Make blend tests more specific

Following the lead set by r209324, I'm making these tests match the whole
instruction, so we can be sure we're lowering them correctly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@209947 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/test/CodeGen/X86/avx-blend.ll b/test/CodeGen/X86/avx-blend.ll
index a66dc549b484cd78fec561bcc03bf61e16c92da3..43cdf7edf70a23039515fff0c067f9d1f9477bf0 100644 (file)
--- a/test/CodeGen/X86/avx-blend.ll
+++ b/test/CodeGen/X86/avx-blend.ll
@@ -162,7 +162,16 @@ declare <4 x double> @llvm.x86.avx.blendv.pd.256(<4 x double>, <4 x double>, <4
 ;; 4 tests for shufflevectors that optimize to blend + immediate
 ; CHECK-LABEL: @blend_shufflevector_4xfloat
 define <4 x float> @blend_shufflevector_4xfloat(<4 x float> %a, <4 x float> %b) {
-; CHECK: vblendps
+; Equivalent select mask is <i1 true, i1 false, i1 true, i1 false>.
+; Big endian representation is 0101 = 5.
+; '1' means takes the first argument, '0' means takes the second argument.
+; This is the opposite of the intel syntax, thus we expect
+; Inverted mask: 1010 = 10.
+; According to the ABI:
+; a is in xmm0 => first argument is xmm0.
+; b is in xmm1 => second argument is xmm1.
+; Result is in xmm0 => destination argument.
+; CHECK: vblendps $10, %xmm1, %xmm0, %xmm0
 ; CHECK: ret
   %1 = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
   ret <4 x float> %1
@@ -170,7 +179,7 @@ define <4 x float> @blend_shufflevector_4xfloat(<4 x float> %a, <4 x float> %b)
 
 ; CHECK-LABEL: @blend_shufflevector_8xfloat
 define <8 x float> @blend_shufflevector_8xfloat(<8 x float> %a, <8 x float> %b) {
-; CHECK: vblendps
+; CHECK: vblendps $190, %ymm1, %ymm0, %ymm0
 ; CHECK: ret
   %1 = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 12, i32 13, i32 6, i32 15>
   ret <8 x float> %1
@@ -178,7 +187,7 @@ define <8 x float> @blend_shufflevector_8xfloat(<8 x float> %a, <8 x float> %b)
 
 ; CHECK-LABEL: @blend_shufflevector_4xdouble
 define <4 x double> @blend_shufflevector_4xdouble(<4 x double> %a, <4 x double> %b) {
-; CHECK: vblendpd
+; CHECK: vblendpd $2, %ymm1, %ymm0, %ymm0
 ; CHECK: ret
   %1 = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 5, i32 2, i32 3>
   ret <4 x double> %1
@@ -186,7 +195,7 @@ define <4 x double> @blend_shufflevector_4xdouble(<4 x double> %a, <4 x double>
 
 ; CHECK-LABEL: @blend_shufflevector_4xi64
 define <4 x i64> @blend_shufflevector_4xi64(<4 x i64> %a, <4 x i64> %b) {
-; CHECK: vblendpd
+; CHECK: vblendpd $13, %ymm1, %ymm0, %ymm0
 ; CHECK: ret
   %1 = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 4, i32 1, i32 6, i32 7>
   ret <4 x i64> %1

diff --git a/test/CodeGen/X86/avx2-shuffle.ll b/test/CodeGen/X86/avx2-shuffle.ll
index 5b2e7a40bdb9de1359603cf5c59d756b6e1c0e80..185b989458ae0345a35be70591ef50143d77bf78 100644 (file)
--- a/test/CodeGen/X86/avx2-shuffle.ll
+++ b/test/CodeGen/X86/avx2-shuffle.ll
@@ -62,7 +62,7 @@ define <4 x i64> @blend_test4(<4 x i64> %a, <4 x i64> %b) nounwind alwaysinline
 
 ;; 2 tests for shufflevectors that optimize to blend + immediate
 ; CHECK-LABEL: @blend_test5
-; CHECK: vpblendd
+; CHECK: vpblendd $10, %xmm1, %xmm0, %xmm0
 ; CHECK: ret
 define <4 x i32> @blend_test5(<4 x i32> %a, <4 x i32> %b) {
   %1 = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
@@ -70,7 +70,7 @@ define <4 x i32> @blend_test5(<4 x i32> %a, <4 x i32> %b) {
 }
 
 ; CHECK-LABEL: @blend_test6
-; CHECK: vpblendw
+; CHECK: vpblendw $134, %ymm1, %ymm0, %ymm0
 ; CHECK: ret
 define <16 x i16> @blend_test6(<16 x i16> %a, <16 x i16> %b) {
   %1 = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 17, i32 18, i32  3, i32  4, i32  5, i32  6, i32 23,

diff --git a/test/CodeGen/X86/sse41-blend.ll b/test/CodeGen/X86/sse41-blend.ll
index ca106e158e74cc6e8b0d6c3d252be4e15233d6f1..3a4812119f8ac042d3d9b63eee111ae9ecbfc481 100644 (file)
--- a/test/CodeGen/X86/sse41-blend.ll
+++ b/test/CodeGen/X86/sse41-blend.ll
@@ -124,7 +124,7 @@ declare <2 x double> @llvm.x86.sse41.blendvpd(<2 x double>, <2 x double>, <2 x d
 
 ;; 2 tests for shufflevectors that optimize to blend + immediate
 ; CHECK-LABEL: @blend_shufflevector_4xfloat
-; CHECK: blendps
+; CHECK: blendps $6, %xmm1, %xmm0
 ; CHECK: ret
 define <4 x float> @blend_shufflevector_4xfloat(<4 x float> %a, <4 x float> %b) {
   %1 = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 5, i32 6, i32 3>
@@ -132,7 +132,7 @@ define <4 x float> @blend_shufflevector_4xfloat(<4 x float> %a, <4 x float> %b)
 }
 
 ; CHECK-LABEL: @blend_shufflevector_8xi16
-; CHECK: pblendw
+; CHECK: pblendw $134, %xmm1, %xmm0
 ; CHECK: ret
 define <8 x i16> @blend_shufflevector_8xi16(<8 x i16> %a, <8 x i16> %b) {
   %1 = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 9, i32 10, i32 3, i32 4, i32 5, i32 6, i32 15>