From 7dbfd07e321dfdc90501d71a144749c69263659f Mon Sep 17 00:00:00 2001
From: Bruno Cardoso Lopes <bruno.cardoso@gmail.com>
Date: Wed, 23 Jun 2010 20:07:15 +0000
Subject: [PATCH] Add AVX SHUF{PS,PD}{rr,rm} instructions

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@106672 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/X86/X86InstrSSE.td           | 56 +++++++++++++------------
 test/MC/AsmParser/X86/x86_32-encoding.s | 16 +++++++
 test/MC/AsmParser/X86/x86_64-encoding.s | 16 +++++++
 3 files changed, 62 insertions(+), 26 deletions(-)

diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index 5e892796efd..3ecdc1c2468 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -1394,33 +1394,37 @@ def : Pat<(v2i64 (X86cmppd (v2f64 VR128:$src1), (memop addr:$src2), imm:$cc)),
           (CMPPDrmi VR128:$src1, addr:$src2, imm:$cc)>;
 
 // Shuffle and unpack instructions
+multiclass sse12_shuffle<RegisterClass RC, X86MemOperand x86memop,
+                         ValueType vt, string asm, PatFrag mem_frag,
+                         Domain d, bit IsConvertibleToThreeAddress = 0> {
+  def rmi : PIi8<0xC6, MRMSrcMem, (outs VR128:$dst),
+                   (ins VR128:$src1, f128mem:$src2, i8imm:$src3), asm,
+                   [(set VR128:$dst, (vt (shufp:$src3
+                            VR128:$src1, (mem_frag addr:$src2))))], d>;
+  let isConvertibleToThreeAddress = IsConvertibleToThreeAddress in
+    def rri : PIi8<0xC6, MRMSrcReg, (outs VR128:$dst),
+                   (ins VR128:$src1, VR128:$src2, i8imm:$src3), asm,
+                   [(set VR128:$dst,
+                            (vt (shufp:$src3 VR128:$src1, VR128:$src2)))], d>;
+}
+
 let Constraints = "$src1 = $dst" in {
-  let isConvertibleToThreeAddress = 1 in // Convert to pshufd
-    def SHUFPSrri : PSIi8<0xC6, MRMSrcReg,
-                          (outs VR128:$dst), (ins VR128:$src1,
-                           VR128:$src2, i8imm:$src3),
-                          "shufps\t{$src3, $src2, $dst|$dst, $src2, $src3}",
-                          [(set VR128:$dst,
-                            (v4f32 (shufp:$src3 VR128:$src1, VR128:$src2)))]>;
-  def SHUFPSrmi : PSIi8<0xC6, MRMSrcMem,
-                        (outs VR128:$dst), (ins VR128:$src1,
-                         f128mem:$src2, i8imm:$src3),
-                        "shufps\t{$src3, $src2, $dst|$dst, $src2, $src3}",
-                        [(set VR128:$dst,
-                          (v4f32 (shufp:$src3
-                                  VR128:$src1, (memopv4f32 addr:$src2))))]>;
-  def SHUFPDrri : PDIi8<0xC6, MRMSrcReg,
-                 (outs VR128:$dst), (ins VR128:$src1, VR128:$src2, i8imm:$src3),
-                 "shufpd\t{$src3, $src2, $dst|$dst, $src2, $src3}",
-                 [(set VR128:$dst,
-                   (v2f64 (shufp:$src3 VR128:$src1, VR128:$src2)))]>;
-  def SHUFPDrmi : PDIi8<0xC6, MRMSrcMem,
-                        (outs VR128:$dst), (ins VR128:$src1,
-                         f128mem:$src2, i8imm:$src3),
-                        "shufpd\t{$src3, $src2, $dst|$dst, $src2, $src3}",
-                        [(set VR128:$dst,
-                          (v2f64 (shufp:$src3
-                                  VR128:$src1, (memopv2f64 addr:$src2))))]>;
+  defm SHUFPS : sse12_shuffle<VR128, f128mem, v4f32,
+                    "shufps\t{$src3, $src2, $dst|$dst, $src2, $src3}",
+                    memopv4f32, SSEPackedSingle, 1 /* cvt to pshufd */>,
+                    TB;
+  defm SHUFPD : sse12_shuffle<VR128, f128mem, v2f64,
+                    "shufpd\t{$src3, $src2, $dst|$dst, $src2, $src3}",
+                    memopv2f64, SSEPackedDouble>, TB, OpSize;
+
+  let Constraints = "", isAsmParserOnly = 1 in {
+    defm VSHUFPS : sse12_shuffle<VR128, f128mem, v4f32,
+              "shufps\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
+              memopv4f32, SSEPackedSingle>, VEX_4V;
+    defm VSHUFPD : sse12_shuffle<VR128, f128mem, v2f64,
+              "shufpd\t{$src3, $src2, $src1, $dst|$dst, $src2, $src2, $src3}",
+              memopv2f64, SSEPackedDouble>, OpSize, VEX_4V;
+  }
 
   let AddedComplexity = 10 in {
     let Constraints = "", isAsmParserOnly = 1 in {
diff --git a/test/MC/AsmParser/X86/x86_32-encoding.s b/test/MC/AsmParser/X86/x86_32-encoding.s
index c180ee0dc42..d8d39e3a430 100644
--- a/test/MC/AsmParser/X86/x86_32-encoding.s
+++ b/test/MC/AsmParser/X86/x86_32-encoding.s
@@ -10374,3 +10374,19 @@
 // CHECK: encoding: [0xc5,0xc9,0xc2,0xc8,0x07]
           vcmppd  $7, %xmm0, %xmm6, %xmm1
 
+// CHECK: vshufps  $8, %xmm1, %xmm2, %xmm3
+// CHECK: encoding: [0xc5,0xe8,0xc6,0xd9,0x08]
+          vshufps  $8, %xmm1, %xmm2, %xmm3
+
+// CHECK: vshufps  $8, -4(%ebx,%ecx,8), %xmm2, %xmm3
+// CHECK: encoding: [0xc5,0xe8,0xc6,0x5c,0xcb,0xfc,0x08]
+          vshufps  $8, -4(%ebx,%ecx,8), %xmm2, %xmm3
+
+// CHECK: vshufpd  $8, %xmm1, %xmm2, %xmm3
+// CHECK: encoding: [0xc5,0xe9,0xc6,0xd9,0x08]
+          vshufpd  $8, %xmm1, %xmm2, %xmm3
+
+// CHECK: vshufpd  $8, -4(%ebx,%ecx,8), %xmm2, %xmm3
+// CHECK: encoding: [0xc5,0xe9,0xc6,0x5c,0xcb,0xfc,0x08]
+          vshufpd  $8, -4(%ebx,%ecx,8), %xmm2, %xmm3
+
diff --git a/test/MC/AsmParser/X86/x86_64-encoding.s b/test/MC/AsmParser/X86/x86_64-encoding.s
index 19efb6d19c3..be698b6699e 100644
--- a/test/MC/AsmParser/X86/x86_64-encoding.s
+++ b/test/MC/AsmParser/X86/x86_64-encoding.s
@@ -438,3 +438,19 @@ pshufb	CPI1_0(%rip), %xmm1
 // CHECK: encoding: [0xc4,0x41,0x19,0xc2,0xfa,0x07]
           vcmppd  $7, %xmm10, %xmm12, %xmm15
 
+// CHECK: vshufps  $8, %xmm11, %xmm12, %xmm13
+// CHECK: encoding: [0xc4,0x41,0x18,0xc6,0xeb,0x08]
+          vshufps  $8, %xmm11, %xmm12, %xmm13
+
+// CHECK: vshufps  $8, -4(%rbx,%rcx,8), %xmm12, %xmm13
+// CHECK: encoding: [0xc5,0x18,0xc6,0x6c,0xcb,0xfc,0x08]
+          vshufps  $8, -4(%rbx,%rcx,8), %xmm12, %xmm13
+
+// CHECK: vshufpd  $8, %xmm11, %xmm12, %xmm13
+// CHECK: encoding: [0xc4,0x41,0x19,0xc6,0xeb,0x08]
+          vshufpd  $8, %xmm11, %xmm12, %xmm13
+
+// CHECK: vshufpd  $8, -4(%rbx,%rcx,8), %xmm12, %xmm13
+// CHECK: encoding: [0xc5,0x19,0xc6,0x6c,0xcb,0xfc,0x08]
+          vshufpd  $8, -4(%rbx,%rcx,8), %xmm12, %xmm13
+