AVX-512: Added all forms of FP compare instructions for KNL and SKX.

Added intrinsics for the instructions. CC parameter of the intrinsics was changed from i8 to i32 according to the spec.

By Igor Breger (igor.breger@intel.com)

llvm-svn: 236714

3 files changed, 414 insertions, 189 deletions

diff --git a/llvm/test/CodeGen/X86/avx512-intrinsics.ll b/llvm/test/CodeGen/X86/avx512-intrinsics.ll
index 59a9c71a336..d4129e14c95 100644
--- a/llvm/test/CodeGen/X86/avx512-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512-intrinsics.ll
@@ -392,17 +392,17 @@ declare <8 x i64> @llvm.x86.avx512.mask.blend.q.512(<8 x i64>, <8 x i64>, i8) no
 
  define i16 @test_cmpps(<16 x float> %a, <16 x float> %b) {
  ;CHECK: vcmpleps {sae}{{.*}}encoding: [0x62,0xf1,0x7c,0x18,0xc2,0xc1,0x02]
-   %res = call i16 @llvm.x86.avx512.mask.cmp.ps.512(<16 x float> %a, <16 x float> %b, i8 2, i16 -1, i32 8)
+   %res = call i16 @llvm.x86.avx512.mask.cmp.ps.512(<16 x float> %a, <16 x float> %b, i32 2, i16 -1, i32 8)
    ret i16 %res
  }
- declare i16 @llvm.x86.avx512.mask.cmp.ps.512(<16 x float> , <16 x float> , i8, i16, i32)
+ declare i16 @llvm.x86.avx512.mask.cmp.ps.512(<16 x float> , <16 x float> , i32, i16, i32)
 
  define i8 @test_cmppd(<8 x double> %a, <8 x double> %b) {
  ;CHECK: vcmpneqpd %zmm{{.*}}encoding: [0x62,0xf1,0xfd,0x48,0xc2,0xc1,0x04]
-   %res = call i8 @llvm.x86.avx512.mask.cmp.pd.512(<8 x double> %a, <8 x double> %b, i8 4, i8 -1, i32 4)
+   %res = call i8 @llvm.x86.avx512.mask.cmp.pd.512(<8 x double> %a, <8 x double> %b, i32 4, i8 -1, i32 4)
    ret i8 %res
  }
- declare i8 @llvm.x86.avx512.mask.cmp.pd.512(<8 x double> , <8 x double> , i8, i8, i32)
+ declare i8 @llvm.x86.avx512.mask.cmp.pd.512(<8 x double> , <8 x double> , i32, i8, i32)
 
  ; cvt intrinsics
  define <16 x float> @test_cvtdq2ps(<16 x i32> %a) {
diff --git a/llvm/test/CodeGen/X86/avx512-vec-cmp.ll b/llvm/test/CodeGen/X86/avx512-vec-cmp.ll
index 4808ea9f854..e1f6276c6ef 100644
--- a/llvm/test/CodeGen/X86/avx512-vec-cmp.ll
+++ b/llvm/test/CodeGen/X86/avx512-vec-cmp.ll
@@ -1,36 +1,37 @@
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s --check-prefix=KNL
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=skx | FileCheck %s --check-prefix=SKX
 
 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
+; KNL-LABEL: test1:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vcmpleps %zmm1, %zmm0, %k1
+; KNL-NEXT:    vmovaps %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-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
 }
 
 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
+; KNL-LABEL: test2:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vcmplepd %zmm1, %zmm0, %k1
+; KNL-NEXT:    vmovapd %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-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
 }
 
 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
+; KNL-LABEL: test3:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpeqd (%rdi), %zmm0, %k1
+; KNL-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    retq
   %y = load <16 x i32>, <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
@@ -38,98 +39,120 @@ define <16 x i32> @test3(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %yp) nounwin
 }
 
 define <16 x i32> @test4_unsigned(<16 x i32> %x, <16 x i32> %y, <16 x i32> %x1) nounwind {
-; CHECK-LABEL: test4_unsigned:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpnltud %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm2, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
-; CHECK-NEXT:    retq
+; KNL-LABEL: test4_unsigned:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpnltud %zmm1, %zmm0, %k1
+; KNL-NEXT:    vmovdqa32 %zmm2, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    retq
   %mask = icmp uge <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x1, <16 x i32> %y
   ret <16 x i32> %max
 }
 
 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
+; KNL-LABEL: test5:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpeqq %zmm1, %zmm0, %k1
+; KNL-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-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
 }
 
 define <8 x i64> @test6_unsigned(<8 x i64> %x, <8 x i64> %y, <8 x i64> %x1) nounwind {
-; CHECK-LABEL: test6_unsigned:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
-; CHECK-NEXT:    retq
+; KNL-LABEL: test6_unsigned:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k1
+; KNL-NEXT:    vmovdqa64 %zmm2, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    retq
   %mask = icmp ugt <8 x i64> %x, %y
   %max = select <8 x i1> %mask, <8 x i64> %x1, <8 x i64> %y
   ret <8 x i64> %max
 }
 
 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
+; KNL-LABEL: test7:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vxorps %xmm2, %xmm2, %xmm2
+; KNL-NEXT:    vcmpltps %xmm2, %xmm0, %xmm2
+; KNL-NEXT:    vblendvps %xmm2, %xmm0, %xmm1, %xmm0
+; KNL-NEXT:    retq
+; SKX-LABEL: test7:
+; SKX:       ## BB#0:
+; SKX:    vxorps   %xmm2, %xmm2, %xmm2
+; SKX:    vcmpltps %xmm2, %xmm0, %k1 
+; SKX:    vmovaps  %xmm0, %xmm1 {%k1}
+; SKX:    vmovaps  %zmm1, %zmm0
+; SKX:    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
 }
 
 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
+; KNL-LABEL: test8:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vxorpd %xmm2, %xmm2, %xmm2
+; KNL-NEXT:    vcmpltpd %xmm2, %xmm0, %xmm2
+; KNL-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
+; KNL-NEXT:    retq
+; SKX-LABEL: test8:
+; SKX:       ## BB#0:
+; SKX: vxorpd  %xmm2, %xmm2, %xmm2
+; SKX: vcmpltpd    %xmm2, %xmm0, %k1 
+; SKX: vmovapd %xmm0, %xmm1 {%k1}
+; SKX: vmovaps %zmm1, %zmm0
+; SKX: 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
 }
 
 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
+; KNL-LABEL: test9:
+; KNL:       ## BB#0:
+; KNL-NEXT:      ## kill: YMM1<def> YMM1<kill> ZMM1<def>
+; KNL-NEXT:      ## kill: YMM0<def> YMM0<kill> ZMM0<def>
+; KNL-NEXT:    vpcmpeqd %zmm1, %zmm0, %k1
+; KNL-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
+; KNL-NEXT:      ## kill: YMM0<def> YMM0<kill> ZMM0<kill>
+; KNL-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
 }
 
 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
+; KNL-LABEL: test10:
+; KNL:       ## BB#0:
+; KNL-NEXT:      ## kill: YMM1<def> YMM1<kill> ZMM1<def>
+; KNL-NEXT:      ## kill: YMM0<def> YMM0<kill> ZMM0<def>
+; KNL-NEXT:    vcmpeqps %zmm1, %zmm0, %k1
+; KNL-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
+; KNL-NEXT:      ## kill: YMM0<def> YMM0<kill> ZMM0<kill>
+; KNL-NEXT:    retq
+; SKX-LABEL: test10:
+; SKX:       ## BB#0:
+; SKX: vcmpeqps    %ymm1, %ymm0, %k1 
+; SKX: vmovaps %ymm0, %ymm1 {%k1}
+; SKX: vmovaps %zmm1, %zmm0
+; SKX: 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
 }
 
 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
+; KNL-LABEL: test11_unsigned:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpmaxud %ymm1, %ymm0, %ymm0
+; KNL-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
@@ -137,25 +160,25 @@ define <8 x i32> @test11_unsigned(<8 x i32> %x, <8 x i32> %y) nounwind {
 
 
 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
+; KNL-LABEL: test12:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpeqq %zmm2, %zmm0, %k0
+; KNL-NEXT:    vpcmpeqq %zmm3, %zmm1, %k1
+; KNL-NEXT:    kunpckbw %k0, %k1, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:      ## kill: AX<def> AX<kill> EAX<kill>
+; KNL-NEXT:    retq
   %res = icmp eq <16 x i64> %a, %b
   %res1 = bitcast <16 x i1> %res to i16
   ret i16 %res1
 }
 
 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
+; KNL-LABEL: test13:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vcmpeqps %zmm1, %zmm0, %k1
+; KNL-NEXT:    vpbroadcastd {{.*}}(%rip), %zmm0 {%k1} {z}
+; KNL-NEXT:    retq
 {
   %cmpvector_i = fcmp oeq <16 x float> %a, %b
   %conv = zext <16 x i1> %cmpvector_i to <16 x i32>
@@ -163,14 +186,14 @@ define <16 x i32> @test13(<16 x float>%a, <16 x float>%b)
 }
 
 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
+; KNL-LABEL: test14:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpsubd %zmm1, %zmm0, %zmm1
+; KNL-NEXT:    vpcmpgtd %zmm0, %zmm1, %k0
+; KNL-NEXT:    knotw %k0, %k0
+; KNL-NEXT:    knotw %k0, %k1
+; KNL-NEXT:    vmovdqu32 %zmm1, %zmm0 {%k1} {z}
+; KNL-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>
@@ -180,14 +203,14 @@ define <16 x i32> @test14(<16 x i32>%a, <16 x i32>%b) {
 }
 
 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
+; KNL-LABEL: test15:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpsubq %zmm1, %zmm0, %zmm1
+; KNL-NEXT:    vpcmpgtq %zmm0, %zmm1, %k0
+; KNL-NEXT:    knotw %k0, %k0
+; KNL-NEXT:    knotw %k0, %k1
+; KNL-NEXT:    vmovdqu64 %zmm1, %zmm0 {%k1} {z}
+; KNL-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>
@@ -197,24 +220,24 @@ define <8 x i64> @test15(<8 x i64>%a, <8 x i64>%b) {
 }
 
 define <16 x i32> @test16(<16 x i32> %x, <16 x i32> %y, <16 x i32> %x1) nounwind {
-; CHECK-LABEL: test16:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpled %zmm0, %zmm1, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm2, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
-; CHECK-NEXT:    retq
+; KNL-LABEL: test16:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpled %zmm0, %zmm1, %k1
+; KNL-NEXT:    vmovdqa32 %zmm2, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    retq
   %mask = icmp sge <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x1, <16 x i32> %y
   ret <16 x i32> %max
 }
 
 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
+; KNL-LABEL: test17:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpgtd (%rdi), %zmm0, %k1
+; KNL-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    retq
   %y = load <16 x i32>, <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
@@ -222,12 +245,12 @@ define <16 x i32> @test17(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %y.ptr) nou
 }
 
 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
+; KNL-LABEL: test18:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpled (%rdi), %zmm0, %k1
+; KNL-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    retq
   %y = load <16 x i32>, <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
@@ -235,12 +258,12 @@ define <16 x i32> @test18(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %y.ptr) nou
 }
 
 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
+; KNL-LABEL: test19:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpleud (%rdi), %zmm0, %k1
+; KNL-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    retq
   %y = load <16 x i32>, <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
@@ -248,13 +271,13 @@ define <16 x i32> @test19(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %y.ptr) nou
 }
 
 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
+; KNL-LABEL: test20:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpeqd %zmm1, %zmm0, %k1
+; KNL-NEXT:    vpcmpeqd %zmm3, %zmm2, %k1 {%k1}
+; KNL-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-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
@@ -263,13 +286,13 @@ define <16 x i32> @test20(<16 x i32> %x, <16 x i32> %y, <16 x i32> %x1, <16 x i3
 }
 
 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
+; KNL-LABEL: test21:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpleq %zmm1, %zmm0, %k1
+; KNL-NEXT:    vpcmpleq %zmm2, %zmm3, %k1 {%k1}
+; KNL-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k1}
+; KNL-NEXT:    vmovaps %zmm2, %zmm0
+; KNL-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
@@ -278,13 +301,13 @@ define <8 x i64> @test21(<8 x i64> %x, <8 x i64> %y, <8 x i64> %x1, <8 x i64> %y
 }
 
 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
+; KNL-LABEL: test22:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpgtq %zmm2, %zmm1, %k1
+; KNL-NEXT:    vpcmpgtq (%rdi), %zmm0, %k1 {%k1}
+; KNL-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    retq
   %mask1 = icmp sgt <8 x i64> %x1, %y1
   %y = load <8 x i64>, <8 x i64>* %y.ptr, align 4
   %mask0 = icmp sgt <8 x i64> %x, %y
@@ -294,13 +317,13 @@ define <8 x i64> @test22(<8 x i64> %x, <8 x i64>* %y.ptr, <8 x i64> %x1, <8 x i6
 }
 
 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
+; KNL-LABEL: test23:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpled %zmm1, %zmm2, %k1
+; KNL-NEXT:    vpcmpleud (%rdi), %zmm0, %k1 {%k1}
+; KNL-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    retq
   %mask1 = icmp sge <16 x i32> %x1, %y1
   %y = load <16 x i32>, <16 x i32>* %y.ptr, align 4
   %mask0 = icmp ule <16 x i32> %x, %y
@@ -310,12 +333,12 @@ define <16 x i32> @test23(<16 x i32> %x, <16 x i32>* %y.ptr, <16 x i32> %x1, <16
 }
 
 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
+; KNL-LABEL: test24:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpeqq (%rdi){1to8}, %zmm0, %k1
+; KNL-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    retq
   %yb = load i64, 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
@@ -325,12 +348,12 @@ define <8 x i64> @test24(<8 x i64> %x, <8 x i64> %x1, i64* %yb.ptr) nounwind {
 }
 
 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
+; KNL-LABEL: test25:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpled (%rdi){1to16}, %zmm0, %k1
+; KNL-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    retq
   %yb = load i32, 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
@@ -340,13 +363,13 @@ define <16 x i32> @test25(<16 x i32> %x, i32* %yb.ptr, <16 x i32> %x1) nounwind
 }
 
 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
+; KNL-LABEL: test26:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpled %zmm1, %zmm2, %k1
+; KNL-NEXT:    vpcmpgtd (%rdi){1to16}, %zmm0, %k1 {%k1}
+; KNL-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    retq
   %mask1 = icmp sge <16 x i32> %x1, %y1
   %yb = load i32, i32* %yb.ptr, align 4
   %y.0 = insertelement <16 x i32> undef, i32 %yb, i32 0
@@ -358,13 +381,13 @@ define <16 x i32> @test26(<16 x i32> %x, i32* %yb.ptr, <16 x i32> %x1, <16 x i32
 }
 
 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
+; KNL-LABEL: test27:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vpcmpleq        %zmm1, %zmm2, %k1
+; KNL-NEXT:    vpcmpleq        (%rdi){1to8}, %zmm0, %k1 {%k1}
+; KNL-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    retq
   %mask1 = icmp sge <8 x i64> %x1, %y1
   %yb = load i64, i64* %yb.ptr, align 4
   %y.0 = insertelement <8 x i64> undef, i64 %yb, i32 0
@@ -375,10 +398,10 @@ define <8 x i64> @test27(<8 x i64> %x, i64* %yb.ptr, <8 x i64> %x1, <8 x i64> %y
   ret <8 x i64> %max
 }
 
-; CHECK-LABEL: test28
-; CHECK: vpcmpgtq
-; CHECK: vpcmpgtq
-; CHECK: kxorw
+; KNL-LABEL: test28
+; KNL: vpcmpgtq
+; KNL: vpcmpgtq
+; KNL: kxorw
 define <8 x i32>@test28(<8 x i64> %x, <8 x i64> %y, <8 x i64> %x1, <8 x i64> %y1) {
   %x_gt_y = icmp sgt <8 x i64> %x, %y
   %x1_gt_y1 = icmp sgt <8 x i64> %x1, %y1
@@ -387,14 +410,188 @@ define <8 x i32>@test28(<8 x i64> %x, <8 x i64> %y, <8 x i64> %x1, <8 x i64> %y1
   ret <8 x i32> %resse
 }
 
-; CHECK-LABEL: test29
-; CHECK: vpcmpgtd
-; CHECK: vpcmpgtd
-; CHECK: kxnorw
+; KNL-LABEL: test29
+; KNL: vpcmpgtd
+; KNL: vpcmpgtd
+; KNL: kxnorw
 define <16 x i8>@test29(<16 x i32> %x, <16 x i32> %y, <16 x i32> %x1, <16 x i32> %y1) {
   %x_gt_y = icmp sgt <16 x i32> %x, %y
   %x1_gt_y1 = icmp sgt <16 x i32> %x1, %y1
   %res = icmp ne <16 x i1>%x_gt_y, %x1_gt_y1
   %resse = sext <16 x i1>%res to <16 x i8>
   ret <16 x i8> %resse
-}
\ No newline at end of file
+}
+
+define <4 x double> @test30(<4 x double> %x, <4 x double> %y) nounwind {
+; SKX-LABEL: test30:
+; SKX: vcmpeqpd   %ymm1, %ymm0, %k1 
+; SKX: vmovapd    %ymm0, %ymm1 {%k1}
+
+  %mask = fcmp oeq <4 x double> %x, %y
+  %max = select <4 x i1> %mask, <4 x double> %x, <4 x double> %y
+  ret <4 x double> %max
+}
+
+define <2 x double> @test31(<2 x double> %x, <2 x double> %x1, <2 x double>* %yp) nounwind {
+; SKX-LABEL: test31:     
+; SKX: vcmpltpd        (%rdi), %xmm0, %k1 
+; SKX: vmovapd %xmm0, %xmm1 {%k1}
+
+  %y = load <2 x double>, <2 x double>* %yp, align 4
+  %mask = fcmp olt <2 x double> %x, %y
+  %max = select <2 x i1> %mask, <2 x double> %x, <2 x double> %x1
+  ret <2 x double> %max
+}
+
+define <4 x double> @test32(<4 x double> %x, <4 x double> %x1, <4 x double>* %yp) nounwind {
+; SKX-LABEL: test32:
+; SKX: vcmpltpd        (%rdi), %ymm0, %k1 
+; SKX: vmovapd %ymm0, %ymm1 {%k1}
+
+  %y = load <4 x double>, <4 x double>* %yp, align 4
+  %mask = fcmp ogt <4 x double> %y, %x
+  %max = select <4 x i1> %mask, <4 x double> %x, <4 x double> %x1
+  ret <4 x double> %max
+}
+
+define <8 x double> @test33(<8 x double> %x, <8 x double> %x1, <8 x double>* %yp) nounwind {
+; SKX-LABEL: test33:     
+; SKX: vcmpltpd        (%rdi), %zmm0, %k1 
+; SKX: vmovapd %zmm0, %zmm1 {%k1}
+  %y = load <8 x double>, <8 x double>* %yp, align 4
+  %mask = fcmp olt <8 x double> %x, %y
+  %max = select <8 x i1> %mask, <8 x double> %x, <8 x double> %x1
+  ret <8 x double> %max
+}
+
+define <4 x float> @test34(<4 x float> %x, <4 x float> %x1, <4 x float>* %yp) nounwind {
+; SKX-LABEL: test34:     
+; SKX: vcmpltps        (%rdi), %xmm0, %k1 
+; SKX: vmovaps %xmm0, %xmm1 {%k1}
+  %y = load <4 x float>, <4 x float>* %yp, align 4
+  %mask = fcmp olt <4 x float> %x, %y
+  %max = select <4 x i1> %mask, <4 x float> %x, <4 x float> %x1
+  ret <4 x float> %max
+}
+
+define <8 x float> @test35(<8 x float> %x, <8 x float> %x1, <8 x float>* %yp) nounwind {
+; SKX-LABEL: test35:
+; SKX: vcmpltps        (%rdi), %ymm0, %k1 
+; SKX: vmovaps %ymm0, %ymm1 {%k1}
+
+  %y = load <8 x float>, <8 x float>* %yp, align 4
+  %mask = fcmp ogt <8 x float> %y, %x
+  %max = select <8 x i1> %mask, <8 x float> %x, <8 x float> %x1
+  ret <8 x float> %max
+}
+
+define <16 x float> @test36(<16 x float> %x, <16 x float> %x1, <16 x float>* %yp) nounwind {
+; SKX-LABEL: test36:     
+; SKX: vcmpltps        (%rdi), %zmm0, %k1 
+; SKX: vmovaps %zmm0, %zmm1 {%k1}
+  %y = load <16 x float>, <16 x float>* %yp, align 4
+  %mask = fcmp olt <16 x float> %x, %y
+  %max = select <16 x i1> %mask, <16 x float> %x, <16 x float> %x1
+  ret <16 x float> %max
+}
+
+define <8 x double> @test37(<8 x double> %x, <8 x double> %x1, double* %ptr) nounwind {
+; SKX-LABEL: test37:                                
+; SKX: vcmpltpd  (%rdi){1to8}, %zmm0, %k1 
+; SKX: vmovapd %zmm0, %zmm1 {%k1}
+
+  %a = load double, double* %ptr
+  %v = insertelement <8 x double> undef, double %a, i32 0
+  %shuffle = shufflevector <8 x double> %v, <8 x double> undef, <8 x i32> zeroinitializer
+
+  %mask = fcmp ogt <8 x double> %shuffle, %x
+  %max = select <8 x i1> %mask, <8 x double> %x, <8 x double> %x1
+  ret <8 x double> %max
+}
+
+define <4 x double> @test38(<4 x double> %x, <4 x double> %x1, double* %ptr) nounwind {
+; SKX-LABEL: test38:                                
+; SKX: vcmpltpd  (%rdi){1to4}, %ymm0, %k1 
+; SKX: vmovapd %ymm0, %ymm1 {%k1}
+
+  %a = load double, double* %ptr
+  %v = insertelement <4 x double> undef, double %a, i32 0
+  %shuffle = shufflevector <4 x double> %v, <4 x double> undef, <4 x i32> zeroinitializer
+  
+  %mask = fcmp ogt <4 x double> %shuffle, %x
+  %max = select <4 x i1> %mask, <4 x double> %x, <4 x double> %x1
+  ret <4 x double> %max
+}
+
+define <2 x double> @test39(<2 x double> %x, <2 x double> %x1, double* %ptr) nounwind {
+; SKX-LABEL: test39:                                
+; SKX: vcmpltpd  (%rdi){1to2}, %xmm0, %k1 
+; SKX: vmovapd %xmm0, %xmm1 {%k1}
+
+  %a = load double, double* %ptr
+  %v = insertelement <2 x double> undef, double %a, i32 0
+  %shuffle = shufflevector <2 x double> %v, <2 x double> undef, <2 x i32> <i32 0, i32 0>
+  
+  %mask = fcmp ogt <2 x double> %shuffle, %x
+  %max = select <2 x i1> %mask, <2 x double> %x, <2 x double> %x1
+  ret <2 x double> %max
+}
+
+
+define <16  x float> @test40(<16  x float> %x, <16  x float> %x1, float* %ptr) nounwind {
+; SKX-LABEL: test40:                                
+; SKX: vcmpltps  (%rdi){1to16}, %zmm0, %k1 
+; SKX: vmovaps %zmm0, %zmm1 {%k1}
+
+  %a = load float, float* %ptr
+  %v = insertelement <16  x float> undef, float %a, i32 0
+  %shuffle = shufflevector <16  x float> %v, <16  x float> undef, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
+  
+  %mask = fcmp ogt <16  x float> %shuffle, %x
+  %max = select <16 x i1> %mask, <16  x float> %x, <16  x float> %x1
+  ret <16  x float> %max
+}
+
+define <8  x float> @test41(<8  x float> %x, <8  x float> %x1, float* %ptr) nounwind {
+; SKX-LABEL: test41:                                
+; SKX: vcmpltps  (%rdi){1to8}, %ymm0, %k1 
+; SKX: vmovaps %ymm0, %ymm1 {%k1}
+
+  %a = load float, float* %ptr
+  %v = insertelement <8  x float> undef, float %a, i32 0
+  %shuffle = shufflevector <8  x float> %v, <8  x float> undef, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
+  
+  %mask = fcmp ogt <8  x float> %shuffle, %x
+  %max = select <8 x i1> %mask, <8  x float> %x, <8  x float> %x1
+  ret <8  x float> %max
+}
+
+define <4  x float> @test42(<4  x float> %x, <4  x float> %x1, float* %ptr) nounwind {
+; SKX-LABEL: test42:                                
+; SKX: vcmpltps  (%rdi){1to4}, %xmm0, %k1 
+; SKX: vmovaps %xmm0, %xmm1 {%k1}
+  
+  %a = load float, float* %ptr
+  %v = insertelement <4  x float> undef, float %a, i32 0
+  %shuffle = shufflevector <4  x float> %v, <4  x float> undef, <4 x i32> <i32 0, i32 0, i32 0, i32 0>
+  
+  %mask = fcmp ogt <4  x float> %shuffle, %x
+  %max = select <4 x i1> %mask, <4  x float> %x, <4  x float> %x1
+  ret <4  x float> %max
+}
+
+define <8 x double> @test43(<8 x double> %x, <8 x double> %x1, double* %ptr,<8 x i1> %mask_in) nounwind {
+; SKX-LABEL: test43:                                
+; SKX: vpmovw2m  %xmm2, %k1
+; SKX: vcmpltpd  (%rdi){1to8}, %zmm0, %k1 {%k1}
+; SKX: vmovapd %zmm0, %zmm1 {%k1}
+
+  %a = load double, double* %ptr
+  %v = insertelement <8 x double> undef, double %a, i32 0
+  %shuffle = shufflevector <8 x double> %v, <8 x double> undef, <8 x i32> zeroinitializer
+  
+  %mask_cmp = fcmp ogt <8 x double> %shuffle, %x
+  %mask = and <8 x i1> %mask_cmp, %mask_in
+  %max = select <8 x i1> %mask, <8 x double> %x, <8 x double> %x1
+  ret <8 x double> %max
+}
diff --git a/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll b/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll
index 4db5df53848..64ec8166d50 100644
--- a/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll
@@ -2261,3 +2261,31 @@ define <4 x i64> @test_mask_andnot_epi64_rmbkz_256(<4 x i64> %a, i64* %ptr_b, i8
 }
 
 declare <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64>, <4 x i64>, <4 x i64>, i8)
+
+define i8 @test_cmpps_256(<8 x float> %a, <8 x float> %b) {
+ ;CHECK: vcmpleps  %ymm1, %ymm0, %k0  ## encoding: [0x62,0xf1,0x7c,0x28,0xc2,0xc1,0x02]
+   %res = call i8 @llvm.x86.avx512.mask.cmp.ps.256(<8 x float> %a, <8 x float> %b, i32 2, i8 -1)
+   ret i8 %res
+ }
+ declare i8 @llvm.x86.avx512.mask.cmp.ps.256(<8 x float> , <8 x float> , i32, i8)
+
+define i8 @test_cmpps_128(<4 x float> %a, <4 x float> %b) {
+ ;CHECK: vcmpleps  %xmm1, %xmm0, %k0  ## encoding: [0x62,0xf1,0x7c,0x08,0xc2,0xc1,0x02]
+   %res = call i8 @llvm.x86.avx512.mask.cmp.ps.128(<4 x float> %a, <4 x float> %b, i32 2, i8 -1)
+   ret i8 %res
+ }
+ declare i8 @llvm.x86.avx512.mask.cmp.ps.128(<4 x float> , <4 x float> , i32, i8)
+
+define i8 @test_cmppd_256(<4 x double> %a, <4 x double> %b) {
+ ;CHECK: vcmplepd  %ymm1, %ymm0, %k0  ## encoding: [0x62,0xf1,0xfd,0x28,0xc2,0xc1,0x02]
+   %res = call i8 @llvm.x86.avx512.mask.cmp.pd.256(<4 x double> %a, <4 x double> %b, i32 2, i8 -1)
+   ret i8 %res
+ }
+ declare i8 @llvm.x86.avx512.mask.cmp.pd.256(<4 x double> , <4 x double> , i32, i8)
+
+define i8 @test_cmppd_128(<2 x double> %a, <2 x double> %b) {
+ ;CHECK: vcmplepd  %xmm1, %xmm0, %k0  ## encoding: [0x62,0xf1,0xfd,0x08,0xc2,0xc1,0x02]
+   %res = call i8 @llvm.x86.avx512.mask.cmp.pd.128(<2 x double> %a, <2 x double> %b, i32 2, i8 -1)
+   ret i8 %res
+ }
+ declare i8 @llvm.x86.avx512.mask.cmp.pd.128(<2 x double> , <2 x double> , i32, i8)