[InstCombine][SSE] Demanded vector elements for scalar intrinsics (Part 2 of 2)

Split from D17490. This patch improves support for determining the demanded vector elements through SSE scalar intrinsics:

1 - demanded vector element support for unary and some extra binary scalar intrinsics (RCP/RSQRT/SQRT/FRCZ and ADD/CMP/DIV/ROUND).

2 - addss/addsd get simplified to a fadd call if we aren't interested in the pass through elements

3 - if we don't need the lowest element of a scalar operation then just use the first argument (the pass through elements) directly

We can add support for propagating demanded elements through any equivalent packed SSE intrinsics in a future patch (these wouldn't use the pass through patterns).

Differential Revision: http://reviews.llvm.org/D19318

llvm-svn: 267357

4 files changed, 60 insertions, 182 deletions

diff --git a/llvm/test/Transforms/InstCombine/x86-sse.ll b/llvm/test/Transforms/InstCombine/x86-sse.ll
index 270c0fb5a4d..831c55e340c 100644
--- a/llvm/test/Transforms/InstCombine/x86-sse.ll
+++ b/llvm/test/Transforms/InstCombine/x86-sse.ll
@@ -5,12 +5,9 @@ target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 define float @test_rcp_ss_0(float %a) {
 ; CHECK-LABEL: @test_rcp_ss_0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = tail call <4 x float> @llvm.x86.sse.rcp.ss(<4 x float> [[TMP4]])
-; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x float> [[TMP5]], i32 0
-; CHECK-NEXT:    ret float [[TMP6]]
+; CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.x86.sse.rcp.ss(<4 x float> [[TMP1]])
+; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x float> [[TMP2]], i32 0
+; CHECK-NEXT:    ret float [[TMP3]]
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -23,13 +20,7 @@ define float @test_rcp_ss_0(float %a) {
 
 define float @test_rcp_ss_1(float %a) {
 ; CHECK-LABEL: @test_rcp_ss_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = tail call <4 x float> @llvm.x86.sse.rcp.ss(<4 x float> [[TMP4]])
-; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x float> [[TMP5]], i32 1
-; CHECK-NEXT:    ret float [[TMP6]]
+; CHECK-NEXT:    ret float 1.000000e+00
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -43,12 +34,9 @@ define float @test_rcp_ss_1(float %a) {
 define float @test_sqrt_ss_0(float %a) {
 ; CHECK-LABEL: @test_sqrt_ss_0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = tail call <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float> [[TMP4]])
-; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x float> [[TMP5]], i32 0
-; CHECK-NEXT:    ret float [[TMP6]]
+; CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float> [[TMP1]])
+; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x float> [[TMP2]], i32 0
+; CHECK-NEXT:    ret float [[TMP3]]
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -61,13 +49,7 @@ define float @test_sqrt_ss_0(float %a) {
 
 define float @test_sqrt_ss_2(float %a) {
 ; CHECK-LABEL: @test_sqrt_ss_2(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = tail call <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float> [[TMP4]])
-; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x float> [[TMP5]], i32 2
-; CHECK-NEXT:    ret float [[TMP6]]
+; CHECK-NEXT:    ret float 2.000000e+00
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -81,12 +63,9 @@ define float @test_sqrt_ss_2(float %a) {
 define float @test_rsqrt_ss_0(float %a) {
 ; CHECK-LABEL: @test_rsqrt_ss_0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = tail call <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float> [[TMP4]])
-; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x float> [[TMP5]], i32 0
-; CHECK-NEXT:    ret float [[TMP6]]
+; CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float> [[TMP1]])
+; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x float> [[TMP2]], i32 0
+; CHECK-NEXT:    ret float [[TMP3]]
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -99,13 +78,7 @@ define float @test_rsqrt_ss_0(float %a) {
 
 define float @test_rsqrt_ss_3(float %a) {
 ; CHECK-LABEL: @test_rsqrt_ss_3(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = tail call <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float> [[TMP4]])
-; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x float> [[TMP5]], i32 3
-; CHECK-NEXT:    ret float [[TMP6]]
+; CHECK-NEXT:    ret float 3.000000e+00
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -130,14 +103,8 @@ define <4 x float> @test_add_ss(<4 x float> %a, <4 x float> %b) {
 
 define float @test_add_ss_0(float %a, float %b) {
 ; CHECK-LABEL: @test_add_ss_0(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x float> undef, float %b, i32 0
-; CHECK-NEXT:    [[TMP6:%.*]] = tail call <4 x float> @llvm.x86.sse.add.ss(<4 x float> [[TMP4]], <4 x float> [[TMP5]])
-; CHECK-NEXT:    [[R:%.*]] = extractelement <4 x float> [[TMP6]], i32 0
-; CHECK-NEXT:    ret float [[R]]
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd float %a, %b
+; CHECK-NEXT:    ret float [[TMP1]]
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -154,14 +121,7 @@ define float @test_add_ss_0(float %a, float %b) {
 
 define float @test_add_ss_1(float %a, float %b) {
 ; CHECK-LABEL: @test_add_ss_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x float> undef, float %b, i32 0
-; CHECK-NEXT:    [[TMP6:%.*]] = tail call <4 x float> @llvm.x86.sse.add.ss(<4 x float> [[TMP4]], <4 x float> [[TMP5]])
-; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <4 x float> [[TMP6]], i32 1
-; CHECK-NEXT:    ret float [[TMP7]]
+; CHECK-NEXT:    ret float 1.000000e+00
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -205,9 +165,7 @@ define float @test_sub_ss_0(float %a, float %b) {
 
 define float @test_sub_ss_2(float %a, float %b) {
 ; CHECK-LABEL: @test_sub_ss_2(
-; CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.x86.sse.sub.ss(<4 x float> <float undef, float undef, float 2.000000e+00, float undef>, <4 x float> undef)
-; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i32 2
-; CHECK-NEXT:    ret float [[TMP2]]
+; CHECK-NEXT:    ret float 2.000000e+00
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -251,9 +209,7 @@ define float @test_mul_ss_0(float %a, float %b) {
 
 define float @test_mul_ss_3(float %a, float %b) {
 ; CHECK-LABEL: @test_mul_ss_3(
-; CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.x86.sse.mul.ss(<4 x float> <float undef, float undef, float undef, float 3.000000e+00>, <4 x float> undef)
-; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i32 3
-; CHECK-NEXT:    ret float [[TMP2]]
+; CHECK-NEXT:    ret float 3.000000e+00
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -280,12 +236,9 @@ define <4 x float> @test_div_ss(<4 x float> %a, <4 x float> %b) {
 define float @test_div_ss_0(float %a, float %b) {
 ; CHECK-LABEL: @test_div_ss_0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x float> undef, float %b, i32 0
-; CHECK-NEXT:    [[TMP6:%.*]] = tail call <4 x float> @llvm.x86.sse.div.ss(<4 x float> [[TMP4]], <4 x float> [[TMP5]])
-; CHECK-NEXT:    [[R:%.*]] = extractelement <4 x float> [[TMP6]], i32 0
+; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> undef, float %b, i32 0
+; CHECK-NEXT:    [[TMP3:%.*]] = tail call <4 x float> @llvm.x86.sse.div.ss(<4 x float> [[TMP1]], <4 x float> [[TMP2]])
+; CHECK-NEXT:    [[R:%.*]] = extractelement <4 x float> [[TMP3]], i32 0
 ; CHECK-NEXT:    ret float [[R]]
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
@@ -303,14 +256,7 @@ define float @test_div_ss_0(float %a, float %b) {
 
 define float @test_div_ss_1(float %a, float %b) {
 ; CHECK-LABEL: @test_div_ss_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x float> undef, float %b, i32 0
-; CHECK-NEXT:    [[TMP6:%.*]] = tail call <4 x float> @llvm.x86.sse.div.ss(<4 x float> [[TMP4]], <4 x float> [[TMP5]])
-; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <4 x float> [[TMP6]], i32 1
-; CHECK-NEXT:    ret float [[TMP7]]
+; CHECK-NEXT:    ret float 1.000000e+00
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -357,9 +303,7 @@ define float @test_min_ss_0(float %a, float %b) {
 
 define float @test_min_ss_2(float %a, float %b) {
 ; CHECK-LABEL: @test_min_ss_2(
-; CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.x86.sse.min.ss(<4 x float> <float undef, float undef, float 2.000000e+00, float undef>, <4 x float> undef)
-; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i32 2
-; CHECK-NEXT:    ret float [[TMP2]]
+; CHECK-NEXT:    ret float 2.000000e+00
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -406,9 +350,7 @@ define float @test_max_ss_0(float %a, float %b) {
 
 define float @test_max_ss_3(float %a, float %b) {
 ; CHECK-LABEL: @test_max_ss_3(
-; CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.x86.sse.max.ss(<4 x float> <float undef, float undef, float undef, float 3.000000e+00>, <4 x float> undef)
-; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i32 3
-; CHECK-NEXT:    ret float [[TMP2]]
+; CHECK-NEXT:    ret float 3.000000e+00
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -435,12 +377,9 @@ define <4 x float> @test_cmp_ss(<4 x float> %a, <4 x float> %b) {
 define float @test_cmp_ss_0(float %a, float %b) {
 ; CHECK-LABEL: @test_cmp_ss_0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x float> undef, float %b, i32 0
-; CHECK-NEXT:    [[TMP6:%.*]] = tail call <4 x float> @llvm.x86.sse.cmp.ss(<4 x float> [[TMP4]], <4 x float> [[TMP5]], i8 0)
-; CHECK-NEXT:    [[R:%.*]] = extractelement <4 x float> [[TMP6]], i32 0
+; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> undef, float %b, i32 0
+; CHECK-NEXT:    [[TMP3:%.*]] = tail call <4 x float> @llvm.x86.sse.cmp.ss(<4 x float> [[TMP1]], <4 x float> [[TMP2]], i8 0)
+; CHECK-NEXT:    [[R:%.*]] = extractelement <4 x float> [[TMP3]], i32 0
 ; CHECK-NEXT:    ret float [[R]]
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
@@ -458,14 +397,7 @@ define float @test_cmp_ss_0(float %a, float %b) {
 
 define float @test_cmp_ss_1(float %a, float %b) {
 ; CHECK-LABEL: @test_cmp_ss_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x float> undef, float %b, i32 0
-; CHECK-NEXT:    [[TMP6:%.*]] = tail call <4 x float> @llvm.x86.sse.cmp.ss(<4 x float> [[TMP4]], <4 x float> [[TMP5]], i8 0)
-; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <4 x float> [[TMP6]], i32 1
-; CHECK-NEXT:    ret float [[TMP7]]
+; CHECK-NEXT:    ret float 1.000000e+00
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
diff --git a/llvm/test/Transforms/InstCombine/x86-sse2.ll b/llvm/test/Transforms/InstCombine/x86-sse2.ll
index 45ce1666e94..f0c0947ac18 100644
--- a/llvm/test/Transforms/InstCombine/x86-sse2.ll
+++ b/llvm/test/Transforms/InstCombine/x86-sse2.ll
@@ -5,10 +5,9 @@ target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 define double @test_sqrt_sd_0(double %a) {
 ; CHECK-LABEL: @test_sqrt_sd_0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> undef, double %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> [[TMP1]], double 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = tail call <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double> [[TMP2]])
-; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x double> [[TMP3]], i32 0
-; CHECK-NEXT:    ret double [[TMP4]]
+; CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double> [[TMP1]])
+; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x double> [[TMP2]], i32 0
+; CHECK-NEXT:    ret double [[TMP3]]
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -19,11 +18,7 @@ define double @test_sqrt_sd_0(double %a) {
 
 define double @test_sqrt_sd_1(double %a) {
 ; CHECK-LABEL: @test_sqrt_sd_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> undef, double %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> [[TMP1]], double 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = tail call <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double> [[TMP2]])
-; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x double> [[TMP3]], i32 1
-; CHECK-NEXT:    ret double [[TMP4]]
+; CHECK-NEXT:    ret double 1.000000e+00
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -44,12 +39,8 @@ define <2 x double> @test_add_sd(<2 x double> %a, <2 x double> %b) {
 
 define double @test_add_sd_0(double %a, double %b) {
 ; CHECK-LABEL: @test_add_sd_0(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> undef, double %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> [[TMP1]], double 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x double> undef, double %b, i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = tail call <2 x double> @llvm.x86.sse2.add.sd(<2 x double> [[TMP2]], <2 x double> [[TMP3]])
-; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x double> [[TMP4]], i32 0
-; CHECK-NEXT:    ret double [[TMP5]]
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd double %a, %b
+; CHECK-NEXT:    ret double [[TMP1]]
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -62,12 +53,7 @@ define double @test_add_sd_0(double %a, double %b) {
 
 define double @test_add_sd_1(double %a, double %b) {
 ; CHECK-LABEL: @test_add_sd_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> undef, double %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> [[TMP1]], double 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x double> undef, double %b, i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = tail call <2 x double> @llvm.x86.sse2.add.sd(<2 x double> [[TMP2]], <2 x double> [[TMP3]])
-; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x double> [[TMP4]], i32 1
-; CHECK-NEXT:    ret double [[TMP5]]
+; CHECK-NEXT:    ret double 1.000000e+00
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -104,9 +90,7 @@ define double @test_sub_sd_0(double %a, double %b) {
 
 define double @test_sub_sd_1(double %a, double %b) {
 ; CHECK-LABEL: @test_sub_sd_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.x86.sse2.sub.sd(<2 x double> <double undef, double 1.000000e+00>, <2 x double> undef)
-; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x double> [[TMP1]], i32 1
-; CHECK-NEXT:    ret double [[TMP2]]
+; CHECK-NEXT:    ret double 1.000000e+00
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -143,9 +127,7 @@ define double @test_mul_sd_0(double %a, double %b) {
 
 define double @test_mul_sd_1(double %a, double %b) {
 ; CHECK-LABEL: @test_mul_sd_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.x86.sse2.mul.sd(<2 x double> <double undef, double 1.000000e+00>, <2 x double> undef)
-; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x double> [[TMP1]], i32 1
-; CHECK-NEXT:    ret double [[TMP2]]
+; CHECK-NEXT:    ret double 1.000000e+00
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -169,11 +151,10 @@ define <2 x double> @test_div_sd(<2 x double> %a, <2 x double> %b) {
 define double @test_div_sd_0(double %a, double %b) {
 ; CHECK-LABEL: @test_div_sd_0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> undef, double %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> [[TMP1]], double 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x double> undef, double %b, i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = tail call <2 x double> @llvm.x86.sse2.div.sd(<2 x double> [[TMP2]], <2 x double> [[TMP3]])
-; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x double> [[TMP4]], i32 0
-; CHECK-NEXT:    ret double [[TMP5]]
+; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> undef, double %b, i32 0
+; CHECK-NEXT:    [[TMP3:%.*]] = tail call <2 x double> @llvm.x86.sse2.div.sd(<2 x double> [[TMP1]], <2 x double> [[TMP2]])
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x double> [[TMP3]], i32 0
+; CHECK-NEXT:    ret double [[TMP4]]
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -186,12 +167,7 @@ define double @test_div_sd_0(double %a, double %b) {
 
 define double @test_div_sd_1(double %a, double %b) {
 ; CHECK-LABEL: @test_div_sd_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> undef, double %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> [[TMP1]], double 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x double> undef, double %b, i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = tail call <2 x double> @llvm.x86.sse2.div.sd(<2 x double> [[TMP2]], <2 x double> [[TMP3]])
-; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x double> [[TMP4]], i32 1
-; CHECK-NEXT:    ret double [[TMP5]]
+; CHECK-NEXT:    ret double 1.000000e+00
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -231,9 +207,7 @@ define double @test_min_sd_0(double %a, double %b) {
 
 define double @test_min_sd_1(double %a, double %b) {
 ; CHECK-LABEL: @test_min_sd_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.x86.sse2.min.sd(<2 x double> <double undef, double 1.000000e+00>, <2 x double> undef)
-; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x double> [[TMP1]], i32 1
-; CHECK-NEXT:    ret double [[TMP2]]
+; CHECK-NEXT:    ret double 1.000000e+00
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -273,9 +247,7 @@ define double @test_max_sd_0(double %a, double %b) {
 
 define double @test_max_sd_1(double %a, double %b) {
 ; CHECK-LABEL: @test_max_sd_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.x86.sse2.max.sd(<2 x double> <double undef, double 1.000000e+00>, <2 x double> undef)
-; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x double> [[TMP1]], i32 1
-; CHECK-NEXT:    ret double [[TMP2]]
+; CHECK-NEXT:    ret double 1.000000e+00
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -299,11 +271,10 @@ define <2 x double> @test_cmp_sd(<2 x double> %a, <2 x double> %b) {
 define double @test_cmp_sd_0(double %a, double %b) {
 ; CHECK-LABEL: @test_cmp_sd_0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> undef, double %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> [[TMP1]], double 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x double> undef, double %b, i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = tail call <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double> [[TMP2]], <2 x double> [[TMP3]], i8 0)
-; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x double> [[TMP4]], i32 0
-; CHECK-NEXT:    ret double [[TMP5]]
+; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> undef, double %b, i32 0
+; CHECK-NEXT:    [[TMP3:%.*]] = tail call <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double> [[TMP1]], <2 x double> [[TMP2]], i8 0)
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x double> [[TMP3]], i32 0
+; CHECK-NEXT:    ret double [[TMP4]]
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -316,12 +287,7 @@ define double @test_cmp_sd_0(double %a, double %b) {
 
 define double @test_cmp_sd_1(double %a, double %b) {
 ; CHECK-LABEL: @test_cmp_sd_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> undef, double %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> [[TMP1]], double 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x double> undef, double %b, i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = tail call <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double> [[TMP2]], <2 x double> [[TMP3]], i8 0)
-; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x double> [[TMP4]], i32 1
-; CHECK-NEXT:    ret double [[TMP5]]
+; CHECK-NEXT:    ret double 1.000000e+00
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
diff --git a/llvm/test/Transforms/InstCombine/x86-sse41.ll b/llvm/test/Transforms/InstCombine/x86-sse41.ll
index 771db65327d..16975471b9e 100644
--- a/llvm/test/Transforms/InstCombine/x86-sse41.ll
+++ b/llvm/test/Transforms/InstCombine/x86-sse41.ll
@@ -16,7 +16,7 @@ define <2 x double> @test_round_sd(<2 x double> %a, <2 x double> %b) {
 define double @test_round_sd_0(double %a, double %b) {
 ; CHECK-LABEL: @test_round_sd_0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> undef, double %b, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> <double undef, double 1.000000e+00>, <2 x double> [[TMP1]], i32 10)
+; CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> undef, <2 x double> [[TMP1]], i32 10)
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x double> [[TMP2]], i32 0
 ; CHECK-NEXT:    ret double [[TMP3]]
 ;
@@ -31,10 +31,7 @@ define double @test_round_sd_0(double %a, double %b) {
 
 define double @test_round_sd_1(double %a, double %b) {
 ; CHECK-LABEL: @test_round_sd_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> undef, double %b, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> <double undef, double 1.000000e+00>, <2 x double> [[TMP1]], i32 10)
-; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x double> [[TMP2]], i32 1
-; CHECK-NEXT:    ret double [[TMP3]]
+; CHECK-NEXT:    ret double 1.000000e+00
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -63,7 +60,7 @@ define <4 x float> @test_round_ss(<4 x float> %a, <4 x float> %b) {
 define float @test_round_ss_0(float %a, float %b) {
 ; CHECK-LABEL: @test_round_ss_0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %b, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> <float undef, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>, <4 x float> [[TMP1]], i32 10)
+; CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> undef, <4 x float> [[TMP1]], i32 10)
 ; CHECK-NEXT:    [[R:%.*]] = extractelement <4 x float> [[TMP2]], i32 0
 ; CHECK-NEXT:    ret float [[R]]
 ;
@@ -82,10 +79,7 @@ define float @test_round_ss_0(float %a, float %b) {
 
 define float @test_round_ss_2(float %a, float %b) {
 ; CHECK-LABEL: @test_round_ss_2(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %b, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> <float undef, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>, <4 x float> [[TMP1]], i32 10)
-; CHECK-NEXT:    [[R:%.*]] = extractelement <4 x float> [[TMP2]], i32 2
-; CHECK-NEXT:    ret float [[R]]
+; CHECK-NEXT:    ret float 2.000000e+00
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
diff --git a/llvm/test/Transforms/InstCombine/x86-xop.ll b/llvm/test/Transforms/InstCombine/x86-xop.ll
index 132ad55fc64..015d511ac4d 100644
--- a/llvm/test/Transforms/InstCombine/x86-xop.ll
+++ b/llvm/test/Transforms/InstCombine/x86-xop.ll
@@ -4,10 +4,9 @@
 define double @test_vfrcz_sd_0(double %a) {
 ; CHECK-LABEL: @test_vfrcz_sd_0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> undef, double %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> [[TMP1]], double 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = tail call <2 x double> @llvm.x86.xop.vfrcz.sd(<2 x double> [[TMP2]])
-; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x double> [[TMP3]], i32 0
-; CHECK-NEXT:    ret double [[TMP4]]
+; CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x double> @llvm.x86.xop.vfrcz.sd(<2 x double> [[TMP1]])
+; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x double> [[TMP2]], i32 0
+; CHECK-NEXT:    ret double [[TMP3]]
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -18,11 +17,7 @@ define double @test_vfrcz_sd_0(double %a) {
 
 define double @test_vfrcz_sd_1(double %a) {
 ; CHECK-LABEL: @test_vfrcz_sd_1(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> undef, double %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> [[TMP1]], double 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = tail call <2 x double> @llvm.x86.xop.vfrcz.sd(<2 x double> [[TMP2]])
-; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x double> [[TMP3]], i32 1
-; CHECK-NEXT:    ret double [[TMP4]]
+; CHECK-NEXT:    ret double 1.000000e+00
 ;
   %1 = insertelement <2 x double> undef, double %a, i32 0
   %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1
@@ -34,12 +29,9 @@ define double @test_vfrcz_sd_1(double %a) {
 define float @test_vfrcz_ss_0(float %a) {
 ; CHECK-LABEL: @test_vfrcz_ss_0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = tail call <4 x float> @llvm.x86.xop.vfrcz.ss(<4 x float> [[TMP4]])
-; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x float> [[TMP5]], i32 0
-; CHECK-NEXT:    ret float [[TMP6]]
+; CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.x86.xop.vfrcz.ss(<4 x float> [[TMP1]])
+; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x float> [[TMP2]], i32 0
+; CHECK-NEXT:    ret float [[TMP3]]
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1
@@ -52,13 +44,7 @@ define float @test_vfrcz_ss_0(float %a) {
 
 define float @test_vfrcz_ss_3(float %a) {
 ; CHECK-LABEL: @test_vfrcz_ss_3(
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float 1.000000e+00, i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float 2.000000e+00, i32 2
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> [[TMP3]], float 3.000000e+00, i32 3
-; CHECK-NEXT:    [[TMP5:%.*]] = tail call <4 x float> @llvm.x86.xop.vfrcz.ss(<4 x float> [[TMP4]])
-; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x float> [[TMP5]], i32 3
-; CHECK-NEXT:    ret float [[TMP6]]
+; CHECK-NEXT:    ret float 3.000000e+00
 ;
   %1 = insertelement <4 x float> undef, float %a, i32 0
   %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1