[x86] regenerate checks

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@272396 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/test/CodeGen/X86/commute-fcmp.ll b/test/CodeGen/X86/commute-fcmp.ll
index 6f43ebe1fcd7fb7387adea48806fcd18aaad9816..b2fa970ab9a84fced6caa52e576f64a4991962f7 100644 (file)
--- a/test/CodeGen/X86/commute-fcmp.ll
+++ b/test/CodeGen/X86/commute-fcmp.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: llc -O3 -mtriple=x86_64-unknown -mcpu=x86-64 -mattr=+sse2 < %s | FileCheck %s --check-prefix=SSE
 ; RUN: llc -O3 -mtriple=x86_64-unknown -mcpu=x86-64 -mattr=+avx2 < %s | FileCheck %s --check-prefix=AVX
 
@@ -6,164 +7,226 @@
 ; Only equal/not-equal/ordered/unordered can be safely commuted
 ;
 
-define <4 x i32> @commute_cmpps_eq(<4 x float>* %a0, <4 x float> %a1) #0 {
-  ;SSE-LABEL: commute_cmpps_eq
-  ;SSE:       cmpeqps (%rdi), %xmm0
-  ;SSE-NEXT:  retq
-
-  ;AVX-LABEL: commute_cmpps_eq
-  ;AVX:       vcmpeqps (%rdi), %xmm0, %xmm0
-  ;AVX-NEXT:  retq
-
+define <4 x i32> @commute_cmpps_eq(<4 x float>* %a0, <4 x float> %a1) {
+; SSE-LABEL: commute_cmpps_eq:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpeqps (%rdi), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmpps_eq:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpeqps (%rdi), %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
   %1 = load <4 x float>, <4 x float>* %a0
   %2 = fcmp oeq <4 x float> %1, %a1
   %3 = sext <4 x i1> %2 to <4 x i32>
   ret <4 x i32> %3
 }
 
-define <4 x i32> @commute_cmpps_ne(<4 x float>* %a0, <4 x float> %a1) #0 {
-  ;SSE-LABEL: commute_cmpps_ne
-  ;SSE:       cmpneqps (%rdi), %xmm0
-  ;SSE-NEXT:  retq
-
-  ;AVX-LABEL: commute_cmpps_ne
-  ;AVX:       vcmpneqps (%rdi), %xmm0, %xmm0
-  ;AVX-NEXT:  retq
-
+define <4 x i32> @commute_cmpps_ne(<4 x float>* %a0, <4 x float> %a1) {
+; SSE-LABEL: commute_cmpps_ne:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpneqps (%rdi), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmpps_ne:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpneqps (%rdi), %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
   %1 = load <4 x float>, <4 x float>* %a0
   %2 = fcmp une <4 x float> %1, %a1
   %3 = sext <4 x i1> %2 to <4 x i32>
   ret <4 x i32> %3
 }
 
-define <4 x i32> @commute_cmpps_ord(<4 x float>* %a0, <4 x float> %a1) #0 {
-  ;SSE-LABEL: commute_cmpps_ord
-  ;SSE:       cmpordps (%rdi), %xmm0
-  ;SSE-NEXT:  retq
-
-  ;AVX-LABEL: commute_cmpps_ord
-  ;AVX:       vcmpordps (%rdi), %xmm0, %xmm0
-  ;AVX-NEXT:  retq
-
+define <4 x i32> @commute_cmpps_ord(<4 x float>* %a0, <4 x float> %a1) {
+; SSE-LABEL: commute_cmpps_ord:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpordps (%rdi), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmpps_ord:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpordps (%rdi), %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
   %1 = load <4 x float>, <4 x float>* %a0
   %2 = fcmp ord <4 x float> %1, %a1
   %3 = sext <4 x i1> %2 to <4 x i32>
   ret <4 x i32> %3
 }
 
-define <4 x i32> @commute_cmpps_uno(<4 x float>* %a0, <4 x float> %a1) #0 {
-  ;SSE-LABEL: commute_cmpps_uno
-  ;SSE:       cmpunordps (%rdi), %xmm0
-  ;SSE-NEXT:  retq
-
-  ;AVX-LABEL: commute_cmpps_uno
-  ;AVX:       vcmpunordps (%rdi), %xmm0, %xmm0
-  ;AVX-NEXT:  retq
-
+define <4 x i32> @commute_cmpps_uno(<4 x float>* %a0, <4 x float> %a1) {
+; SSE-LABEL: commute_cmpps_uno:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpunordps (%rdi), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmpps_uno:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpunordps (%rdi), %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
   %1 = load <4 x float>, <4 x float>* %a0
   %2 = fcmp uno <4 x float> %1, %a1
   %3 = sext <4 x i1> %2 to <4 x i32>
   ret <4 x i32> %3
 }
 
-define <4 x i32> @commute_cmpps_lt(<4 x float>* %a0, <4 x float> %a1) #0 {
-  ;SSE-LABEL: commute_cmpps_lt
-  ;SSE:       movaps (%rdi), %xmm1
-  ;SSE-NEXT:  cmpltps %xmm0, %xmm1
-  ;SSE-NEXT:  movaps %xmm1, %xmm0
-  ;SSE-NEXT:  retq
-
-  ;AVX-LABEL: commute_cmpps_lt
-  ;AVX:       vmovaps (%rdi), %xmm1
-  ;AVX-NEXT:  vcmpltps %xmm0, %xmm1, %xmm0
-  ;AVX-NEXT:  retq
-
+define <4 x i32> @commute_cmpps_lt(<4 x float>* %a0, <4 x float> %a1) {
+; SSE-LABEL: commute_cmpps_lt:
+; SSE:       # BB#0:
+; SSE-NEXT:    movaps (%rdi), %xmm1
+; SSE-NEXT:    cmpltps %xmm0, %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmpps_lt:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovaps (%rdi), %xmm1
+; AVX-NEXT:    vcmpltps %xmm0, %xmm1, %xmm0
+; AVX-NEXT:    retq
+;
   %1 = load <4 x float>, <4 x float>* %a0
   %2 = fcmp olt <4 x float> %1, %a1
   %3 = sext <4 x i1> %2 to <4 x i32>
   ret <4 x i32> %3
 }
 
-define <4 x i32> @commute_cmpps_le(<4 x float>* %a0, <4 x float> %a1) #0 {
-  ;SSE-LABEL: commute_cmpps_le
-  ;SSE:       movaps (%rdi), %xmm1
-  ;SSE-NEXT:  cmpleps %xmm0, %xmm1
-  ;SSE-NEXT:  movaps %xmm1, %xmm0
-  ;SSE-NEXT:  retq
-
-  ;AVX-LABEL: commute_cmpps_le
-  ;AVX:       vmovaps (%rdi), %xmm1
-  ;AVX-NEXT:  vcmpleps %xmm0, %xmm1, %xmm0
-  ;AVX-NEXT:  retq
-
+define <4 x i32> @commute_cmpps_le(<4 x float>* %a0, <4 x float> %a1) {
+; SSE-LABEL: commute_cmpps_le:
+; SSE:       # BB#0:
+; SSE-NEXT:    movaps (%rdi), %xmm1
+; SSE-NEXT:    cmpleps %xmm0, %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmpps_le:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovaps (%rdi), %xmm1
+; AVX-NEXT:    vcmpleps %xmm0, %xmm1, %xmm0
+; AVX-NEXT:    retq
+;
   %1 = load <4 x float>, <4 x float>* %a0
   %2 = fcmp ole <4 x float> %1, %a1
   %3 = sext <4 x i1> %2 to <4 x i32>
   ret <4 x i32> %3
 }
 
-define <8 x i32> @commute_cmpps_eq_ymm(<8 x float>* %a0, <8 x float> %a1) #0 {
-  ;AVX-LABEL: commute_cmpps_eq_ymm
-  ;AVX:       vcmpeqps (%rdi), %ymm0, %ymm0
-  ;AVX-NEXT:  retq
-
+define <8 x i32> @commute_cmpps_eq_ymm(<8 x float>* %a0, <8 x float> %a1) {
+; SSE-LABEL: commute_cmpps_eq_ymm:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpeqps (%rdi), %xmm0
+; SSE-NEXT:    cmpeqps 16(%rdi), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmpps_eq_ymm:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpeqps (%rdi), %ymm0, %ymm0
+; AVX-NEXT:    retq
+;
   %1 = load <8 x float>, <8 x float>* %a0
   %2 = fcmp oeq <8 x float> %1, %a1
   %3 = sext <8 x i1> %2 to <8 x i32>
   ret <8 x i32> %3
 }
 
-define <8 x i32> @commute_cmpps_ne_ymm(<8 x float>* %a0, <8 x float> %a1) #0 {
-  ;AVX-LABEL: commute_cmpps_ne_ymm
-  ;AVX:       vcmpneqps (%rdi), %ymm0, %ymm0
-  ;AVX-NEXT:  retq
-
+define <8 x i32> @commute_cmpps_ne_ymm(<8 x float>* %a0, <8 x float> %a1) {
+; SSE-LABEL: commute_cmpps_ne_ymm:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpneqps (%rdi), %xmm0
+; SSE-NEXT:    cmpneqps 16(%rdi), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmpps_ne_ymm:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpneqps (%rdi), %ymm0, %ymm0
+; AVX-NEXT:    retq
+;
   %1 = load <8 x float>, <8 x float>* %a0
   %2 = fcmp une <8 x float> %1, %a1
   %3 = sext <8 x i1> %2 to <8 x i32>
   ret <8 x i32> %3
 }
 
-define <8 x i32> @commute_cmpps_ord_ymm(<8 x float>* %a0, <8 x float> %a1) #0 {
-  ;AVX-LABEL: commute_cmpps_ord_ymm
-  ;AVX:       vcmpordps (%rdi), %ymm0, %ymm0
-  ;AVX-NEXT:  retq
-
+define <8 x i32> @commute_cmpps_ord_ymm(<8 x float>* %a0, <8 x float> %a1) {
+; SSE-LABEL: commute_cmpps_ord_ymm:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpordps (%rdi), %xmm0
+; SSE-NEXT:    cmpordps 16(%rdi), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmpps_ord_ymm:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpordps (%rdi), %ymm0, %ymm0
+; AVX-NEXT:    retq
+;
   %1 = load <8 x float>, <8 x float>* %a0
   %2 = fcmp ord <8 x float> %1, %a1
   %3 = sext <8 x i1> %2 to <8 x i32>
   ret <8 x i32> %3
 }
 
-define <8 x i32> @commute_cmpps_uno_ymm(<8 x float>* %a0, <8 x float> %a1) #0 {
-  ;AVX-LABEL: commute_cmpps_uno_ymm
-  ;AVX:       vcmpunordps (%rdi), %ymm0, %ymm0
-  ;AVX-NEXT:  retq
-
+define <8 x i32> @commute_cmpps_uno_ymm(<8 x float>* %a0, <8 x float> %a1) {
+; SSE-LABEL: commute_cmpps_uno_ymm:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpunordps (%rdi), %xmm0
+; SSE-NEXT:    cmpunordps 16(%rdi), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmpps_uno_ymm:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpunordps (%rdi), %ymm0, %ymm0
+; AVX-NEXT:    retq
+;
   %1 = load <8 x float>, <8 x float>* %a0
   %2 = fcmp uno <8 x float> %1, %a1
   %3 = sext <8 x i1> %2 to <8 x i32>
   ret <8 x i32> %3
 }
 
-define <8 x i32> @commute_cmpps_lt_ymm(<8 x float>* %a0, <8 x float> %a1) #0 {
-  ;AVX-LABEL: commute_cmpps_lt_ymm
-  ;AVX:       vmovaps (%rdi), %ymm1
-  ;AVX-NEXT:  vcmpltps %ymm0, %ymm1, %ymm0
-  ;AVX-NEXT:  retq
-
+define <8 x i32> @commute_cmpps_lt_ymm(<8 x float>* %a0, <8 x float> %a1) {
+; SSE-LABEL: commute_cmpps_lt_ymm:
+; SSE:       # BB#0:
+; SSE-NEXT:    movaps (%rdi), %xmm2
+; SSE-NEXT:    movaps 16(%rdi), %xmm3
+; SSE-NEXT:    cmpltps %xmm0, %xmm2
+; SSE-NEXT:    cmpltps %xmm1, %xmm3
+; SSE-NEXT:    movaps %xmm2, %xmm0
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmpps_lt_ymm:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovaps (%rdi), %ymm1
+; AVX-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
+; AVX-NEXT:    retq
+;
   %1 = load <8 x float>, <8 x float>* %a0
   %2 = fcmp olt <8 x float> %1, %a1
   %3 = sext <8 x i1> %2 to <8 x i32>
   ret <8 x i32> %3
 }
 
-define <8 x i32> @commute_cmpps_le_ymm(<8 x float>* %a0, <8 x float> %a1) #0 {
-  ;AVX-LABEL: commute_cmpps_le_ymm
-  ;AVX:       vmovaps (%rdi), %ymm1
-  ;AVX-NEXT:  vcmpleps %ymm0, %ymm1, %ymm0
-  ;AVX-NEXT:  retq
-
+define <8 x i32> @commute_cmpps_le_ymm(<8 x float>* %a0, <8 x float> %a1) {
+; SSE-LABEL: commute_cmpps_le_ymm:
+; SSE:       # BB#0:
+; SSE-NEXT:    movaps (%rdi), %xmm2
+; SSE-NEXT:    movaps 16(%rdi), %xmm3
+; SSE-NEXT:    cmpleps %xmm0, %xmm2
+; SSE-NEXT:    cmpleps %xmm1, %xmm3
+; SSE-NEXT:    movaps %xmm2, %xmm0
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmpps_le_ymm:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovaps (%rdi), %ymm1
+; AVX-NEXT:    vcmpleps %ymm0, %ymm1, %ymm0
+; AVX-NEXT:    retq
+;
   %1 = load <8 x float>, <8 x float>* %a0
   %2 = fcmp ole <8 x float> %1, %a1
   %3 = sext <8 x i1> %2 to <8 x i32>
@@ -175,164 +238,226 @@ define <8 x i32> @commute_cmpps_le_ymm(<8 x float>* %a0, <8 x float> %a1) #0 {
 ; Only equal/not-equal/ordered/unordered can be safely commuted
 ;
 
-define <2 x i64> @commute_cmppd_eq(<2 x double>* %a0, <2 x double> %a1) #0 {
-  ;SSE-LABEL: commute_cmppd_eq
-  ;SSE:       cmpeqpd (%rdi), %xmm0
-  ;SSE-NEXT:  retq
-
-  ;AVX-LABEL: commute_cmppd_eq
-  ;AVX:       vcmpeqpd (%rdi), %xmm0, %xmm0
-  ;AVX-NEXT:  retq
-
+define <2 x i64> @commute_cmppd_eq(<2 x double>* %a0, <2 x double> %a1) {
+; SSE-LABEL: commute_cmppd_eq:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpeqpd (%rdi), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmppd_eq:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpeqpd (%rdi), %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
   %1 = load <2 x double>, <2 x double>* %a0
   %2 = fcmp oeq <2 x double> %1, %a1
   %3 = sext <2 x i1> %2 to <2 x i64>
   ret <2 x i64> %3
 }
 
-define <2 x i64> @commute_cmppd_ne(<2 x double>* %a0, <2 x double> %a1) #0 {
-  ;SSE-LABEL: commute_cmppd_ne
-  ;SSE:       cmpneqpd (%rdi), %xmm0
-  ;SSE-NEXT:  retq
-
-  ;AVX-LABEL: commute_cmppd_ne
-  ;AVX:       vcmpneqpd (%rdi), %xmm0, %xmm0
-  ;AVX-NEXT:  retq
-
+define <2 x i64> @commute_cmppd_ne(<2 x double>* %a0, <2 x double> %a1) {
+; SSE-LABEL: commute_cmppd_ne:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpneqpd (%rdi), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmppd_ne:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpneqpd (%rdi), %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
   %1 = load <2 x double>, <2 x double>* %a0
   %2 = fcmp une <2 x double> %1, %a1
   %3 = sext <2 x i1> %2 to <2 x i64>
   ret <2 x i64> %3
 }
 
-define <2 x i64> @commute_cmppd_ord(<2 x double>* %a0, <2 x double> %a1) #0 {
-  ;SSE-LABEL: commute_cmppd_ord
-  ;SSE:       cmpordpd (%rdi), %xmm0
-  ;SSE-NEXT:  retq
-
-  ;AVX-LABEL: commute_cmppd_ord
-  ;AVX:       vcmpordpd (%rdi), %xmm0, %xmm0
-  ;AVX-NEXT:  retq
-
+define <2 x i64> @commute_cmppd_ord(<2 x double>* %a0, <2 x double> %a1) {
+; SSE-LABEL: commute_cmppd_ord:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpordpd (%rdi), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmppd_ord:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpordpd (%rdi), %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
   %1 = load <2 x double>, <2 x double>* %a0
   %2 = fcmp ord <2 x double> %1, %a1
   %3 = sext <2 x i1> %2 to <2 x i64>
   ret <2 x i64> %3
 }
 
-define <2 x i64> @commute_cmppd_uno(<2 x double>* %a0, <2 x double> %a1) #0 {
-  ;SSE-LABEL: commute_cmppd_uno
-  ;SSE:       cmpunordpd (%rdi), %xmm0
-  ;SSE-NEXT:  retq
-
-  ;AVX-LABEL: commute_cmppd_uno
-  ;AVX:       vcmpunordpd (%rdi), %xmm0, %xmm0
-  ;AVX-NEXT:  retq
-
+define <2 x i64> @commute_cmppd_uno(<2 x double>* %a0, <2 x double> %a1) {
+; SSE-LABEL: commute_cmppd_uno:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpunordpd (%rdi), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmppd_uno:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpunordpd (%rdi), %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
   %1 = load <2 x double>, <2 x double>* %a0
   %2 = fcmp uno <2 x double> %1, %a1
   %3 = sext <2 x i1> %2 to <2 x i64>
   ret <2 x i64> %3
 }
 
-define <2 x i64> @commute_cmppd_lt(<2 x double>* %a0, <2 x double> %a1) #0 {
-  ;SSE-LABEL: commute_cmppd_lt
-  ;SSE:       movapd (%rdi), %xmm1
-  ;SSE-NEXT:  cmpltpd %xmm0, %xmm1
-  ;SSE-NEXT:  movapd %xmm1, %xmm0
-  ;SSE-NEXT:  retq
-
-  ;AVX-LABEL: commute_cmppd_lt
-  ;AVX:       vmovapd (%rdi), %xmm1
-  ;AVX-NEXT:  vcmpltpd %xmm0, %xmm1, %xmm0
-  ;AVX-NEXT:  retq
-
+define <2 x i64> @commute_cmppd_lt(<2 x double>* %a0, <2 x double> %a1) {
+; SSE-LABEL: commute_cmppd_lt:
+; SSE:       # BB#0:
+; SSE-NEXT:    movapd (%rdi), %xmm1
+; SSE-NEXT:    cmpltpd %xmm0, %xmm1
+; SSE-NEXT:    movapd %xmm1, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmppd_lt:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovapd (%rdi), %xmm1
+; AVX-NEXT:    vcmpltpd %xmm0, %xmm1, %xmm0
+; AVX-NEXT:    retq
+;
   %1 = load <2 x double>, <2 x double>* %a0
   %2 = fcmp olt <2 x double> %1, %a1
   %3 = sext <2 x i1> %2 to <2 x i64>
   ret <2 x i64> %3
 }
 
-define <2 x i64> @commute_cmppd_le(<2 x double>* %a0, <2 x double> %a1) #0 {
-  ;SSE-LABEL: commute_cmppd_le
-  ;SSE:       movapd (%rdi), %xmm1
-  ;SSE-NEXT:  cmplepd %xmm0, %xmm1
-  ;SSE-NEXT:  movapd %xmm1, %xmm0
-  ;SSE-NEXT:  retq
-
-  ;AVX-LABEL: commute_cmppd_le
-  ;AVX:       vmovapd (%rdi), %xmm1
-  ;AVX-NEXT:  vcmplepd %xmm0, %xmm1, %xmm0
-  ;AVX-NEXT:  retq
-
+define <2 x i64> @commute_cmppd_le(<2 x double>* %a0, <2 x double> %a1) {
+; SSE-LABEL: commute_cmppd_le:
+; SSE:       # BB#0:
+; SSE-NEXT:    movapd (%rdi), %xmm1
+; SSE-NEXT:    cmplepd %xmm0, %xmm1
+; SSE-NEXT:    movapd %xmm1, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmppd_le:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovapd (%rdi), %xmm1
+; AVX-NEXT:    vcmplepd %xmm0, %xmm1, %xmm0
+; AVX-NEXT:    retq
+;
   %1 = load <2 x double>, <2 x double>* %a0
   %2 = fcmp ole <2 x double> %1, %a1
   %3 = sext <2 x i1> %2 to <2 x i64>
   ret <2 x i64> %3
 }
 
-define <4 x i64> @commute_cmppd_eq_ymmm(<4 x double>* %a0, <4 x double> %a1) #0 {
-  ;AVX-LABEL: commute_cmppd_eq
-  ;AVX:       vcmpeqpd (%rdi), %ymm0, %ymm0
-  ;AVX-NEXT:  retq
-
+define <4 x i64> @commute_cmppd_eq_ymmm(<4 x double>* %a0, <4 x double> %a1) {
+; SSE-LABEL: commute_cmppd_eq_ymmm:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpeqpd (%rdi), %xmm0
+; SSE-NEXT:    cmpeqpd 16(%rdi), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmppd_eq_ymmm:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpeqpd (%rdi), %ymm0, %ymm0
+; AVX-NEXT:    retq
+;
   %1 = load <4 x double>, <4 x double>* %a0
   %2 = fcmp oeq <4 x double> %1, %a1
   %3 = sext <4 x i1> %2 to <4 x i64>
   ret <4 x i64> %3
 }
 
-define <4 x i64> @commute_cmppd_ne_ymmm(<4 x double>* %a0, <4 x double> %a1) #0 {
-  ;AVX-LABEL: commute_cmppd_ne
-  ;AVX:       vcmpneqpd (%rdi), %ymm0, %ymm0
-  ;AVX-NEXT:  retq
-
+define <4 x i64> @commute_cmppd_ne_ymmm(<4 x double>* %a0, <4 x double> %a1) {
+; SSE-LABEL: commute_cmppd_ne_ymmm:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpneqpd (%rdi), %xmm0
+; SSE-NEXT:    cmpneqpd 16(%rdi), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmppd_ne_ymmm:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpneqpd (%rdi), %ymm0, %ymm0
+; AVX-NEXT:    retq
+;
   %1 = load <4 x double>, <4 x double>* %a0
   %2 = fcmp une <4 x double> %1, %a1
   %3 = sext <4 x i1> %2 to <4 x i64>
   ret <4 x i64> %3
 }
 
-define <4 x i64> @commute_cmppd_ord_ymmm(<4 x double>* %a0, <4 x double> %a1) #0 {
-  ;AVX-LABEL: commute_cmppd_ord
-  ;AVX:       vcmpordpd (%rdi), %ymm0, %ymm0
-  ;AVX-NEXT:  retq
-
+define <4 x i64> @commute_cmppd_ord_ymmm(<4 x double>* %a0, <4 x double> %a1) {
+; SSE-LABEL: commute_cmppd_ord_ymmm:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpordpd (%rdi), %xmm0
+; SSE-NEXT:    cmpordpd 16(%rdi), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmppd_ord_ymmm:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpordpd (%rdi), %ymm0, %ymm0
+; AVX-NEXT:    retq
+;
   %1 = load <4 x double>, <4 x double>* %a0
   %2 = fcmp ord <4 x double> %1, %a1
   %3 = sext <4 x i1> %2 to <4 x i64>
   ret <4 x i64> %3
 }
 
-define <4 x i64> @commute_cmppd_uno_ymmm(<4 x double>* %a0, <4 x double> %a1) #0 {
-  ;AVX-LABEL: commute_cmppd_uno
-  ;AVX:       vcmpunordpd (%rdi), %ymm0, %ymm0
-  ;AVX-NEXT:  retq
-
+define <4 x i64> @commute_cmppd_uno_ymmm(<4 x double>* %a0, <4 x double> %a1) {
+; SSE-LABEL: commute_cmppd_uno_ymmm:
+; SSE:       # BB#0:
+; SSE-NEXT:    cmpunordpd (%rdi), %xmm0
+; SSE-NEXT:    cmpunordpd 16(%rdi), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmppd_uno_ymmm:
+; AVX:       # BB#0:
+; AVX-NEXT:    vcmpunordpd (%rdi), %ymm0, %ymm0
+; AVX-NEXT:    retq
+;
   %1 = load <4 x double>, <4 x double>* %a0
   %2 = fcmp uno <4 x double> %1, %a1
   %3 = sext <4 x i1> %2 to <4 x i64>
   ret <4 x i64> %3
 }
 
-define <4 x i64> @commute_cmppd_lt_ymmm(<4 x double>* %a0, <4 x double> %a1) #0 {
-  ;AVX-LABEL: commute_cmppd_lt
-  ;AVX:       vmovapd (%rdi), %ymm1
-  ;AVX-NEXT:  vcmpltpd %ymm0, %ymm1, %ymm0
-  ;AVX-NEXT:  retq
-
+define <4 x i64> @commute_cmppd_lt_ymmm(<4 x double>* %a0, <4 x double> %a1) {
+; SSE-LABEL: commute_cmppd_lt_ymmm:
+; SSE:       # BB#0:
+; SSE-NEXT:    movapd (%rdi), %xmm2
+; SSE-NEXT:    movapd 16(%rdi), %xmm3
+; SSE-NEXT:    cmpltpd %xmm0, %xmm2
+; SSE-NEXT:    cmpltpd %xmm1, %xmm3
+; SSE-NEXT:    movapd %xmm2, %xmm0
+; SSE-NEXT:    movapd %xmm3, %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmppd_lt_ymmm:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovapd (%rdi), %ymm1
+; AVX-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
+; AVX-NEXT:    retq
+;
   %1 = load <4 x double>, <4 x double>* %a0
   %2 = fcmp olt <4 x double> %1, %a1
   %3 = sext <4 x i1> %2 to <4 x i64>
   ret <4 x i64> %3
 }
 
-define <4 x i64> @commute_cmppd_le_ymmm(<4 x double>* %a0, <4 x double> %a1) #0 {
-  ;AVX-LABEL: commute_cmppd_le
-  ;AVX:       vmovapd (%rdi), %ymm1
-  ;AVX-NEXT:  vcmplepd %ymm0, %ymm1, %ymm0
-  ;AVX-NEXT:  retq
-
+define <4 x i64> @commute_cmppd_le_ymmm(<4 x double>* %a0, <4 x double> %a1) {
+; SSE-LABEL: commute_cmppd_le_ymmm:
+; SSE:       # BB#0:
+; SSE-NEXT:    movapd (%rdi), %xmm2
+; SSE-NEXT:    movapd 16(%rdi), %xmm3
+; SSE-NEXT:    cmplepd %xmm0, %xmm2
+; SSE-NEXT:    cmplepd %xmm1, %xmm3
+; SSE-NEXT:    movapd %xmm2, %xmm0
+; SSE-NEXT:    movapd %xmm3, %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: commute_cmppd_le_ymmm:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovapd (%rdi), %ymm1
+; AVX-NEXT:    vcmplepd %ymm0, %ymm1, %ymm0
+; AVX-NEXT:    retq
+;
   %1 = load <4 x double>, <4 x double>* %a0
   %2 = fcmp ole <4 x double> %1, %a1
   %3 = sext <4 x i1> %2 to <4 x i64>