From: Craig Topper <craig.topper@gmail.com>
Date: Thu, 16 Feb 2017 06:31:54 +0000 (+0000)
Subject: [AVX-512] Remove masked packss/packus intrinsics and autoupgrade to unmasked intrinsi... 
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=fdf52147044c334284dbde5b5f0b605f22781b68;p=llvm

[AVX-512] Remove masked packss/packus intrinsics and autoupgrade to unmasked intrinsics with select instructions. For 512-bit add new unmasked intrinsics.

The new 512-bit unmasked intrinsics will make it easy to handle these with the SSE/AVX intrinsics in InstCombine where we currently have a TODO.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@295290 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/include/llvm/IR/IntrinsicsX86.td b/include/llvm/IR/IntrinsicsX86.td
index 4802eb50c17..6545435d671 100644
--- a/include/llvm/IR/IntrinsicsX86.td
+++ b/include/llvm/IR/IntrinsicsX86.td
@@ -3913,42 +3913,18 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
 
 // Pack ops.
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
-  def int_x86_avx512_mask_packsswb_128 : GCCBuiltin<"__builtin_ia32_packsswb128_mask">,
-              Intrinsic<[llvm_v16i8_ty], [llvm_v8i16_ty, llvm_v8i16_ty,
-                         llvm_v16i8_ty, llvm_i16_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_packsswb_256 : GCCBuiltin<"__builtin_ia32_packsswb256_mask">,
-              Intrinsic<[llvm_v32i8_ty], [llvm_v16i16_ty,llvm_v16i16_ty,
-                         llvm_v32i8_ty, llvm_i32_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_packsswb_512 : GCCBuiltin<"__builtin_ia32_packsswb512_mask">,
-              Intrinsic<[llvm_v64i8_ty], [llvm_v32i16_ty,llvm_v32i16_ty,
-                         llvm_v64i8_ty, llvm_i64_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_packssdw_128 : GCCBuiltin<"__builtin_ia32_packssdw128_mask">,
-              Intrinsic<[llvm_v8i16_ty], [llvm_v4i32_ty, llvm_v4i32_ty,
-                         llvm_v8i16_ty, llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_packssdw_256 : GCCBuiltin<"__builtin_ia32_packssdw256_mask">,
-              Intrinsic<[llvm_v16i16_ty], [llvm_v8i32_ty, llvm_v8i32_ty,
-                         llvm_v16i16_ty, llvm_i16_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_packssdw_512 : GCCBuiltin<"__builtin_ia32_packssdw512_mask">,
-              Intrinsic<[llvm_v32i16_ty], [llvm_v16i32_ty, llvm_v16i32_ty,
-                         llvm_v32i16_ty, llvm_i32_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_packuswb_128 : GCCBuiltin<"__builtin_ia32_packuswb128_mask">,
-              Intrinsic<[llvm_v16i8_ty], [llvm_v8i16_ty, llvm_v8i16_ty,
-                         llvm_v16i8_ty, llvm_i16_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_packuswb_256 : GCCBuiltin<"__builtin_ia32_packuswb256_mask">,
-              Intrinsic<[llvm_v32i8_ty], [llvm_v16i16_ty,llvm_v16i16_ty,
-                         llvm_v32i8_ty, llvm_i32_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_packuswb_512 : GCCBuiltin<"__builtin_ia32_packuswb512_mask">,
-              Intrinsic<[llvm_v64i8_ty], [llvm_v32i16_ty,llvm_v32i16_ty,
-                         llvm_v64i8_ty, llvm_i64_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_packusdw_128 : GCCBuiltin<"__builtin_ia32_packusdw128_mask">,
-              Intrinsic<[llvm_v8i16_ty], [llvm_v4i32_ty, llvm_v4i32_ty,
-                         llvm_v8i16_ty, llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_packusdw_256 : GCCBuiltin<"__builtin_ia32_packusdw256_mask">,
-              Intrinsic<[llvm_v16i16_ty], [llvm_v8i32_ty, llvm_v8i32_ty,
-                         llvm_v16i16_ty, llvm_i16_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_packusdw_512 : GCCBuiltin<"__builtin_ia32_packusdw512_mask">,
-              Intrinsic<[llvm_v32i16_ty], [llvm_v16i32_ty, llvm_v16i32_ty,
-                         llvm_v32i16_ty, llvm_i32_ty], [IntrNoMem]>;
+  def int_x86_avx512_packsswb_512 : GCCBuiltin<"__builtin_ia32_packsswb512">,
+              Intrinsic<[llvm_v64i8_ty], [llvm_v32i16_ty,llvm_v32i16_ty],
+                        [IntrNoMem]>;
+  def int_x86_avx512_packssdw_512 : GCCBuiltin<"__builtin_ia32_packssdw512">,
+              Intrinsic<[llvm_v32i16_ty], [llvm_v16i32_ty, llvm_v16i32_ty],
+                         [IntrNoMem]>;
+  def int_x86_avx512_packuswb_512 : GCCBuiltin<"__builtin_ia32_packuswb512">,
+              Intrinsic<[llvm_v64i8_ty], [llvm_v32i16_ty,llvm_v32i16_ty],
+                         [IntrNoMem]>;
+  def int_x86_avx512_packusdw_512 : GCCBuiltin<"__builtin_ia32_packusdw512">,
+              Intrinsic<[llvm_v32i16_ty], [llvm_v16i32_ty, llvm_v16i32_ty],
+                         [IntrNoMem]>;
 }
 
 // Vector convert
diff --git a/lib/IR/AutoUpgrade.cpp b/lib/IR/AutoUpgrade.cpp
index 41f4159c559..a76c3a039f5 100644
--- a/lib/IR/AutoUpgrade.cpp
+++ b/lib/IR/AutoUpgrade.cpp
@@ -335,6 +335,10 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
          Name.startswith("avx512.mask.cvtudq2pd.") || // Added in 4.0
          Name.startswith("avx512.mask.pmul.dq.") || // Added in 4.0
          Name.startswith("avx512.mask.pmulu.dq.") || // Added in 4.0
+         Name.startswith("avx512.mask.packsswb.") || // Added in 4.0
+         Name.startswith("avx512.mask.packssdw.") || // Added in 4.0
+         Name.startswith("avx512.mask.packuswb.") || // Added in 4.0
+         Name.startswith("avx512.mask.packusdw.") || // Added in 4.0
          Name == "avx512.mask.add.pd.128" || // Added in 4.0
          Name == "avx512.mask.add.pd.256" || // Added in 4.0
          Name == "avx512.mask.add.ps.128" || // Added in 4.0
@@ -1538,6 +1542,42 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
       else
         llvm_unreachable("Unexpected intrinsic");
 
+      Rep = Builder.CreateCall(Intrinsic::getDeclaration(F->getParent(), IID),
+                               { CI->getArgOperand(0), CI->getArgOperand(1) });
+      Rep = EmitX86Select(Builder, CI->getArgOperand(3), Rep,
+                          CI->getArgOperand(2));
+    } else if (IsX86 && Name.startswith("avx512.mask.pack")) {
+      bool IsUnsigned = Name[16] == 'u';
+      bool IsDW = Name[18] == 'd';
+      VectorType *VecTy = cast<VectorType>(CI->getType());
+      Intrinsic::ID IID;
+      if (!IsUnsigned && !IsDW && VecTy->getPrimitiveSizeInBits() == 128)
+        IID = Intrinsic::x86_sse2_packsswb_128;
+      else if (!IsUnsigned && !IsDW && VecTy->getPrimitiveSizeInBits() == 256)
+        IID = Intrinsic::x86_avx2_packsswb;
+      else if (!IsUnsigned && !IsDW && VecTy->getPrimitiveSizeInBits() == 512)
+        IID = Intrinsic::x86_avx512_packsswb_512;
+      else if (!IsUnsigned && IsDW && VecTy->getPrimitiveSizeInBits() == 128)
+        IID = Intrinsic::x86_sse2_packssdw_128;
+      else if (!IsUnsigned && IsDW && VecTy->getPrimitiveSizeInBits() == 256)
+        IID = Intrinsic::x86_avx2_packssdw;
+      else if (!IsUnsigned && IsDW && VecTy->getPrimitiveSizeInBits() == 512)
+        IID = Intrinsic::x86_avx512_packssdw_512;
+      else if (IsUnsigned && !IsDW && VecTy->getPrimitiveSizeInBits() == 128)
+        IID = Intrinsic::x86_sse2_packuswb_128;
+      else if (IsUnsigned && !IsDW && VecTy->getPrimitiveSizeInBits() == 256)
+        IID = Intrinsic::x86_avx2_packuswb;
+      else if (IsUnsigned && !IsDW && VecTy->getPrimitiveSizeInBits() == 512)
+        IID = Intrinsic::x86_avx512_packuswb_512;
+      else if (IsUnsigned && IsDW && VecTy->getPrimitiveSizeInBits() == 128)
+        IID = Intrinsic::x86_sse41_packusdw;
+      else if (IsUnsigned && IsDW && VecTy->getPrimitiveSizeInBits() == 256)
+        IID = Intrinsic::x86_avx2_packusdw;
+      else if (IsUnsigned && IsDW && VecTy->getPrimitiveSizeInBits() == 512)
+        IID = Intrinsic::x86_avx512_packusdw_512;
+      else
+        llvm_unreachable("Unexpected intrinsic");
+
       Rep = Builder.CreateCall(Intrinsic::getDeclaration(F->getParent(), IID),
                                { CI->getArgOperand(0), CI->getArgOperand(1) });
       Rep = EmitX86Select(Builder, CI->getArgOperand(3), Rep,
diff --git a/lib/Target/X86/X86IntrinsicsInfo.h b/lib/Target/X86/X86IntrinsicsInfo.h
index 63a02af02fa..6365594334b 100644
--- a/lib/Target/X86/X86IntrinsicsInfo.h
+++ b/lib/Target/X86/X86IntrinsicsInfo.h
@@ -851,18 +851,6 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_mask_pabs_w_128, INTR_TYPE_1OP_MASK, X86ISD::ABS, 0),
   X86_INTRINSIC_DATA(avx512_mask_pabs_w_256, INTR_TYPE_1OP_MASK, X86ISD::ABS, 0),
   X86_INTRINSIC_DATA(avx512_mask_pabs_w_512, INTR_TYPE_1OP_MASK, X86ISD::ABS, 0),
-  X86_INTRINSIC_DATA(avx512_mask_packssdw_128, INTR_TYPE_2OP_MASK, X86ISD::PACKSS, 0),
-  X86_INTRINSIC_DATA(avx512_mask_packssdw_256, INTR_TYPE_2OP_MASK, X86ISD::PACKSS, 0),
-  X86_INTRINSIC_DATA(avx512_mask_packssdw_512, INTR_TYPE_2OP_MASK, X86ISD::PACKSS, 0),
-  X86_INTRINSIC_DATA(avx512_mask_packsswb_128, INTR_TYPE_2OP_MASK, X86ISD::PACKSS, 0),
-  X86_INTRINSIC_DATA(avx512_mask_packsswb_256, INTR_TYPE_2OP_MASK, X86ISD::PACKSS, 0),
-  X86_INTRINSIC_DATA(avx512_mask_packsswb_512, INTR_TYPE_2OP_MASK, X86ISD::PACKSS, 0),
-  X86_INTRINSIC_DATA(avx512_mask_packusdw_128, INTR_TYPE_2OP_MASK, X86ISD::PACKUS, 0),
-  X86_INTRINSIC_DATA(avx512_mask_packusdw_256, INTR_TYPE_2OP_MASK, X86ISD::PACKUS, 0),
-  X86_INTRINSIC_DATA(avx512_mask_packusdw_512, INTR_TYPE_2OP_MASK, X86ISD::PACKUS, 0),
-  X86_INTRINSIC_DATA(avx512_mask_packuswb_128, INTR_TYPE_2OP_MASK, X86ISD::PACKUS, 0),
-  X86_INTRINSIC_DATA(avx512_mask_packuswb_256, INTR_TYPE_2OP_MASK, X86ISD::PACKUS, 0),
-  X86_INTRINSIC_DATA(avx512_mask_packuswb_512, INTR_TYPE_2OP_MASK, X86ISD::PACKUS, 0),
   X86_INTRINSIC_DATA(avx512_mask_padds_b_128, INTR_TYPE_2OP_MASK, X86ISD::ADDS, 0),
   X86_INTRINSIC_DATA(avx512_mask_padds_b_256, INTR_TYPE_2OP_MASK, X86ISD::ADDS, 0),
   X86_INTRINSIC_DATA(avx512_mask_padds_b_512, INTR_TYPE_2OP_MASK, X86ISD::ADDS, 0),
@@ -1486,6 +1474,10 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
                      X86ISD::VPMADD52L, 0),
   X86_INTRINSIC_DATA(avx512_maskz_vpmadd52l_uq_512, FMA_OP_MASKZ,
                      X86ISD::VPMADD52L, 0),
+  X86_INTRINSIC_DATA(avx512_packssdw_512, INTR_TYPE_2OP, X86ISD::PACKSS, 0),
+  X86_INTRINSIC_DATA(avx512_packsswb_512, INTR_TYPE_2OP, X86ISD::PACKSS, 0),
+  X86_INTRINSIC_DATA(avx512_packusdw_512, INTR_TYPE_2OP, X86ISD::PACKUS, 0),
+  X86_INTRINSIC_DATA(avx512_packuswb_512, INTR_TYPE_2OP, X86ISD::PACKUS, 0),
   X86_INTRINSIC_DATA(avx512_pmul_dq_512, INTR_TYPE_2OP, X86ISD::PMULDQ, 0),
   X86_INTRINSIC_DATA(avx512_pmulu_dq_512, INTR_TYPE_2OP, X86ISD::PMULUDQ, 0),
   X86_INTRINSIC_DATA(avx512_psad_bw_512, INTR_TYPE_2OP, X86ISD::PSADBW, 0),
diff --git a/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll b/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll
index 507205ceb4f..0cda27a8349 100644
--- a/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll
+++ b/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll
@@ -992,3 +992,539 @@ define <64 x i8>@test_int_x86_avx512_mask_pshuf_b_512(<64 x i8> %x0, <64 x i8> %
   ret <64 x i8> %res2
 }
 
+define <32 x i16> @test_mask_packs_epi32_rr_512(<16 x i32> %a, <16 x i32> %b) {
+; AVX512BW-LABEL: test_mask_packs_epi32_rr_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpackssdw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rr_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpackssdw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packs_epi32_rrk_512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask) {
+; AVX512BW-LABEL: test_mask_packs_epi32_rrk_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %edi, %k1
+; AVX512BW-NEXT:    vpackssdw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rrk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackssdw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packs_epi32_rrkz_512(<16 x i32> %a, <16 x i32> %b, i32 %mask) {
+; AVX512BW-LABEL: test_mask_packs_epi32_rrkz_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %edi, %k1
+; AVX512BW-NEXT:    vpackssdw %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rrkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackssdw %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packs_epi32_rm_512(<16 x i32> %a, <16 x i32>* %ptr_b) {
+; AVX512BW-LABEL: test_mask_packs_epi32_rm_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpackssdw (%rdi), %zmm0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rm_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpackssdw (%eax), %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %b = load <16 x i32>, <16 x i32>* %ptr_b
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packs_epi32_rmk_512(<16 x i32> %a, <16 x i32>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
+; AVX512BW-LABEL: test_mask_packs_epi32_rmk_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %esi, %k1
+; AVX512BW-NEXT:    vpackssdw (%rdi), %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rmk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackssdw (%eax), %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
+  %b = load <16 x i32>, <16 x i32>* %ptr_b
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packs_epi32_rmkz_512(<16 x i32> %a, <16 x i32>* %ptr_b, i32 %mask) {
+; AVX512BW-LABEL: test_mask_packs_epi32_rmkz_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %esi, %k1
+; AVX512BW-NEXT:    vpackssdw (%rdi), %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rmkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackssdw (%eax), %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
+  %b = load <16 x i32>, <16 x i32>* %ptr_b
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packs_epi32_rmb_512(<16 x i32> %a, i32* %ptr_b) {
+; AVX512BW-LABEL: test_mask_packs_epi32_rmb_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpackssdw (%rdi){1to16}, %zmm0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rmb_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpackssdw (%eax){1to16}, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packs_epi32_rmbk_512(<16 x i32> %a, i32* %ptr_b, <32 x i16> %passThru, i32 %mask) {
+; AVX512BW-LABEL: test_mask_packs_epi32_rmbk_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %esi, %k1
+; AVX512BW-NEXT:    vpackssdw (%rdi){1to16}, %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rmbk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackssdw (%eax){1to16}, %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packs_epi32_rmbkz_512(<16 x i32> %a, i32* %ptr_b, i32 %mask) {
+; AVX512BW-LABEL: test_mask_packs_epi32_rmbkz_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %esi, %k1
+; AVX512BW-NEXT:    vpackssdw (%rdi){1to16}, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rmbkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackssdw (%eax){1to16}, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
+  ret <32 x i16> %res
+}
+
+declare <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32>, <16 x i32>, <32 x i16>, i32)
+
+define <64 x i8> @test_mask_packs_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
+; AVX512BW-LABEL: test_mask_packs_epi16_rr_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpacksswb %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi16_rr_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpacksswb %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 -1)
+  ret <64 x i8> %res
+}
+
+define <64 x i8> @test_mask_packs_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask) {
+; AVX512BW-LABEL: test_mask_packs_epi16_rrk_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovq %rdi, %k1
+; AVX512BW-NEXT:    vpacksswb %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi16_rrk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpacksswb %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
+  %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask)
+  ret <64 x i8> %res
+}
+
+define <64 x i8> @test_mask_packs_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i64 %mask) {
+; AVX512BW-LABEL: test_mask_packs_epi16_rrkz_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovq %rdi, %k1
+; AVX512BW-NEXT:    vpacksswb %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi16_rrkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpacksswb %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
+  %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 %mask)
+  ret <64 x i8> %res
+}
+
+define <64 x i8> @test_mask_packs_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
+; AVX512BW-LABEL: test_mask_packs_epi16_rm_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpacksswb (%rdi), %zmm0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi16_rm_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpacksswb (%eax), %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %b = load <32 x i16>, <32 x i16>* %ptr_b
+  %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 -1)
+  ret <64 x i8> %res
+}
+
+define <64 x i8> @test_mask_packs_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <64 x i8> %passThru, i64 %mask) {
+; AVX512BW-LABEL: test_mask_packs_epi16_rmk_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovq %rsi, %k1
+; AVX512BW-NEXT:    vpacksswb (%rdi), %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi16_rmk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpacksswb (%eax), %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
+  %b = load <32 x i16>, <32 x i16>* %ptr_b
+  %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask)
+  ret <64 x i8> %res
+}
+
+define <64 x i8> @test_mask_packs_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i64 %mask) {
+; AVX512BW-LABEL: test_mask_packs_epi16_rmkz_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovq %rsi, %k1
+; AVX512BW-NEXT:    vpacksswb (%rdi), %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi16_rmkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpacksswb (%eax), %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
+  %b = load <32 x i16>, <32 x i16>* %ptr_b
+  %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 %mask)
+  ret <64 x i8> %res
+}
+
+declare <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16>, <32 x i16>, <64 x i8>, i64)
+
+
+define <32 x i16> @test_mask_packus_epi32_rr_512(<16 x i32> %a, <16 x i32> %b) {
+; AVX512BW-LABEL: test_mask_packus_epi32_rr_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpackusdw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rr_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpackusdw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packus_epi32_rrk_512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask) {
+; AVX512BW-LABEL: test_mask_packus_epi32_rrk_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %edi, %k1
+; AVX512BW-NEXT:    vpackusdw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rrk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackusdw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packus_epi32_rrkz_512(<16 x i32> %a, <16 x i32> %b, i32 %mask) {
+; AVX512BW-LABEL: test_mask_packus_epi32_rrkz_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %edi, %k1
+; AVX512BW-NEXT:    vpackusdw %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rrkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackusdw %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packus_epi32_rm_512(<16 x i32> %a, <16 x i32>* %ptr_b) {
+; AVX512BW-LABEL: test_mask_packus_epi32_rm_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpackusdw (%rdi), %zmm0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rm_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpackusdw (%eax), %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %b = load <16 x i32>, <16 x i32>* %ptr_b
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packus_epi32_rmk_512(<16 x i32> %a, <16 x i32>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
+; AVX512BW-LABEL: test_mask_packus_epi32_rmk_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %esi, %k1
+; AVX512BW-NEXT:    vpackusdw (%rdi), %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rmk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackusdw (%eax), %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
+  %b = load <16 x i32>, <16 x i32>* %ptr_b
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packus_epi32_rmkz_512(<16 x i32> %a, <16 x i32>* %ptr_b, i32 %mask) {
+; AVX512BW-LABEL: test_mask_packus_epi32_rmkz_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %esi, %k1
+; AVX512BW-NEXT:    vpackusdw (%rdi), %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rmkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackusdw (%eax), %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
+  %b = load <16 x i32>, <16 x i32>* %ptr_b
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packus_epi32_rmb_512(<16 x i32> %a, i32* %ptr_b) {
+; AVX512BW-LABEL: test_mask_packus_epi32_rmb_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpackusdw (%rdi){1to16}, %zmm0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rmb_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpackusdw (%eax){1to16}, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packus_epi32_rmbk_512(<16 x i32> %a, i32* %ptr_b, <32 x i16> %passThru, i32 %mask) {
+; AVX512BW-LABEL: test_mask_packus_epi32_rmbk_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %esi, %k1
+; AVX512BW-NEXT:    vpackusdw (%rdi){1to16}, %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rmbk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackusdw (%eax){1to16}, %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
+  ret <32 x i16> %res
+}
+
+define <32 x i16> @test_mask_packus_epi32_rmbkz_512(<16 x i32> %a, i32* %ptr_b, i32 %mask) {
+; AVX512BW-LABEL: test_mask_packus_epi32_rmbkz_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %esi, %k1
+; AVX512BW-NEXT:    vpackusdw (%rdi){1to16}, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rmbkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackusdw (%eax){1to16}, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
+  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
+  ret <32 x i16> %res
+}
+
+declare <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32>, <16 x i32>, <32 x i16>, i32)
+
+define <64 x i8> @test_mask_packus_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
+; AVX512BW-LABEL: test_mask_packus_epi16_rr_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpackuswb %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi16_rr_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpackuswb %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 -1)
+  ret <64 x i8> %res
+}
+
+define <64 x i8> @test_mask_packus_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask) {
+; AVX512BW-LABEL: test_mask_packus_epi16_rrk_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovq %rdi, %k1
+; AVX512BW-NEXT:    vpackuswb %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi16_rrk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackuswb %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
+  %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask)
+  ret <64 x i8> %res
+}
+
+define <64 x i8> @test_mask_packus_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i64 %mask) {
+; AVX512BW-LABEL: test_mask_packus_epi16_rrkz_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovq %rdi, %k1
+; AVX512BW-NEXT:    vpackuswb %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi16_rrkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackuswb %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
+  %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 %mask)
+  ret <64 x i8> %res
+}
+
+define <64 x i8> @test_mask_packus_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
+; AVX512BW-LABEL: test_mask_packus_epi16_rm_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpackuswb (%rdi), %zmm0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi16_rm_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpackuswb (%eax), %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %b = load <32 x i16>, <32 x i16>* %ptr_b
+  %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 -1)
+  ret <64 x i8> %res
+}
+
+define <64 x i8> @test_mask_packus_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <64 x i8> %passThru, i64 %mask) {
+; AVX512BW-LABEL: test_mask_packus_epi16_rmk_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovq %rsi, %k1
+; AVX512BW-NEXT:    vpackuswb (%rdi), %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi16_rmk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackuswb (%eax), %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
+  %b = load <32 x i16>, <32 x i16>* %ptr_b
+  %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask)
+  ret <64 x i8> %res
+}
+
+define <64 x i8> @test_mask_packus_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i64 %mask) {
+; AVX512BW-LABEL: test_mask_packus_epi16_rmkz_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovq %rsi, %k1
+; AVX512BW-NEXT:    vpackuswb (%rdi), %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi16_rmkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackuswb (%eax), %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
+  %b = load <32 x i16>, <32 x i16>* %ptr_b
+  %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 %mask)
+  ret <64 x i8> %res
+}
+
+declare <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16>, <32 x i16>, <64 x i8>, i64)
diff --git a/test/CodeGen/X86/avx512bw-intrinsics.ll b/test/CodeGen/X86/avx512bw-intrinsics.ll
index d9a9a2d655b..5a6b49dbb0e 100644
--- a/test/CodeGen/X86/avx512bw-intrinsics.ll
+++ b/test/CodeGen/X86/avx512bw-intrinsics.ll
@@ -660,8 +660,8 @@ define <32 x i16> @test_mask_packs_epi32_rr_512(<16 x i32> %a, <16 x i32> %b) {
 ; AVX512F-32:       # BB#0:
 ; AVX512F-32-NEXT:    vpackssdw %zmm1, %zmm0, %zmm0
 ; AVX512F-32-NEXT:    retl
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> %a, <16 x i32> %b)
+  ret <32 x i16> %1
 }
 
 define <32 x i16> @test_mask_packs_epi32_rrk_512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask) {
@@ -678,8 +678,10 @@ define <32 x i16> @test_mask_packs_epi32_rrk_512(<16 x i32> %a, <16 x i32> %b, <
 ; AVX512F-32-NEXT:    vpackssdw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512F-32-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; AVX512F-32-NEXT:    retl
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> %a, <16 x i32> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> %passThru
+  ret <32 x i16> %3
 }
 
 define <32 x i16> @test_mask_packs_epi32_rrkz_512(<16 x i32> %a, <16 x i32> %b, i32 %mask) {
@@ -694,8 +696,10 @@ define <32 x i16> @test_mask_packs_epi32_rrkz_512(<16 x i32> %a, <16 x i32> %b,
 ; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
 ; AVX512F-32-NEXT:    vpackssdw %zmm1, %zmm0, %zmm0 {%k1} {z}
 ; AVX512F-32-NEXT:    retl
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> %a, <16 x i32> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> zeroinitializer
+  ret <32 x i16> %3
 }
 
 define <32 x i16> @test_mask_packs_epi32_rm_512(<16 x i32> %a, <16 x i32>* %ptr_b) {
@@ -710,8 +714,8 @@ define <32 x i16> @test_mask_packs_epi32_rm_512(<16 x i32> %a, <16 x i32>* %ptr_
 ; AVX512F-32-NEXT:    vpackssdw (%eax), %zmm0, %zmm0
 ; AVX512F-32-NEXT:    retl
   %b = load <16 x i32>, <16 x i32>* %ptr_b
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> %a, <16 x i32> %b)
+  ret <32 x i16> %1
 }
 
 define <32 x i16> @test_mask_packs_epi32_rmk_512(<16 x i32> %a, <16 x i32>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
@@ -730,8 +734,10 @@ define <32 x i16> @test_mask_packs_epi32_rmk_512(<16 x i32> %a, <16 x i32>* %ptr
 ; AVX512F-32-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; AVX512F-32-NEXT:    retl
   %b = load <16 x i32>, <16 x i32>* %ptr_b
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> %a, <16 x i32> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> %passThru
+  ret <32 x i16> %3
 }
 
 define <32 x i16> @test_mask_packs_epi32_rmkz_512(<16 x i32> %a, <16 x i32>* %ptr_b, i32 %mask) {
@@ -748,8 +754,10 @@ define <32 x i16> @test_mask_packs_epi32_rmkz_512(<16 x i32> %a, <16 x i32>* %pt
 ; AVX512F-32-NEXT:    vpackssdw (%eax), %zmm0, %zmm0 {%k1} {z}
 ; AVX512F-32-NEXT:    retl
   %b = load <16 x i32>, <16 x i32>* %ptr_b
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> %a, <16 x i32> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> zeroinitializer
+  ret <32 x i16> %3
 }
 
 define <32 x i16> @test_mask_packs_epi32_rmb_512(<16 x i32> %a, i32* %ptr_b) {
@@ -766,8 +774,8 @@ define <32 x i16> @test_mask_packs_epi32_rmb_512(<16 x i32> %a, i32* %ptr_b) {
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
   %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> %a, <16 x i32> %b)
+  ret <32 x i16> %1
 }
 
 define <32 x i16> @test_mask_packs_epi32_rmbk_512(<16 x i32> %a, i32* %ptr_b, <32 x i16> %passThru, i32 %mask) {
@@ -788,8 +796,10 @@ define <32 x i16> @test_mask_packs_epi32_rmbk_512(<16 x i32> %a, i32* %ptr_b, <3
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
   %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> %a, <16 x i32> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> %passThru
+  ret <32 x i16> %3
 }
 
 define <32 x i16> @test_mask_packs_epi32_rmbkz_512(<16 x i32> %a, i32* %ptr_b, i32 %mask) {
@@ -808,11 +818,13 @@ define <32 x i16> @test_mask_packs_epi32_rmbkz_512(<16 x i32> %a, i32* %ptr_b, i
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
   %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> %a, <16 x i32> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> zeroinitializer
+  ret <32 x i16> %3
 }
 
-declare <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32>, <16 x i32>, <32 x i16>, i32)
+declare <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32>, <16 x i32>)
 
 define <64 x i8> @test_mask_packs_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: test_mask_packs_epi16_rr_512:
@@ -824,8 +836,8 @@ define <64 x i8> @test_mask_packs_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512F-32:       # BB#0:
 ; AVX512F-32-NEXT:    vpacksswb %zmm1, %zmm0, %zmm0
 ; AVX512F-32-NEXT:    retl
-  %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 -1)
-  ret <64 x i8> %res
+  %1 = call <64 x i8> @llvm.x86.avx512.packsswb.512(<32 x i16> %a, <32 x i16> %b)
+  ret <64 x i8> %1
 }
 
 define <64 x i8> @test_mask_packs_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask) {
@@ -838,14 +850,14 @@ define <64 x i8> @test_mask_packs_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <6
 ;
 ; AVX512F-32-LABEL: test_mask_packs_epi16_rrk_512:
 ; AVX512F-32:       # BB#0:
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
-; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
 ; AVX512F-32-NEXT:    vpacksswb %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512F-32-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; AVX512F-32-NEXT:    retl
-  %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask)
-  ret <64 x i8> %res
+  %1 = call <64 x i8> @llvm.x86.avx512.packsswb.512(<32 x i16> %a, <32 x i16> %b)
+  %2 = bitcast i64 %mask to <64 x i1>
+  %3 = select <64 x i1> %2, <64 x i8> %1, <64 x i8> %passThru
+  ret <64 x i8> %3
 }
 
 define <64 x i8> @test_mask_packs_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i64 %mask) {
@@ -857,13 +869,13 @@ define <64 x i8> @test_mask_packs_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i
 ;
 ; AVX512F-32-LABEL: test_mask_packs_epi16_rrkz_512:
 ; AVX512F-32:       # BB#0:
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
-; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
 ; AVX512F-32-NEXT:    vpacksswb %zmm1, %zmm0, %zmm0 {%k1} {z}
 ; AVX512F-32-NEXT:    retl
-  %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 %mask)
-  ret <64 x i8> %res
+  %1 = call <64 x i8> @llvm.x86.avx512.packsswb.512(<32 x i16> %a, <32 x i16> %b)
+  %2 = bitcast i64 %mask to <64 x i1>
+  %3 = select <64 x i1> %2, <64 x i8> %1, <64 x i8> zeroinitializer
+  ret <64 x i8> %3
 }
 
 define <64 x i8> @test_mask_packs_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
@@ -878,8 +890,8 @@ define <64 x i8> @test_mask_packs_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b
 ; AVX512F-32-NEXT:    vpacksswb (%eax), %zmm0, %zmm0
 ; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
-  %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 -1)
-  ret <64 x i8> %res
+  %1 = call <64 x i8> @llvm.x86.avx512.packsswb.512(<32 x i16> %a, <32 x i16> %b)
+  ret <64 x i8> %1
 }
 
 define <64 x i8> @test_mask_packs_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <64 x i8> %passThru, i64 %mask) {
@@ -893,15 +905,15 @@ define <64 x i8> @test_mask_packs_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_
 ; AVX512F-32-LABEL: test_mask_packs_epi16_rmk_512:
 ; AVX512F-32:       # BB#0:
 ; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
-; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
 ; AVX512F-32-NEXT:    vpacksswb (%eax), %zmm0, %zmm1 {%k1}
 ; AVX512F-32-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
-  %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask)
-  ret <64 x i8> %res
+  %1 = call <64 x i8> @llvm.x86.avx512.packsswb.512(<32 x i16> %a, <32 x i16> %b)
+  %2 = bitcast i64 %mask to <64 x i1>
+  %3 = select <64 x i1> %2, <64 x i8> %1, <64 x i8> %passThru
+  ret <64 x i8> %3
 }
 
 define <64 x i8> @test_mask_packs_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i64 %mask) {
@@ -914,17 +926,17 @@ define <64 x i8> @test_mask_packs_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr
 ; AVX512F-32-LABEL: test_mask_packs_epi16_rmkz_512:
 ; AVX512F-32:       # BB#0:
 ; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
-; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
 ; AVX512F-32-NEXT:    vpacksswb (%eax), %zmm0, %zmm0 {%k1} {z}
 ; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
-  %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 %mask)
-  ret <64 x i8> %res
+  %1 = call <64 x i8> @llvm.x86.avx512.packsswb.512(<32 x i16> %a, <32 x i16> %b)
+  %2 = bitcast i64 %mask to <64 x i1>
+  %3 = select <64 x i1> %2, <64 x i8> %1, <64 x i8> zeroinitializer
+  ret <64 x i8> %3
 }
 
-declare <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16>, <32 x i16>, <64 x i8>, i64)
+declare <64 x i8> @llvm.x86.avx512.packsswb.512(<32 x i16>, <32 x i16>)
 
 
 define <32 x i16> @test_mask_packus_epi32_rr_512(<16 x i32> %a, <16 x i32> %b) {
@@ -937,8 +949,8 @@ define <32 x i16> @test_mask_packus_epi32_rr_512(<16 x i32> %a, <16 x i32> %b) {
 ; AVX512F-32:       # BB#0:
 ; AVX512F-32-NEXT:    vpackusdw %zmm1, %zmm0, %zmm0
 ; AVX512F-32-NEXT:    retl
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32> %a, <16 x i32> %b)
+  ret <32 x i16> %1
 }
 
 define <32 x i16> @test_mask_packus_epi32_rrk_512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask) {
@@ -955,8 +967,10 @@ define <32 x i16> @test_mask_packus_epi32_rrk_512(<16 x i32> %a, <16 x i32> %b,
 ; AVX512F-32-NEXT:    vpackusdw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512F-32-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; AVX512F-32-NEXT:    retl
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32> %a, <16 x i32> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> %passThru
+  ret <32 x i16> %3
 }
 
 define <32 x i16> @test_mask_packus_epi32_rrkz_512(<16 x i32> %a, <16 x i32> %b, i32 %mask) {
@@ -971,8 +985,10 @@ define <32 x i16> @test_mask_packus_epi32_rrkz_512(<16 x i32> %a, <16 x i32> %b,
 ; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
 ; AVX512F-32-NEXT:    vpackusdw %zmm1, %zmm0, %zmm0 {%k1} {z}
 ; AVX512F-32-NEXT:    retl
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32> %a, <16 x i32> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> zeroinitializer
+  ret <32 x i16> %3
 }
 
 define <32 x i16> @test_mask_packus_epi32_rm_512(<16 x i32> %a, <16 x i32>* %ptr_b) {
@@ -987,8 +1003,8 @@ define <32 x i16> @test_mask_packus_epi32_rm_512(<16 x i32> %a, <16 x i32>* %ptr
 ; AVX512F-32-NEXT:    vpackusdw (%eax), %zmm0, %zmm0
 ; AVX512F-32-NEXT:    retl
   %b = load <16 x i32>, <16 x i32>* %ptr_b
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32> %a, <16 x i32> %b)
+  ret <32 x i16> %1
 }
 
 define <32 x i16> @test_mask_packus_epi32_rmk_512(<16 x i32> %a, <16 x i32>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
@@ -1007,8 +1023,10 @@ define <32 x i16> @test_mask_packus_epi32_rmk_512(<16 x i32> %a, <16 x i32>* %pt
 ; AVX512F-32-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; AVX512F-32-NEXT:    retl
   %b = load <16 x i32>, <16 x i32>* %ptr_b
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32> %a, <16 x i32> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> %passThru
+  ret <32 x i16> %3
 }
 
 define <32 x i16> @test_mask_packus_epi32_rmkz_512(<16 x i32> %a, <16 x i32>* %ptr_b, i32 %mask) {
@@ -1025,8 +1043,10 @@ define <32 x i16> @test_mask_packus_epi32_rmkz_512(<16 x i32> %a, <16 x i32>* %p
 ; AVX512F-32-NEXT:    vpackusdw (%eax), %zmm0, %zmm0 {%k1} {z}
 ; AVX512F-32-NEXT:    retl
   %b = load <16 x i32>, <16 x i32>* %ptr_b
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32> %a, <16 x i32> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> zeroinitializer
+  ret <32 x i16> %3
 }
 
 define <32 x i16> @test_mask_packus_epi32_rmb_512(<16 x i32> %a, i32* %ptr_b) {
@@ -1043,8 +1063,8 @@ define <32 x i16> @test_mask_packus_epi32_rmb_512(<16 x i32> %a, i32* %ptr_b) {
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
   %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32> %a, <16 x i32> %b)
+  ret <32 x i16> %1
 }
 
 define <32 x i16> @test_mask_packus_epi32_rmbk_512(<16 x i32> %a, i32* %ptr_b, <32 x i16> %passThru, i32 %mask) {
@@ -1065,8 +1085,10 @@ define <32 x i16> @test_mask_packus_epi32_rmbk_512(<16 x i32> %a, i32* %ptr_b, <
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
   %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32> %a, <16 x i32> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> %passThru
+  ret <32 x i16> %3
 }
 
 define <32 x i16> @test_mask_packus_epi32_rmbkz_512(<16 x i32> %a, i32* %ptr_b, i32 %mask) {
@@ -1085,11 +1107,13 @@ define <32 x i16> @test_mask_packus_epi32_rmbkz_512(<16 x i32> %a, i32* %ptr_b,
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
   %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
-  %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
-  ret <32 x i16> %res
+  %1 = call <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32> %a, <16 x i32> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> zeroinitializer
+  ret <32 x i16> %3
 }
 
-declare <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32>, <16 x i32>, <32 x i16>, i32)
+declare <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32>, <16 x i32>)
 
 define <64 x i8> @test_mask_packus_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: test_mask_packus_epi16_rr_512:
@@ -1101,8 +1125,8 @@ define <64 x i8> @test_mask_packus_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512F-32:       # BB#0:
 ; AVX512F-32-NEXT:    vpackuswb %zmm1, %zmm0, %zmm0
 ; AVX512F-32-NEXT:    retl
-  %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 -1)
-  ret <64 x i8> %res
+  %1 = call <64 x i8> @llvm.x86.avx512.packuswb.512(<32 x i16> %a, <32 x i16> %b)
+  ret <64 x i8> %1
 }
 
 define <64 x i8> @test_mask_packus_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask) {
@@ -1115,14 +1139,14 @@ define <64 x i8> @test_mask_packus_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <
 ;
 ; AVX512F-32-LABEL: test_mask_packus_epi16_rrk_512:
 ; AVX512F-32:       # BB#0:
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
-; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
 ; AVX512F-32-NEXT:    vpackuswb %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512F-32-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; AVX512F-32-NEXT:    retl
-  %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask)
-  ret <64 x i8> %res
+  %1 = call <64 x i8> @llvm.x86.avx512.packuswb.512(<32 x i16> %a, <32 x i16> %b)
+  %2 = bitcast i64 %mask to <64 x i1>
+  %3 = select <64 x i1> %2, <64 x i8> %1, <64 x i8> %passThru
+  ret <64 x i8> %3
 }
 
 define <64 x i8> @test_mask_packus_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i64 %mask) {
@@ -1134,13 +1158,13 @@ define <64 x i8> @test_mask_packus_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b,
 ;
 ; AVX512F-32-LABEL: test_mask_packus_epi16_rrkz_512:
 ; AVX512F-32:       # BB#0:
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
-; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
 ; AVX512F-32-NEXT:    vpackuswb %zmm1, %zmm0, %zmm0 {%k1} {z}
 ; AVX512F-32-NEXT:    retl
-  %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 %mask)
-  ret <64 x i8> %res
+  %1 = call <64 x i8> @llvm.x86.avx512.packuswb.512(<32 x i16> %a, <32 x i16> %b)
+  %2 = bitcast i64 %mask to <64 x i1>
+  %3 = select <64 x i1> %2, <64 x i8> %1, <64 x i8> zeroinitializer
+  ret <64 x i8> %3
 }
 
 define <64 x i8> @test_mask_packus_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
@@ -1155,8 +1179,8 @@ define <64 x i8> @test_mask_packus_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_
 ; AVX512F-32-NEXT:    vpackuswb (%eax), %zmm0, %zmm0
 ; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
-  %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 -1)
-  ret <64 x i8> %res
+  %1 = call <64 x i8> @llvm.x86.avx512.packuswb.512(<32 x i16> %a, <32 x i16> %b)
+  ret <64 x i8> %1
 }
 
 define <64 x i8> @test_mask_packus_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <64 x i8> %passThru, i64 %mask) {
@@ -1170,15 +1194,15 @@ define <64 x i8> @test_mask_packus_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr
 ; AVX512F-32-LABEL: test_mask_packus_epi16_rmk_512:
 ; AVX512F-32:       # BB#0:
 ; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
-; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
 ; AVX512F-32-NEXT:    vpackuswb (%eax), %zmm0, %zmm1 {%k1}
 ; AVX512F-32-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
-  %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask)
-  ret <64 x i8> %res
+  %1 = call <64 x i8> @llvm.x86.avx512.packuswb.512(<32 x i16> %a, <32 x i16> %b)
+  %2 = bitcast i64 %mask to <64 x i1>
+  %3 = select <64 x i1> %2, <64 x i8> %1, <64 x i8> %passThru
+  ret <64 x i8> %3
 }
 
 define <64 x i8> @test_mask_packus_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i64 %mask) {
@@ -1191,17 +1215,17 @@ define <64 x i8> @test_mask_packus_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %pt
 ; AVX512F-32-LABEL: test_mask_packus_epi16_rmkz_512:
 ; AVX512F-32:       # BB#0:
 ; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
-; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    kmovq {{[0-9]+}}(%esp), %k1
 ; AVX512F-32-NEXT:    vpackuswb (%eax), %zmm0, %zmm0 {%k1} {z}
 ; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
-  %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 %mask)
-  ret <64 x i8> %res
+  %1 = call <64 x i8> @llvm.x86.avx512.packuswb.512(<32 x i16> %a, <32 x i16> %b)
+  %2 = bitcast i64 %mask to <64 x i1>
+  %3 = select <64 x i1> %2, <64 x i8> %1, <64 x i8> zeroinitializer
+  ret <64 x i8> %3
 }
 
-declare <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16>, <32 x i16>, <64 x i8>, i64)
+declare <64 x i8> @llvm.x86.avx512.packuswb.512(<32 x i16>, <32 x i16>)
 
 define <32 x i16> @test_mask_adds_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: test_mask_adds_epi16_rr_512:
diff --git a/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll b/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll
index 0a8e1445be8..efc544fd3e9 100644
--- a/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll
+++ b/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll
@@ -1863,3 +1863,680 @@ define <4 x i64>@test_int_x86_avx512_mask_pmovsxd_q_256(<4 x i32> %x0, <4 x i64>
   ret <4 x i64> %res4
 }
 
+define <8 x i16> @test_mask_packs_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
+; CHECK-LABEL: test_mask_packs_epi32_rr_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packs_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask) {
+; CHECK-LABEL: test_mask_packs_epi32_rrk_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpackssdw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x6b,0xd1]
+; CHECK-NEXT:    vmovdqa %xmm2, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packs_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
+; CHECK-LABEL: test_mask_packs_epi32_rrkz_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x6b,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packs_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
+; CHECK-LABEL: test_mask_packs_epi32_rm_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackssdw (%rdi), %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <4 x i32>, <4 x i32>* %ptr_b
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packs_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
+; CHECK-LABEL: test_mask_packs_epi32_rmk_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackssdw (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x6b,0x0f]
+; CHECK-NEXT:    vmovdqa %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <4 x i32>, <4 x i32>* %ptr_b
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packs_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
+; CHECK-LABEL: test_mask_packs_epi32_rmkz_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackssdw (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x6b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <4 x i32>, <4 x i32>* %ptr_b
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packs_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
+; CHECK-LABEL: test_mask_packs_epi32_rmb_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackssdw (%rdi){1to4}, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x18,0x6b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packs_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <8 x i16> %passThru, i8 %mask) {
+; CHECK-LABEL: test_mask_packs_epi32_rmbk_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackssdw (%rdi){1to4}, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x19,0x6b,0x0f]
+; CHECK-NEXT:    vmovdqa %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packs_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
+; CHECK-LABEL: test_mask_packs_epi32_rmbkz_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackssdw (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x99,0x6b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
+  ret <8 x i16> %res
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32>, <4 x i32>, <8 x i16>, i8)
+
+define <16 x i16> @test_mask_packs_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
+; CHECK-LABEL: test_mask_packs_epi32_rr_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packs_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask) {
+; CHECK-LABEL: test_mask_packs_epi32_rrk_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpackssdw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x6b,0xd1]
+; CHECK-NEXT:    vmovdqa %ymm2, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packs_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i16 %mask) {
+; CHECK-LABEL: test_mask_packs_epi32_rrkz_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packs_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
+; CHECK-LABEL: test_mask_packs_epi32_rm_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackssdw (%rdi), %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <8 x i32>, <8 x i32>* %ptr_b
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packs_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
+; CHECK-LABEL: test_mask_packs_epi32_rmk_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackssdw (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x6b,0x0f]
+; CHECK-NEXT:    vmovdqa %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <8 x i32>, <8 x i32>* %ptr_b
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packs_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i16 %mask) {
+; CHECK-LABEL: test_mask_packs_epi32_rmkz_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackssdw (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <8 x i32>, <8 x i32>* %ptr_b
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packs_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
+; CHECK-LABEL: test_mask_packs_epi32_rmb_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackssdw (%rdi){1to8}, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x38,0x6b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packs_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <16 x i16> %passThru, i16 %mask) {
+; CHECK-LABEL: test_mask_packs_epi32_rmbk_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackssdw (%rdi){1to8}, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x39,0x6b,0x0f]
+; CHECK-NEXT:    vmovdqa %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packs_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i16 %mask) {
+; CHECK-LABEL: test_mask_packs_epi32_rmbkz_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackssdw (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xb9,0x6b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
+  ret <16 x i16> %res
+}
+
+declare <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32>, <8 x i32>, <16 x i16>, i16)
+
+define <16 x i8> @test_mask_packs_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
+; CHECK-LABEL: test_mask_packs_epi16_rr_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
+  ret <16 x i8> %res
+}
+
+define <16 x i8> @test_mask_packs_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask) {
+; CHECK-LABEL: test_mask_packs_epi16_rrk_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpacksswb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x63,0xd1]
+; CHECK-NEXT:    vmovdqa %xmm2, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
+  ret <16 x i8> %res
+}
+
+define <16 x i8> @test_mask_packs_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i16 %mask) {
+; CHECK-LABEL: test_mask_packs_epi16_rrkz_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x63,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
+  ret <16 x i8> %res
+}
+
+define <16 x i8> @test_mask_packs_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
+; CHECK-LABEL: test_mask_packs_epi16_rm_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpacksswb (%rdi), %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <8 x i16>, <8 x i16>* %ptr_b
+  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
+  ret <16 x i8> %res
+}
+
+define <16 x i8> @test_mask_packs_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
+; CHECK-LABEL: test_mask_packs_epi16_rmk_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpacksswb (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x63,0x0f]
+; CHECK-NEXT:    vmovdqa %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <8 x i16>, <8 x i16>* %ptr_b
+  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
+  ret <16 x i8> %res
+}
+
+define <16 x i8> @test_mask_packs_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i16 %mask) {
+; CHECK-LABEL: test_mask_packs_epi16_rmkz_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpacksswb (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x63,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <8 x i16>, <8 x i16>* %ptr_b
+  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
+  ret <16 x i8> %res
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16>, <8 x i16>, <16 x i8>, i16)
+
+define <32 x i8> @test_mask_packs_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
+; CHECK-LABEL: test_mask_packs_epi16_rr_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpacksswb %ymm1, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
+  ret <32 x i8> %res
+}
+
+define <32 x i8> @test_mask_packs_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask) {
+; CHECK-LABEL: test_mask_packs_epi16_rrk_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpacksswb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x63,0xd1]
+; CHECK-NEXT:    vmovdqa %ymm2, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
+  ret <32 x i8> %res
+}
+
+define <32 x i8> @test_mask_packs_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i32 %mask) {
+; CHECK-LABEL: test_mask_packs_epi16_rrkz_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpacksswb %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x63,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
+  ret <32 x i8> %res
+}
+
+define <32 x i8> @test_mask_packs_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
+; CHECK-LABEL: test_mask_packs_epi16_rm_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpacksswb (%rdi), %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <16 x i16>, <16 x i16>* %ptr_b
+  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
+  ret <32 x i8> %res
+}
+
+define <32 x i8> @test_mask_packs_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
+; CHECK-LABEL: test_mask_packs_epi16_rmk_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
+; CHECK-NEXT:    vpacksswb (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x63,0x0f]
+; CHECK-NEXT:    vmovdqa %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <16 x i16>, <16 x i16>* %ptr_b
+  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
+  ret <32 x i8> %res
+}
+
+define <32 x i8> @test_mask_packs_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i32 %mask) {
+; CHECK-LABEL: test_mask_packs_epi16_rmkz_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
+; CHECK-NEXT:    vpacksswb (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x63,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <16 x i16>, <16 x i16>* %ptr_b
+  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
+  ret <32 x i8> %res
+}
+
+declare <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16>, <16 x i16>, <32 x i8>, i32)
+
+
+define <8 x i16> @test_mask_packus_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
+; CHECK-LABEL: test_mask_packus_epi32_rr_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packus_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask) {
+; CHECK-LABEL: test_mask_packus_epi32_rrk_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpackusdw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x2b,0xd1]
+; CHECK-NEXT:    vmovdqa %xmm2, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packus_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
+; CHECK-LABEL: test_mask_packus_epi32_rrkz_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x2b,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packus_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
+; CHECK-LABEL: test_mask_packus_epi32_rm_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <4 x i32>, <4 x i32>* %ptr_b
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packus_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
+; CHECK-LABEL: test_mask_packus_epi32_rmk_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackusdw (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x2b,0x0f]
+; CHECK-NEXT:    vmovdqa %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <4 x i32>, <4 x i32>* %ptr_b
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packus_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
+; CHECK-LABEL: test_mask_packus_epi32_rmkz_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x2b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <4 x i32>, <4 x i32>* %ptr_b
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packus_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
+; CHECK-LABEL: test_mask_packus_epi32_rmb_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackusdw (%rdi){1to4}, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x18,0x2b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packus_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <8 x i16> %passThru, i8 %mask) {
+; CHECK-LABEL: test_mask_packus_epi32_rmbk_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackusdw (%rdi){1to4}, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x19,0x2b,0x0f]
+; CHECK-NEXT:    vmovdqa %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
+  ret <8 x i16> %res
+}
+
+define <8 x i16> @test_mask_packus_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
+; CHECK-LABEL: test_mask_packus_epi32_rmbkz_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackusdw (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x99,0x2b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
+  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
+  ret <8 x i16> %res
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32>, <4 x i32>, <8 x i16>, i8)
+
+define <16 x i16> @test_mask_packus_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
+; CHECK-LABEL: test_mask_packus_epi32_rr_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packus_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask) {
+; CHECK-LABEL: test_mask_packus_epi32_rrk_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpackusdw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x2b,0xd1]
+; CHECK-NEXT:    vmovdqa %ymm2, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packus_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i16 %mask) {
+; CHECK-LABEL: test_mask_packus_epi32_rrkz_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packus_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
+; CHECK-LABEL: test_mask_packus_epi32_rm_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackusdw (%rdi), %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <8 x i32>, <8 x i32>* %ptr_b
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packus_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
+; CHECK-LABEL: test_mask_packus_epi32_rmk_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackusdw (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x2b,0x0f]
+; CHECK-NEXT:    vmovdqa %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <8 x i32>, <8 x i32>* %ptr_b
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packus_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i16 %mask) {
+; CHECK-LABEL: test_mask_packus_epi32_rmkz_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackusdw (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <8 x i32>, <8 x i32>* %ptr_b
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packus_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
+; CHECK-LABEL: test_mask_packus_epi32_rmb_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackusdw (%rdi){1to8}, %ymm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x38,0x2b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packus_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <16 x i16> %passThru, i16 %mask) {
+; CHECK-LABEL: test_mask_packus_epi32_rmbk_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackusdw (%rdi){1to8}, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x39,0x2b,0x0f]
+; CHECK-NEXT:    vmovdqa %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
+  ret <16 x i16> %res
+}
+
+define <16 x i16> @test_mask_packus_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i16 %mask) {
+; CHECK-LABEL: test_mask_packus_epi32_rmbkz_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackusdw (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xb9,0x2b,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %q = load i32, i32* %ptr_b
+  %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
+  %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
+  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
+  ret <16 x i16> %res
+}
+
+declare <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32>, <8 x i32>, <16 x i16>, i16)
+
+define <16 x i8> @test_mask_packus_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
+; CHECK-LABEL: test_mask_packus_epi16_rr_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
+  ret <16 x i8> %res
+}
+
+define <16 x i8> @test_mask_packus_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask) {
+; CHECK-LABEL: test_mask_packus_epi16_rrk_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpackuswb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x67,0xd1]
+; CHECK-NEXT:    vmovdqa %xmm2, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
+  ret <16 x i8> %res
+}
+
+define <16 x i8> @test_mask_packus_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i16 %mask) {
+; CHECK-LABEL: test_mask_packus_epi16_rrkz_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x67,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
+  ret <16 x i8> %res
+}
+
+define <16 x i8> @test_mask_packus_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
+; CHECK-LABEL: test_mask_packus_epi16_rm_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackuswb (%rdi), %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <8 x i16>, <8 x i16>* %ptr_b
+  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
+  ret <16 x i8> %res
+}
+
+define <16 x i8> @test_mask_packus_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
+; CHECK-LABEL: test_mask_packus_epi16_rmk_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackuswb (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x67,0x0f]
+; CHECK-NEXT:    vmovdqa %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <8 x i16>, <8 x i16>* %ptr_b
+  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
+  ret <16 x i8> %res
+}
+
+define <16 x i8> @test_mask_packus_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i16 %mask) {
+; CHECK-LABEL: test_mask_packus_epi16_rmkz_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vpackuswb (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x67,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <8 x i16>, <8 x i16>* %ptr_b
+  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
+  ret <16 x i8> %res
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16>, <8 x i16>, <16 x i8>, i16)
+
+define <32 x i8> @test_mask_packus_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
+; CHECK-LABEL: test_mask_packus_epi16_rr_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackuswb %ymm1, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
+  ret <32 x i8> %res
+}
+
+define <32 x i8> @test_mask_packus_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask) {
+; CHECK-LABEL: test_mask_packus_epi16_rrk_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpackuswb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x67,0xd1]
+; CHECK-NEXT:    vmovdqa %ymm2, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
+  ret <32 x i8> %res
+}
+
+define <32 x i8> @test_mask_packus_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i32 %mask) {
+; CHECK-LABEL: test_mask_packus_epi16_rrkz_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpackuswb %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x67,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
+  ret <32 x i8> %res
+}
+
+define <32 x i8> @test_mask_packus_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
+; CHECK-LABEL: test_mask_packus_epi16_rm_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpackuswb (%rdi), %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <16 x i16>, <16 x i16>* %ptr_b
+  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
+  ret <32 x i8> %res
+}
+
+define <32 x i8> @test_mask_packus_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
+; CHECK-LABEL: test_mask_packus_epi16_rmk_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
+; CHECK-NEXT:    vpackuswb (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x67,0x0f]
+; CHECK-NEXT:    vmovdqa %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <16 x i16>, <16 x i16>* %ptr_b
+  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
+  ret <32 x i8> %res
+}
+
+define <32 x i8> @test_mask_packus_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i32 %mask) {
+; CHECK-LABEL: test_mask_packus_epi16_rmkz_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
+; CHECK-NEXT:    vpackuswb (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x67,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %b = load <16 x i16>, <16 x i16>* %ptr_b
+  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
+  ret <32 x i8> %res
+}
+
+declare <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16>, <16 x i16>, <32 x i8>, i32)
+
diff --git a/test/CodeGen/X86/avx512bwvl-intrinsics.ll b/test/CodeGen/X86/avx512bwvl-intrinsics.ll
index 79d8e53c514..7f0385ee38c 100644
--- a/test/CodeGen/X86/avx512bwvl-intrinsics.ll
+++ b/test/CodeGen/X86/avx512bwvl-intrinsics.ll
@@ -882,8 +882,8 @@ define <8 x i16> @test_mask_packs_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
+  ret <8 x i16> %1
 }
 
 define <8 x i16> @test_mask_packs_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask) {
@@ -893,8 +893,10 @@ define <8 x i16> @test_mask_packs_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8 x
 ; CHECK-NEXT:    vpackssdw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x6b,0xd1]
 ; CHECK-NEXT:    vmovdqa %xmm2, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
+  %2 = bitcast i8 %mask to <8 x i1>
+  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
+  ret <8 x i16> %3
 }
 
 define <8 x i16> @test_mask_packs_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
@@ -903,8 +905,10 @@ define <8 x i16> @test_mask_packs_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8
 ; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
 ; CHECK-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x6b,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
+  %2 = bitcast i8 %mask to <8 x i1>
+  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
+  ret <8 x i16> %3
 }
 
 define <8 x i16> @test_mask_packs_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
@@ -913,8 +917,8 @@ define <8 x i16> @test_mask_packs_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b)
 ; CHECK-NEXT:    vpackssdw (%rdi), %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <4 x i32>, <4 x i32>* %ptr_b
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
+  ret <8 x i16> %1
 }
 
 define <8 x i16> @test_mask_packs_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
@@ -925,8 +929,10 @@ define <8 x i16> @test_mask_packs_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b,
 ; CHECK-NEXT:    vmovdqa %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <4 x i32>, <4 x i32>* %ptr_b
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
+  %2 = bitcast i8 %mask to <8 x i1>
+  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
+  ret <8 x i16> %3
 }
 
 define <8 x i16> @test_mask_packs_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
@@ -936,8 +942,10 @@ define <8 x i16> @test_mask_packs_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b
 ; CHECK-NEXT:    vpackssdw (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x6b,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <4 x i32>, <4 x i32>* %ptr_b
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
+  %2 = bitcast i8 %mask to <8 x i1>
+  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
+  ret <8 x i16> %3
 }
 
 define <8 x i16> @test_mask_packs_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
@@ -948,8 +956,8 @@ define <8 x i16> @test_mask_packs_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
   %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
+  ret <8 x i16> %1
 }
 
 define <8 x i16> @test_mask_packs_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <8 x i16> %passThru, i8 %mask) {
@@ -962,8 +970,10 @@ define <8 x i16> @test_mask_packs_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <8 x
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
   %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
+  %2 = bitcast i8 %mask to <8 x i1>
+  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
+  ret <8 x i16> %3
 }
 
 define <8 x i16> @test_mask_packs_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
@@ -975,19 +985,21 @@ define <8 x i16> @test_mask_packs_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
   %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
+  %2 = bitcast i8 %mask to <8 x i1>
+  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
+  ret <8 x i16> %3
 }
 
-declare <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32>, <4 x i32>, <8 x i16>, i8)
+declare <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32>, <4 x i32>)
 
 define <16 x i16> @test_mask_packs_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
 ; CHECK-LABEL: test_mask_packs_epi32_rr_256:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
+  ret <16 x i16> %1
 }
 
 define <16 x i16> @test_mask_packs_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask) {
@@ -997,8 +1009,10 @@ define <16 x i16> @test_mask_packs_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <16
 ; CHECK-NEXT:    vpackssdw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x6b,0xd1]
 ; CHECK-NEXT:    vmovdqa %ymm2, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
+  ret <16 x i16> %3
 }
 
 define <16 x i16> @test_mask_packs_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i16 %mask) {
@@ -1007,8 +1021,10 @@ define <16 x i16> @test_mask_packs_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i1
 ; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
 ; CHECK-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
+  ret <16 x i16> %3
 }
 
 define <16 x i16> @test_mask_packs_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
@@ -1017,8 +1033,8 @@ define <16 x i16> @test_mask_packs_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b)
 ; CHECK-NEXT:    vpackssdw (%rdi), %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <8 x i32>, <8 x i32>* %ptr_b
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
+  ret <16 x i16> %1
 }
 
 define <16 x i16> @test_mask_packs_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
@@ -1029,8 +1045,10 @@ define <16 x i16> @test_mask_packs_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b
 ; CHECK-NEXT:    vmovdqa %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <8 x i32>, <8 x i32>* %ptr_b
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
+  ret <16 x i16> %3
 }
 
 define <16 x i16> @test_mask_packs_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i16 %mask) {
@@ -1040,8 +1058,10 @@ define <16 x i16> @test_mask_packs_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_
 ; CHECK-NEXT:    vpackssdw (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <8 x i32>, <8 x i32>* %ptr_b
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
+  ret <16 x i16> %3
 }
 
 define <16 x i16> @test_mask_packs_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
@@ -1052,8 +1072,8 @@ define <16 x i16> @test_mask_packs_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
   %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
+  ret <16 x i16> %1
 }
 
 define <16 x i16> @test_mask_packs_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <16 x i16> %passThru, i16 %mask) {
@@ -1066,8 +1086,10 @@ define <16 x i16> @test_mask_packs_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <16
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
   %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
+  ret <16 x i16> %3
 }
 
 define <16 x i16> @test_mask_packs_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i16 %mask) {
@@ -1079,19 +1101,21 @@ define <16 x i16> @test_mask_packs_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i1
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
   %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
+  ret <16 x i16> %3
 }
 
-declare <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32>, <8 x i32>, <16 x i16>, i16)
+declare <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32>, <8 x i32>)
 
 define <16 x i8> @test_mask_packs_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
 ; CHECK-LABEL: test_mask_packs_epi16_rr_128:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
-  ret <16 x i8> %res
+  %1 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a, <8 x i16> %b)
+  ret <16 x i8> %1
 }
 
 define <16 x i8> @test_mask_packs_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask) {
@@ -1101,8 +1125,10 @@ define <16 x i8> @test_mask_packs_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16
 ; CHECK-NEXT:    vpacksswb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x63,0xd1]
 ; CHECK-NEXT:    vmovdqa %xmm2, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
-  ret <16 x i8> %res
+  %1 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a, <8 x i16> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
+  ret <16 x i8> %3
 }
 
 define <16 x i8> @test_mask_packs_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i16 %mask) {
@@ -1111,8 +1137,10 @@ define <16 x i8> @test_mask_packs_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i16
 ; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
 ; CHECK-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x63,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
-  ret <16 x i8> %res
+  %1 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a, <8 x i16> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
+  ret <16 x i8> %3
 }
 
 define <16 x i8> @test_mask_packs_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
@@ -1121,8 +1149,8 @@ define <16 x i8> @test_mask_packs_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b)
 ; CHECK-NEXT:    vpacksswb (%rdi), %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <8 x i16>, <8 x i16>* %ptr_b
-  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
-  ret <16 x i8> %res
+  %1 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a, <8 x i16> %b)
+  ret <16 x i8> %1
 }
 
 define <16 x i8> @test_mask_packs_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
@@ -1133,8 +1161,10 @@ define <16 x i8> @test_mask_packs_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b,
 ; CHECK-NEXT:    vmovdqa %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <8 x i16>, <8 x i16>* %ptr_b
-  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
-  ret <16 x i8> %res
+  %1 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a, <8 x i16> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
+  ret <16 x i8> %3
 }
 
 define <16 x i8> @test_mask_packs_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i16 %mask) {
@@ -1144,19 +1174,21 @@ define <16 x i8> @test_mask_packs_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b
 ; CHECK-NEXT:    vpacksswb (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x63,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <8 x i16>, <8 x i16>* %ptr_b
-  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
-  ret <16 x i8> %res
+  %1 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a, <8 x i16> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
+  ret <16 x i8> %3
 }
 
-declare <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16>, <8 x i16>, <16 x i8>, i16)
+declare <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16>, <8 x i16>)
 
 define <32 x i8> @test_mask_packs_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
 ; CHECK-LABEL: test_mask_packs_epi16_rr_256:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpacksswb %ymm1, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
-  ret <32 x i8> %res
+  %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a, <16 x i16> %b)
+  ret <32 x i8> %1
 }
 
 define <32 x i8> @test_mask_packs_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask) {
@@ -1166,8 +1198,10 @@ define <32 x i8> @test_mask_packs_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <3
 ; CHECK-NEXT:    vpacksswb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x63,0xd1]
 ; CHECK-NEXT:    vmovdqa %ymm2, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
-  ret <32 x i8> %res
+  %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a, <16 x i16> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
+  ret <32 x i8> %3
 }
 
 define <32 x i8> @test_mask_packs_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i32 %mask) {
@@ -1176,8 +1210,10 @@ define <32 x i8> @test_mask_packs_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i
 ; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
 ; CHECK-NEXT:    vpacksswb %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x63,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
-  ret <32 x i8> %res
+  %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a, <16 x i16> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
+  ret <32 x i8> %3
 }
 
 define <32 x i8> @test_mask_packs_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
@@ -1186,8 +1222,8 @@ define <32 x i8> @test_mask_packs_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b
 ; CHECK-NEXT:    vpacksswb (%rdi), %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <16 x i16>, <16 x i16>* %ptr_b
-  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
-  ret <32 x i8> %res
+  %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a, <16 x i16> %b)
+  ret <32 x i8> %1
 }
 
 define <32 x i8> @test_mask_packs_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
@@ -1198,8 +1234,10 @@ define <32 x i8> @test_mask_packs_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_
 ; CHECK-NEXT:    vmovdqa %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <16 x i16>, <16 x i16>* %ptr_b
-  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
-  ret <32 x i8> %res
+  %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a, <16 x i16> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
+  ret <32 x i8> %3
 }
 
 define <32 x i8> @test_mask_packs_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i32 %mask) {
@@ -1209,11 +1247,13 @@ define <32 x i8> @test_mask_packs_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr
 ; CHECK-NEXT:    vpacksswb (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x63,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <16 x i16>, <16 x i16>* %ptr_b
-  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
-  ret <32 x i8> %res
+  %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a, <16 x i16> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
+  ret <32 x i8> %3
 }
 
-declare <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16>, <16 x i16>, <32 x i8>, i32)
+declare <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16>, <16 x i16>)
 
 
 define <8 x i16> @test_mask_packus_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
@@ -1221,8 +1261,8 @@ define <8 x i16> @test_mask_packus_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
+  ret <8 x i16> %1
 }
 
 define <8 x i16> @test_mask_packus_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask) {
@@ -1232,8 +1272,10 @@ define <8 x i16> @test_mask_packus_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8
 ; CHECK-NEXT:    vpackusdw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x2b,0xd1]
 ; CHECK-NEXT:    vmovdqa %xmm2, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
+  %2 = bitcast i8 %mask to <8 x i1>
+  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
+  ret <8 x i16> %3
 }
 
 define <8 x i16> @test_mask_packus_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
@@ -1242,8 +1284,10 @@ define <8 x i16> @test_mask_packus_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8
 ; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
 ; CHECK-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x2b,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
+  %2 = bitcast i8 %mask to <8 x i1>
+  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
+  ret <8 x i16> %3
 }
 
 define <8 x i16> @test_mask_packus_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
@@ -1252,8 +1296,8 @@ define <8 x i16> @test_mask_packus_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b)
 ; CHECK-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <4 x i32>, <4 x i32>* %ptr_b
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
+  ret <8 x i16> %1
 }
 
 define <8 x i16> @test_mask_packus_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
@@ -1264,8 +1308,10 @@ define <8 x i16> @test_mask_packus_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b
 ; CHECK-NEXT:    vmovdqa %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <4 x i32>, <4 x i32>* %ptr_b
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
+  %2 = bitcast i8 %mask to <8 x i1>
+  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
+  ret <8 x i16> %3
 }
 
 define <8 x i16> @test_mask_packus_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
@@ -1275,8 +1321,10 @@ define <8 x i16> @test_mask_packus_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_
 ; CHECK-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x2b,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <4 x i32>, <4 x i32>* %ptr_b
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
+  %2 = bitcast i8 %mask to <8 x i1>
+  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
+  ret <8 x i16> %3
 }
 
 define <8 x i16> @test_mask_packus_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
@@ -1287,8 +1335,8 @@ define <8 x i16> @test_mask_packus_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
   %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
+  ret <8 x i16> %1
 }
 
 define <8 x i16> @test_mask_packus_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <8 x i16> %passThru, i8 %mask) {
@@ -1301,8 +1349,10 @@ define <8 x i16> @test_mask_packus_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <8
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
   %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
+  %2 = bitcast i8 %mask to <8 x i1>
+  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
+  ret <8 x i16> %3
 }
 
 define <8 x i16> @test_mask_packus_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
@@ -1314,19 +1364,21 @@ define <8 x i16> @test_mask_packus_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
   %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
-  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
-  ret <8 x i16> %res
+  %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
+  %2 = bitcast i8 %mask to <8 x i1>
+  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
+  ret <8 x i16> %3
 }
 
-declare <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32>, <4 x i32>, <8 x i16>, i8)
+declare <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32>, <4 x i32>)
 
 define <16 x i16> @test_mask_packus_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
 ; CHECK-LABEL: test_mask_packus_epi32_rr_256:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
+  ret <16 x i16> %1
 }
 
 define <16 x i16> @test_mask_packus_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask) {
@@ -1336,8 +1388,10 @@ define <16 x i16> @test_mask_packus_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <1
 ; CHECK-NEXT:    vpackusdw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x2b,0xd1]
 ; CHECK-NEXT:    vmovdqa %ymm2, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
+  ret <16 x i16> %3
 }
 
 define <16 x i16> @test_mask_packus_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i16 %mask) {
@@ -1346,8 +1400,10 @@ define <16 x i16> @test_mask_packus_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i
 ; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
 ; CHECK-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
+  ret <16 x i16> %3
 }
 
 define <16 x i16> @test_mask_packus_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
@@ -1356,8 +1412,8 @@ define <16 x i16> @test_mask_packus_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b
 ; CHECK-NEXT:    vpackusdw (%rdi), %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <8 x i32>, <8 x i32>* %ptr_b
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
+  ret <16 x i16> %1
 }
 
 define <16 x i16> @test_mask_packus_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
@@ -1368,8 +1424,10 @@ define <16 x i16> @test_mask_packus_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_
 ; CHECK-NEXT:    vmovdqa %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <8 x i32>, <8 x i32>* %ptr_b
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
+  ret <16 x i16> %3
 }
 
 define <16 x i16> @test_mask_packus_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i16 %mask) {
@@ -1379,8 +1437,10 @@ define <16 x i16> @test_mask_packus_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr
 ; CHECK-NEXT:    vpackusdw (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <8 x i32>, <8 x i32>* %ptr_b
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
+  ret <16 x i16> %3
 }
 
 define <16 x i16> @test_mask_packus_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
@@ -1391,8 +1451,8 @@ define <16 x i16> @test_mask_packus_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
   %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
+  ret <16 x i16> %1
 }
 
 define <16 x i16> @test_mask_packus_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <16 x i16> %passThru, i16 %mask) {
@@ -1405,8 +1465,10 @@ define <16 x i16> @test_mask_packus_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <1
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
   %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
+  ret <16 x i16> %3
 }
 
 define <16 x i16> @test_mask_packus_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i16 %mask) {
@@ -1418,19 +1480,21 @@ define <16 x i16> @test_mask_packus_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
   %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
-  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
-  ret <16 x i16> %res
+  %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
+  ret <16 x i16> %3
 }
 
-declare <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32>, <8 x i32>, <16 x i16>, i16)
+declare <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32>, <8 x i32>)
 
 define <16 x i8> @test_mask_packus_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
 ; CHECK-LABEL: test_mask_packus_epi16_rr_128:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
-  ret <16 x i8> %res
+  %1 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a, <8 x i16> %b)
+  ret <16 x i8> %1
 }
 
 define <16 x i8> @test_mask_packus_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask) {
@@ -1440,8 +1504,10 @@ define <16 x i8> @test_mask_packus_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16
 ; CHECK-NEXT:    vpackuswb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x67,0xd1]
 ; CHECK-NEXT:    vmovdqa %xmm2, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
-  ret <16 x i8> %res
+  %1 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a, <8 x i16> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
+  ret <16 x i8> %3
 }
 
 define <16 x i8> @test_mask_packus_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i16 %mask) {
@@ -1450,8 +1516,10 @@ define <16 x i8> @test_mask_packus_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i1
 ; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
 ; CHECK-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x67,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
-  ret <16 x i8> %res
+  %1 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a, <8 x i16> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
+  ret <16 x i8> %3
 }
 
 define <16 x i8> @test_mask_packus_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
@@ -1460,8 +1528,8 @@ define <16 x i8> @test_mask_packus_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b)
 ; CHECK-NEXT:    vpackuswb (%rdi), %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <8 x i16>, <8 x i16>* %ptr_b
-  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
-  ret <16 x i8> %res
+  %1 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a, <8 x i16> %b)
+  ret <16 x i8> %1
 }
 
 define <16 x i8> @test_mask_packus_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
@@ -1472,8 +1540,10 @@ define <16 x i8> @test_mask_packus_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b
 ; CHECK-NEXT:    vmovdqa %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <8 x i16>, <8 x i16>* %ptr_b
-  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
-  ret <16 x i8> %res
+  %1 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a, <8 x i16> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
+  ret <16 x i8> %3
 }
 
 define <16 x i8> @test_mask_packus_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i16 %mask) {
@@ -1483,19 +1553,21 @@ define <16 x i8> @test_mask_packus_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_
 ; CHECK-NEXT:    vpackuswb (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x67,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <8 x i16>, <8 x i16>* %ptr_b
-  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
-  ret <16 x i8> %res
+  %1 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a, <8 x i16> %b)
+  %2 = bitcast i16 %mask to <16 x i1>
+  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
+  ret <16 x i8> %3
 }
 
-declare <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16>, <8 x i16>, <16 x i8>, i16)
+declare <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16>, <8 x i16>)
 
 define <32 x i8> @test_mask_packus_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
 ; CHECK-LABEL: test_mask_packus_epi16_rr_256:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpackuswb %ymm1, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
-  ret <32 x i8> %res
+  %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b)
+  ret <32 x i8> %1
 }
 
 define <32 x i8> @test_mask_packus_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask) {
@@ -1505,8 +1577,10 @@ define <32 x i8> @test_mask_packus_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <
 ; CHECK-NEXT:    vpackuswb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x67,0xd1]
 ; CHECK-NEXT:    vmovdqa %ymm2, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
-  ret <32 x i8> %res
+  %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
+  ret <32 x i8> %3
 }
 
 define <32 x i8> @test_mask_packus_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i32 %mask) {
@@ -1515,8 +1589,10 @@ define <32 x i8> @test_mask_packus_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b,
 ; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
 ; CHECK-NEXT:    vpackuswb %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x67,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
-  ret <32 x i8> %res
+  %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
+  ret <32 x i8> %3
 }
 
 define <32 x i8> @test_mask_packus_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
@@ -1525,8 +1601,8 @@ define <32 x i8> @test_mask_packus_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_
 ; CHECK-NEXT:    vpackuswb (%rdi), %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <16 x i16>, <16 x i16>* %ptr_b
-  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
-  ret <32 x i8> %res
+  %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b)
+  ret <32 x i8> %1
 }
 
 define <32 x i8> @test_mask_packus_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
@@ -1537,8 +1613,10 @@ define <32 x i8> @test_mask_packus_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr
 ; CHECK-NEXT:    vmovdqa %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <16 x i16>, <16 x i16>* %ptr_b
-  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
-  ret <32 x i8> %res
+  %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
+  ret <32 x i8> %3
 }
 
 define <32 x i8> @test_mask_packus_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i32 %mask) {
@@ -1548,11 +1626,13 @@ define <32 x i8> @test_mask_packus_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %pt
 ; CHECK-NEXT:    vpackuswb (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x67,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <16 x i16>, <16 x i16>* %ptr_b
-  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
-  ret <32 x i8> %res
+  %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b)
+  %2 = bitcast i32 %mask to <32 x i1>
+  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
+  ret <32 x i8> %3
 }
 
-declare <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16>, <16 x i16>, <32 x i8>, i32)
+declare <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16>, <16 x i16>)
 
 define <8 x i16> @test_mask_adds_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
 ; CHECK-LABEL: test_mask_adds_epi16_rr_128: