[llvm] d2e7ee7 - [SLP]Do not check for clustered loads only

[Alexey Bataev via llvm-commits]

Author: Alexey Bataev
Date: 2024-10-24T08:16:42-07:00
New Revision: d2e7ee77d33e8b3be3b1d4e9bc5bc4c60b62b554

URL: https://github.com/llvm/llvm-project/commit/d2e7ee77d33e8b3be3b1d4e9bc5bc4c60b62b554
DIFF: https://github.com/llvm/llvm-project/commit/d2e7ee77d33e8b3be3b1d4e9bc5bc4c60b62b554.diff

LOG: [SLP]Do not check for clustered loads only

Since SLP support "clusterization" of the non-load instructions, the
restriction for reduced values for loads only should be removed to avoid
compiler crash.

Fixes #113516

Added: 
    llvm/test/Transforms/SLPVectorizer/X86/non-load-reduced-as-part-of-bv.ll

Modified: 
    llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index f76e913a57d52c..012d85353f085e 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -16118,7 +16118,7 @@ BoUpSLP::vectorizeTree(const ExtraValueToDebugLocsMap &ExternallyUsedValues,
     for (Instruction *I : RemovedInsts) {
       const TreeEntry *IE = getTreeEntry(I);
       if (IE->Idx != 0 &&
-          !(VectorizableTree.front()->isGather() && isa<LoadInst>(I) &&
+          !(VectorizableTree.front()->isGather() &&
             !IE->UserTreeIndices.empty() &&
             any_of(IE->UserTreeIndices,
                    [&](const EdgeInfo &EI) {

diff  --git a/llvm/test/Transforms/SLPVectorizer/X86/non-load-reduced-as-part-of-bv.ll b/llvm/test/Transforms/SLPVectorizer/X86/non-load-reduced-as-part-of-bv.ll
new file mode 100644
index 00000000000000..2f49a2e6a212e6
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/X86/non-load-reduced-as-part-of-bv.ll
@@ -0,0 +1,38 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s | FileCheck %s
+
+define i1 @foo() {
+; CHECK-LABEL: define i1 @foo() {
+; CHECK-NEXT:  [[ENTRY:.*:]]
+; CHECK-NEXT:    [[TOBOOL_NOT_NOT509_I_2329_I_I:%.*]] = icmp ne i32 0, 0
+; CHECK-NEXT:    [[STOREMERGE_2333_I_I:%.*]] = select i1 [[TOBOOL_NOT_NOT509_I_2329_I_I]], i32 0, i32 0
+; CHECK-NEXT:    [[TOBOOL_NOT_NOT509_I_1_2_I_I:%.*]] = icmp ne i32 [[STOREMERGE_2333_I_I]], 0
+; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <8 x i1> poison, i1 [[TOBOOL_NOT_NOT509_I_1_2_I_I]], i32 4
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <8 x i1> [[TMP0]], i1 [[TOBOOL_NOT_NOT509_I_2329_I_I]], i32 5
+; CHECK-NEXT:    [[TMP2:%.*]] = call <8 x i1> @llvm.vector.insert.v8i1.v4i1(<8 x i1> [[TMP1]], <4 x i1> zeroinitializer, i64 0)
+; CHECK-NEXT:    [[TMP3:%.*]] = call <8 x i1> @llvm.vector.insert.v8i1.v2i1(<8 x i1> [[TMP2]], <2 x i1> zeroinitializer, i64 6)
+; CHECK-NEXT:    [[TMP4:%.*]] = freeze <8 x i1> [[TMP3]]
+; CHECK-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.and.v8i1(<8 x i1> [[TMP4]])
+; CHECK-NEXT:    [[OP_RDX:%.*]] = select i1 false, i1 [[TMP5]], i1 false
+; CHECK-NEXT:    ret i1 [[OP_RDX]]
+;
+entry:
+  %tobool.not.not509.i.1311.i.i = icmp ne i32 0, 0
+  %cmp80.1319.i.i189 = icmp sgt i32 0, 0
+  %tobool.not.not509.i.1.1.i.i = icmp ne i32 0, 0
+  %cmp80.1.1.i.i190 = icmp sgt i32 0, 0
+  %tobool.not.not509.i.2329.i.i = icmp ne i32 0, 0
+  %storemerge.2333.i.i = select i1 %tobool.not.not509.i.2329.i.i, i32 0, i32 0
+  %cmp80.2337.i.i192 = icmp sgt i32 0, 0
+  %tobool.not.not509.i.1.2.i.i = icmp ne i32 %storemerge.2333.i.i, 0
+  %cmp80.1.2.i.i193 = icmp sgt i32 0, 0
+  %cmp80.1.2.i.i = select i1 %tobool.not.not509.i.1.2.i.i, i1 %cmp80.1.2.i.i193, i1 false
+  %0 = select i1 %cmp80.1.2.i.i, i1 %tobool.not.not509.i.2329.i.i, i1 false
+  %1 = select i1 %0, i1 %cmp80.2337.i.i192, i1 false
+  %2 = select i1 %1, i1 false, i1 false
+  %3 = select i1 %2, i1 %tobool.not.not509.i.1.1.i.i, i1 false
+  %4 = select i1 %3, i1 %cmp80.1.1.i.i190, i1 false
+  %5 = select i1 %4, i1 %tobool.not.not509.i.1311.i.i, i1 false
+  %6 = select i1 %5, i1 %cmp80.1319.i.i189, i1 false
+  ret i1 %6
+}