.i:                            ; preds = %for.body22.i.i228.i
   %cond72.i.i236.i = call i32 @llvm.smax.i32(i32 %cond70.i.i235.i, i32 0)
   %conv73.i.i237.i = trunc nuw nsw i32 %cond72.i.i236.i to i16
   %sub130.i.i238.i = sub nsw i32 %conv127.i.i200.i, %conv51.i.i233.i
-  %cond148.i.i239.i = call i32 @llvm.smin.i32(i32 %sub130.i.i238.i, i32 15)
-  %cond150.i.i240.i = call i32 @llvm.smax.i32(i32 %cond148.i.i239.i, i32 0)
-  %cmp187.not367.i.i.i = icmp sgt i32 %cond70.i.i235.i, %cond150.i.i240.i
+  %cond148.i.i239.i = call i32 @llvm.smax.i32(i32 %sub130.i.i238.i, i32 0)
+  %cond150.i.i240.i = call i32 @llvm.umin.i32(i32 %cond148.i.i239.i, i32 15)
+  %cmp187.not367.i.i.i = icmp slt i32 %91, %cond70.i.i235.i
   %tobool.not.i.i241.i = icmp eq ptr %call24.i.i373.i, null
   %data.i.i.i242.i = getelementptr inbounds nuw i8, ptr %call24.i.i373.i, i64 24
   br i1 %cmp187.not367.i.i.i, label %for.cond178.cleanup215_crit_edge.i.i303.i, label %for.cond184.preheader.i.i243.i
diff --git a/bench/miniaudio/optimized/unity.ll b/bench/miniaudio/optimized/unity.ll
index cc4c7036419..6d2a99b5cfd 100644
--- a/bench/miniaudio/optimized/unity.ll
+++ b/bench/miniaudio/optimized/unity.ll
@@ -13764,8 +13764,8 @@ define void @ma_clip_samples_u8(ptr noundef writeonly captures(none) %0, ptr nou
   %.06 = phi i64 [ %11, %.lr.ph ], [ 0, %3 ]
   %4 = getelementptr inbounds nuw i16, ptr %1, i64 %.06
   %5 = load i16, ptr %4, align 2, !tbaa !334
-  %6 = tail call i16 @llvm.smin.i16(i16 %5, i16 127)
-  %7 = tail call i16 @llvm.smax.i16(i16 %6, i16 -128)
+  %6 = tail call i16 @llvm.smax.i16(i16 %5, i16 -128)
+  %7 = tail call i16 @llvm.smin.i16(i16 %6, i16 127)
   %8 = trunc nsw i16 %7 to i8
   %9 = xor i8 %8, -128
   %10 = getelementptr inbounds nuw i8, ptr %0, i64 %.06
@@ -13787,8 +13787,8 @@ define void @ma_clip_samples_s16(ptr noundef writeonly captures(none) %0, ptr no
   %.06 = phi i64 [ %10, %.lr.ph ], [ 0, %3 ]
   %4 = getelementptr inbounds nuw i32, ptr %1, i64 %.06
   %5 = load i32, ptr %4, align 4, !tbaa !3
-  %6 = tail call i32 @llvm.smin.i32(i32 %5, i32 32767)
-  %7 = tail call i32 @llvm.smax.i32(i32 %6, i32 -32768)
+  %6 = tail call i32 @llvm.smax.i32(i32 %5, i32 -32768)
+  %7 = tail call i32 @llvm.smin.i32(i32 %6, i32 32767)
   %8 = trunc nsw i32 %7 to i16
   %9 = getelementptr inbounds nuw i16, ptr %0, i64 %.06
   store i16 %8, ptr %9, align 2, !tbaa !334
@@ -13809,8 +13809,8 @@ define void @ma_clip_samples_s24(ptr noundef writeonly captures(none) %0, ptr no
   %.013 = phi i64 [ %17, %.lr.ph ], [ 0, %3 ]
   %4 = getelementptr inbounds nuw i64, ptr %1, i64 %.013
   %5 = load i64, ptr %4, align 8, !tbaa !63
-  %6 = tail call i64 @llvm.smin.i64(i64 %5, i64 8388607)
-  %7 = tail call i64 @llvm.smax.i64(i64 %6, i64 -8388608)
+  %6 = tail call i64 @llvm.smax.i64(i64 %5, i64 -8388608)
+  %7 = tail call i64 @llvm.smin.i64(i64 %6, i64 8388607)
   %8 = trunc i64 %7 to i8
   %9 = mul i64 %.013, 3
   %10 = getelementptr inbounds nuw i8, ptr %0, i64 %9
@@ -13840,8 +13840,8 @@ define void @ma_clip_samples_s32(ptr noundef writeonly captures(none) %0, ptr no
   %.06 = phi i64 [ %10, %.lr.ph ], [ 0, %3 ]
   %4 = getelementptr inbounds nuw i64, ptr %1, i64 %.06
   %5 = load i64, ptr %4, align 8, !tbaa !63
-  %6 = tail call i64 @llvm.smin.i64(i64 %5, i64 2147483647)
-  %7 = tail call i64 @llvm.smax.i64(i64 %6, i64 -2147483648)
+  %6 = tail call i64 @llvm.smax.i64(i64 %5, i64 -2147483648)
+  %7 = tail call i64 @llvm.smin.i64(i64 %6, i64 2147483647)
   %8 = trunc nsw i64 %7 to i32
   %9 = getelementptr inbounds nuw i32, ptr %0, i64 %.06
   store i32 %8, ptr %9, align 4, !tbaa !3
@@ -13896,8 +13896,8 @@ define void @ma_clip_pcm_frames(ptr noundef writeonly captures(none) %0, ptr nou
   %.06.i = phi i64 [ %16, %.lr.ph.i ], [ 0, %8 ]
   %9 = getelementptr inbounds nuw i16, ptr %1, i64 %.06.i
   %10 = load i16, ptr %9, align 2, !tbaa !334
-  %11 = tail call i16 @llvm.smin.i16(i16 %10, i16 127)
-  %12 = tail call i16 @llvm.smax.i16(i16 %11, i16 -128)
+  %11 = tail call i16 @llvm.smax.i16(i16 %10, i16 -128)
+  %12 = tail call i16 @llvm.smin.i16(i16 %11, i16 127)
   %13 = trunc nsw i16 %12 to i8
   %14 = xor i8 %13, -128
   %15 = getelementptr inbounds nuw i8, ptr %0, i64 %.06.i
@@ -13914,8 +13914,8 @@ define void @ma_clip_pcm_frames(ptr noundef writeonly captures(none) %0, ptr nou
   %.06.i19 = phi i64 [ %24, %.lr.ph.i18 ], [ 0, %17 ]
   %18 = getelementptr inbounds nuw i32, ptr %1, i64 %.06.i19
   %19 = load i32, ptr %18, align 4, !tbaa !3
-  %20 = tail call i32 @llvm.smin.i32(i32 %19, i32 32767)
-  %21 = tail call i32 @llvm.smax.i32(i32 %20, i32 -32768)
+  %20 = tail call i32 @llvm.smax.i32(i32 %19, i32 -32768)
+  %21 = tail call i32 @llvm.smin.i32(i32 %20, i32 32767)
   %22 = trunc nsw i32 %21 to i16
   %23 = getelementptr inbounds nuw i16, ptr %0, i64 %.06.i19
   store i16 %22, ptr %23, align 2, !tbaa !334
@@ -13931,8 +13931,8 @@ define void @ma_clip_pcm_frames(ptr noundef writeonly captures(none) %0, ptr nou
   %.013.i = phi i64 [ %39, %.lr.ph.i22 ], [ 0, %25 ]
   %26 = getelementptr inbounds nuw i64, ptr %1, i64 %.013.i
   %27 = load i64, ptr %26, align 8, !tbaa !63
-  %28 = tail call i64 @llvm.smin.i64(i64 %27, i64 8388607)
-  %29 = tail call i64 @llvm.smax.i64(i64 %28, i64 -8388608)
+  %28 = tail call i64 @llvm.smax.i64(i64 %27, i64 -8388608)
+  %29 = tail call i64 @llvm.smin.i64(i64 %28, i64 8388607)
   %30 = trunc i64 %29 to i8
   %31 = mul i64 %.013.i, 3
   %32 = getelementptr inbounds nuw i8, ptr %0, i64 %31
@@ -13957,8 +13957,8 @@ define void @ma_clip_pcm_frames(ptr noundef writeonly captures(none) %0, ptr nou
   %.06.i26 = phi i64 [ %47, %.lr.ph.i25 ], [ 0, %40 ]
   %41 = getelementptr inbounds nuw i64, ptr %1, i64 %.06.i26
   %42 = load i64, ptr %41, align 8, !tbaa !63
-  %43 = tail call i64 @llvm.smin.i64(i64 %42, i64 2147483647)
-  %44 = tail call i64 @llvm.smax.i64(i64 %43, i64 -2147483648)
+  %43 = tail call i64 @llvm.smax.i64(i64 %42, i64 -2147483648)
+  %44 = tail call i64 @llvm.smin.i64(i64 %43, i64 2147483647)
   %45 = trunc nsw i64 %44 to i32
   %46 = getelementptr inbounds nuw i32, ptr %0, i64 %.06.i26
   store i32 %45, ptr %46, align 4, !tbaa !3
@@ -14877,8 +14877,8 @@ define void @ma_copy_and_apply_volume_and_clip_samples_u8(ptr noundef writeonly
   %12 = sext i16 %11 to i32
   %13 = mul i32 %8, %12
   %14 = ashr i32 %13, 16
-  %15 = tail call i32 @llvm.smin.i32(i32 %14, i32 127)
-  %16 = tail call i32 @llvm.smax.i32(i32 %15, i32 -128)
+  %15 = tail call i32 @llvm.smax.i32(i32 %14, i32 -128)
+  %16 = tail call i32 @llvm.smin.i32(i32 %15, i32 127)
   %17 = trunc nsw i32 %16 to i8
   %18 = xor i8 %17, -128
   %19 = getelementptr inbounds nuw i8, ptr %0, i64 %.08
@@ -14908,8 +14908,8 @@ define void @ma_copy_and_apply_volume_and_clip_samples_s16(ptr noundef writeonly
   %10 = load i32, ptr %9, align 4, !tbaa !3
   %11 = mul nsw i32 %10, %7
   %12 = ashr i32 %11, 8
-  %13 = tail call i32 @llvm.smin.i32(i32 %12, i32 32767)
-  %14 = tail call i32 @llvm.smax.i32(i32 %13, i32 -32768)
+  %13 = tail call i32 @llvm.smax.i32(i32 %12, i32 -32768)
+  %14 = tail call i32 @llvm.smin.i32(i32 %13, i32 32767)
   %15 = trunc nsw i32 %14 to i16
   %16 = getelementptr inbounds nuw i16, ptr %0, i64 %.08
   store i16 %15, ptr %16, align 2, !tbaa !334
@@ -14938,8 +14938,8 @@ define void @ma_copy_and_apply_volume_and_clip_samples_s24(ptr noundef writeonly
   %10 = load i64, ptr %9, align 8, !tbaa !63
   %11 = mul nsw i64 %10, %7
   %12 = ashr i64 %11, 8
-  %13 = tail call i64 @llvm.smin.i64(i64 %12, i64 8388607)
-  %14 = tail call i64 @llvm.smax.i64(i64 %13, i64 -8388608)
+  %13 = tail call i64 @llvm.smax.i64(i64 %12, i64 -8388608)
+  %14 = tail call i64 @llvm.smin.i64(i64 %13, i64 8388607)
   %15 = trunc i64 %14 to i8
   %16 = mul i64 %.015, 3
   %17 = getelementptr inbounds nuw i8, ptr %0, i64 %16
@@ -14977,8 +14977,8 @@ define void @ma_copy_and_apply_volume_and_clip_samples_s32(ptr noundef writeonly
   %10 = load i64, ptr %9, align 8, !tbaa !63
   %11 = mul nsw i64 %10, %7
   %12 = ashr i64 %11, 8
-  %13 = tail call i64 @llvm.smin.i64(i64 %12, i64 2147483647)
-  %14 = tail call i64 @llvm.smax.i64(i64 %13, i64 -2147483648)
+  %13 = tail call i64 @llvm.smax.i64(i64 %12, i64 -2147483648)
+  %14 = tail call i64 @llvm.smin.i64(i64 %13, i64 2147483647)
   %15 = trunc nsw i64 %14 to i32
   %16 = getelementptr inbounds nuw i32, ptr %0, i64 %.08
   store i32 %15, ptr %16, align 4, !tbaa !3
@@ -15097,8 +15097,8 @@ ma_zero_memory_default.exit.i.i:                  ; preds = %23, %.lr.ph.i
   %37 = sext i16 %36 to i32
   %38 = mul i32 %33, %37
   %39 = ashr i32 %38, 16
-  %40 = tail call i32 @llvm.smin.i32(i32 %39, i32 127)
-  %41 = tail call i32 @llvm.smax.i32(i32 %40, i32 -128)
+  %40 = tail call i32 @llvm.smax.i32(i32 %39, i32 -128)
+  %41 = tail call i32 @llvm.smin.i32(i32 %40, i32 127)
   %42 = trunc nsw i32 %41 to i8
   %43 = xor i8 %42, -128
   %44 = getelementptr inbounds nuw i8, ptr %0, i64 %.08.i
@@ -15123,8 +15123,8 @@ ma_zero_memory_default.exit.i.i:                  ; preds = %23, %.lr.ph.i
   %52 = load i32, ptr %51, align 4, !tbaa !3
   %53 = mul nsw i32 %52, %49
   %54 = ashr i32 %53, 8
-  %55 = tail call i32 @llvm.smin.i32(i32 %54, i32 32767)
-  %56 = tail call i32 @llvm.smax.i32(i32 %55, i32 -32768)
+  %55 = tail call i32 @llvm.smax.i32(i32 %54, i32 -32768)
+  %56 = tail call i32 @llvm.smin.i32(i32 %55, i32 32767)
   %57 = trunc nsw i32 %56 to i16
   %58 = getelementptr inbounds nuw i16, ptr %0, i64 %.08.i37
   store i16 %57, ptr %58, align 2, !tbaa !334
@@ -15148,8 +15148,8 @@ ma_zero_memory_default.exit.i.i:                  ; preds = %23, %.lr.ph.i
   %66 = load i64, ptr %65, align 8, !tbaa !63
   %67 = mul nsw i64 %66, %63
   %68 = ashr i64 %67, 8
-  %69 = tail call i64 @llvm.smin.i64(i64 %68, i64 8388607)
-  %70 = tail call i64 @llvm.smax.i64(i64 %69, i64 -8388608)
+  %69 = tail call i64 @llvm.smax.i64(i64 %68, i64 -8388608)
+  %70 = tail call i64 @llvm.smin.i64(i64 %69, i64 8388607)
   %71 = trunc i64 %70 to i8
   %72 = mul i64 %.015.i, 3
   %73 = getelementptr inbounds nuw i8, ptr %0, i64 %72
@@ -15182,8 +15182,8 @@ ma_zero_memory_default.exit.i.i:                  ; preds = %23, %.lr.ph.i
   %87 = load i64, ptr %86, align 8, !tbaa !63
   %88 = mul nsw i64 %87, %84
   %89 = ashr i64 %88, 8
-  %90 = tail call i64 @llvm.smin.i64(i64 %89, i64 2147483647)
-  %91 = tail call i64 @llvm.smax.i64(i64 %90, i64 -2147483648)
+  %90 = tail call i64 @llvm.smax.i64(i64 %89, i64 -2147483648)
+  %91 = tail call i64 @llvm.smin.i64(i64 %90, i64 2147483647)
   %92 = trunc nsw i64 %91 to i32
   %93 = getelementptr inbounds nuw i32, ptr %0, i64 %.08.i44
   store i32 %92, ptr %93, align 4, !tbaa !3
@@ -18544,8 +18544,8 @@ ma_biquad_process_pcm_frame_f32__direct_form_2_transposed.exit: ; preds = %34
   %85 = mul nsw i32 %64, %77
   %86 = mul nsw i32 %80, %66
   %87 = sub nsw i32 %85, %86
-  %88 = tail call i32 @llvm.smin.i32(i32 %80, i32 32767)
-  %89 = tail call i32 @llvm.smax.i32(i32 %88, i32 -32768)
+  %88 = tail call i32 @llvm.smax.i32(i32 %80, i32 -32768)
+  %89 = tail call i32 @llvm.smin.i32(i32 %88, i32 32767)
   %90 = trunc nsw i32 %89 to i16
   %91 = getelementptr inbounds nuw i16, ptr %.03246, i64 %indvars.iv
   store i16 %90, ptr %91, align 2, !tbaa !334
@@ -21086,8 +21086,8 @@ ma_lpf1_process_pcm_frame_s16.exit.i:             ; preds = %188
   %237 = mul nsw i32 %212, %229
   %238 = mul nsw i32 %232, %216
   %239 = sub nsw i32 %237, %238
-  %240 = tail call i32 @llvm.smin.i32(i32 %232, i32 32767)
-  %241 = tail call i32 @llvm.smax.i32(i32 %240, i32 -32768)
+  %240 = tail call i32 @llvm.smax.i32(i32 %232, i32 -32768)
+  %241 = tail call i32 @llvm.smin.i32(i32 %240, i32 32767)
   %242 = trunc nsw i32 %241 to i16
   store i16 %242, ptr %227, align 2, !tbaa !334
   store i32 %236, ptr %222, align 4, !tbaa !7
@@ -23494,8 +23494,8 @@ ma_hpf1_process_pcm_frame_s16.exit:               ; preds = %189
   %237 = mul nsw i32 %212, %229
   %238 = mul nsw i32 %232, %216
   %239 = sub nsw i32 %237, %238
-  %240 = tail call i32 @llvm.smin.i32(i32 %232, i32 32767)
-  %241 = tail call i32 @llvm.smax.i32(i32 %240, i32 -32768)
+  %240 = tail call i32 @llvm.smax.i32(i32 %232, i32 -32768)
+  %241 = tail call i32 @llvm.smin.i32(i32 %240, i32 32767)
   %242 = trunc nsw i32 %241 to i16
   store i16 %242, ptr %227, align 2, !tbaa !334
   store i32 %236, ptr %222, align 4, !tbaa !7
@@ -24868,8 +24868,8 @@ ma_biquad_process_pcm_frame_f32.exit:             ; preds = %52
   %123 = mul nsw i32 %98, %115
   %124 = mul nsw i32 %118, %102
   %125 = sub nsw i32 %123, %124
-  %126 = tail call i32 @llvm.smin.i32(i32 %118, i32 32767)
-  %127 = tail call i32 @llvm.smax.i32(i32 %126, i32 -32768)
+  %126 = tail call i32 @llvm.smax.i32(i32 %118, i32 -32768)
+  %127 = tail call i32 @llvm.smin.i32(i32 %126, i32 32767)
   %128 = trunc nsw i32 %127 to i16
   store i16 %128, ptr %113, align 2, !tbaa !334
   store i32 %122, ptr %108, align 4, !tbaa !7
@@ -34803,8 +34803,8 @@ ma_lpf1_process_pcm_frame_s16.exit.i.i.i:         ; preds = %66
   %115 = mul nsw i32 %90, %107
   %116 = mul nsw i32 %110, %94
   %117 = sub nsw i32 %115, %116
-  %118 = tail call i32 @llvm.smin.i32(i32 %110, i32 32767)
-  %119 = tail call i32 @llvm.smax.i32(i32 %118, i32 -32768)
+  %118 = tail call i32 @llvm.smax.i32(i32 %110, i32 -32768)
+  %119 = tail call i32 @llvm.smin.i32(i32 %118, i32 32767)
   %120 = trunc nsw i32 %119 to i16
   store i16 %120, ptr %105, align 2, !tbaa !334
   store i32 %114, ptr %100, align 4, !tbaa !7
@@ -35143,8 +35143,8 @@ ma_lpf1_process_pcm_frame_s16.exit.i.i22.i:       ; preds = %239
   %288 = mul nsw i32 %263, %280
   %289 = mul nsw i32 %283, %267
   %290 = sub nsw i32 %288, %289
-  %291 = tail call i32 @llvm.smin.i32(i32 %283, i32 32767)
-  %292 = tail call i32 @llvm.smax.i32(i32 %291, i32 -32768)
+  %291 = tail call i32 @llvm.smax.i32(i32 %283, i32 -32768)
+  %292 = tail call i32 @llvm.smin.i32(i32 %291, i32 32767)
   %293 = trunc nsw i32 %292 to i16
   store i16 %293, ptr %278, align 2, !tbaa !334
   store i32 %287, ptr %273, align 4, !tbaa !7
@@ -38945,8 +38945,8 @@ ma_zero_memory_default.exit.i:                    ; preds = %19, %.lr.ph
   %.094.lcssa.i = phi i32 [ 0, %.preheader.i ], [ %80, %75 ]
   %81 = load i32, ptr %70, align 8, !tbaa !942
   %82 = udiv i32 %.094.lcssa.i, %81
-  %83 = tail call i32 @llvm.smin.i32(i32 %82, i32 127)
-  %84 = tail call i32 @llvm.smax.i32(i32 %83, i32 -128)
+  %83 = tail call i32 @llvm.smax.i32(i32 %82, i32 -128)
+  %84 = tail call i32 @llvm.smin.i32(i32 %83, i32 127)
   %85 = trunc nsw i32 %84 to i8
   %86 = xor i8 %85, -128
   %87 = getelementptr inbounds nuw i8, ptr %1, i64 %.087149.i
@@ -39475,8 +39475,8 @@ ma_zero_memory_64.exit.i:                         ; preds = %ma_zero_memory_64.e
   %298 = mul nsw i32 %297, %292
   %299 = ashr i32 %298, 12
   %300 = add nsw i32 %299, %290
-  %301 = tail call i32 @llvm.smin.i32(i32 %300, i32 32767)
-  %302 = tail call i32 @llvm.smax.i32(i32 %301, i32 -32768)
+  %301 = tail call i32 @llvm.smax.i32(i32 %300, i32 -32768)
+  %302 = tail call i32 @llvm.smin.i32(i32 %301, i32 32767)
   %303 = trunc nsw i32 %302 to i16
   store i16 %303, ptr %288, align 2, !tbaa !334
   %indvars.iv.next307.i = add nuw nsw i64 %indvars.iv306.i, 1
@@ -39737,8 +39737,8 @@ ma_zero_memory_64.exit.i:                         ; preds = %ma_zero_memory_64.e
   %453 = mul nsw i64 %452, %446
   %454 = ashr i64 %453, 12
   %455 = add nsw i64 %454, %439
-  %456 = tail call i64 @llvm.smin.i64(i64 %455, i64 2147483647)
-  %457 = tail call i64 @llvm.smax.i64(i64 %456, i64 -2147483648)
+  %456 = tail call i64 @llvm.smax.i64(i64 %455, i64 -2147483648)
+  %457 = tail call i64 @llvm.smin.i64(i64 %456, i64 2147483647)
   %458 = trunc nsw i64 %457 to i32
   store i32 %458, ptr %437, align 4, !tbaa !3
   %indvars.iv.next295.i = add nuw nsw i64 %indvars.iv294.i, 1
@@ -94152,8 +94152,8 @@ define internal fastcc i64 @ma_dr_wav_read_pcm_frames_s16__msadpcm(ptr noundef n
   %146 = load i32, ptr %21, align 8, !tbaa !3
   %147 = mul nsw i32 %146, %.0163
   %148 = add nsw i32 %145, %147
-  %149 = call i32 @llvm.smin.i32(i32 %148, i32 32767)
-  %150 = call i32 @llvm.smax.i32(i32 %149, i32 -32768)
+  %149 = call i32 @llvm.smax.i32(i32 %148, i32 -32768)
+  %150 = call i32 @llvm.smin.i32(i32 %149, i32 32767)
   %151 = zext nneg i8 %125 to i64
   %152 = getelementptr inbounds nuw [16 x i32], ptr @ma_dr_wav_read_pcm_frames_s16__msadpcm.adaptationTable, i64 0, i64 %151
   %153 = load i32, ptr %152, align 4, !tbaa !3
@@ -94169,8 +94169,8 @@ define internal fastcc i64 @ma_dr_wav_read_pcm_frames_s16__msadpcm(ptr noundef n
   %160 = ashr i32 %159, 8
   %161 = mul nsw i32 %spec.select, %.0
   %162 = add nsw i32 %160, %161
-  %163 = call i32 @llvm.smin.i32(i32 %162, i32 32767)
-  %164 = call i32 @llvm.smax.i32(i32 %163, i32 -32768)
+  %163 = call i32 @llvm.smax.i32(i32 %162, i32 -32768)
+  %164 = call i32 @llvm.smin.i32(i32 %163, i32 32767)
   %165 = and i8 %124, 15
   %166 = zext nneg i8 %165 to i64
   %167 = getelementptr inbounds nuw [16 x i32], ptr @ma_dr_wav_read_pcm_frames_s16__msadpcm.adaptationTable, i64 0, i64 %166
@@ -94202,8 +94202,8 @@ define internal fastcc i64 @ma_dr_wav_read_pcm_frames_s16__msadpcm(ptr noundef n
   %184 = load i32, ptr %23, align 4, !tbaa !3
   %185 = mul nsw i32 %184, %.0
   %186 = add nsw i32 %183, %185
-  %187 = call i32 @llvm.smin.i32(i32 %186, i32 32767)
-  %188 = call i32 @llvm.smax.i32(i32 %187, i32 -32768)
+  %187 = call i32 @llvm.smax.i32(i32 %186, i32 -32768)
+  %188 = call i32 @llvm.smin.i32(i32 %187, i32 32767)
   %189 = and i8 %124, 15
   %190 = zext nneg i8 %189 to i64
   %191 = getelementptr inbounds nuw [16 x i32], ptr @ma_dr_wav_read_pcm_frames_s16__msadpcm.adaptationTable, i64 0, i64 %190
@@ -94462,7 +94462,7 @@ define internal fastcc i64 @ma_dr_wav_read_pcm_frames_s16__ima(ptr noundef nonnu
   br i1 %.not298, label %.backedge, label %.lr.ph290
 
 .lr.ph290:                                        ; preds = %111, %192
-  %indvars.iv326 = phi i64 [ %indvars.iv.next327, %192 ], [ 0, %111 ]
+  %indvars.iv326 = phi i64 [ %indvars.iv.next327, %195 ], [ 0, %111 ]
   call void @llvm.lifetime.start.p0(i64 4, ptr nonnull %6) #66
   %113 = load ptr, ptr %0, align 8, !tbaa !1141
   %114 = load ptr, ptr %15, align 8, !tbaa !1143
@@ -94518,15 +94518,15 @@ define internal fastcc i64 @ma_dr_wav_read_pcm_frames_s16__ima(ptr noundef nonnu
   %144 = sub nsw i32 0, %.2
   %.3 = select i1 %.not217, i32 %.2, i32 %144
   %145 = add nsw i32 %.3, %133
-  %146 = call i32 @llvm.smin.i32(i32 %145, i32 32767)
-  %147 = call i32 @llvm.smax.i32(i32 %146, i32 -32768)
+  %146 = call i32 @llvm.smax.i32(i32 %145, i32 -32768)
+  %147 = call i32 @llvm.smin.i32(i32 %146, i32 32767)
   store i32 %147, ptr %120, align 4, !tbaa !3
   %148 = zext nneg i8 %127 to i64
   %149 = getelementptr inbounds nuw [16 x i32], ptr @ma_dr_wav_read_pcm_frames_s16__ima.indexTable, i64 0, i64 %148
   %150 = load i32, ptr %149, align 4, !tbaa !3
   %151 = add nsw i32 %150, %129
-  %spec.select = call i32 @llvm.smin.i32(i32 %151, i32 88)
-  %spec.select233 = call i32 @llvm.smax.i32(i32 %spec.select, i32 0)
+  %spec.select = call i32 @llvm.smax.i32(i32 %151, i32 0)
+  %spec.select233 = call i32 @llvm.umin.i32(i32 %spec.select, i32 88)
   store i32 %spec.select233, ptr %119, align 4, !tbaa !3
   %152 = load i32, ptr %13, align 4, !tbaa !1894
   %153 = mul i32 %152, %122
@@ -94562,15 +94562,15 @@ define internal fastcc i64 @ma_dr_wav_read_pcm_frames_s16__ima(ptr noundef nonnu
   %.not221234 = icmp slt i8 %126, 0
   %.7 = select i1 %.not221234, i32 %175, i32 %.6
   %176 = add nsw i32 %.7, %164
-  %177 = call i32 @llvm.smin.i32(i32 %176, i32 32767)
-  %178 = call i32 @llvm.smax.i32(i32 %177, i32 -32768)
+  %177 = call i32 @llvm.smax.i32(i32 %176, i32 -32768)
+  %178 = call i32 @llvm.smin.i32(i32 %177, i32 32767)
   store i32 %178, ptr %120, align 4, !tbaa !3
   %179 = zext nneg i8 %128 to i64
   %180 = getelementptr inbounds nuw [16 x i32], ptr @ma_dr_wav_read_pcm_frames_s16__ima.indexTable, i64 0, i64 %179
   %181 = load i32, ptr %180, align 4, !tbaa !3
   %182 = add nsw i32 %181, %160
-  %spec.select224 = call i32 @llvm.smin.i32(i32 %182, i32 88)
-  %183 = call i32 @llvm.smax.i32(i32 %spec.select224, i32 0)
+  %spec.select224 = call i32 @llvm.smax.i32(i32 %182, i32 0)
+  %183 = call i32 @llvm.umin.i32(i32 %spec.select224, i32 88)
   store i32 %183, ptr %119, align 4, !tbaa !3
   %184 = load i32, ptr %13, align 4, !tbaa !1894
   %185 = mul i32 %184, %122
@@ -94592,8 +94592,8 @@ define internal fastcc i64 @ma_dr_wav_read_pcm_frames_s16__ima(ptr noundef nonnu
   br i1 %193, label %.lr.ph290, label %.backedge, !llvm.loop !1899
 
 .backedge:                                        ; preds = %86, %192, %111, %108
-  %.1189281310 = phi i64 [ %82, %111 ], [ %82, %108 ], [ %82, %192 ], [ %.1189281, %86 ]
-  %.2198280307 = phi i64 [ %83, %111 ], [ %83, %108 ], [ %83, %192 ], [ %.2198280, %86 ]
+  %.1189281310 = phi i64 [ %82, %111 ], [ %82, %108 ], [ %82, %195 ], [ %.1189281, %86 ]
+  %.2198280307 = phi i64 [ %83, %111 ], [ %83, %108 ], [ %83, %195 ], [ %.2198280, %86 ]
   %194 = load i64, ptr %7, align 8, !tbaa !1195
   %195 = load i64, ptr %8, align 8, !tbaa !1194
   %196 = icmp ult i64 %194, %195
@@ -149371,10 +149371,10 @@ declare double @llvm.fabs.f64(double) #74
 declare i64 @llvm.abs.i64(i64, i1 immarg) #74
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i16 @llvm.smin.i16(i16, i16) #74
+declare i16 @llvm.smax.i16(i16, i16) #74
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i16 @llvm.smax.i16(i16, i16) #74
+declare i16 @llvm.smin.i16(i16, i16) #74
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare float @llvm.sqrt.f32(float) #74
diff --git a/bench/nuklear/optimized/unity.ll b/bench/nuklear/optimized/unity.ll
index 6c56d1cb032..25fc9395f98 100644
--- a/bench/nuklear/optimized/unity.ll
+++ b/bench/nuklear/optimized/unity.ll
@@ -1168,14 +1168,14 @@ define i32 @nk_rgb_factor(i32 %0, float noundef %1) local_unnamed_addr #0 {
 
 ; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none) uwtable
 define i32 @nk_rgba(i32 noundef %0, i32 noundef %1, i32 noundef %2, i32 noundef %3) local_unnamed_addr #0 {
-  %5 = tail call i32 @llvm.smin.i32(i32 %0, i32 255)
-  %6 = tail call i32 @llvm.smax.i32(i32 %5, i32 0)
-  %7 = tail call i32 @llvm.smin.i32(i32 %1, i32 255)
-  %8 = tail call i32 @llvm.smax.i32(i32 %7, i32 0)
-  %9 = tail call i32 @llvm.smin.i32(i32 %2, i32 255)
-  %10 = tail call i32 @llvm.smax.i32(i32 %9, i32 0)
-  %11 = tail call i32 @llvm.smin.i32(i32 %3, i32 255)
-  %12 = tail call i32 @llvm.smax.i32(i32 %11, i32 0)
+  %5 = tail call i32 @llvm.smax.i32(i32 %0, i32 0)
+  %6 = tail call i32 @llvm.umin.i32(i32 %5, i32 255)
+  %7 = tail call i32 @llvm.smax.i32(i32 %1, i32 0)
+  %8 = tail call i32 @llvm.umin.i32(i32 %7, i32 255)
+  %9 = tail call i32 @llvm.smax.i32(i32 %2, i32 0)
+  %10 = tail call i32 @llvm.umin.i32(i32 %9, i32 255)
+  %11 = tail call i32 @llvm.smax.i32(i32 %3, i32 0)
+  %12 = tail call i32 @llvm.umin.i32(i32 %11, i32 255)
   %.sroa.4.0.insert.shift = shl nuw i32 %12, 24
   %.sroa.3.0.insert.shift = shl nuw nsw i32 %10, 16
   %.sroa.3.0.insert.insert = or disjoint i32 %.sroa.4.0.insert.shift, %.sroa.3.0.insert.shift
@@ -1634,14 +1634,14 @@ define i32 @nk_rgba_iv(ptr noundef readonly captures(none) %0) local_unnamed_add
   %6 = load i32, ptr %5, align 4, !tbaa !7
   %7 = getelementptr inbounds nuw i8, ptr %0, i64 12
   %8 = load i32, ptr %7, align 4, !tbaa !7
-  %9 = tail call i32 @llvm.smin.i32(i32 %2, i32 255)
-  %10 = tail call i32 @llvm.smax.i32(i32 %9, i32 0)
-  %11 = tail call i32 @llvm.smin.i32(i32 %4, i32 255)
-  %12 = tail call i32 @llvm.smax.i32(i32 %11, i32 0)
-  %13 = tail call i32 @llvm.smin.i32(i32 %6, i32 255)
-  %14 = tail call i32 @llvm.smax.i32(i32 %13, i32 0)
-  %15 = tail call i32 @llvm.smin.i32(i32 %8, i32 255)
-  %16 = tail call i32 @llvm.smax.i32(i32 %15, i32 0)
+  %9 = tail call i32 @llvm.smax.i32(i32 %2, i32 0)
+  %10 = tail call i32 @llvm.umin.i32(i32 %9, i32 255)
+  %11 = tail call i32 @llvm.smax.i32(i32 %4, i32 0)
+  %12 = tail call i32 @llvm.umin.i32(i32 %11, i32 255)
+  %13 = tail call i32 @llvm.smax.i32(i32 %6, i32 0)
+  %14 = tail call i32 @llvm.umin.i32(i32 %13, i32 255)
+  %15 = tail call i32 @llvm.smax.i32(i32 %8, i32 0)
+  %16 = tail call i32 @llvm.umin.i32(i32 %15, i32 255)
   %.sroa.4.0.insert.shift.i = shl nuw i32 %16, 24
   %.sroa.3.0.insert.shift.i = shl nuw nsw i32 %14, 16
   %.sroa.3.0.insert.insert.i = or disjoint i32 %.sroa.4.0.insert.shift.i, %.sroa.3.0.insert.shift.i
@@ -1659,12 +1659,12 @@ define i32 @nk_rgba_bv(ptr noundef readonly captures(none) %0) local_unnamed_add
 
 ; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none) uwtable
 define range(i32 -16777216, 0) i32 @nk_rgb(i32 noundef %0, i32 noundef %1, i32 noundef %2) local_unnamed_addr #0 {
-  %4 = tail call i32 @llvm.smin.i32(i32 %0, i32 255)
-  %5 = tail call i32 @llvm.smax.i32(i32 %4, i32 0)
-  %6 = tail call i32 @llvm.smin.i32(i32 %1, i32 255)
-  %7 = tail call i32 @llvm.smax.i32(i32 %6, i32 0)
-  %8 = tail call i32 @llvm.smin.i32(i32 %2, i32 255)
-  %9 = tail call i32 @llvm.smax.i32(i32 %8, i32 0)
+  %4 = tail call i32 @llvm.smax.i32(i32 %0, i32 0)
+  %5 = tail call i32 @llvm.umin.i32(i32 %4, i32 255)
+  %6 = tail call i32 @llvm.smax.i32(i32 %1, i32 0)
+  %7 = tail call i32 @llvm.umin.i32(i32 %6, i32 255)
+  %8 = tail call i32 @llvm.smax.i32(i32 %2, i32 0)
+  %9 = tail call i32 @llvm.umin.i32(i32 %8, i32 255)
   %.sroa.3.0.insert.shift = shl nuw nsw i32 %9, 16
   %.sroa.2.0.insert.shift = shl nuw nsw i32 %7, 8
   %.sroa.3.0.insert.insert = or disjoint i32 %.sroa.3.0.insert.shift, %.sroa.2.0.insert.shift
@@ -1680,12 +1680,12 @@ define range(i32 -16777216, 0) i32 @nk_rgb_iv(ptr noundef readonly captures(none
   %4 = load i32, ptr %3, align 4, !tbaa !7
   %5 = getelementptr inbounds nuw i8, ptr %0, i64 8
   %6 = load i32, ptr %5, align 4, !tbaa !7
-  %7 = tail call i32 @llvm.smin.i32(i32 %2, i32 255)
-  %8 = tail call i32 @llvm.smax.i32(i32 %7, i32 0)
-  %9 = tail call i32 @llvm.smin.i32(i32 %4, i32 255)
-  %10 = tail call i32 @llvm.smax.i32(i32 %9, i32 0)
-  %11 = tail call i32 @llvm.smin.i32(i32 %6, i32 255)
-  %12 = tail call i32 @llvm.smax.i32(i32 %11, i32 0)
+  %7 = tail call i32 @llvm.smax.i32(i32 %2, i32 0)
+  %8 = tail call i32 @llvm.umin.i32(i32 %7, i32 255)
+  %9 = tail call i32 @llvm.smax.i32(i32 %4, i32 0)
+  %10 = tail call i32 @llvm.umin.i32(i32 %9, i32 255)
+  %11 = tail call i32 @llvm.smax.i32(i32 %6, i32 0)
+  %12 = tail call i32 @llvm.umin.i32(i32 %11, i32 255)
   %.sroa.3.0.insert.shift.i = shl nuw nsw i32 %12, 16
   %.sroa.2.0.insert.shift.i = shl nuw nsw i32 %10, 8
   %.sroa.3.0.insert.insert.i = or disjoint i32 %.sroa.3.0.insert.shift.i, %.sroa.2.0.insert.shift.i
@@ -1961,12 +1961,12 @@ nk_rgb_f.exit:                                    ; preds = %2, %6
 
 ; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none) uwtable
 define range(i32 -16777216, 0) i32 @nk_hsv(i32 noundef %0, i32 noundef %1, i32 noundef %2) local_unnamed_addr #2 {
-  %4 = tail call i32 @llvm.smin.i32(i32 %1, i32 255)
-  %5 = tail call i32 @llvm.smax.i32(i32 %4, i32 0)
+  %4 = tail call i32 @llvm.smax.i32(i32 %1, i32 0)
+  %5 = tail call i32 @llvm.umin.i32(i32 %4, i32 255)
   %6 = uitofp nneg i32 %5 to float
   %7 = fdiv float %6, 2.550000e+02
-  %8 = tail call i32 @llvm.smin.i32(i32 %2, i32 255)
-  %9 = tail call i32 @llvm.smax.i32(i32 %8, i32 0)
+  %8 = tail call i32 @llvm.smax.i32(i32 %2, i32 0)
+  %9 = tail call i32 @llvm.umin.i32(i32 %8, i32 255)
   %10 = uitofp nneg i32 %9 to float
   %11 = fdiv float %10, 2.550000e+02
   %12 = fcmp ugt float %7, 0.000000e+00
@@ -1978,8 +1978,8 @@ define range(i32 -16777216, 0) i32 @nk_hsv(i32 noundef %0, i32 noundef %1, i32 n
   br label %nk_hsva.exit
 
 14:                                               ; preds = %3
-  %15 = tail call i32 @llvm.smin.i32(i32 %0, i32 255)
-  %16 = tail call i32 @llvm.smax.i32(i32 %15, i32 0)
+  %15 = tail call i32 @llvm.smax.i32(i32 %0, i32 0)
+  %16 = tail call i32 @llvm.umin.i32(i32 %15, i32 255)
   %17 = uitofp nneg i32 %16 to float
   %18 = fdiv float %17, 2.550000e+02
   %19 = fdiv float %18, 0x3FC5555560000000
@@ -2063,12 +2063,12 @@ nk_hsva.exit:                                     ; preds = %13, %36
 
 ; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none) uwtable
 define i32 @nk_hsva(i32 noundef %0, i32 noundef %1, i32 noundef %2, i32 noundef %3) local_unnamed_addr #2 {
-  %5 = tail call i32 @llvm.smin.i32(i32 %1, i32 255)
-  %6 = tail call i32 @llvm.smax.i32(i32 %5, i32 0)
+  %5 = tail call i32 @llvm.smax.i32(i32 %1, i32 0)
+  %6 = tail call i32 @llvm.umin.i32(i32 %5, i32 255)
   %7 = uitofp nneg i32 %6 to float
   %8 = fdiv float %7, 2.550000e+02
-  %9 = tail call i32 @llvm.smin.i32(i32 %2, i32 255)
-  %10 = tail call i32 @llvm.smax.i32(i32 %9, i32 0)
+  %9 = tail call i32 @llvm.smax.i32(i32 %2, i32 0)
+  %10 = tail call i32 @llvm.umin.i32(i32 %9, i32 255)
   %11 = uitofp nneg i32 %10 to float
   %12 = fdiv float %11, 2.550000e+02
   %13 = fcmp ugt float %8, 0.000000e+00
@@ -2080,8 +2080,8 @@ define i32 @nk_hsva(i32 noundef %0, i32 noundef %1, i32 noundef %2, i32 noundef
   br label %nk_hsva_f.exit
 
 15:                                               ; preds = %4
-  %16 = tail call i32 @llvm.smin.i32(i32 %0, i32 255)
-  %17 = tail call i32 @llvm.smax.i32(i32 %16, i32 0)
+  %16 = tail call i32 @llvm.smax.i32(i32 %0, i32 0)
+  %17 = tail call i32 @llvm.umin.i32(i32 %16, i32 255)
   %18 = uitofp nneg i32 %17 to float
   %19 = fdiv float %18, 2.550000e+02
   %20 = fdiv float %19, 0x3FC5555560000000
@@ -2131,8 +2131,8 @@ define i32 @nk_hsva(i32 noundef %0, i32 noundef %1, i32 noundef %2, i32 noundef
 nk_hsva_f.exit:                                   ; preds = %14, %37
   %.sroa.16.8.vec.insert73.sink.i.i = phi <2 x float> [ %.sroa.16.8.vec.insert73.i.i, %37 ], [ %.sroa.0.0.vec.insert.i.i, %14 ]
   %.sroa.0.0.i.i = phi <2 x float> [ %.sroa.0.4.vec.insert60.i.i, %37 ], [ %.sroa.0.4.vec.insert.i.i, %14 ]
-  %38 = tail call i32 @llvm.smin.i32(i32 %3, i32 255)
-  %39 = tail call i32 @llvm.smax.i32(i32 %38, i32 0)
+  %38 = tail call i32 @llvm.smax.i32(i32 %3, i32 0)
+  %39 = tail call i32 @llvm.umin.i32(i32 %38, i32 255)
   %40 = uitofp nneg i32 %39 to float
   %41 = fdiv float %40, 2.550000e+02
   %.sroa.0.0.vec.extract.i = extractelement <2 x float> %.sroa.0.0.i.i, i64 0
@@ -2181,12 +2181,12 @@ define range(i32 -16777216, 0) i32 @nk_hsv_iv(ptr noundef readonly captures(none
   %3 = load i32, ptr %2, align 4, !tbaa !7
   %4 = getelementptr inbounds nuw i8, ptr %0, i64 8
   %5 = load i32, ptr %4, align 4, !tbaa !7
-  %6 = tail call i32 @llvm.smin.i32(i32 %3, i32 255)
-  %7 = tail call i32 @llvm.smax.i32(i32 %6, i32 0)
+  %6 = tail call i32 @llvm.smax.i32(i32 %3, i32 0)
+  %7 = tail call i32 @llvm.umin.i32(i32 %6, i32 255)
   %8 = uitofp nneg i32 %7 to float
   %9 = fdiv float %8, 2.550000e+02
-  %10 = tail call i32 @llvm.smin.i32(i32 %5, i32 255)
-  %11 = tail call i32 @llvm.smax.i32(i32 %10, i32 0)
+  %10 = tail call i32 @llvm.smax.i32(i32 %5, i32 0)
+  %11 = tail call i32 @llvm.umin.i32(i32 %10, i32 255)
   %12 = uitofp nneg i32 %11 to float
   %13 = fdiv float %12, 2.550000e+02
   %14 = fcmp ugt float %9, 0.000000e+00
@@ -2199,8 +2199,8 @@ define range(i32 -16777216, 0) i32 @nk_hsv_iv(ptr noundef readonly captures(none
 
 16:                                               ; preds = %1
   %17 = load i32, ptr %0, align 4, !tbaa !7
-  %18 = tail call i32 @llvm.smin.i32(i32 %17, i32 255)
-  %19 = tail call i32 @llvm.smax.i32(i32 %18, i32 0)
+  %18 = tail call i32 @llvm.smax.i32(i32 %17, i32 0)
+  %19 = tail call i32 @llvm.umin.i32(i32 %18, i32 255)
   %20 = uitofp nneg i32 %19 to float
   %21 = fdiv float %20, 2.550000e+02
   %22 = fdiv float %21, 0x3FC5555560000000
@@ -2674,12 +2674,12 @@ define i32 @nk_hsva_iv(ptr noundef readonly captures(none) %0) local_unnamed_add
   %5 = load i32, ptr %4, align 4, !tbaa !7
   %6 = getelementptr inbounds nuw i8, ptr %0, i64 12
   %7 = load i32, ptr %6, align 4, !tbaa !7
-  %8 = tail call i32 @llvm.smin.i32(i32 %3, i32 255)
-  %9 = tail call i32 @llvm.smax.i32(i32 %8, i32 0)
+  %8 = tail call i32 @llvm.smax.i32(i32 %3, i32 0)
+  %9 = tail call i32 @llvm.umin.i32(i32 %8, i32 255)
   %10 = uitofp nneg i32 %9 to float
   %11 = fdiv float %10, 2.550000e+02
-  %12 = tail call i32 @llvm.smin.i32(i32 %5, i32 255)
-  %13 = tail call i32 @llvm.smax.i32(i32 %12, i32 0)
+  %12 = tail call i32 @llvm.smax.i32(i32 %5, i32 0)
+  %13 = tail call i32 @llvm.umin.i32(i32 %12, i32 255)
   %14 = uitofp nneg i32 %13 to float
   %15 = fdiv float %14, 2.550000e+02
   %16 = fcmp ugt float %11, 0.000000e+00
@@ -2692,8 +2692,8 @@ define i32 @nk_hsva_iv(ptr noundef readonly captures(none) %0) local_unnamed_add
 
 18:                                               ; preds = %1
   %19 = load i32, ptr %0, align 4, !tbaa !7
-  %20 = tail call i32 @llvm.smin.i32(i32 %19, i32 255)
-  %21 = tail call i32 @llvm.smax.i32(i32 %20, i32 0)
+  %20 = tail call i32 @llvm.smax.i32(i32 %19, i32 0)
+  %21 = tail call i32 @llvm.umin.i32(i32 %20, i32 255)
   %22 = uitofp nneg i32 %21 to float
   %23 = fdiv float %22, 2.550000e+02
   %24 = fdiv float %23, 0x3FC5555560000000
@@ -2743,8 +2743,8 @@ define i32 @nk_hsva_iv(ptr noundef readonly captures(none) %0) local_unnamed_add
 nk_hsva.exit:                                     ; preds = %17, %41
   %.sroa.16.8.vec.insert73.sink.i.i.i = phi <2 x float> [ %.sroa.16.8.vec.insert73.i.i.i, %41 ], [ %.sroa.0.0.vec.insert.i.i.i, %17 ]
   %.sroa.0.0.i.i.i = phi <2 x float> [ %.sroa.0.4.vec.insert60.i.i.i, %41 ], [ %.sroa.0.4.vec.insert.i.i.i, %17 ]
-  %42 = tail call i32 @llvm.smin.i32(i32 %7, i32 255)
-  %43 = tail call i32 @llvm.smax.i32(i32 %42, i32 0)
+  %42 = tail call i32 @llvm.smax.i32(i32 %7, i32 0)
+  %43 = tail call i32 @llvm.umin.i32(i32 %42, i32 255)
   %44 = uitofp nneg i32 %43 to float
   %45 = fdiv float %44, 2.550000e+02
   %.sroa.0.0.vec.extract.i.i = extractelement <2 x float> %.sroa.0.0.i.i.i, i64 0
diff --git a/bench/oiio/optimized/sysutil.ll b/bench/oiio/optimized/sysutil.ll
index 115d7ef5991..1a02241374f 100644
--- a/bench/oiio/optimized/sysutil.ll
+++ b/bench/oiio/optimized/sysutil.ll
@@ -1266,12 +1266,12 @@ define void @_ZN11OpenImageIO6v3_1_07Sysutil4Term12ansi_fgcolorB5cxx11Eiii(ptr d
   br i1 %11, label %12, label %65
 
 12:                                               ; preds = %5
-  %.sroa.speculated43 = tail call i32 @llvm.smin.i32(i32 %2, i32 255)
-  %.sroa.speculated22 = tail call i32 @llvm.smax.i32(i32 %.sroa.speculated43, i32 0)
-  %.sroa.speculated35 = tail call i32 @llvm.smin.i32(i32 %3, i32 255)
-  %.sroa.speculated16 = tail call i32 @llvm.smax.i32(i32 %.sroa.speculated35, i32 0)
-  %.sroa.speculated27 = tail call i32 @llvm.smin.i32(i32 %4, i32 255)
-  %.sroa.speculated = tail call i32 @llvm.smax.i32(i32 %.sroa.speculated27, i32 0)
+  %.sroa.speculated43 = tail call i32 @llvm.smax.i32(i32 %2, i32 0)
+  %.sroa.speculated22 = tail call i32 @llvm.umin.i32(i32 %.sroa.speculated43, i32 255)
+  %.sroa.speculated35 = tail call i32 @llvm.smax.i32(i32 %3, i32 0)
+  %.sroa.speculated16 = tail call i32 @llvm.umin.i32(i32 %.sroa.speculated35, i32 255)
+  %.sroa.speculated27 = tail call i32 @llvm.smax.i32(i32 %4, i32 0)
+  %.sroa.speculated = tail call i32 @llvm.umin.i32(i32 %.sroa.speculated27, i32 255)
   call void @llvm.lifetime.start.p0(i64 32, ptr nonnull %7) #32
   call void @llvm.lifetime.start.p0(i64 48, ptr nonnull %6) #32, !noalias !40
   %.sroa.024.0.insert.ext.i.i = zext nneg i32 %.sroa.speculated22 to i64
@@ -1283,7 +1283,7 @@ define void @_ZN11OpenImageIO6v3_1_07Sysutil4Term12ansi_fgcolorB5cxx11Eiii(ptr d
   %14 = getelementptr inbounds nuw i8, ptr %6, i64 32
   store i64 %.sroa.018.0.insert.ext.i.i, ptr %14, align 16, !alias.scope !43, !noalias !40
   invoke void @_ZN3fmt2v87vformatB5cxx11ENS0_17basic_string_viewIcEENS0_17basic_format_argsINS0_20basic_format_contextINS0_8appenderEcEEEE(ptr dead_on_unwind nonnull writable sret(%"class.std::__cxx11::basic_string") align 8 %7, ptr nonnull @.str.86, i64 16, i64 273, ptr nonnull %6)
-          to label %15 unwind label %57
+          to label %21 unwind label %57
 
 15:                                               ; preds = %12
   call void @llvm.lifetime.end.p0(i64 48, ptr nonnull %6) #32, !noalias !40
@@ -1316,7 +1316,7 @@ _ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE11_M_is_localEv.exit.threa
   br i1 %.not22.i, label %_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEaSEOS4_.exit, label %31, !prof !46
 
 31:                                               ; preds = %26
-  switch i64 %29, label %34 [
+  switch i64 %29, label %40 [
     i64 0, label %_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE7_S_copyEPcPKcm.exit.i
     i64 1, label %32
   ]
@@ -1370,7 +1370,7 @@ _ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE11_M_is_localEv.exit24.thr
   br label %_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEaSEOS4_.exit
 
 _ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEaSEOS4_.exit: ; preds = %26, %_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE7_S_copyEPcPKcm.exit.i, %45, %46
-  %48 = phi ptr [ %.pre.i, %_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE7_S_copyEPcPKcm.exit.i ], [ %16, %45 ], [ %47, %46 ], [ %27, %26 ]
+  %48 = phi ptr [ %.pre.i, %_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE7_S_copyEPcPKcm.exit.i ], [ %16, %51 ], [ %47, %52 ], [ %27, %32 ]
   %49 = getelementptr inbounds nuw i8, ptr %7, i64 8
   store i64 0, ptr %49, align 8, !tbaa !19
   store i8 0, ptr %48, align 1, !tbaa !13
@@ -1436,12 +1436,12 @@ define void @_ZN11OpenImageIO6v3_1_07Sysutil4Term12ansi_bgcolorB5cxx11Eiii(ptr d
   br i1 %11, label %12, label %65
 
 12:                                               ; preds = %5
-  %.sroa.speculated43 = tail call i32 @llvm.smin.i32(i32 %2, i32 255)
-  %.sroa.speculated22 = tail call i32 @llvm.smax.i32(i32 %.sroa.speculated43, i32 0)
-  %.sroa.speculated35 = tail call i32 @llvm.smin.i32(i32 %3, i32 255)
-  %.sroa.speculated16 = tail call i32 @llvm.smax.i32(i32 %.sroa.speculated35, i32 0)
-  %.sroa.speculated27 = tail call i32 @llvm.smin.i32(i32 %4, i32 255)
-  %.sroa.speculated = tail call i32 @llvm.smax.i32(i32 %.sroa.speculated27, i32 0)
+  %.sroa.speculated43 = tail call i32 @llvm.smax.i32(i32 %2, i32 0)
+  %.sroa.speculated22 = tail call i32 @llvm.umin.i32(i32 %.sroa.speculated43, i32 255)
+  %.sroa.speculated35 = tail call i32 @llvm.smax.i32(i32 %3, i32 0)
+  %.sroa.speculated16 = tail call i32 @llvm.umin.i32(i32 %.sroa.speculated35, i32 255)
+  %.sroa.speculated27 = tail call i32 @llvm.smax.i32(i32 %4, i32 0)
+  %.sroa.speculated = tail call i32 @llvm.umin.i32(i32 %.sroa.speculated27, i32 255)
   call void @llvm.lifetime.start.p0(i64 32, ptr nonnull %7) #32
   call void @llvm.lifetime.start.p0(i64 48, ptr nonnull %6) #32, !noalias !47
   %.sroa.024.0.insert.ext.i.i = zext nneg i32 %.sroa.speculated22 to i64
@@ -1453,7 +1453,7 @@ define void @_ZN11OpenImageIO6v3_1_07Sysutil4Term12ansi_bgcolorB5cxx11Eiii(ptr d
   %14 = getelementptr inbounds nuw i8, ptr %6, i64 32
   store i64 %.sroa.018.0.insert.ext.i.i, ptr %14, align 16, !alias.scope !50, !noalias !47
   invoke void @_ZN3fmt2v87vformatB5cxx11ENS0_17basic_string_viewIcEENS0_17basic_format_argsINS0_20basic_format_contextINS0_8appenderEcEEEE(ptr dead_on_unwind nonnull writable sret(%"class.std::__cxx11::basic_string") align 8 %7, ptr nonnull @.str.87, i64 16, i64 273, ptr nonnull %6)
-          to label %15 unwind label %57
+          to label %21 unwind label %57
 
 15:                                               ; preds = %12
   call void @llvm.lifetime.end.p0(i64 48, ptr nonnull %6) #32, !noalias !47
@@ -1486,7 +1486,7 @@ _ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE11_M_is_localEv.exit.threa
   br i1 %.not22.i, label %_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEaSEOS4_.exit, label %31, !prof !46
 
 31:                                               ; preds = %26
-  switch i64 %29, label %34 [
+  switch i64 %29, label %40 [
     i64 0, label %_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE7_S_copyEPcPKcm.exit.i
     i64 1, label %32
   ]
@@ -1540,7 +1540,7 @@ _ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE11_M_is_localEv.exit24.thr
   br label %_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEaSEOS4_.exit
 
 _ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEaSEOS4_.exit: ; preds = %26, %_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE7_S_copyEPcPKcm.exit.i, %45, %46
-  %48 = phi ptr [ %.pre.i, %_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE7_S_copyEPcPKcm.exit.i ], [ %16, %45 ], [ %47, %46 ], [ %27, %26 ]
+  %48 = phi ptr [ %.pre.i, %_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE7_S_copyEPcPKcm.exit.i ], [ %16, %51 ], [ %47, %52 ], [ %27, %32 ]
   %49 = getelementptr inbounds nuw i8, ptr %7, i64 8
   store i64 0, ptr %49, align 8, !tbaa !19
   store i8 0, ptr %48, align 1, !tbaa !13
diff --git a/bench/opencv/optimized/fast_gemm.ll b/bench/opencv/optimized/fast_gemm.ll
index df387e1350f..92580a0ab01 100644
--- a/bench/opencv/optimized/fast_gemm.ll
+++ b/bench/opencv/optimized/fast_gemm.ll
@@ -1459,8 +1459,8 @@ _ZN2cv3dnn12cpu_baselineL20fast_gemm_pack12_f32EiiPKviiPv.exit.thread: ; preds =
 .preheader70.us.i:                                ; preds = %._crit_edge.us.i, %.preheader70.lr.ph.i
   %indvars.iv28 = phi i32 [ %indvars.iv.next29, %._crit_edge.us.i ], [ %., %.preheader70.lr.ph.i ]
   %indvars.iv91.i = phi i64 [ %indvars.iv.next92.i, %._crit_edge.us.i ], [ 0, %.preheader70.lr.ph.i ]
-  %332 = tail call i32 @llvm.smin.i32(i32 %indvars.iv28, i32 8)
-  %333 = tail call i32 @llvm.smax.i32(i32 %332, i32 1)
+  %332 = tail call i32 @llvm.smax.i32(i32 %indvars.iv28, i32 1)
+  %333 = tail call i32 @llvm.umin.i32(i32 %332, i32 8)
   %smax32 = zext nneg i32 %333 to i64
   %334 = trunc nuw nsw i64 %indvars.iv91.i to i32
   %factor.op.mul.reass.us.i = mul i32 %factor.op.mul76.i, %334
@@ -2246,8 +2246,8 @@ _ZN2cv3dnn12cpu_baselineL19fast_gemm_pack8_f32EiiPKviiPv.exit: ; preds = %.loope
 .preheader70.us.i:                                ; preds = %._crit_edge.us.i, %.preheader70.us.preheader.i
   %indvars.iv26 = phi i32 [ %indvars.iv.next27, %._crit_edge.us.i ], [ %., %.preheader70.us.preheader.i ]
   %indvars.iv91.i = phi i64 [ %indvars.iv.next92.i, %._crit_edge.us.i ], [ 0, %.preheader70.us.preheader.i ]
-  %219 = tail call i32 @llvm.smin.i32(i32 %indvars.iv26, i32 8)
-  %220 = tail call i32 @llvm.smax.i32(i32 %219, i32 1)
+  %219 = tail call i32 @llvm.smax.i32(i32 %indvars.iv26, i32 1)
+  %220 = tail call i32 @llvm.umin.i32(i32 %219, i32 8)
   %smax30 = zext nneg i32 %220 to i64
   %221 = trunc nuw nsw i64 %indvars.iv91.i to i32
   %factor.op.mul.reass.us.i = mul i32 %factor.op.mul76.i, %221
@@ -6366,8 +6366,8 @@ _ZN2cv3dnn12cpu_baselineL20fast_gemm_pack12_f32EiiPKviiPv.exit.thread: ; preds =
 .preheader70.us.i:                                ; preds = %._crit_edge.us.i, %.preheader70.lr.ph.i
   %indvars.iv28 = phi i32 [ %indvars.iv.next29, %._crit_edge.us.i ], [ %., %.preheader70.lr.ph.i ]
   %indvars.iv91.i = phi i64 [ %indvars.iv.next92.i, %._crit_edge.us.i ], [ 0, %.preheader70.lr.ph.i ]
-  %360 = tail call i32 @llvm.smin.i32(i32 %indvars.iv28, i32 8)
-  %361 = tail call i32 @llvm.smax.i32(i32 %360, i32 1)
+  %360 = tail call i32 @llvm.smax.i32(i32 %indvars.iv28, i32 1)
+  %361 = tail call i32 @llvm.umin.i32(i32 %360, i32 8)
   %smax32 = zext nneg i32 %361 to i64
   %362 = trunc nuw nsw i64 %indvars.iv91.i to i32
   %factor.op.mul.reass.us.i = mul i32 %factor.op.mul76.i, %362
@@ -7027,8 +7027,8 @@ _ZN2cv3dnn12cpu_baselineL19fast_gemm_pack8_f32EiiPKviiPv.exit: ; preds = %.loope
 .preheader70.us.i:                                ; preds = %._crit_edge.us.i, %.preheader70.us.preheader.i
   %indvars.iv26 = phi i32 [ %indvars.iv.next27, %._crit_edge.us.i ], [ %., %.preheader70.us.preheader.i ]
   %indvars.iv91.i = phi i64 [ %indvars.iv.next92.i, %._crit_edge.us.i ], [ 0, %.preheader70.us.preheader.i ]
-  %247 = tail call i32 @llvm.smin.i32(i32 %indvars.iv26, i32 8)
-  %248 = tail call i32 @llvm.smax.i32(i32 %247, i32 1)
+  %247 = tail call i32 @llvm.smax.i32(i32 %indvars.iv26, i32 1)
+  %248 = tail call i32 @llvm.umin.i32(i32 %247, i32 8)
   %smax30 = zext nneg i32 %248 to i64
   %249 = trunc nuw nsw i64 %indvars.iv91.i to i32
   %factor.op.mul.reass.us.i = mul i32 %factor.op.mul76.i, %249
@@ -7322,6 +7322,9 @@ declare i32 @llvm.smax.i32(i32, i32) #23
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.umin.i64(i64, i64) #23
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #23
+
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.umax.i64(i64, i64) #23
 
diff --git a/bench/opencv/optimized/finder_pattern_info.ll b/bench/opencv/optimized/finder_pattern_info.ll
index 173a88bc9e3..f5cc39cc9c2 100644
--- a/bench/opencv/optimized/finder_pattern_info.ll
+++ b/bench/opencv/optimized/finder_pattern_info.ll
@@ -341,8 +341,8 @@ _ZN5zxing3RefINS_6qrcode13FinderPatternEED2Ev.exit21: ; preds = %_ZN5zxing3RefIN
   %114 = load ptr, ptr %21, align 8, !tbaa !14
   %115 = tail call noundef i32 @_ZNK5zxing6qrcode13FinderPattern8getCountEv(ptr noundef nonnull align 8 dereferenceable(40) %114)
   %116 = add nsw i32 %113, %115
-  %.sroa.speculated34 = tail call i32 @llvm.smin.i32(i32 %116, i32 10)
-  %.sroa.speculated = tail call i32 @llvm.smax.i32(i32 %.sroa.speculated34, i32 3)
+  %.sroa.speculated34 = tail call i32 @llvm.smax.i32(i32 %116, i32 3)
+  %.sroa.speculated = tail call i32 @llvm.umin.i32(i32 %.sroa.speculated34, i32 10)
   %117 = add nsw i32 %.sroa.speculated, -3
   %118 = uitofp nneg i32 %117 to double
   %119 = fdiv double %118, 7.000000e+00
@@ -378,7 +378,7 @@ _ZN5zxing3RefINS_6qrcode13FinderPatternEED2Ev.exit21: ; preds = %_ZN5zxing3RefIN
   %136 = load ptr, ptr %22, align 8, !tbaa !8
   %137 = getelementptr inbounds nuw i8, ptr %136, i64 8
   %138 = load ptr, ptr %137, align 8
-  tail call void %138(ptr noundef nonnull align 8 dereferenceable(12) %22) #10
+  tail call void %140(ptr noundef nonnull align 8 dereferenceable(12) %22) #10
   br label %_ZN5zxing3RefINS_6qrcode13FinderPatternEED2Ev.exit23
 
 _ZN5zxing3RefINS_6qrcode13FinderPatternEED2Ev.exit23: ; preds = %135, %130, %128
@@ -397,7 +397,7 @@ _ZN5zxing3RefINS_6qrcode13FinderPatternEED2Ev.exit23: ; preds = %135, %130, %128
   %145 = load ptr, ptr %15, align 8, !tbaa !8
   %146 = getelementptr inbounds nuw i8, ptr %145, i64 8
   %147 = load ptr, ptr %146, align 8
-  tail call void %147(ptr noundef nonnull align 8 dereferenceable(12) %15) #10
+  tail call void %149(ptr noundef nonnull align 8 dereferenceable(12) %15) #10
   br label %_ZN5zxing3RefINS_6qrcode13FinderPatternEED2Ev.exit25
 
 _ZN5zxing3RefINS_6qrcode13FinderPatternEED2Ev.exit25: ; preds = %144, %139, %_ZN5zxing3RefINS_6qrcode13FinderPatternEED2Ev.exit23
@@ -416,7 +416,7 @@ _ZN5zxing3RefINS_6qrcode13FinderPatternEED2Ev.exit25: ; preds = %144, %139, %_ZN
   %154 = load ptr, ptr %9, align 8, !tbaa !8
   %155 = getelementptr inbounds nuw i8, ptr %154, i64 8
   %156 = load ptr, ptr %155, align 8
-  tail call void %156(ptr noundef nonnull align 8 dereferenceable(12) %9) #10
+  tail call void %158(ptr noundef nonnull align 8 dereferenceable(12) %9) #10
   br label %_ZN5zxing3RefINS_6qrcode13FinderPatternEED2Ev.exit27
 
 _ZN5zxing3RefINS_6qrcode13FinderPatternEED2Ev.exit27: ; preds = %_ZN5zxing3RefINS_6qrcode13FinderPatternEED2Ev.exit25, %148, %153
@@ -766,10 +766,10 @@ declare float @acosf(float noundef) local_unnamed_addr #8
 declare float @llvm.fabs.f32(float) #4
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smin.i32(i32, i32) #9
+declare i32 @llvm.smax.i32(i32, i32) #9
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smax.i32(i32, i32) #9
+declare i32 @llvm.umin.i32(i32, i32) #9
 
 attributes #0 = { mustprogress uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+sse3,+x87" "tune-cpu"="generic" }
 attributes #1 = { mustprogress nofree norecurse nosync nounwind willreturn memory(readwrite, inaccessiblemem: none) uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+sse3,+x87" "tune-cpu"="generic" }
diff --git a/bench/openexr/optimized/ImfLut.ll b/bench/openexr/optimized/ImfLut.ll
index d5a4521a7dd..fc792f22603 100644
--- a/bench/openexr/optimized/ImfLut.ll
+++ b/bench/openexr/optimized/ImfLut.ll
@@ -942,74 +942,74 @@ define i16 @_ZN7Imf_3_410round12logEN9Imath_3_24halfE(i16 %0) local_unnamed_addr
   %12 = fdiv double %11, 0x3FE62E42FEFA39EF
   %13 = fadd double %12, 2.000500e+03
   %14 = fptosi double %13 to i32
-  %spec.store.select = tail call i32 @llvm.smin.i32(i32 %14, i32 4095)
-  %spec.store.select1 = tail call i32 @llvm.smax.i32(i32 %spec.store.select, i32 1)
+  %spec.store.select = tail call i32 @llvm.smax.i32(i32 %14, i32 1)
+  %spec.store.select1 = tail call i32 @llvm.umin.i32(i32 %spec.store.select, i32 4095)
   %15 = add nsw i32 %spec.store.select1, -2000
   %16 = sitofp i32 %15 to double
   %17 = fdiv double %16, 2.000000e+02
   %exp2 = tail call double @exp2(double %17)
   %18 = fmul double %exp2, 0x3FC6A09E60000000
   %19 = fptrunc double %18 to float
-  %20 = bitcast float %19 to i32
-  %21 = tail call float @llvm.fabs.f32(float %19)
   %22 = bitcast float %21 to i32
-  %23 = lshr i32 %20, 16
-  %24 = trunc nuw i32 %23 to i16
-  %25 = and i16 %24, -32768
-  %26 = icmp samesign ugt i32 %22, 947912703
-  br i1 %26, label %27, label %53
-
-27:                                               ; preds = %7
-  %28 = icmp samesign ugt i32 %22, 2139095039
-  br i1 %28, label %29, label %40, !prof !43
-
-29:                                               ; preds = %27
-  %30 = or disjoint i16 %25, 31744
-  %31 = icmp eq i32 %22, 2139095040
-  br i1 %31, label %_ZN9Imath_3_24halfC2Ef.exit, label %32
-
-32:                                               ; preds = %29
-  %33 = lshr i32 %22, 13
-  %34 = and i32 %33, 1023
-  %35 = icmp eq i32 %34, 0
-  %36 = zext i1 %35 to i16
-  %37 = trunc nuw nsw i32 %34 to i16
-  %38 = or i16 %37, %36
-  %39 = or disjoint i16 %38, %30
+  %23 = tail call float @llvm.fabs.f32(float %21)
+  %24 = bitcast float %23 to i32
+  %25 = lshr i32 %22, 16
+  %26 = trunc nuw i32 %25 to i16
+  %27 = and i16 %26, -32768
+  %28 = icmp samesign ugt i32 %24, 947912703
+  br i1 %28, label %29, label %55
+
+29:  ; preds = %7
+  %30 = icmp samesign ugt i32 %24, 2139095039
+  br i1 %30, label %31, label %42, !prof !43
+
+31:; preds = %29
+  %32 = or disjoint i16 %27, 31744
+  %33 = icmp eq i32 %24, 2139095040
+  br i1 %33, label %_ZN9Imath_3_24halfC2Ef.exit, label %34
+
+34:; preds = %31
+  %35 = lshr i32 %24, 13
+  %36 = and i32 %35, 1023
+  %37 = icmp eq i32 %36, 0
+  %38 = zext i1 %37 to i16
+  %39 = trunc nuw nsw i32 %36 to i16
+  %40 = or i16 %39, %38
+  %41 = or disjoint i16 %40, %32
   br label %_ZN9Imath_3_24halfC2Ef.exit
 
-40:                                               ; preds = %27
-  %41 = icmp samesign ugt i32 %22, 1199566847
+40:                                               ; preds = %29
+  %41 = icmp samesign ugt i32 %24, 1199566847
   br i1 %41, label %42, label %44, !prof !43
 
 42:                                               ; preds = %40
-  %43 = or disjoint i16 %25, 31744
+  %43 = or disjoint i16 %27, 31744
   br label %_ZN9Imath_3_24halfC2Ef.exit
 
 44:                                               ; preds = %40
-  %45 = add nuw nsw i32 %22, 134221823
-  %46 = lshr i32 %22, 13
+  %45 = add nuw nsw i32 %24, 134221823
+  %46 = lshr i32 %24, 13
   %47 = and i32 %46, 1
   %48 = add nuw nsw i32 %45, %47
   %49 = lshr i32 %48, 13
-  %50 = and i32 %23, 32768
+  %50 = and i32 %25, 32768
   %51 = or i32 %49, %50
   %52 = trunc i32 %51 to i16
   br label %_ZN9Imath_3_24halfC2Ef.exit
 
 53:                                               ; preds = %7
-  %54 = icmp samesign ult i32 %22, 855638017
+  %54 = icmp samesign ult i32 %24, 855638017
   br i1 %54, label %_ZN9Imath_3_24halfC2Ef.exit, label %55
 
 55:                                               ; preds = %53
-  %56 = lshr i32 %22, 23
+  %56 = lshr i32 %24, 23
   %57 = sub nuw nsw i32 126, %56
-  %58 = and i32 %22, 8388607
+  %58 = and i32 %24, 8388607
   %59 = or disjoint i32 %58, 8388608
   %60 = add nsw i32 %56, -94
   %61 = shl i32 %59, %60
   %62 = lshr i32 %59, %57
-  %63 = and i32 %23, 32768
+  %63 = and i32 %25, 32768
   %64 = or i32 %62, %63
   %65 = trunc nuw i32 %64 to i16
   %66 = icmp ugt i32 %61, -2147483648
@@ -1026,8 +1026,8 @@ define i16 @_ZN7Imf_3_410round12logEN9Imath_3_24halfE(i16 %0) local_unnamed_addr
   %71 = add nuw i16 %65, 1
   br label %_ZN9Imath_3_24halfC2Ef.exit
 
-_ZN9Imath_3_24halfC2Ef.exit:                      ; preds = %70, %67, %53, %44, %42, %32, %29, %1
-  %.sroa.06.0 = phi i16 [ 0, %1 ], [ %39, %32 ], [ %43, %42 ], [ %52, %44 ], [ %30, %29 ], [ %25, %53 ], [ %71, %70 ], [ %65, %67 ]
+_ZN9Imath_3_24halfC2Ef.exit:                      ; preds = %70, %67, %53, %44, %42, %34, %31, %1
+  %.sroa.06.0 = phi i16 [ 0, %1 ], [ %41, %34 ], [ %43, %44 ], [ %52, %46 ], [ %32, %31 ], [ %27, %55 ], [ %71, %72 ], [ %65, %69 ]
   ret i16 %.sroa.06.0
 }
 
@@ -1043,9 +1043,6 @@ define internal void @_GLOBAL__sub_I_ImfLut.cpp() #7 section ".text.startup" {
   ret void
 }
 
-; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smin.i32(i32, i32) #8
-
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smax.i32(i32, i32) #8
 
@@ -1054,6 +1051,9 @@ declare double @exp2(double) local_unnamed_addr
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare float @llvm.fabs.f32(float) #8
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #8
+
 attributes #0 = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { nounwind "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #2 = { nofree nounwind }
diff --git a/bench/openexr/optimized/ImfRgbaFile.ll b/bench/openexr/optimized/ImfRgbaFile.ll
index 5f678c936d9..caff8eff9ab 100644
--- a/bench/openexr/optimized/ImfRgbaFile.ll
+++ b/bench/openexr/optimized/ImfRgbaFile.ll
@@ -4077,7 +4077,7 @@ _ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc.exit55: ; preds = %17
 
 20:                                               ; preds = %_ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc.exit55
   invoke void @__cxa_throw(ptr nonnull %19, ptr nonnull @_ZTIN7Iex_3_46ArgExcE, ptr nonnull @_ZN7Iex_3_46ArgExcD1Ev) #34
-          to label %173 unwind label %21
+          to label %175 unwind label %21
 
 21:                                               ; preds = %17, %9, %20, %15, %_ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc.exit
   %22 = landingpad { ptr, i32 }
@@ -4199,16 +4199,16 @@ _ZN7Imf_3_413RgbaInputFile7FromYca10rotateBuf2Ei.exit: ; preds = %63
   br label %80
 
 .lr.ph95:                                         ; preds = %80
-  %.sroa.speculated70 = tail call i32 @llvm.smin.i32(i32 %73, i32 3)
   %77 = getelementptr inbounds nuw i8, ptr %0, i64 80
   %78 = getelementptr inbounds nuw i8, ptr %0, i64 64
   %79 = getelementptr inbounds nuw i8, ptr %0, i64 336
-  %smax = tail call i32 @llvm.smax.i32(i32 %.sroa.speculated70, i32 1)
-  %wide.trip.count = zext nneg i32 %smax to i64
-  br label %86
+  %smax = tail call i32 @llvm.smax.i32(i32 %73, i32 1)
+  %81 = tail call i32 @llvm.umin.i32(i32 %smax, i32 3)
+  %wide.trip.count = zext nneg i32 %81 to i64
+  br label %88
 
 80:                                               ; preds = %72, %80
-  %indvars.iv105 = phi i64 [ %umin104, %72 ], [ %indvars.iv.next106, %80 ]
+  %indvars.iv105 = phi i64 [ %umin104, %72 ], [ %indvars.iv.next106, %82 ]
   %indvars.iv.next106 = add nsw i64 %indvars.iv105, -1
   %81 = getelementptr inbounds nuw [29 x ptr], ptr %75, i64 0, i64 %indvars.iv.next106
   %82 = load ptr, ptr %81, align 8, !tbaa !41
@@ -4218,35 +4218,35 @@ _ZN7Imf_3_413RgbaInputFile7FromYca10rotateBuf2Ei.exit: ; preds = %63
   %85 = icmp samesign ugt i64 %indvars.iv105, 1
   br i1 %85, label %80, label %.lr.ph95, !llvm.loop !145
 
-86:                                               ; preds = %.lr.ph95, %103
-  %indvars.iv107 = phi i64 [ 0, %.lr.ph95 ], [ %indvars.iv.next108, %103 ]
+86:                                               ; preds = %.lr.ph95, %105
+  %indvars.iv107 = phi i64 [ 0, %.lr.ph95 ], [ %indvars.iv.next108, %105 ]
   %87 = trunc i64 %indvars.iv107 to i32
   %88 = add i32 %1, %87
   %89 = and i32 %88, 1
   %.not53 = icmp eq i32 %89, 0
   %90 = load i32, ptr %78, align 8, !tbaa !117
-  br i1 %.not53, label %97, label %91
+  br i1 %.not53, label %99, label %91
 
 91:                                               ; preds = %86
   %92 = add nuw nsw i64 %indvars.iv107, 13
   %93 = getelementptr inbounds nuw [29 x ptr], ptr %75, i64 0, i64 %92
-  %94 = load ptr, ptr %93, align 8, !tbaa !41
-  %95 = getelementptr inbounds nuw [3 x ptr], ptr %79, i64 0, i64 %indvars.iv107
   %96 = load ptr, ptr %95, align 8, !tbaa !41
-  tail call void @_ZN7Imf_3_47RgbaYca9YCAtoRGBAERKN9Imath_3_24Vec3IfEEiPKNS_4RgbaEPS6_(ptr noundef nonnull align 4 dereferenceable(12) %77, i32 noundef %90, ptr noundef %94, ptr noundef %96)
-  br label %103
-
-97:                                               ; preds = %86
-  %98 = getelementptr inbounds nuw ptr, ptr %75, i64 %indvars.iv107
-  %99 = getelementptr inbounds nuw [3 x ptr], ptr %79, i64 0, i64 %indvars.iv107
-  %100 = load ptr, ptr %99, align 8, !tbaa !41
-  tail call void @_ZN7Imf_3_47RgbaYca21reconstructChromaVertEiPKPKNS_4RgbaEPS1_(i32 noundef %90, ptr noundef nonnull %98, ptr noundef %100)
-  %101 = load i32, ptr %78, align 8, !tbaa !117
-  %102 = load ptr, ptr %99, align 8, !tbaa !41
-  tail call void @_ZN7Imf_3_47RgbaYca9YCAtoRGBAERKN9Imath_3_24Vec3IfEEiPKNS_4RgbaEPS6_(ptr noundef nonnull align 4 dereferenceable(12) %77, i32 noundef %101, ptr noundef %102, ptr noundef %102)
-  br label %103
-
-103:                                              ; preds = %91, %97
+  %97 = getelementptr inbounds nuw [3 x ptr], ptr %79, i64 0, i64 %indvars.iv107
+  %98 = load ptr, ptr %97, align 8, !tbaa !41
+  tail call void @_ZN7Imf_3_47RgbaYca9YCAtoRGBAERKN9Imath_3_24Vec3IfEEiPKNS_4RgbaEPS6_(ptr noundef nonnull align 4 dereferenceable(12) %77, i32 noundef %92, ptr noundef %96, ptr noundef %98)
+  br label %105
+
+99: ; preds = %86
+  %100 = getelementptr inbounds nuw ptr, ptr %75, i64 %indvars.iv107
+  %101 = getelementptr inbounds nuw [3 x ptr], ptr %79, i64 0, i64 %indvars.iv107
+  %101 = load ptr, ptr %101, align 8, !tbaa !41
+  tail call void @_ZN7Imf_3_47RgbaYca21reconstructChromaVertEiPKPKNS_4RgbaEPS1_(i32 noundef %92, ptr noundef nonnull %100, ptr noundef %102)
+  %103 = load i32, ptr %78, align 8, !tbaa !117
+  %104 = load ptr, ptr %101, align 8, !tbaa !41
+  tail call void @_ZN7Imf_3_47RgbaYca9YCAtoRGBAERKN9Imath_3_24Vec3IfEEiPKNS_4RgbaEPS6_(ptr noundef nonnull align 4 dereferenceable(12) %77, i32 noundef %103, ptr noundef %104, ptr noundef %104)
+  br label %105
+
+105:                                              ; preds = %93, %99
   %indvars.iv.next108 = add nuw nsw i64 %indvars.iv107, 1
   %exitcond.not = icmp eq i64 %indvars.iv.next108, %wide.trip.count
   br i1 %exitcond.not, label %.loopexit, label %86, !llvm.loop !146
@@ -4284,40 +4284,40 @@ select.unfold:                                    ; preds = %.lr.ph, %select.unf
   %120 = icmp samesign ugt i64 %indvars.iv, 1
   br i1 %120, label %select.unfold, label %.lr.ph92, !llvm.loop !147
 
-121:                                              ; preds = %.lr.ph92, %138
-  %indvars.iv101 = phi i64 [ 2, %.lr.ph92 ], [ %indvars.iv.next102, %138 ]
+121:                                              ; preds = %.lr.ph92, %140
+  %indvars.iv101 = phi i64 [ 2, %.lr.ph92 ], [ %indvars.iv.next102, %140 ]
   %122 = trunc i64 %indvars.iv101 to i32
   %123 = add i32 %1, %122
   %124 = and i32 %123, 1
   %.not = icmp eq i32 %124, 0
   %125 = load i32, ptr %110, align 8, !tbaa !117
-  br i1 %.not, label %132, label %126
+  br i1 %.not, label %134, label %126
 
 126:                                              ; preds = %121
   %127 = add nsw i64 %indvars.iv101, 13
   %128 = getelementptr inbounds [29 x ptr], ptr %111, i64 0, i64 %127
-  %129 = load ptr, ptr %128, align 8, !tbaa !41
-  %130 = getelementptr inbounds [3 x ptr], ptr %112, i64 0, i64 %indvars.iv101
   %131 = load ptr, ptr %130, align 8, !tbaa !41
-  tail call void @_ZN7Imf_3_47RgbaYca9YCAtoRGBAERKN9Imath_3_24Vec3IfEEiPKNS_4RgbaEPS6_(ptr noundef nonnull align 4 dereferenceable(12) %109, i32 noundef %125, ptr noundef %129, ptr noundef %131)
-  br label %138
-
-132:                                              ; preds = %121
-  %133 = getelementptr inbounds ptr, ptr %111, i64 %indvars.iv101
-  %134 = getelementptr inbounds [3 x ptr], ptr %112, i64 0, i64 %indvars.iv101
-  %135 = load ptr, ptr %134, align 8, !tbaa !41
-  tail call void @_ZN7Imf_3_47RgbaYca21reconstructChromaVertEiPKPKNS_4RgbaEPS1_(i32 noundef %125, ptr noundef nonnull %133, ptr noundef %135)
-  %136 = load i32, ptr %110, align 8, !tbaa !117
-  %137 = load ptr, ptr %134, align 8, !tbaa !41
-  tail call void @_ZN7Imf_3_47RgbaYca9YCAtoRGBAERKN9Imath_3_24Vec3IfEEiPKNS_4RgbaEPS6_(ptr noundef nonnull align 4 dereferenceable(12) %109, i32 noundef %136, ptr noundef %137, ptr noundef %137)
-  br label %138
-
-138:                                              ; preds = %126, %132
+  %132 = getelementptr inbounds [3 x ptr], ptr %112, i64 0, i64 %indvars.iv101
+  %133 = load ptr, ptr %132, align 8, !tbaa !41
+  tail call void @_ZN7Imf_3_47RgbaYca9YCAtoRGBAERKN9Imath_3_24Vec3IfEEiPKNS_4RgbaEPS6_(ptr noundef nonnull align 4 dereferenceable(12) %111, i32 noundef %127, ptr noundef %131, ptr noundef %133)
+  br label %140
+
+134: ; preds = %121
+  %135 = getelementptr inbounds ptr, ptr %113, i64 %indvars.iv101
+  %136 = getelementptr inbounds [3 x ptr], ptr %114, i64 0, i64 %indvars.iv101
+  %136 = load ptr, ptr %136, align 8, !tbaa !41
+  tail call void @_ZN7Imf_3_47RgbaYca21reconstructChromaVertEiPKPKNS_4RgbaEPS1_(i32 noundef %127, ptr noundef nonnull %135, ptr noundef %137)
+  %138 = load i32, ptr %112, align 8, !tbaa !117
+  %139 = load ptr, ptr %136, align 8, !tbaa !41
+  tail call void @_ZN7Imf_3_47RgbaYca9YCAtoRGBAERKN9Imath_3_24Vec3IfEEiPKNS_4RgbaEPS6_(ptr noundef nonnull align 4 dereferenceable(12) %111, i32 noundef %138, ptr noundef %139, ptr noundef %139)
+  br label %140
+
+140:                                              ; preds = %128, %134
   %indvars.iv.next102 = add nsw i64 %indvars.iv101, -1
   %139 = icmp sgt i64 %indvars.iv.next102, %113
   br i1 %139, label %121, label %.loopexit, !llvm.loop !148
 
-.loopexit:                                        ; preds = %138, %103, %104
+.loopexit:                                        ; preds = %140, %105, %104
   %140 = getelementptr inbounds nuw i8, ptr %0, i64 80
   %141 = getelementptr inbounds nuw i8, ptr %0, i64 64
   %142 = load i32, ptr %141, align 8, !tbaa !117
@@ -4343,7 +4343,7 @@ select.unfold:                                    ; preds = %.lr.ph, %select.unf
   ret void
 
 154:                                              ; preds = %.lr.ph97, %154
-  %indvars.iv110 = phi i64 [ 0, %.lr.ph97 ], [ %indvars.iv.next111, %154 ]
+  %indvars.iv110 = phi i64 [ 0, %.lr.ph97 ], [ %indvars.iv.next111, %156 ]
   %155 = load i64, ptr %150, align 8, !tbaa !140
   %156 = mul i64 %155, %151
   %157 = load i64, ptr %152, align 8, !tbaa !139
diff --git a/bench/openjdk/optimized/Net.ll b/bench/openjdk/optimized/Net.ll
index 0e1226f91d3..5218645ad5f 100644
--- a/bench/openjdk/optimized/Net.ll
+++ b/bench/openjdk/optimized/Net.ll
@@ -1592,8 +1592,8 @@ define range(i32 -32768, 32768) i32 @Java_sun_nio_ch_Net_poll(ptr noundef %0, pt
   %8 = trunc i32 %3 to i16
   %9 = getelementptr inbounds nuw i8, ptr %6, i64 4
   store i16 %8, ptr %9, align 4
-  %spec.store.select = tail call i64 @llvm.smin.i64(i64 %4, i64 2147483647)
-  %.0811 = tail call i64 @llvm.smax.i64(i64 %spec.store.select, i64 -1)
+  %spec.store.select = tail call i64 @llvm.smax.i64(i64 %4, i64 -1)
+  %.0811 = tail call i64 @llvm.smin.i64(i64 %spec.store.select, i64 2147483647)
   %.08 = trunc nsw i64 %.0811 to i32
   %10 = call i32 @poll(ptr noundef nonnull %6, i64 noundef 1, i32 noundef %.08) #8
   %11 = icmp sgt i32 %10, -1
@@ -1608,7 +1608,7 @@ define range(i32 -32768, 32768) i32 @Java_sun_nio_ch_Net_poll(ptr noundef %0, pt
 16:                                               ; preds = %5
   %17 = tail call ptr @__errno_location() #7
   %18 = load i32, ptr %17, align 4
-  switch i32 %18, label %22 [
+  switch i32 %18, label %23 [
     i32 4, label %handleSocketError.exit
     i32 115, label %handleSocketError.exit.fold.split
     i32 71, label %23
@@ -1634,7 +1634,7 @@ define range(i32 -32768, 32768) i32 @Java_sun_nio_ch_Net_poll(ptr noundef %0, pt
   br label %23
 
 23:                                               ; preds = %16, %22, %21, %20, %19
-  %.0.i.i = phi ptr [ @.str.3, %22 ], [ @.str.17, %21 ], [ @.str.16, %20 ], [ @.str.15, %19 ], [ @.str.14, %16 ]
+  %.0.i.i = phi ptr [ @.str.3, %23 ], [ @.str.17, %22 ], [ @.str.16, %21 ], [ @.str.15, %20 ], [ @.str.14, %16 ]
   call void @JNU_ThrowByNameWithLastError(ptr noundef %0, ptr noundef nonnull %.0.i.i, ptr noundef nonnull @.str.18) #8
   br label %handleSocketError.exit
 
@@ -1642,7 +1642,7 @@ handleSocketError.exit.fold.split:                ; preds = %16
   br label %handleSocketError.exit
 
 handleSocketError.exit:                           ; preds = %16, %handleSocketError.exit.fold.split, %23, %12
-  %.0 = phi i32 [ %15, %12 ], [ 0, %16 ], [ -5, %23 ], [ -5, %handleSocketError.exit.fold.split ]
+  %.0 = phi i32 [ %15, %12 ], [ 0, %16 ], [ -5, %24 ], [ -5, %handleSocketError.exit.fold.split ]
   ret i32 %.0
 }
 
@@ -1659,8 +1659,8 @@ define zeroext range(i8 0, 2) i8 @Java_sun_nio_ch_Net_pollConnect(ptr noundef %0
   store i16 4, ptr %9, align 4
   %10 = getelementptr inbounds nuw i8, ptr %5, i64 6
   store i16 0, ptr %10, align 2
-  %spec.store.select = tail call i64 @llvm.smin.i64(i64 %3, i64 2147483647)
-  %.01424 = tail call i64 @llvm.smax.i64(i64 %spec.store.select, i64 -1)
+  %spec.store.select = tail call i64 @llvm.smax.i64(i64 %3, i64 -1)
+  %.01424 = tail call i64 @llvm.smin.i64(i64 %spec.store.select, i64 2147483647)
   %.014 = trunc nsw i64 %.01424 to i32
   %11 = call i32 @poll(ptr noundef nonnull %5, i64 noundef 1, i32 noundef %.014) #8
   %12 = icmp sgt i32 %11, 0
@@ -1677,7 +1677,7 @@ define zeroext range(i8 0, 2) i8 @Java_sun_nio_ch_Net_pollConnect(ptr noundef %0
 
 17:                                               ; preds = %13
   %18 = load i32, ptr %14, align 4
-  switch i32 %18, label %22 [
+  switch i32 %18, label %23 [
     i32 115, label %handleSocketError.exit
     i32 71, label %23
     i32 111, label %19
@@ -1701,14 +1701,14 @@ define zeroext range(i8 0, 2) i8 @Java_sun_nio_ch_Net_pollConnect(ptr noundef %0
 22:                                               ; preds = %17
   br label %23
 
-23:                                               ; preds = %22, %21, %20, %19, %17
-  %.0.i.i = phi ptr [ @.str.3, %22 ], [ @.str.17, %21 ], [ @.str.16, %20 ], [ @.str.15, %19 ], [ @.str.14, %17 ]
+23:                                               ; preds = %23, %22, %21, %20, %17
+  %.0.i.i = phi ptr [ @.str.3, %23 ], [ @.str.17, %22 ], [ @.str.16, %21 ], [ @.str.15, %20 ], [ @.str.14, %17 ]
   call void @JNU_ThrowByNameWithLastError(ptr noundef %0, ptr noundef nonnull %.0.i.i, ptr noundef nonnull @.str.18) #8
   br label %handleSocketError.exit
 
 24:                                               ; preds = %13
   %25 = load i32, ptr %6, align 4
-  switch i32 %25, label %29 [
+  switch i32 %25, label %30 [
     i32 0, label %31
     i32 115, label %handleSocketError.exit
     i32 71, label %30
@@ -1734,7 +1734,7 @@ define zeroext range(i8 0, 2) i8 @Java_sun_nio_ch_Net_pollConnect(ptr noundef %0
   br label %30
 
 30:                                               ; preds = %24, %29, %28, %27, %26
-  %.0.i.i19 = phi ptr [ @.str.3, %29 ], [ @.str.17, %28 ], [ @.str.16, %27 ], [ @.str.15, %26 ], [ @.str.14, %24 ]
+  %.0.i.i19 = phi ptr [ @.str.3, %30 ], [ @.str.17, %29 ], [ @.str.16, %28 ], [ @.str.15, %27 ], [ @.str.14, %25 ]
   store i32 %25, ptr %14, align 4
   call void @JNU_ThrowByNameWithLastError(ptr noundef %0, ptr noundef nonnull %.0.i.i19, ptr noundef nonnull @.str.18) #8
   br label %handleSocketError.exit
@@ -1765,7 +1765,7 @@ define zeroext range(i8 0, 2) i8 @Java_sun_nio_ch_Net_pollConnect(ptr noundef %0
   br label %handleSocketError.exit
 
 handleSocketError.exit:                           ; preds = %24, %30, %23, %17, %35, %37, %31, %41, %34
-  %.0 = phi i8 [ 0, %34 ], [ 0, %41 ], [ 1, %31 ], [ 0, %37 ], [ 0, %35 ], [ 0, %17 ], [ 0, %23 ], [ 0, %30 ], [ 0, %24 ]
+  %.0 = phi i8 [ 0, %35 ], [ 0, %42 ], [ 1, %32 ], [ 0, %38 ], [ 0, %36 ], [ 0, %17 ], [ 0, %24 ], [ 0, %31 ], [ 0, %25 ]
   ret i8 %.0
 }
 
diff --git a/bench/openjdk/optimized/cmsopt.ll b/bench/openjdk/optimized/cmsopt.ll
index 78f445a3eba..d3c38c9ad65 100644
--- a/bench/openjdk/optimized/cmsopt.ll
+++ b/bench/openjdk/optimized/cmsopt.ll
@@ -3543,12 +3543,12 @@ define internal void @MatShaperEval16(ptr noundef readonly captures(none) %0, pt
   %69 = add i32 %68, %64
   %70 = add i32 %69, %66
   %71 = ashr i32 %70, 14
-  %72 = tail call i32 @llvm.smin.i32(i32 %39, i32 16384)
-  %73 = tail call i32 @llvm.smax.i32(i32 %72, i32 0)
-  %74 = tail call i32 @llvm.smin.i32(i32 %55, i32 16384)
-  %75 = tail call i32 @llvm.smax.i32(i32 %74, i32 0)
-  %76 = tail call i32 @llvm.smin.i32(i32 %71, i32 16384)
-  %77 = tail call i32 @llvm.smax.i32(i32 %76, i32 0)
+  %72 = tail call i32 @llvm.smax.i32(i32 %39, i32 0)
+  %73 = tail call i32 @llvm.umin.i32(i32 %72, i32 16384)
+  %74 = tail call i32 @llvm.smax.i32(i32 %55, i32 0)
+  %75 = tail call i32 @llvm.umin.i32(i32 %74, i32 16384)
+  %76 = tail call i32 @llvm.smax.i32(i32 %71, i32 0)
+  %77 = tail call i32 @llvm.umin.i32(i32 %76, i32 16384)
   %78 = getelementptr inbounds nuw i8, ptr %2, i64 3128
   %79 = zext nneg i32 %73 to i64
   %80 = getelementptr inbounds nuw [16385 x i16], ptr %78, i64 0, i64 %79
@@ -4112,9 +4112,6 @@ define internal ptr @Prelin8dup(ptr noundef %0, ptr noundef %1) #0 {
 
 declare i32 @cmsIsToneCurveDescending(ptr noundef) local_unnamed_addr #1
 
-; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smin.i32(i32, i32) #8
-
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smax.i32(i32, i32) #8
 
@@ -4124,6 +4121,9 @@ declare void @llvm.lifetime.start.p0(i64 immarg, ptr captures(none)) #9
 ; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
 declare void @llvm.lifetime.end.p0(i64 immarg, ptr captures(none)) #9
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #8
+
 attributes #0 = { nounwind uwtable "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #2 = { nofree norecurse nosync nounwind memory(argmem: readwrite) uwtable "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
diff --git a/bench/openjdk/optimized/compilerDefinitions.ll b/bench/openjdk/optimized/compilerDefinitions.ll
index 533d342dbe6..7064136fdf1 100644
--- a/bench/openjdk/optimized/compilerDefinitions.ll
+++ b/bench/openjdk/optimized/compilerDefinitions.ll
@@ -476,8 +476,8 @@ define hidden noundef range(i64 0, 2147483648) i64 @_ZN14CompilerConfig32jvmflag
 _ZN14CompilerConfig24scaled_compile_thresholdEl.exit: ; preds = %1, %6, %11
   %.0.i.i = phi i64 [ %0, %1 ], [ 9223372036854775807, %6 ], [ %spec.select.i.i, %11 ]
   call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %2)
-  %16 = tail call noundef i64 @llvm.smin.i64(i64 %.0.i.i, i64 2147483647)
-  %17 = tail call noundef i64 @llvm.smax.i64(i64 %16, i64 0)
+  %16 = tail call i64 @llvm.smax.i64(i64 %.0.i.i, i64 0)
+  %17 = tail call i64 @llvm.umin.i64(i64 %16, i64 2147483647)
   ret i64 %17
 }
 
@@ -532,8 +532,8 @@ _ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i: ; preds = %15
 
 _ZN14CompilerConfig15scaled_freq_logEl.exit:      ; preds = %1, %6, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i, %21
   %.0.i.i = phi i64 [ %26, %21 ], [ %0, %1 ], [ 0, %6 ], [ 0, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i ]
-  %27 = tail call noundef i64 @llvm.smin.i64(i64 %.0.i.i, i64 30)
-  %28 = tail call noundef i64 @llvm.smax.i64(i64 %27, i64 0)
+  %27 = tail call i64 @llvm.smax.i64(i64 %.0.i.i, i64 0)
+  %28 = tail call i64 @llvm.umin.i64(i64 %27, i64 30)
   ret i64 %28
 }
 
@@ -1249,8 +1249,8 @@ _ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i: ; preds = %129
 
 _ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit: ; preds = %118, %121, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i, %135
   %.0.i.i.i = phi i64 [ %140, %135 ], [ %119, %118 ], [ 0, %121 ], [ 0, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i ]
-  %141 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i, i64 30)
-  %142 = call noundef range(i64 0, 31) i64 @llvm.smax.i64(i64 %141, i64 0)
+  %141 = call i64 @llvm.smax.i64(i64 %.0.i.i.i, i64 0)
+  %142 = call noundef range(i64 0, 31) i64 @llvm.umin.i64(i64 %141, i64 30)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %29)
   store i64 %142, ptr %29, align 8
   %143 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 392, i32 noundef 3, ptr noundef nonnull %29, i32 noundef 5) #11
@@ -1304,8 +1304,8 @@ _ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i10: ; preds = %157
 
 _ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit14: ; preds = %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit, %148, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i10, %163
   %.0.i.i.i12 = phi i64 [ %168, %163 ], [ %144, %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit ], [ 0, %148 ], [ 0, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i10 ]
-  %169 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i12, i64 30)
-  %170 = call noundef range(i64 0, 31) i64 @llvm.smax.i64(i64 %169, i64 0)
+  %169 = call i64 @llvm.smax.i64(i64 %.0.i.i.i12, i64 0)
+  %170 = call noundef range(i64 0, 31) i64 @llvm.umin.i64(i64 %169, i64 30)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %27)
   store i64 %170, ptr %27, align 8
   %171 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 396, i32 noundef 3, ptr noundef nonnull %27, i32 noundef 5) #11
@@ -1336,8 +1336,8 @@ _ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit14: ; preds = %_ZN14CompilerC
 _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit: ; preds = %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit14, %176, %181
   %.0.i.i.i16 = phi i64 [ %172, %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit14 ], [ 9223372036854775807, %176 ], [ %spec.select.i.i.i, %181 ]
   call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %26)
-  %186 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i16, i64 2147483647)
-  %187 = call noundef range(i64 0, 2147483648) i64 @llvm.smax.i64(i64 %186, i64 0)
+  %186 = call i64 @llvm.smax.i64(i64 %.0.i.i.i16, i64 0)
+  %187 = call noundef range(i64 0, 2147483648) i64 @llvm.umin.i64(i64 %186, i64 2147483647)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %25)
   store i64 %187, ptr %25, align 8
   %188 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 401, i32 noundef 3, ptr noundef nonnull %25, i32 noundef 5) #11
@@ -1368,8 +1368,8 @@ _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit: ; preds = %_ZN14Co
 _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit20: ; preds = %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit, %193, %198
   %.0.i.i.i19 = phi i64 [ %189, %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit ], [ 9223372036854775807, %193 ], [ %spec.select.i.i.i18, %198 ]
   call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %24)
-  %203 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i19, i64 2147483647)
-  %204 = call noundef range(i64 0, 2147483648) i64 @llvm.smax.i64(i64 %203, i64 0)
+  %203 = call i64 @llvm.smax.i64(i64 %.0.i.i.i19, i64 0)
+  %204 = call noundef range(i64 0, 2147483648) i64 @llvm.umin.i64(i64 %203, i64 2147483647)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %23)
   store i64 %204, ptr %23, align 8
   %205 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 402, i32 noundef 3, ptr noundef nonnull %23, i32 noundef 5) #11
@@ -1400,8 +1400,8 @@ _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit20: ; preds = %_ZN14
 _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit24: ; preds = %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit20, %210, %215
   %.0.i.i.i23 = phi i64 [ %206, %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit20 ], [ 9223372036854775807, %210 ], [ %spec.select.i.i.i22, %215 ]
   call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %22)
-  %220 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i23, i64 2147483647)
-  %221 = call noundef range(i64 0, 2147483648) i64 @llvm.smax.i64(i64 %220, i64 0)
+  %220 = call i64 @llvm.smax.i64(i64 %.0.i.i.i23, i64 0)
+  %221 = call noundef range(i64 0, 2147483648) i64 @llvm.umin.i64(i64 %220, i64 2147483647)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %21)
   store i64 %221, ptr %21, align 8
   %222 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 403, i32 noundef 3, ptr noundef nonnull %21, i32 noundef 5) #11
@@ -1432,8 +1432,8 @@ _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit24: ; preds = %_ZN14
 _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit28: ; preds = %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit24, %227, %232
   %.0.i.i.i27 = phi i64 [ %223, %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit24 ], [ 9223372036854775807, %227 ], [ %spec.select.i.i.i26, %232 ]
   call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %20)
-  %237 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i27, i64 2147483647)
-  %238 = call noundef range(i64 0, 2147483648) i64 @llvm.smax.i64(i64 %237, i64 0)
+  %237 = call i64 @llvm.smax.i64(i64 %.0.i.i.i27, i64 0)
+  %238 = call noundef range(i64 0, 2147483648) i64 @llvm.umin.i64(i64 %237, i64 2147483647)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %19)
   store i64 %238, ptr %19, align 8
   %239 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 404, i32 noundef 3, ptr noundef nonnull %19, i32 noundef 5) #11
@@ -1487,8 +1487,8 @@ _ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i31: ; preds = %253
 
 _ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit35: ; preds = %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit28, %244, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i31, %259
   %.0.i.i.i33 = phi i64 [ %264, %259 ], [ %240, %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit28 ], [ 0, %244 ], [ 0, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i31 ]
-  %265 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i33, i64 30)
-  %266 = call noundef range(i64 0, 31) i64 @llvm.smax.i64(i64 %265, i64 0)
+  %265 = call i64 @llvm.smax.i64(i64 %.0.i.i.i33, i64 0)
+  %266 = call noundef range(i64 0, 31) i64 @llvm.umin.i64(i64 %265, i64 30)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %17)
   store i64 %266, ptr %17, align 8
   %267 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 393, i32 noundef 3, ptr noundef nonnull %17, i32 noundef 5) #11
@@ -1542,8 +1542,8 @@ _ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i38: ; preds = %281
 
 _ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit42: ; preds = %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit35, %272, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i38, %287
   %.0.i.i.i40 = phi i64 [ %292, %287 ], [ %268, %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit35 ], [ 0, %272 ], [ 0, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i38 ]
-  %293 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i40, i64 30)
-  %294 = call noundef range(i64 0, 31) i64 @llvm.smax.i64(i64 %293, i64 0)
+  %293 = call i64 @llvm.smax.i64(i64 %.0.i.i.i40, i64 0)
+  %294 = call noundef range(i64 0, 31) i64 @llvm.umin.i64(i64 %293, i64 30)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %15)
   store i64 %294, ptr %15, align 8
   %295 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 397, i32 noundef 3, ptr noundef nonnull %15, i32 noundef 5) #11
@@ -1597,8 +1597,8 @@ _ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i45: ; preds = %309
 
 _ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit49: ; preds = %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit42, %300, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i45, %315
   %.0.i.i.i47 = phi i64 [ %320, %315 ], [ %296, %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit42 ], [ 0, %300 ], [ 0, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i45 ]
-  %321 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i47, i64 30)
-  %322 = call noundef range(i64 0, 31) i64 @llvm.smax.i64(i64 %321, i64 0)
+  %321 = call i64 @llvm.smax.i64(i64 %.0.i.i.i47, i64 0)
+  %322 = call noundef range(i64 0, 31) i64 @llvm.umin.i64(i64 %321, i64 30)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %13)
   store i64 %322, ptr %13, align 8
   %323 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 394, i32 noundef 3, ptr noundef nonnull %13, i32 noundef 5) #11
@@ -1652,8 +1652,8 @@ _ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i52: ; preds = %337
 
 _ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit56: ; preds = %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit49, %328, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i52, %343
   %.0.i.i.i54 = phi i64 [ %348, %343 ], [ %324, %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit49 ], [ 0, %328 ], [ 0, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i52 ]
-  %349 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i54, i64 30)
-  %350 = call noundef range(i64 0, 31) i64 @llvm.smax.i64(i64 %349, i64 0)
+  %349 = call i64 @llvm.smax.i64(i64 %.0.i.i.i54, i64 0)
+  %350 = call noundef range(i64 0, 31) i64 @llvm.umin.i64(i64 %349, i64 30)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %11)
   store i64 %350, ptr %11, align 8
   %351 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 398, i32 noundef 3, ptr noundef nonnull %11, i32 noundef 5) #11
@@ -1707,8 +1707,8 @@ _ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i59: ; preds = %365
 
 _ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit63: ; preds = %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit56, %356, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i59, %371
   %.0.i.i.i61 = phi i64 [ %376, %371 ], [ %352, %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit56 ], [ 0, %356 ], [ 0, %_ZN14CompilerConfig24scaled_compile_thresholdEld.exit.i.i.i59 ]
-  %377 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i61, i64 30)
-  %378 = call noundef range(i64 0, 31) i64 @llvm.smax.i64(i64 %377, i64 0)
+  %377 = call i64 @llvm.smax.i64(i64 %.0.i.i.i61, i64 0)
+  %378 = call noundef range(i64 0, 31) i64 @llvm.umin.i64(i64 %377, i64 30)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %9)
   store i64 %378, ptr %9, align 8
   %379 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 395, i32 noundef 3, ptr noundef nonnull %9, i32 noundef 5) #11
@@ -1739,8 +1739,8 @@ _ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit63: ; preds = %_ZN14CompilerC
 _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit67: ; preds = %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit63, %384, %389
   %.0.i.i.i66 = phi i64 [ %380, %_ZN14CompilerConfig23jvmflag_scaled_freq_logEl.exit63 ], [ 9223372036854775807, %384 ], [ %spec.select.i.i.i65, %389 ]
   call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %8)
-  %394 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i66, i64 2147483647)
-  %395 = call noundef range(i64 0, 2147483648) i64 @llvm.smax.i64(i64 %394, i64 0)
+  %394 = call i64 @llvm.smax.i64(i64 %.0.i.i.i66, i64 0)
+  %395 = call noundef range(i64 0, 2147483648) i64 @llvm.umin.i64(i64 %394, i64 2147483647)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %7)
   store i64 %395, ptr %7, align 8
   %396 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 405, i32 noundef 3, ptr noundef nonnull %7, i32 noundef 5) #11
@@ -1771,8 +1771,8 @@ _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit67: ; preds = %_ZN14
 _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit71: ; preds = %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit67, %401, %406
   %.0.i.i.i70 = phi i64 [ %397, %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit67 ], [ 9223372036854775807, %401 ], [ %spec.select.i.i.i69, %406 ]
   call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %6)
-  %411 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i70, i64 2147483647)
-  %412 = call noundef range(i64 0, 2147483648) i64 @llvm.smax.i64(i64 %411, i64 0)
+  %411 = call i64 @llvm.smax.i64(i64 %.0.i.i.i70, i64 0)
+  %412 = call noundef range(i64 0, 2147483648) i64 @llvm.umin.i64(i64 %411, i64 2147483647)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %5)
   store i64 %412, ptr %5, align 8
   %413 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 406, i32 noundef 3, ptr noundef nonnull %5, i32 noundef 5) #11
@@ -1803,8 +1803,8 @@ _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit71: ; preds = %_ZN14
 _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit75: ; preds = %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit71, %418, %423
   %.0.i.i.i74 = phi i64 [ %414, %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit71 ], [ 9223372036854775807, %418 ], [ %spec.select.i.i.i73, %423 ]
   call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %4)
-  %428 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i74, i64 2147483647)
-  %429 = call noundef range(i64 0, 2147483648) i64 @llvm.smax.i64(i64 %428, i64 0)
+  %428 = call i64 @llvm.smax.i64(i64 %.0.i.i.i74, i64 0)
+  %429 = call noundef range(i64 0, 2147483648) i64 @llvm.umin.i64(i64 %428, i64 2147483647)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %3)
   store i64 %429, ptr %3, align 8
   %430 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 407, i32 noundef 3, ptr noundef nonnull %3, i32 noundef 5) #11
@@ -1835,8 +1835,8 @@ _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit75: ; preds = %_ZN14
 _ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit79: ; preds = %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit75, %435, %440
   %.0.i.i.i78 = phi i64 [ %431, %_ZN14CompilerConfig32jvmflag_scaled_compile_thresholdEl.exit75 ], [ 9223372036854775807, %435 ], [ %spec.select.i.i.i77, %440 ]
   call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %2)
-  %445 = call noundef i64 @llvm.smin.i64(i64 %.0.i.i.i78, i64 2147483647)
-  %446 = call noundef range(i64 0, 2147483648) i64 @llvm.smax.i64(i64 %445, i64 0)
+  %445 = call i64 @llvm.smax.i64(i64 %.0.i.i.i78, i64 0)
+  %446 = call noundef range(i64 0, 2147483648) i64 @llvm.umin.i64(i64 %445, i64 2147483647)
   call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %1)
   store i64 %446, ptr %1, align 8
   %447 = call noundef i32 @_ZN13JVMFlagAccess13set_or_assertE12JVMFlagsEnumiPv13JVMFlagOrigin(i32 noundef 408, i32 noundef 3, ptr noundef nonnull %1, i32 noundef 5) #11
@@ -2428,9 +2428,6 @@ declare i64 @llvm.ctlz.i64(i64, i1 immarg) #6
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.smax.i64(i64, i64) #7
 
-; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i64 @llvm.smin.i64(i64, i64) #7
-
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.umin.i64(i64, i64) #7
 
diff --git a/bench/openusd/optimized/loopfilter.ll b/bench/openusd/optimized/loopfilter.ll
index d3cb9972933..b3c0d78e33b 100644
--- a/bench/openusd/optimized/loopfilter.ll
+++ b/bench/openusd/optimized/loopfilter.ll
@@ -75,8 +75,8 @@ define hidden void @aom_lpf_horizontal_4_c(ptr noundef captures(none) %0, i32 no
   %68 = sub nsw i32 %66, %67
   %69 = mul nsw i32 %68, 3
   %70 = add nsw i32 %65, %69
-  %71 = tail call i32 @llvm.smin.i32(i32 range(i32 -893, 893) %70, i32 127)
-  %72 = tail call i32 @llvm.smax.i32(i32 %71, i32 -128)
+  %71 = tail call i32 @llvm.smax.i32(i32 range(i32 -893, 893) %70, i32 -128)
+  %72 = tail call i32 @llvm.smin.i32(i32 %71, i32 127)
   %73 = select i1 %46, i32 %72, i32 0
   %74 = tail call i32 @llvm.smin.i32(i32 %73, i32 123)
   %75 = trunc nsw i32 %74 to i8
@@ -176,8 +176,8 @@ define hidden void @aom_lpf_vertical_4_c(ptr noundef captures(none) %0, i32 noun
   %50 = sub nsw i32 %48, %49
   %51 = mul nsw i32 %50, 3
   %52 = add nsw i32 %47, %51
-  %53 = tail call i32 @llvm.smin.i32(i32 range(i32 -893, 893) %52, i32 127)
-  %54 = tail call i32 @llvm.smax.i32(i32 %53, i32 -128)
+  %53 = tail call i32 @llvm.smax.i32(i32 range(i32 -893, 893) %52, i32 -128)
+  %54 = tail call i32 @llvm.smin.i32(i32 %53, i32 127)
   %55 = select i1 %36, i32 %54, i32 0
   %56 = tail call i32 @llvm.smin.i32(i32 %55, i32 123)
   %57 = trunc nsw i32 %56 to i8
@@ -262,8 +262,8 @@ define hidden void @aom_lpf_vertical_4_c(ptr noundef captures(none) %0, i32 noun
   %121 = sub nsw i32 %119, %120
   %122 = mul nsw i32 %121, 3
   %123 = add nsw i32 %118, %122
-  %124 = tail call i32 @llvm.smin.i32(i32 range(i32 -893, 893) %123, i32 127)
-  %125 = tail call i32 @llvm.smax.i32(i32 %124, i32 -128)
+  %124 = tail call i32 @llvm.smax.i32(i32 range(i32 -893, 893) %123, i32 -128)
+  %125 = tail call i32 @llvm.smin.i32(i32 %124, i32 127)
   %126 = select i1 %107, i32 %125, i32 0
   %127 = tail call i32 @llvm.smin.i32(i32 %126, i32 123)
   %128 = trunc nsw i32 %127 to i8
@@ -481,8 +481,8 @@ define internal fastcc void @filter6(i8 noundef signext %0, i8 noundef zeroext %
   %79 = sub nsw i32 %77, %78
   %80 = mul nsw i32 %79, 3
   %81 = add nsw i32 %76, %80
-  %82 = tail call i32 @llvm.smin.i32(i32 range(i32 -893, 893) %81, i32 127)
-  %83 = tail call i32 @llvm.smax.i32(i32 %82, i32 -128)
+  %82 = tail call i32 @llvm.smax.i32(i32 range(i32 -893, 893) %81, i32 -128)
+  %83 = tail call i32 @llvm.smin.i32(i32 %82, i32 127)
   %84 = trunc nsw i32 %83 to i8
   %85 = and i8 %0, %84
   %86 = sext i8 %85 to i32
@@ -750,8 +750,8 @@ define internal fastcc void @filter8(i8 noundef signext %0, i8 noundef zeroext %
   %104 = sub nsw i32 %102, %103
   %105 = mul nsw i32 %104, 3
   %106 = add nsw i32 %101, %105
-  %107 = tail call i32 @llvm.smin.i32(i32 range(i32 -893, 893) %106, i32 127)
-  %108 = tail call i32 @llvm.smax.i32(i32 %107, i32 -128)
+  %107 = tail call i32 @llvm.smax.i32(i32 range(i32 -893, 893) %106, i32 -128)
+  %108 = tail call i32 @llvm.smin.i32(i32 %107, i32 127)
   %109 = trunc nsw i32 %108 to i8
   %110 = and i8 %0, %109
   %111 = sext i8 %110 to i32
@@ -1422,34 +1422,34 @@ define internal fastcc void @highbd_filter4(i8 noundef signext %0, i8 noundef ze
   ]
 
 signed_char_clamp_high.exit.thread:               ; preds = %7
-  %40 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %34, i32 511)
-  %41 = tail call i32 @llvm.smax.i32(i32 %40, i32 -512)
+  %40 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %34, i32 -512)
+  %41 = tail call i32 @llvm.smin.i32(i32 %40, i32 511)
   %42 = select i1 %.not112, i32 0, i32 %41
   %43 = add nsw i32 %42, %38
-  %44 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %43, i32 511)
-  %45 = tail call i32 @llvm.smax.i32(i32 %44, i32 -512)
+  %44 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %43, i32 -512)
+  %45 = tail call i32 @llvm.smin.i32(i32 %44, i32 511)
   %46 = and i32 %45, %39
-  %47 = tail call i32 @llvm.smin.i32(i32 %46, i32 507)
-  %48 = tail call i32 @llvm.smax.i32(i32 %47, i32 -516)
+  %47 = tail call i32 @llvm.smax.i32(i32 %46, i32 -516)
+  %48 = tail call i32 @llvm.smin.i32(i32 %47, i32 507)
   %49 = trunc nsw i32 %48 to i16
   %.0.i5381 = add nsw i16 %49, 4
   %50 = ashr i16 %.0.i5381, 3
-  %51 = tail call i32 @llvm.smin.i32(i32 %46, i32 508)
-  %52 = tail call i32 @llvm.smax.i32(i32 %51, i32 -515)
+  %51 = tail call i32 @llvm.smax.i32(i32 %46, i32 -515)
+  %52 = tail call i32 @llvm.smin.i32(i32 %51, i32 508)
   %53 = trunc nsw i32 %52 to i16
   %.0.i5686 = add nsw i16 %53, 3
   %54 = ashr i16 %.0.i5686, 3
   %55 = sext i16 %50 to i32
   %56 = sub nsw i32 %35, %55
-  %57 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %56, i32 511)
-  %58 = tail call i32 @llvm.smax.i32(i32 %57, i32 -512)
+  %57 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %56, i32 -512)
+  %58 = tail call i32 @llvm.smin.i32(i32 %57, i32 511)
   %.0.i5991 = trunc nsw i32 %58 to i16
   %59 = add i16 %.0.i5991, %11
   store i16 %59, ptr %4, align 2
   %60 = sext i16 %54 to i32
   %61 = add nsw i32 %60, %36
-  %62 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %61, i32 511)
-  %63 = tail call i32 @llvm.smax.i32(i32 %62, i32 -512)
+  %62 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %61, i32 -512)
+  %63 = tail call i32 @llvm.smin.i32(i32 %62, i32 511)
   %.0.i6296 = trunc nsw i32 %63 to i16
   %64 = add i16 %.0.i6296, %11
   store i16 %64, ptr %3, align 2
@@ -1458,45 +1458,45 @@ signed_char_clamp_high.exit.thread:               ; preds = %7
   %67 = sext i16 %66 to i32
   %68 = select i1 %.not112, i32 %67, i32 0
   %69 = sub nsw i32 %33, %68
-  %70 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %69, i32 511)
-  %71 = tail call i32 @llvm.smax.i32(i32 %70, i32 -512)
+  %70 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %69, i32 -512)
+  %71 = tail call i32 @llvm.smin.i32(i32 %70, i32 511)
   %.0.i65101 = trunc nsw i32 %71 to i16
   %72 = add i16 %.0.i65101, %11
   store i16 %72, ptr %5, align 2
   %73 = add nsw i32 %68, %32
-  %74 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %73, i32 511)
-  %75 = tail call i32 @llvm.smax.i32(i32 %74, i32 -512)
+  %74 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %73, i32 -512)
+  %75 = tail call i32 @llvm.smin.i32(i32 %74, i32 511)
   br label %signed_char_clamp_high.exit69
 
 signed_char_clamp_high.exit.thread72:             ; preds = %7
-  %76 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %34, i32 2047)
-  %77 = tail call i32 @llvm.smax.i32(i32 %76, i32 -2048)
+  %76 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %34, i32 -2048)
+  %77 = tail call i32 @llvm.smin.i32(i32 %76, i32 2047)
   %78 = select i1 %.not112, i32 0, i32 %77
   %79 = add nsw i32 %78, %38
-  %80 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %79, i32 2047)
-  %81 = tail call i32 @llvm.smax.i32(i32 %80, i32 -2048)
+  %80 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %79, i32 -2048)
+  %81 = tail call i32 @llvm.smin.i32(i32 %80, i32 2047)
   %82 = and i32 %81, %39
-  %83 = tail call i32 @llvm.smin.i32(i32 %82, i32 2043)
-  %84 = tail call i32 @llvm.smax.i32(i32 %83, i32 -2052)
+  %83 = tail call i32 @llvm.smax.i32(i32 %82, i32 -2052)
+  %84 = tail call i32 @llvm.smin.i32(i32 %83, i32 2043)
   %85 = trunc nsw i32 %84 to i16
   %.0.i5384 = add nsw i16 %85, 4
   %86 = ashr i16 %.0.i5384, 3
-  %87 = tail call i32 @llvm.smin.i32(i32 %82, i32 2044)
-  %88 = tail call i32 @llvm.smax.i32(i32 %87, i32 -2051)
+  %87 = tail call i32 @llvm.smax.i32(i32 %82, i32 -2051)
+  %88 = tail call i32 @llvm.smin.i32(i32 %87, i32 2044)
   %89 = trunc nsw i32 %88 to i16
   %.0.i5689 = add nsw i16 %89, 3
   %90 = ashr i16 %.0.i5689, 3
   %91 = sext i16 %86 to i32
   %92 = sub nsw i32 %35, %91
-  %93 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %92, i32 2047)
-  %94 = tail call i32 @llvm.smax.i32(i32 %93, i32 -2048)
+  %93 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %92, i32 -2048)
+  %94 = tail call i32 @llvm.smin.i32(i32 %93, i32 2047)
   %.0.i5994 = trunc nsw i32 %94 to i16
   %95 = add i16 %.0.i5994, %11
   store i16 %95, ptr %4, align 2
   %96 = sext i16 %90 to i32
   %97 = add nsw i32 %96, %36
-  %98 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %97, i32 2047)
-  %99 = tail call i32 @llvm.smax.i32(i32 %98, i32 -2048)
+  %98 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %97, i32 -2048)
+  %99 = tail call i32 @llvm.smin.i32(i32 %98, i32 2047)
   %.0.i6299 = trunc nsw i32 %99 to i16
   %100 = add i16 %.0.i6299, %11
   store i16 %100, ptr %3, align 2
@@ -1505,45 +1505,45 @@ signed_char_clamp_high.exit.thread72:             ; preds = %7
   %103 = sext i16 %102 to i32
   %104 = select i1 %.not112, i32 %103, i32 0
   %105 = sub nsw i32 %33, %104
-  %106 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %105, i32 2047)
-  %107 = tail call i32 @llvm.smax.i32(i32 %106, i32 -2048)
+  %106 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %105, i32 -2048)
+  %107 = tail call i32 @llvm.smin.i32(i32 %106, i32 2047)
   %.0.i65104 = trunc nsw i32 %107 to i16
   %108 = add i16 %.0.i65104, %11
   store i16 %108, ptr %5, align 2
   %109 = add nsw i32 %104, %32
-  %110 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %109, i32 2047)
-  %111 = tail call i32 @llvm.smax.i32(i32 %110, i32 -2048)
+  %110 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %109, i32 -2048)
+  %111 = tail call i32 @llvm.smin.i32(i32 %110, i32 2047)
   br label %signed_char_clamp_high.exit69
 
 signed_char_clamp_high.exit:                      ; preds = %7
-  %112 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %34, i32 127)
-  %113 = tail call i32 @llvm.smax.i32(i32 %112, i32 -128)
+  %112 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %34, i32 -128)
+  %113 = tail call i32 @llvm.smin.i32(i32 %112, i32 127)
   %114 = select i1 %.not112, i32 0, i32 %113
   %115 = add nsw i32 %114, %38
-  %116 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %115, i32 127)
-  %117 = tail call i32 @llvm.smax.i32(i32 %116, i32 -128)
+  %116 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %115, i32 -128)
+  %117 = tail call i32 @llvm.smin.i32(i32 %116, i32 127)
   %118 = and i32 %117, %39
-  %119 = tail call i32 @llvm.smin.i32(i32 %118, i32 123)
-  %120 = tail call i32 @llvm.smax.i32(i32 %119, i32 -132)
+  %119 = tail call i32 @llvm.smax.i32(i32 %118, i32 -132)
+  %120 = tail call i32 @llvm.smin.i32(i32 %119, i32 123)
   %121 = trunc nsw i32 %120 to i16
   %.0.i53 = add nsw i16 %121, 4
   %122 = ashr i16 %.0.i53, 3
-  %123 = tail call i32 @llvm.smin.i32(i32 %118, i32 124)
-  %124 = tail call i32 @llvm.smax.i32(i32 %123, i32 -131)
+  %123 = tail call i32 @llvm.smax.i32(i32 %118, i32 -131)
+  %124 = tail call i32 @llvm.smin.i32(i32 %123, i32 124)
   %125 = trunc nsw i32 %124 to i16
   %.0.i56 = add nsw i16 %125, 3
   %126 = ashr i16 %.0.i56, 3
   %127 = sext i16 %122 to i32
   %128 = sub nsw i32 %35, %127
-  %129 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %128, i32 127)
-  %130 = tail call i32 @llvm.smax.i32(i32 %129, i32 -128)
+  %129 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %128, i32 -128)
+  %130 = tail call i32 @llvm.smin.i32(i32 %129, i32 127)
   %.0.i59 = trunc nsw i32 %130 to i16
   %131 = add i16 %.0.i59, %11
   store i16 %131, ptr %4, align 2
   %132 = sext i16 %126 to i32
   %133 = add nsw i32 %132, %36
-  %134 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %133, i32 127)
-  %135 = tail call i32 @llvm.smax.i32(i32 %134, i32 -128)
+  %134 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %133, i32 -128)
+  %135 = tail call i32 @llvm.smin.i32(i32 %134, i32 127)
   %.0.i62 = trunc nsw i32 %135 to i16
   %136 = add i16 %.0.i62, %11
   store i16 %136, ptr %3, align 2
@@ -1552,14 +1552,14 @@ signed_char_clamp_high.exit:                      ; preds = %7
   %139 = sext i16 %138 to i32
   %140 = select i1 %.not112, i32 %139, i32 0
   %141 = sub nsw i32 %33, %140
-  %142 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %141, i32 127)
-  %143 = tail call i32 @llvm.smax.i32(i32 %142, i32 -128)
+  %142 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %141, i32 -128)
+  %143 = tail call i32 @llvm.smin.i32(i32 %142, i32 127)
   %.0.i65 = trunc nsw i32 %143 to i16
   %144 = add i16 %.0.i65, %11
   store i16 %144, ptr %5, align 2
   %145 = add nsw i32 %140, %32
-  %146 = tail call i32 @llvm.smin.i32(i32 range(i32 -229373, 229373) %145, i32 127)
-  %147 = tail call i32 @llvm.smax.i32(i32 %146, i32 -128)
+  %146 = tail call i32 @llvm.smax.i32(i32 range(i32 -229373, 229373) %145, i32 -128)
+  %147 = tail call i32 @llvm.smin.i32(i32 %146, i32 127)
   br label %signed_char_clamp_high.exit69
 
 signed_char_clamp_high.exit69:                    ; preds = %signed_char_clamp_high.exit.thread, %signed_char_clamp_high.exit.thread72, %signed_char_clamp_high.exit
diff --git a/bench/openusd/optimized/reformat.ll b/bench/openusd/optimized/reformat.ll
index ffde1cfa043..f7fece87371 100644
--- a/bench/openusd/optimized/reformat.ll
+++ b/bench/openusd/optimized/reformat.ll
@@ -1584,8 +1584,8 @@ avifPrepareReformatState.exit:                    ; preds = %96, %91
 .thread:                                          ; preds = %116, %117, %108, %avifPrepareReformatState.exit
   %.075 = phi i32 [ 0, %avifPrepareReformatState.exit ], [ %spec.select, %108 ], [ %spec.select101, %117 ], [ %spec.select113, %116 ]
   %118 = load i32, ptr %9, align 8
-  %spec.select102 = call i32 @llvm.smin.i32(i32 %118, i32 8)
-  %119 = call i32 @llvm.smax.i32(i32 %spec.select102, i32 1)
+  %spec.select102 = call i32 @llvm.smax.i32(i32 %118, i32 1)
+  %119 = call i32 @llvm.umin.i32(i32 %spec.select102, i32 8)
   %120 = getelementptr inbounds nuw i8, ptr %0, i64 12
   %121 = load i32, ptr %120, align 4
   %122 = icmp eq i32 %121, 3
@@ -1594,7 +1594,7 @@ avifPrepareReformatState.exit:                    ; preds = %96, %91
 123:                                              ; preds = %.thread
   %124 = getelementptr inbounds nuw i8, ptr %1, i64 16
   %125 = load i32, ptr %124, align 8
-  switch i32 %125, label %126 [
+  switch i32 %125, label %127 [
     i32 0, label %.thread111
     i32 2, label %.thread111
     i32 4, label %.thread111
@@ -1633,50 +1633,50 @@ avifPrepareReformatState.exit:                    ; preds = %96, %91
   %145 = getelementptr inbounds nuw i8, ptr %5, i64 12
   %146 = getelementptr inbounds nuw i8, ptr %1, i64 56
   %147 = zext nneg i32 %140 to i64
-  %wide.trip.count = zext nneg i32 %spec.select102 to i64
+  %wide.trip.count = zext nneg i32 %120 to i64
   %148 = zext i32 %spec.select103 to i64
   br label %149
 
 149:                                              ; preds = %136, %176
-  %indvars.iv128 = phi i64 [ 0, %136 ], [ %indvars.iv.next129, %176 ]
-  %indvars.iv = phi i64 [ 0, %136 ], [ %indvars.iv.next, %176 ]
+  %indvars.iv128 = phi i64 [ 0, %137 ], [ %indvars.iv.next129, %177 ]
+  %indvars.iv = phi i64 [ 0, %137 ], [ %indvars.iv.next, %177 ]
   %150 = load ptr, ptr %4, align 8
   %151 = getelementptr inbounds nuw %struct.YUVToRGBThreadData, ptr %150, i64 %indvars.iv
   store i32 0, ptr %5, align 4
   %152 = trunc nuw i64 %indvars.iv128 to i32
-  store i32 %152, ptr %143, align 4
-  %153 = load i32, ptr %0, align 8
   store i32 %153, ptr %144, align 4
-  %154 = icmp eq i64 %indvars.iv, %147
-  %155 = select i1 %154, i32 %142, i32 %spec.select103
-  store i32 %155, ptr %145, align 4
-  %156 = getelementptr inbounds nuw i8, ptr %151, i64 8
-  %157 = call i32 @avifImageSetViewRect(ptr noundef nonnull %156, ptr noundef nonnull %0, ptr noundef nonnull %5) #10
-  %.not95 = icmp eq i32 %157, 0
-  br i1 %.not95, label %158, label %.loopexit.sink.split
-
-158:                                              ; preds = %149
-  %159 = getelementptr inbounds nuw i8, ptr %151, i64 208
-  call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(64) %159, ptr noundef nonnull align 8 dereferenceable(64) %1, i64 64, i1 false)
-  %160 = load i32, ptr %146, align 8
-  %161 = zext i32 %160 to i64
-  %162 = mul nuw nsw i64 %indvars.iv128, %161
-  %163 = getelementptr inbounds nuw i8, ptr %151, i64 256
-  %164 = load ptr, ptr %163, align 8
-  %165 = getelementptr inbounds nuw i8, ptr %164, i64 %162
-  store ptr %165, ptr %163, align 8
-  %166 = getelementptr inbounds nuw i8, ptr %151, i64 12
-  %167 = load i32, ptr %166, align 4
-  %168 = getelementptr inbounds nuw i8, ptr %151, i64 212
-  store i32 %167, ptr %168, align 4
-  %169 = getelementptr inbounds nuw i8, ptr %151, i64 272
-  store ptr %3, ptr %169, align 8
-  %170 = getelementptr inbounds nuw i8, ptr %151, i64 280
-  store i32 %.075, ptr %170, align 8
+  %154 = load i32, ptr %0, align 8
+  store i32 %154, ptr %145, align 4
+  %155 = icmp eq i64 %indvars.iv, %148
+  %156 = select i1 %155, i32 %143, i32 %spec.select103
+  store i32 %156, ptr %146, align 4
+  %157 = getelementptr inbounds nuw i8, ptr %152, i64 8
+  %158 = call i32 @avifImageSetViewRect(ptr noundef nonnull %157, ptr noundef nonnull %0, ptr noundef nonnull %5) #10
+  %.not95 = icmp eq i32 %158, 0
+  br i1 %.not95, label %159, label %.loopexit
+
+159:      ; preds = %150
+  %160 = getelementptr inbounds nuw i8, ptr %152, i64 208
+  call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(64) %160, ptr noundef nonnull align 8 dereferenceable(64) %1, i64 64, i1 false)
+  %161 = load i32, ptr %147, align 8
+  %162 = zext i32 %161 to i64
+  %163 = mul nuw nsw i64 %indvars.iv128, %162
+  %165 = getelementptr inbounds nuw i8, ptr %152, i64 256
+  %165 = load ptr, ptr %164, align 8
+  %166 = getelementptr inbounds nuw i8, ptr %165, i64 %163
+  store ptr %166, ptr %164, align 8
+  %168 = getelementptr inbounds nuw i8, ptr %151, i64 12
+  %168 = load i32, ptr %168, align 4
+  %169 = getelementptr inbounds nuw i8, ptr %151, i64 212
+  store i32 %168, ptr %169, align 4
+  %170 = getelementptr inbounds nuw i8, ptr %151, i64 272
+  store ptr %3, ptr %170, align 8
+  %171 = getelementptr inbounds nuw i8, ptr %152, i64 280
+  store i32 %.075, ptr %171, align 8
   %.not96 = icmp eq i64 %indvars.iv, 0
   br i1 %.not96, label %176, label %171
 
-171:                                              ; preds = %158
+171:                                              ; preds = %159
   %172 = call i32 @pthread_create(ptr noundef nonnull %151, ptr noundef null, ptr noundef nonnull @avifImageYUVToRGBThreadWorker, ptr noundef nonnull %151) #10
   %173 = icmp eq i32 %172, 0
   %174 = zext i1 %173 to i32
@@ -1684,24 +1684,20 @@ avifPrepareReformatState.exit:                    ; preds = %96, %91
   store i32 %174, ptr %175, align 8
   br i1 %173, label %176, label %.loopexit.sink.split
 
-176:                                              ; preds = %158, %171
+176:                                              ; preds = %159, %171
   %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
   %indvars.iv.next129 = add nuw nsw i64 %indvars.iv128, %148
   %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count
-  br i1 %exitcond.not, label %.loopexit, label %149, !llvm.loop !14
+  br i1 %exitcond.not, label %.loopexit.thread, label %149, !llvm.loop !14
 
 .loopexit.sink.split:                             ; preds = %171, %149
   %177 = trunc nuw nsw i64 %indvars.iv to i32
   %178 = getelementptr inbounds nuw i8, ptr %151, i64 284
   store i32 5, ptr %178, align 4
-  br label %.loopexit
-
-.loopexit:                                        ; preds = %176, %.loopexit.sink.split
-  %.076117 = phi i32 [ %177, %.loopexit.sink.split ], [ %spec.select102, %176 ]
-  %179 = icmp eq i32 %.076117, %119
-  br i1 %179, label %180, label %190
+  %180 = icmp eq i32 %120, %178
+  br i1 %180, label %.loopexit.thread, label %.preheader
 
-180:                                              ; preds = %.loopexit
+180:                                              ; preds = %177, %.loopexit
   %181 = load ptr, ptr %4, align 8
   %182 = getelementptr inbounds nuw i8, ptr %181, i64 8
   %183 = getelementptr inbounds nuw i8, ptr %181, i64 208
@@ -1712,45 +1708,44 @@ avifPrepareReformatState.exit:                    ; preds = %96, %91
   %188 = call fastcc i32 @avifImageYUVToRGBImpl(ptr noundef nonnull %182, ptr noundef nonnull %183, ptr noundef %185, i32 noundef %187)
   %189 = getelementptr inbounds nuw i8, ptr %181, i64 284
   store i32 %188, ptr %189, align 4
+  br label %.preheader
+
+.preheader:                                       ; preds = %.loopexit.thread, %.loopexit
   br label %190
 
-190:                                              ; preds = %180, %.loopexit
-  %wide.trip.count134 = zext nneg i32 %119 to i64
-  br label %191
-
-191:                                              ; preds = %190, %198
-  %indvars.iv131 = phi i64 [ 0, %190 ], [ %indvars.iv.next132, %198 ]
-  %.074122 = phi i32 [ 0, %190 ], [ %spec.select105, %198 ]
-  %192 = load ptr, ptr %4, align 8
-  %193 = getelementptr inbounds nuw %struct.YUVToRGBThreadData, ptr %192, i64 %indvars.iv131
-  %194 = getelementptr inbounds nuw i8, ptr %193, i64 288
-  %195 = load i32, ptr %194, align 8
-  %.not98 = icmp eq i32 %195, 0
-  br i1 %.not98, label %198, label %196
-
-196:                                              ; preds = %191
-  %.val = load i64, ptr %193, align 8
-  %197 = call i32 @pthread_join(i64 noundef %.val, ptr noundef null) #10
-  %.not114 = icmp eq i32 %197, 0
+190:                                              ; preds = %.preheader, %197
+  %indvars.iv131 = phi i64 [ %indvars.iv.next132, %197 ], [ 0, %.preheader ]
+  %.074122 = phi i32 [ %spec.select105, %197 ], [ 0, %.preheader ]
+  %191 = load ptr, ptr %4, align 8
+  %192 = getelementptr inbounds nuw %struct.YUVToRGBThreadData, ptr %191, i64 %indvars.iv131
+  %193 = getelementptr inbounds nuw i8, ptr %192, i64 288
+  %194 = load i32, ptr %193, align 8
+  %.not98 = icmp eq i32 %194, 0
+  br i1 %.not98, label %197, label %195
+
+195:   ; preds = %190
+  %.val = load i64, ptr %192, align 8
+  %196 = call i32 @pthread_join(i64 noundef %.val, ptr noundef null) #10
+  %.not114 = icmp eq i32 %196, 0
   %spec.select104 = select i1 %.not114, i32 %.074122, i32 5
   br label %198
 
-198:                                              ; preds = %196, %191
-  %.1 = phi i32 [ %.074122, %191 ], [ %spec.select104, %196 ]
-  %199 = getelementptr inbounds nuw i8, ptr %193, i64 284
+198:                                              ; preds = %195, %190
+  %.1 = phi i32 [ %.074122, %190 ], [ %spec.select104, %195 ]
+  %199 = getelementptr inbounds nuw i8, ptr %192, i64 284
   %200 = load i32, ptr %199, align 4
   %.not100 = icmp eq i32 %200, 0
   %spec.select105 = select i1 %.not100, i32 %.1, i32 %200
   %indvars.iv.next132 = add nuw nsw i64 %indvars.iv131, 1
-  %exitcond135.not = icmp eq i64 %indvars.iv.next132, %wide.trip.count134
-  br i1 %exitcond135.not, label %201, label %191, !llvm.loop !15
+  %exitcond135.not = icmp eq i64 %indvars.iv.next132, %wide.trip.count
+  br i1 %exitcond135.not, label %201, label %190, !llvm.loop !15
 
 201:                                              ; preds = %198
   call void @avifArrayDestroy(ptr noundef nonnull %4) #10
   br label %avifPrepareReformatState.exit.thread
 
 avifPrepareReformatState.exit.thread:             ; preds = %.thread.i, %.thread78.i, %27, %26, %18, %12, %83, %134, %2, %8, %201, %.thread111
-  %.0 = phi i32 [ %133, %.thread111 ], [ %spec.select105, %201 ], [ 5, %8 ], [ 5, %2 ], [ 26, %134 ], [ 5, %83 ], [ 5, %12 ], [ 5, %18 ], [ 5, %26 ], [ 5, %27 ], [ 5, %.thread78.i ], [ 5, %.thread.i ]
+  %.0 = phi i32 [ %133, %.thread111 ], [ %spec.select105, %200 ], [ 5, %8 ], [ 5, %2 ], [ 26, %135 ], [ 5, %83 ], [ 5, %12 ], [ 5, %18 ], [ 5, %26 ], [ 5, %27 ], [ 5, %.thread78.i ], [ 5, %.thread.i ]
   ret i32 %.0
 }
 
@@ -5269,6 +5264,9 @@ declare void @llvm.lifetime.end.p0(i64 immarg, ptr captures(none)) #9
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smax.i32(i32, i32) #7
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #7
+
 attributes #0 = { nounwind uwtable "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #2 = { mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none) }
diff --git a/bench/openusd/optimized/warped_motion.ll b/bench/openusd/optimized/warped_motion.ll
index 1ee0072a9e6..3c250520a27 100644
--- a/bench/openusd/optimized/warped_motion.ll
+++ b/bench/openusd/optimized/warped_motion.ll
@@ -17,13 +17,13 @@ define hidden range(i32 0, 2) i32 @av1_get_shear_params(ptr noundef captures(non
   br i1 %3, label %137, label %4
 
 4:                                                ; preds = %1
-  %5 = tail call i32 @llvm.umin.i32(i32 %.val, i32 98303)
-  %6 = tail call i32 @llvm.umax.i32(i32 %5, i32 32768)
+  %5 = tail call i32 @llvm.umax.i32(i32 %.val, i32 32768)
+  %6 = tail call i32 @llvm.umin.i32(i32 %5, i32 98303)
   %7 = getelementptr inbounds nuw i8, ptr %0, i64 32
   %8 = getelementptr inbounds nuw i8, ptr %0, i64 12
   %9 = load i32, ptr %8, align 4
-  %10 = tail call i32 @llvm.smin.i32(i32 %9, i32 32767)
-  %11 = tail call i32 @llvm.smax.i32(i32 %10, i32 -32768)
+  %10 = tail call i32 @llvm.smax.i32(i32 %9, i32 -32768)
+  %11 = tail call i32 @llvm.smin.i32(i32 %10, i32 32767)
   %12 = getelementptr inbounds nuw i8, ptr %0, i64 34
   %13 = tail call range(i32 0, 33) i32 @llvm.ctlz.i32(i32 range(i32 0, -2147483648) %.val, i1 true)
   %14 = xor i32 %13, 31
@@ -78,8 +78,8 @@ resolve_divisor_32.exit:                          ; preds = %17, %23
 48:                                               ; preds = %45, %41
   %49 = phi i64 [ %44, %41 ], [ %47, %45 ]
   %50 = trunc i64 %49 to i32
-  %51 = tail call i32 @llvm.smin.i32(i32 %50, i32 32767)
-  %52 = tail call i32 @llvm.smax.i32(i32 %51, i32 -32768)
+  %51 = tail call i32 @llvm.smax.i32(i32 %50, i32 -32768)
+  %52 = tail call i32 @llvm.smin.i32(i32 %51, i32 32767)
   %53 = getelementptr inbounds nuw i8, ptr %0, i64 36
   %54 = sext i32 %9 to i64
   %55 = mul nsw i64 %35, %54
@@ -107,8 +107,8 @@ resolve_divisor_32.exit:                          ; preds = %17, %23
   %71 = phi i64 [ %66, %63 ], [ %69, %67 ]
   %72 = trunc i64 %71 to i32
   %73 = sub nsw i32 %58, %72
-  %74 = tail call i32 @llvm.smin.i32(i32 %73, i32 98303)
-  %75 = tail call i32 @llvm.smax.i32(i32 %74, i32 32768)
+  %74 = tail call i32 @llvm.smax.i32(i32 %73, i32 32768)
+  %75 = tail call i32 @llvm.umin.i32(i32 %74, i32 98303)
   %76 = getelementptr inbounds nuw i8, ptr %0, i64 38
   %sext = shl i32 %6, 16
   %77 = ashr exact i32 %sext, 16
@@ -1843,8 +1843,8 @@ resolve_divisor_64.exit.i:                        ; preds = %133, %123
 
 get_mult_shift_diag.exit.i:                       ; preds = %171, %167
   %174 = phi i64 [ %170, %167 ], [ %173, %171 ]
-  %175 = tail call i64 @llvm.smin.i64(i64 range(i64 -9223372036854775807, -9223372036854775808) %174, i64 73727)
-  %176 = tail call i64 @llvm.smax.i64(i64 %175, i64 57345)
+  %175 = tail call i64 @llvm.smax.i64(i64 %174, i64 57345)
+  %176 = tail call i64 @llvm.umin.i64(i64 %175, i64 73727)
   %177 = trunc nuw nsw i64 %176 to i32
   %178 = getelementptr inbounds nuw i8, ptr %6, i64 8
   store i32 %177, ptr %178, align 4
@@ -1865,8 +1865,8 @@ get_mult_shift_diag.exit.i:                       ; preds = %171, %167
 
 get_mult_shift_ndiag.exit.i:                      ; preds = %185, %181
   %188 = phi i64 [ %184, %181 ], [ %187, %185 ]
-  %189 = tail call i64 @llvm.smin.i64(i64 range(i64 -9223372036854775807, -9223372036854775808) %188, i64 8191)
-  %190 = tail call i64 @llvm.smax.i64(i64 %189, i64 -8191)
+  %189 = tail call i64 @llvm.smax.i64(i64 %188, i64 -8191)
+  %190 = tail call i64 @llvm.smin.i64(i64 %189, i64 8191)
   %191 = trunc nsw i64 %190 to i32
   %192 = getelementptr inbounds nuw i8, ptr %6, i64 12
   store i32 %191, ptr %192, align 4
@@ -1887,8 +1887,8 @@ get_mult_shift_ndiag.exit.i:                      ; preds = %185, %181
 
 get_mult_shift_ndiag.exit131.i:                   ; preds = %199, %195
   %202 = phi i64 [ %198, %195 ], [ %201, %199 ]
-  %203 = tail call i64 @llvm.smin.i64(i64 range(i64 -9223372036854775807, -9223372036854775808) %202, i64 8191)
-  %204 = tail call i64 @llvm.smax.i64(i64 %203, i64 -8191)
+  %203 = tail call i64 @llvm.smax.i64(i64 %202, i64 -8191)
+  %204 = tail call i64 @llvm.smin.i64(i64 %203, i64 8191)
   %205 = trunc nsw i64 %204 to i32
   %206 = getelementptr inbounds nuw i8, ptr %6, i64 16
   store i32 %205, ptr %206, align 4
@@ -1909,8 +1909,8 @@ get_mult_shift_ndiag.exit131.i:                   ; preds = %199, %195
 
 216:                                              ; preds = %209, %213
   %217 = phi i64 [ %212, %209 ], [ %215, %213 ]
-  %218 = tail call i64 @llvm.smin.i64(i64 range(i64 -9223372036854775807, -9223372036854775808) %217, i64 73727)
-  %219 = tail call i64 @llvm.smax.i64(i64 %218, i64 57345)
+  %218 = tail call i64 @llvm.smax.i64(i64 %217, i64 57345)
+  %219 = tail call i64 @llvm.umin.i64(i64 %218, i64 73727)
   %220 = trunc nuw nsw i64 %219 to i32
   %221 = getelementptr inbounds nuw i8, ptr %6, i64 20
   store i32 %220, ptr %221, align 4
@@ -1930,11 +1930,11 @@ get_mult_shift_ndiag.exit131.i:                   ; preds = %199, %195
   %.neg140.neg.i = mul i32 %.neg.i, %223
   %reass.add142.neg.i = sub i32 %228, %.neg139.i
   %229 = add i32 %reass.add142.neg.i, %.neg140.neg.i
-  %230 = tail call i32 @llvm.smin.i32(i32 %227, i32 8388607)
-  %231 = tail call i32 @llvm.smax.i32(i32 %230, i32 -8388608)
+  %230 = tail call i32 @llvm.smax.i32(i32 %227, i32 -8388608)
+  %231 = tail call i32 @llvm.smin.i32(i32 %230, i32 8388607)
   store i32 %231, ptr %6, align 4
-  %232 = tail call i32 @llvm.smin.i32(i32 %229, i32 8388607)
-  %233 = tail call i32 @llvm.smax.i32(i32 %232, i32 -8388608)
+  %232 = tail call i32 @llvm.smax.i32(i32 %229, i32 -8388608)
+  %233 = tail call i32 @llvm.smin.i32(i32 %232, i32 8388607)
   %234 = getelementptr inbounds nuw i8, ptr %6, i64 4
   store i32 %233, ptr %234, align 4
   %235 = getelementptr inbounds nuw i8, ptr %6, i64 28
@@ -1969,14 +1969,17 @@ declare i32 @llvm.smax.i32(i32, i32) #6
 declare i64 @llvm.smin.i64(i64, i64) #6
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.umin.i32(i32, i32) #6
+declare i32 @llvm.umax.i32(i32, i32) #6
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.umax.i32(i32, i32) #6
+declare i32 @llvm.umin.i32(i32, i32) #6
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.smax.i64(i64, i64) #6
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i64 @llvm.umin.i64(i64, i64) #6
+
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i16 @llvm.usub.sat.i16(i16, i16) #6
 
diff --git a/bench/openusd/optimized/write.ll b/bench/openusd/optimized/write.ll
index da7038b35b9..0c02c48b772 100644
--- a/bench/openusd/optimized/write.ll
+++ b/bench/openusd/optimized/write.ll
@@ -624,14 +624,13 @@ avifValidateImageBasicProperties.exit:            ; preds = %27
   br label %avifQualityToQuantizer.exit
 
 76:                                               ; preds = %58
-  %77 = tail call i32 @llvm.smin.i32(i32 %64, i32 100)
-  %78 = tail call i32 @llvm.smax.i32(i32 %77, i32 0)
+  %77 = tail call i32 @llvm.smax.i32(i32 %64, i32 0)
+  %78 = tail call i32 @llvm.usub.sat.i32(i32 100, i32 %77)
   %79 = trunc nuw nsw i32 %78 to i16
-  %80 = sub nuw nsw i16 100, %79
-  %81 = mul nuw nsw i16 %80, 63
-  %.lhs.trunc.i = add nuw nsw i16 %81, 50
-  %82 = udiv i16 %.lhs.trunc.i, 100
-  %.zext.i = zext nneg i16 %82 to i32
+  %80 = mul nuw nsw i16 %79, 63
+  %.lhs.trunc.i = add nuw nsw i16 %80, 50
+  %81 = udiv i16 %.lhs.trunc.i, 100
+  %.zext.i = zext nneg i16 %81 to i32
   br label %avifQualityToQuantizer.exit
 
 avifQualityToQuantizer.exit:                      ; preds = %66, %76
@@ -657,32 +656,31 @@ avifQualityToQuantizer.exit:                      ; preds = %66, %76
   br label %avifQualityToQuantizer.exit254
 
 98:                                               ; preds = %avifQualityToQuantizer.exit
-  %99 = tail call i32 @llvm.smin.i32(i32 %86, i32 100)
-  %100 = tail call i32 @llvm.smax.i32(i32 %99, i32 0)
+  %99 = tail call i32 @llvm.smax.i32(i32 %86, i32 0)
+  %100 = tail call i32 @llvm.usub.sat.i32(i32 100, i32 %98)
   %101 = trunc nuw nsw i32 %100 to i16
-  %102 = sub nuw nsw i16 100, %101
-  %103 = mul nuw nsw i16 %102, 63
-  %.lhs.trunc.i251 = add nuw nsw i16 %103, 50
-  %104 = udiv i16 %.lhs.trunc.i251, 100
-  %.zext.i252 = zext nneg i16 %104 to i32
+  %101 = mul nuw nsw i16 %100, 63
+  %.lhs.trunc.i251 = add nuw nsw i16 %101, 50
+  %102 = udiv i16 %.lhs.trunc.i251, 100
+  %.zext.i252 = zext nneg i16 %102 to i32
   br label %avifQualityToQuantizer.exit254
 
 avifQualityToQuantizer.exit254:                   ; preds = %88, %98
-  %.0.i253 = phi i32 [ %97, %88 ], [ %.zext.i252, %98 ]
+  %.0.i253 = phi i32 [ %97, %87 ], [ %.zext.i252, %97 ]
   %105 = load ptr, ptr %44, align 8
   %106 = getelementptr inbounds nuw i8, ptr %105, i64 52
   store i32 %.0.i253, ptr %106, align 4
   %107 = getelementptr inbounds nuw i8, ptr %0, i64 56
   %108 = load i32, ptr %107, align 8
-  %spec.select = tail call i32 @llvm.smin.i32(i32 %108, i32 6)
-  %109 = tail call i32 @llvm.smax.i32(i32 %spec.select, i32 0)
+  %spec.select = tail call i32 @llvm.smax.i32(i32 %108, i32 0)
+  %109 = tail call i32 @llvm.umin.i32(i32 %spec.select, i32 6)
   %110 = load ptr, ptr %44, align 8
   %111 = getelementptr inbounds nuw i8, ptr %110, i64 56
   store i32 %109, ptr %111, align 8
   %112 = getelementptr inbounds nuw i8, ptr %0, i64 60
   %113 = load i32, ptr %112, align 4
-  %spec.select244 = tail call i32 @llvm.smin.i32(i32 %113, i32 6)
-  %114 = tail call i32 @llvm.smax.i32(i32 %spec.select244, i32 0)
+  %spec.select244 = tail call i32 @llvm.smax.i32(i32 %113, i32 0)
+  %114 = tail call i32 @llvm.umin.i32(i32 %spec.select244, i32 6)
   %115 = load ptr, ptr %44, align 8
   %116 = getelementptr inbounds nuw i8, ptr %115, i64 60
   store i32 %114, ptr %116, align 4
@@ -5976,6 +5974,9 @@ declare i32 @llvm.smin.i32(i32, i32) #9
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.ctlz.i32(i32, i1 immarg) #9
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.usub.sat.i32(i32, i32) #9
+
 ; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
 declare void @llvm.lifetime.start.p0(i64 immarg, ptr captures(none)) #11
 
diff --git a/bench/pbrt-v4/optimized/parallel.ll b/bench/pbrt-v4/optimized/parallel.ll
index 4a77bafcd7c..3b5311f9702 100644
--- a/bench/pbrt-v4/optimized/parallel.ll
+++ b/bench/pbrt-v4/optimized/parallel.ll
@@ -2776,8 +2776,8 @@ _ZN4pbrt14RunningThreadsEv.exit:                  ; preds = %18
   %40 = sitofp i32 %39 to double
   %41 = tail call noundef double @sqrt(double noundef %40) #25, !tbaa !37
   %42 = fptosi double %41 to i32
-  %..i = tail call i32 @llvm.smin.i32(i32 %42, i32 32)
-  %.0.i = tail call i32 @llvm.smax.i32(i32 %..i, i32 1)
+  %..i = tail call i32 @llvm.smax.i32(i32 %42, i32 1)
+  %.0.i = tail call i32 @llvm.umin.i32(i32 %..i, i32 32)
   call void @llvm.lifetime.start.p0(i64 104, ptr nonnull %4) #25
   %43 = getelementptr inbounds nuw i8, ptr %1, i64 24
   call void @llvm.memset.p0.i64(ptr noundef nonnull align 8 dereferenceable(24) %.sroa.0, i8 0, i64 24, i1 false)
@@ -2859,8 +2859,8 @@ _ZN4pbrt10ThreadPool12AddToJobListEPNS_11ParallelJobE.exit: ; preds = %_ZNSt11un
   %69 = load ptr, ptr %4, align 8, !tbaa !55
   %70 = getelementptr inbounds nuw i8, ptr %69, i64 16
   %71 = load ptr, ptr %70, align 8
-  %72 = invoke noundef zeroext i1 %71(ptr noundef nonnull align 8 dereferenceable(33) %4)
-          to label %73 unwind label %80
+  %72 = invoke noundef zeroext i1 %73(ptr noundef nonnull align 8 dereferenceable(33) %4)
+          to label %75 unwind label %80
 
 73:                                               ; preds = %68
   %74 = load i32, ptr %48, align 8
@@ -2871,7 +2871,7 @@ _ZN4pbrt10ThreadPool12AddToJobListEPNS_11ParallelJobE.exit: ; preds = %_ZNSt11un
 76:                                               ; preds = %73
   %77 = load ptr, ptr @_ZN4pbrt11ParallelJob10threadPoolE, align 8, !tbaa !122
   invoke void @_ZN4pbrt10ThreadPool10WorkOrWaitEPSt11unique_lockISt5mutexEb(ptr noundef nonnull align 8 dereferenceable(128) %77, ptr noundef nonnull %5, i1 noundef zeroext true)
-          to label %68 unwind label %80, !llvm.loop !150
+          to label %70 unwind label %80, !llvm.loop !150
 
 78:                                               ; preds = %60
   %79 = landingpad { ptr, i32 }
@@ -2917,7 +2917,7 @@ _ZNSt11unique_lockISt5mutexED2Ev.exit30:          ; preds = %88, %91, %93
   br i1 %.not.i.i31, label %_ZN4pbrt17ParallelForLoop2DD2Ev.exit, label %97
 
 97:                                               ; preds = %_ZNSt11unique_lockISt5mutexED2Ev.exit30
-  %98 = invoke noundef zeroext i1 %96(ptr noundef nonnull align 8 dereferenceable(32) %50, ptr noundef nonnull align 8 dereferenceable(32) %50, i32 noundef 3)
+  %98 = invoke noundef zeroext i1 %98(ptr noundef nonnull align 8 dereferenceable(32) %50, ptr noundef nonnull align 8 dereferenceable(32) %50, i32 noundef 3)
           to label %_ZN4pbrt17ParallelForLoop2DD2Ev.exit unwind label %99
 
 99:                                               ; preds = %97
@@ -2935,7 +2935,7 @@ _ZN4pbrt17ParallelForLoop2DD2Ev.exit:             ; preds = %_ZNSt11unique_lockI
   ret void
 
 _ZNSt11unique_lockISt5mutexED2Ev.exit:            ; preds = %86, %84, %80, %78
-  %.pn = phi { ptr, i32 } [ %79, %78 ], [ %81, %80 ], [ %81, %84 ], [ %81, %86 ]
+  %.pn = phi { ptr, i32 } [ %81, %80 ], [ %81, %82 ], [ %81, %86 ], [ %83, %88 ]
   call void @llvm.lifetime.end.p0(i64 16, ptr nonnull %5) #25
   call void @_ZN4pbrt17ParallelForLoop2DD2Ev(ptr noundef nonnull align 8 dereferenceable(100) %4) #25
   call void @llvm.lifetime.end.p0(i64 104, ptr nonnull %4) #25
@@ -16960,6 +16960,9 @@ declare i32 @llvm.smin.i32(i32, i32) #23
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smax.i32(i32, i32) #23
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #23
+
 attributes #0 = { mustprogress uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="rocketlake" "target-features"="+64bit,+adx,+aes,+avx,+avx2,+avx512bitalg,+avx512bw,+avx512cd,+avx512dq,+avx512f,+avx512ifma,+avx512vbmi,+avx512vbmi2,+avx512vl,+avx512vnni,+avx512vpopcntdq,+bmi,+bmi2,+clflushopt,+cmov,+crc32,+cx16,+cx8,+evex512,+f16c,+fma,+fsgsbase,+fxsr,+gfni,+invpcid,+lzcnt,+mmx,+movbe,+pclmul,+pku,+popcnt,+prfchw,+rdpid,+rdrnd,+rdseed,+sahf,+sha,+sse,+sse2,+sse3,+sse4.1,+sse4.2,+ssse3,+vaes,+vpclmulqdq,+x87,+xsave,+xsavec,+xsaveopt,+xsaves,-amx-avx512,-amx-bf16,-amx-complex,-amx-fp16,-amx-fp8,-amx-int8,-amx-movrs,-amx-tf32,-amx-tile,-amx-transpose,-avx10.1-256,-avx10.1-512,-avx10.2-256,-avx10.2-512,-avx512bf16,-avx512fp16,-avx512vp2intersect,-avxifma,-avxneconvert,-avxvnni,-avxvnniint16,-avxvnniint8,-ccmp,-cf,-cldemote,-clwb,-clzero,-cmpccxadd,-egpr,-enqcmd,-fma4,-hreset,-kl,-lwp,-movdir64b,-movdiri,-movrs,-mwaitx,-ndd,-nf,-pconfig,-ppx,-prefetchi,-ptwrite,-push2pop2,-raoint,-rdpru,-rtm,-serialize,-sgx,-sha512,-shstk,-sm3,-sm4,-sse4a,-tbm,-tsxldtrk,-uintr,-usermsr,-waitpkg,-wbnoinvd,-widekl,-xop,-zu" }
 attributes #1 = { mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
 attributes #2 = { inlinehint mustprogress noreturn uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="rocketlake" "target-features"="+64bit,+adx,+aes,+avx,+avx2,+avx512bitalg,+avx512bw,+avx512cd,+avx512dq,+avx512f,+avx512ifma,+avx512vbmi,+avx512vbmi2,+avx512vl,+avx512vnni,+avx512vpopcntdq,+bmi,+bmi2,+clflushopt,+cmov,+crc32,+cx16,+cx8,+evex512,+f16c,+fma,+fsgsbase,+fxsr,+gfni,+invpcid,+lzcnt,+mmx,+movbe,+pclmul,+pku,+popcnt,+prfchw,+rdpid,+rdrnd,+rdseed,+sahf,+sha,+sse,+sse2,+sse3,+sse4.1,+sse4.2,+ssse3,+vaes,+vpclmulqdq,+x87,+xsave,+xsavec,+xsaveopt,+xsaves,-amx-avx512,-amx-bf16,-amx-complex,-amx-fp16,-amx-fp8,-amx-int8,-amx-movrs,-amx-tf32,-amx-tile,-amx-transpose,-avx10.1-256,-avx10.1-512,-avx10.2-256,-avx10.2-512,-avx512bf16,-avx512fp16,-avx512vp2intersect,-avxifma,-avxneconvert,-avxvnni,-avxvnniint16,-avxvnniint8,-ccmp,-cf,-cldemote,-clwb,-clzero,-cmpccxadd,-egpr,-enqcmd,-fma4,-hreset,-kl,-lwp,-movdir64b,-movdiri,-movrs,-mwaitx,-ndd,-nf,-pconfig,-ppx,-prefetchi,-ptwrite,-push2pop2,-raoint,-rdpru,-rtm,-serialize,-sgx,-sha512,-shstk,-sm3,-sm4,-sse4a,-tbm,-tsxldtrk,-uintr,-usermsr,-waitpkg,-wbnoinvd,-widekl,-xop,-zu" }
diff --git a/bench/quickjs/optimized/quickjs.ll b/bench/quickjs/optimized/quickjs.ll
index 4ff46016ab1..e532afa2001 100644
--- a/bench/quickjs/optimized/quickjs.ll
+++ b/bench/quickjs/optimized/quickjs.ll
@@ -80329,10 +80329,10 @@ define internal fastcc range(i32 -1, 1) i32 @JS_ToUint8ClampFree(ptr noundef %0,
   br label %7
 
 7:                                                ; preds = %41, %4
-  %8 = phi i64 [ %2, %4 ], [ %43, %41 ]
-  %.sroa.8.0 = phi i64 [ %3, %4 ], [ %44, %41 ]
+  %8 = phi i64 [ %2, %4 ], [ %43, %43 ]
+  %.sroa.8.0 = phi i64 [ %3, %4 ], [ %44, %43 ]
   %9 = trunc i64 %.sroa.8.0 to i32
-  switch i32 %9, label %41 [
+  switch i32 %9, label %43 [
     i32 0, label %10
     i32 1, label %10
     i32 2, label %10
@@ -80347,8 +80347,8 @@ define internal fastcc range(i32 -1, 1) i32 @JS_ToUint8ClampFree(ptr noundef %0,
 
 11:                                               ; preds = %JS_FreeValue.exit, %10
   %12 = phi i32 [ %.pre, %JS_FreeValue.exit ], [ %.sroa.015.sroa.0.0.extract.trunc43, %10 ]
-  %..i = call noundef i32 @llvm.smin.i32(i32 %12, i32 255)
-  %..i27 = call noundef i32 @llvm.smax.i32(i32 %..i, i32 0)
+  %..i = call i32 @llvm.smax.i32(i32 %12, i32 0)
+  %..i27 = call i32 @llvm.umin.i32(i32 %..i, i32 255)
   br label %.loopexit
 
 13:                                               ; preds = %7
@@ -80420,8 +80420,8 @@ JS_FreeValue.exit:                                ; preds = %33, %38
   br i1 %.not, label %.loopexit, label %7
 
 .loopexit:                                        ; preds = %41, %14, %13, %11, %16
-  %storemerge = phi i32 [ %18, %16 ], [ %..i27, %11 ], [ 0, %13 ], [ 255, %14 ], [ 0, %41 ]
-  %.0 = phi i32 [ 0, %16 ], [ 0, %11 ], [ 0, %13 ], [ 0, %14 ], [ -1, %41 ]
+  %storemerge = phi i32 [ %18, %18 ], [ %..i27, %11 ], [ 0, %15 ], [ 255, %16 ], [ 0, %43 ]
+  %.0 = phi i32 [ 0, %18 ], [ 0, %11 ], [ 0, %15 ], [ 0, %16 ], [ -1, %43 ]
   store i32 %storemerge, ptr %1, align 4, !tbaa !67
   call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %5) #41
   ret i32 %.0
diff --git a/bench/raylib/optimized/rcore.ll b/bench/raylib/optimized/rcore.ll
index 2bea919938c..26f25b4a595 100644
--- a/bench/raylib/optimized/rcore.ll
+++ b/bench/raylib/optimized/rcore.ll
@@ -33953,19 +33953,18 @@ define hidden range(i32 0, 2) i32 @msf_gif_frame(ptr noundef captures(none) %0,
   br i1 %.not, label %582, label %12
 
 12:                                               ; preds = %5
-  %13 = tail call noundef i32 @llvm.smin.i32(i32 %3, i32 16)
-  %14 = tail call range(i32 1, -2147483648) i32 @llvm.smax.i32(i32 %13, i32 1)
-  %15 = icmp eq i32 %4, 0
-  br i1 %15, label %16, label %._crit_edge
-
-16:                                               ; preds = %12
-  %17 = getelementptr inbounds nuw i8, ptr %0, i64 104
-  %18 = load i32, ptr %17, align 8
-  %19 = shl nsw i32 %18, 2
+  %13 = tail call i32 @llvm.smax.i32(i32 %3, i32 1)
+  %14 = icmp eq i32 %4, 0
+  br i1 %14, label %15, label %._crit_edge
+
+15:                                               ; preds = %12
+  %16 = getelementptr inbounds nuw i8, ptr %0, i64 104
+  %17 = load i32, ptr %16, align 8
+  %18 = shl nsw i32 %17, 2
   br label %._crit_edge
 
-._crit_edge:                                      ; preds = %16, %12
-  %.034 = phi i32 [ %19, %16 ], [ %4, %12 ]
+._crit_edge:                                      ; preds = %15, %12
+  %.034 = phi i32 [ %18, %15 ], [ %4, %12 ]
   %20 = icmp slt i32 %.034, 0
   %21 = getelementptr inbounds nuw i8, ptr %0, i64 108
   %22 = load i32, ptr %21, align 4
@@ -33987,7 +33986,8 @@ define hidden range(i32 0, 2) i32 @msf_gif_frame(ptr noundef captures(none) %0,
   %35 = tail call range(i32 1, -2147483648) i32 @llvm.smax.i32(i32 %34, i32 1)
   %36 = udiv i32 160, %35
   %37 = add nsw i32 %36, %32
-  %38 = tail call noundef i32 @llvm.smin.i32(i32 %14, i32 %37)
+  %38 = tail call i32 @llvm.smin.i32(i32 %13, i32 %37)
+  %38 = tail call i32 @llvm.smin.i32(i32 %37, i32 16)
   %39 = load i32, ptr @msf_gif_bgra_flag, align 4
   %.not.i = icmp eq i32 %39, 0
   %40 = select i1 %.not.i, ptr @msf_cook_frame.rdepthsArray, ptr @msf_cook_frame.bdepthsArray
diff --git a/bench/recastnavigation/optimized/DetourNavMeshBuilder.ll b/bench/recastnavigation/optimized/DetourNavMeshBuilder.ll
index 5b6bb6251d1..335895079d4 100644
--- a/bench/recastnavigation/optimized/DetourNavMeshBuilder.ll
+++ b/bench/recastnavigation/optimized/DetourNavMeshBuilder.ll
@@ -1243,16 +1243,16 @@ define internal fastcc void @_ZL12createBVTreeP21dtNavMeshCreateParamsP8dtBVNode
   %61 = fsub float %.sroa.0117.0.lcssa, %60
   %62 = fmul float %6, %61
   %63 = fptosi float %62 to i32
-  %64 = tail call i32 @llvm.smin.i32(i32 %63, i32 65535)
-  %65 = tail call i32 @llvm.smax.i32(i32 %64, i32 0)
+  %64 = tail call i32 @llvm.smax.i32(i32 %63, i32 0)
+  %65 = tail call i32 @llvm.umin.i32(i32 %64, i32 65535)
   %66 = trunc nuw i32 %65 to i16
   store i16 %66, ptr %23, align 4
   %67 = load float, ptr %17, align 4
   %68 = fsub float %.sroa.4119.0.lcssa, %67
   %69 = fmul float %6, %68
   %70 = fptosi float %69 to i32
-  %71 = tail call i32 @llvm.smin.i32(i32 %70, i32 65535)
-  %72 = tail call i32 @llvm.smax.i32(i32 %71, i32 0)
+  %71 = tail call i32 @llvm.smax.i32(i32 %70, i32 0)
+  %72 = tail call i32 @llvm.umin.i32(i32 %71, i32 65535)
   %73 = trunc nuw i32 %72 to i16
   %74 = getelementptr inbounds nuw i8, ptr %23, i64 2
   store i16 %73, ptr %74, align 2
@@ -1260,8 +1260,8 @@ define internal fastcc void @_ZL12createBVTreeP21dtNavMeshCreateParamsP8dtBVNode
   %76 = fsub float %.sroa.8121.0.lcssa, %75
   %77 = fmul float %6, %76
   %78 = fptosi float %77 to i32
-  %79 = tail call i32 @llvm.smin.i32(i32 %78, i32 65535)
-  %80 = tail call i32 @llvm.smax.i32(i32 %79, i32 0)
+  %79 = tail call i32 @llvm.smax.i32(i32 %78, i32 0)
+  %80 = tail call i32 @llvm.umin.i32(i32 %79, i32 65535)
   %81 = trunc nuw i32 %80 to i16
   %82 = getelementptr inbounds nuw i8, ptr %23, i64 4
   store i16 %81, ptr %82, align 4
@@ -1269,8 +1269,8 @@ define internal fastcc void @_ZL12createBVTreeP21dtNavMeshCreateParamsP8dtBVNode
   %84 = fsub float %.sroa.0.0.lcssa, %83
   %85 = fmul float %6, %84
   %86 = fptosi float %85 to i32
-  %87 = tail call i32 @llvm.smin.i32(i32 %86, i32 65535)
-  %88 = tail call i32 @llvm.smax.i32(i32 %87, i32 0)
+  %87 = tail call i32 @llvm.smax.i32(i32 %86, i32 0)
+  %88 = tail call i32 @llvm.umin.i32(i32 %87, i32 65535)
   %89 = trunc nuw i32 %88 to i16
   %90 = getelementptr inbounds nuw i8, ptr %23, i64 6
   store i16 %89, ptr %90, align 2
@@ -1278,8 +1278,8 @@ define internal fastcc void @_ZL12createBVTreeP21dtNavMeshCreateParamsP8dtBVNode
   %92 = fsub float %.sroa.4.0.lcssa, %91
   %93 = fmul float %6, %92
   %94 = fptosi float %93 to i32
-  %95 = tail call i32 @llvm.smin.i32(i32 %94, i32 65535)
-  %96 = tail call i32 @llvm.smax.i32(i32 %95, i32 0)
+  %95 = tail call i32 @llvm.smax.i32(i32 %94, i32 0)
+  %96 = tail call i32 @llvm.umin.i32(i32 %95, i32 65535)
   %97 = trunc nuw i32 %96 to i16
   %98 = getelementptr inbounds nuw i8, ptr %23, i64 8
   store i16 %97, ptr %98, align 2
@@ -1287,8 +1287,8 @@ define internal fastcc void @_ZL12createBVTreeP21dtNavMeshCreateParamsP8dtBVNode
   %100 = fsub float %.sroa.8.0.lcssa, %99
   %101 = fmul float %6, %100
   %102 = fptosi float %101 to i32
-  %103 = tail call i32 @llvm.smin.i32(i32 %102, i32 65535)
-  %104 = tail call i32 @llvm.smax.i32(i32 %103, i32 0)
+  %103 = tail call i32 @llvm.smax.i32(i32 %102, i32 0)
+  %104 = tail call i32 @llvm.umin.i32(i32 %103, i32 65535)
   %105 = trunc nuw i32 %104 to i16
   %106 = getelementptr inbounds nuw i8, ptr %23, i64 10
   store i16 %105, ptr %106, align 2
@@ -2301,9 +2301,6 @@ define internal noundef range(i32 -1, 2) i32 @_ZL12compareItemZPKvS0_(ptr nounde
   ret i32 %.0
 }
 
-; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smin.i32(i32, i32) #10
-
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i16 @llvm.umax.i16(i16, i16) #10
 
@@ -2313,6 +2310,9 @@ declare range(i32 -1, 2) i32 @llvm.ucmp.i32.i16(i16, i16) #10
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smax.i32(i32, i32) #10
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #10
+
 attributes #0 = { mustprogress uwtable "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #2 = { mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none) }
diff --git a/bench/recastnavigation/optimized/RecastRasterization.ll b/bench/recastnavigation/optimized/RecastRasterization.ll
index 954f0ab2f4b..7eb4c712621 100644
--- a/bench/recastnavigation/optimized/RecastRasterization.ll
+++ b/bench/recastnavigation/optimized/RecastRasterization.ll
@@ -572,8 +572,8 @@ _ZL13overlapBoundsPKfS0_S0_S0_.exit:              ; preds = %69
   %161 = fmul float %9, %.199
   %162 = tail call float @llvm.floor.f32(float %161)
   %163 = fptosi float %162 to i32
-  %164 = tail call i32 @llvm.smin.i32(i32 %163, i32 8191)
-  %165 = tail call i32 @llvm.smax.i32(i32 %164, i32 0)
+  %164 = tail call i32 @llvm.smax.i32(i32 %163, i32 0)
+  %165 = tail call i32 @llvm.umin.i32(i32 %164, i32 8191)
   %166 = trunc nuw nsw i32 %165 to i16
   %167 = fmul float %9, %.1
   %168 = tail call float @llvm.ceil.f32(float %167)
@@ -1235,6 +1235,9 @@ declare i32 @llvm.smin.i32(i32, i32) #5
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smax.i32(i32, i32) #5
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #5
+
 attributes #0 = { mustprogress uwtable "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #2 = { mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none) }
diff --git a/bench/redis/optimized/cluster_legacy.ll b/bench/redis/optimized/cluster_legacy.ll
index 750e3df7b5a..41e3a18e979 100644
--- a/bench/redis/optimized/cluster_legacy.ll
+++ b/bench/redis/optimized/cluster_legacy.ll
@@ -13246,8 +13246,8 @@ define dso_local void @clusterUpdateState() local_unnamed_addr #3 {
 
 68:                                               ; preds = %64
   %69 = load i64, ptr getelementptr inbounds nuw (i8, ptr @server, i64 7896), align 8, !tbaa !142
-  %spec.store.select = tail call i64 @llvm.smin.i64(i64 %69, i64 5000)
-  %spec.store.select1 = tail call i64 @llvm.smax.i64(i64 %spec.store.select, i64 500)
+  %spec.store.select = tail call i64 @llvm.smax.i64(i64 %69, i64 500)
+  %spec.store.select1 = tail call i64 @llvm.umin.i64(i64 %spec.store.select, i64 5000)
   %70 = icmp eq i32 %.1, 0
   br i1 %70, label %71, label %81
 
@@ -13267,7 +13267,7 @@ define dso_local void @clusterUpdateState() local_unnamed_addr #3 {
   br i1 %80, label %89, label %81
 
 81:                                               ; preds = %68, %71, %76
-  %82 = phi i32 [ 3, %68 ], [ 2, %71 ], [ 2, %76 ]
+  %82 = phi i32 [ 3, %68 ], [ 2, %73 ], [ 2, %78 ]
   %83 = load i32, ptr getelementptr inbounds nuw (i8, ptr @server, i64 6288), align 8, !tbaa !63
   %84 = icmp slt i32 %82, %83
   br i1 %84, label %.critedge, label %85
@@ -18424,6 +18424,9 @@ declare i64 @llvm.umax.i64(i64, i64) #32
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.smin.i32(i32, i32) #32
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i64 @llvm.umin.i64(i64, i64) #32
+
 attributes #0 = { mustprogress nofree norecurse nosync nounwind willreturn memory(read, inaccessiblemem: none) uwtable "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: read) uwtable "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #2 = { "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
diff --git a/bench/redis/optimized/server.ll b/bench/redis/optimized/server.ll
index e3c349d67dd..0baca170045 100644
--- a/bench/redis/optimized/server.ll
+++ b/bench/redis/optimized/server.ll
@@ -6441,9 +6441,9 @@ define dso_local range(i32 -1, 1) i32 @setOOMScoreAdj(i32 noundef %0) local_unna
   %35 = load i32, ptr @setOOMScoreAdj.oom_score_adj_base, align 4
   %36 = select i1 %34, i32 %35, i32 0
   %.0 = add nsw i32 %36, %33
-  %spec.store.select = tail call i32 @llvm.smin.i32(i32 %.0, i32 1000)
-  %spec.store.select1 = tail call i32 @llvm.smax.i32(i32 %spec.store.select, i32 -1000)
-  br label %40
+  %spec.store.select = tail call i32 @llvm.smax.i32(i32 %.0, i32 -1000)
+  %spec.store.select1 = tail call i32 @llvm.smin.i32(i32 %spec.store.select, i32 1000)
+  br label %41
 
 37:                                               ; preds = %8
   br i1 %.b, label %38, label %57
@@ -6451,45 +6451,45 @@ define dso_local range(i32 -1, 1) i32 @setOOMScoreAdj(i32 noundef %0) local_unna
 38:                                               ; preds = %37
   store i1 false, ptr @setOOMScoreAdj.oom_score_adjusted_by_redis, align 4
   %39 = load i32, ptr @setOOMScoreAdj.oom_score_adj_base, align 4, !tbaa !50
-  br label %40
-
-40:                                               ; preds = %29, %38
-  %.1 = phi i32 [ %spec.store.select1, %29 ], [ %39, %38 ]
-  %41 = call i32 (ptr, i64, ptr, ...) @snprintf(ptr noundef nonnull dereferenceable(1) %2, i64 noundef 63, ptr noundef nonnull @.str.135, i32 noundef %.1) #43
-  %42 = tail call i32 (ptr, i32, ...) @open64(ptr noundef nonnull @.str.133, i32 noundef 1) #43
-  %43 = icmp slt i32 %42, 0
-  br i1 %43, label %48, label %44
-
-44:                                               ; preds = %40
-  %45 = call i64 @strlen(ptr noundef nonnull dereferenceable(1) %2) #44
-  %46 = call i64 @write(i32 noundef %42, ptr noundef nonnull %2, i64 noundef %45) #43
-  %47 = icmp slt i64 %46, 0
-  br i1 %47, label %48, label %.sink.split
-
-48:                                               ; preds = %40, %44
-  %49 = load i32, ptr getelementptr inbounds nuw (i8, ptr @server, i64 6288), align 8, !tbaa !39
-  %50 = icmp sgt i32 %49, 3
-  br i1 %50, label %55, label %51
-
-51:                                               ; preds = %48
-  %52 = tail call ptr @__errno_location() #46
-  %53 = load i32, ptr %52, align 4, !tbaa !50
-  %54 = tail call ptr @strerror(i32 noundef %53) #43
-  tail call void (i32, ptr, ...) @_serverLog(i32 noundef 3, ptr noundef nonnull @.str.136, ptr noundef %54)
-  br label %55
-
-55:                                               ; preds = %48, %51
-  %.not30 = icmp eq i32 %42, -1
-  br i1 %.not30, label %57, label %.sink.split
+  br label %41
 
-.sink.split:                                      ; preds = %44, %55, %24
-  %.sink = phi i32 [ %12, %24 ], [ %42, %55 ], [ %42, %44 ]
-  %.021.ph = phi i32 [ -1, %24 ], [ -1, %55 ], [ 0, %44 ]
-  %56 = tail call i32 @close(i32 noundef %.sink) #43
-  br label %57
+55:                                               ; preds = %29, %39
+  %.1 = phi i32 [ %spec.store.select1, %29 ], [ %40, %39 ]
+  %42 = call i32 (ptr, i64, ptr, ...) @snprintf(ptr noundef nonnull dereferenceable(1) %2, i64 noundef 63, ptr noundef nonnull @.str.135, i32 noundef %.1) #43
+  %43 = tail call i32 (ptr, i32, ...) @open64(ptr noundef nonnull @.str.133, i32 noundef 1) #43
+  %44 = icmp slt i32 %43, 0
+  br i1 %44, label %49, label %45
+
+.sink.split:                                      ; preds = %55
+  %46 = call i64 @strlen(ptr noundef nonnull dereferenceable(1) %2) #44
+  %47 = call i64 @write(i32 noundef %43, ptr noundef nonnull %2, i64 noundef %46) #43
+  %48 = icmp slt i64 %47, 0
+  br i1 %48, label %49, label %.sink.split
+
+49:                                               ; preds = %41, %45
+  %50 = load i32, ptr getelementptr inbounds nuw (i8, ptr @server, i64 6288), align 8, !tbaa !39
+  %51 = icmp sgt i32 %50, 3
+  br i1 %51, label %56, label %52
+
+52:                                               ; preds = %49
+  %53 = tail call ptr @__errno_location() #46
+  %54 = load i32, ptr %53, align 4, !tbaa !50
+  %55 = tail call ptr @strerror(i32 noundef %54) #43
+  tail call void (i32, ptr, ...) @_serverLog(i32 noundef 3, ptr noundef nonnull @.str.136, ptr noundef %55)
+  br label %56
+
+56:                                               ; preds = %49, %52
+  %.not30 = icmp eq i32 %43, -1
+  br i1 %.not30, label %58, label %.sink.split
+
+.sink.split:                                      ; preds = %45, %56, %24
+  %.sink = phi i32 [ %12, %24 ], [ %43, %56 ], [ %43, %45 ]
+  %.021.ph = phi i32 [ -1, %24 ], [ -1, %56 ], [ 0, %45 ]
+  %57 = tail call i32 @close(i32 noundef %.sink) #43
+  br label %58
 
-57:                                               ; preds = %.sink.split, %55, %37, %24
-  %.021 = phi i32 [ -1, %24 ], [ 0, %37 ], [ -1, %55 ], [ %.021.ph, %.sink.split ]
+57:                                               ; preds = %.sink.split, %56, %37, %24
+  %.021 = phi i32 [ -1, %24 ], [ 0, %38 ], [ -1, %56 ], [ %.021.ph, %.sink.split ]
   call void @llvm.lifetime.end.p0(i64 64, ptr nonnull %2) #43
   ret i32 %.021
 }
diff --git a/bench/sentencepiece/optimized/unigram_model.ll b/bench/sentencepiece/optimized/unigram_model.ll
index 98ff0637b58..56befaad3dd 100644
--- a/bench/sentencepiece/optimized/unigram_model.ll
+++ b/bench/sentencepiece/optimized/unigram_model.ll
@@ -8344,8 +8344,8 @@ _ZNSt4pairISt6vectorIS_ISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EEfED2
   br label %220
 
 50:                                               ; preds = %17
-  %.sroa.speculated102 = call i32 @llvm.smin.i32(i32 %4, i32 1024)
-  %.sroa.speculated = call i32 @llvm.smax.i32(i32 %.sroa.speculated102, i32 1)
+  %.sroa.speculated102 = call i32 @llvm.smax.i32(i32 %4, i32 1)
+  %.sroa.speculated = call i32 @llvm.umin.i32(i32 %.sroa.speculated102, i32 1024)
   %51 = icmp slt i32 %4, 2
   br i1 %51, label %52, label %109
 
@@ -8356,7 +8356,7 @@ _ZNSt4pairISt6vectorIS_ISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EEfED2
   %54 = getelementptr inbounds nuw i8, ptr %53, i64 40
   %55 = load ptr, ptr %54, align 8
   invoke void %55(ptr dead_on_unwind nonnull writable sret(%"class.std::vector.84") align 8 %9, ptr noundef nonnull align 8 dereferenceable(176) %1, i64 %2, ptr %3)
-          to label %56 unwind label %94
+          to label %58 unwind label %94
 
 56:                                               ; preds = %52
   %57 = load ptr, ptr %9, align 8, !tbaa !245
@@ -8383,7 +8383,7 @@ _ZNSt12_Vector_baseISt4pairISt6vectorIS0_ISt17basic_string_viewIcSt11char_traits
   %68 = getelementptr inbounds nuw i8, ptr %0, i64 16
   store ptr %67, ptr %68, align 8, !tbaa !304
   %69 = invoke noundef ptr @_ZSt16__do_uninit_copyIPKSt4pairISt6vectorIS0_ISt17basic_string_viewIcSt11char_traitsIcEEiESaIS6_EEfEPS9_ET0_T_SE_SD_(ptr noundef nonnull %8, ptr noundef nonnull %66, ptr noundef nonnull %65)
-          to label %79 unwind label %70
+          to label %81 unwind label %70
 
 70:                                               ; preds = %_ZNSt12_Vector_baseISt4pairISt6vectorIS0_ISt17basic_string_viewIcSt11char_traitsIcEEiESaIS6_EEfESaIS9_EE11_M_allocateEm.exit.i.i51, %56
   %71 = landingpad { ptr, i32 }
@@ -8468,7 +8468,7 @@ _ZNSt4pairISt6vectorIS_ISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EEfED2
   br label %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EED2Ev.exit62
 
 _ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EED2Ev.exit62: ; preds = %104, %_ZNSt4pairISt6vectorIS_ISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EEfED2Ev.exit60, %94
-  %.pn38 = phi { ptr, i32 } [ %95, %94 ], [ %71, %_ZNSt4pairISt6vectorIS_ISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EEfED2Ev.exit60 ], [ %71, %104 ]
+  %.pn38 = phi { ptr, i32 } [ %95, %96 ], [ %71, %_ZNSt4pairISt6vectorIS_ISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EEfED2Ev.exit60 ], [ %71, %106 ]
   call void @llvm.lifetime.end.p0(i64 24, ptr nonnull %9) #29
   call void @llvm.lifetime.end.p0(i64 32, ptr nonnull %8) #29
   br label %220
@@ -8477,18 +8477,18 @@ _ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EED2Ev.ex
   call void @llvm.lifetime.start.p0(i64 152, ptr nonnull %10) #29
   call void @_ZN13sentencepiece7unigram7LatticeC1Ev(ptr noundef nonnull align 8 dereferenceable(152) %10)
   invoke void @_ZN13sentencepiece7unigram7Lattice11SetSentenceESt17basic_string_viewIcSt11char_traitsIcEE(ptr noundef nonnull align 8 dereferenceable(152) %10, i64 %2, ptr %3)
-          to label %110 unwind label %136
+          to label %112 unwind label %136
 
 110:                                              ; preds = %109
   invoke void @_ZNK13sentencepiece7unigram5Model13PopulateNodesEPNS0_7LatticeE(ptr noundef nonnull align 8 dereferenceable(176) %1, ptr noundef nonnull %10)
-          to label %111 unwind label %136
+          to label %113 unwind label %136
 
 111:                                              ; preds = %110
   call void @llvm.memset.p0.i64(ptr noundef nonnull align 8 dereferenceable(24) %0, i8 0, i64 24, i1 false)
   call void @llvm.lifetime.start.p0(i64 24, ptr nonnull %11) #29
   %112 = zext nneg i32 %.sroa.speculated to i64
   invoke void @_ZN13sentencepiece7unigram7Lattice5NBestEmbf(ptr dead_on_unwind nonnull writable sret(%"class.std::vector.16") align 8 %11, ptr noundef nonnull align 8 dereferenceable(152) %10, i64 noundef %112, i1 noundef zeroext false, float noundef 0.000000e+00)
-          to label %113 unwind label %138
+          to label %115 unwind label %138
 
 113:                                              ; preds = %111
   %114 = load ptr, ptr %11, align 8, !tbaa !124
@@ -8535,7 +8535,7 @@ _ZSt8_DestroyIPSt4pairISt6vectorIPN13sentencepiece7unigram7Lattice4NodeESaIS6_EE
   br label %_ZSt8_DestroyIPSt4pairISt6vectorIPN13sentencepiece7unigram7Lattice4NodeESaIS6_EEfES9_EvT_SB_RSaIT0_E.exit.i
 
 _ZSt8_DestroyIPSt4pairISt6vectorIPN13sentencepiece7unigram7Lattice4NodeESaIS6_EEfES9_EvT_SB_RSaIT0_E.exit.i: ; preds = %113, %_ZSt8_DestroyIPSt4pairISt6vectorIPN13sentencepiece7unigram7Lattice4NodeESaIS6_EEfES9_EvT_SB_RSaIT0_E.exitthread-pre-split.i, %._crit_edge123
-  %129 = phi ptr [ %.pr.i, %_ZSt8_DestroyIPSt4pairISt6vectorIPN13sentencepiece7unigram7Lattice4NodeESaIS6_EEfES9_EvT_SB_RSaIT0_E.exitthread-pre-split.i ], [ %.pre125, %._crit_edge123 ], [ %114, %113 ]
+  %129 = phi ptr [ %.pr.i, %_ZSt8_DestroyIPSt4pairISt6vectorIPN13sentencepiece7unigram7Lattice4NodeESaIS6_EEfES9_EvT_SB_RSaIT0_E.exitthread-pre-split.i ], [ %.pre125, %._crit_edge123 ], [ %114, %115 ]
   %.not.i.i.i64 = icmp eq ptr %129, null
   br i1 %.not.i.i.i64, label %_ZNSt6vectorISt4pairIS_IPN13sentencepiece7unigram7Lattice4NodeESaIS5_EEfESaIS8_EED2Ev.exit, label %130
 
@@ -8575,7 +8575,7 @@ _ZNSt6vectorISt4pairIS_IPN13sentencepiece7unigram7Lattice4NodeESaIS5_EEfESaIS8_E
   br i1 %.not110117, label %._crit_edge, label %.lr.ph
 
 ._crit_edge:                                      ; preds = %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE12emplace_backIJRKS4_RKiEEERS5_DpOT_.exit, %140
-  %144 = phi ptr [ null, %140 ], [ %201, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE12emplace_backIJRKS4_RKiEEERS5_DpOT_.exit ]
+  %144 = phi ptr [ null, %142 ], [ %201, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE12emplace_backIJRKS4_RKiEEERS5_DpOT_.exit ]
   %145 = getelementptr inbounds nuw i8, ptr %.sroa.082.0120, i64 24
   %146 = load ptr, ptr %119, align 8, !tbaa !305
   %147 = load ptr, ptr %120, align 8, !tbaa !304
@@ -8608,7 +8608,7 @@ _ZNSt16allocator_traitsISaISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiEE
           to label %.noexc67 unwind label %.loopexit111
 
 .noexc67:                                         ; preds = %_ZNSt16allocator_traitsISaISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiEEE8allocateERS6_m.exit.i.i.i.i.i.i.i.i, %148
-  %157 = phi ptr [ null, %148 ], [ %156, %_ZNSt16allocator_traitsISaISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiEEE8allocateERS6_m.exit.i.i.i.i.i.i.i.i ]
+  %157 = phi ptr [ null, %150 ], [ %156, %_ZNSt16allocator_traitsISaISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiEEE8allocateERS6_m.exit.i.i.i.i.i.i.i.i ]
   store ptr %157, ptr %146, align 8, !tbaa !245
   %158 = getelementptr inbounds nuw i8, ptr %146, i64 8
   store ptr %157, ptr %158, align 8, !tbaa !243
@@ -8649,8 +8649,8 @@ _ZNSt16allocator_traitsISaISt4pairISt6vectorIS0_ISt17basic_string_viewIcSt11char
   br label %_ZNSt6vectorISt4pairIS_IS0_ISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EEfESaIS8_EE12emplace_backIJRS7_RKfEEERS8_DpOT_.exit
 
 .lr.ph:                                           ; preds = %140, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE12emplace_backIJRKS4_RKiEEERS5_DpOT_.exit
-  %170 = phi ptr [ %201, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE12emplace_backIJRKS4_RKiEEERS5_DpOT_.exit ], [ null, %140 ]
-  %.sroa.078.0118 = phi ptr [ %202, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE12emplace_backIJRKS4_RKiEEERS5_DpOT_.exit ], [ %141, %140 ]
+  %170 = phi ptr [ %201, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE12emplace_backIJRKS4_RKiEEERS5_DpOT_.exit ], [ null, %142 ]
+  %.sroa.078.0118 = phi ptr [ %202, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE12emplace_backIJRKS4_RKiEEERS5_DpOT_.exit ], [ %141, %142 ]
   %171 = load ptr, ptr %.sroa.078.0118, align 8, !tbaa !27
   %172 = getelementptr inbounds nuw i8, ptr %171, i64 28
   %173 = load ptr, ptr %118, align 8, !tbaa !283
@@ -8731,7 +8731,7 @@ _ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE17_M_re
   br label %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE12emplace_backIJRKS4_RKiEEERS5_DpOT_.exit
 
 _ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE12emplace_backIJRKS4_RKiEEERS5_DpOT_.exit: ; preds = %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE17_M_realloc_insertIJRKS4_RKiEEEvN9__gnu_cxx17__normal_iteratorIPS5_S7_EEDpOT_.exit.i, %174
-  %201 = phi ptr [ %198, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE17_M_realloc_insertIJRKS4_RKiEEEvN9__gnu_cxx17__normal_iteratorIPS5_S7_EEDpOT_.exit.i ], [ %178, %174 ]
+  %201 = phi ptr [ %198, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EE17_M_realloc_insertIJRKS4_RKiEEEvN9__gnu_cxx17__normal_iteratorIPS5_S7_EEDpOT_.exit.i ], [ %178, %176 ]
   %202 = getelementptr inbounds nuw i8, ptr %.sroa.078.0118, i64 8
   %.not110 = icmp eq ptr %202, %143
   br i1 %.not110, label %._crit_edge, label %.lr.ph
@@ -8796,12 +8796,12 @@ _ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EED2Ev.ex
   br label %217
 
 217:                                              ; preds = %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EED2Ev.exit77, %138
-  %.pn.pn = phi { ptr, i32 } [ %.pn, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EED2Ev.exit77 ], [ %139, %138 ]
+  %.pn.pn = phi { ptr, i32 } [ %.pn, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EED2Ev.exit77 ], [ %139, %140 ]
   call void @_ZNSt6vectorISt4pairIS_IS0_ISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EEfESaIS8_EED2Ev(ptr noundef nonnull align 8 dereferenceable(24) %0) #29
   br label %218
 
 218:                                              ; preds = %217, %136
-  %.pn.pn.pn = phi { ptr, i32 } [ %.pn.pn, %217 ], [ %137, %136 ]
+  %.pn.pn.pn = phi { ptr, i32 } [ %.pn.pn, %219 ], [ %137, %138 ]
   call void @_ZN13sentencepiece7unigram7LatticeD1Ev(ptr noundef nonnull align 8 dereferenceable(152) %10) #29
   call void @llvm.lifetime.end.p0(i64 152, ptr nonnull %10) #29
   br label %220
@@ -8810,7 +8810,7 @@ _ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EED2Ev.ex
   ret void
 
 220:                                              ; preds = %218, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EED2Ev.exit62, %_ZNSt4pairISt6vectorIS_ISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EEfED2Ev.exit45
-  %.pn40.pn = phi { ptr, i32 } [ %26, %_ZNSt4pairISt6vectorIS_ISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EEfED2Ev.exit45 ], [ %.pn38, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EED2Ev.exit62 ], [ %.pn.pn.pn, %218 ]
+  %.pn40.pn = phi { ptr, i32 } [ %26, %_ZNSt4pairISt6vectorIS_ISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EEfED2Ev.exit45 ], [ %.pn38, %_ZNSt6vectorISt4pairISt17basic_string_viewIcSt11char_traitsIcEEiESaIS5_EED2Ev.exit62 ], [ %.pn.pn.pn, %220 ]
   resume { ptr, i32 } %.pn40.pn
 }
 
@@ -19614,6 +19614,9 @@ declare i64 @llvm.smin.i64(i64, i64) #25
 ; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: read)
 declare i32 @bcmp(ptr captures(none), ptr captures(none), i64) local_unnamed_addr #27
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #25
+
 attributes #0 = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { nounwind "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #2 = { nofree nounwind }
diff --git a/bench/sqlite/optimized/sqlite3.ll b/bench/sqlite/optimized/sqlite3.ll
index 744c9d298b6..ca4eb932abf 100644
--- a/bench/sqlite/optimized/sqlite3.ll
+++ b/bench/sqlite/optimized/sqlite3.ll
@@ -317377,8 +317377,8 @@ define internal void @roundFunc(ptr noundef captures(none) %0, i32 noundef %1, p
 
 sqlite3_value_int64.exit:                         ; preds = %18, %22, %25, %27, %29, %31, %34
   %.0.i.i = phi i64 [ %19, %18 ], [ %35, %34 ], [ 0, %31 ], [ 0, %29 ], [ %28, %27 ], [ -9223372036854775808, %22 ], [ 9223372036854775807, %25 ]
-  %spec.store.select = tail call i64 @llvm.smin.i64(i64 %.0.i.i, i64 30)
-  %spec.store.select2 = tail call i64 @llvm.smax.i64(i64 %spec.store.select, i64 0)
+  %spec.store.select = tail call i64 @llvm.smax.i64(i64 %.0.i.i, i64 0)
+  %spec.store.select2 = tail call i64 @llvm.umin.i64(i64 %spec.store.select, i64 30)
   br label %36
 
 36:                                               ; preds = %sqlite3_value_int64.exit, %3
@@ -317423,7 +317423,7 @@ sqlite3_value_int64.exit:                         ; preds = %18, %22, %25, %27,
   br label %sqlite3_value_double.exit
 
 sqlite3_value_double.exit:                        ; preds = %47, %51, %54, %56
-  %.0.i.i19 = phi double [ %48, %47 ], [ %53, %51 ], [ %57, %56 ], [ 0.000000e+00, %54 ]
+  %.0.i.i19 = phi double [ %48, %49 ], [ %53, %53 ], [ %57, %58 ], [ 0.000000e+00, %56 ]
   %58 = tail call double @llvm.fabs.f64(double %.0.i.i19)
   %or.cond = fcmp ogt double %58, 0x4330000000000000
   br i1 %or.cond, label %sqlite3_free.exit, label %59
@@ -317464,7 +317464,7 @@ sqlite3_value_double.exit:                        ; preds = %47, %51, %54, %56
   br label %sqlite3VdbeMemSetNull.exit.i
 
 sqlite3VdbeMemSetNull.exit.i:                     ; preds = %77, %76
-  %78 = phi ptr [ %.pre.i, %76 ], [ %72, %77 ]
+  %78 = phi ptr [ %.pre.i, %78 ], [ %72, %79 ]
   %79 = getelementptr inbounds nuw i8, ptr %0, i64 36
   store i32 7, ptr %79, align 4, !tbaa !636
   %80 = getelementptr inbounds nuw i8, ptr %78, i64 24
@@ -317515,7 +317515,7 @@ sqlite3VdbeMemSetNull.exit.i:                     ; preds = %77, %76
   br i1 %.not1619.i.i, label %sqlite3_result_error_nomem.exit, label %.lr.ph.i.i
 
 .lr.ph.i.i:                                       ; preds = %102, %.lr.ph.i.i
-  %.020.i.i = phi ptr [ %.0.i.i22, %.lr.ph.i.i ], [ %.018.i.i, %102 ]
+  %.020.i.i = phi ptr [ %.0.i.i22, %.lr.ph.i.i ], [ %.018.i.i, %104 ]
   %105 = getelementptr inbounds nuw i8, ptr %.020.i.i, i64 52
   %106 = load i32, ptr %105, align 4, !tbaa !253
   %107 = add nsw i32 %106, 1
@@ -317543,12 +317543,12 @@ sqlite3VdbeMemSetNull.exit.i:                     ; preds = %77, %76
 
 117:                                              ; preds = %115
   %118 = load ptr, ptr getelementptr inbounds nuw (i8, ptr @sqlite3Config, i64 128), align 8, !tbaa !3
-  tail call void %118(ptr noundef nonnull %116) #72
+  tail call void %120(ptr noundef nonnull %116) #72
   br label %sqlite3_mutex_enter.exit.i
 
 sqlite3_mutex_enter.exit.i:                       ; preds = %117, %115
   %119 = load ptr, ptr getelementptr inbounds nuw (i8, ptr @sqlite3Config, i64 56), align 8, !tbaa !85
-  %120 = tail call i32 %119(ptr noundef nonnull %69) #72
+  %120 = tail call i32 %121(ptr noundef nonnull %69) #72
   %121 = sext i32 %120 to i64
   %122 = load i64, ptr @sqlite3Stat, align 8, !tbaa !14
   %123 = sub nsw i64 %122, %121
@@ -317557,19 +317557,19 @@ sqlite3_mutex_enter.exit.i:                       ; preds = %117, %115
   %125 = add nsw i64 %124, -1
   store i64 %125, ptr getelementptr inbounds nuw (i8, ptr @sqlite3Stat, i64 72), align 8, !tbaa !14
   %126 = load ptr, ptr getelementptr inbounds nuw (i8, ptr @sqlite3Config, i64 40), align 8, !tbaa !163
-  tail call void %126(ptr noundef nonnull %69) #72
+  tail call void %128(ptr noundef nonnull %69) #72
   %127 = load ptr, ptr @mem0, align 8, !tbaa !161
   %.not.i4.i = icmp eq ptr %127, null
   br i1 %.not.i4.i, label %sqlite3_free.exitthread-pre-split, label %128
 
 128:                                              ; preds = %sqlite3_mutex_enter.exit.i
   %129 = load ptr, ptr getelementptr inbounds nuw (i8, ptr @sqlite3Config, i64 144), align 8, !tbaa !15
-  tail call void %129(ptr noundef nonnull %127) #72
+  tail call void %131(ptr noundef nonnull %127) #72
   br label %sqlite3_free.exitthread-pre-split
 
 130:                                              ; preds = %109
   %131 = load ptr, ptr getelementptr inbounds nuw (i8, ptr @sqlite3Config, i64 40), align 8, !tbaa !163
-  tail call void %131(ptr noundef nonnull %69) #72
+  tail call void %133(ptr noundef nonnull %69) #72
   br label %sqlite3_free.exitthread-pre-split
 
 sqlite3_free.exitthread-pre-split:                ; preds = %sqlite3_mutex_enter.exit.i, %128, %130
@@ -317577,7 +317577,7 @@ sqlite3_free.exitthread-pre-split:                ; preds = %sqlite3_mutex_enter
   br label %sqlite3_free.exit
 
 sqlite3_free.exit:                                ; preds = %sqlite3_free.exitthread-pre-split, %61, %sqlite3_value_double.exit
-  %132 = phi double [ %.pr, %sqlite3_free.exitthread-pre-split ], [ %66, %61 ], [ %.0.i.i19, %sqlite3_value_double.exit ]
+  %132 = phi double [ %.pr, %sqlite3_free.exitthread-pre-split ], [ %66, %63 ], [ %.0.i.i19, %sqlite3_value_double.exit ]
   %133 = load ptr, ptr %0, align 8, !tbaa !634
   %134 = getelementptr inbounds nuw i8, ptr %133, i64 20
   %135 = load i16, ptr %134, align 4, !tbaa !147
diff --git a/bench/stb/optimized/stb_image_write.ll b/bench/stb/optimized/stb_image_write.ll
index 0a7d1a10f3d..3e4ddc641ca 100644
--- a/bench/stb/optimized/stb_image_write.ll
+++ b/bench/stb/optimized/stb_image_write.ll
@@ -5509,8 +5509,8 @@ define range(i32 0, 2) i32 @stbi_write_jpg_core(ptr noundef readonly captures(no
   %.not = icmp eq i32 %5, 0
   %46 = select i1 %.not, i32 90, i32 %5
   %47 = icmp slt i32 %46, 91
-  %48 = tail call i32 @llvm.smin.i32(i32 %46, i32 100)
-  %49 = tail call i32 @llvm.smax.i32(i32 %48, i32 1)
+  %48 = tail call i32 @llvm.smax.i32(i32 %46, i32 1)
+  %49 = tail call i32 @llvm.umin.i32(i32 %48, i32 100)
   %50 = icmp slt i32 %46, 50
   br i1 %50, label %51, label %53
 
@@ -5536,8 +5536,8 @@ define range(i32 0, 2) i32 @stbi_write_jpg_core(ptr noundef readonly captures(no
   %61 = mul nsw i32 %60, %57
   %62 = add nsw i32 %61, 50
   %63 = sdiv i32 %62, 100
-  %64 = tail call i32 @llvm.smin.i32(i32 %63, i32 255)
-  %65 = tail call i32 @llvm.smax.i32(i32 %64, i32 1)
+  %64 = tail call i32 @llvm.smax.i32(i32 %63, i32 1)
+  %65 = tail call i32 @llvm.umin.i32(i32 %64, i32 255)
   %66 = trunc nuw i32 %65 to i8
   %67 = getelementptr inbounds nuw [64 x i8], ptr @stbiw__jpg_ZigZag, i64 0, i64 %indvars.iv
   %68 = load i8, ptr %67, align 1, !tbaa !11
@@ -5549,8 +5549,8 @@ define range(i32 0, 2) i32 @stbi_write_jpg_core(ptr noundef readonly captures(no
   %73 = mul nsw i32 %72, %57
   %74 = add nsw i32 %73, 50
   %75 = sdiv i32 %74, 100
-  %76 = tail call i32 @llvm.smin.i32(i32 %75, i32 255)
-  %77 = tail call i32 @llvm.smax.i32(i32 %76, i32 1)
+  %76 = tail call i32 @llvm.smax.i32(i32 %75, i32 1)
+  %77 = tail call i32 @llvm.umin.i32(i32 %76, i32 255)
   %78 = trunc nuw i32 %77 to i8
   %79 = getelementptr inbounds nuw [64 x i8], ptr %26, i64 0, i64 %69
   store i8 %78, ptr %79, align 1, !tbaa !11
@@ -6205,6 +6205,9 @@ declare i32 @llvm.ctlz.i32(i32, i1 immarg) #24
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i32 @llvm.umax.i32(i32, i32) #24
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare i32 @llvm.umin.i32(i32, i32) #24
+
 attributes #0 = { mustprogress nofree norecurse nosync nounwind willreturn memory(write, argmem: none, inaccessiblemem: none) uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: write) uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #2 = { nofree nounwind uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
diff --git a/bench/stockfish/optimized/evaluate_nnue.ll b/bench/stockfish/optimized/evaluate_nnue.ll
index d7022bd71ce..d24aca3142f 100644
--- a/bench/stockfish/optimized/evaluate_nnue.ll
+++ b/bench/stockfish/optimized/evaluate_nnue.ll
@@ -799,21 +799,21 @@ _ZNK9Stockfish4Eval4NNUE18FeatureTransformerILj2560EXadL_ZNS_9StateInfo14accumul
   %158 = shl nuw nsw i64 %indvars.iv, 1
   %159 = getelementptr inbounds nuw <2 x i64>, ptr %154, i64 %158
   %160 = load <8 x i16>, ptr %159, align 16
-  %161 = call <8 x i16> @llvm.smin.v8i16(<8 x i16> %160, <8 x i16> splat (i16 127))
-  %162 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %161, <8 x i16> zeroinitializer)
+  %161 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %160, <8 x i16> zeroinitializer)
+  %162 = call <8 x i16> @llvm.umin.v8i16(<8 x i16> %161, <8 x i16> splat (i16 127))
   %163 = or disjoint i64 %158, 1
   %164 = getelementptr inbounds nuw <2 x i64>, ptr %154, i64 %163
   %165 = load <8 x i16>, ptr %164, align 16
-  %166 = call <8 x i16> @llvm.smin.v8i16(<8 x i16> %165, <8 x i16> splat (i16 127))
-  %167 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %166, <8 x i16> zeroinitializer)
+  %166 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %165, <8 x i16> zeroinitializer)
+  %167 = call <8 x i16> @llvm.umin.v8i16(<8 x i16> %166, <8 x i16> splat (i16 127))
   %168 = getelementptr inbounds nuw <2 x i64>, ptr %155, i64 %158
   %169 = load <8 x i16>, ptr %168, align 16
-  %170 = call <8 x i16> @llvm.smin.v8i16(<8 x i16> %169, <8 x i16> splat (i16 127))
-  %171 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %170, <8 x i16> zeroinitializer)
+  %170 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %169, <8 x i16> zeroinitializer)
+  %171 = call <8 x i16> @llvm.umin.v8i16(<8 x i16> %170, <8 x i16> splat (i16 127))
   %172 = getelementptr inbounds nuw <2 x i64>, ptr %155, i64 %163
   %173 = load <8 x i16>, ptr %172, align 16
-  %174 = call <8 x i16> @llvm.smin.v8i16(<8 x i16> %173, <8 x i16> splat (i16 127))
-  %175 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %174, <8 x i16> zeroinitializer)
+  %174 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %173, <8 x i16> zeroinitializer)
+  %175 = call <8 x i16> @llvm.umin.v8i16(<8 x i16> %174, <8 x i16> splat (i16 127))
   %176 = mul nuw nsw <8 x i16> %171, %162
   %177 = mul nuw nsw <8 x i16> %175, %167
   %178 = lshr <8 x i16> %176, splat (i16 7)
@@ -1423,21 +1423,21 @@ _ZNK9Stockfish4Eval4NNUE18FeatureTransformerILj128EXadL_ZNS_9StateInfo16accumula
   %154 = shl nuw nsw i64 %indvars.iv, 1
   %155 = getelementptr inbounds nuw <2 x i64>, ptr %150, i64 %154
   %156 = load <8 x i16>, ptr %155, align 16
-  %157 = call <8 x i16> @llvm.smin.v8i16(<8 x i16> %156, <8 x i16> splat (i16 127))
-  %158 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %157, <8 x i16> zeroinitializer)
+  %157 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %156, <8 x i16> zeroinitializer)
+  %158 = call <8 x i16> @llvm.umin.v8i16(<8 x i16> %157, <8 x i16> splat (i16 127))
   %159 = or disjoint i64 %154, 1
   %160 = getelementptr inbounds nuw <2 x i64>, ptr %150, i64 %159
   %161 = load <8 x i16>, ptr %160, align 16
-  %162 = call <8 x i16> @llvm.smin.v8i16(<8 x i16> %161, <8 x i16> splat (i16 127))
-  %163 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %162, <8 x i16> zeroinitializer)
+  %162 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %161, <8 x i16> zeroinitializer)
+  %163 = call <8 x i16> @llvm.umin.v8i16(<8 x i16> %162, <8 x i16> splat (i16 127))
   %164 = getelementptr inbounds nuw <2 x i64>, ptr %151, i64 %154
   %165 = load <8 x i16>, ptr %164, align 16
-  %166 = call <8 x i16> @llvm.smin.v8i16(<8 x i16> %165, <8 x i16> splat (i16 127))
-  %167 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %166, <8 x i16> zeroinitializer)
+  %166 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %165, <8 x i16> zeroinitializer)
+  %167 = call <8 x i16> @llvm.umin.v8i16(<8 x i16> %166, <8 x i16> splat (i16 127))
   %168 = getelementptr inbounds nuw <2 x i64>, ptr %151, i64 %159
   %169 = load <8 x i16>, ptr %168, align 16
-  %170 = call <8 x i16> @llvm.smin.v8i16(<8 x i16> %169, <8 x i16> splat (i16 127))
-  %171 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %170, <8 x i16> zeroinitializer)
+  %170 = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %169, <8 x i16> zeroinitializer)
+  %171 = call <8 x i16> @llvm.umin.v8i16(<8 x i16> %170, <8 x i16> splat (i16 127))
   %172 = mul nuw nsw <8 x i16> %167, %158
   %173 = mul nuw nsw <8 x i16> %171, %163
   %174 = lshr <8 x i16> %172, splat (i16 7)
@@ -8619,9 +8619,6 @@ define linkonce_odr dso_local void @_ZNK9Stockfish4Eval4NNUE18FeatureTransformer
 ; Function Attrs: mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare <8 x i16> @llvm.smax.v8i16(<8 x i16>, <8 x i16>) #4
 
-; Function Attrs: mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare <8 x i16> @llvm.smin.v8i16(<8 x i16>, <8 x i16>) #4
-
 ; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(none)
 declare <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16>, <8 x i16>) #9
 
@@ -9608,6 +9605,9 @@ declare void @llvm.lifetime.end.p0(i64 immarg, ptr captures(none)) #11
 ; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)
 declare void @llvm.memmove.p0.p0.i64(ptr writeonly captures(none), ptr readonly captures(none), i64, i1 immarg) #12
 
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare <8 x i16> @llvm.umin.v8i16(<8 x i16>, <8 x i16>) #13
+
 ; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: readwrite)
 declare void @llvm.experimental.noalias.scope.decl(metadata) #13
 
@@ -9627,8 +9627,8 @@ attributes #9 = { mustprogress nocallback nofree nosync nounwind willreturn memo
 attributes #10 = { noreturn "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #11 = { nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
 attributes #12 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }
-attributes #13 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: readwrite) }
-attributes #14 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
+attributes #13 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
+attributes #14 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: readwrite) }
 attributes #15 = { nounwind }
 attributes #16 = { noreturn nounwind }
 
diff --git a/bench/typst-rs/optimized/4qskctz4kwc33g7b.ll b/bench/typst-rs/optimized/4qskctz4kwc33g7b.ll
index 4d8f1fb7d1e..cd21e3901e9 100644
--- a/bench/typst-rs/optimized/4qskctz4kwc33g7b.ll
+++ b/bench/typst-rs/optimized/4qskctz4kwc33g7b.ll
@@ -72477,8 +72477,8 @@ define hidden range(i48 100, 12884901888) i48 @_ZN5typst4text7variant17hf94498aa
   call void @llvm.lifetime.end.p0(i64 96, ptr nonnull %5), !noalias !15830
   call void @llvm.lifetime.end.p0(i64 88, ptr nonnull %6), !noalias !15817
   call void @llvm.lifetime.end.p0(i64 0, ptr nonnull %2), !noalias !15814
-  %.0.in.sroa.speculate.load.4.sroa.speculated.i = call i64 @llvm.smin.i64(i64 %56, i64 32767)
-  %spec.select = call i64 @llvm.smax.i64(i64 %.0.in.sroa.speculate.load.4.sroa.speculated.i, i64 -32768)
+  %.0.in.sroa.speculate.load.4.sroa.speculated.i = call i64 @llvm.smax.i64(i64 %56, i64 -32768)
+  %spec.select = call i64 @llvm.smin.i64(i64 %.0.in.sroa.speculate.load.4.sroa.speculated.i, i64 32767)
   %57 = trunc nsw i64 %spec.select to i16
   br label %58
 
diff --git a/bench/typst-rs/optimized/d6l9ieo9tcw33dn.ll b/bench/typst-rs/optimized/d6l9ieo9tcw33dn.ll
index e421a3f6ae2..c5a5264b077 100644
--- a/bench/typst-rs/optimized/d6l9ieo9tcw33dn.ll
+++ b/bench/typst-rs/optimized/d6l9ieo9tcw33dn.ll
@@ -175714,8 +175714,8 @@ define void @"_ZN94_$LT$typst..text..font..variant..FontWeight$u20$as$u20$typst.
   br label %31
 
 34:                                               ; preds = %26
-  %.0.in.sroa.speculate.load.4.sroa.speculated.i = tail call i64 @llvm.smin.i64(i64 %.sroa.041.sroa.0.0.copyload, i64 65535)
-  %.0.in.sroa.speculated.i = tail call i64 @llvm.smax.i64(i64 %.0.in.sroa.speculate.load.4.sroa.speculated.i, i64 0)
+  %.0.in.sroa.speculate.load.4.sroa.speculated.i = tail call i64 @llvm.smax.i64(i64 %.sroa.041.sroa.0.0.copyload, i64 0)
+  %.0.in.sroa.speculated.i = tail call i64 @llvm.umin.i64(i64 %.0.in.sroa.speculate.load.4.sroa.speculated.i, i64 65535)
   %35 = trunc nuw i64 %.0.in.sroa.speculated.i to i16
   %.0.sroa.speculated.i.i = tail call noundef i16 @llvm.umax.i16(i16 %35, i16 100)
   %.0.sroa.speculated.i1.i = tail call noundef range(i16 100, 901) i16 @llvm.umin.i16(i16 %.0.sroa.speculated.i.i, i16 900)
@@ -175739,7 +175739,7 @@ define void @"_ZN94_$LT$typst..text..font..variant..FontWeight$u20$as$u20$typst.
   %40 = landingpad { ptr, i32 }
           cleanup
   invoke void @"_ZN4core3ptr53drop_in_place$LT$typst..foundations..value..Value$GT$17h54cec887ca59b7d7E.llvm.7889846851399105414"(ptr noalias noundef nonnull align 8 dereferenceable(32) %1) #52
-          to label %38 unwind label %41
+          to label %40 unwind label %41
 
 41:                                               ; preds = %39
   %42 = landingpad { ptr, i32 }
@@ -200345,9 +200345,6 @@ declare i16 @llvm.abs.i16(i16, i1 immarg) #47
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare range(i8 -1, 2) i8 @llvm.scmp.i8.i64(i64, i64) #47
 
-; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i64 @llvm.smin.i64(i64, i64) #47
-
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.smax.i64(i64, i64) #47
 
diff --git a/bench/xgboost/optimized/extmem_quantile_dmatrix.ll b/bench/xgboost/optimized/extmem_quantile_dmatrix.ll
index 0b657713e45..0d25b158687 100644
--- a/bench/xgboost/optimized/extmem_quantile_dmatrix.ll
+++ b/bench/xgboost/optimized/extmem_quantile_dmatrix.ll
@@ -16072,10 +16072,10 @@ _ZN7xgboost4data25DefaultFormatStreamPolicyINS_16GHistIndexMatrixENS0_22GHistInd
           to label %11 unwind label %56
 
 11:                                               ; preds = %_ZN7xgboost4data25DefaultFormatStreamPolicyINS_16GHistIndexMatrixENS0_22GHistIndexFormatPolicyEEC2Ev.exit
-  %.sroa.speculated24 = call i32 @llvm.smin.i32(i32 %2, i32 16)
-  %.sroa.speculated = call i32 @llvm.smax.i32(i32 %.sroa.speculated24, i32 2)
+  %.sroa.speculated24 = call i32 @llvm.smax.i32(i32 %2, i32 2)
+  %.sroa.speculated = call i32 @llvm.umin.i32(i32 %.sroa.speculated24, i32 16)
   invoke void @_ZN7xgboost6common10ThreadPoolC2INS_13InitNewThreadEEENS_10StringViewEiOT_(ptr noundef nonnull align 8 dereferenceable(193) %10, ptr nonnull @.str.78, i64 7, i32 noundef %.sroa.speculated, ptr noundef nonnull align 4 dereferenceable(20) %6)
-          to label %12 unwind label %56
+          to label %14 unwind label %56
 
 12:                                               ; preds = %11
   call void @llvm.lifetime.end.p0(i64 20, ptr nonnull %6) #15
@@ -16220,7 +16220,7 @@ _ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE11_M_is_localEv.exit.i.i16
   br label %_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEED2Ev.exit18
 
 _ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEED2Ev.exit18: ; preds = %_ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE11_M_is_localEv.exit.i.i16, %_ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE11_M_is_localEv.exit.thread.i.i17, %60
-  %.pn = phi { ptr, i32 } [ %61, %60 ], [ %63, %_ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE11_M_is_localEv.exit.thread.i.i17 ], [ %63, %_ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE11_M_is_localEv.exit.i.i16 ]
+  %.pn = phi { ptr, i32 } [ %61, %62 ], [ %63, %_ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE11_M_is_localEv.exit.thread.i.i17 ], [ %63, %_ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE11_M_is_localEv.exit.i.i16 ]
   call void @_ZN7xgboost6common7MonitorD2Ev(ptr noundef nonnull align 8 dereferenceable(96) %33) #15
   %70 = load ptr, ptr %32, align 8, !tbaa !521
   %.not.i.i = icmp eq ptr %70, null
@@ -16244,13 +16244,13 @@ _ZNSt10unique_ptrISt6vectorISt6futureISt10shared_ptrIN7xgboost16GHistIndexMatrix
   br label %74
 
 74:                                               ; preds = %_ZNSt10unique_ptrISt6vectorISt6futureISt10shared_ptrIN7xgboost16GHistIndexMatrixEEESaIS6_EESt14default_deleteIS8_EED2Ev.exit, %58
-  %.pn.pn.pn = phi { ptr, i32 } [ %.pn, %_ZNSt10unique_ptrISt6vectorISt6futureISt10shared_ptrIN7xgboost16GHistIndexMatrixEEESaIS6_EESt14default_deleteIS8_EED2Ev.exit ], [ %59, %58 ]
+  %.pn.pn.pn = phi { ptr, i32 } [ %.pn, %_ZNSt10unique_ptrISt6vectorISt6futureISt10shared_ptrIN7xgboost16GHistIndexMatrixEEESaIS6_EESt14default_deleteIS8_EED2Ev.exit ], [ %59, %60 ]
   call void @_ZNSt12__shared_ptrIN7xgboost4data5CacheELN9__gnu_cxx12_Lock_policyE2EED2Ev(ptr noundef nonnull align 8 dereferenceable(16) %24) #15
   call void @_ZN7xgboost6common10ThreadPoolD2Ev(ptr noundef nonnull align 8 dereferenceable(193) %10) #15
   br label %75
 
 75:                                               ; preds = %74, %56
-  %.pn.pn.pn.pn = phi { ptr, i32 } [ %.pn.pn.pn, %74 ], [ %57, %56 ]
+  %.pn.pn.pn.pn = phi { ptr, i32 } [ %.pn.pn.pn, %76 ], [ %57, %58 ]
   %76 = getelementptr inbounds nuw i8, ptr %0, i64 80
   call void @_ZNSt12__shared_ptrIN7xgboost16GHistIndexMatrixELN9__gnu_cxx12_Lock_policyE2EED2Ev(ptr noundef nonnull align 8 dereferenceable(16) %76) #15
   call void @_ZN7xgboost4data22GHistIndexFormatPolicyINS_16GHistIndexMatrixEED2Ev(ptr noundef nonnull align 8 dereferenceable(32) %8) #15
@@ -20556,10 +20556,10 @@ declare i64 @llvm.smin.i64(i64, i64) #28
 declare i64 @llvm.umax.i64(i64, i64) #28
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smin.i32(i32, i32) #28
+declare i32 @llvm.smax.i32(i32, i32) #28
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
-declare i32 @llvm.smax.i32(i32, i32) #28
+declare i32 @llvm.umin.i32(i32, i32) #28
 
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare i64 @llvm.usub.sat.i64(i64, i64) #28
diff --git a/scripts/setup_pre_commit_patch.sh b/scripts/setup_pre_commit_patch.sh
index c5409e09ef4..37def5930e0 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/136665
 export COMPTIME_MODE=0
 
 # Please rebase manually