summaryrefslogtreecommitdiffstats
path: root/llvm/test/CodeGen/X86/vshift-6.ll
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/CodeGen/X86/vshift-6.ll')
-rw-r--r--llvm/test/CodeGen/X86/vshift-6.ll79
1 files changed, 74 insertions, 5 deletions
diff --git a/llvm/test/CodeGen/X86/vshift-6.ll b/llvm/test/CodeGen/X86/vshift-6.ll
index 551a1385003..36d428cb9cf 100644
--- a/llvm/test/CodeGen/X86/vshift-6.ll
+++ b/llvm/test/CodeGen/X86/vshift-6.ll
@@ -1,4 +1,6 @@
-; RUN: llc < %s -march=x86-64 -mattr=+sse2 | FileCheck %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=i686-unknown -mattr=+sse2 | FileCheck %s --check-prefix=CHECK --check-prefix=X32
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2 | FileCheck %s --check-prefix=CHECK --check-prefix=X64
; This test makes sure that the compiler does not crash with an
; assertion failure when trying to fold a vector shift left
@@ -23,6 +25,77 @@
; 'count' is the vector shift count.
define <16 x i8> @do_not_crash(i8*, i32*, i64*, i32, i64, i8) {
+; X32-LABEL: do_not_crash:
+; X32: # BB#0: # %entry
+; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT: movb %al, (%ecx)
+; X32-NEXT: movd %eax, %xmm0
+; X32-NEXT: psllq $56, %xmm0
+; X32-NEXT: movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255]
+; X32-NEXT: movdqa %xmm2, %xmm1
+; X32-NEXT: pandn %xmm0, %xmm1
+; X32-NEXT: por %xmm2, %xmm1
+; X32-NEXT: pcmpeqd %xmm2, %xmm2
+; X32-NEXT: psllw $5, %xmm1
+; X32-NEXT: pxor %xmm3, %xmm3
+; X32-NEXT: pxor %xmm0, %xmm0
+; X32-NEXT: pcmpgtb %xmm1, %xmm0
+; X32-NEXT: pxor %xmm0, %xmm2
+; X32-NEXT: pand {{\.LCPI.*}}, %xmm0
+; X32-NEXT: por %xmm2, %xmm0
+; X32-NEXT: paddb %xmm1, %xmm1
+; X32-NEXT: pxor %xmm2, %xmm2
+; X32-NEXT: pcmpgtb %xmm1, %xmm2
+; X32-NEXT: movdqa %xmm2, %xmm4
+; X32-NEXT: pandn %xmm0, %xmm4
+; X32-NEXT: psllw $2, %xmm0
+; X32-NEXT: pand {{\.LCPI.*}}, %xmm0
+; X32-NEXT: pand %xmm2, %xmm0
+; X32-NEXT: por %xmm4, %xmm0
+; X32-NEXT: paddb %xmm1, %xmm1
+; X32-NEXT: pcmpgtb %xmm1, %xmm3
+; X32-NEXT: movdqa %xmm3, %xmm1
+; X32-NEXT: pandn %xmm0, %xmm1
+; X32-NEXT: paddb %xmm0, %xmm0
+; X32-NEXT: pand %xmm3, %xmm0
+; X32-NEXT: por %xmm1, %xmm0
+; X32-NEXT: retl
+;
+; X64-LABEL: do_not_crash:
+; X64: # BB#0: # %entry
+; X64-NEXT: movb %r9b, (%rdi)
+; X64-NEXT: movd %r9d, %xmm0
+; X64-NEXT: psllq $56, %xmm0
+; X64-NEXT: movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255]
+; X64-NEXT: movdqa %xmm2, %xmm1
+; X64-NEXT: pandn %xmm0, %xmm1
+; X64-NEXT: por %xmm2, %xmm1
+; X64-NEXT: pcmpeqd %xmm2, %xmm2
+; X64-NEXT: psllw $5, %xmm1
+; X64-NEXT: pxor %xmm3, %xmm3
+; X64-NEXT: pxor %xmm0, %xmm0
+; X64-NEXT: pcmpgtb %xmm1, %xmm0
+; X64-NEXT: pxor %xmm0, %xmm2
+; X64-NEXT: pand {{.*}}(%rip), %xmm0
+; X64-NEXT: por %xmm2, %xmm0
+; X64-NEXT: paddb %xmm1, %xmm1
+; X64-NEXT: pxor %xmm2, %xmm2
+; X64-NEXT: pcmpgtb %xmm1, %xmm2
+; X64-NEXT: movdqa %xmm2, %xmm4
+; X64-NEXT: pandn %xmm0, %xmm4
+; X64-NEXT: psllw $2, %xmm0
+; X64-NEXT: pand {{.*}}(%rip), %xmm0
+; X64-NEXT: pand %xmm2, %xmm0
+; X64-NEXT: por %xmm4, %xmm0
+; X64-NEXT: paddb %xmm1, %xmm1
+; X64-NEXT: pcmpgtb %xmm1, %xmm3
+; X64-NEXT: movdqa %xmm3, %xmm1
+; X64-NEXT: pandn %xmm0, %xmm1
+; X64-NEXT: paddb %xmm0, %xmm0
+; X64-NEXT: pand %xmm3, %xmm0
+; X64-NEXT: por %xmm1, %xmm0
+; X64-NEXT: retq
entry:
store i8 %5, i8* %0
%L5 = load i8, i8* %0
@@ -30,7 +103,3 @@ entry:
%B51 = shl <16 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, %I8
ret <16 x i8> %B51
}
-
-; CHECK-LABEL: do_not_crash
-; CHECK: ret
-
OpenPOWER on IntegriCloud