Merge "Update aosp/master LLVM for rebase to r230699."

1 files changed, 403 insertions, 152 deletions

diff --git a/test/CodeGen/X86/avx512vl-intrinsics.ll b/test/CodeGen/X86/avx512vl-intrinsics.ll
index fa19084..fe347bd 100644
--- a/test/CodeGen/X86/avx512vl-intrinsics.ll
+++ b/test/CodeGen/X86/avx512vl-intrinsics.ll
@@ -67,244 +67,244 @@ define i8 @test_mask_pcmpgt_q_256(<4 x i64> %a, <4 x i64> %b, i8 %mask) {
 declare i8 @llvm.x86.avx512.mask.pcmpgt.q.256(<4 x i64>, <4 x i64>, i8)
 
 define <8 x i8> @test_cmp_d_256(<8 x i32> %a0, <8 x i32> %a1) {
-; CHECK_LABEL: test_cmp_d_256
+; CHECK-LABEL: test_cmp_d_256
 ; CHECK: vpcmpeqd %ymm1, %ymm0, %k0 ##
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 -1)
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 0, i8 -1)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltd %ymm1, %ymm0, %k0 ##
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 -1)
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 1, i8 -1)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpled %ymm1, %ymm0, %k0 ##
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 -1)
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 2, i8 -1)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunordd %ymm1, %ymm0, %k0 ##
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 -1)
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 3, i8 -1)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpneqd %ymm1, %ymm0, %k0 ##
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 -1)
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 4, i8 -1)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltd %ymm1, %ymm0, %k0 ##
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 -1)
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 5, i8 -1)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnled %ymm1, %ymm0, %k0 ##
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 -1)
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 6, i8 -1)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmpordd %ymm1, %ymm0, %k0 ##
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 7, i8 -1)
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 7, i8 -1)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
 define <8 x i8> @test_mask_cmp_d_256(<8 x i32> %a0, <8 x i32> %a1, i8 %mask) {
-; CHECK_LABEL: test_mask_cmp_d_256
+; CHECK-LABEL: test_mask_cmp_d_256
 ; CHECK: vpcmpeqd %ymm1, %ymm0, %k0 {%k1} ##
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 %mask)
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltd %ymm1, %ymm0, %k0 {%k1} ##
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 %mask)
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 1, i8 %mask)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpled %ymm1, %ymm0, %k0 {%k1} ##
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 %mask)
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 2, i8 %mask)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunordd %ymm1, %ymm0, %k0 {%k1} ##
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 %mask)
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 3, i8 %mask)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpneqd %ymm1, %ymm0, %k0 {%k1} ##
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 %mask)
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 4, i8 %mask)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltd %ymm1, %ymm0, %k0 {%k1} ##
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 %mask)
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 5, i8 %mask)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnled %ymm1, %ymm0, %k0 {%k1} ##
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 %mask)
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 6, i8 %mask)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmpordd %ymm1, %ymm0, %k0 {%k1} ##
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 7, i8 %mask)
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 7, i8 %mask)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
-declare i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32>, <8 x i32>, i32, i8) nounwind readnone
+declare i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32>, <8 x i32>, i8, i8) nounwind readnone
 
 define <8 x i8> @test_ucmp_d_256(<8 x i32> %a0, <8 x i32> %a1) {
-; CHECK_LABEL: test_ucmp_d_256
+; CHECK-LABEL: test_ucmp_d_256
 ; CHECK: vpcmpequd %ymm1, %ymm0, %k0 ##
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 -1)
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 0, i8 -1)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltud %ymm1, %ymm0, %k0 ##
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 -1)
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 1, i8 -1)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpleud %ymm1, %ymm0, %k0 ##
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 -1)
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 2, i8 -1)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunordud %ymm1, %ymm0, %k0 ##
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 -1)
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 3, i8 -1)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpnequd %ymm1, %ymm0, %k0 ##
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 -1)
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 4, i8 -1)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltud %ymm1, %ymm0, %k0 ##
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 -1)
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 5, i8 -1)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnleud %ymm1, %ymm0, %k0 ##
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 -1)
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 6, i8 -1)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmpordud %ymm1, %ymm0, %k0 ##
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 7, i8 -1)
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 7, i8 -1)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
 define <8 x i8> @test_mask_ucmp_d_256(<8 x i32> %a0, <8 x i32> %a1, i8 %mask) {
-; CHECK_LABEL: test_mask_ucmp_d_256
+; CHECK-LABEL: test_mask_ucmp_d_256
 ; CHECK: vpcmpequd %ymm1, %ymm0, %k0 {%k1} ##
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 %mask)
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltud %ymm1, %ymm0, %k0 {%k1} ##
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 %mask)
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 1, i8 %mask)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpleud %ymm1, %ymm0, %k0 {%k1} ##
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 %mask)
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 2, i8 %mask)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunordud %ymm1, %ymm0, %k0 {%k1} ##
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 %mask)
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 3, i8 %mask)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpnequd %ymm1, %ymm0, %k0 {%k1} ##
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 %mask)
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 4, i8 %mask)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltud %ymm1, %ymm0, %k0 {%k1} ##
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 %mask)
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 5, i8 %mask)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnleud %ymm1, %ymm0, %k0 {%k1} ##
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 %mask)
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 6, i8 %mask)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmpordud %ymm1, %ymm0, %k0 {%k1} ##
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 7, i8 %mask)
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i8 7, i8 %mask)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
-declare i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32>, <8 x i32>, i32, i8) nounwind readnone
+declare i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32>, <8 x i32>, i8, i8) nounwind readnone
 
 define <8 x i8> @test_cmp_q_256(<4 x i64> %a0, <4 x i64> %a1) {
-; CHECK_LABEL: test_cmp_q_256
+; CHECK-LABEL: test_cmp_q_256
 ; CHECK: vpcmpeqq %ymm1, %ymm0, %k0 ##
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 -1)
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 0, i8 -1)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltq %ymm1, %ymm0, %k0 ##
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 -1)
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 1, i8 -1)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpleq %ymm1, %ymm0, %k0 ##
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 -1)
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 2, i8 -1)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunordq %ymm1, %ymm0, %k0 ##
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 -1)
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 3, i8 -1)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpneqq %ymm1, %ymm0, %k0 ##
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 -1)
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 4, i8 -1)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltq %ymm1, %ymm0, %k0 ##
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 -1)
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 5, i8 -1)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnleq %ymm1, %ymm0, %k0 ##
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 -1)
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 6, i8 -1)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmpordq %ymm1, %ymm0, %k0 ##
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 7, i8 -1)
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 7, i8 -1)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
 define <8 x i8> @test_mask_cmp_q_256(<4 x i64> %a0, <4 x i64> %a1, i8 %mask) {
-; CHECK_LABEL: test_mask_cmp_q_256
+; CHECK-LABEL: test_mask_cmp_q_256
 ; CHECK: vpcmpeqq %ymm1, %ymm0, %k0 {%k1} ##
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 %mask)
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltq %ymm1, %ymm0, %k0 {%k1} ##
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 %mask)
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 1, i8 %mask)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpleq %ymm1, %ymm0, %k0 {%k1} ##
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 %mask)
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 2, i8 %mask)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunordq %ymm1, %ymm0, %k0 {%k1} ##
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 %mask)
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 3, i8 %mask)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpneqq %ymm1, %ymm0, %k0 {%k1} ##
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 %mask)
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 4, i8 %mask)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltq %ymm1, %ymm0, %k0 {%k1} ##
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 %mask)
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 5, i8 %mask)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnleq %ymm1, %ymm0, %k0 {%k1} ##
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 %mask)
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 6, i8 %mask)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmpordq %ymm1, %ymm0, %k0 {%k1} ##
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 7, i8 %mask)
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 7, i8 %mask)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
-declare i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64>, <4 x i64>, i32, i8) nounwind readnone
+declare i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64>, <4 x i64>, i8, i8) nounwind readnone
 
 define <8 x i8> @test_ucmp_q_256(<4 x i64> %a0, <4 x i64> %a1) {
-; CHECK_LABEL: test_ucmp_q_256
+; CHECK-LABEL: test_ucmp_q_256
 ; CHECK: vpcmpequq %ymm1, %ymm0, %k0 ##
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 -1)
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 0, i8 -1)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltuq %ymm1, %ymm0, %k0 ##
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 -1)
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 1, i8 -1)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpleuq %ymm1, %ymm0, %k0 ##
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 -1)
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 2, i8 -1)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunorduq %ymm1, %ymm0, %k0 ##
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 -1)
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 3, i8 -1)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpnequq %ymm1, %ymm0, %k0 ##
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 -1)
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 4, i8 -1)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltuq %ymm1, %ymm0, %k0 ##
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 -1)
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 5, i8 -1)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnleuq %ymm1, %ymm0, %k0 ##
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 -1)
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 6, i8 -1)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmporduq %ymm1, %ymm0, %k0 ##
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 7, i8 -1)
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 7, i8 -1)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
 define <8 x i8> @test_mask_ucmp_q_256(<4 x i64> %a0, <4 x i64> %a1, i8 %mask) {
-; CHECK_LABEL: test_mask_ucmp_q_256
+; CHECK-LABEL: test_mask_ucmp_q_256
 ; CHECK: vpcmpequq %ymm1, %ymm0, %k0 {%k1} ##
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 %mask)
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltuq %ymm1, %ymm0, %k0 {%k1} ##
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 %mask)
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 1, i8 %mask)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpleuq %ymm1, %ymm0, %k0 {%k1} ##
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 %mask)
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 2, i8 %mask)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunorduq %ymm1, %ymm0, %k0 {%k1} ##
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 %mask)
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 3, i8 %mask)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpnequq %ymm1, %ymm0, %k0 {%k1} ##
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 %mask)
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 4, i8 %mask)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltuq %ymm1, %ymm0, %k0 {%k1} ##
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 %mask)
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 5, i8 %mask)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnleuq %ymm1, %ymm0, %k0 {%k1} ##
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 %mask)
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 6, i8 %mask)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmporduq %ymm1, %ymm0, %k0 {%k1} ##
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 7, i8 %mask)
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i8 7, i8 %mask)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
-declare i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64>, <4 x i64>, i32, i8) nounwind readnone
+declare i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64>, <4 x i64>, i8, i8) nounwind readnone
 
 ; 128-bit
 
@@ -373,241 +373,492 @@ define i8 @test_mask_pcmpgt_q_128(<2 x i64> %a, <2 x i64> %b, i8 %mask) {
 declare i8 @llvm.x86.avx512.mask.pcmpgt.q.128(<2 x i64>, <2 x i64>, i8)
 
 define <8 x i8> @test_cmp_d_128(<4 x i32> %a0, <4 x i32> %a1) {
-; CHECK_LABEL: test_cmp_d_128
+; CHECK-LABEL: test_cmp_d_128
 ; CHECK: vpcmpeqd %xmm1, %xmm0, %k0 ##
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 -1)
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 0, i8 -1)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltd %xmm1, %xmm0, %k0 ##
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 -1)
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 1, i8 -1)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpled %xmm1, %xmm0, %k0 ##
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 -1)
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 2, i8 -1)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunordd %xmm1, %xmm0, %k0 ##
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 -1)
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 3, i8 -1)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpneqd %xmm1, %xmm0, %k0 ##
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 -1)
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 4, i8 -1)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltd %xmm1, %xmm0, %k0 ##
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 -1)
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 5, i8 -1)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnled %xmm1, %xmm0, %k0 ##
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 -1)
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 6, i8 -1)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmpordd %xmm1, %xmm0, %k0 ##
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 7, i8 -1)
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 7, i8 -1)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
 define <8 x i8> @test_mask_cmp_d_128(<4 x i32> %a0, <4 x i32> %a1, i8 %mask) {
-; CHECK_LABEL: test_mask_cmp_d_128
+; CHECK-LABEL: test_mask_cmp_d_128
 ; CHECK: vpcmpeqd %xmm1, %xmm0, %k0 {%k1} ##
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 %mask)
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltd %xmm1, %xmm0, %k0 {%k1} ##
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 %mask)
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 1, i8 %mask)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpled %xmm1, %xmm0, %k0 {%k1} ##
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 %mask)
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 2, i8 %mask)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunordd %xmm1, %xmm0, %k0 {%k1} ##
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 %mask)
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 3, i8 %mask)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpneqd %xmm1, %xmm0, %k0 {%k1} ##
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 %mask)
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 4, i8 %mask)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltd %xmm1, %xmm0, %k0 {%k1} ##
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 %mask)
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 5, i8 %mask)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnled %xmm1, %xmm0, %k0 {%k1} ##
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 %mask)
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 6, i8 %mask)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmpordd %xmm1, %xmm0, %k0 {%k1} ##
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 7, i8 %mask)
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 7, i8 %mask)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
-declare i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32>, <4 x i32>, i32, i8) nounwind readnone
+declare i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32>, <4 x i32>, i8, i8) nounwind readnone
 
 define <8 x i8> @test_ucmp_d_128(<4 x i32> %a0, <4 x i32> %a1) {
-; CHECK_LABEL: test_ucmp_d_128
+; CHECK-LABEL: test_ucmp_d_128
 ; CHECK: vpcmpequd %xmm1, %xmm0, %k0 ##
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 -1)
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 0, i8 -1)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltud %xmm1, %xmm0, %k0 ##
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 -1)
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 1, i8 -1)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpleud %xmm1, %xmm0, %k0 ##
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 -1)
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 2, i8 -1)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunordud %xmm1, %xmm0, %k0 ##
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 -1)
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 3, i8 -1)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpnequd %xmm1, %xmm0, %k0 ##
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 -1)
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 4, i8 -1)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltud %xmm1, %xmm0, %k0 ##
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 -1)
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 5, i8 -1)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnleud %xmm1, %xmm0, %k0 ##
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 -1)
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 6, i8 -1)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmpordud %xmm1, %xmm0, %k0 ##
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 7, i8 -1)
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 7, i8 -1)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
 define <8 x i8> @test_mask_ucmp_d_128(<4 x i32> %a0, <4 x i32> %a1, i8 %mask) {
-; CHECK_LABEL: test_mask_ucmp_d_128
+; CHECK-LABEL: test_mask_ucmp_d_128
 ; CHECK: vpcmpequd %xmm1, %xmm0, %k0 {%k1} ##
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 %mask)
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltud %xmm1, %xmm0, %k0 {%k1} ##
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 %mask)
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 1, i8 %mask)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpleud %xmm1, %xmm0, %k0 {%k1} ##
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 %mask)
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 2, i8 %mask)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunordud %xmm1, %xmm0, %k0 {%k1} ##
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 %mask)
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 3, i8 %mask)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpnequd %xmm1, %xmm0, %k0 {%k1} ##
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 %mask)
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 4, i8 %mask)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltud %xmm1, %xmm0, %k0 {%k1} ##
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 %mask)
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 5, i8 %mask)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnleud %xmm1, %xmm0, %k0 {%k1} ##
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 %mask)
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 6, i8 %mask)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmpordud %xmm1, %xmm0, %k0 {%k1} ##
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 7, i8 %mask)
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i8 7, i8 %mask)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
-declare i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32>, <4 x i32>, i32, i8) nounwind readnone
+declare i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32>, <4 x i32>, i8, i8) nounwind readnone
 
 define <8 x i8> @test_cmp_q_128(<2 x i64> %a0, <2 x i64> %a1) {
-; CHECK_LABEL: test_cmp_q_128
+; CHECK-LABEL: test_cmp_q_128
 ; CHECK: vpcmpeqq %xmm1, %xmm0, %k0 ##
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 -1)
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 0, i8 -1)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltq %xmm1, %xmm0, %k0 ##
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 -1)
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 1, i8 -1)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpleq %xmm1, %xmm0, %k0 ##
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 -1)
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 2, i8 -1)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunordq %xmm1, %xmm0, %k0 ##
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 -1)
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 3, i8 -1)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpneqq %xmm1, %xmm0, %k0 ##
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 -1)
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 4, i8 -1)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltq %xmm1, %xmm0, %k0 ##
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 -1)
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 5, i8 -1)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnleq %xmm1, %xmm0, %k0 ##
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 -1)
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 6, i8 -1)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmpordq %xmm1, %xmm0, %k0 ##
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 7, i8 -1)
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 7, i8 -1)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
 define <8 x i8> @test_mask_cmp_q_128(<2 x i64> %a0, <2 x i64> %a1, i8 %mask) {
-; CHECK_LABEL: test_mask_cmp_q_128
+; CHECK-LABEL: test_mask_cmp_q_128
 ; CHECK: vpcmpeqq %xmm1, %xmm0, %k0 {%k1} ##
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 %mask)
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltq %xmm1, %xmm0, %k0 {%k1} ##
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 %mask)
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 1, i8 %mask)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpleq %xmm1, %xmm0, %k0 {%k1} ##
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 %mask)
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 2, i8 %mask)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunordq %xmm1, %xmm0, %k0 {%k1} ##
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 %mask)
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 3, i8 %mask)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpneqq %xmm1, %xmm0, %k0 {%k1} ##
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 %mask)
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 4, i8 %mask)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltq %xmm1, %xmm0, %k0 {%k1} ##
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 %mask)
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 5, i8 %mask)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnleq %xmm1, %xmm0, %k0 {%k1} ##
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 %mask)
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 6, i8 %mask)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmpordq %xmm1, %xmm0, %k0 {%k1} ##
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 7, i8 %mask)
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 7, i8 %mask)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
-declare i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64>, <2 x i64>, i32, i8) nounwind readnone
+declare i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64>, <2 x i64>, i8, i8) nounwind readnone
 
 define <8 x i8> @test_ucmp_q_128(<2 x i64> %a0, <2 x i64> %a1) {
-; CHECK_LABEL: test_ucmp_q_128
+; CHECK-LABEL: test_ucmp_q_128
 ; CHECK: vpcmpequq %xmm1, %xmm0, %k0 ##
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 -1)
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 0, i8 -1)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltuq %xmm1, %xmm0, %k0 ##
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 -1)
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 1, i8 -1)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpleuq %xmm1, %xmm0, %k0 ##
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 -1)
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 2, i8 -1)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunorduq %xmm1, %xmm0, %k0 ##
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 -1)
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 3, i8 -1)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpnequq %xmm1, %xmm0, %k0 ##
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 -1)
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 4, i8 -1)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltuq %xmm1, %xmm0, %k0 ##
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 -1)
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 5, i8 -1)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnleuq %xmm1, %xmm0, %k0 ##
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 -1)
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 6, i8 -1)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmporduq %xmm1, %xmm0, %k0 ##
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 7, i8 -1)
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 7, i8 -1)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
 define <8 x i8> @test_mask_ucmp_q_128(<2 x i64> %a0, <2 x i64> %a1, i8 %mask) {
-; CHECK_LABEL: test_mask_ucmp_q_128
+; CHECK-LABEL: test_mask_ucmp_q_128
 ; CHECK: vpcmpequq %xmm1, %xmm0, %k0 {%k1} ##
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 %mask)
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
 ; CHECK: vpcmpltuq %xmm1, %xmm0, %k0 {%k1} ##
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 %mask)
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 1, i8 %mask)
   %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
 ; CHECK: vpcmpleuq %xmm1, %xmm0, %k0 {%k1} ##
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 %mask)
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 2, i8 %mask)
   %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
 ; CHECK: vpcmpunorduq %xmm1, %xmm0, %k0 {%k1} ##
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 %mask)
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 3, i8 %mask)
   %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
 ; CHECK: vpcmpnequq %xmm1, %xmm0, %k0 {%k1} ##
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 %mask)
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 4, i8 %mask)
   %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
 ; CHECK: vpcmpnltuq %xmm1, %xmm0, %k0 {%k1} ##
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 %mask)
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 5, i8 %mask)
   %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
 ; CHECK: vpcmpnleuq %xmm1, %xmm0, %k0 {%k1} ##
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 %mask)
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 6, i8 %mask)
   %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
 ; CHECK: vpcmporduq %xmm1, %xmm0, %k0 {%k1} ##
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 7, i8 %mask)
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i8 7, i8 %mask)
   %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
   ret <8 x i8> %vec7
 }
 
-declare i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64>, <2 x i64>, i32, i8) nounwind readnone
+declare i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64>, <2 x i64>, i8, i8) nounwind readnone
+
+; CHECK-LABEL: compr1
+; CHECK: vcompresspd %zmm0
+define void @compr1(i8* %addr, <8 x double> %data, i8 %mask) {
+  call void @llvm.x86.avx512.mask.compress.store.pd.512(i8* %addr, <8 x double> %data, i8 %mask)
+  ret void
+}
+
+declare void @llvm.x86.avx512.mask.compress.store.pd.512(i8* %addr, <8 x double> %data, i8 %mask)
+
+; CHECK-LABEL: compr2
+; CHECK: vcompresspd %ymm0
+define void @compr2(i8* %addr, <4 x double> %data, i8 %mask) {
+  call void @llvm.x86.avx512.mask.compress.store.pd.256(i8* %addr, <4 x double> %data, i8 %mask)
+  ret void
+}
+
+declare void @llvm.x86.avx512.mask.compress.store.pd.256(i8* %addr, <4 x double> %data, i8 %mask)
+
+; CHECK-LABEL: compr3
+; CHECK: vcompressps %xmm0
+define void @compr3(i8* %addr, <4 x float> %data, i8 %mask) {
+  call void @llvm.x86.avx512.mask.compress.store.ps.128(i8* %addr, <4 x float> %data, i8 %mask)
+  ret void
+}
+
+declare void @llvm.x86.avx512.mask.compress.store.ps.128(i8* %addr, <4 x float> %data, i8 %mask)
+
+; CHECK-LABEL: compr4
+; CHECK: vcompresspd %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf2,0xfd,0xc9,0x8a,0xc0]
+define <8 x double> @compr4(i8* %addr, <8 x double> %data, i8 %mask) {
+  %res = call <8 x double> @llvm.x86.avx512.mask.compress.pd.512(<8 x double> %data, <8 x double> zeroinitializer, i8 %mask)
+  ret <8 x double> %res
+}
+
+declare <8 x double> @llvm.x86.avx512.mask.compress.pd.512(<8 x double> %data, <8 x double> %src0, i8 %mask)
+
+; CHECK-LABEL: compr5
+; CHECK: vcompresspd %ymm0, %ymm1 {%k1}  ## encoding: [0x62,0xf2,0xfd,0x29,0x8a,0xc1]
+define <4 x double> @compr5(<4 x double> %data, <4 x double> %src0, i8 %mask) {
+  %res = call <4 x double> @llvm.x86.avx512.mask.compress.pd.256( <4 x double> %data, <4 x double> %src0, i8 %mask)
+  ret <4 x double> %res
+}
+
+declare <4 x double> @llvm.x86.avx512.mask.compress.pd.256(<4 x double> %data, <4 x double> %src0, i8 %mask)
+
+; CHECK-LABEL: compr6
+; CHECK: vcompressps %xmm0
+define <4 x float> @compr6(<4 x float> %data, i8 %mask) {
+  %res = call <4 x float> @llvm.x86.avx512.mask.compress.ps.128(<4 x float> %data, <4 x float>zeroinitializer, i8 %mask)
+  ret <4 x float> %res
+}
+
+declare <4 x float> @llvm.x86.avx512.mask.compress.ps.128(<4 x float> %data, <4 x float> %src0, i8 %mask)
+
+; CHECK-LABEL: compr7
+; CHECK-NOT: vcompress
+; CHECK: vmovapd
+define void @compr7(i8* %addr, <8 x double> %data) {
+  call void @llvm.x86.avx512.mask.compress.store.pd.512(i8* %addr, <8 x double> %data, i8 -1)
+  ret void
+}
+
+; CHECK-LABEL: compr8
+; CHECK-NOT: vcompressps %xmm0
+define <4 x float> @compr8(<4 x float> %data) {
+  %res = call <4 x float> @llvm.x86.avx512.mask.compress.ps.128(<4 x float> %data, <4 x float>zeroinitializer, i8 -1)
+  ret <4 x float> %res
+}
+
+; CHECK-LABEL: compr9
+; CHECK: vpcompressq %zmm0, (%rdi) {%k1}  ## encoding: [0x62,0xf2,0xfd,0x49,0x8b,0x07]
+define void @compr9(i8* %addr, <8 x i64> %data, i8 %mask) {
+  call void @llvm.x86.avx512.mask.compress.store.q.512(i8* %addr, <8 x i64> %data, i8 %mask)
+  ret void
+}
+
+declare void @llvm.x86.avx512.mask.compress.store.q.512(i8* %addr, <8 x i64> %data, i8 %mask)
+
+; CHECK-LABEL: compr10
+; CHECK: vpcompressd %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x8b,0xc0]
+define <4 x i32> @compr10(<4 x i32> %data, i8 %mask) {
+  %res = call <4 x i32> @llvm.x86.avx512.mask.compress.d.128(<4 x i32> %data, <4 x i32>zeroinitializer, i8 %mask)
+  ret <4 x i32> %res
+}
+
+declare <4 x i32> @llvm.x86.avx512.mask.compress.d.128(<4 x i32> %data, <4 x i32> %src0, i8 %mask)
+
+; Expand
+
+; CHECK-LABEL: expand1
+; CHECK: vexpandpd (%rdi), %zmm0 {%k1}  ## encoding: [0x62,0xf2,0xfd,0x49,0x88,0x07]
+define <8 x double> @expand1(i8* %addr, <8 x double> %data, i8 %mask) {
+  %res = call <8 x double> @llvm.x86.avx512.mask.expand.load.pd.512(i8* %addr, <8 x double> %data, i8 %mask)
+  ret <8 x double> %res
+}
+
+declare <8 x double> @llvm.x86.avx512.mask.expand.load.pd.512(i8* %addr, <8 x double> %data, i8 %mask)
+
+; CHECK-LABEL: expand2
+; CHECK: vexpandpd (%rdi), %ymm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x88,0x07]
+define <4 x double> @expand2(i8* %addr, <4 x double> %data, i8 %mask) {
+  %res = call <4 x double> @llvm.x86.avx512.mask.expand.load.pd.256(i8* %addr, <4 x double> %data, i8 %mask)
+  ret <4 x double> %res
+}
+
+declare <4 x double> @llvm.x86.avx512.mask.expand.load.pd.256(i8* %addr, <4 x double> %data, i8 %mask)
+
+; CHECK-LABEL: expand3
+; CHECK: vexpandps (%rdi), %xmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x88,0x07]
+define <4 x float> @expand3(i8* %addr, <4 x float> %data, i8 %mask) {
+  %res = call <4 x float> @llvm.x86.avx512.mask.expand.load.ps.128(i8* %addr, <4 x float> %data, i8 %mask)
+  ret <4 x float> %res
+}
+
+declare <4 x float> @llvm.x86.avx512.mask.expand.load.ps.128(i8* %addr, <4 x float> %data, i8 %mask)
+
+; CHECK-LABEL: expand4
+; CHECK: vexpandpd %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf2,0xfd,0xc9,0x88,0xc0]
+define <8 x double> @expand4(i8* %addr, <8 x double> %data, i8 %mask) {
+  %res = call <8 x double> @llvm.x86.avx512.mask.expand.pd.512(<8 x double> %data, <8 x double> zeroinitializer, i8 %mask)
+  ret <8 x double> %res
+}
+
+declare <8 x double> @llvm.x86.avx512.mask.expand.pd.512(<8 x double> %data, <8 x double> %src0, i8 %mask)
+
+; CHECK-LABEL: expand5
+; CHECK: vexpandpd %ymm0, %ymm1 {%k1}  ## encoding: [0x62,0xf2,0xfd,0x29,0x88,0xc8]
+define <4 x double> @expand5(<4 x double> %data, <4 x double> %src0, i8 %mask) {
+  %res = call <4 x double> @llvm.x86.avx512.mask.expand.pd.256( <4 x double> %data, <4 x double> %src0, i8 %mask)
+  ret <4 x double> %res
+}
+
+declare <4 x double> @llvm.x86.avx512.mask.expand.pd.256(<4 x double> %data, <4 x double> %src0, i8 %mask)
+
+; CHECK-LABEL: expand6
+; CHECK: vexpandps %xmm0
+define <4 x float> @expand6(<4 x float> %data, i8 %mask) {
+  %res = call <4 x float> @llvm.x86.avx512.mask.expand.ps.128(<4 x float> %data, <4 x float>zeroinitializer, i8 %mask)
+  ret <4 x float> %res
+}
+
+declare <4 x float> @llvm.x86.avx512.mask.expand.ps.128(<4 x float> %data, <4 x float> %src0, i8 %mask)
+
+; CHECK-LABEL: expand7
+; CHECK-NOT: vexpand
+; CHECK: vmovapd
+define <8 x double> @expand7(i8* %addr, <8 x double> %data) {
+  %res = call <8 x double> @llvm.x86.avx512.mask.expand.load.pd.512(i8* %addr, <8 x double> %data, i8 -1)
+  ret <8 x double> %res
+}
+
+; CHECK-LABEL: expand8
+; CHECK-NOT: vexpandps %xmm0
+define <4 x float> @expand8(<4 x float> %data) {
+  %res = call <4 x float> @llvm.x86.avx512.mask.expand.ps.128(<4 x float> %data, <4 x float>zeroinitializer, i8 -1)
+  ret <4 x float> %res
+}
+
+; CHECK-LABEL: expand9
+; CHECK: vpexpandq (%rdi), %zmm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x89,0x07]
+define <8 x i64> @expand9(i8* %addr, <8 x i64> %data, i8 %mask) {
+  %res = call <8 x i64> @llvm.x86.avx512.mask.expand.load.q.512(i8* %addr, <8 x i64> %data, i8 %mask)
+  ret <8 x i64> %res
+}
+
+declare <8 x i64> @llvm.x86.avx512.mask.expand.load.q.512(i8* %addr, <8 x i64> %data, i8 %mask)
+
+; CHECK-LABEL: expand10
+; CHECK: vpexpandd %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x89,0xc0]
+define <4 x i32> @expand10(<4 x i32> %data, i8 %mask) {
+  %res = call <4 x i32> @llvm.x86.avx512.mask.expand.d.128(<4 x i32> %data, <4 x i32>zeroinitializer, i8 %mask)
+  ret <4 x i32> %res
+}
+
+declare <4 x i32> @llvm.x86.avx512.mask.expand.d.128(<4 x i32> %data, <4 x i32> %src0, i8 %mask)
+
+define <8 x float> @test_x86_mask_blend_ps_256(i8 %a0, <8 x float> %a1, <8 x float> %a2) {
+  ; CHECK: vblendmps %ymm1, %ymm0
+  %res = call <8 x float> @llvm.x86.avx512.mask.blend.ps.256(<8 x float> %a1, <8 x float> %a2, i8 %a0) ; <<8 x float>> [#uses=1]
+  ret <8 x float> %res
+}
+
+declare <8 x float> @llvm.x86.avx512.mask.blend.ps.256(<8 x float>, <8 x float>, i8) nounwind readonly
+
+define <4 x double> @test_x86_mask_blend_pd_256(i8 %a0, <4 x double> %a1, <4 x double> %a2) {
+  ; CHECK: vblendmpd %ymm1, %ymm0
+  %res = call <4 x double> @llvm.x86.avx512.mask.blend.pd.256(<4 x double> %a1, <4 x double> %a2, i8 %a0) ; <<4 x double>> [#uses=1]
+  ret <4 x double> %res
+}
+
+define <4 x double> @test_x86_mask_blend_pd_256_memop(<4 x double> %a, <4 x double>* %ptr, i8 %mask) {
+  ; CHECK-LABEL: test_x86_mask_blend_pd_256_memop
+  ; CHECK: vblendmpd (%
+  %b = load <4 x double>* %ptr
+  %res = call <4 x double> @llvm.x86.avx512.mask.blend.pd.256(<4 x double> %a, <4 x double> %b, i8 %mask) ; <<4 x double>> [#uses=1]
+  ret <4 x double> %res
+}
+declare <4 x double> @llvm.x86.avx512.mask.blend.pd.256(<4 x double>, <4 x double>, i8) nounwind readonly
+
+; CHECK-LABEL: test_x86_mask_blend_d_256
+; CHECK: vpblendmd
+define <8 x i32> @test_x86_mask_blend_d_256(i8 %a0, <8 x i32> %a1, <8 x i32> %a2) {
+  %res = call <8 x i32> @llvm.x86.avx512.mask.blend.d.256(<8 x i32> %a1, <8 x i32> %a2, i8 %a0) ; <<8 x i32>> [#uses=1]
+  ret <8 x i32> %res
+}
+declare <8 x i32> @llvm.x86.avx512.mask.blend.d.256(<8 x i32>, <8 x i32>, i8) nounwind readonly
+
+define <4 x i64> @test_x86_mask_blend_q_256(i8 %a0, <4 x i64> %a1, <4 x i64> %a2) {
+  ; CHECK: vpblendmq
+  %res = call <4 x i64> @llvm.x86.avx512.mask.blend.q.256(<4 x i64> %a1, <4 x i64> %a2, i8 %a0) ; <<4 x i64>> [#uses=1]
+  ret <4 x i64> %res
+}
+declare <4 x i64> @llvm.x86.avx512.mask.blend.q.256(<4 x i64>, <4 x i64>, i8) nounwind readonly
+
+define <4 x float> @test_x86_mask_blend_ps_128(i8 %a0, <4 x float> %a1, <4 x float> %a2) {
+  ; CHECK: vblendmps %xmm1, %xmm0
+  %res = call <4 x float> @llvm.x86.avx512.mask.blend.ps.128(<4 x float> %a1, <4 x float> %a2, i8 %a0) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+
+declare <4 x float> @llvm.x86.avx512.mask.blend.ps.128(<4 x float>, <4 x float>, i8) nounwind readonly
+
+define <2 x double> @test_x86_mask_blend_pd_128(i8 %a0, <2 x double> %a1, <2 x double> %a2) {
+  ; CHECK: vblendmpd %xmm1, %xmm0
+  %res = call <2 x double> @llvm.x86.avx512.mask.blend.pd.128(<2 x double> %a1, <2 x double> %a2, i8 %a0) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+
+define <2 x double> @test_x86_mask_blend_pd_128_memop(<2 x double> %a, <2 x double>* %ptr, i8 %mask) {
+  ; CHECK-LABEL: test_x86_mask_blend_pd_128_memop
+  ; CHECK: vblendmpd (%
+  %b = load <2 x double>* %ptr
+  %res = call <2 x double> @llvm.x86.avx512.mask.blend.pd.128(<2 x double> %a, <2 x double> %b, i8 %mask) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.avx512.mask.blend.pd.128(<2 x double>, <2 x double>, i8) nounwind readonly
+
+define <4 x i32> @test_x86_mask_blend_d_128(i8 %a0, <4 x i32> %a1, <4 x i32> %a2) {
+  ; CHECK: vpblendmd
+  %res = call <4 x i32> @llvm.x86.avx512.mask.blend.d.128(<4 x i32> %a1, <4 x i32> %a2, i8 %a0) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.avx512.mask.blend.d.128(<4 x i32>, <4 x i32>, i8) nounwind readonly
+
+define <2 x i64> @test_x86_mask_blend_q_128(i8 %a0, <2 x i64> %a1, <2 x i64> %a2) {
+  ; CHECK: vpblendmq
+  %res = call <2 x i64> @llvm.x86.avx512.mask.blend.q.128(<2 x i64> %a1, <2 x i64> %a2, i8 %a0) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.avx512.mask.blend.q.128(<2 x i64>, <2 x i64>, i8) nounwind readonly