98d3a25f74
This follows similar reasoning as 45ce88758d24 (https://github.com/llvm/llvm-project/pull/159556): LV does not preserve LCSSA, it constructs it just before processing a loop to vectorize. Runtime check expressions are invariant to that loop, so expanding them should not break LCSSA form for the loop we are about to vectorize. LV creates SCEV and memory runtime checks early on and then disconnects the blocks temporarily. The patch fixes a mis-compile, where previously LCSSA construction during SCEV expand may replace uses in currently unreachable SCEV/memory check blocks. Fixes https://github.com/llvm/llvm-project/issues/162512 PR: https://github.com/llvm/llvm-project/pull/165505
82 lines
3.8 KiB
LLVM
82 lines
3.8 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 3
|
|
; RUN: opt -passes="print<scalar-evolution>,loop-vectorize" --verify-scev -S < %s -force-vector-width=4 2>/dev/null | FileCheck %s
|
|
|
|
; Make sure users of SCEVUnknowns from the scalar loop are invalidated.
|
|
|
|
define void @pr66616(ptr %ptr) {
|
|
; CHECK-LABEL: define void @pr66616(
|
|
; CHECK-SAME: ptr [[PTR:%.*]]) {
|
|
; CHECK-NEXT: entry:
|
|
; CHECK-NEXT: br label [[VECTOR_PH:%.*]]
|
|
; CHECK: vector.ph:
|
|
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
|
|
; CHECK: vector.body:
|
|
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
|
|
; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[PTR]], align 4
|
|
; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[TMP0]], 1
|
|
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
|
|
; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
|
|
; CHECK-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
|
|
; CHECK: middle.block:
|
|
; CHECK-NEXT: br label [[LOOP_1:%.*]]
|
|
; CHECK: preheader:
|
|
; CHECK-NEXT: [[TMP4:%.*]] = sub i32 -1, [[TMP0]]
|
|
; CHECK-NEXT: [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
|
|
; CHECK-NEXT: [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 1
|
|
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP6]], 4
|
|
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH1:%.*]], label [[VECTOR_PH2:%.*]]
|
|
; CHECK: vector.ph1:
|
|
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP6]], 4
|
|
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP6]], [[N_MOD_VF]]
|
|
; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i32
|
|
; CHECK-NEXT: [[TMP8:%.*]] = add i32 [[TMP3]], [[DOTCAST]]
|
|
; CHECK-NEXT: [[IND_END5:%.*]] = getelementptr i8, ptr [[PTR]], i64 [[N_VEC]]
|
|
; CHECK-NEXT: br label [[VECTOR_BODY3:%.*]]
|
|
; CHECK: vector.body2:
|
|
; CHECK-NEXT: [[INDEX8:%.*]] = phi i64 [ 0, [[VECTOR_PH2]] ], [ [[INDEX_NEXT9:%.*]], [[VECTOR_BODY3]] ]
|
|
; CHECK-NEXT: [[INDEX_NEXT9]] = add nuw i64 [[INDEX8]], 4
|
|
; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT9]], [[N_VEC]]
|
|
; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK5:%.*]], label [[VECTOR_BODY3]], !llvm.loop [[LOOP3:![0-9]+]]
|
|
; CHECK: middle.block5:
|
|
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP6]], [[N_VEC]]
|
|
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH1]]
|
|
; CHECK: scalar.ph:
|
|
; CHECK-NEXT: [[BC_RESUME_VAL4:%.*]] = phi i32 [ [[TMP8]], [[MIDDLE_BLOCK5]] ], [ [[TMP3]], [[LOOP_1]] ]
|
|
; CHECK-NEXT: [[BC_RESUME_VAL6:%.*]] = phi ptr [ [[IND_END5]], [[MIDDLE_BLOCK5]] ], [ [[PTR]], [[LOOP_1]] ]
|
|
; CHECK-NEXT: br label [[LOOP_2:%.*]]
|
|
; CHECK: loop.2:
|
|
; CHECK-NEXT: [[IV_2:%.*]] = phi i32 [ [[IV_2_I:%.*]], [[LOOP_2]] ], [ [[BC_RESUME_VAL4]], [[SCALAR_PH1]] ]
|
|
; CHECK-NEXT: [[IV_3:%.*]] = phi ptr [ [[IV_3_I:%.*]], [[LOOP_2]] ], [ [[BC_RESUME_VAL6]], [[SCALAR_PH1]] ]
|
|
; CHECK-NEXT: [[IV_2_I]] = add i32 [[IV_2]], 1
|
|
; CHECK-NEXT: [[IV_3_I]] = getelementptr i8, ptr [[IV_3]], i64 1
|
|
; CHECK-NEXT: [[COND2:%.*]] = icmp eq i32 [[IV_2]], 0
|
|
; CHECK-NEXT: br i1 [[COND2]], label [[EXIT]], label [[LOOP_2]], !llvm.loop [[LOOP4:![0-9]+]]
|
|
; CHECK: exit:
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
entry:
|
|
br label %loop.1
|
|
|
|
loop.1:
|
|
%iv.1 = phi i8 [ 0, %entry ], [ %inc, %loop.1 ]
|
|
%load = load i32, ptr %ptr, align 4
|
|
%add3 = add i32 %load, 1
|
|
%inc = add i8 %iv.1, 1
|
|
%cond1 = icmp eq i8 %inc, 0
|
|
br i1 %cond1, label %preheader, label %loop.1
|
|
|
|
preheader:
|
|
br label %loop.2
|
|
|
|
loop.2:
|
|
%iv.2 = phi i32 [ %iv.2.i, %loop.2 ], [ %add3, %preheader ]
|
|
%iv.3 = phi ptr [ %iv.3.i, %loop.2 ], [ %ptr, %preheader ]
|
|
%iv.2.i = add i32 %iv.2, 1
|
|
%iv.3.i = getelementptr i8, ptr %iv.3, i64 1
|
|
%cond2 = icmp eq i32 %iv.2, 0
|
|
br i1 %cond2, label %exit, label %loop.2
|
|
|
|
exit:
|
|
ret void
|
|
}
|