[LV] Add additional SCEV expansion tests for #147824.

Add additional test coverage for
llvm/llvm-project#147824.

Author: fhahn

llvm/test/Transforms/LoopVectorize/reuse-lcssa-phi-scev-expansion.ll

@@ -194,3 +194,215 @@ loop.2.latch:
 exit:
   ret void
 }
+
+
+declare void @foo()
+declare void @bar()
+
+define void @expand_diff_scev_unknown(ptr %dst, i1 %invar.c, i32 %step) mustprogress {
+; CHECK-LABEL: define void @expand_diff_scev_unknown(
+; CHECK-SAME: ptr [[DST:%.*]], i1 [[INVAR_C:%.*]], i32 [[STEP:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    br label %[[LOOP_1:.*]]
+; CHECK:       [[LOOP_1]]:
+; CHECK-NEXT:    [[INDVAR:%.*]] = phi i32 [ [[INDVAR_NEXT:%.*]], %[[LOOP_1]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[IV_1:%.*]] = phi i32 [ [[STEP]], %[[ENTRY]] ], [ [[IV_1_NEXT:%.*]], %[[LOOP_1]] ]
+; CHECK-NEXT:    call void @foo()
+; CHECK-NEXT:    [[IV_1_NEXT]] = add i32 [[IV_1]], 1
+; CHECK-NEXT:    [[INDVAR_NEXT]] = add i32 [[INDVAR]], 1
+; CHECK-NEXT:    br i1 [[INVAR_C]], label %[[LOOP_2_PREHEADER:.*]], label %[[LOOP_1]]
+; CHECK:       [[LOOP_2_PREHEADER]]:
+; CHECK-NEXT:    [[INDVAR_LCSSA1:%.*]] = phi i32 [ [[INDVAR]], %[[LOOP_1]] ]
+; CHECK-NEXT:    [[INDVAR_LCSSA:%.*]] = phi i32 [ [[INDVAR]], %[[LOOP_1]] ]
+; CHECK-NEXT:    [[IV_1_LCSSA:%.*]] = phi i32 [ [[IV_1]], %[[LOOP_1]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = shl i32 [[STEP]], 1
+; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[INDVAR_LCSSA]], [[TMP0]]
+; CHECK-NEXT:    [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[TMP1]], i32 0)
+; CHECK-NEXT:    [[TMP2:%.*]] = mul i32 [[STEP]], -2
+; CHECK-NEXT:    [[TMP3:%.*]] = mul i32 [[INDVAR_LCSSA1]], -1
+; CHECK-NEXT:    [[TMP4:%.*]] = add i32 [[TMP3]], [[TMP2]]
+; CHECK-NEXT:    [[TMP5:%.*]] = add i32 [[SMAX]], [[TMP4]]
+; CHECK-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP5]], i32 1)
+; CHECK-NEXT:    [[TMP6:%.*]] = add i32 [[UMIN]], 1
+; CHECK-NEXT:    [[TMP7:%.*]] = sub i32 [[TMP5]], [[UMIN]]
+; CHECK-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[STEP]], i32 1)
+; CHECK-NEXT:    [[TMP8:%.*]] = udiv i32 [[TMP7]], [[UMAX]]
+; CHECK-NEXT:    [[TMP9:%.*]] = add i32 [[TMP6]], [[TMP8]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP9]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_SCEVCHECK:.*]]
+; CHECK:       [[VECTOR_SCEVCHECK]]:
+; CHECK-NEXT:    [[IDENT_CHECK:%.*]] = icmp ne i32 [[STEP]], 1
+; CHECK-NEXT:    br i1 [[IDENT_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP9]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP9]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[IV_1_LCSSA]], [[N_VEC]]
+; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
+; CHECK:       [[VECTOR_BODY]]:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[IV_1_LCSSA]], [[INDEX]]
+; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[DST]], i32 [[OFFSET_IDX]]
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
+; CHECK-NEXT:    store <2 x i32> zeroinitializer, ptr [[TMP12]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK:       [[MIDDLE_BLOCK]]:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP9]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK:       [[SCALAR_PH]]:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[TMP10]], %[[MIDDLE_BLOCK]] ], [ [[IV_1_LCSSA]], %[[LOOP_2_PREHEADER]] ], [ [[IV_1_LCSSA]], %[[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    br label %[[LOOP_2:.*]]
+; CHECK:       [[LOOP_2]]:
+; CHECK-NEXT:    [[IV_2:%.*]] = phi i32 [ [[IV_2_NEXT:%.*]], %[[LOOP_2]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
+; CHECK-NEXT:    [[IV_2_NEXT]] = add nsw i32 [[IV_2]], [[STEP]]
+; CHECK-NEXT:    [[GEP_DST:%.*]] = getelementptr inbounds i32, ptr [[DST]], i32 [[IV_2]]
+; CHECK-NEXT:    store i32 0, ptr [[GEP_DST]], align 4
+; CHECK-NEXT:    [[EC_2:%.*]] = icmp slt i32 [[IV_2_NEXT]], 0
+; CHECK-NEXT:    br i1 [[EC_2]], label %[[LOOP_2]], label %[[EXIT]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK:       [[EXIT]]:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %loop.1
+
+loop.1:
+  %iv.1 = phi i32 [ %step, %entry ], [ %iv.1.next, %loop.1 ]
+  call void @foo()
+  %iv.1.next = add i32 %iv.1, 1
+  br i1 %invar.c, label %loop.2, label %loop.1
+
+loop.2:
+  %iv.2 = phi i32 [ %iv.1, %loop.1 ], [ %iv.2.next, %loop.2 ]
+  %iv.2.next = add nsw i32 %iv.2, %step
+  %gep.dst = getelementptr inbounds i32, ptr %dst, i32 %iv.2
+  store i32 0, ptr %gep.dst
+  %ec.2 = icmp slt i32 %iv.2.next, 0
+  br i1 %ec.2, label %loop.2, label %exit
+
+exit:
+  ret void
+}
+
+define void @expand_diff_neg_ptrtoint_expr(ptr %src, ptr %start) {
+; CHECK-LABEL: define void @expand_diff_neg_ptrtoint_expr(
+; CHECK-SAME: ptr [[SRC:%.*]], ptr [[START:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    [[SRC2:%.*]] = ptrtoint ptr [[SRC]] to i64
+; CHECK-NEXT:    [[START1:%.*]] = ptrtoint ptr [[START]] to i64
+; CHECK-NEXT:    br label %[[LOOP_1:.*]]
+; CHECK:       [[LOOP_1]]:
+; CHECK-NEXT:    [[INDVAR:%.*]] = phi i64 [ [[INDVAR_NEXT:%.*]], %[[LOOP_1]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 1, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP_1]] ]
+; CHECK-NEXT:    [[PTR_IV_1:%.*]] = phi ptr [ [[START]], %[[ENTRY]] ], [ [[PTR_IV_1_NEXT:%.*]], %[[LOOP_1]] ]
+; CHECK-NEXT:    [[PTR_IV_1_NEXT]] = getelementptr i8, ptr [[PTR_IV_1]], i64 8
+; CHECK-NEXT:    call void @foo()
+; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
+; CHECK-NEXT:    [[EC_1:%.*]] = icmp eq i64 [[IV_NEXT]], 32
+; CHECK-NEXT:    [[INDVAR_NEXT]] = add i64 [[INDVAR]], 1
+; CHECK-NEXT:    br i1 [[EC_1]], label %[[LOOP_2_PREHEADER:.*]], label %[[LOOP_1]]
+; CHECK:       [[LOOP_2_PREHEADER]]:
+; CHECK-NEXT:    [[INDVAR_LCSSA:%.*]] = phi i64 [ [[INDVAR]], %[[LOOP_1]] ]
+; CHECK-NEXT:    [[PTR_IV_1_NEXT_LCSSA:%.*]] = phi ptr [ [[PTR_IV_1_NEXT]], %[[LOOP_1]] ]
+; CHECK-NEXT:    br label %[[LOOP_2:.*]]
+; CHECK:       [[LOOP_2]]:
+; CHECK-NEXT:    [[INDVAR3:%.*]] = phi i64 [ 0, %[[LOOP_2_PREHEADER]] ], [ [[INDVAR_NEXT4:%.*]], %[[LOOP_2]] ]
+; CHECK-NEXT:    [[IV_1:%.*]] = phi i64 [ [[IV_NEXT_1:%.*]], %[[LOOP_2]] ], [ 1, %[[LOOP_2_PREHEADER]] ]
+; CHECK-NEXT:    [[PTR_IV_2:%.*]] = phi ptr [ [[PTR_IV_2_NEXT:%.*]], %[[LOOP_2]] ], [ [[PTR_IV_1_NEXT_LCSSA]], %[[LOOP_2_PREHEADER]] ]
+; CHECK-NEXT:    call void @bar()
+; CHECK-NEXT:    [[PTR_IV_2_NEXT]] = getelementptr i8, ptr [[PTR_IV_2]], i64 8
+; CHECK-NEXT:    [[IV_NEXT_1]] = add i64 [[IV_1]], 1
+; CHECK-NEXT:    [[EC_2:%.*]] = icmp eq i64 [[IV_NEXT_1]], 32
+; CHECK-NEXT:    [[INDVAR_NEXT4]] = add i64 [[INDVAR3]], 1
+; CHECK-NEXT:    br i1 [[EC_2]], label %[[LOOP_3_PREHEADER:.*]], label %[[LOOP_2]]
+; CHECK:       [[LOOP_3_PREHEADER]]:
+; CHECK-NEXT:    [[INDVAR3_LCSSA:%.*]] = phi i64 [ [[INDVAR3]], %[[LOOP_2]] ]
+; CHECK-NEXT:    [[PTR_IV_2_NEXT_LCSSA:%.*]] = phi ptr [ [[PTR_IV_2_NEXT]], %[[LOOP_2]] ]
+; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; CHECK:       [[VECTOR_MEMCHECK]]:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[START1]], 16
+; CHECK-NEXT:    [[TMP1:%.*]] = sub i64 [[TMP0]], [[SRC2]]
+; CHECK-NEXT:    [[TMP2:%.*]] = shl nuw nsw i64 [[INDVAR_LCSSA]], 3
+; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[TMP2]], [[TMP1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = shl nuw nsw i64 [[INDVAR3_LCSSA]], 3
+; CHECK-NEXT:    [[TMP5:%.*]] = add i64 [[TMP4]], [[TMP3]]
+; CHECK-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP5]], 16
+; CHECK-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[PTR_IV_2_NEXT_LCSSA]], i64 -16
+; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
+; CHECK:       [[VECTOR_BODY]]:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
+; CHECK-NEXT:    [[OFFSET_IDX5:%.*]] = mul i64 [[INDEX]], 8
+; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[PTR_IV_2_NEXT_LCSSA]], i64 [[OFFSET_IDX5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = add i64 [[OFFSET_IDX]], -1
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr double, ptr [[SRC]], i64 [[TMP7]]
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[TMP8]], i32 0
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP9]], align 8
+; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
+; CHECK-NEXT:    store <2 x i64> [[WIDE_LOAD]], ptr [[TMP10]], align 8
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], -2
+; CHECK-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK:       [[MIDDLE_BLOCK]]:
+; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK:       [[SCALAR_PH]]:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, %[[MIDDLE_BLOCK]] ], [ 1, %[[LOOP_3_PREHEADER]] ], [ 1, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL6:%.*]] = phi ptr [ [[TMP6]], %[[MIDDLE_BLOCK]] ], [ [[PTR_IV_2_NEXT_LCSSA]], %[[LOOP_3_PREHEADER]] ], [ [[PTR_IV_2_NEXT_LCSSA]], %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    br label %[[LOOP_3:.*]]
+; CHECK:       [[LOOP_3]]:
+; CHECK-NEXT:    [[IV_2:%.*]] = phi i64 [ [[IV_NEXT_2:%.*]], %[[LOOP_3]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
+; CHECK-NEXT:    [[PTR_IV_3:%.*]] = phi ptr [ [[PTR_IV_3_NEXT:%.*]], %[[LOOP_3]] ], [ [[BC_RESUME_VAL6]], %[[SCALAR_PH]] ]
+; CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[IV_2]], -1
+; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr double, ptr [[SRC]], i64 [[TMP12]]
+; CHECK-NEXT:    [[L:%.*]] = load i64, ptr [[GEP_SRC]], align 8
+; CHECK-NEXT:    [[PTR_IV_3_NEXT]] = getelementptr i8, ptr [[PTR_IV_3]], i64 8
+; CHECK-NEXT:    store i64 [[L]], ptr [[PTR_IV_3]], align 8
+; CHECK-NEXT:    [[IV_NEXT_2]] = add i64 [[IV_2]], 1
+; CHECK-NEXT:    [[EC_3:%.*]] = icmp eq i64 [[IV_NEXT_2]], 0
+; CHECK-NEXT:    br i1 [[EC_3]], label %[[EXIT]], label %[[LOOP_3]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK:       [[EXIT]]:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %loop.1
+
+loop.1:
+  %iv = phi i64 [ 1, %entry ], [ %iv.next, %loop.1 ]
+  %ptr.iv.1 = phi ptr [ %start, %entry ], [ %ptr.iv.1.next, %loop.1 ]
+  %ptr.iv.1.next = getelementptr i8, ptr %ptr.iv.1, i64 8
+  call void @foo()
+  %iv.next = add i64 %iv, 1
+  %ec.1 = icmp eq i64 %iv.next, 32
+  br i1 %ec.1, label %loop.2, label %loop.1
+
+loop.2:
+  %iv.1 = phi i64 [ 1, %loop.1 ], [ %iv.next.1, %loop.2 ]
+  %ptr.iv.2 = phi ptr [ %ptr.iv.1.next, %loop.1 ], [ %ptr.iv.2.next, %loop.2 ]
+  call void @bar()
+  %ptr.iv.2.next = getelementptr i8, ptr %ptr.iv.2, i64 8
+  %iv.next.1 = add i64 %iv.1, 1
+  %ec.2 = icmp eq i64 %iv.next.1, 32
+  br i1 %ec.2, label %loop.3, label %loop.2
+
+loop.3:
+  %iv.2 = phi i64 [ 1, %loop.2 ], [ %iv.next.2, %loop.3 ]
+  %ptr.iv.3 = phi ptr [ %ptr.iv.2.next, %loop.2 ], [ %ptr.iv.3.next, %loop.3 ]
+  %6 = add i64 %iv.2, -1
+  %gep.src = getelementptr double, ptr %src, i64 %6
+  %l = load i64, ptr %gep.src, align 8
+  %ptr.iv.3.next = getelementptr i8, ptr %ptr.iv.3, i64 8
+  store i64 %l, ptr %ptr.iv.3, align 8
+  %iv.next.2 = add i64 %iv.2, 1
+  %ec.3 = icmp eq i64 %iv.next.2, 0
+  br i1 %ec.3, label %exit, label %loop.3
+
+exit:
+  ret void
+}
+
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare double @llvm.cos.f64(double) #0
+
+attributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }