From: Julian Seward <jseward@acm.org>
Date: Sat, 20 Nov 2004 13:17:04 +0000 (+0000)
Subject: Some more MMX insns.
X-Git-Tag: svn/VALGRIND_3_0_1^2~765
X-Git-Url: http://git.ipfire.org/gitweb.cgi?a=commitdiff_plain;h=ab7ff28c2281cbfcfda24a7683fa47e379840c65;p=thirdparty%2Fvalgrind.git

Some more MMX insns.


git-svn-id: svn://svn.valgrind.org/vex/trunk@569
---

diff --git a/VEX/priv/guest-x86/gdefs.h b/VEX/priv/guest-x86/gdefs.h
index 80144bfd9d..4dd73b3bdf 100644
--- a/VEX/priv/guest-x86/gdefs.h
+++ b/VEX/priv/guest-x86/gdefs.h
@@ -114,6 +114,16 @@ extern ULong calculate_vqsub8Ux8  ( ULong, ULong );
 extern ULong calculate_vmulhi16x4 ( ULong, ULong );
 extern ULong calculate_vmullo16x4 ( ULong, ULong );
 
+extern ULong calculate_pmaddwd ( ULong, ULong );
+
+extern ULong calculate_vcmpeq32x2  ( ULong, ULong );
+extern ULong calculate_vcmpeq16x4  ( ULong, ULong );
+extern ULong calculate_vcmpeq8x8   ( ULong, ULong );
+extern ULong calculate_vcmpge32Sx2 ( ULong, ULong );
+extern ULong calculate_vcmpge16Sx4 ( ULong, ULong );
+extern ULong calculate_vcmpge8Sx8  ( ULong, ULong );
+
+
 /* --- DIRTY HELPERS --- */
 
 extern ULong loadF80le  ( UInt );
diff --git a/VEX/priv/guest-x86/ghelpers.c b/VEX/priv/guest-x86/ghelpers.c
index 95ff692314..7d9f817d5c 100644
--- a/VEX/priv/guest-x86/ghelpers.c
+++ b/VEX/priv/guest-x86/ghelpers.c
@@ -1785,16 +1785,47 @@ static inline Short mullo16S ( Short xx, Short yy )
    return (Short)t;
 }
 
+static inline UInt cmpeq32 ( UInt xx, UInt yy )
+{
+   return xx==yy ? 0xFFFFFFFF : 0;
+}
+
+static inline UShort cmpeq16 ( UShort xx, UShort yy )
+{
+   return xx==yy ? 0xFFFF : 0;
+}
+
+static inline UChar cmpeq8 ( UChar xx, UChar yy )
+{
+   return xx==yy ? 0xFF : 0;
+}
+
+static inline UInt cmpge32S ( Int xx, Int yy )
+{
+   return xx>yy ? 0xFFFFFFFF : 0;
+}
+
+static inline UShort cmpge16S ( Short xx, Short yy )
+{
+   return xx>yy ? 0xFFFF : 0;
+}
+
+static inline UChar cmpge8S ( SChar xx, SChar yy )
+{
+   return xx>yy ? 0xFF : 0;
+}
+
+
 
 /* Actually do something interesting! */
 
 /* Normal adds. */
 
-ULong calculate_vadd32x2 ( ULong xxL, ULong yyL )
+ULong calculate_vadd32x2 ( ULong xx, ULong yy )
 {
    return mk32x2(
-             sel32x2_1(xxL) + sel32x2_1(yyL),
-             sel32x2_0(xxL) + sel32x2_0(yyL)
+             sel32x2_1(xx) + sel32x2_1(yy),
+             sel32x2_0(xx) + sel32x2_0(yy)
           );
 }
 
@@ -1874,11 +1905,11 @@ ULong calculate_vqadd8Ux8 ( ULong xx, ULong yy )
 
 /* Normal subtracts. */
 
-ULong calculate_vsub32x2 ( ULong xxL, ULong yyL )
+ULong calculate_vsub32x2 ( ULong xx, ULong yy )
 {
    return mk32x2(
-             sel32x2_1(xxL) - sel32x2_1(yyL),
-             sel32x2_0(xxL) - sel32x2_0(yyL)
+             sel32x2_1(xx) - sel32x2_1(yy),
+             sel32x2_0(xx) - sel32x2_0(yy)
           );
 }
 
@@ -1978,6 +2009,84 @@ ULong calculate_vmullo16x4 ( ULong xx, ULong yy )
           );
 }
 
+ULong calculate_pmaddwd ( ULong xx, ULong yy )
+{
+   return
+      mk32x2( 
+         (((Int)(Short)sel16x4_3(xx)) * ((Int)(Short)sel16x4_3(yy)))
+            + (((Int)(Short)sel16x4_2(xx)) * ((Int)(Short)sel16x4_2(yy))),
+         (((Int)(Short)sel16x4_1(xx)) * ((Int)(Short)sel16x4_1(yy)))
+            + (((Int)(Short)sel16x4_0(xx)) * ((Int)(Short)sel16x4_0(yy)))
+      );
+}
+
+/* Comparison. */
+
+ULong calculate_vcmpeq32x2 ( ULong xx, ULong yy )
+{
+   return mk32x2(
+             cmpeq32( sel32x2_1(xx), sel32x2_1(yy) ),
+             cmpeq32( sel32x2_0(xx), sel32x2_0(yy) )
+          );
+}
+
+ULong calculate_vcmpeq16x4 ( ULong xx, ULong yy )
+{
+   return mk16x4(
+             cmpeq16( sel16x4_3(xx), sel16x4_3(yy) ),
+             cmpeq16( sel16x4_2(xx), sel16x4_2(yy) ),
+             cmpeq16( sel16x4_1(xx), sel16x4_1(yy) ),
+             cmpeq16( sel16x4_0(xx), sel16x4_0(yy) )
+          );
+}
+
+ULong calculate_vcmpeq8x8 ( ULong xx, ULong yy )
+{
+   return mk8x8(
+             cmpeq8( sel8x8_7(xx), sel8x8_7(yy) ),
+             cmpeq8( sel8x8_6(xx), sel8x8_6(yy) ),
+             cmpeq8( sel8x8_5(xx), sel8x8_5(yy) ),
+             cmpeq8( sel8x8_4(xx), sel8x8_4(yy) ),
+             cmpeq8( sel8x8_3(xx), sel8x8_3(yy) ),
+             cmpeq8( sel8x8_2(xx), sel8x8_2(yy) ),
+             cmpeq8( sel8x8_1(xx), sel8x8_1(yy) ),
+             cmpeq8( sel8x8_0(xx), sel8x8_0(yy) )
+          );
+}
+
+ULong calculate_vcmpge32Sx2 ( ULong xx, ULong yy )
+{
+   return mk32x2(
+             cmpge32S( sel32x2_1(xx), sel32x2_1(yy) ),
+             cmpge32S( sel32x2_0(xx), sel32x2_0(yy) )
+          );
+}
+
+ULong calculate_vcmpge16Sx4 ( ULong xx, ULong yy )
+{
+   return mk16x4(
+             cmpge16S( sel16x4_3(xx), sel16x4_3(yy) ),
+             cmpge16S( sel16x4_2(xx), sel16x4_2(yy) ),
+             cmpge16S( sel16x4_1(xx), sel16x4_1(yy) ),
+             cmpge16S( sel16x4_0(xx), sel16x4_0(yy) )
+          );
+}
+
+ULong calculate_vcmpge8Sx8 ( ULong xx, ULong yy )
+{
+   return mk8x8(
+             cmpge8S( sel8x8_7(xx), sel8x8_7(yy) ),
+             cmpge8S( sel8x8_6(xx), sel8x8_6(yy) ),
+             cmpge8S( sel8x8_5(xx), sel8x8_5(yy) ),
+             cmpge8S( sel8x8_4(xx), sel8x8_4(yy) ),
+             cmpge8S( sel8x8_3(xx), sel8x8_3(yy) ),
+             cmpge8S( sel8x8_2(xx), sel8x8_2(yy) ),
+             cmpge8S( sel8x8_1(xx), sel8x8_1(yy) ),
+             cmpge8S( sel8x8_0(xx), sel8x8_0(yy) )
+          );
+}
+
+
 
 /*-----------------------------------------------------------*/
 /*--- Describing the x86 guest state, for the benefit     ---*/
diff --git a/VEX/priv/guest-x86/toIR.c b/VEX/priv/guest-x86/toIR.c
index 1e506a816d..bb32060144 100644
--- a/VEX/priv/guest-x86/toIR.c
+++ b/VEX/priv/guest-x86/toIR.c
@@ -4667,6 +4667,15 @@ UInt dis_MMXop_regmem_to_reg ( UChar sorb,
 
       case 0xE5: XXX(calculate_vmulhi16x4); break;
       case 0xD5: XXX(calculate_vmullo16x4); break;
+      case 0xF5: XXX(calculate_pmaddwd); break;
+
+      case 0x74: XXX(calculate_vcmpeq8x8); break;
+      case 0x75: XXX(calculate_vcmpeq16x4); break;
+      case 0x76: XXX(calculate_vcmpeq32x2); break;
+
+      case 0x64: XXX(calculate_vcmpge8Sx8); break;
+      case 0x65: XXX(calculate_vcmpge16Sx4); break;
+      case 0x66: XXX(calculate_vcmpge32Sx2); break;
 
       default: 
          vex_printf("\n0x%x\n", (Int)opc);
@@ -4769,38 +4778,40 @@ vassert(0);
          }
          break;
 
-      case 0xFC: case 0xFD: case 0xFE: 
-         /* PADDgg (src)mmxreg-or-mem, (dst)mmxreg */
+      case 0xFC: 
+      case 0xFD: 
+      case 0xFE: /* PADDgg (src)mmxreg-or-mem, (dst)mmxreg */
          vassert(sz == 4);
          delta = dis_MMXop_regmem_to_reg ( sorb, delta, opc, "padd", True );
          break;
 
-      case 0xEC: case 0xED:
-         /* PADDSgg (src)mmxreg-or-mem, (dst)mmxreg */
+      case 0xEC: 
+      case 0xED: /* PADDSgg (src)mmxreg-or-mem, (dst)mmxreg */
          vassert(sz == 4);
          delta = dis_MMXop_regmem_to_reg ( sorb, delta, opc, "padds", True );
          break;
 
-      case 0xDC: case 0xDD:
-         /* PADDUSgg (src)mmxreg-or-mem, (dst)mmxreg */
+      case 0xDC: 
+      case 0xDD: /* PADDUSgg (src)mmxreg-or-mem, (dst)mmxreg */
          vassert(sz == 4);
          delta = dis_MMXop_regmem_to_reg ( sorb, delta, opc, "paddus", True );
          break;
 
-      case 0xF8: case 0xF9: case 0xFA:
-         /* PSUBgg (src)mmxreg-or-mem, (dst)mmxreg */
+      case 0xF8: 
+      case 0xF9: 
+      case 0xFA: /* PSUBgg (src)mmxreg-or-mem, (dst)mmxreg */
          vassert(sz == 4);
          delta = dis_MMXop_regmem_to_reg ( sorb, delta, opc, "psub", True );
          break;
 
-      case 0xE8: case 0xE9:
-         /* PSUBSgg (src)mmxreg-or-mem, (dst)mmxreg */
+      case 0xE8: 
+      case 0xE9: /* PSUBSgg (src)mmxreg-or-mem, (dst)mmxreg */
          vassert(sz == 4);
          delta = dis_MMXop_regmem_to_reg ( sorb, delta, opc, "psubs", True );
          break;
 
-      case 0xD8: case 0xD9:
-         /* PSUBUSgg (src)mmxreg-or-mem, (dst)mmxreg */
+      case 0xD8: 
+      case 0xD9: /* PSUBUSgg (src)mmxreg-or-mem, (dst)mmxreg */
          vassert(sz == 4);
          delta = dis_MMXop_regmem_to_reg ( sorb, delta, opc, "psubus", True );
          break;
@@ -4815,6 +4826,25 @@ vassert(0);
          delta = dis_MMXop_regmem_to_reg ( sorb, delta, opc, "pmullw", False );
          break;
 
+      case 0xF5: /* PMADDWD (src)mmxreg-or-mem, (dst)mmxreg */
+         vassert(sz == 4);
+         delta = dis_MMXop_regmem_to_reg ( sorb, delta, opc, "pmaddwd", False );
+         break;
+
+      case 0x74: 
+      case 0x75: 
+      case 0x76: /* PCMPEQgg (src)mmxreg-or-mem, (dst)mmxreg */
+         vassert(sz == 4);
+         delta = dis_MMXop_regmem_to_reg ( sorb, delta, opc, "pcmpeq", True );
+         break;
+
+      case 0x64: 
+      case 0x65: 
+      case 0x66: /* PCMPGTgg (src)mmxreg-or-mem, (dst)mmxreg */
+         vassert(sz == 4);
+         delta = dis_MMXop_regmem_to_reg ( sorb, delta, opc, "pcmpgt", True );
+         break;
+
       default:
          *decode_ok = False;
          return delta; /* ignored */
@@ -9448,6 +9478,16 @@ static DisResult disInstr ( /*IN*/  Bool    resteerOK,
       case 0xE5: /* PMULHW (src)mmxreg-or-mem, (dst)mmxreg */
       case 0xD5: /* PMULLW (src)mmxreg-or-mem, (dst)mmxreg */
 
+      case 0xF5: /* PMADDWD (src)mmxreg-or-mem, (dst)mmxreg */
+
+      case 0x74: 
+      case 0x75: 
+      case 0x76: /* PCMPEQgg (src)mmxreg-or-mem, (dst)mmxreg */
+
+      case 0x64: 
+      case 0x65: 
+      case 0x66: /* PCMPGTgg (src)mmxreg-or-mem, (dst)mmxreg */
+
       case 0x6F: /* MOVQ (src)mmxreg-or-mem, (dst)mmxreg */
       {
          UInt delta0    = delta-1;