From d19760a40c60a48f3b7619183fc6fef33c3c2d10 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Jan=20Sj=C3=B6din?= <jan_sjodin@yahoo.com>
Date: Thu, 8 Dec 2011 14:43:19 +0000
Subject: [PATCH] Src2 and src3 were accidentally swapped for the FMA4 rr
 patterns. Undo this and fix the encoding.

llvm-svn: 146151
---
 .../X86/MCTargetDesc/X86MCCodeEmitter.cpp     | 19 ++++++++-----------
 llvm/lib/Target/X86/X86InstrFMA.td            |  6 +++---
 2 files changed, 11 insertions(+), 14 deletions(-)

diff --git a/llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp b/llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp
index dbd52078d801..0c00f11be26d 100644
--- a/llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp
+++ b/llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp
@@ -973,13 +973,14 @@ EncodeInstruction(const MCInst &MI, raw_ostream &OS,
     if (HasVEX_4V) // Skip 1st src (which is encoded in VEX_VVVV)
       SrcRegNum++;
 
-    // GAS sets the XOP_W even with register operands, we want to match this.
-    // XOP_W is ignored, so there is no swapping of the operands
-    XOP_W_I8IMMOperand = 3;
+    if(HasXOP_W) // Skip 2nd src (which is encoded in I8IMM)
+      SrcRegNum++;
 
     EmitRegModRMByte(MI.getOperand(SrcRegNum),
                      GetX86RegNum(MI.getOperand(CurOp)), CurByte, OS);
-    CurOp = SrcRegNum + 1;
+
+    // 2 operands skipped with HasXOP_W, comensate accordingly
+    CurOp = HasXOP_W ? SrcRegNum : SrcRegNum + 1;
     if (HasVEX_4VOp3)
       ++CurOp;
     break;
@@ -998,13 +999,9 @@ EncodeInstruction(const MCInst &MI, raw_ostream &OS,
 
     EmitMemModRMByte(MI, FirstMemOp, GetX86RegNum(MI.getOperand(CurOp)),
                      TSFlags, CurByte, OS, Fixups);
-    if(HasXOP_W) {
-      CurOp = NumOps - 1; // We have consumed all except one operand (third)
-    } else {
-      CurOp += AddrOperands + 1;
-      if (HasVEX_4VOp3)
-        ++CurOp;
-    }
+    CurOp += AddrOperands + 1;
+    if (HasVEX_4VOp3)
+      ++CurOp;
     break;
   }
 
diff --git a/llvm/lib/Target/X86/X86InstrFMA.td b/llvm/lib/Target/X86/X86InstrFMA.td
index 015b01ecffd7..f4430887a1ff 100644
--- a/llvm/lib/Target/X86/X86InstrFMA.td
+++ b/llvm/lib/Target/X86/X86InstrFMA.td
@@ -68,7 +68,7 @@ multiclass fma4s<bits<8> opc, string OpcodeStr> {
   def rr : FMA4<opc, MRMSrcReg, (outs VR128:$dst),
            (ins VR128:$src1, VR128:$src2, VR128:$src3),
            !strconcat(OpcodeStr,
-           "\t{$src2, $src3, $src1, $dst|$dst, $src1, $src3, $src2}"),
+           "\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
            []>, XOP_W;
   def rm : FMA4<opc, MRMSrcMem, (outs VR128:$dst),
            (ins VR128:$src1, VR128:$src2, f128mem:$src3),
@@ -87,7 +87,7 @@ multiclass fma4p<bits<8> opc, string OpcodeStr> {
   def rr : FMA4<opc, MRMSrcReg, (outs VR128:$dst),
            (ins VR128:$src1, VR128:$src2, VR128:$src3),
            !strconcat(OpcodeStr,
-           "\t{$src2, $src3, $src1, $dst|$dst, $src1, $src3, $src2}"),
+           "\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
            []>, XOP_W;
   def rm : FMA4<opc, MRMSrcMem, (outs VR128:$dst),
            (ins VR128:$src1, VR128:$src2, f128mem:$src3),
@@ -102,7 +102,7 @@ multiclass fma4p<bits<8> opc, string OpcodeStr> {
   def rrY : FMA4<opc, MRMSrcReg, (outs VR256:$dst),
            (ins VR256:$src1, VR256:$src2, VR256:$src3),
            !strconcat(OpcodeStr,
-           "\t{$src2, $src3, $src1, $dst|$dst, $src1, $src3, $src2}"),
+           "\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
            []>, XOP_W;
   def rmY : FMA4<opc, MRMSrcMem, (outs VR256:$dst),
            (ins VR256:$src1, VR256:$src2, f256mem:$src3),