From cb49278ad60224dd7ede8f159bc9ac4ec41316e3 Mon Sep 17 00:00:00 2001
From: Bruno Cardoso Lopes <bruno.cardoso@gmail.com>
Date: Wed, 13 Jul 2011 01:15:33 +0000
Subject: [PATCH] AVX Codegen support for 256-bit versions of vandps, vandpd,
 vorps, vorpd, vxorps, vxorpd

llvm-svn: 135023
---
 lib/Target/X86/X86InstrFragmentsSIMD.td |   1 +
 lib/Target/X86/X86InstrSSE.td           |  34 +++++--
 test/CodeGen/X86/avx-256-logic.ll       | 116 ++++++++++++++++++++++++
 3 files changed, 143 insertions(+), 8 deletions(-)
 create mode 100644 test/CodeGen/X86/avx-256-logic.ll

diff --git a/lib/Target/X86/X86InstrFragmentsSIMD.td b/lib/Target/X86/X86InstrFragmentsSIMD.td
index 7c9a9f7e8c5..0d6ff38a91d 100644
--- a/lib/Target/X86/X86InstrFragmentsSIMD.td
+++ b/lib/Target/X86/X86InstrFragmentsSIMD.td
@@ -301,6 +301,7 @@ def bc_v2i64 : PatFrag<(ops node:$in), (v2i64 (bitconvert node:$in))>;
 
 // 256-bit bitconvert pattern fragments
 def bc_v8i32 : PatFrag<(ops node:$in), (v8i32 (bitconvert node:$in))>;
+def bc_v4i64 : PatFrag<(ops node:$in), (v4i64 (bitconvert node:$in))>;
 
 def vzmovl_v2i64 : PatFrag<(ops node:$src),
                            (bitconvert (v2i64 (X86vzmovl
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index 6d22f2b60ed..93902d171ab 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -1522,20 +1522,35 @@ multiclass sse12_fp_packed_logical<bits<8> opc, string OpcodeStr,
 
 /// sse12_fp_packed_logical_y - AVX 256-bit SSE 1 & 2 logical ops forms
 ///
-multiclass sse12_fp_packed_logical_y<bits<8> opc, string OpcodeStr> {
+multiclass sse12_fp_packed_logical_y<bits<8> opc, string OpcodeStr,
+                                     SDNode OpNode, int HasNoPat = 0> {
     defm PSY : sse12_fp_packed_logical_rm<opc, VR256, SSEPackedSingle,
-          !strconcat(OpcodeStr, "ps"), f256mem, [], [], 0>, VEX_4V;
+          !strconcat(OpcodeStr, "ps"), f256mem,
+          !if(HasNoPat, []<dag>, // rr
+              [(set VR256:$dst, (v4i64 (OpNode VR256:$src1,
+                                               VR256:$src2)))]),
+          !if(HasNoPat, []<dag>, // rm
+              [(set VR256:$dst, (OpNode (bc_v4i64 (v8f32 VR256:$src1)),
+                                        (memopv4i64 addr:$src2)))]), 0>, VEX_4V;
 
     defm PDY : sse12_fp_packed_logical_rm<opc, VR256, SSEPackedDouble,
-          !strconcat(OpcodeStr, "pd"), f256mem, [], [], 0>, OpSize, VEX_4V;
+          !strconcat(OpcodeStr, "pd"), f256mem,
+          !if(HasNoPat, []<dag>, // rr
+              [(set VR256:$dst, (OpNode (bc_v4i64 (v4f64 VR256:$src1)),
+                                        (bc_v4i64 (v4f64 VR256:$src2))))]),
+          !if(HasNoPat, []<dag>, // rm
+              [(set VR256:$dst, (OpNode (bc_v4i64 (v4f64 VR256:$src1)),
+                                        (memopv4i64 addr:$src2)))]), 0>,
+                                        OpSize, VEX_4V;
 }
 
 // AVX 256-bit packed logical ops forms
-defm VAND : sse12_fp_packed_logical_y<0x54, "and">;
-defm VOR  : sse12_fp_packed_logical_y<0x56, "or">;
-defm VXOR : sse12_fp_packed_logical_y<0x57, "xor">;
-let isCommutable = 0 in
-  defm VANDN : sse12_fp_packed_logical_y<0x55, "andn">;
+defm VAND : sse12_fp_packed_logical_y<0x54, "and", and>;
+defm VOR  : sse12_fp_packed_logical_y<0x56, "or", or>;
+defm VXOR : sse12_fp_packed_logical_y<0x57, "xor", xor>;
+let isCommutable = 0 in {
+  defm VANDN : sse12_fp_packed_logical_y<0x55, "andn", undef /* dummy */, 1>;
+}
 
 defm AND  : sse12_fp_packed_logical<0x54, "and", and>;
 defm OR   : sse12_fp_packed_logical<0x56, "or", or>;
@@ -3660,6 +3675,9 @@ let Predicates = [HasXMMInt] in {
 
 let Predicates = [HasAVX] in {
   def : Pat<(v4f64 (bitconvert (v8f32 VR256:$src))), (v4f64 VR256:$src)>;
+  def : Pat<(v4f64 (bitconvert (v4i64 VR256:$src))), (v4f64 VR256:$src)>;
+  def : Pat<(v8f32 (bitconvert (v4i64 VR256:$src))), (v8f32 VR256:$src)>;
+  def : Pat<(v4i64 (bitconvert (v8f32 VR256:$src))), (v4i64 VR256:$src)>;
 }
 
 // Move scalar to XMM zero-extended
diff --git a/test/CodeGen/X86/avx-256-logic.ll b/test/CodeGen/X86/avx-256-logic.ll
new file mode 100644
index 00000000000..05e82895fec
--- /dev/null
+++ b/test/CodeGen/X86/avx-256-logic.ll
@@ -0,0 +1,116 @@
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7-avx -mattr=+avx | FileCheck %s
+
+; CHECK: vandpd
+define <4 x double> @andpd256(<4 x double> %y, <4 x double> %x) nounwind uwtable readnone ssp {
+entry:
+  %0 = bitcast <4 x double> %x to <4 x i64>
+  %1 = bitcast <4 x double> %y to <4 x i64>
+  %and.i = and <4 x i64> %0, %1
+  %2 = bitcast <4 x i64> %and.i to <4 x double>
+  ret <4 x double> %2
+}
+
+; CHECK: vandpd LCP{{.*}}(%rip)
+define <4 x double> @andpd256fold(<4 x double> %y) nounwind uwtable readnone ssp {
+entry:
+  %0 = bitcast <4 x double> %y to <4 x i64>
+  %and.i = and <4 x i64> %0, <i64 4616752568008179712, i64 4614838538166547251, i64 4612361558371493478, i64 4608083138725491507>
+  %1 = bitcast <4 x i64> %and.i to <4 x double>
+  ret <4 x double> %1
+}
+
+; CHECK: vandps
+define <8 x float> @andps256(<8 x float> %y, <8 x float> %x) nounwind uwtable readnone ssp {
+entry:
+  %0 = bitcast <8 x float> %x to <8 x i32>
+  %1 = bitcast <8 x float> %y to <8 x i32>
+  %and.i = and <8 x i32> %0, %1
+  %2 = bitcast <8 x i32> %and.i to <8 x float>
+  ret <8 x float> %2
+}
+
+; CHECK: vandps LCP{{.*}}(%rip)
+define <8 x float> @andps256fold(<8 x float> %y) nounwind uwtable readnone ssp {
+entry:
+  %0 = bitcast <8 x float> %y to <8 x i32>
+  %and.i = and <8 x i32> %0, <i32 1083179008, i32 1079613850, i32 1075000115, i32 1067030938, i32 1083179008, i32 1079613850, i32 1075000115, i32 1067030938>
+  %1 = bitcast <8 x i32> %and.i to <8 x float>
+  ret <8 x float> %1
+}
+
+; CHECK: vxorpd
+define <4 x double> @xorpd256(<4 x double> %y, <4 x double> %x) nounwind uwtable readnone ssp {
+entry:
+  %0 = bitcast <4 x double> %x to <4 x i64>
+  %1 = bitcast <4 x double> %y to <4 x i64>
+  %xor.i = xor <4 x i64> %0, %1
+  %2 = bitcast <4 x i64> %xor.i to <4 x double>
+  ret <4 x double> %2
+}
+
+; CHECK: vxorpd LCP{{.*}}(%rip)
+define <4 x double> @xorpd256fold(<4 x double> %y) nounwind uwtable readnone ssp {
+entry:
+  %0 = bitcast <4 x double> %y to <4 x i64>
+  %xor.i = xor <4 x i64> %0, <i64 4616752568008179712, i64 4614838538166547251, i64 4612361558371493478, i64 4608083138725491507>
+  %1 = bitcast <4 x i64> %xor.i to <4 x double>
+  ret <4 x double> %1
+}
+
+; CHECK: vxorps
+define <8 x float> @xorps256(<8 x float> %y, <8 x float> %x) nounwind uwtable readnone ssp {
+entry:
+  %0 = bitcast <8 x float> %x to <8 x i32>
+  %1 = bitcast <8 x float> %y to <8 x i32>
+  %xor.i = xor <8 x i32> %0, %1
+  %2 = bitcast <8 x i32> %xor.i to <8 x float>
+  ret <8 x float> %2
+}
+
+; CHECK: vxorps LCP{{.*}}(%rip)
+define <8 x float> @xorps256fold(<8 x float> %y) nounwind uwtable readnone ssp {
+entry:
+  %0 = bitcast <8 x float> %y to <8 x i32>
+  %xor.i = xor <8 x i32> %0, <i32 1083179008, i32 1079613850, i32 1075000115, i32 1067030938, i32 1083179008, i32 1079613850, i32 1075000115, i32 1067030938>
+  %1 = bitcast <8 x i32> %xor.i to <8 x float>
+  ret <8 x float> %1
+}
+
+; CHECK: vorpd
+define <4 x double> @orpd256(<4 x double> %y, <4 x double> %x) nounwind uwtable readnone ssp {
+entry:
+  %0 = bitcast <4 x double> %x to <4 x i64>
+  %1 = bitcast <4 x double> %y to <4 x i64>
+  %or.i = or <4 x i64> %0, %1
+  %2 = bitcast <4 x i64> %or.i to <4 x double>
+  ret <4 x double> %2
+}
+
+; CHECK: vorpd LCP{{.*}}(%rip)
+define <4 x double> @orpd256fold(<4 x double> %y) nounwind uwtable readnone ssp {
+entry:
+  %0 = bitcast <4 x double> %y to <4 x i64>
+  %or.i = or <4 x i64> %0, <i64 4616752568008179712, i64 4614838538166547251, i64 4612361558371493478, i64 4608083138725491507>
+  %1 = bitcast <4 x i64> %or.i to <4 x double>
+  ret <4 x double> %1
+}
+
+; CHECK: vorps
+define <8 x float> @orps256(<8 x float> %y, <8 x float> %x) nounwind uwtable readnone ssp {
+entry:
+  %0 = bitcast <8 x float> %x to <8 x i32>
+  %1 = bitcast <8 x float> %y to <8 x i32>
+  %or.i = or <8 x i32> %0, %1
+  %2 = bitcast <8 x i32> %or.i to <8 x float>
+  ret <8 x float> %2
+}
+
+; CHECK: vorps LCP{{.*}}(%rip)
+define <8 x float> @orps256fold(<8 x float> %y) nounwind uwtable readnone ssp {
+entry:
+  %0 = bitcast <8 x float> %y to <8 x i32>
+  %or.i = or <8 x i32> %0, <i32 1083179008, i32 1079613850, i32 1075000115, i32 1067030938, i32 1083179008, i32 1079613850, i32 1075000115, i32 1067030938>
+  %1 = bitcast <8 x i32> %or.i to <8 x float>
+  ret <8 x float> %1
+}
+