[AVX512] Add support for 512b variable bit shift intrinsics.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@224028 91177308-0d34-0410-b5e6-96231b3b80d8

5 files changed, 215 insertions, 39 deletions

diff --git a/include/llvm/IR/IntrinsicsX86.td b/include/llvm/IR/IntrinsicsX86.td
index 9456cb642ec..40b2ba28f2e 100644
--- a/include/llvm/IR/IntrinsicsX86.td
+++ b/include/llvm/IR/IntrinsicsX86.td
@@ -1930,6 +1930,31 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_avx2_psrav_d_256 : GCCBuiltin<"__builtin_ia32_psrav8si">,
               Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty, llvm_v8i32_ty],
                         [IntrNoMem]>;
+
+  def int_x86_avx512_mask_psllv_d : GCCBuiltin<"__builtin_ia32_psllv16si_mask">,
+              Intrinsic<[llvm_v16i32_ty], [llvm_v16i32_ty,
+                         llvm_v16i32_ty, llvm_v16i32_ty, llvm_i16_ty],
+                        [IntrNoMem]>;
+  def int_x86_avx512_mask_psllv_q : GCCBuiltin<"__builtin_ia32_psllv8di_mask">,
+              Intrinsic<[llvm_v8i64_ty], [llvm_v8i64_ty,
+                         llvm_v8i64_ty, llvm_v8i64_ty, llvm_i8_ty], 
+                        [IntrNoMem]>;
+  def int_x86_avx512_mask_psrav_d : GCCBuiltin<"__builtin_ia32_psrav16si_mask">,
+              Intrinsic<[llvm_v16i32_ty], [llvm_v16i32_ty,
+                         llvm_v16i32_ty, llvm_v16i32_ty, llvm_i16_ty],
+                        [IntrNoMem]>;
+  def int_x86_avx512_mask_psrav_q : GCCBuiltin<"__builtin_ia32_psrav8di_mask">,
+              Intrinsic<[llvm_v8i64_ty], [llvm_v8i64_ty,
+                         llvm_v8i64_ty, llvm_v8i64_ty, llvm_i8_ty],
+                        [IntrNoMem]>;
+  def int_x86_avx512_mask_psrlv_d : GCCBuiltin<"__builtin_ia32_psrlv16si_mask">,
+              Intrinsic<[llvm_v16i32_ty], [llvm_v16i32_ty,
+                         llvm_v16i32_ty, llvm_v16i32_ty, llvm_i16_ty],
+                        [IntrNoMem]>;
+  def int_x86_avx512_mask_psrlv_q : GCCBuiltin<"__builtin_ia32_psrlv8di_mask">,
+              Intrinsic<[llvm_v8i64_ty], [llvm_v8i64_ty,
+                         llvm_v8i64_ty, llvm_v8i64_ty, llvm_i8_ty], 
+                        [IntrNoMem]>;
 }
 
 // Gather ops
diff --git a/lib/Target/X86/X86InstrAVX512.td b/lib/Target/X86/X86InstrAVX512.td
index b512305d5dd..26b71c32b97 100644
--- a/lib/Target/X86/X86InstrAVX512.td
+++ b/lib/Target/X86/X86InstrAVX512.td
@@ -3306,16 +3306,16 @@ multiclass avx512_shift_rrm<bits<8> opc, string OpcodeStr, SDNode OpNode,
                    " ",  SSE_INTSHIFT_ITINS_P.rm>, AVX512BIBase, EVEX_4V;
 }
 
-multiclass avx512_varshift_sizes<bits<8> opc, string OpcodeStr, SDNode OpNode,
+multiclass avx512_shift_sizes<bits<8> opc, string OpcodeStr, SDNode OpNode,
                                   ValueType SrcVT, PatFrag bc_frag, X86VectorVTInfo _> {
   defm Z : avx512_shift_rrm<opc, OpcodeStr, OpNode, SrcVT, bc_frag, _>, EVEX_V512;
 }
 
-multiclass avx512_varshift_types<bits<8> opcd, bits<8> opcq, string OpcodeStr,
+multiclass avx512_shift_types<bits<8> opcd, bits<8> opcq, string OpcodeStr,
                                  SDNode OpNode> {
-  defm D : avx512_varshift_sizes<opcd, OpcodeStr#"d", OpNode, v4i32, bc_v4i32,
+  defm D : avx512_shift_sizes<opcd, OpcodeStr#"d", OpNode, v4i32, bc_v4i32,
                                  v16i32_info>, EVEX_CD8<32, CD8VQ>;
-  defm Q : avx512_varshift_sizes<opcq, OpcodeStr#"q", OpNode, v2i64, bc_v2i64,
+  defm Q : avx512_shift_sizes<opcq, OpcodeStr#"q", OpNode, v2i64, bc_v2i64,
                                  v8i64_info>, EVEX_CD8<64, CD8VQ>, VEX_W;
 }
 
@@ -3340,48 +3340,43 @@ defm VPSRAQZ : avx512_shift_rmi<0x72, MRM4r, MRM4m, "vpsraq", X86vsrai,
                            v8i64_info>, EVEX_V512,
                            EVEX_CD8<64, CD8VF>, VEX_W;
 
-defm VPSRL : avx512_varshift_types<0xD2, 0xD3, "vpsrl", X86vsrl>;
-defm VPSLL : avx512_varshift_types<0xF2, 0xF3, "vpsll", X86vshl>;
-defm VPSRA : avx512_varshift_types<0xE2, 0xE2, "vpsra", X86vsra>;
+defm VPSLL : avx512_shift_types<0xF2, 0xF3, "vpsll", X86vshl>;
+defm VPSRA : avx512_shift_types<0xE2, 0xE2, "vpsra", X86vsra>;
+defm VPSRL : avx512_shift_types<0xD2, 0xD3, "vpsrl", X86vsrl>;
 
 //===-------------------------------------------------------------------===//
 // Variable Bit Shifts
 //===-------------------------------------------------------------------===//
 multiclass avx512_var_shift<bits<8> opc, string OpcodeStr, SDNode OpNode,
-                           RegisterClass RC, ValueType vt,
-                           X86MemOperand x86memop, PatFrag mem_frag> {
-  def rr  : AVX5128I<opc, MRMSrcReg, (outs RC:$dst),
-             (ins RC:$src1, RC:$src2),
-             !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-             [(set RC:$dst,
-               (vt (OpNode RC:$src1, (vt RC:$src2))))]>,
-             EVEX_4V;
-  def rm  : AVX5128I<opc, MRMSrcMem, (outs RC:$dst),
-             (ins RC:$src1, x86memop:$src2),
-             !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-             [(set RC:$dst,
-               (vt (OpNode RC:$src1, (mem_frag addr:$src2))))]>,
-             EVEX_4V;
+                            X86VectorVTInfo _> {
+  defm rr : AVX512_maskable<opc, MRMSrcReg, _, (outs _.RC:$dst),
+                   (ins _.RC:$src1, _.RC:$src2), OpcodeStr,
+                      "$src2, $src1", "$src1, $src2",
+                   (_.VT (OpNode _.RC:$src1, (_.VT _.RC:$src2))),
+                   " ",  SSE_INTSHIFT_ITINS_P.rr>, AVX5128IBase, EVEX_4V;
+  defm rm : AVX512_maskable<opc, MRMSrcMem, _, (outs _.RC:$dst),
+                   (ins _.RC:$src1, _.MemOp:$src2), OpcodeStr,
+                       "$src2, $src1", "$src1, $src2",
+                   (_.VT (OpNode _.RC:$src1, (_.MemOpFrag addr:$src2))),
+                   " ",  SSE_INTSHIFT_ITINS_P.rm>, AVX5128IBase, EVEX_4V;
+}
+
+multiclass avx512_var_shift_sizes<bits<8> opc, string OpcodeStr, SDNode OpNode,
+                                  AVX512VLVectorVTInfo _> {
+  defm Z : avx512_var_shift<opc, OpcodeStr, OpNode, _.info512>, EVEX_V512;
+}
+
+multiclass avx512_var_shift_types<bits<8> opc, string OpcodeStr,
+                                 SDNode OpNode> {
+  defm D : avx512_var_shift_sizes<opc, OpcodeStr#"d", OpNode,
+                                 avx512vl_i32_info>, EVEX_CD8<32, CD8VQ>;
+  defm Q : avx512_var_shift_sizes<opc, OpcodeStr#"q", OpNode,
+                                 avx512vl_i64_info>, EVEX_CD8<64, CD8VQ>, VEX_W;
 }
 
-defm VPSLLVDZ : avx512_var_shift<0x47, "vpsllvd", shl, VR512, v16i32,
-                               i512mem, memopv16i32>, EVEX_V512,
-                               EVEX_CD8<32, CD8VF>;
-defm VPSLLVQZ : avx512_var_shift<0x47, "vpsllvq", shl, VR512, v8i64,
-                               i512mem, memopv8i64>, EVEX_V512, VEX_W,
-                               EVEX_CD8<64, CD8VF>;
-defm VPSRLVDZ : avx512_var_shift<0x45, "vpsrlvd", srl, VR512, v16i32,
-                               i512mem, memopv16i32>, EVEX_V512,
-                               EVEX_CD8<32, CD8VF>;
-defm VPSRLVQZ : avx512_var_shift<0x45, "vpsrlvq", srl, VR512, v8i64,
-                               i512mem, memopv8i64>, EVEX_V512, VEX_W,
-                               EVEX_CD8<64, CD8VF>;
-defm VPSRAVDZ : avx512_var_shift<0x46, "vpsravd", sra, VR512, v16i32,
-                               i512mem, memopv16i32>, EVEX_V512,
-                               EVEX_CD8<32, CD8VF>;
-defm VPSRAVQZ : avx512_var_shift<0x46, "vpsravq", sra, VR512, v8i64,
-                               i512mem, memopv8i64>, EVEX_V512, VEX_W,
-                               EVEX_CD8<64, CD8VF>;
+defm VPSLLV : avx512_var_shift_types<0x47, "vpsllv", shl>;
+defm VPSRAV : avx512_var_shift_types<0x46, "vpsrav", sra>;
+defm VPSRLV : avx512_var_shift_types<0x45, "vpsrlv", srl>;
 
 //===----------------------------------------------------------------------===//
 // AVX-512 - MOVDDUP
diff --git a/lib/Target/X86/X86InstrFormats.td b/lib/Target/X86/X86InstrFormats.td
index d7de9219811..4746d0c4aad 100644
--- a/lib/Target/X86/X86InstrFormats.td
+++ b/lib/Target/X86/X86InstrFormats.td
@@ -699,6 +699,9 @@ class AVX5128I<bits<8> o, Format F, dag outs, dag ins, string asm,
             list<dag> pattern, InstrItinClass itin = NoItinerary>
       : I<o, F, outs, ins, asm, pattern, itin, SSEPackedInt>, T8PD,
         Requires<[HasAVX512]>;
+class AVX5128IBase : T8PD {
+  Domain ExeDomain = SSEPackedInt;
+}
 class AVX512XS8I<bits<8> o, Format F, dag outs, dag ins, string asm,
             list<dag> pattern, InstrItinClass itin = NoItinerary>
       : I<o, F, outs, ins, asm, pattern, itin, SSEPackedInt>, T8XS,
diff --git a/lib/Target/X86/X86IntrinsicsInfo.h b/lib/Target/X86/X86IntrinsicsInfo.h
index e7ad044d1e5..c1255224db2 100644
--- a/lib/Target/X86/X86IntrinsicsInfo.h
+++ b/lib/Target/X86/X86IntrinsicsInfo.h
@@ -285,14 +285,20 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_mask_psll_q,        INTR_TYPE_2OP_MASK, X86ISD::VSHL, 0),
   X86_INTRINSIC_DATA(avx512_mask_pslli_d,       VSHIFT_MASK, X86ISD::VSHLI, 0),
   X86_INTRINSIC_DATA(avx512_mask_pslli_q,       VSHIFT_MASK, X86ISD::VSHLI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_psllv_d,       INTR_TYPE_2OP_MASK, ISD::SHL, 0),
+  X86_INTRINSIC_DATA(avx512_mask_psllv_q,       INTR_TYPE_2OP_MASK, ISD::SHL, 0),
   X86_INTRINSIC_DATA(avx512_mask_psra_d,        INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0),
   X86_INTRINSIC_DATA(avx512_mask_psra_q,        INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0),
   X86_INTRINSIC_DATA(avx512_mask_psrai_d,       VSHIFT_MASK, X86ISD::VSRAI, 0),
   X86_INTRINSIC_DATA(avx512_mask_psrai_q,       VSHIFT_MASK, X86ISD::VSRAI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_psrav_d,       INTR_TYPE_2OP_MASK, ISD::SRA, 0),
+  X86_INTRINSIC_DATA(avx512_mask_psrav_q,       INTR_TYPE_2OP_MASK, ISD::SRA, 0),
   X86_INTRINSIC_DATA(avx512_mask_psrl_d,        INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0),
   X86_INTRINSIC_DATA(avx512_mask_psrl_q,        INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0),
   X86_INTRINSIC_DATA(avx512_mask_psrli_d,       VSHIFT_MASK, X86ISD::VSRLI, 0),
   X86_INTRINSIC_DATA(avx512_mask_psrli_q,       VSHIFT_MASK, X86ISD::VSRLI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_psrlv_d,       INTR_TYPE_2OP_MASK, ISD::SRL, 0),
+  X86_INTRINSIC_DATA(avx512_mask_psrlv_q,       INTR_TYPE_2OP_MASK, ISD::SRL, 0),
   X86_INTRINSIC_DATA(avx512_mask_ucmp_b_128,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
   X86_INTRINSIC_DATA(avx512_mask_ucmp_b_256,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
   X86_INTRINSIC_DATA(avx512_mask_ucmp_b_512,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
diff --git a/test/CodeGen/X86/avx512-intrinsics.ll b/test/CodeGen/X86/avx512-intrinsics.ll
index a4d479280f8..8e5165e0a06 100644
--- a/test/CodeGen/X86/avx512-intrinsics.ll
+++ b/test/CodeGen/X86/avx512-intrinsics.ll
@@ -1226,3 +1226,150 @@ define <8 x i64> @test_x86_avx512_maskz_psra_q(<8 x i64> %a0, <2 x i64> %a1, i8
 }
 
 declare <8 x i64> @llvm.x86.avx512.mask.psra.q(<8 x i64>, <2 x i64>, <8 x i64>, i8) nounwind readnone
+
+define <16 x i32> @test_x86_avx512_psllv_d(<16 x i32> %a0, <16 x i32> %a1) {
+  ; CHECK-LABEL: test_x86_avx512_psllv_d
+  ; CHECK: vpsllvd
+  %res = call <16 x i32> @llvm.x86.avx512.mask.psllv.d(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> zeroinitializer, i16 -1)
+  ret <16 x i32> %res
+}
+
+define <16 x i32> @test_x86_avx512_mask_psllv_d(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, i16 %mask) {
+  ; CHECK-LABEL: test_x86_avx512_mask_psllv_d
+  ; CHECK: vpsllvd %zmm1, %zmm0, %zmm2 {%k1}
+  %res = call <16 x i32> @llvm.x86.avx512.mask.psllv.d(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, i16 %mask)
+  ret <16 x i32> %res
+}
+
+define <16 x i32> @test_x86_avx512_maskz_psllv_d(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ; CHECK-LABEL: test_x86_avx512_maskz_psllv_d
+  ; CHECK: vpsllvd %zmm1, %zmm0, %zmm0 {%k1} {z}
+  %res = call <16 x i32> @llvm.x86.avx512.mask.psllv.d(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> zeroinitializer, i16 %mask)
+  ret <16 x i32> %res
+}
+
+declare <16 x i32> @llvm.x86.avx512.mask.psllv.d(<16 x i32>, <16 x i32>, <16 x i32>, i16) nounwind readnone
+
+define <8 x i64> @test_x86_avx512_psllv_q(<8 x i64> %a0, <8 x i64> %a1) {
+  ; CHECK-LABEL: test_x86_avx512_psllv_q
+  ; CHECK: vpsllvq
+  %res = call <8 x i64> @llvm.x86.avx512.mask.psllv.q(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> zeroinitializer, i8 -1)
+  ret <8 x i64> %res
+}
+
+define <8 x i64> @test_x86_avx512_mask_psllv_q(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> %a2, i8 %mask) {
+  ; CHECK-LABEL: test_x86_avx512_mask_psllv_q
+  ; CHECK: vpsllvq %zmm1, %zmm0, %zmm2 {%k1}
+  %res = call <8 x i64> @llvm.x86.avx512.mask.psllv.q(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> %a2, i8 %mask)
+  ret <8 x i64> %res
+}
+
+define <8 x i64> @test_x86_avx512_maskz_psllv_q(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ; CHECK-LABEL: test_x86_avx512_maskz_psllv_q
+  ; CHECK: vpsllvq %zmm1, %zmm0, %zmm0 {%k1} {z}
+  %res = call <8 x i64> @llvm.x86.avx512.mask.psllv.q(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> zeroinitializer, i8 %mask)
+  ret <8 x i64> %res
+}
+
+declare <8 x i64> @llvm.x86.avx512.mask.psllv.q(<8 x i64>, <8 x i64>, <8 x i64>, i8) nounwind readnone
+
+
+define <16 x i32> @test_x86_avx512_psrav_d(<16 x i32> %a0, <16 x i32> %a1) {
+  ; CHECK-LABEL: test_x86_avx512_psrav_d
+  ; CHECK: vpsravd
+  %res = call <16 x i32> @llvm.x86.avx512.mask.psrav.d(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> zeroinitializer, i16 -1)
+  ret <16 x i32> %res
+}
+
+define <16 x i32> @test_x86_avx512_mask_psrav_d(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, i16 %mask) {
+  ; CHECK-LABEL: test_x86_avx512_mask_psrav_d
+  ; CHECK: vpsravd %zmm1, %zmm0, %zmm2 {%k1}
+  %res = call <16 x i32> @llvm.x86.avx512.mask.psrav.d(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, i16 %mask)
+  ret <16 x i32> %res
+}
+
+define <16 x i32> @test_x86_avx512_maskz_psrav_d(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ; CHECK-LABEL: test_x86_avx512_maskz_psrav_d
+  ; CHECK: vpsravd %zmm1, %zmm0, %zmm0 {%k1} {z}
+  %res = call <16 x i32> @llvm.x86.avx512.mask.psrav.d(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> zeroinitializer, i16 %mask)
+  ret <16 x i32> %res
+}
+
+declare <16 x i32> @llvm.x86.avx512.mask.psrav.d(<16 x i32>, <16 x i32>, <16 x i32>, i16) nounwind readnone
+
+define <8 x i64> @test_x86_avx512_psrav_q(<8 x i64> %a0, <8 x i64> %a1) {
+  ; CHECK-LABEL: test_x86_avx512_psrav_q
+  ; CHECK: vpsravq
+  %res = call <8 x i64> @llvm.x86.avx512.mask.psrav.q(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> zeroinitializer, i8 -1)
+  ret <8 x i64> %res
+}
+
+define <8 x i64> @test_x86_avx512_mask_psrav_q(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> %a2, i8 %mask) {
+  ; CHECK-LABEL: test_x86_avx512_mask_psrav_q
+  ; CHECK: vpsravq %zmm1, %zmm0, %zmm2 {%k1}
+  %res = call <8 x i64> @llvm.x86.avx512.mask.psrav.q(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> %a2, i8 %mask)
+  ret <8 x i64> %res
+}
+
+define <8 x i64> @test_x86_avx512_maskz_psrav_q(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ; CHECK-LABEL: test_x86_avx512_maskz_psrav_q
+  ; CHECK: vpsravq %zmm1, %zmm0, %zmm0 {%k1} {z}
+  %res = call <8 x i64> @llvm.x86.avx512.mask.psrav.q(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> zeroinitializer, i8 %mask)
+  ret <8 x i64> %res
+}
+
+declare <8 x i64> @llvm.x86.avx512.mask.psrav.q(<8 x i64>, <8 x i64>, <8 x i64>, i8) nounwind readnone
+
+define <16 x i32> @test_x86_avx512_psrlv_d(<16 x i32> %a0, <16 x i32> %a1) {
+  ; CHECK-LABEL: test_x86_avx512_psrlv_d
+  ; CHECK: vpsrlvd
+  %res = call <16 x i32> @llvm.x86.avx512.mask.psrlv.d(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> zeroinitializer, i16 -1)
+  ret <16 x i32> %res
+}
+
+define <16 x i32> @test_x86_avx512_mask_psrlv_d(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, i16 %mask) {
+  ; CHECK-LABEL: test_x86_avx512_mask_psrlv_d
+  ; CHECK: vpsrlvd %zmm1, %zmm0, %zmm2 {%k1}
+  %res = call <16 x i32> @llvm.x86.avx512.mask.psrlv.d(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, i16 %mask)
+  ret <16 x i32> %res
+}
+
+define <16 x i32> @test_x86_avx512_maskz_psrlv_d(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ; CHECK-LABEL: test_x86_avx512_maskz_psrlv_d
+  ; CHECK: vpsrlvd %zmm1, %zmm0, %zmm0 {%k1} {z}
+  %res = call <16 x i32> @llvm.x86.avx512.mask.psrlv.d(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> zeroinitializer, i16 %mask)
+  ret <16 x i32> %res
+}
+
+declare <16 x i32> @llvm.x86.avx512.mask.psrlv.d(<16 x i32>, <16 x i32>, <16 x i32>, i16) nounwind readnone
+
+define <8 x i64> @test_x86_avx512_psrlv_q(<8 x i64> %a0, <8 x i64> %a1) {
+  ; CHECK-LABEL: test_x86_avx512_psrlv_q
+  ; CHECK: vpsrlvq
+  %res = call <8 x i64> @llvm.x86.avx512.mask.psrlv.q(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> zeroinitializer, i8 -1)
+  ret <8 x i64> %res
+}
+
+define <8 x i64> @test_x86_avx512_mask_psrlv_q(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> %a2, i8 %mask) {
+  ; CHECK-LABEL: test_x86_avx512_mask_psrlv_q
+  ; CHECK: vpsrlvq %zmm1, %zmm0, %zmm2 {%k1}
+  %res = call <8 x i64> @llvm.x86.avx512.mask.psrlv.q(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> %a2, i8 %mask)
+  ret <8 x i64> %res
+}
+
+define <8 x i64> @test_x86_avx512_maskz_psrlv_q(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ; CHECK-LABEL: test_x86_avx512_maskz_psrlv_q
+  ; CHECK: vpsrlvq %zmm1, %zmm0, %zmm0 {%k1} {z}
+  %res = call <8 x i64> @llvm.x86.avx512.mask.psrlv.q(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> zeroinitializer, i8 %mask)
+  ret <8 x i64> %res
+}
+
+declare <8 x i64> @llvm.x86.avx512.mask.psrlv.q(<8 x i64>, <8 x i64>, <8 x i64>, i8) nounwind readnone
+
+define <8 x i64> @test_x86_avx512_psrlv_q_memop(<8 x i64> %a0, <8 x i64>* %ptr) {
+  ; CHECK-LABEL: test_x86_avx512_psrlv_q_memop
+  ; CHECK: vpsrlvq (%
+  %b = load <8 x i64>* %ptr
+  %res = call <8 x i64> @llvm.x86.avx512.mask.psrlv.q(<8 x i64> %a0, <8 x i64> %b, <8 x i64> zeroinitializer, i8 -1)
+  ret <8 x i64> %res
+}