commit 6c07dca0af0527226ec9784fbb703526887f09c0
Author: sewardj <sewardj@8f6e269a-dfd6-0310-a8e1-e2731360e62c>
Date:   Tue Mar 26 10:27:39 2013 +0000

    Implement SSE4 MOVNTDQA insn.  Fixes #316503.
    (Patrick J. LoPresti, lopresti@gmail.com)
    
    
    git-svn-id: svn://svn.valgrind.org/vex/trunk@2700 8f6e269a-dfd6-0310-a8e1-e2731360e62c

diff --git a/VEX/priv/guest_amd64_toIR.c b/VEX/priv/guest_amd64_toIR.c
index 1034971..7e98e76 100644
--- a/VEX/priv/guest_amd64_toIR.c
+++ b/VEX/priv/guest_amd64_toIR.c
@@ -16378,6 +16378,25 @@ Long dis_ESC_0F38__SSE4 ( Bool* decode_OK,
       }
       break;
 
+   case 0x2A:
+      /* 66 0F 38 2A = MOVNTDQA
+         "non-temporal" "streaming" load
+         Handle like MOVDQA but only memory operand is allowed */
+      if (have66noF2noF3(pfx) && sz == 2) {
+         modrm = getUChar(delta);
+         if (!epartIsReg(modrm)) {
+            addr = disAMode ( &alen, vbi, pfx, delta, dis_buf, 0 );
+            gen_SEGV_if_not_16_aligned( addr );
+            putXMMReg( gregOfRexRM(pfx,modrm),
+                       loadLE(Ity_V128, mkexpr(addr)) );
+            DIP("movntdqa %s,%s\n", dis_buf,
+                                    nameXMMReg(gregOfRexRM(pfx,modrm)));
+            delta += alen;
+            goto decode_success;
+         }
+      }
+      break;
+
    case 0x2B:
       /* 66 0f 38 2B /r = PACKUSDW xmm1, xmm2/m128
          2x 32x4 S->U saturating narrow from xmm2/m128 to xmm1 */