1 files changed, 43 insertions, 9 deletions
diff --git a/test/CodeGen/X86/avx-cvt-2.ll b/test/CodeGen/X86/avx-cvt-2.ll
index 8cc7190..583c7d5 100644
--- a/test/CodeGen/X86/avx-cvt-2.ll
+++ b/test/CodeGen/X86/avx-cvt-2.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7-avx | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx | FileCheck %s
 
 ; Check that we generate vector conversion from float to narrower int types
 
@@ -8,8 +8,16 @@
 
 define void @fptoui16(%f32vec_t %a, %i16vec_t *%p) {
 ; CHECK-LABEL: fptoui16:
-; CHECK: vcvttps2dq %ymm
-; CHECK-NOT: vcvttss2si
+; CHECK:       # BB#0:
+; CHECK-NEXT:    vcvttps2dq %ymm0, %ymm0
+; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; CHECK-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; CHECK-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; CHECK-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
+; CHECK-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; CHECK-NEXT:    vmovdqa %xmm0, (%rdi)
+; CHECK-NEXT:    vzeroupper
+; CHECK-NEXT:    retq
   %b = fptoui %f32vec_t %a to %i16vec_t
   store %i16vec_t %b, %i16vec_t * %p
   ret void
@@ -17,8 +25,16 @@ define void @fptoui16(%f32vec_t %a, %i16vec_t *%p) {
 
 define void @fptosi16(%f32vec_t %a, %i16vec_t *%p) {
 ; CHECK-LABEL: fptosi16:
-; CHECK: vcvttps2dq %ymm
-; CHECK-NOT: vcvttss2si
+; CHECK:       # BB#0:
+; CHECK-NEXT:    vcvttps2dq %ymm0, %ymm0
+; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; CHECK-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; CHECK-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; CHECK-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
+; CHECK-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; CHECK-NEXT:    vmovdqa %xmm0, (%rdi)
+; CHECK-NEXT:    vzeroupper
+; CHECK-NEXT:    retq
   %b = fptosi %f32vec_t %a to %i16vec_t
   store %i16vec_t %b, %i16vec_t * %p
   ret void
@@ -26,8 +42,17 @@ define void @fptosi16(%f32vec_t %a, %i16vec_t *%p) {
 
 define void @fptoui8(%f32vec_t %a, %i8vec_t *%p) {
 ; CHECK-LABEL: fptoui8:
-; CHECK: vcvttps2dq %ymm
-; CHECK-NOT: vcvttss2si
+; CHECK:       # BB#0:
+; CHECK-NEXT:    vcvttps2dq %ymm0, %ymm0
+; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; CHECK-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; CHECK-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; CHECK-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
+; CHECK-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; CHECK-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; CHECK-NEXT:    vmovq %xmm0, (%rdi)
+; CHECK-NEXT:    vzeroupper
+; CHECK-NEXT:    retq
   %b = fptoui %f32vec_t %a to %i8vec_t
   store %i8vec_t %b, %i8vec_t * %p
   ret void
@@ -35,8 +60,17 @@ define void @fptoui8(%f32vec_t %a, %i8vec_t *%p) {
 
 define void @fptosi8(%f32vec_t %a, %i8vec_t *%p) {
 ; CHECK-LABEL: fptosi8:
-; CHECK: vcvttps2dq %ymm
-; CHECK-NOT: vcvttss2si
+; CHECK:       # BB#0:
+; CHECK-NEXT:    vcvttps2dq %ymm0, %ymm0
+; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; CHECK-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; CHECK-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; CHECK-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
+; CHECK-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; CHECK-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; CHECK-NEXT:    vmovq %xmm0, (%rdi)
+; CHECK-NEXT:    vzeroupper
+; CHECK-NEXT:    retq
   %b = fptosi %f32vec_t %a to %i8vec_t
   store %i8vec_t %b, %i8vec_t * %p
   ret void