[llvm-branch-commits] [llvm] release/23.x: [VPlan] Bail out on FindLast reductions with extra phi users. (#230676) (PR #239574)
via llvm-branch-commits
llvm-branch-commits at lists.llvm.org
Fri Nov 6 04:01:41 PST 2026
llvmorg-github-actions[bot] wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-llvm-vectorizers
@llvm/pr-subscribers-llvm-transforms
Author: Nikita Popov (nikic)
<details>
<summary>Changes</summary>
(cherry picked from commit 2f272fc322cdc1fe1fbb667cd2cb6abddeb0eb9b)
(cherry picked from commit 79caa03343df4da7b8ae0338d932125ef4bcf20b)
Fixes a miscompile encountered in the wild.
---
Patch is 29.96 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/239574.diff
2 Files Affected:
- (modified) llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp (+8)
- (modified) llvm/test/Transforms/LoopVectorize/find-last.ll (+234-161)
``````````diff
diff --git a/llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp b/llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp
index 1a135b85495146..4e68c58048204b 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp
@@ -1864,6 +1864,14 @@ bool VPlanTransforms::handleFindLastReductions(VPlan &Plan) {
!MatchBlend(SelectR))
return false;
+ // Bail out if PhiR has users other than the find-last select/blend and the
+ // header mask select, e.g. nested blends. Then the data operand may be PhiR
+ // on some paths and Cond does not determine the lanes to update.
+ if (any_of(PhiR->users(), [&](VPUser *U) {
+ return U != SelectR && U != BackedgeSelect->getDefiningRecipe();
+ }))
+ return false;
+
assert(Cond != HeaderMask && "Cond must not be HeaderMask");
// Find final reduction computation and replace it with an
diff --git a/llvm/test/Transforms/LoopVectorize/find-last.ll b/llvm/test/Transforms/LoopVectorize/find-last.ll
index 1da5657c281ef2..78070f6cedd403 100644
--- a/llvm/test/Transforms/LoopVectorize/find-last.ll
+++ b/llvm/test/Transforms/LoopVectorize/find-last.ll
@@ -1,22 +1,23 @@
-; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s | FileCheck %s
-; RUN: opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -prefer-inloop-reductions -S %s | FileCheck %s
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --filter-out-after "^loop(\.header)?:" --version 6
+; RUN: opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -force-target-supports-masked-memory-ops -S %s | FileCheck %s
+; RUN: opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -force-target-supports-masked-memory-ops -prefer-inloop-reductions -S %s | FileCheck %s
; Test vectorization of a "find last" pattern where a scalar value is
; conditionally updated based on comparing elements from two arrays.
define i32 @find_last_with_select(ptr noalias %a, ptr noalias %b) {
-; CHECK-LABEL: @find_last_with_select(
-; 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 i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 1), [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP0:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-LABEL: define i32 @find_last_with_select(
+; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]]) {
+; 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 i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 1), %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP0:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
-; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
+; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
; CHECK-NEXT: [[TMP3:%.*]] = icmp sge <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
; CHECK-NEXT: [[TMP4:%.*]] = freeze <4 x i1> [[TMP3]]
@@ -25,11 +26,11 @@ define i32 @find_last_with_select(ptr noalias %a, ptr noalias %b) {
; CHECK-NEXT: [[TMP7]] = select i1 [[TMP5]], <4 x i32> [[WIDE_LOAD]], <4 x i32> [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
-; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
-; CHECK: middle.block:
+; CHECK-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP9:%.*]] = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> [[TMP7]], <4 x i1> [[TMP6]], i32 1)
-; CHECK-NEXT: br label [[EXIT:%.*]]
-; CHECK: exit:
+; CHECK-NEXT: br label %[[EXIT:.*]]
+; CHECK: [[EXIT]]:
; CHECK-NEXT: ret i32 [[TMP9]]
;
entry:
@@ -54,18 +55,19 @@ exit:
; Keep when a[i] > b[i] (signed), i.e., update when a[i] <= b[i].
define i32 @find_last_sgt_inverted(ptr noalias %a, ptr noalias %b) {
-; CHECK-LABEL: @find_last_sgt_inverted(
-; 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 i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 1), [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP0:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-LABEL: define i32 @find_last_sgt_inverted(
+; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]]) {
+; 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 i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 1), %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP0:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
-; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
+; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
; CHECK-NEXT: [[TMP3:%.*]] = icmp sle <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
; CHECK-NEXT: [[TMP4:%.*]] = freeze <4 x i1> [[TMP3]]
@@ -74,11 +76,11 @@ define i32 @find_last_sgt_inverted(ptr noalias %a, ptr noalias %b) {
; CHECK-NEXT: [[TMP7]] = select i1 [[TMP5]], <4 x i32> [[WIDE_LOAD]], <4 x i32> [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
-; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
-; CHECK: middle.block:
+; CHECK-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP9:%.*]] = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> [[TMP7]], <4 x i1> [[TMP6]], i32 1)
-; CHECK-NEXT: br label [[EXIT:%.*]]
-; CHECK: exit:
+; CHECK-NEXT: br label %[[EXIT:.*]]
+; CHECK: [[EXIT]]:
; CHECK-NEXT: ret i32 [[TMP9]]
;
entry:
@@ -103,18 +105,19 @@ exit:
; Keep when a[i] < b[i] (unsigned), i.e., update when a[i] >= b[i].
define i32 @find_last_ult_inverted(ptr noalias %a, ptr noalias %b) {
-; CHECK-LABEL: @find_last_ult_inverted(
-; 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 i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 1), [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP0:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-LABEL: define i32 @find_last_ult_inverted(
+; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]]) {
+; 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 i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 1), %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP0:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
-; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
+; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
; CHECK-NEXT: [[TMP3:%.*]] = icmp uge <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
; CHECK-NEXT: [[TMP4:%.*]] = freeze <4 x i1> [[TMP3]]
@@ -123,11 +126,11 @@ define i32 @find_last_ult_inverted(ptr noalias %a, ptr noalias %b) {
; CHECK-NEXT: [[TMP7]] = select i1 [[TMP5]], <4 x i32> [[WIDE_LOAD]], <4 x i32> [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
-; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
-; CHECK: middle.block:
+; CHECK-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP9:%.*]] = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> [[TMP7]], <4 x i1> [[TMP6]], i32 1)
-; CHECK-NEXT: br label [[EXIT:%.*]]
-; CHECK: exit:
+; CHECK-NEXT: br label %[[EXIT:.*]]
+; CHECK: [[EXIT]]:
; CHECK-NEXT: ret i32 [[TMP9]]
;
entry:
@@ -152,18 +155,19 @@ exit:
; Keep when a[i] != b[i], i.e., update when a[i] == b[i].
define i32 @find_last_ne_inverted(ptr noalias %a, ptr noalias %b) {
-; CHECK-LABEL: @find_last_ne_inverted(
-; 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 i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 1), [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP0:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-LABEL: define i32 @find_last_ne_inverted(
+; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]]) {
+; 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 i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 1), %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP0:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
-; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
+; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
; CHECK-NEXT: [[TMP3:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
; CHECK-NEXT: [[TMP4:%.*]] = freeze <4 x i1> [[TMP3]]
@@ -172,11 +176,11 @@ define i32 @find_last_ne_inverted(ptr noalias %a, ptr noalias %b) {
; CHECK-NEXT: [[TMP7]] = select i1 [[TMP5]], <4 x i32> [[WIDE_LOAD]], <4 x i32> [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
-; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
-; CHECK: middle.block:
+; CHECK-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP9:%.*]] = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> [[TMP7]], <4 x i1> [[TMP6]], i32 1)
-; CHECK-NEXT: br label [[EXIT:%.*]]
-; CHECK: exit:
+; CHECK-NEXT: br label %[[EXIT:.*]]
+; CHECK: [[EXIT]]:
; CHECK-NEXT: ret i32 [[TMP9]]
;
entry:
@@ -200,56 +204,48 @@ exit:
}
define i32 @loop_inv_select_condition_issue_182152(i32 %iv.start, i32 %a, i32 %b) {
-; CHECK-LABEL: @loop_inv_select_condition_issue_182152(
-; CHECK-NEXT: entry:
-; CHECK-NEXT: [[INV_COND:%.*]] = icmp eq i32 [[A:%.*]], [[B:%.*]]
-; CHECK-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[IV_START:%.*]], i32 1)
-; CHECK-NEXT: [[TMP1:%.*]] = sub i32 [[SMAX]], [[IV_START]]
-; CHECK-NEXT: [[TMP11:%.*]] = add i32 [[TMP1]], 1
-; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP11]], 4
-; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
-; CHECK: vector.ph:
-; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP11]], 4
-; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP11]], [[N_MOD_VF]]
+; CHECK-LABEL: define i32 @loop_inv_select_condition_issue_182152(
+; CHECK-SAME: i32 [[IV_START:%.*]], i32 [[A:%.*]], i32 [[B:%.*]]) {
+; CHECK-NEXT: [[ENTRY:.*]]:
+; CHECK-NEXT: [[INV_COND:%.*]] = icmp eq i32 [[A]], [[B]]
+; CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.smax.i32(i32 [[IV_START]], i32 1)
+; CHECK-NEXT: [[TMP1:%.*]] = sub i32 [[TMP0]], [[IV_START]]
+; CHECK-NEXT: [[TMP2:%.*]] = add i32 [[TMP1]], 1
+; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP2]], 4
+; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK: [[VECTOR_PH]]:
+; CHECK-NEXT: [[TMP3:%.*]] = urem i32 [[TMP2]], 4
+; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP2]], [[TMP3]]
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[INV_COND]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i1> [[BROADCAST_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP2:%.*]] = add i32 [[IV_START]], [[N_VEC]]
-; CHECK-NEXT: [[TMP3:%.*]] = xor <4 x i1> [[BROADCAST_SPLAT]], splat (i1 true)
-; CHECK-NEXT: [[TMP4:%.*]] = freeze <4 x i1> [[TMP3]]
-; CHECK-NEXT: [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP4]])
+; CHECK-NEXT: [[TMP4:%.*]] = add i32 [[IV_START]], [[N_VEC]]
+; CHECK-NEXT: [[TMP5:%.*]] = xor <4 x i1> [[BROADCAST_SPLAT]], splat (i1 true)
+; CHECK-NEXT: [[TMP6:%.*]] = freeze <4 x i1> [[TMP5]]
+; CHECK-NEXT: [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
; CHECK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i32> poison, i32 [[IV_START]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT1]], <4 x i32> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[INDUCTION:%.*]] = add <4 x i32> [[BROADCAST_SPLAT2]], <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT: br label [[LOOP:%.*]]
-; CHECK: vector.body:
-; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[LOOP]] ]
-; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[LOOP]] ]
-; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[LOOP]] ]
-; CHECK-NEXT: [[TMP6:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[LOOP]] ]
-; CHECK-NEXT: [[TMP7]] = select i1 [[TMP5]], <4 x i1> [[TMP3]], <4 x i1> [[TMP6]]
-; CHECK-NEXT: [[TMP8]] = select i1 [[TMP5]], <4 x i32> [[VEC_IND]], <4 x i32> [[VEC_PHI]]
+; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
+; CHECK: [[VECTOR_BODY]]:
+; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP10:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP8:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP9:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP9]] = select i1 [[TMP7]], <4 x i1> [[TMP5]], <4 x i1> [[TMP8]]
+; CHECK-NEXT: [[TMP10]] = select i1 [[TMP7]], <4 x i32> [[VEC_IND]], <4 x i32> [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
-; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP]], !llvm.loop [[LOOP6:![0-9]+]]
-; CHECK: middle.block:
-; CHECK-NEXT: [[TMP10:%.*]] = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> [[TMP8]], <4 x i1> [[TMP7]], i32 0)
-; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP11]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
-; CHECK: scalar.ph:
-; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[TMP2]], [[MIDDLE_BLOCK]] ], [ [[IV_START]], [[ENTRY:%.*]] ]
-; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP10]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
-; CHECK-NEXT: br label [[LOOP1:%.*]]
-; CHECK: loop:
-; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP1]] ]
-; CHECK-NEXT: [[RDX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[SELECT:%.*]], [[LOOP1]] ]
-; CHECK-NEXT: [[SELECT]] = select i1 [[INV_COND]], i32 [[RDX]], i32 [[IV]]
-; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1
-; CHECK-NEXT: [[EXIT_COND:%.*]] = icmp sgt i32 [[IV]], 0
-; CHECK-NEXT: br i1 [[EXIT_COND]], label [[EXIT]], label [[LOOP1]], !llvm.loop [[LOOP7:![0-9]+]]
-; CHECK: exit:
-; CHECK-NEXT: [[SELECT_LCSSA:%.*]] = phi i32 [ [[SELECT]], [[LOOP1]] ], [ [[TMP10]], [[MIDDLE_BLOCK]] ]
-; CHECK-NEXT: ret i32 [[SELECT_LCSSA]]
+; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK: [[MIDDLE_BLOCK]]:
+; CHECK-NEXT: [[TMP12:%.*]] = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> [[TMP10]], <4 x i1> [[TMP9]], i32 0)
+; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP2]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[CMP_N]], [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK: [[SCALAR_PH]]:
+; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[TMP4]], %[[MIDDLE_BLOCK]] ], [ [[IV_START]], %[[ENTRY]] ]
+; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP12]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT: br label %[[LOOP:.*]]
+; CHECK: [[LOOP]]:
;
entry:
%inv_cond = icmp eq i32 %a, %b
@@ -268,25 +264,26 @@ exit:
}
define ptr @loop_inv_select_condition_issue_185682(i32 %a, i32 %b) {
-; CHECK-LABEL: @loop_inv_select_cond...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/239574
More information about the llvm-branch-commits
mailing list