From b4173a5a7094907818ebc54f30abc45c9c3b0d25 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@gmail.com>
Date: Sun, 13 Nov 2016 07:26:19 +0000
Subject: [PATCH] [InstCombine][AVX-512] Teach InstCombineCalls to handle the
 new unmasked AVX-512 variable shift intrinsics.

llvm-svn: 286755
---
 .../Transforms/InstCombine/InstCombineCalls.cpp    |  22 +-
 .../Transforms/InstCombine/x86-vector-shifts.ll    | 326 +++++++++++++++++++++
 2 files changed, 344 insertions(+), 4 deletions(-)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index d109df3..9de103a 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -258,8 +258,7 @@ static Value *simplifyX86immShift(const IntrinsicInst &II,
   bool ShiftLeft = false;
 
   switch (II.getIntrinsicID()) {
-  default:
-    return nullptr;
+  default: llvm_unreachable("Unexpected intrinsic!");
   case Intrinsic::x86_sse2_psra_d:
   case Intrinsic::x86_sse2_psra_w:
   case Intrinsic::x86_sse2_psrai_d:
@@ -393,10 +392,13 @@ static Value *simplifyX86varShift(const IntrinsicInst &II,
   bool ShiftLeft = false;
 
   switch (II.getIntrinsicID()) {
-  default:
-    return nullptr;
+  default: llvm_unreachable("Unexpected intrinsic!");
   case Intrinsic::x86_avx2_psrav_d:
   case Intrinsic::x86_avx2_psrav_d_256:
+  case Intrinsic::x86_avx512_psrav_q_128:
+  case Intrinsic::x86_avx512_psrav_q_256:
+  case Intrinsic::x86_avx512_psrav_d_512:
+  case Intrinsic::x86_avx512_psrav_q_512:
     LogicalShift = false;
     ShiftLeft = false;
     break;
@@ -404,6 +406,8 @@ static Value *simplifyX86varShift(const IntrinsicInst &II,
   case Intrinsic::x86_avx2_psrlv_d_256:
   case Intrinsic::x86_avx2_psrlv_q:
   case Intrinsic::x86_avx2_psrlv_q_256:
+  case Intrinsic::x86_avx512_psrlv_d_512:
+  case Intrinsic::x86_avx512_psrlv_q_512:
     LogicalShift = true;
     ShiftLeft = false;
     break;
@@ -411,6 +415,8 @@ static Value *simplifyX86varShift(const IntrinsicInst &II,
   case Intrinsic::x86_avx2_psllv_d_256:
   case Intrinsic::x86_avx2_psllv_q:
   case Intrinsic::x86_avx2_psllv_q_256:
+  case Intrinsic::x86_avx512_psllv_d_512:
+  case Intrinsic::x86_avx512_psllv_q_512:
     LogicalShift = true;
     ShiftLeft = true;
     break;
@@ -1873,12 +1879,20 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
   case Intrinsic::x86_avx2_psllv_d_256:
   case Intrinsic::x86_avx2_psllv_q:
   case Intrinsic::x86_avx2_psllv_q_256:
+  case Intrinsic::x86_avx512_psllv_d_512:
+  case Intrinsic::x86_avx512_psllv_q_512:
   case Intrinsic::x86_avx2_psrav_d:
   case Intrinsic::x86_avx2_psrav_d_256:
+  case Intrinsic::x86_avx512_psrav_q_128:
+  case Intrinsic::x86_avx512_psrav_q_256:
+  case Intrinsic::x86_avx512_psrav_d_512:
+  case Intrinsic::x86_avx512_psrav_q_512:
   case Intrinsic::x86_avx2_psrlv_d:
   case Intrinsic::x86_avx2_psrlv_d_256:
   case Intrinsic::x86_avx2_psrlv_q:
   case Intrinsic::x86_avx2_psrlv_q_256:
+  case Intrinsic::x86_avx512_psrlv_d_512:
+  case Intrinsic::x86_avx512_psrlv_q_512:
     if (Value *V = simplifyX86varShift(*II, *Builder))
       return replaceInstUsesWith(*II, V);
     break;
diff --git a/llvm/test/Transforms/InstCombine/x86-vector-shifts.ll b/llvm/test/Transforms/InstCombine/x86-vector-shifts.ll
index 51bf6ed..eac7fd7 100644
--- a/llvm/test/Transforms/InstCombine/x86-vector-shifts.ll
+++ b/llvm/test/Transforms/InstCombine/x86-vector-shifts.ll
@@ -1564,6 +1564,14 @@ define <8 x i32> @avx2_psrav_d_256_0(<8 x i32> %v) {
   ret <8 x i32> %1
 }
 
+define <16 x i32> @avx512_psrav_d_512_0(<16 x i32> %v) {
+; CHECK-LABEL: @avx512_psrav_d_512_0(
+; CHECK-NEXT:    ret <16 x i32> %v
+;
+  %1 = tail call <16 x i32> @llvm.x86.avx512.psrav.d.512(<16 x i32> %v, <16 x i32> zeroinitializer)
+  ret <16 x i32> %1
+}
+
 define <4 x i32> @avx2_psrav_d_128_var(<4 x i32> %v) {
 ; CHECK-LABEL: @avx2_psrav_d_128_var(
 ; CHECK-NEXT:    [[TMP1:%.*]] = ashr <4 x i32> %v, <i32 0, i32 8, i32 16, i32 31>
@@ -1582,6 +1590,15 @@ define <8 x i32> @avx2_psrav_d_256_var(<8 x i32> %v) {
   ret <8 x i32> %1
 }
 
+define <16 x i32> @avx512_psrav_d_512_var(<16 x i32> %v) {
+; CHECK-LABEL: @avx512_psrav_d_512_var(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <16 x i32> %v, <i32 0, i32 8, i32 16, i32 24, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 24, i32 31, i32 24, i32 8, i32 0>
+; CHECK-NEXT:    ret <16 x i32> [[TMP1]]
+;
+  %1 = tail call <16 x i32> @llvm.x86.avx512.psrav.d.512(<16 x i32> %v, <16 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 24, i32 32, i32 24, i32 8, i32 0>)
+  ret <16 x i32> %1
+}
+
 define <4 x i32> @avx2_psrav_d_128_allbig(<4 x i32> %v) {
 ; CHECK-LABEL: @avx2_psrav_d_128_allbig(
 ; CHECK-NEXT:    [[TMP1:%.*]] = ashr <4 x i32> %v, <i32 31, i32 31, i32 31, i32 undef>
@@ -1600,6 +1617,15 @@ define <8 x i32> @avx2_psrav_d_256_allbig(<8 x i32> %v) {
   ret <8 x i32> %1
 }
 
+define <16 x i32> @avx512_psrav_d_512_allbig(<16 x i32> %v) {
+; CHECK-LABEL: @avx512_psrav_d_512_allbig(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <16 x i32> %v, <i32 undef, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 undef, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31>
+; CHECK-NEXT:    ret <16 x i32> [[TMP1]]
+;
+  %1 = tail call <16 x i32> @llvm.x86.avx512.psrav.d.512(<16 x i32> %v, <16 x i32> <i32 undef, i32 100, i32 255, i32 55555, i32 -32, i32 -100, i32 -255, i32 -55555, i32 undef, i32 100, i32 255, i32 55555, i32 -32, i32 -100, i32 -255, i32 -55555>)
+  ret <16 x i32> %1
+}
+
 define <4 x i32> @avx2_psrav_d_128_undef(<4 x i32> %v) {
 ; CHECK-LABEL: @avx2_psrav_d_128_undef(
 ; CHECK-NEXT:    [[TMP1:%.*]] = ashr <4 x i32> %v, <i32 undef, i32 8, i32 16, i32 31>
@@ -1620,6 +1646,122 @@ define <8 x i32> @avx2_psrav_d_256_undef(<8 x i32> %v) {
   ret <8 x i32> %2
 }
 
+define <16 x i32> @avx512_psrav_d_512_undef(<16 x i32> %v) {
+; CHECK-LABEL: @avx512_psrav_d_512_undef(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <16 x i32> %v, <i32 0, i32 undef, i32 16, i32 24, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 24, i32 31, i32 24, i32 8, i32 0>
+; CHECK-NEXT:    ret <16 x i32> [[TMP1]]
+;
+  %1 = insertelement <16 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 24, i32 32, i32 24, i32 8, i32 0>, i32 undef, i32 1
+  %2 = tail call <16 x i32> @llvm.x86.avx512.psrav.d.512(<16 x i32> %v, <16 x i32> %1)
+  ret <16 x i32> %2
+}
+
+define <2 x i64> @avx512_psrav_q_128_0(<2 x i64> %v) {
+; CHECK-LABEL: @avx512_psrav_q_128_0(
+; CHECK-NEXT:    ret <2 x i64> %v
+;
+  %1 = tail call <2 x i64> @llvm.x86.avx512.psrav.q.128(<2 x i64> %v, <2 x i64> zeroinitializer)
+  ret <2 x i64> %1
+}
+
+define <4 x i64> @avx512_psrav_q_256_0(<4 x i64> %v) {
+; CHECK-LABEL: @avx512_psrav_q_256_0(
+; CHECK-NEXT:    ret <4 x i64> %v
+;
+  %1 = tail call <4 x i64> @llvm.x86.avx512.psrav.q.256(<4 x i64> %v, <4 x i64> zeroinitializer)
+  ret <4 x i64> %1
+}
+
+define <2 x i64> @avx512_psrav_q_128_var(<2 x i64> %v) {
+; CHECK-LABEL: @avx512_psrav_q_128_var(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <2 x i64> %v, <i64 0, i64 8>
+  %1 = tail call <2 x i64> @llvm.x86.avx512.psrav.q.128(<2 x i64> %v, <2 x i64> <i64 0, i64 8>)
+  ret <2 x i64> %1
+}
+
+define <4 x i64> @avx512_psrav_q_256_var(<4 x i64> %v) {
+; CHECK-LABEL: @avx512_psrav_q_256_var(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <4 x i64> %v, <i64 0, i64 8, i64 16, i64 31>
+; CHECK-NEXT:    ret <4 x i64> [[TMP1]]
+;
+  %1 = tail call <4 x i64> @llvm.x86.avx512.psrav.q.256(<4 x i64> %v, <4 x i64> <i64 0, i64 8, i64 16, i64 31>)
+  ret <4 x i64> %1
+}
+
+define <2 x i64> @avx512_psrav_q_128_allbig(<2 x i64> %v) {
+; CHECK-LABEL: @avx512_psrav_q_128_allbig(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <2 x i64> %v, <i64 63, i64 undef>
+; CHECK-NEXT:    ret <2 x i64> [[TMP1]]
+;
+  %1 = tail call <2 x i64> @llvm.x86.avx512.psrav.q.128(<2 x i64> %v, <2 x i64> <i64 64, i64 undef>)
+  ret <2 x i64> %1
+}
+
+define <4 x i64> @avx512_psrav_q_256_allbig(<4 x i64> %v) {
+; CHECK-LABEL: @avx512_psrav_q_256_allbig(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <4 x i64> %v, <i64 63, i64 undef, i64 63, i64 63>
+; CHECK-NEXT:    ret <4 x i64> [[TMP1]]
+;
+  %1 = tail call <4 x i64> @llvm.x86.avx512.psrav.q.256(<4 x i64> %v, <4 x i64> <i64 64, i64 undef, i64 -128, i64 -60>)
+  ret <4 x i64> %1
+}
+
+define <2 x i64> @avx512_psrav_q_128_undef(<2 x i64> %v) {
+; CHECK-LABEL: @avx512_psrav_q_128_undef(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <2 x i64> %v, <i64 undef, i64 8>
+; CHECK-NEXT:    ret <2 x i64> [[TMP1]]
+;
+  %1 = insertelement <2 x i64> <i64 0, i64 8>, i64 undef, i64 0
+  %2 = tail call <2 x i64> @llvm.x86.avx512.psrav.q.128(<2 x i64> %v, <2 x i64> %1)
+  ret <2 x i64> %2
+}
+
+define <4 x i64> @avx512_psrav_q_256_undef(<4 x i64> %v) {
+; CHECK-LABEL: @avx512_psrav_q_256_undef(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <4 x i64> %v, <i64 undef, i64 8, i64 16, i64 31>
+; CHECK-NEXT:    ret <4 x i64> [[TMP1]]
+;
+  %1 = insertelement <4 x i64> <i64 0, i64 8, i64 16, i64 31>, i64 undef, i64 0
+  %2 = tail call <4 x i64> @llvm.x86.avx512.psrav.q.256(<4 x i64> %v, <4 x i64> %1)
+  ret <4 x i64> %2
+}
+
+define <8 x i64> @avx512_psrav_q_512_0(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psrav_q_512_0(
+; CHECK-NEXT:    ret <8 x i64> %v
+;
+  %1 = tail call <8 x i64> @llvm.x86.avx512.psrav.q.512(<8 x i64> %v, <8 x i64> zeroinitializer)
+  ret <8 x i64> %1
+}
+
+define <8 x i64> @avx512_psrav_q_512_var(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psrav_q_512_var(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <8 x i64> %v, <i64 0, i64 8, i64 16, i64 31, i64 0, i64 8, i64 16, i64 31>
+; CHECK-NEXT:    ret <8 x i64> [[TMP1]]
+;
+  %1 = tail call <8 x i64> @llvm.x86.avx512.psrav.q.512(<8 x i64> %v, <8 x i64> <i64 0, i64 8, i64 16, i64 31, i64 0, i64 8, i64 16, i64 31>)
+  ret <8 x i64> %1
+}
+
+define <8 x i64> @avx512_psrav_q_512_allbig(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psrav_q_512_allbig(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <8 x i64> %v, <i64 63, i64 undef, i64 63, i64 63, i64 63, i64 undef, i64 63, i64 63>
+; CHECK-NEXT:    ret <8 x i64> [[TMP1]]
+;
+  %1 = tail call <8 x i64> @llvm.x86.avx512.psrav.q.512(<8 x i64> %v, <8 x i64> <i64 64, i64 undef, i64 -128, i64 -60, i64 64, i64 undef, i64 -128, i64 -60>)
+  ret <8 x i64> %1
+}
+
+define <8 x i64> @avx512_psrav_q_512_undef(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psrav_q_512_undef(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <8 x i64> %v, <i64 undef, i64 8, i64 16, i64 31, i64 0, i64 8, i64 16, i64 31>
+; CHECK-NEXT:    ret <8 x i64> [[TMP1]]
+;
+  %1 = insertelement <8 x i64> <i64 0, i64 8, i64 16, i64 31, i64 0, i64 8, i64 16, i64 31>, i64 undef, i64 0
+  %2 = tail call <8 x i64> @llvm.x86.avx512.psrav.q.512(<8 x i64> %v, <8 x i64> %1)
+  ret <8 x i64> %2
+}
+
 ;
 ; LSHR - Constant Per-Element Vector
 ;
@@ -1800,6 +1942,94 @@ define <4 x i64> @avx2_psrlv_q_256_undef(<4 x i64> %v) {
   ret <4 x i64> %2
 }
 
+define <16 x i32> @avx2_psrlv_d_512_0(<16 x i32> %v) {
+; CHECK-LABEL: @avx2_psrlv_d_512_0(
+; CHECK-NEXT:    ret <16 x i32> %v
+;
+  %1 = tail call <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32> %v, <16 x i32> zeroinitializer)
+  ret <16 x i32> %1
+}
+
+define <16 x i32> @avx512_psrlv_d_512_var(<16 x i32> %v) {
+; CHECK-LABEL: @avx512_psrlv_d_512_var(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <16 x i32> %v, <i32 0, i32 8, i32 16, i32 24, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 24, i32 31, i32 24, i32 8, i32 0>
+; CHECK-NEXT:    ret <16 x i32> [[TMP1]]
+;
+  %1 = tail call <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32> %v, <16 x i32> <i32 0, i32 8, i32 16, i32 24, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 24, i32 31, i32 24, i32 8, i32 0>)
+  ret <16 x i32> %1
+}
+
+define <16 x i32> @avx512_psrlv_d_512_big(<16 x i32> %v) {
+; CHECK-LABEL: @avx512_psrlv_d_512_big(
+; CHECK-NEXT:    [[TMP1:%.*]] = tail call <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32> %v, <16 x i32> <i32 0, i32 8, i32 16, i32 64, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 64, i32 31, i32 24, i32 8, i32 0>)
+; CHECK-NEXT:    ret <16 x i32> [[TMP1]]
+;
+  %1 = tail call <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32> %v, <16 x i32> <i32 0, i32 8, i32 16, i32 64, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 64, i32 31, i32 24, i32 8, i32 0>)
+  ret <16 x i32> %1
+}
+
+define <16 x i32> @avx512_psrlv_d_512_allbig(<16 x i32> %v) {
+; CHECK-LABEL: @avx512_psrlv_d_512_allbig(
+; CHECK-NEXT:    ret <16 x i32> <i32 undef, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 undef, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
+;
+  %1 = tail call <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32> %v, <16 x i32> <i32 undef, i32 100, i32 255, i32 55555, i32 -32, i32 -100, i32 -255, i32 -55555, i32 undef, i32 100, i32 255, i32 55555, i32 -32, i32 -100, i32 -255, i32 -55555>)
+  ret <16 x i32> %1
+}
+
+define <16 x i32> @avx512_psrlv_d_512_undef(<16 x i32> %v) {
+; CHECK-LABEL: @avx512_psrlv_d_512_undef(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <16 x i32> %v, <i32 0, i32 undef, i32 16, i32 31, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 31, i32 31, i32 24, i32 8, i32 0>
+; CHECK-NEXT:    ret <16 x i32> [[TMP1]]
+;
+  %1 = insertelement <16 x i32> <i32 0, i32 8, i32 16, i32 31, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 31, i32 31, i32 24, i32 8, i32 0>, i32 undef, i32 1
+  %2 = tail call <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32> %v, <16 x i32> %1)
+  ret <16 x i32> %2
+}
+
+define <8 x i64> @avx512_psrlv_q_512_0(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psrlv_q_512_0(
+; CHECK-NEXT:    ret <8 x i64> %v
+;
+  %1 = tail call <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64> %v, <8 x i64> zeroinitializer)
+  ret <8 x i64> %1
+}
+
+define <8 x i64> @avx512_psrlv_q_512_var(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psrlv_q_512_var(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <8 x i64> %v, <i64 0, i64 8, i64 16, i64 31, i64 0, i64 8, i64 16, i64 31>
+; CHECK-NEXT:    ret <8 x i64> [[TMP1]]
+;
+  %1 = tail call <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64> %v, <8 x i64> <i64 0, i64 8, i64 16, i64 31, i64 0, i64 8, i64 16, i64 31>)
+  ret <8 x i64> %1
+}
+
+define <8 x i64> @avx512_psrlv_q_512_big(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psrlv_q_512_big(
+; CHECK-NEXT:    [[TMP1:%.*]] = tail call <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64> %v, <8 x i64> <i64 0, i64 8, i64 16, i64 64, i64 0, i64 8, i64 16, i64 64>)
+; CHECK-NEXT:    ret <8 x i64> [[TMP1]]
+;
+  %1 = tail call <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64> %v, <8 x i64> <i64 0, i64 8, i64 16, i64 64, i64 0, i64 8, i64 16, i64 64>)
+  ret <8 x i64> %1
+}
+
+define <8 x i64> @avx512_psrlv_q_512_allbig(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psrlv_q_512_allbig(
+; CHECK-NEXT:    ret <8 x i64> <i64 0, i64 undef, i64 0, i64 0, i64 0, i64 undef, i64 0, i64 0>
+;
+  %1 = tail call <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64> %v, <8 x i64> <i64 64, i64 undef, i64 -128, i64 -60, i64 64, i64 undef, i64 -128, i64 -60>)
+  ret <8 x i64> %1
+}
+
+define <8 x i64> @avx512_psrlv_q_512_undef(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psrlv_q_512_undef(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <8 x i64> %v, <i64 undef, i64 8, i64 16, i64 31, i64 0, i64 8, i64 16, i64 31>
+; CHECK-NEXT:    ret <8 x i64> [[TMP1]]
+;
+  %1 = insertelement <8 x i64> <i64 0, i64 8, i64 16, i64 31, i64 0, i64 8, i64 16, i64 31>, i64 undef, i64 0
+  %2 = tail call <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64> %v, <8 x i64> %1)
+  ret <8 x i64> %2
+}
+
 ;
 ; SHL - Constant Per-Element Vector
 ;
@@ -1980,6 +2210,94 @@ define <4 x i64> @avx2_psllv_q_256_undef(<4 x i64> %v) {
   ret <4 x i64> %2
 }
 
+define <16 x i32> @avx512_psllv_d_512_0(<16 x i32> %v) {
+; CHECK-LABEL: @avx512_psllv_d_512_0(
+; CHECK-NEXT:    ret <16 x i32> %v
+;
+  %1 = tail call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> %v, <16 x i32> zeroinitializer)
+  ret <16 x i32> %1
+}
+
+define <16 x i32> @avx512_psllv_d_512_var(<16 x i32> %v) {
+; CHECK-LABEL: @avx512_psllv_d_512_var(
+; CHECK-NEXT:    [[TMP1:%.*]] = shl <16 x i32> %v, <i32 0, i32 8, i32 16, i32 24, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 24, i32 31, i32 24, i32 8, i32 0>
+; CHECK-NEXT:    ret <16 x i32> [[TMP1]]
+;
+  %1 = tail call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> %v, <16 x i32> <i32 0, i32 8, i32 16, i32 24, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 24, i32 31, i32 24, i32 8, i32 0>)
+  ret <16 x i32> %1
+}
+
+define <16 x i32> @avx512_psllv_d_512_big(<16 x i32> %v) {
+; CHECK-LABEL: @avx512_psllv_d_512_big(
+; CHECK-NEXT:    [[TMP1:%.*]] = tail call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> %v, <16 x i32> <i32 0, i32 8, i32 16, i32 64, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 64, i32 31, i32 24, i32 8, i32 0>)
+; CHECK-NEXT:    ret <16 x i32> [[TMP1]]
+;
+  %1 = tail call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> %v, <16 x i32> <i32 0, i32 8, i32 16, i32 64, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 64, i32 31, i32 24, i32 8, i32 0>)
+  ret <16 x i32> %1
+}
+
+define <16 x i32> @avx512_psllv_d_512_allbig(<16 x i32> %v) {
+; CHECK-LABEL: @avx512_psllv_d_512_allbig(
+; CHECK-NEXT:    ret <16 x i32> <i32 undef, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 undef, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
+;
+  %1 = tail call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> %v, <16 x i32> <i32 undef, i32 100, i32 255, i32 55555, i32 -32, i32 -100, i32 -255, i32 -55555, i32 undef, i32 100, i32 255, i32 55555, i32 -32, i32 -100, i32 -255, i32 -55555>)
+  ret <16 x i32> %1
+}
+
+define <16 x i32> @avx512_psllv_d_512_undef(<16 x i32> %v) {
+; CHECK-LABEL: @avx512_psllv_d_512_undef(
+; CHECK-NEXT:    [[TMP1:%.*]] = shl <16 x i32> %v, <i32 0, i32 undef, i32 16, i32 31, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 31, i32 31, i32 24, i32 8, i32 0>
+; CHECK-NEXT:    ret <16 x i32> [[TMP1]]
+;
+  %1 = insertelement <16 x i32> <i32 0, i32 8, i32 16, i32 31, i32 31, i32 24, i32 8, i32 0, i32 0, i32 8, i32 16, i32 31, i32 31, i32 24, i32 8, i32 0>, i32 undef, i32 1
+  %2 = tail call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> %v, <16 x i32> %1)
+  ret <16 x i32> %2
+}
+
+define <8 x i64> @avx512_psllv_q_512_0(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psllv_q_512_0(
+; CHECK-NEXT:    ret <8 x i64> %v
+;
+  %1 = tail call <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64> %v, <8 x i64> zeroinitializer)
+  ret <8 x i64> %1
+}
+
+define <8 x i64> @avx512_psllv_q_512_var(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psllv_q_512_var(
+; CHECK-NEXT:    [[TMP1:%.*]] = shl <8 x i64> %v, <i64 0, i64 8, i64 16, i64 31, i64 0, i64 8, i64 16, i64 31>
+; CHECK-NEXT:    ret <8 x i64> [[TMP1]]
+;
+  %1 = tail call <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64> %v, <8 x i64> <i64 0, i64 8, i64 16, i64 31, i64 0, i64 8, i64 16, i64 31>)
+  ret <8 x i64> %1
+}
+
+define <8 x i64> @avx512_psllv_q_512_big(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psllv_q_512_big(
+; CHECK-NEXT:    [[TMP1:%.*]] = tail call <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64> %v, <8 x i64> <i64 0, i64 8, i64 16, i64 64, i64 0, i64 8, i64 16, i64 64>)
+; CHECK-NEXT:    ret <8 x i64> [[TMP1]]
+;
+  %1 = tail call <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64> %v, <8 x i64> <i64 0, i64 8, i64 16, i64 64, i64 0, i64 8, i64 16, i64 64>)
+  ret <8 x i64> %1
+}
+
+define <8 x i64> @avx512_psllv_q_512_allbig(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psllv_q_512_allbig(
+; CHECK-NEXT:    ret <8 x i64> <i64 0, i64 undef, i64 0, i64 0, i64 0, i64 undef, i64 0, i64 0>
+;
+  %1 = tail call <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64> %v, <8 x i64> <i64 64, i64 undef, i64 -128, i64 -60, i64 64, i64 undef, i64 -128, i64 -60>)
+  ret <8 x i64> %1
+}
+
+define <8 x i64> @avx512_psllv_q_512_undef(<8 x i64> %v) {
+; CHECK-LABEL: @avx512_psllv_q_512_undef(
+; CHECK-NEXT:    [[TMP1:%.*]] = shl <8 x i64> %v, <i64 undef, i64 8, i64 16, i64 31, i64 0, i64 8, i64 16, i64 31>
+; CHECK-NEXT:    ret <8 x i64> [[TMP1]]
+;
+  %1 = insertelement <8 x i64> <i64 0, i64 8, i64 16, i64 31, i64 0, i64 8, i64 16, i64 31>, i64 undef, i64 0
+  %2 = tail call <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64> %v, <8 x i64> %1)
+  ret <8 x i64> %2
+}
+
 ;
 ; Vector Demanded Bits
 ;
@@ -2710,15 +3028,23 @@ declare <8 x i16> @llvm.x86.sse2.psra.w(<8 x i16>, <8 x i16>) #1
 
 declare <4 x i32> @llvm.x86.avx2.psrav.d(<4 x i32>, <4 x i32>) #1
 declare <8 x i32> @llvm.x86.avx2.psrav.d.256(<8 x i32>, <8 x i32>) #1
+declare <16 x i32> @llvm.x86.avx512.psrav.d.512(<16 x i32>, <16 x i32>) #1
+declare <2 x i64> @llvm.x86.avx512.psrav.q.128(<2 x i64>, <2 x i64>) #1
+declare <4 x i64> @llvm.x86.avx512.psrav.q.256(<4 x i64>, <4 x i64>) #1
+declare <8 x i64> @llvm.x86.avx512.psrav.q.512(<8 x i64>, <8 x i64>) #1
 
 declare <4 x i32> @llvm.x86.avx2.psrlv.d(<4 x i32>, <4 x i32>) #1
 declare <8 x i32> @llvm.x86.avx2.psrlv.d.256(<8 x i32>, <8 x i32>) #1
 declare <2 x i64> @llvm.x86.avx2.psrlv.q(<2 x i64>, <2 x i64>) #1
 declare <4 x i64> @llvm.x86.avx2.psrlv.q.256(<4 x i64>, <4 x i64>) #1
+declare <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32>, <16 x i32>) #1
+declare <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64>, <8 x i64>) #1
 
 declare <4 x i32> @llvm.x86.avx2.psllv.d(<4 x i32>, <4 x i32>) #1
 declare <8 x i32> @llvm.x86.avx2.psllv.d.256(<8 x i32>, <8 x i32>) #1
 declare <2 x i64> @llvm.x86.avx2.psllv.q(<2 x i64>, <2 x i64>) #1
 declare <4 x i64> @llvm.x86.avx2.psllv.q.256(<4 x i64>, <4 x i64>) #1
+declare <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32>, <16 x i32>) #1
+declare <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64>, <8 x i64>) #1
 
 attributes #1 = { nounwind readnone }
-- 
2.7.4