|
From: <sv...@va...> - 2006-04-08 16:16:00
|
Author: sewardj
Date: 2006-04-08 17:15:53 +0100 (Sat, 08 Apr 2006)
New Revision: 1602
Log:
Implement amd64 pmaddwd for SSE2.
Modified:
trunk/priv/guest-amd64/toIR.c
Modified: trunk/priv/guest-amd64/toIR.c
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
--- trunk/priv/guest-amd64/toIR.c 2006-03-20 12:05:42 UTC (rev 1601)
+++ trunk/priv/guest-amd64/toIR.c 2006-04-08 16:15:53 UTC (rev 1602)
@@ -10883,6 +10883,54 @@
goto decode_success;
}
=20
+ /* 66 0F F5 =3D PMADDWD -- Multiply and add packed integers from
+ E(xmm or mem) to G(xmm) */
+ if (have66noF2noF3(pfx) && sz =3D=3D 2=20
+ && insn[0] =3D=3D 0x0F && insn[1] =3D=3D 0xF5) {
+ IRTemp s1V =3D newTemp(Ity_V128);
+ IRTemp s2V =3D newTemp(Ity_V128);
+ IRTemp dV =3D newTemp(Ity_V128);
+ IRTemp s1Hi =3D newTemp(Ity_I64);
+ IRTemp s1Lo =3D newTemp(Ity_I64);
+ IRTemp s2Hi =3D newTemp(Ity_I64);
+ IRTemp s2Lo =3D newTemp(Ity_I64);
+ IRTemp dHi =3D newTemp(Ity_I64);
+ IRTemp dLo =3D newTemp(Ity_I64);
+ modrm =3D insn[2];
+ if (epartIsReg(modrm)) {
+ assign( s1V, getXMMReg(eregOfRexRM(pfx,modrm)) );
+ delta +=3D 2+1;
+ DIP("pmaddwd %s,%s\n", nameXMMReg(eregOfRexRM(pfx,modrm)),
+ nameXMMReg(gregOfRexRM(pfx,modrm)));
+ } else {
+ addr =3D disAMode ( &alen, pfx, delta+2, dis_buf, 0 );
+ assign( s1V, loadLE(Ity_V128, mkexpr(addr)) );
+ delta +=3D 2+alen;
+ DIP("pmaddwd %s,%s\n", dis_buf,
+ nameXMMReg(gregOfRexRM(pfx,modrm)));
+ }
+ assign( s2V, getXMMReg(gregOfRexRM(pfx,modrm)) );
+ assign( s1Hi, unop(Iop_V128HIto64, mkexpr(s1V)) );
+ assign( s1Lo, unop(Iop_V128to64, mkexpr(s1V)) );
+ assign( s2Hi, unop(Iop_V128HIto64, mkexpr(s2V)) );
+ assign( s2Lo, unop(Iop_V128to64, mkexpr(s2V)) );
+ assign( dHi, mkIRExprCCall(
+ Ity_I64, 0/*regparms*/,
+ "amd64g_calculate_mmx_pmaddwd",=20
+ &amd64g_calculate_mmx_pmaddwd,
+ mkIRExprVec_2( mkexpr(s1Hi), mkexpr(s2Hi))
+ ));
+ assign( dLo, mkIRExprCCall(
+ Ity_I64, 0/*regparms*/,
+ "amd64g_calculate_mmx_pmaddwd",=20
+ &amd64g_calculate_mmx_pmaddwd,
+ mkIRExprVec_2( mkexpr(s1Lo), mkexpr(s2Lo))
+ ));
+ assign( dV, binop(Iop_64HLtoV128, mkexpr(dHi), mkexpr(dLo))) ;
+ putXMMReg(gregOfRexRM(pfx,modrm), mkexpr(dV));
+ goto decode_success;
+ }
+
/* 66 0F EE =3D PMAXSW -- 16x8 signed max */
if (have66noF2noF3(pfx) && sz =3D=3D 2=20
&& insn[0] =3D=3D 0x0F && insn[1] =3D=3D 0xEE) {
|