From ebf01d63b0ae24b29aa3e917d51458231c203b6f Mon Sep 17 00:00:00 2001
From: Evan Cheng <evan.cheng@apple.com>
Date: Thu, 16 Nov 2006 23:33:25 +0000
Subject: [PATCH] Correct instructions for moving data between GR64 and SSE
 registers; also correct load i64 / store i64 from v2i64.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@31795 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/X86/X86InstrSSE.td      | 11 ++++++-----
 lib/Target/X86/X86InstrX86-64.td   | 26 ++++++++++++++++++++++++++
 lib/Target/X86/X86RegisterInfo.cpp |  3 ++-
 3 files changed, 34 insertions(+), 6 deletions(-)

diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index bbea10737a1..2de8081ea7c 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -1682,17 +1682,18 @@ def MOVDI2PDIrm : PDI<0x6E, MRMSrcMem, (ops VR128:$dst, i32mem:$src),
                       "movd {$src, $dst|$dst, $src}",
                       [(set VR128:$dst,
                         (v4i32 (scalar_to_vector (loadi32 addr:$src))))]>;
+
 // SSE2 instructions with XS prefix
-def MOVQI2PQIrr : I<0x7E, MRMSrcReg, (ops VR128:$dst, VR64:$src),
-                    "movq {$src, $dst|$dst, $src}",
-                    [(set VR128:$dst,
-                      (v2i64 (scalar_to_vector VR64:$src)))]>, XS,
-                  Requires<[HasSSE2]>;
 def MOVQI2PQIrm : I<0x7E, MRMSrcMem, (ops VR128:$dst, i64mem:$src),
                     "movq {$src, $dst|$dst, $src}",
                     [(set VR128:$dst,
                       (v2i64 (scalar_to_vector (loadi64 addr:$src))))]>, XS,
                   Requires<[HasSSE2]>;
+def MOVPQI2QImr : PDI<0xD6, MRMDestMem, (ops i64mem:$dst, VR128:$src),
+                      "movq {$src, $dst|$dst, $src}",
+                      [(store (i64 (vector_extract (v2i64 VR128:$src),
+                                    (iPTR 0))), addr:$dst)]>;
+
 // FIXME: may not be able to eliminate this movss with coalescing the src and
 // dest register classes are different. We really want to write this pattern
 // like this:
diff --git a/lib/Target/X86/X86InstrX86-64.td b/lib/Target/X86/X86InstrX86-64.td
index b4dd1dbe6c9..3042802b778 100644
--- a/lib/Target/X86/X86InstrX86-64.td
+++ b/lib/Target/X86/X86InstrX86-64.td
@@ -59,6 +59,8 @@ class RSSI<bits<8> o, Format F, dag ops, string asm, list<dag> pattern>
       : SSI<o, F, ops, asm, pattern>, REX_W;
 class RSDI<bits<8> o, Format F, dag ops, string asm, list<dag> pattern>
       : SDI<o, F, ops, asm, pattern>, REX_W;
+class RPDI<bits<8> o, Format F, dag ops, string asm, list<dag> pattern>
+      : PDI<o, F, ops, asm, pattern>, REX_W;
 
 //===----------------------------------------------------------------------===//
 // Pattern fragments...
@@ -1076,3 +1078,27 @@ def : Pat<(or (shl GR64:$src1, CL:$amt),
 def : Pat<(store (or (shl (loadi64 addr:$dst), CL:$amt),
                      (srl GR64:$src2, (sub 64, CL:$amt))), addr:$dst),
           (SHLD64mrCL addr:$dst, GR64:$src2)>;
+
+//===----------------------------------------------------------------------===//
+// X86-64 SSE Instructions
+//===----------------------------------------------------------------------===//
+
+// Move instructions...
+
+def MOV64toPQIrr : RPDI<0x6E, MRMSrcReg, (ops VR128:$dst, GR64:$src),
+                        "movq {$src, $dst|$dst, $src}",
+                        [(set VR128:$dst,
+                          (v2i64 (scalar_to_vector GR64:$src)))]>;
+def MOV64toPQIrm : RPDI<0x6E, MRMSrcMem, (ops VR128:$dst, i64mem:$src),
+                        "movq {$src, $dst|$dst, $src}",
+                        [(set VR128:$dst,
+                          (v2i64 (scalar_to_vector (loadi64 addr:$src))))]>;
+
+def MOVPQIto64rr  : RPDI<0x7E, MRMDestReg, (ops GR64:$dst, VR128:$src),
+                         "movd {$src, $dst|$dst, $src}",
+                         [(set GR64:$dst, (vector_extract (v2i64 VR128:$src),
+                                           (iPTR 0)))]>;
+def MOVPQIto64mr  : RPDI<0x7E, MRMDestMem, (ops i64mem:$dst, VR128:$src),
+                         "movd {$src, $dst|$dst, $src}",
+                         [(store (i64 (vector_extract (v2i64 VR128:$src),
+                                       (iPTR 0))), addr:$dst)]>;
diff --git a/lib/Target/X86/X86RegisterInfo.cpp b/lib/Target/X86/X86RegisterInfo.cpp
index 353a159d181..f061aeed7ee 100644
--- a/lib/Target/X86/X86RegisterInfo.cpp
+++ b/lib/Target/X86/X86RegisterInfo.cpp
@@ -509,6 +509,7 @@ MachineInstr* X86RegisterInfo::foldMemoryOperand(MachineInstr *MI,
       { X86::MOVAPDrr,    X86::MOVAPDmr },
       { X86::MOVAPSrr,    X86::MOVAPSmr },
       { X86::MOVPDI2DIrr, X86::MOVPDI2DImr },
+      { X86::MOVPQIto64rr,X86::MOVPQIto64mr },
       { X86::MOVPS2SSrr,  X86::MOVPS2SSmr },
       { X86::MOVSDrr,     X86::MOVSDmr },
       { X86::MOVSSrr,     X86::MOVSSmr },
@@ -606,12 +607,12 @@ MachineInstr* X86RegisterInfo::foldMemoryOperand(MachineInstr *MI,
       { X86::MOV16rr,         X86::MOV16rm },
       { X86::MOV32rr,         X86::MOV32rm },
       { X86::MOV64rr,         X86::MOV64rm },
+      { X86::MOV64toPQIrr,    X86::MOV64toPQIrm },
       { X86::MOV8rr,          X86::MOV8rm },
       { X86::MOVAPDrr,        X86::MOVAPDrm },
       { X86::MOVAPSrr,        X86::MOVAPSrm },
       { X86::MOVDDUPrr,       X86::MOVDDUPrm },
       { X86::MOVDI2PDIrr,     X86::MOVDI2PDIrm },
-      { X86::MOVQI2PQIrr,     X86::MOVQI2PQIrm },
       { X86::MOVSD2PDrr,      X86::MOVSD2PDrm },
       { X86::MOVSDrr,         X86::MOVSDrm },
       { X86::MOVSHDUPrr,      X86::MOVSHDUPrm },