AVX-512: fixed extended load to 512-bit register

[oota-llvm.git] / test / CodeGen / X86 / pr12312.ll

diff --git a/test/CodeGen/X86/pr12312.ll b/test/CodeGen/X86/pr12312.ll
index 74733acc617a331ced2601ae09e1f736a4c54972..81aaf91f26887e4bd1b99eb8060899e5ab400915 100644 (file)
--- a/test/CodeGen/X86/pr12312.ll
+++ b/test/CodeGen/X86/pr12312.ll
@@ -1,7 +1,7 @@
-; RUN: llc -march=x86-64 -mattr=+sse41,-avx < %s | FileCheck %s --check-prefix SSE41
-; RUN: llc -march=x86-64 -mattr=+avx < %s | FileCheck %s --check-prefix AVX
+; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=+sse4.1,-avx < %s | FileCheck %s --check-prefix SSE41
+; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=+avx,-avx2 < %s | FileCheck %s --check-prefix AVX
 
-define i32 @veccond(<4 x i32> %input) {
+define i32 @veccond128(<4 x i32> %input) {
 entry:
   %0 = bitcast <4 x i32> %input to i128
   %1 = icmp ne i128 %0, 0
@@ -11,38 +11,145 @@ if-true-block:                                    ; preds = %entry
   ret i32 0
 endif-block:                                      ; preds = %entry,
   ret i32 1
-; SSE41: veccond
+; SSE41: veccond128
 ; SSE41: ptest
 ; SSE41: ret
-; AVX:   veccond
-; AVX:   vptest
+; AVX:   veccond128
+; AVX:   vptest %xmm{{.*}}, %xmm{{.*}}
 ; AVX:   ret
 }
 
-define i32 @vectest(<4 x i32> %input) {
+define i32 @veccond256(<8 x i32> %input) {
+entry:
+  %0 = bitcast <8 x i32> %input to i256
+  %1 = icmp ne i256 %0, 0
+  br i1 %1, label %if-true-block, label %endif-block
+
+if-true-block:                                    ; preds = %entry
+  ret i32 0
+endif-block:                                      ; preds = %entry,
+  ret i32 1
+; SSE41: veccond256
+; SSE41: por
+; SSE41: ptest
+; SSE41: ret
+; AVX:   veccond256
+; AVX:   vptest %ymm{{.*}}, %ymm{{.*}}
+; AVX:   ret
+}
+
+define i32 @veccond512(<16 x i32> %input) {
+entry:
+  %0 = bitcast <16 x i32> %input to i512
+  %1 = icmp ne i512 %0, 0
+  br i1 %1, label %if-true-block, label %endif-block
+
+if-true-block:                                    ; preds = %entry
+  ret i32 0
+endif-block:                                      ; preds = %entry,
+  ret i32 1
+; SSE41: veccond512
+; SSE41: por
+; SSE41: por
+; SSE41: por
+; SSE41: ptest
+; SSE41: ret
+; AVX:   veccond512
+; AVX:   vorps
+; AVX:   vptest %ymm{{.*}}, %ymm{{.*}}
+; AVX:   ret
+}
+
+define i32 @vectest128(<4 x i32> %input) {
 entry:
   %0 = bitcast <4 x i32> %input to i128
   %1 = icmp ne i128 %0, 0
   %2 = zext i1 %1 to i32
   ret i32 %2
-; SSE41: vectest
+; SSE41: vectest128
+; SSE41: ptest
+; SSE41: ret
+; AVX:   vectest128
+; AVX:   vptest %xmm{{.*}}, %xmm{{.*}}
+; AVX:   ret
+}
+
+define i32 @vectest256(<8 x i32> %input) {
+entry:
+  %0 = bitcast <8 x i32> %input to i256
+  %1 = icmp ne i256 %0, 0
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+; SSE41: vectest256
+; SSE41: por
+; SSE41: ptest
+; SSE41: ret
+; AVX:   vectest256
+; AVX:   vptest %ymm{{.*}}, %ymm{{.*}}
+; AVX:   ret
+}
+
+define i32 @vectest512(<16 x i32> %input) {
+entry:
+  %0 = bitcast <16 x i32> %input to i512
+  %1 = icmp ne i512 %0, 0
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+; SSE41: vectest512
+; SSE41: por
+; SSE41: por
+; SSE41: por
 ; SSE41: ptest
 ; SSE41: ret
-; AVX:   vectest
-; AVX:   vptest
+; AVX:   vectest512
+; AVX:   vorps
+; AVX:   vptest %ymm{{.*}}, %ymm{{.*}}
 ; AVX:   ret
 }
 
-define i32 @vecsel(<4 x i32> %input, i32 %a, i32 %b) {
+define i32 @vecsel128(<4 x i32> %input, i32 %a, i32 %b) {
 entry:
   %0 = bitcast <4 x i32> %input to i128
   %1 = icmp ne i128 %0, 0
   %2 = select i1 %1, i32 %a, i32 %b
   ret i32 %2
-; SSE41: vecsel
+; SSE41: vecsel128
+; SSE41: ptest
+; SSE41: ret
+; AVX:   vecsel128
+; AVX:   vptest %xmm{{.*}}, %xmm{{.*}}
+; AVX:   ret
+}
+
+define i32 @vecsel256(<8 x i32> %input, i32 %a, i32 %b) {
+entry:
+  %0 = bitcast <8 x i32> %input to i256
+  %1 = icmp ne i256 %0, 0
+  %2 = select i1 %1, i32 %a, i32 %b
+  ret i32 %2
+; SSE41: vecsel256
+; SSE41: por
+; SSE41: ptest
+; SSE41: ret
+; AVX:   vecsel256
+; AVX:   vptest %ymm{{.*}}, %ymm{{.*}}
+; AVX:   ret
+}
+
+define i32 @vecsel512(<16 x i32> %input, i32 %a, i32 %b) {
+entry:
+  %0 = bitcast <16 x i32> %input to i512
+  %1 = icmp ne i512 %0, 0
+  %2 = select i1 %1, i32 %a, i32 %b
+  ret i32 %2
+; SSE41: vecsel512
+; SSE41: por
+; SSE41: por
+; SSE41: por
 ; SSE41: ptest
 ; SSE41: ret
-; AVX:   vecsel
-; AVX:   vptest
+; AVX:   vecsel512
+; AVX:   vorps
+; AVX:   vptest %ymm{{.*}}, %ymm{{.*}}
 ; AVX:   ret
 }