From 28f0327565edb3e9c189b1b2225f489dc42ddb8f Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@gmail.com>
Date: Mon, 7 Nov 2011 08:26:24 +0000
Subject: [PATCH] Add AVX2 variable shift instructions and intrinsics.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@143915 91177308-0d34-0410-b5e6-96231b3b80d8
---
 include/llvm/IntrinsicsX86.td           | 36 +++++++++++++++
 lib/Target/X86/X86InstrSSE.td           | 40 +++++++++++++++++
 test/CodeGen/X86/avx2-intrinsics-x86.ll | 80 +++++++++++++++++++++++++++++++++
 3 files changed, 156 insertions(+)

diff --git a/include/llvm/IntrinsicsX86.td b/include/llvm/IntrinsicsX86.td
index 42209b8291d..cba6599d660 100644
--- a/include/llvm/IntrinsicsX86.td
+++ b/include/llvm/IntrinsicsX86.td
@@ -1771,6 +1771,42 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
         Intrinsic<[], [llvm_ptr_ty, llvm_v4i64_ty, llvm_v4i64_ty], []>;
 }
 
+// Variable bit shift ops
+let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
+  def int_x86_avx2_psllv_d : GCCBuiltin<"__builtin_ia32_psllv4si">,
+              Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v4i32_ty],
+                        [IntrNoMem]>;
+  def int_x86_avx2_psllv_d_256 : GCCBuiltin<"__builtin_ia32_psllv8si">,
+              Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty, llvm_v8i32_ty],
+                        [IntrNoMem]>;
+  def int_x86_avx2_psllv_q : GCCBuiltin<"__builtin_ia32_psllv2di">,
+              Intrinsic<[llvm_v2i64_ty], [llvm_v2i64_ty, llvm_v2i64_ty],
+                        [IntrNoMem]>;
+  def int_x86_avx2_psllv_q_256 : GCCBuiltin<"__builtin_ia32_psllv4di">,
+              Intrinsic<[llvm_v4i64_ty], [llvm_v4i64_ty, llvm_v4i64_ty],
+                        [IntrNoMem]>;
+
+  def int_x86_avx2_psrlv_d : GCCBuiltin<"__builtin_ia32_psrlv4si">,
+              Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v4i32_ty],
+                        [IntrNoMem]>;
+  def int_x86_avx2_psrlv_d_256 : GCCBuiltin<"__builtin_ia32_psrlv8si">,
+              Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty, llvm_v8i32_ty],
+                        [IntrNoMem]>;
+  def int_x86_avx2_psrlv_q : GCCBuiltin<"__builtin_ia32_psrlv2di">,
+              Intrinsic<[llvm_v2i64_ty], [llvm_v2i64_ty, llvm_v2i64_ty],
+                        [IntrNoMem]>;
+  def int_x86_avx2_psrlv_q_256 : GCCBuiltin<"__builtin_ia32_psrlv4di">,
+              Intrinsic<[llvm_v4i64_ty], [llvm_v4i64_ty, llvm_v4i64_ty],
+                        [IntrNoMem]>;
+
+  def int_x86_avx2_psrav_d : GCCBuiltin<"__builtin_ia32_psrav4si">,
+              Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v4i32_ty],
+                        [IntrNoMem]>;
+  def int_x86_avx2_psrav_d_256 : GCCBuiltin<"__builtin_ia32_psrav8si">,
+              Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty, llvm_v8i32_ty],
+                        [IntrNoMem]>;
+}
+
 // Misc.
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_avx2_pmovmskb : GCCBuiltin<"__builtin_ia32_pmovmskb256">,
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index b46e5d10e04..db4382a0c1a 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -7601,3 +7601,43 @@ defm VPMASKMOVQ : avx2_pmovmask<"vpmaskmovq",
                                 int_x86_avx2_maskstore_q,
                                 int_x86_avx2_maskstore_q_256,
                                 memopv2i64, memopv4i64>, VEX_W;
+
+
+//===----------------------------------------------------------------------===//
+// Variable Bit Shifts
+//
+multiclass avx2_var_shift<bits<8> opc, string OpcodeStr,
+                          PatFrag pf128, PatFrag pf256,
+                          Intrinsic Int128, Intrinsic Int256> {
+  def rr  : AVX28I<opc, MRMSrcReg, (outs VR128:$dst),
+             (ins VR128:$src1, VR128:$src2),
+             !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+             [(set VR128:$dst, (Int128 VR128:$src1, VR128:$src2))]>, VEX_4V;
+  def rm  : AVX28I<opc, MRMSrcMem, (outs VR128:$dst),
+             (ins VR128:$src1, i128mem:$src2),
+             !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+             [(set VR128:$dst, (Int128 VR128:$src1, (pf128 addr:$src2)))]>,
+             VEX_4V;
+  def Yrr : AVX28I<opc, MRMSrcReg, (outs VR256:$dst),
+             (ins VR256:$src1, VR256:$src2),
+             !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+             [(set VR256:$dst, (Int256 VR256:$src1, VR256:$src2))]>, VEX_4V;
+  def Yrm : AVX28I<opc, MRMSrcMem, (outs VR256:$dst),
+             (ins VR256:$src1, i256mem:$src2),
+             !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+             [(set VR256:$dst, (Int256 VR256:$src1, (pf256 addr:$src2)))]>,
+             VEX_4V;
+}
+
+defm VPSLLVD : avx2_var_shift<0x47, "vpsllvd", memopv4i32, memopv8i32,
+                              int_x86_avx2_psllv_d, int_x86_avx2_psllv_d_256>;
+defm VPSLLVQ : avx2_var_shift<0x47, "vpsllvq", memopv2i64, memopv4i64,
+                              int_x86_avx2_psllv_q, int_x86_avx2_psllv_q_256>,
+                              VEX_W;
+defm VPSRLVD : avx2_var_shift<0x45, "vpsrlvd", memopv4i32, memopv8i32,
+                              int_x86_avx2_psrlv_d, int_x86_avx2_psrlv_d_256>;
+defm VPSRLVQ : avx2_var_shift<0x45, "vpsrlvq", memopv2i64, memopv4i64,
+                              int_x86_avx2_psrlv_q, int_x86_avx2_psrlv_q_256>,
+                              VEX_W;
+defm VPSRAVD : avx2_var_shift<0x46, "vpsravd", memopv4i32, memopv8i32,
+                              int_x86_avx2_psrav_d, int_x86_avx2_psrav_d_256>;
diff --git a/test/CodeGen/X86/avx2-intrinsics-x86.ll b/test/CodeGen/X86/avx2-intrinsics-x86.ll
index da03c386273..bab7fb81e53 100644
--- a/test/CodeGen/X86/avx2-intrinsics-x86.ll
+++ b/test/CodeGen/X86/avx2-intrinsics-x86.ll
@@ -966,3 +966,83 @@ define void @test_x86_avx2_maskstore_d_256(i8* %a0, <8 x i32> %a1, <8 x i32> %a2
   ret void
 }
 declare void @llvm.x86.avx2.maskstore.d.256(i8*, <8 x i32>, <8 x i32>) nounwind
+
+
+define <4 x i32> @test_x86_avx2_psllv_d(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpsllvd
+  %res = call <4 x i32> @llvm.x86.avx2.psllv.d(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.avx2.psllv.d(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <8 x i32> @test_x86_avx2_psllv_d_256(<8 x i32> %a0, <8 x i32> %a1) {
+  ; CHECK: vpsllvd
+  %res = call <8 x i32> @llvm.x86.avx2.psllv.d.256(<8 x i32> %a0, <8 x i32> %a1) ; <<8 x i32>> [#uses=1]
+  ret <8 x i32> %res
+}
+declare <8 x i32> @llvm.x86.avx2.psllv.d.256(<8 x i32>, <8 x i32>) nounwind readnone
+
+
+define <2 x i64> @test_x86_avx2_psllv_q(<2 x i64> %a0, <2 x i64> %a1) {
+  ; CHECK: vpsllvq
+  %res = call <2 x i64> @llvm.x86.avx2.psllv.q(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.avx2.psllv.q(<2 x i64>, <2 x i64>) nounwind readnone
+
+
+define <4 x i64> @test_x86_avx2_psllv_q_256(<4 x i64> %a0, <4 x i64> %a1) {
+  ; CHECK: vpsllvq
+  %res = call <4 x i64> @llvm.x86.avx2.psllv.q.256(<4 x i64> %a0, <4 x i64> %a1) ; <<4 x i64>> [#uses=1]
+  ret <4 x i64> %res
+}
+declare <4 x i64> @llvm.x86.avx2.psllv.q.256(<4 x i64>, <4 x i64>) nounwind readnone
+
+
+define <4 x i32> @test_x86_avx2_psrlv_d(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpsrlvd
+  %res = call <4 x i32> @llvm.x86.avx2.psrlv.d(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.avx2.psrlv.d(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <8 x i32> @test_x86_avx2_psrlv_d_256(<8 x i32> %a0, <8 x i32> %a1) {
+  ; CHECK: vpsrlvd
+  %res = call <8 x i32> @llvm.x86.avx2.psrlv.d.256(<8 x i32> %a0, <8 x i32> %a1) ; <<8 x i32>> [#uses=1]
+  ret <8 x i32> %res
+}
+declare <8 x i32> @llvm.x86.avx2.psrlv.d.256(<8 x i32>, <8 x i32>) nounwind readnone
+
+
+define <2 x i64> @test_x86_avx2_psrlv_q(<2 x i64> %a0, <2 x i64> %a1) {
+  ; CHECK: vpsrlvq
+  %res = call <2 x i64> @llvm.x86.avx2.psrlv.q(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.avx2.psrlv.q(<2 x i64>, <2 x i64>) nounwind readnone
+
+
+define <4 x i64> @test_x86_avx2_psrlv_q_256(<4 x i64> %a0, <4 x i64> %a1) {
+  ; CHECK: vpsrlvq
+  %res = call <4 x i64> @llvm.x86.avx2.psrlv.q.256(<4 x i64> %a0, <4 x i64> %a1) ; <<4 x i64>> [#uses=1]
+  ret <4 x i64> %res
+}
+declare <4 x i64> @llvm.x86.avx2.psrlv.q.256(<4 x i64>, <4 x i64>) nounwind readnone
+
+
+define <4 x i32> @test_x86_avx2_psrav_d(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpsravd
+  %res = call <4 x i32> @llvm.x86.avx2.psrav.d(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.avx2.psrav.d(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <8 x i32> @test_x86_avx2_psrav_d_256(<8 x i32> %a0, <8 x i32> %a1) {
+  ; CHECK: vpsravd
+  %res = call <8 x i32> @llvm.x86.avx2.psrav.d.256(<8 x i32> %a0, <8 x i32> %a1) ; <<8 x i32>> [#uses=1]
+  ret <8 x i32> %res
+}
+declare <8 x i32> @llvm.x86.avx2.psrav.d.256(<8 x i32>, <8 x i32>) nounwind readnone
-- 
2.11.0