From b1ef950217f810d458fa5ec80d370c7fbfa4bc44 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@gmail.com>
Date: Mon, 7 Nov 2011 03:20:35 +0000
Subject: [PATCH] Add AVX2 VPMOVMASK instructions and intrinsics.

llvm-svn: 143904
---
 include/llvm/IntrinsicsX86.td           | 27 +++++++++++
 lib/Target/X86/X86InstrSSE.td           | 38 +++++++++++++++
 test/CodeGen/X86/avx2-intrinsics-x86.ll | 64 +++++++++++++++++++++++++
 3 files changed, 129 insertions(+)

diff --git a/include/llvm/IntrinsicsX86.td b/include/llvm/IntrinsicsX86.td
index 345c2803f4b..42209b8291d 100644
--- a/include/llvm/IntrinsicsX86.td
+++ b/include/llvm/IntrinsicsX86.td
@@ -1744,6 +1744,33 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
               Intrinsic<[llvm_v4i64_ty], [llvm_v4i64_ty,
                          llvm_v2i64_ty, llvm_i8_ty], [IntrNoMem]>;
 }
+
+// Conditional load ops
+let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
+  def int_x86_avx2_maskload_d : GCCBuiltin<"__builtin_ia32_maskloadd">,
+        Intrinsic<[llvm_v4i32_ty], [llvm_ptr_ty, llvm_v4i32_ty], [IntrReadMem]>;
+  def int_x86_avx2_maskload_q : GCCBuiltin<"__builtin_ia32_maskloadq">,
+        Intrinsic<[llvm_v2i64_ty], [llvm_ptr_ty, llvm_v2i64_ty], [IntrReadMem]>;
+  def int_x86_avx2_maskload_d_256 : GCCBuiltin<"__builtin_ia32_maskloadd256">,
+        Intrinsic<[llvm_v8i32_ty], [llvm_ptr_ty, llvm_v8i32_ty], [IntrReadMem]>;
+  def int_x86_avx2_maskload_q_256 : GCCBuiltin<"__builtin_ia32_maskloadq256">,
+        Intrinsic<[llvm_v4i64_ty], [llvm_ptr_ty, llvm_v4i64_ty], [IntrReadMem]>;
+}
+
+// Conditional store ops
+let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
+  def int_x86_avx2_maskstore_d : GCCBuiltin<"__builtin_ia32_maskstored">,
+        Intrinsic<[], [llvm_ptr_ty, llvm_v4i32_ty, llvm_v4i32_ty], []>;
+  def int_x86_avx2_maskstore_q : GCCBuiltin<"__builtin_ia32_maskstoreq">,
+        Intrinsic<[], [llvm_ptr_ty, llvm_v2i64_ty, llvm_v2i64_ty], []>;
+  def int_x86_avx2_maskstore_d_256 :
+        GCCBuiltin<"__builtin_ia32_maskstored256">,
+        Intrinsic<[], [llvm_ptr_ty, llvm_v8i32_ty, llvm_v8i32_ty], []>;
+  def int_x86_avx2_maskstore_q_256 :
+        GCCBuiltin<"__builtin_ia32_maskstoreq256">,
+        Intrinsic<[], [llvm_ptr_ty, llvm_v4i64_ty, llvm_v4i64_ty], []>;
+}
+
 // 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 fc36884bf1f..b46e5d10e04 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -7563,3 +7563,41 @@ let neverHasSideEffects = 1, mayStore = 1 in
 def VEXTRACTI128mr : AVX2AIi8<0x39, MRMDestMem, (outs),
           (ins i128mem:$dst, VR256:$src1, i8imm:$src2),
           "vextracti128\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>, VEX;
+
+//===----------------------------------------------------------------------===//
+// VPMASKMOV - Conditional SIMD Integer Packed Loads and Stores
+//
+multiclass avx2_pmovmask<string OpcodeStr,
+                         Intrinsic IntLd128, Intrinsic IntLd256,
+                         Intrinsic IntSt128, Intrinsic IntSt256,
+                         PatFrag pf128, PatFrag pf256> {
+  def rm  : AVX28I<0x8c, MRMSrcMem, (outs VR128:$dst),
+             (ins VR128:$src1, i128mem:$src2),
+             !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+             [(set VR128:$dst, (IntLd128 addr:$src2, VR128:$src1))]>, VEX_4V;
+  def Yrm : AVX28I<0x8c, MRMSrcMem, (outs VR256:$dst),
+             (ins VR256:$src1, i256mem:$src2),
+             !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+             [(set VR256:$dst, (IntLd256 addr:$src2, VR256:$src1))]>, VEX_4V;
+  def mr  : AVX28I<0x8e, MRMDestMem, (outs),
+             (ins i128mem:$dst, VR128:$src1, VR128:$src2),
+             !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+             [(IntSt128 addr:$dst, VR128:$src1, VR128:$src2)]>, VEX_4V;
+  def Ymr : AVX28I<0x8e, MRMDestMem, (outs),
+             (ins i256mem:$dst, VR256:$src1, VR256:$src2),
+             !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+             [(IntSt256 addr:$dst, VR256:$src1, VR256:$src2)]>, VEX_4V;
+}
+
+defm VPMASKMOVD : avx2_pmovmask<"vpmaskmovd",
+                                int_x86_avx2_maskload_d,
+                                int_x86_avx2_maskload_d_256,
+                                int_x86_avx2_maskstore_d,
+                                int_x86_avx2_maskstore_d_256,
+                                memopv4i32, memopv8i32>;
+defm VPMASKMOVQ : avx2_pmovmask<"vpmaskmovq",
+                                int_x86_avx2_maskload_q,
+                                int_x86_avx2_maskload_q_256,
+                                int_x86_avx2_maskstore_q,
+                                int_x86_avx2_maskstore_q_256,
+                                memopv2i64, memopv4i64>, VEX_W;
diff --git a/test/CodeGen/X86/avx2-intrinsics-x86.ll b/test/CodeGen/X86/avx2-intrinsics-x86.ll
index 4ab0884b5b7..da03c386273 100644
--- a/test/CodeGen/X86/avx2-intrinsics-x86.ll
+++ b/test/CodeGen/X86/avx2-intrinsics-x86.ll
@@ -902,3 +902,67 @@ define <4 x i64> @test_x86_avx2_vinserti128(<4 x i64> %a0, <2 x i64> %a1) {
   ret <4 x i64> %res
 }
 declare <4 x i64> @llvm.x86.avx2.vinserti128(<4 x i64>, <2 x i64>, i8) nounwind readnone
+
+
+define <2 x i64> @test_x86_avx2_maskload_q(i8* %a0, <2 x i64> %a1) {
+  ; CHECK: vpmaskmovq
+  %res = call <2 x i64> @llvm.x86.avx2.maskload.q(i8* %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.avx2.maskload.q(i8*, <2 x i64>) nounwind readonly
+
+
+define <4 x i64> @test_x86_avx2_maskload_q_256(i8* %a0, <4 x i64> %a1) {
+  ; CHECK: vpmaskmovq
+  %res = call <4 x i64> @llvm.x86.avx2.maskload.q.256(i8* %a0, <4 x i64> %a1) ; <<4 x i64>> [#uses=1]
+  ret <4 x i64> %res
+}
+declare <4 x i64> @llvm.x86.avx2.maskload.q.256(i8*, <4 x i64>) nounwind readonly
+
+
+define <4 x i32> @test_x86_avx2_maskload_d(i8* %a0, <4 x i32> %a1) {
+  ; CHECK: vpmaskmovd
+  %res = call <4 x i32> @llvm.x86.avx2.maskload.d(i8* %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.avx2.maskload.d(i8*, <4 x i32>) nounwind readonly
+
+
+define <8 x i32> @test_x86_avx2_maskload_d_256(i8* %a0, <8 x i32> %a1) {
+  ; CHECK: vpmaskmovd
+  %res = call <8 x i32> @llvm.x86.avx2.maskload.d.256(i8* %a0, <8 x i32> %a1) ; <<8 x i32>> [#uses=1]
+  ret <8 x i32> %res
+}
+declare <8 x i32> @llvm.x86.avx2.maskload.d.256(i8*, <8 x i32>) nounwind readonly
+
+
+define void @test_x86_avx2_maskstore_q(i8* %a0, <2 x i64> %a1, <2 x i64> %a2) {
+  ; CHECK: vpmaskmovq
+  call void @llvm.x86.avx2.maskstore.q(i8* %a0, <2 x i64> %a1, <2 x i64> %a2)
+  ret void
+}
+declare void @llvm.x86.avx2.maskstore.q(i8*, <2 x i64>, <2 x i64>) nounwind
+
+
+define void @test_x86_avx2_maskstore_q_256(i8* %a0, <4 x i64> %a1, <4 x i64> %a2) {
+  ; CHECK: vpmaskmovq
+  call void @llvm.x86.avx2.maskstore.q.256(i8* %a0, <4 x i64> %a1, <4 x i64> %a2)
+  ret void
+}
+declare void @llvm.x86.avx2.maskstore.q.256(i8*, <4 x i64>, <4 x i64>) nounwind
+
+
+define void @test_x86_avx2_maskstore_d(i8* %a0, <4 x i32> %a1, <4 x i32> %a2) {
+  ; CHECK: vpmaskmovd
+  call void @llvm.x86.avx2.maskstore.d(i8* %a0, <4 x i32> %a1, <4 x i32> %a2)
+  ret void
+}
+declare void @llvm.x86.avx2.maskstore.d(i8*, <4 x i32>, <4 x i32>) nounwind
+
+
+define void @test_x86_avx2_maskstore_d_256(i8* %a0, <8 x i32> %a1, <8 x i32> %a2) {
+  ; CHECK: vpmaskmovd
+  call void @llvm.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