[llvm] f49e971 - [LV] Add historgram multi-use tests (NFC). (#214831)
via llvm-commits
llvm-commits at lists.llvm.org
Fri Aug 7 12:46:29 PDT 2026
Author: Florian Hahn
Date: 2026-08-07T19:46:24Z
New Revision: f49e97198399a4e900f37b3d5045eb89ab74c0f0
URL: https://github.com/llvm/llvm-project/commit/f49e97198399a4e900f37b3d5045eb89ab74c0f0
DIFF: https://github.com/llvm/llvm-project/commit/f49e97198399a4e900f37b3d5045eb89ab74c0f0.diff
LOG: [LV] Add historgram multi-use tests (NFC). (#214831)
Add test coverage for miscompiles with histograms and multiple uses of
bucket value and increment.
While touching the file, modernize the existing test + check lines as
well.
Added:
Modified:
llvm/test/Transforms/LoopVectorize/histograms.ll
Removed:
################################################################################
diff --git a/llvm/test/Transforms/LoopVectorize/histograms.ll b/llvm/test/Transforms/LoopVectorize/histograms.ll
index d15db60949746..78e315e015856 100644
--- a/llvm/test/Transforms/LoopVectorize/histograms.ll
+++ b/llvm/test/Transforms/LoopVectorize/histograms.ll
@@ -1,18 +1,15 @@
-; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 3
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 6
; RUN: opt < %s -passes=loop-vectorize -enable-histogram-loop-vectorization -force-vector-width=2 -S | FileCheck %s
-define void @simple_histogram(ptr noalias %buckets, ptr readonly %indices, i64 %N) {
+define void @simple_histogram(ptr noalias %buckets, ptr readonly %indices) {
; CHECK-LABEL: define void @simple_histogram(
-; CHECK-SAME: ptr noalias [[BUCKETS:%.*]], ptr readonly [[INDICES:%.*]], i64 [[N:%.*]]) {
-; CHECK-NEXT: entry:
-; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 2
-; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
-; CHECK: vector.ph:
-; CHECK-NEXT: [[N_MOD_VF:%.*]] = and i64 [[N]], 1
-; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-NEXT: br label [[FOR_BODY:%.*]]
-; CHECK: vector.body:
-; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-SAME: ptr noalias [[BUCKETS:%.*]], ptr readonly [[INDICES:%.*]]) {
+; CHECK-NEXT: [[ENTRY1:.*:]]
+; CHECK-NEXT: br label %[[ENTRY:.*]]
+; CHECK: [[ENTRY]]:
+; CHECK-NEXT: br label %[[FOR_BODY:.*]]
+; CHECK: [[FOR_BODY]]:
+; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[FOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[INDICES]], i64 [[IV]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP0]], align 4
; CHECK-NEXT: [[TMP1:%.*]] = zext <2 x i32> [[WIDE_LOAD]] to <2 x i64>
@@ -24,34 +21,18 @@ define void @simple_histogram(ptr noalias %buckets, ptr readonly %indices, i64 %
; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x ptr> [[TMP6]], ptr [[TMP5]], i32 1
; CHECK-NEXT: call void @llvm.experimental.vector.histogram.add.v2p0.i32(<2 x ptr> [[TMP7]], i32 1, <2 x i1> splat (i1 true))
; CHECK-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 2
-; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
-; CHECK: middle.block:
-; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
-; CHECK: scalar.ph:
-; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
-; CHECK-NEXT: br label [[FOR_BODY1:%.*]]
-; CHECK: for.body:
-; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ]
-; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[INDICES]], i64 [[IV1]]
-; CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
-; CHECK-NEXT: [[IDXPROM1:%.*]] = zext i32 [[TMP12]] to i64
-; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[BUCKETS]], i64 [[IDXPROM1]]
-; CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
-; CHECK-NEXT: [[INC:%.*]] = add nsw i32 [[TMP13]], 1
-; CHECK-NEXT: store i32 [[INC]], ptr [[ARRAYIDX2]], align 4
-; CHECK-NEXT: [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1
-; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N]]
-; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_EXIT]], label [[FOR_BODY1]], !llvm.loop [[LOOP3:![0-9]+]]
-; CHECK: for.exit:
+; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT: br i1 [[TMP8]], label %[[SCALAR_PH:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK: [[SCALAR_PH]]:
+; CHECK-NEXT: br label %[[FOR_BODY1:.*]]
+; CHECK: [[FOR_BODY1]]:
; CHECK-NEXT: ret void
;
entry:
- br label %for.body
+ br label %loop
-for.body:
- %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+loop:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%gep.indices = getelementptr inbounds i32, ptr %indices, i64 %iv
%l.idx = load i32, ptr %gep.indices, align 4
%idxprom1 = zext i32 %l.idx to i64
@@ -60,9 +41,124 @@ for.body:
%inc = add nsw i32 %l.bucket, 1
store i32 %inc, ptr %gep.bucket, align 4
%iv.next = add nuw nsw i64 %iv, 1
- %exitcond = icmp eq i64 %iv.next, %N
- br i1 %exitcond, label %for.exit, label %for.body
+ %exitcond = icmp eq i64 %iv.next, 1000
+ br i1 %exitcond, label %exit, label %loop
-for.exit:
+exit:
+ ret void
+}
+
+; FIXME: currently this gets vectorized incorrectly due to uses outside the
+; histogram chain.
+define void @histogram_extra_use_of_bucket_value(ptr noalias %buckets, ptr readonly %indices, ptr noalias %out) {
+; CHECK-LABEL: define void @histogram_extra_use_of_bucket_value(
+; CHECK-SAME: ptr noalias [[BUCKETS:%.*]], ptr readonly [[INDICES:%.*]], ptr noalias [[OUT:%.*]]) {
+; 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: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[INDICES]], i64 [[INDEX]]
+; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
+; CHECK-NEXT: [[TMP2:%.*]] = zext <2 x i32> [[WIDE_LOAD]] to <2 x i64>
+; CHECK-NEXT: [[TMP3:%.*]] = extractelement <2 x i64> [[TMP2]], i64 0
+; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[BUCKETS]], i64 [[TMP3]]
+; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x i64> [[TMP2]], i64 1
+; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[BUCKETS]], i64 [[TMP5]]
+; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x ptr> poison, ptr [[TMP4]], i32 0
+; CHECK-NEXT: [[TMP8:%.*]] = insertelement <2 x ptr> [[TMP7]], ptr [[TMP6]], i32 1
+; CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[TMP4]], align 4
+; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP6]], align 4
+; CHECK-NEXT: [[TMP11:%.*]] = insertelement <2 x i32> poison, i32 [[TMP9]], i32 0
+; CHECK-NEXT: [[TMP12:%.*]] = insertelement <2 x i32> [[TMP11]], i32 [[TMP10]], i32 1
+; CHECK-NEXT: call void @llvm.experimental.vector.histogram.add.v2p0.i32(<2 x ptr> [[TMP8]], i32 1, <2 x i1> splat (i1 true))
+; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[OUT]], i64 [[INDEX]]
+; CHECK-NEXT: store <2 x i32> [[TMP12]], ptr [[TMP13]], align 4
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT: br i1 [[TMP14]], label %[[SCALAR_PH:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK: [[SCALAR_PH]]:
+; CHECK-NEXT: br label %[[LOOP:.*]]
+; CHECK: [[LOOP]]:
+; CHECK-NEXT: ret void
+;
+entry:
+ br label %loop
+
+loop:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+ %gep.indices = getelementptr inbounds i32, ptr %indices, i64 %iv
+ %l.idx = load i32, ptr %gep.indices, align 4
+ %idxprom1 = zext i32 %l.idx to i64
+ %gep.bucket = getelementptr inbounds i32, ptr %buckets, i64 %idxprom1
+ %l.bucket = load i32, ptr %gep.bucket, align 4
+ %inc = add nsw i32 %l.bucket, 1
+ store i32 %inc, ptr %gep.bucket, align 4
+ %gep.out = getelementptr inbounds i32, ptr %out, i64 %iv
+ store i32 %l.bucket, ptr %gep.out, align 4
+ %iv.next = add nuw nsw i64 %iv, 1
+ %exitcond = icmp eq i64 %iv.next, 1000
+ br i1 %exitcond, label %exit, label %loop
+
+exit:
+ ret void
+}
+
+; FIXME: currently this gets vectorized incorrectly due to uses outside the
+; histogram chain.
+define void @histogram_extra_use_of_update(ptr noalias %buckets, ptr readonly %indices, ptr noalias %out) {
+; CHECK-LABEL: define void @histogram_extra_use_of_update(
+; CHECK-SAME: ptr noalias [[BUCKETS:%.*]], ptr readonly [[INDICES:%.*]], ptr noalias [[OUT:%.*]]) {
+; 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: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[INDICES]], i64 [[INDEX]]
+; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
+; CHECK-NEXT: [[TMP2:%.*]] = zext <2 x i32> [[WIDE_LOAD]] to <2 x i64>
+; CHECK-NEXT: [[TMP3:%.*]] = extractelement <2 x i64> [[TMP2]], i64 0
+; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[BUCKETS]], i64 [[TMP3]]
+; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x i64> [[TMP2]], i64 1
+; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[BUCKETS]], i64 [[TMP5]]
+; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x ptr> poison, ptr [[TMP4]], i32 0
+; CHECK-NEXT: [[TMP8:%.*]] = insertelement <2 x ptr> [[TMP7]], ptr [[TMP6]], i32 1
+; CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[TMP4]], align 4
+; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP6]], align 4
+; CHECK-NEXT: [[TMP11:%.*]] = insertelement <2 x i32> poison, i32 [[TMP9]], i32 0
+; CHECK-NEXT: [[TMP12:%.*]] = insertelement <2 x i32> [[TMP11]], i32 [[TMP10]], i32 1
+; CHECK-NEXT: [[TMP13:%.*]] = add nsw <2 x i32> [[TMP12]], splat (i32 1)
+; CHECK-NEXT: call void @llvm.experimental.vector.histogram.add.v2p0.i32(<2 x ptr> [[TMP8]], i32 1, <2 x i1> splat (i1 true))
+; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[OUT]], i64 [[INDEX]]
+; CHECK-NEXT: store <2 x i32> [[TMP13]], ptr [[TMP14]], align 4
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT: br i1 [[TMP15]], label %[[SCALAR_PH:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK: [[SCALAR_PH]]:
+; CHECK-NEXT: br label %[[LOOP:.*]]
+; CHECK: [[LOOP]]:
+; CHECK-NEXT: ret void
+;
+entry:
+ br label %loop
+
+loop:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+ %gep.indices = getelementptr inbounds i32, ptr %indices, i64 %iv
+ %l.idx = load i32, ptr %gep.indices, align 4
+ %idxprom1 = zext i32 %l.idx to i64
+ %gep.bucket = getelementptr inbounds i32, ptr %buckets, i64 %idxprom1
+ %l.bucket = load i32, ptr %gep.bucket, align 4
+ %inc = add nsw i32 %l.bucket, 1
+ store i32 %inc, ptr %gep.bucket, align 4
+ %gep.out = getelementptr inbounds i32, ptr %out, i64 %iv
+ store i32 %inc, ptr %gep.out, align 4
+ %iv.next = add nuw nsw i64 %iv, 1
+ %exitcond = icmp eq i64 %iv.next, 1000
+ br i1 %exitcond, label %exit, label %loop
+
+exit:
ret void
}
More information about the llvm-commits
mailing list