From b534439b8d181eb21c844d5b5b5087888000c6ee Mon Sep 17 00:00:00 2001
From: Chandler Carruth <chandlerc@gmail.com>
Date: Fri, 3 Oct 2014 00:50:03 +0000
Subject: [PATCH] [x86] Regenerate a bunch more avx512 test cases using my
 script to have tighter, more strict FileCheck assertions. Some of these I
 really like as they show case exactly what instruction sequences come out of
 these microscopic functionality tests.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218936 91177308-0d34-0410-b5e6-96231b3b80d8
---
 test/CodeGen/X86/avx512-build-vector.ll |  36 ++-
 test/CodeGen/X86/avx512-vbroadcast.ll   |  82 ++++---
 test/CodeGen/X86/avx512-vec-cmp.ll      | 286 +++++++++++++++---------
 3 files changed, 252 insertions(+), 152 deletions(-)

diff --git a/test/CodeGen/X86/avx512-build-vector.ll b/test/CodeGen/X86/avx512-build-vector.ll
index b5a2aa80ce1..20f837129eb 100644
--- a/test/CodeGen/X86/avx512-build-vector.ll
+++ b/test/CodeGen/X86/avx512-build-vector.ll
@@ -1,30 +1,44 @@
 ; RUN: llc < %s -march=x86-64 -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s
 
-; CHECK-LABEL: test1
-; CHECK: vpxord
-; CHECK: ret
 define <16 x i32> @test1(i32* %x) {
+; CHECK-LABEL: test1:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vmovd (%rdi), %xmm0
+; CHECK-NEXT:    vmovdqa32 {{.*}}(%rip), %zmm1
+; CHECK-NEXT:    vpxord %zmm2, %zmm2, %zmm2
+; CHECK-NEXT:    vpermt2d %zmm0, %zmm1, %zmm2
+; CHECK-NEXT:    vmovdqa32 {{.*}}(%rip), %zmm0
+; CHECK-NEXT:    vpermd %zmm2, %zmm0, %zmm0
+; CHECK-NEXT:    retq
    %y = load i32* %x, align 4
    %res = insertelement <16 x i32>zeroinitializer, i32 %y, i32 4
    ret <16 x i32>%res
 }
 
-; CHECK-LABEL: test2
-; CHECK: vpaddd LCP{{.*}}(%rip){1to16}
-; CHECK: ret
 define <16 x i32> @test2(<16 x i32> %x) {
+; CHECK-LABEL: test2:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpaddd {{.*}}(%rip){1to16}, %zmm0, %zmm0
+; CHECK-NEXT:    retq
    %res = add <16 x i32><i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>, %x
    ret <16 x i32>%res
 }
 
-; CHECK-LABEL: test3
-; CHECK: vinsertf128
-; CHECK: vinsertf64x4
-; CHECK: ret
 define <16 x float> @test3(<4 x float> %a) {
+; CHECK-LABEL: test3:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vmovhlps %xmm0, %xmm0, %xmm1
+; CHECK-NEXT:    vxorps %xmm2, %xmm2, %xmm2
+; CHECK-NEXT:    vmovss %xmm0, %xmm2, %xmm0
+; CHECK-NEXT:    vmovss %xmm1, %xmm2, %xmm1
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[1,0],xmm0[0,1]
+; CHECK-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
+; CHECK-NEXT:    vxorps %ymm1, %ymm1, %ymm1
+; CHECK-NEXT:    vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
+; CHECK-NEXT:    retq
   %b = extractelement <4 x float> %a, i32 2
   %c = insertelement <16 x float> <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float undef, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00>, float %b, i32 5
   %b1 = extractelement <4 x float> %a, i32 0
   %c1 = insertelement <16 x float> %c, float %b1, i32 6
   ret <16 x float>%c1
-}
\ No newline at end of file
+}
diff --git a/test/CodeGen/X86/avx512-vbroadcast.ll b/test/CodeGen/X86/avx512-vbroadcast.ll
index 93dd677395c..0b0e0fc2bc8 100644
--- a/test/CodeGen/X86/avx512-vbroadcast.ll
+++ b/test/CodeGen/X86/avx512-vbroadcast.ll
@@ -1,60 +1,72 @@
 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s
 
-;CHECK-LABEL: _inreg16xi32:
-;CHECK: vpbroadcastd {{.*}}, %zmm
-;CHECK: ret
 define   <16 x i32> @_inreg16xi32(i32 %a) {
+; CHECK-LABEL: _inreg16xi32:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpbroadcastd %edi, %zmm0
+; CHECK-NEXT:    retq
   %b = insertelement <16 x i32> undef, i32 %a, i32 0
   %c = shufflevector <16 x i32> %b, <16 x i32> undef, <16 x i32> zeroinitializer
   ret <16 x i32> %c
 }
 
-;CHECK-LABEL: _inreg8xi64:
-;CHECK: vpbroadcastq {{.*}}, %zmm
-;CHECK: ret
 define   <8 x i64> @_inreg8xi64(i64 %a) {
+; CHECK-LABEL: _inreg8xi64:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpbroadcastq %rdi, %zmm0
+; CHECK-NEXT:    retq
   %b = insertelement <8 x i64> undef, i64 %a, i32 0
   %c = shufflevector <8 x i64> %b, <8 x i64> undef, <8 x i32> zeroinitializer
   ret <8 x i64> %c
 }
 
-;CHECK-LABEL: _inreg16xfloat:
-;CHECK: vbroadcastss {{.*}}, %zmm
-;CHECK: ret
 define   <16 x float> @_inreg16xfloat(float %a) {
+; CHECK-LABEL: _inreg16xfloat:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vbroadcastss %xmm0, %zmm0
+; CHECK-NEXT:    retq
   %b = insertelement <16 x float> undef, float %a, i32 0
   %c = shufflevector <16 x float> %b, <16 x float> undef, <16 x i32> zeroinitializer
   ret <16 x float> %c
 }
 
-;CHECK-LABEL: _inreg8xdouble:
-;CHECK: vbroadcastsd {{.*}}, %zmm
-;CHECK: ret
 define   <8 x double> @_inreg8xdouble(double %a) {
+; CHECK-LABEL: _inreg8xdouble:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vbroadcastsd %xmm0, %zmm0
+; CHECK-NEXT:    retq
   %b = insertelement <8 x double> undef, double %a, i32 0
   %c = shufflevector <8 x double> %b, <8 x double> undef, <8 x i32> zeroinitializer
   ret <8 x double> %c
 }
 
-;CHECK-LABEL: _xmm16xi32
-;CHECK: vpbroadcastd
-;CHECK: ret
 define   <16 x i32> @_xmm16xi32(<16 x i32> %a) {
+; CHECK-LABEL: _xmm16xi32:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpbroadcastd %xmm0, %zmm0
+; CHECK-NEXT:    retq
   %b = shufflevector <16 x i32> %a, <16 x i32> undef, <16 x i32> zeroinitializer
   ret <16 x i32> %b
 }
 
-;CHECK-LABEL: _xmm16xfloat
-;CHECK: vbroadcastss {{.*}}
-;CHECK: ret
 define   <16 x float> @_xmm16xfloat(<16 x float> %a) {
+; CHECK-LABEL: _xmm16xfloat:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vbroadcastss %xmm0, %zmm0
+; CHECK-NEXT:    retq
   %b = shufflevector <16 x float> %a, <16 x float> undef, <16 x i32> zeroinitializer
   ret <16 x float> %b
 }
 
 define <16 x i32> @test_vbroadcast() {
-; CHECK-LABEL: _test_vbroadcast:
-; CHECK: vpbroadcastd
+; CHECK-LABEL: test_vbroadcast:
+; CHECK:       ## BB#0: ## %entry
+; CHECK-NEXT:    vpxord %zmm0, %zmm0, %zmm0
+; CHECK-NEXT:    vcmpunordps %zmm0, %zmm0, %k1
+; CHECK-NEXT:    vpbroadcastd {{.*}}(%rip), %zmm0 {%k1} {z}
+; CHECK-NEXT:    knotw %k1, %k1
+; CHECK-NEXT:    vmovdqu32 %zmm0, %zmm0 {%k1} {z}
+; CHECK-NEXT:    retq
 entry:
   %0 = sext <16 x i1> zeroinitializer to <16 x i32>
   %1 = fcmp uno <16 x float> undef, zeroinitializer
@@ -66,8 +78,10 @@ entry:
 ; We implement the set1 intrinsics with vector initializers.  Verify that the
 ; IR generated will produce broadcasts at the end.
 define <8 x double> @test_set1_pd(double %d) #2 {
-; CHECK-LABEL: _test_set1_pd:
-; CHECK: vbroadcastsd
+; CHECK-LABEL: test_set1_pd:
+; CHECK:       ## BB#0: ## %entry
+; CHECK-NEXT:    vbroadcastsd %xmm0, %zmm0
+; CHECK-NEXT:    retq
 entry:
   %vecinit.i = insertelement <8 x double> undef, double %d, i32 0
   %vecinit1.i = insertelement <8 x double> %vecinit.i, double %d, i32 1
@@ -81,8 +95,10 @@ entry:
 }
 
 define <8 x i64> @test_set1_epi64(i64 %d) #2 {
-; CHECK-LABEL: _test_set1_epi64:
-; CHECK: vpbroadcastq
+; CHECK-LABEL: test_set1_epi64:
+; CHECK:       ## BB#0: ## %entry
+; CHECK-NEXT:    vpbroadcastq %rdi, %zmm0
+; CHECK-NEXT:    retq
 entry:
   %vecinit.i = insertelement <8 x i64> undef, i64 %d, i32 0
   %vecinit1.i = insertelement <8 x i64> %vecinit.i, i64 %d, i32 1
@@ -96,8 +112,10 @@ entry:
 }
 
 define <16 x float> @test_set1_ps(float %f) #2 {
-; CHECK-LABEL: _test_set1_ps:
-; CHECK: vbroadcastss
+; CHECK-LABEL: test_set1_ps:
+; CHECK:       ## BB#0: ## %entry
+; CHECK-NEXT:    vbroadcastss %xmm0, %zmm0
+; CHECK-NEXT:    retq
 entry:
   %vecinit.i = insertelement <16 x float> undef, float %f, i32 0
   %vecinit1.i = insertelement <16 x float> %vecinit.i, float %f, i32 1
@@ -119,8 +137,10 @@ entry:
 }
 
 define <16 x i32> @test_set1_epi32(i32 %f) #2 {
-; CHECK-LABEL: _test_set1_epi32:
-; CHECK: vpbroadcastd
+; CHECK-LABEL: test_set1_epi32:
+; CHECK:       ## BB#0: ## %entry
+; CHECK-NEXT:    vpbroadcastd %edi, %zmm0
+; CHECK-NEXT:    retq
 entry:
   %vecinit.i = insertelement <16 x i32> undef, i32 %f, i32 0
   %vecinit1.i = insertelement <16 x i32> %vecinit.i, i32 %f, i32 1
@@ -144,8 +164,10 @@ entry:
 ; We implement the scalar broadcast intrinsics with vector initializers.
 ; Verify that the IR generated will produce the broadcast at the end.
 define <8 x double> @test_mm512_broadcastsd_pd(<2 x double> %a) {
-; CHECK-LABEL: _test_mm512_broadcastsd_pd:
-; CHECK: vbroadcastsd %xmm0, %zmm0
+; CHECK-LABEL: test_mm512_broadcastsd_pd:
+; CHECK:       ## BB#0: ## %entry
+; CHECK-NEXT:    vbroadcastsd %xmm0, %zmm0
+; CHECK-NEXT:    retq
 entry:
   %0 = extractelement <2 x double> %a, i32 0
   %vecinit.i = insertelement <8 x double> undef, double %0, i32 0
diff --git a/test/CodeGen/X86/avx512-vec-cmp.ll b/test/CodeGen/X86/avx512-vec-cmp.ll
index d9acc1d325f..d993a0e0f21 100644
--- a/test/CodeGen/X86/avx512-vec-cmp.ll
+++ b/test/CodeGen/X86/avx512-vec-cmp.ll
@@ -1,145 +1,176 @@
 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s
 
-; CHECK-LABEL: test1
-; CHECK: vcmpleps
-; CHECK: vmovaps
-; CHECK: ret
 define <16 x float> @test1(<16 x float> %x, <16 x float> %y) nounwind {
+; CHECK-LABEL: test1:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vcmpleps %zmm1, %zmm0, %k1
+; CHECK-NEXT:    vmovaps %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %mask = fcmp ole <16 x float> %x, %y
   %max = select <16 x i1> %mask, <16 x float> %x, <16 x float> %y
   ret <16 x float> %max
 }
 
-; CHECK-LABEL: test2
-; CHECK: vcmplepd
-; CHECK: vmovapd
-; CHECK: ret
 define <8 x double> @test2(<8 x double> %x, <8 x double> %y) nounwind {
+; CHECK-LABEL: test2:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vcmplepd %zmm1, %zmm0, %k1
+; CHECK-NEXT:    vmovapd %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %mask = fcmp ole <8 x double> %x, %y
   %max = select <8 x i1> %mask, <8 x double> %x, <8 x double> %y
   ret <8 x double> %max
 }
 
-; CHECK-LABEL: test3
-; CHECK: vpcmpeqd  (%rdi)
-; CHECK: vmovdqa32
-; CHECK: ret
 define <16 x i32> @test3(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %yp) nounwind {
+; CHECK-LABEL: test3:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqd (%rdi), %zmm0, %k1
+; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %y = load <16 x i32>* %yp, align 4
   %mask = icmp eq <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %x1
   ret <16 x i32> %max
 }
 
-; CHECK-LABEL: @test4_unsigned
-; CHECK: vpcmpnltud
-; CHECK: vmovdqa32
-; CHECK: ret
 define <16 x i32> @test4_unsigned(<16 x i32> %x, <16 x i32> %y) nounwind {
+; CHECK-LABEL: test4_unsigned:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpnltud %zmm1, %zmm0, %k1
+; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %mask = icmp uge <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %y
   ret <16 x i32> %max
 }
 
-; CHECK-LABEL: test5
-; CHECK: vpcmpeqq {{.*}}%k1
-; CHECK: vmovdqa64 {{.*}}%k1
-; CHECK: ret
 define <8 x i64> @test5(<8 x i64> %x, <8 x i64> %y) nounwind {
+; CHECK-LABEL: test5:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqq %zmm1, %zmm0, %k1
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %mask = icmp eq <8 x i64> %x, %y
   %max = select <8 x i1> %mask, <8 x i64> %x, <8 x i64> %y
   ret <8 x i64> %max
 }
 
-; CHECK-LABEL: test6_unsigned
-; CHECK: vpcmpnleuq {{.*}}%k1
-; CHECK: vmovdqa64 {{.*}}%k1
-; CHECK: ret
 define <8 x i64> @test6_unsigned(<8 x i64> %x, <8 x i64> %y) nounwind {
+; CHECK-LABEL: test6_unsigned:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k1
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %mask = icmp ugt <8 x i64> %x, %y
   %max = select <8 x i1> %mask, <8 x i64> %x, <8 x i64> %y
   ret <8 x i64> %max
 }
 
-; CHECK-LABEL: test7
-; CHECK: xor
-; CHECK: vcmpltps
-; CHECK: vblendvps
-; CHECK: ret
 define <4 x float> @test7(<4 x float> %a, <4 x float> %b) {
+; CHECK-LABEL: test7:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vxorps %xmm2, %xmm2, %xmm2
+; CHECK-NEXT:    vcmpltps %xmm2, %xmm0, %xmm2
+; CHECK-NEXT:    vblendvps %xmm2, %xmm0, %xmm1, %xmm0
+; CHECK-NEXT:    retq
   %mask = fcmp olt <4 x float> %a, zeroinitializer
   %c = select <4 x i1>%mask, <4 x float>%a, <4 x float>%b
   ret <4 x float>%c
 }
 
-; CHECK-LABEL: test8
-; CHECK: xor
-; CHECK: vcmpltpd
-; CHECK: vblendvpd
-; CHECK: ret
 define <2 x double> @test8(<2 x double> %a, <2 x double> %b) {
+; CHECK-LABEL: test8:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vxorpd %xmm2, %xmm2, %xmm2
+; CHECK-NEXT:    vcmpltpd %xmm2, %xmm0, %xmm2
+; CHECK-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
+; CHECK-NEXT:    retq
   %mask = fcmp olt <2 x double> %a, zeroinitializer
   %c = select <2 x i1>%mask, <2 x double>%a, <2 x double>%b
   ret <2 x double>%c
 }
 
-; CHECK-LABEL: test9
-; CHECK: vpcmpeqd
-; CHECK: vpblendmd
-; CHECK: ret
 define <8 x i32> @test9(<8 x i32> %x, <8 x i32> %y) nounwind {
+; CHECK-LABEL: test9:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:      ## kill: YMM1<def> YMM1<kill> ZMM1<def>
+; CHECK-NEXT:      ## kill: YMM0<def> YMM0<kill> ZMM0<def>
+; CHECK-NEXT:    vpcmpeqd %zmm1, %zmm0, %k1
+; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
+; CHECK-NEXT:      ## kill: YMM0<def> YMM0<kill> ZMM0<kill>
+; CHECK-NEXT:    retq
   %mask = icmp eq <8 x i32> %x, %y
   %max = select <8 x i1> %mask, <8 x i32> %x, <8 x i32> %y
   ret <8 x i32> %max
 }
 
-; CHECK-LABEL: test10
-; CHECK: vcmpeqps
-; CHECK: vblendmps
-; CHECK: ret
 define <8 x float> @test10(<8 x float> %x, <8 x float> %y) nounwind {
+; CHECK-LABEL: test10:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:      ## kill: YMM1<def> YMM1<kill> ZMM1<def>
+; CHECK-NEXT:      ## kill: YMM0<def> YMM0<kill> ZMM0<def>
+; CHECK-NEXT:    vcmpeqps %zmm1, %zmm0, %k1
+; CHECK-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
+; CHECK-NEXT:      ## kill: YMM0<def> YMM0<kill> ZMM0<kill>
+; CHECK-NEXT:    retq
   %mask = fcmp oeq <8 x float> %x, %y
   %max = select <8 x i1> %mask, <8 x float> %x, <8 x float> %y
   ret <8 x float> %max
 }
 
-; CHECK-LABEL: test11_unsigned
-; CHECK: vpmaxud
-; CHECK: ret
 define <8 x i32> @test11_unsigned(<8 x i32> %x, <8 x i32> %y) nounwind {
+; CHECK-LABEL: test11_unsigned:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpmaxud %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retq
   %mask = icmp ugt <8 x i32> %x, %y
   %max = select <8 x i1> %mask, <8 x i32> %x, <8 x i32> %y
   ret <8 x i32> %max
 }
 
-; CHECK-LABEL: test12
-; CHECK: vpcmpeqq        %zmm2, %zmm0, [[LO:%k[0-7]]]
-; CHECK: vpcmpeqq        %zmm3, %zmm1, [[HI:%k[0-7]]]
-; CHECK: kunpckbw        [[LO]], [[HI]], {{%k[0-7]}}
 
 define i16 @test12(<16 x i64> %a, <16 x i64> %b) nounwind {
+; CHECK-LABEL: test12:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqq %zmm2, %zmm0, %k0
+; CHECK-NEXT:    vpcmpeqq %zmm3, %zmm1, %k1
+; CHECK-NEXT:    kunpckbw %k0, %k1, %k0
+; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:      ## kill: AX<def> AX<kill> EAX<kill>
+; CHECK-NEXT:    retq
   %res = icmp eq <16 x i64> %a, %b
   %res1 = bitcast <16 x i1> %res to i16
   ret i16 %res1
 }
 
-; CHECK-LABEL: test13
-; CHECK: vcmpeqps        %zmm
-; CHECK: vpbroadcastd
-; CHECK: ret
 define <16 x i32> @test13(<16 x float>%a, <16 x float>%b)
+; CHECK-LABEL: test13:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vcmpeqps %zmm1, %zmm0, %k1
+; CHECK-NEXT:    vpbroadcastd {{.*}}(%rip), %zmm0 {%k1} {z}
+; CHECK-NEXT:    retq
 {
   %cmpvector_i = fcmp oeq <16 x float> %a, %b
   %conv = zext <16 x i1> %cmpvector_i to <16 x i32>
   ret <16 x i32> %conv
 }
 
-; CHECK-LABEL: test14
-; CHECK: vpcmp
-; CHECK-NOT: vpcmp
-; CHECK: vmovdqu32 {{.*}}{%k1} {z}
-; CHECK: ret
 define <16 x i32> @test14(<16 x i32>%a, <16 x i32>%b) {
+; CHECK-LABEL: test14:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpsubd %zmm1, %zmm0, %zmm1
+; CHECK-NEXT:    vpcmpgtd %zmm0, %zmm1, %k0
+; CHECK-NEXT:    knotw %k0, %k0
+; CHECK-NEXT:    knotw %k0, %k1
+; CHECK-NEXT:    vmovdqu32 %zmm1, %zmm0 {%k1} {z}
+; CHECK-NEXT:    retq
   %sub_r = sub <16 x i32> %a, %b
   %cmp.i2.i = icmp sgt <16 x i32> %sub_r, %a
   %sext.i3.i = sext <16 x i1> %cmp.i2.i to <16 x i32>
@@ -148,12 +179,15 @@ define <16 x i32> @test14(<16 x i32>%a, <16 x i32>%b) {
   ret <16 x i32>%res
 }
 
-; CHECK-LABEL: test15
-; CHECK: vpcmpgtq
-; CHECK-NOT: vpcmp
-; CHECK: vmovdqu64 {{.*}}{%k1} {z}
-; CHECK: ret
 define <8 x i64> @test15(<8 x i64>%a, <8 x i64>%b) {
+; CHECK-LABEL: test15:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpsubq %zmm1, %zmm0, %zmm1
+; CHECK-NEXT:    vpcmpgtq %zmm0, %zmm1, %k0
+; CHECK-NEXT:    knotw %k0, %k0
+; CHECK-NEXT:    knotw %k0, %k1
+; CHECK-NEXT:    vmovdqu64 %zmm1, %zmm0 {%k1} {z}
+; CHECK-NEXT:    retq
   %sub_r = sub <8 x i64> %a, %b
   %cmp.i2.i = icmp sgt <8 x i64> %sub_r, %a
   %sext.i3.i = sext <8 x i1> %cmp.i2.i to <8 x i64>
@@ -162,54 +196,65 @@ define <8 x i64> @test15(<8 x i64>%a, <8 x i64>%b) {
   ret <8 x i64>%res
 }
 
-; CHECK-LABEL: @test16
-; CHECK: vpcmpled
-; CHECK: vmovdqa32
-; CHECK: ret
 define <16 x i32> @test16(<16 x i32> %x, <16 x i32> %y) nounwind {
+; CHECK-LABEL: test16:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpled %zmm0, %zmm1, %k1
+; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %mask = icmp sge <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %y
   ret <16 x i32> %max
 }
 
-; CHECK-LABEL: @test17
-; CHECK: vpcmpgtd (%rdi)
-; CHECK: vmovdqa32
-; CHECK: ret
 define <16 x i32> @test17(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %y.ptr) nounwind {
+; CHECK-LABEL: test17:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpgtd (%rdi), %zmm0, %k1
+; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %y = load <16 x i32>* %y.ptr, align 4
   %mask = icmp sgt <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %x1
   ret <16 x i32> %max
 }
 
-; CHECK-LABEL: @test18
-; CHECK: vpcmpled (%rdi)
-; CHECK: vmovdqa32
-; CHECK: ret
 define <16 x i32> @test18(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %y.ptr) nounwind {
+; CHECK-LABEL: test18:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpled (%rdi), %zmm0, %k1
+; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %y = load <16 x i32>* %y.ptr, align 4
   %mask = icmp sle <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %x1
   ret <16 x i32> %max
 }
 
-; CHECK-LABEL: @test19
-; CHECK: vpcmpleud (%rdi)
-; CHECK: vmovdqa32
-; CHECK: ret
 define <16 x i32> @test19(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %y.ptr) nounwind {
+; CHECK-LABEL: test19:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpleud (%rdi), %zmm0, %k1
+; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %y = load <16 x i32>* %y.ptr, align 4
   %mask = icmp ule <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %x1
   ret <16 x i32> %max
 }
 
-; CHECK-LABEL: @test20
-; CHECK: vpcmpeqd %zmm{{.*{%k[1-7]}}}
-; CHECK: vmovdqa32
-; CHECK: ret
 define <16 x i32> @test20(<16 x i32> %x, <16 x i32> %y, <16 x i32> %x1, <16 x i32> %y1) nounwind {
+; CHECK-LABEL: test20:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqd %zmm1, %zmm0, %k1
+; CHECK-NEXT:    vpcmpeqd %zmm3, %zmm2, %k1 {%k1}
+; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %mask1 = icmp eq <16 x i32> %x1, %y1
   %mask0 = icmp eq <16 x i32> %x, %y
   %mask = select <16 x i1> %mask0, <16 x i1> %mask1, <16 x i1> zeroinitializer
@@ -217,11 +262,14 @@ define <16 x i32> @test20(<16 x i32> %x, <16 x i32> %y, <16 x i32> %x1, <16 x i3
   ret <16 x i32> %max
 }
 
-; CHECK-LABEL: @test21
-; CHECK: vpcmpleq %zmm{{.*{%k[1-7]}}}
-; CHECK: vmovdqa64
-; CHECK: ret
 define <8 x i64> @test21(<8 x i64> %x, <8 x i64> %y, <8 x i64> %x1, <8 x i64> %y1) nounwind {
+; CHECK-LABEL: test21:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpleq %zmm1, %zmm0, %k1
+; CHECK-NEXT:    vpcmpleq %zmm2, %zmm3, %k1 {%k1}
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k1}
+; CHECK-NEXT:    vmovaps %zmm2, %zmm0
+; CHECK-NEXT:    retq
   %mask1 = icmp sge <8 x i64> %x1, %y1
   %mask0 = icmp sle <8 x i64> %x, %y
   %mask = select <8 x i1> %mask0, <8 x i1> %mask1, <8 x i1> zeroinitializer
@@ -229,11 +277,14 @@ define <8 x i64> @test21(<8 x i64> %x, <8 x i64> %y, <8 x i64> %x1, <8 x i64> %y
   ret <8 x i64> %max
 }
 
-; CHECK-LABEL: @test22
-; CHECK: vpcmpgtq (%rdi){{.*{%k[1-7]}}}
-; CHECK: vmovdqa64
-; CHECK: ret
 define <8 x i64> @test22(<8 x i64> %x, <8 x i64>* %y.ptr, <8 x i64> %x1, <8 x i64> %y1) nounwind {
+; CHECK-LABEL: test22:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpgtq %zmm2, %zmm1, %k1
+; CHECK-NEXT:    vpcmpgtq (%rdi), %zmm0, %k1 {%k1}
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %mask1 = icmp sgt <8 x i64> %x1, %y1
   %y = load <8 x i64>* %y.ptr, align 4
   %mask0 = icmp sgt <8 x i64> %x, %y
@@ -242,11 +293,14 @@ define <8 x i64> @test22(<8 x i64> %x, <8 x i64>* %y.ptr, <8 x i64> %x1, <8 x i6
   ret <8 x i64> %max
 }
 
-; CHECK-LABEL: @test23
-; CHECK: vpcmpleud (%rdi){{.*{%k[1-7]}}}
-; CHECK: vmovdqa32
-; CHECK: ret
 define <16 x i32> @test23(<16 x i32> %x, <16 x i32>* %y.ptr, <16 x i32> %x1, <16 x i32> %y1) nounwind {
+; CHECK-LABEL: test23:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpled %zmm1, %zmm2, %k1
+; CHECK-NEXT:    vpcmpleud (%rdi), %zmm0, %k1 {%k1}
+; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %mask1 = icmp sge <16 x i32> %x1, %y1
   %y = load <16 x i32>* %y.ptr, align 4
   %mask0 = icmp ule <16 x i32> %x, %y
@@ -255,11 +309,13 @@ define <16 x i32> @test23(<16 x i32> %x, <16 x i32>* %y.ptr, <16 x i32> %x1, <16
   ret <16 x i32> %max
 }
 
-; CHECK-LABEL: test24
-; CHECK: vpcmpeqq (%rdi){1to8}
-; CHECK: vmovdqa64
-; CHECK: ret
 define <8 x i64> @test24(<8 x i64> %x, <8 x i64> %x1, i64* %yb.ptr) nounwind {
+; CHECK-LABEL: test24:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqq (%rdi){1to8}, %zmm0, %k1
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %yb = load i64* %yb.ptr, align 4
   %y.0 = insertelement <8 x i64> undef, i64 %yb, i32 0
   %y = shufflevector <8 x i64> %y.0, <8 x i64> undef, <8 x i32> zeroinitializer
@@ -268,11 +324,13 @@ define <8 x i64> @test24(<8 x i64> %x, <8 x i64> %x1, i64* %yb.ptr) nounwind {
   ret <8 x i64> %max
 }
 
-; CHECK-LABEL: test25
-; CHECK: vpcmpled (%rdi){1to16}
-; CHECK: vmovdqa32
-; CHECK: ret
 define <16 x i32> @test25(<16 x i32> %x, i32* %yb.ptr, <16 x i32> %x1) nounwind {
+; CHECK-LABEL: test25:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpled (%rdi){1to16}, %zmm0, %k1
+; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %yb = load i32* %yb.ptr, align 4
   %y.0 = insertelement <16 x i32> undef, i32 %yb, i32 0
   %y = shufflevector <16 x i32> %y.0, <16 x i32> undef, <16 x i32> zeroinitializer
@@ -281,11 +339,14 @@ define <16 x i32> @test25(<16 x i32> %x, i32* %yb.ptr, <16 x i32> %x1) nounwind
   ret <16 x i32> %max
 }
 
-; CHECK-LABEL: test26
-; CHECK: vpcmpgtd (%rdi){1to16}{{.*{%k[1-7]}}}
-; CHECK: vmovdqa32
-; CHECK: ret
 define <16 x i32> @test26(<16 x i32> %x, i32* %yb.ptr, <16 x i32> %x1, <16 x i32> %y1) nounwind {
+; CHECK-LABEL: test26:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpled %zmm1, %zmm2, %k1
+; CHECK-NEXT:    vpcmpgtd (%rdi){1to16}, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %mask1 = icmp sge <16 x i32> %x1, %y1
   %yb = load i32* %yb.ptr, align 4
   %y.0 = insertelement <16 x i32> undef, i32 %yb, i32 0
@@ -296,11 +357,14 @@ define <16 x i32> @test26(<16 x i32> %x, i32* %yb.ptr, <16 x i32> %x1, <16 x i32
   ret <16 x i32> %max
 }
 
-; CHECK-LABEL: test27
-; CHECK: vpcmpleq (%rdi){1to8}{{.*{%k[1-7]}}}
-; CHECK: vmovdqa64
-; CHECK: ret
 define <8 x i64> @test27(<8 x i64> %x, i64* %yb.ptr, <8 x i64> %x1, <8 x i64> %y1) nounwind {
+; CHECK-LABEL: test27:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpleq %zmm1, %zmm2, %k1
+; CHECK-NEXT:    vpcmpleq (%rdi){1to8}, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    retq
   %mask1 = icmp sge <8 x i64> %x1, %y1
   %yb = load i64* %yb.ptr, align 4
   %y.0 = insertelement <8 x i64> undef, i64 %yb, i32 0