[llvm] 53a858f - [SLP][NFC]Add a test for incorrect skipping of shuffle instruction at

[Alexey Bataev via llvm-commits]

Author: Alexey Bataev
Date: 2023-01-04T10:17:03-08:00
New Revision: 53a858f7fc5325d66443fa5e3ddaa71e2bb08640

URL: https://github.com/llvm/llvm-project/commit/53a858f7fc5325d66443fa5e3ddaa71e2bb08640
DIFF: https://github.com/llvm/llvm-project/commit/53a858f7fc5325d66443fa5e3ddaa71e2bb08640.diff

LOG: [SLP][NFC]Add a test for incorrect skipping of shuffle instruction at
peek-through-shuffles, NFC.

Added: 
    llvm/test/Transforms/SLPVectorizer/X86/peek-through-shuffle.ll

Modified: 
    

Removed: 
    


################################################################################
diff  --git a/llvm/test/Transforms/SLPVectorizer/X86/peek-through-shuffle.ll b/llvm/test/Transforms/SLPVectorizer/X86/peek-through-shuffle.ll
new file mode 100644
index 000000000000..36cdffcbd5d1
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/X86/peek-through-shuffle.ll
@@ -0,0 +1,31 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -S < %s -mtriple=x86_64-unknown-linux-gnu -o - | FileCheck %s
+
+define void @foo(ptr %0, <4 x float> %1) {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr float, ptr null, i64 22
+; CHECK-NEXT:    [[TMP3:%.*]] = load float, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> zeroinitializer, float [[TMP3]], i64 0
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr float, ptr null, i64 23
+; CHECK-NEXT:    [[TMP6:%.*]] = load float, ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x float> [[TMP4]], float [[TMP6]], i64 1
+; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <4 x float> [[TMP1:%.*]], <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 5, i32 6, i32 undef>
+; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <4 x float> [[TMP8]], float [[TMP3]], i64 3
+; CHECK-NEXT:    [[TMP10:%.*]] = fpext <4 x float> [[TMP9]] to <4 x double>
+; CHECK-NEXT:    store <4 x double> [[TMP10]], ptr [[TMP0:%.*]], align 32
+; CHECK-NEXT:    ret void
+;
+entry:
+  %2 = getelementptr float, ptr null, i64 22
+  %3 = load float, ptr %2, align 8
+  %4 = insertelement <4 x float> zeroinitializer, float %3, i64 0
+  %5 = getelementptr float, ptr null, i64 23
+  %6 = load float, ptr %5, align 4
+  %7 = insertelement <4 x float> %4, float %6, i64 1
+  %8 = shufflevector <4 x float> %1, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 5, i32 6, i32 undef>
+  %9 = insertelement <4 x float> %8, float %3, i64 3
+  %10 = fpext <4 x float> %9 to <4 x double>
+  store <4 x double> %10, ptr %0, align 32
+  ret void
+}