AVX-512: Added "pandn" intrinsics set

by Asaf Badouh (asaf.badouh@intel.com)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235971 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/llvm/IR/IntrinsicsX86.td b/include/llvm/IR/IntrinsicsX86.td
index 51f760d05c47ad15474f3751d12f6deb5da5066f..86cf68ae7c976754a40de0b0a53b9e3a8606a452 100644 (file)
--- a/include/llvm/IR/IntrinsicsX86.td
+++ b/include/llvm/IR/IntrinsicsX86.td
@@ -3107,6 +3107,24 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_avx512_mask_pand_q_512 : GCCBuiltin<"__builtin_ia32_pandq512_mask">,
           Intrinsic<[llvm_v8i64_ty], [llvm_v8i64_ty, llvm_v8i64_ty,
                      llvm_v8i64_ty, llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_pandn_d_128 : GCCBuiltin<"__builtin_ia32_pandnd128_mask">,
+          Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v4i32_ty,
+                     llvm_v4i32_ty, llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_pandn_d_256 : GCCBuiltin<"__builtin_ia32_pandnd256_mask">,
+          Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty, llvm_v8i32_ty,
+                     llvm_v8i32_ty, llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_pandn_d_512 : GCCBuiltin<"__builtin_ia32_pandnd512_mask">,
+          Intrinsic<[llvm_v16i32_ty], [llvm_v16i32_ty, llvm_v16i32_ty,
+                     llvm_v16i32_ty, llvm_i16_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_pandn_q_128 : GCCBuiltin<"__builtin_ia32_pandnq128_mask">,
+          Intrinsic<[llvm_v2i64_ty], [llvm_v2i64_ty, llvm_v2i64_ty,
+                     llvm_v2i64_ty, llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_pandn_q_256 : GCCBuiltin<"__builtin_ia32_pandnq256_mask">,
+          Intrinsic<[llvm_v4i64_ty], [llvm_v4i64_ty, llvm_v4i64_ty,
+                     llvm_v4i64_ty, llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_pandn_q_512 : GCCBuiltin<"__builtin_ia32_pandnq512_mask">,
+          Intrinsic<[llvm_v8i64_ty], [llvm_v8i64_ty, llvm_v8i64_ty,
+                     llvm_v8i64_ty, llvm_i8_ty], [IntrNoMem]>;
   def int_x86_avx512_mask_por_d_128 : GCCBuiltin<"__builtin_ia32_pord128_mask">,
           Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v4i32_ty,
                      llvm_v4i32_ty, llvm_i8_ty], [IntrNoMem]>;

diff --git a/lib/Target/X86/X86IntrinsicsInfo.h b/lib/Target/X86/X86IntrinsicsInfo.h
index d0c67a8f8f0197a6a35d536857bf95c7958d7244..2bea4a107fd977e5c4a499a49159e5caeb1eda35 100644 (file)
--- a/lib/Target/X86/X86IntrinsicsInfo.h
+++ b/lib/Target/X86/X86IntrinsicsInfo.h
@@ -371,6 +371,12 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_mask_pand_q_128, INTR_TYPE_2OP_MASK, ISD::AND, 0),
   X86_INTRINSIC_DATA(avx512_mask_pand_q_256, INTR_TYPE_2OP_MASK, ISD::AND, 0),
   X86_INTRINSIC_DATA(avx512_mask_pand_q_512, INTR_TYPE_2OP_MASK, ISD::AND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_pandn_d_128, INTR_TYPE_2OP_MASK, X86ISD::ANDNP, 0),
+  X86_INTRINSIC_DATA(avx512_mask_pandn_d_256, INTR_TYPE_2OP_MASK, X86ISD::ANDNP, 0),
+  X86_INTRINSIC_DATA(avx512_mask_pandn_d_512, INTR_TYPE_2OP_MASK, X86ISD::ANDNP, 0),
+  X86_INTRINSIC_DATA(avx512_mask_pandn_q_128, INTR_TYPE_2OP_MASK, X86ISD::ANDNP, 0),
+  X86_INTRINSIC_DATA(avx512_mask_pandn_q_256, INTR_TYPE_2OP_MASK, X86ISD::ANDNP, 0),
+  X86_INTRINSIC_DATA(avx512_mask_pandn_q_512, INTR_TYPE_2OP_MASK, X86ISD::ANDNP, 0),
   X86_INTRINSIC_DATA(avx512_mask_pcmpeq_b_128,  CMP_MASK,  X86ISD::PCMPEQM, 0),
   X86_INTRINSIC_DATA(avx512_mask_pcmpeq_b_256,  CMP_MASK,  X86ISD::PCMPEQM, 0),
   X86_INTRINSIC_DATA(avx512_mask_pcmpeq_b_512,  CMP_MASK,  X86ISD::PCMPEQM, 0),

diff --git a/test/CodeGen/X86/avx512vl-intrinsics.ll b/test/CodeGen/X86/avx512vl-intrinsics.ll
index e61860df718a895a8337684e2efbbb342edfff2b..4db5df53848d3914fdba8d74bbc2108c6dbb35c2 100644 (file)
--- a/test/CodeGen/X86/avx512vl-intrinsics.ll
+++ b/test/CodeGen/X86/avx512vl-intrinsics.ll
@@ -1952,4 +1952,312 @@ define <8 x i32> @test_mask_xor_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i8 %m
   ret <8 x i32> %res
 }
 
-declare <8 x i32> @llvm.x86.avx512.mask.pxor.d.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)
\ No newline at end of file
+declare <8 x i32> @llvm.x86.avx512.mask.pxor.d.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)
+
+define <4 x i32> @test_mask_andnot_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rr_128
+  ;CHECK: vpandnd  %xmm1, %xmm0, %xmm0     ## encoding: [0x62,0xf1,0x7d,0x08,0xdf,0xc1]
+  %res = call <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
+  ret <4 x i32> %res
+}
+
+define <4 x i32> @test_mask_andnot_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rrk_128
+  ;CHECK: vpandnd  %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xdf,0xd1]
+  %res = call <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
+  ret <4 x i32> %res
+}
+
+define <4 x i32> @test_mask_andnot_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rrkz_128
+  ;CHECK: vpandnd  %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xdf,0xc1]
+  %res = call <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
+  ret <4 x i32> %res
+}
+
+define <4 x i32> @test_mask_andnot_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rm_128
+  ;CHECK: vpandnd  (%rdi), %xmm0, %xmm0    ## encoding: [0x62,0xf1,0x7d,0x08,0xdf,0x07]
+  %b = load <4 x i32>, <4 x i32>* %ptr_b
+  %res = call <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
+  ret <4 x i32> %res
+}
+
+define <4 x i32> @test_mask_andnot_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <4 x i32> %passThru, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rmk_128
+  ;CHECK: vpandnd  (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xdf,0x0f]
+  %b = load <4 x i32>, <4 x i32>* %ptr_b
+  %res = call <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
+  ret <4 x i32> %res
+}
+
+define <4 x i32> @test_mask_andnot_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rmkz_128
+  ;CHECK: vpandnd  (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xdf,0x07]
+  %b = load <4 x i32>, <4 x i32>* %ptr_b
+  %res = call <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
+  ret <4 x i32> %res
+}
+
+define <4 x i32> @test_mask_andnot_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rmb_128
+  ;CHECK: vpandnd  (%rdi){1to4}, %xmm0, %xmm0  ## encoding: [0x62,0xf1,0x7d,0x18,0xdf,0x07]
+  %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 <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
+  ret <4 x i32> %res
+}
+
+define <4 x i32> @test_mask_andnot_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <4 x i32> %passThru, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rmbk_128
+  ;CHECK: vpandnd  (%rdi){1to4}, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x19,0xdf,0x0f]
+  %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 <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
+  ret <4 x i32> %res
+}
+
+define <4 x i32> @test_mask_andnot_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rmbkz_128
+  ;CHECK: vpandnd  (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x99,0xdf,0x07]
+  %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 <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
+  ret <4 x i32> %res
+}
+
+declare <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32>, <4 x i32>, <4 x i32>, i8)
+
+define <8 x i32> @test_mask_andnot_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rr_256
+  ;CHECK: vpandnd  %ymm1, %ymm0, %ymm0     ## encoding: [0x62,0xf1,0x7d,0x28,0xdf,0xc1]
+  %res = call <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
+  ret <8 x i32> %res
+}
+
+define <8 x i32> @test_mask_andnot_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rrk_256
+  ;CHECK: vpandnd  %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xdf,0xd1]
+  %res = call <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
+  ret <8 x i32> %res
+}
+
+define <8 x i32> @test_mask_andnot_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rrkz_256
+  ;CHECK: vpandnd  %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xdf,0xc1]
+  %res = call <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
+  ret <8 x i32> %res
+}
+
+define <8 x i32> @test_mask_andnot_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rm_256
+  ;CHECK: vpandnd  (%rdi), %ymm0, %ymm0    ## encoding: [0x62,0xf1,0x7d,0x28,0xdf,0x07]
+  %b = load <8 x i32>, <8 x i32>* %ptr_b
+  %res = call <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
+  ret <8 x i32> %res
+}
+
+define <8 x i32> @test_mask_andnot_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <8 x i32> %passThru, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rmk_256
+  ;CHECK: vpandnd  (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xdf,0x0f]
+  %b = load <8 x i32>, <8 x i32>* %ptr_b
+  %res = call <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
+  ret <8 x i32> %res
+}
+
+define <8 x i32> @test_mask_andnot_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rmkz_256
+  ;CHECK: vpandnd  (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xdf,0x07]
+  %b = load <8 x i32>, <8 x i32>* %ptr_b
+  %res = call <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
+  ret <8 x i32> %res
+}
+
+define <8 x i32> @test_mask_andnot_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rmb_256
+  ;CHECK: vpandnd  (%rdi){1to8}, %ymm0, %ymm0  ## encoding: [0x62,0xf1,0x7d,0x38,0xdf,0x07]
+  %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 <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
+  ret <8 x i32> %res
+}
+
+define <8 x i32> @test_mask_andnot_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <8 x i32> %passThru, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rmbk_256
+  ;CHECK: vpandnd  (%rdi){1to8}, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x39,0xdf,0x0f]
+  %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 <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
+  ret <8 x i32> %res
+}
+
+define <8 x i32> @test_mask_andnot_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi32_rmbkz_256
+  ;CHECK: vpandnd  (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xb9,0xdf,0x07]
+  %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 <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
+  ret <8 x i32> %res
+}
+
+declare <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)
+
+define <2 x i64> @test_mask_andnot_epi64_rr_128(<2 x i64> %a, <2 x i64> %b) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rr_128
+  ;CHECK: vpandnq  %xmm1, %xmm0, %xmm0     ## encoding: [0x62,0xf1,0xfd,0x08,0xdf,0xc1]
+  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> zeroinitializer, i8 -1)
+  ret <2 x i64> %res
+}
+
+define <2 x i64> @test_mask_andnot_epi64_rrk_128(<2 x i64> %a, <2 x i64> %b, <2 x i64> %passThru, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rrk_128
+  ;CHECK: vpandnq  %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0xfd,0x09,0xdf,0xd1]
+  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> %passThru, i8 %mask)
+  ret <2 x i64> %res
+}
+
+define <2 x i64> @test_mask_andnot_epi64_rrkz_128(<2 x i64> %a, <2 x i64> %b, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rrkz_128
+  ;CHECK: vpandnq  %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0xfd,0x89,0xdf,0xc1]
+  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> zeroinitializer, i8 %mask)
+  ret <2 x i64> %res
+}
+
+define <2 x i64> @test_mask_andnot_epi64_rm_128(<2 x i64> %a, <2 x i64>* %ptr_b) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rm_128
+  ;CHECK: vpandnq  (%rdi), %xmm0, %xmm0    ## encoding: [0x62,0xf1,0xfd,0x08,0xdf,0x07]
+  %b = load <2 x i64>, <2 x i64>* %ptr_b
+  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> zeroinitializer, i8 -1)
+  ret <2 x i64> %res
+}
+
+define <2 x i64> @test_mask_andnot_epi64_rmk_128(<2 x i64> %a, <2 x i64>* %ptr_b, <2 x i64> %passThru, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rmk_128
+  ;CHECK: vpandnq  (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0xfd,0x09,0xdf,0x0f]
+  %b = load <2 x i64>, <2 x i64>* %ptr_b
+  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> %passThru, i8 %mask)
+  ret <2 x i64> %res
+}
+
+define <2 x i64> @test_mask_andnot_epi64_rmkz_128(<2 x i64> %a, <2 x i64>* %ptr_b, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rmkz_128
+  ;CHECK: vpandnq  (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0xfd,0x89,0xdf,0x07]
+  %b = load <2 x i64>, <2 x i64>* %ptr_b
+  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> zeroinitializer, i8 %mask)
+  ret <2 x i64> %res
+}
+
+define <2 x i64> @test_mask_andnot_epi64_rmb_128(<2 x i64> %a, i64* %ptr_b) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rmb_128
+  ;CHECK: vpandnq  (%rdi){1to2}, %xmm0, %xmm0  ## encoding: [0x62,0xf1,0xfd,0x18,0xdf,0x07]
+  %q = load i64, i64* %ptr_b
+  %vecinit.i = insertelement <2 x i64> undef, i64 %q, i32 0
+  %b = shufflevector <2 x i64> %vecinit.i, <2 x i64> undef, <2 x i32> zeroinitializer
+  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> zeroinitializer, i8 -1)
+  ret <2 x i64> %res
+}
+
+define <2 x i64> @test_mask_andnot_epi64_rmbk_128(<2 x i64> %a, i64* %ptr_b, <2 x i64> %passThru, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rmbk_128
+  ;CHECK: vpandnq  (%rdi){1to2}, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0xfd,0x19,0xdf,0x0f]
+  %q = load i64, i64* %ptr_b
+  %vecinit.i = insertelement <2 x i64> undef, i64 %q, i32 0
+  %b = shufflevector <2 x i64> %vecinit.i, <2 x i64> undef, <2 x i32> zeroinitializer
+  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> %passThru, i8 %mask)
+  ret <2 x i64> %res
+}
+
+define <2 x i64> @test_mask_andnot_epi64_rmbkz_128(<2 x i64> %a, i64* %ptr_b, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rmbkz_128
+  ;CHECK: vpandnq  (%rdi){1to2}, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0xfd,0x99,0xdf,0x07]
+  %q = load i64, i64* %ptr_b
+  %vecinit.i = insertelement <2 x i64> undef, i64 %q, i32 0
+  %b = shufflevector <2 x i64> %vecinit.i, <2 x i64> undef, <2 x i32> zeroinitializer
+  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> zeroinitializer, i8 %mask)
+  ret <2 x i64> %res
+}
+
+declare <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64>, <2 x i64>, <2 x i64>, i8)
+
+define <4 x i64> @test_mask_andnot_epi64_rr_256(<4 x i64> %a, <4 x i64> %b) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rr_256
+  ;CHECK: vpandnq  %ymm1, %ymm0, %ymm0     ## encoding: [0x62,0xf1,0xfd,0x28,0xdf,0xc1]
+  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> zeroinitializer, i8 -1)
+  ret <4 x i64> %res
+}
+
+define <4 x i64> @test_mask_andnot_epi64_rrk_256(<4 x i64> %a, <4 x i64> %b, <4 x i64> %passThru, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rrk_256
+  ;CHECK: vpandnq  %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0xfd,0x29,0xdf,0xd1]
+  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> %passThru, i8 %mask)
+  ret <4 x i64> %res
+}
+
+define <4 x i64> @test_mask_andnot_epi64_rrkz_256(<4 x i64> %a, <4 x i64> %b, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rrkz_256
+  ;CHECK: vpandnq  %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0xfd,0xa9,0xdf,0xc1]
+  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> zeroinitializer, i8 %mask)
+  ret <4 x i64> %res
+}
+
+define <4 x i64> @test_mask_andnot_epi64_rm_256(<4 x i64> %a, <4 x i64>* %ptr_b) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rm_256
+  ;CHECK: vpandnq  (%rdi), %ymm0, %ymm0    ## encoding: [0x62,0xf1,0xfd,0x28,0xdf,0x07]
+  %b = load <4 x i64>, <4 x i64>* %ptr_b
+  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> zeroinitializer, i8 -1)
+  ret <4 x i64> %res
+}
+
+define <4 x i64> @test_mask_andnot_epi64_rmk_256(<4 x i64> %a, <4 x i64>* %ptr_b, <4 x i64> %passThru, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rmk_256
+  ;CHECK: vpandnq  (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0xfd,0x29,0xdf,0x0f]
+  %b = load <4 x i64>, <4 x i64>* %ptr_b
+  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> %passThru, i8 %mask)
+  ret <4 x i64> %res
+}
+
+define <4 x i64> @test_mask_andnot_epi64_rmkz_256(<4 x i64> %a, <4 x i64>* %ptr_b, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rmkz_256
+  ;CHECK: vpandnq  (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0xfd,0xa9,0xdf,0x07]
+  %b = load <4 x i64>, <4 x i64>* %ptr_b
+  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> zeroinitializer, i8 %mask)
+  ret <4 x i64> %res
+}
+
+define <4 x i64> @test_mask_andnot_epi64_rmb_256(<4 x i64> %a, i64* %ptr_b) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rmb_256
+  ;CHECK: vpandnq  (%rdi){1to4}, %ymm0, %ymm0  ## encoding: [0x62,0xf1,0xfd,0x38,0xdf,0x07]
+  %q = load i64, i64* %ptr_b
+  %vecinit.i = insertelement <4 x i64> undef, i64 %q, i32 0
+  %b = shufflevector <4 x i64> %vecinit.i, <4 x i64> undef, <4 x i32> zeroinitializer
+  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> zeroinitializer, i8 -1)
+  ret <4 x i64> %res
+}
+
+define <4 x i64> @test_mask_andnot_epi64_rmbk_256(<4 x i64> %a, i64* %ptr_b, <4 x i64> %passThru, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rmbk_256
+  ;CHECK: vpandnq  (%rdi){1to4}, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0xfd,0x39,0xdf,0x0f]
+  %q = load i64, i64* %ptr_b
+  %vecinit.i = insertelement <4 x i64> undef, i64 %q, i32 0
+  %b = shufflevector <4 x i64> %vecinit.i, <4 x i64> undef, <4 x i32> zeroinitializer
+  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> %passThru, i8 %mask)
+  ret <4 x i64> %res
+}
+
+define <4 x i64> @test_mask_andnot_epi64_rmbkz_256(<4 x i64> %a, i64* %ptr_b, i8 %mask) {
+  ;CHECK-LABEL: test_mask_andnot_epi64_rmbkz_256
+  ;CHECK: vpandnq  (%rdi){1to4}, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0xfd,0xb9,0xdf,0x07]
+  %q = load i64, i64* %ptr_b
+  %vecinit.i = insertelement <4 x i64> undef, i64 %q, i32 0
+  %b = shufflevector <4 x i64> %vecinit.i, <4 x i64> undef, <4 x i32> zeroinitializer
+  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> zeroinitializer, i8 %mask)
+  ret <4 x i64> %res
+}
+
+declare <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64>, <4 x i64>, <4 x i64>, i8)