converted SSE/MMX ops to TCG

git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@4441 c046a42c-6fe2-441c-8c8c-71466251a162

6 files changed, 715 insertions, 671 deletions

diff --git a/target-i386/helper.c b/target-i386/helper.c
index c93f10fa42..30a6eccb83 100644
--- a/target-i386/helper.c
+++ b/target-i386/helper.c
@@ -4462,3 +4462,32 @@ void vmexit(uint64_t exit_code, uint64_t exit_info_1)
 }
 
 #endif
+
+/* MMX/SSE */
+/* XXX: optimize by storing fptt and fptags in the static cpu state */
+void helper_enter_mmx(void)
+{
+    env->fpstt = 0;
+    *(uint32_t *)(env->fptags) = 0;
+    *(uint32_t *)(env->fptags + 4) = 0;
+}
+
+void helper_emms(void)
+{
+    /* set to empty state */
+    *(uint32_t *)(env->fptags) = 0x01010101;
+    *(uint32_t *)(env->fptags + 4) = 0x01010101;
+}
+
+/* XXX: suppress */
+void helper_movq(uint64_t *d, uint64_t *s)
+{
+    *d = *s;
+}
+
+#define SHIFT 0
+#include "ops_sse.h"
+
+#define SHIFT 1
+#include "ops_sse.h"
+
diff --git a/target-i386/helper.h b/target-i386/helper.h
index 1b247dbf5b..bc06fa3a8b 100644
--- a/target-i386/helper.h
+++ b/target-i386/helper.h
@@ -2,3 +2,12 @@
 
 void TCG_HELPER_PROTO helper_divl_EAX_T0(target_ulong t0);
 void TCG_HELPER_PROTO helper_idivl_EAX_T0(target_ulong t0);
+void TCG_HELPER_PROTO helper_enter_mmx(void);
+void TCG_HELPER_PROTO helper_emms(void);
+void TCG_HELPER_PROTO helper_movq(uint64_t *d, uint64_t *s);
+
+#define SHIFT 0
+#include "ops_sse_header.h"
+#define SHIFT 1
+#include "ops_sse_header.h"
+
diff --git a/target-i386/op.c b/target-i386/op.c
index 64b64a4fe2..c87a600332 100644
--- a/target-i386/op.c
+++ b/target-i386/op.c
@@ -2144,44 +2144,9 @@ void OPPROTO op_unlock(void)
 }
 
 /* SSE support */
-static inline void memcpy16(void *d, void *s)
+void OPPROTO op_com_dummy(void)
 {
-    ((uint32_t *)d)[0] = ((uint32_t *)s)[0];
-    ((uint32_t *)d)[1] = ((uint32_t *)s)[1];
-    ((uint32_t *)d)[2] = ((uint32_t *)s)[2];
-    ((uint32_t *)d)[3] = ((uint32_t *)s)[3];
-}
-
-void OPPROTO op_movo(void)
-{
-    /* XXX: badly generated code */
-    XMMReg *d, *s;
-    d = (XMMReg *)((char *)env + PARAM1);
-    s = (XMMReg *)((char *)env + PARAM2);
-    memcpy16(d, s);
-}
-
-void OPPROTO op_movq(void)
-{
-    uint64_t *d, *s;
-    d = (uint64_t *)((char *)env + PARAM1);
-    s = (uint64_t *)((char *)env + PARAM2);
-    *d = *s;
-}
-
-void OPPROTO op_movl(void)
-{
-    uint32_t *d, *s;
-    d = (uint32_t *)((char *)env + PARAM1);
-    s = (uint32_t *)((char *)env + PARAM2);
-    *d = *s;
-}
-
-void OPPROTO op_movq_env_0(void)
-{
-    uint64_t *d;
-    d = (uint64_t *)((char *)env + PARAM1);
-    *d = 0;
+    T0 = 0;
 }
 
 void OPPROTO op_fxsave_A0(void)
@@ -2194,27 +2159,6 @@ void OPPROTO op_fxrstor_A0(void)
     helper_fxrstor(A0, PARAM1);
 }
 
-/* XXX: optimize by storing fptt and fptags in the static cpu state */
-void OPPROTO op_enter_mmx(void)
-{
-    env->fpstt = 0;
-    *(uint32_t *)(env->fptags) = 0;
-    *(uint32_t *)(env->fptags + 4) = 0;
-}
-
-void OPPROTO op_emms(void)
-{
-    /* set to empty state */
-    *(uint32_t *)(env->fptags) = 0x01010101;
-    *(uint32_t *)(env->fptags + 4) = 0x01010101;
-}
-
-#define SHIFT 0
-#include "ops_sse.h"
-
-#define SHIFT 1
-#include "ops_sse.h"
-
 /* Secure Virtual Machine ops */
 
 void OPPROTO op_vmrun(void)
diff --git a/target-i386/ops_sse.h b/target-i386/ops_sse.h
index 6610e750f8..010cfea410 100644
--- a/target-i386/ops_sse.h
+++ b/target-i386/ops_sse.h
@@ -35,14 +35,10 @@
 #define SUFFIX _xmm
 #endif
 
-void OPPROTO glue(op_psrlw, SUFFIX)(void)
+void glue(helper_psrlw, SUFFIX)(Reg *d, Reg *s)
 {
-    Reg *d, *s;
     int shift;
 
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-
     if (s->Q(0) > 15) {
         d->Q(0) = 0;
 #if SHIFT == 1
@@ -64,14 +60,10 @@ void OPPROTO glue(op_psrlw, SUFFIX)(void)
     FORCE_RET();
 }
 
-void OPPROTO glue(op_psraw, SUFFIX)(void)
+void glue(helper_psraw, SUFFIX)(Reg *d, Reg *s)
 {
-    Reg *d, *s;
     int shift;
 
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-
     if (s->Q(0) > 15) {
         shift = 15;
     } else {
@@ -89,14 +81,10 @@ void OPPROTO glue(op_psraw, SUFFIX)(void)
 #endif
 }
 
-void OPPROTO glue(op_psllw, SUFFIX)(void)
+void glue(helper_psllw, SUFFIX)(Reg *d, Reg *s)
 {
-    Reg *d, *s;
     int shift;
 
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-
     if (s->Q(0) > 15) {
         d->Q(0) = 0;
 #if SHIFT == 1
@@ -118,14 +106,10 @@ void OPPROTO glue(op_psllw, SUFFIX)(void)
     FORCE_RET();
 }
 
-void OPPROTO glue(op_psrld, SUFFIX)(void)
+void glue(helper_psrld, SUFFIX)(Reg *d, Reg *s)
 {
-    Reg *d, *s;
     int shift;
 
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-
     if (s->Q(0) > 31) {
         d->Q(0) = 0;
 #if SHIFT == 1
@@ -143,14 +127,10 @@ void OPPROTO glue(op_psrld, SUFFIX)(void)
     FORCE_RET();
 }
 
-void OPPROTO glue(op_psrad, SUFFIX)(void)
+void glue(helper_psrad, SUFFIX)(Reg *d, Reg *s)
 {
-    Reg *d, *s;
     int shift;
 
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-
     if (s->Q(0) > 31) {
         shift = 31;
     } else {
@@ -164,14 +144,10 @@ void OPPROTO glue(op_psrad, SUFFIX)(void)
 #endif
 }
 
-void OPPROTO glue(op_pslld, SUFFIX)(void)
+void glue(helper_pslld, SUFFIX)(Reg *d, Reg *s)
 {
-    Reg *d, *s;
     int shift;
 
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-
     if (s->Q(0) > 31) {
         d->Q(0) = 0;
 #if SHIFT == 1
@@ -189,14 +165,10 @@ void OPPROTO glue(op_pslld, SUFFIX)(void)
     FORCE_RET();
 }
 
-void OPPROTO glue(op_psrlq, SUFFIX)(void)
+void glue(helper_psrlq, SUFFIX)(Reg *d, Reg *s)
 {
-    Reg *d, *s;
     int shift;
 
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-
     if (s->Q(0) > 63) {
         d->Q(0) = 0;
 #if SHIFT == 1
@@ -212,14 +184,10 @@ void OPPROTO glue(op_psrlq, SUFFIX)(void)
     FORCE_RET();
 }
 
-void OPPROTO glue(op_psllq, SUFFIX)(void)
+void glue(helper_psllq, SUFFIX)(Reg *d, Reg *s)
 {
-    Reg *d, *s;
     int shift;
 
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-
     if (s->Q(0) > 63) {
         d->Q(0) = 0;
 #if SHIFT == 1
@@ -236,13 +204,10 @@ void OPPROTO glue(op_psllq, SUFFIX)(void)
 }
 
 #if SHIFT == 1
-void OPPROTO glue(op_psrldq, SUFFIX)(void)
+void glue(helper_psrldq, SUFFIX)(Reg *d, Reg *s)
 {
-    Reg *d, *s;
     int shift, i;
 
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
     shift = s->L(0);
     if (shift > 16)
         shift = 16;
@@ -253,13 +218,10 @@ void OPPROTO glue(op_psrldq, SUFFIX)(void)
     FORCE_RET();
 }
 
-void OPPROTO glue(op_pslldq, SUFFIX)(void)
+void glue(helper_pslldq, SUFFIX)(Reg *d, Reg *s)
 {
-    Reg *d, *s;
     int shift, i;
 
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
     shift = s->L(0);
     if (shift > 16)
         shift = 16;
@@ -271,12 +233,9 @@ void OPPROTO glue(op_pslldq, SUFFIX)(void)
 }
 #endif
 
-#define SSE_OP_B(name, F)\
-void OPPROTO glue(name, SUFFIX) (void)\
+#define SSE_HELPER_B(name, F)\
+void glue(name, SUFFIX) (Reg *d, Reg *s)\
 {\
-    Reg *d, *s;\
-    d = (Reg *)((char *)env + PARAM1);\
-    s = (Reg *)((char *)env + PARAM2);\
     d->B(0) = F(d->B(0), s->B(0));\
     d->B(1) = F(d->B(1), s->B(1));\
     d->B(2) = F(d->B(2), s->B(2));\
@@ -297,12 +256,9 @@ void OPPROTO glue(name, SUFFIX) (void)\
     )\
 }
 
-#define SSE_OP_W(name, F)\
-void OPPROTO glue(name, SUFFIX) (void)\
+#define SSE_HELPER_W(name, F)\
+void glue(name, SUFFIX) (Reg *d, Reg *s)\
 {\
-    Reg *d, *s;\
-    d = (Reg *)((char *)env + PARAM1);\
-    s = (Reg *)((char *)env + PARAM2);\
     d->W(0) = F(d->W(0), s->W(0));\
     d->W(1) = F(d->W(1), s->W(1));\
     d->W(2) = F(d->W(2), s->W(2));\
@@ -315,12 +271,9 @@ void OPPROTO glue(name, SUFFIX) (void)\
     )\
 }
 
-#define SSE_OP_L(name, F)\
-void OPPROTO glue(name, SUFFIX) (void)\
+#define SSE_HELPER_L(name, F)\
+void glue(name, SUFFIX) (Reg *d, Reg *s)\
 {\
-    Reg *d, *s;\
-    d = (Reg *)((char *)env + PARAM1);\
-    s = (Reg *)((char *)env + PARAM2);\
     d->L(0) = F(d->L(0), s->L(0));\
     d->L(1) = F(d->L(1), s->L(1));\
     XMM_ONLY(\
@@ -329,12 +282,9 @@ void OPPROTO glue(name, SUFFIX) (void)\
     )\
 }
 
-#define SSE_OP_Q(name, F)\
-void OPPROTO glue(name, SUFFIX) (void)\
+#define SSE_HELPER_Q(name, F)\
+void glue(name, SUFFIX) (Reg *d, Reg *s)\
 {\
-    Reg *d, *s;\
-    d = (Reg *)((char *)env + PARAM1);\
-    s = (Reg *)((char *)env + PARAM2);\
     d->Q(0) = F(d->Q(0), s->Q(0));\
     XMM_ONLY(\
     d->Q(1) = F(d->Q(1), s->Q(1));\
@@ -416,73 +366,66 @@ static inline int satsw(int x)
 #define FAVG(a, b) ((a) + (b) + 1) >> 1
 #endif
 
-SSE_OP_B(op_paddb, FADD)
-SSE_OP_W(op_paddw, FADD)
-SSE_OP_L(op_paddl, FADD)
-SSE_OP_Q(op_paddq, FADD)
+SSE_HELPER_B(helper_paddb, FADD)
+SSE_HELPER_W(helper_paddw, FADD)
+SSE_HELPER_L(helper_paddl, FADD)
+SSE_HELPER_Q(helper_paddq, FADD)
 
-SSE_OP_B(op_psubb, FSUB)
-SSE_OP_W(op_psubw, FSUB)
-SSE_OP_L(op_psubl, FSUB)
-SSE_OP_Q(op_psubq, FSUB)
+SSE_HELPER_B(helper_psubb, FSUB)
+SSE_HELPER_W(helper_psubw, FSUB)
+SSE_HELPER_L(helper_psubl, FSUB)
+SSE_HELPER_Q(helper_psubq, FSUB)
 
-SSE_OP_B(op_paddusb, FADDUB)
-SSE_OP_B(op_paddsb, FADDSB)
-SSE_OP_B(op_psubusb, FSUBUB)
-SSE_OP_B(op_psubsb, FSUBSB)
+SSE_HELPER_B(helper_paddusb, FADDUB)
+SSE_HELPER_B(helper_paddsb, FADDSB)
+SSE_HELPER_B(helper_psubusb, FSUBUB)
+SSE_HELPER_B(helper_psubsb, FSUBSB)
 
-SSE_OP_W(op_paddusw, FADDUW)
-SSE_OP_W(op_paddsw, FADDSW)
-SSE_OP_W(op_psubusw, FSUBUW)
-SSE_OP_W(op_psubsw, FSUBSW)
+SSE_HELPER_W(helper_paddusw, FADDUW)
+SSE_HELPER_W(helper_paddsw, FADDSW)
+SSE_HELPER_W(helper_psubusw, FSUBUW)
+SSE_HELPER_W(helper_psubsw, FSUBSW)
 
-SSE_OP_B(op_pminub, FMINUB)
-SSE_OP_B(op_pmaxub, FMAXUB)
+SSE_HELPER_B(helper_pminub, FMINUB)
+SSE_HELPER_B(helper_pmaxub, FMAXUB)
 
-SSE_OP_W(op_pminsw, FMINSW)
-SSE_OP_W(op_pmaxsw, FMAXSW)
+SSE_HELPER_W(helper_pminsw, FMINSW)
+SSE_HELPER_W(helper_pmaxsw, FMAXSW)
 
-SSE_OP_Q(op_pand, FAND)
-SSE_OP_Q(op_pandn, FANDN)
-SSE_OP_Q(op_por, FOR)
-SSE_OP_Q(op_pxor, FXOR)
+SSE_HELPER_Q(helper_pand, FAND)
+SSE_HELPER_Q(helper_pandn, FANDN)
+SSE_HELPER_Q(helper_por, FOR)
+SSE_HELPER_Q(helper_pxor, FXOR)
 
-SSE_OP_B(op_pcmpgtb, FCMPGTB)
-SSE_OP_W(op_pcmpgtw, FCMPGTW)
-SSE_OP_L(op_pcmpgtl, FCMPGTL)
+SSE_HELPER_B(helper_pcmpgtb, FCMPGTB)
+SSE_HELPER_W(helper_pcmpgtw, FCMPGTW)
+SSE_HELPER_L(helper_pcmpgtl, FCMPGTL)
 
-SSE_OP_B(op_pcmpeqb, FCMPEQ)
-SSE_OP_W(op_pcmpeqw, FCMPEQ)
-SSE_OP_L(op_pcmpeql, FCMPEQ)
+SSE_HELPER_B(helper_pcmpeqb, FCMPEQ)
+SSE_HELPER_W(helper_pcmpeqw, FCMPEQ)
+SSE_HELPER_L(helper_pcmpeql, FCMPEQ)
 
-SSE_OP_W(op_pmullw, FMULLW)
+SSE_HELPER_W(helper_pmullw, FMULLW)
 #if SHIFT == 0
-SSE_OP_W(op_pmulhrw, FMULHRW)
+SSE_HELPER_W(helper_pmulhrw, FMULHRW)
 #endif
-SSE_OP_W(op_pmulhuw, FMULHUW)
-SSE_OP_W(op_pmulhw, FMULHW)
+SSE_HELPER_W(helper_pmulhuw, FMULHUW)
+SSE_HELPER_W(helper_pmulhw, FMULHW)
 
-SSE_OP_B(op_pavgb, FAVG)
-SSE_OP_W(op_pavgw, FAVG)
+SSE_HELPER_B(helper_pavgb, FAVG)
+SSE_HELPER_W(helper_pavgw, FAVG)
 
-void OPPROTO glue(op_pmuludq, SUFFIX) (void)
+void glue(helper_pmuludq, SUFFIX) (Reg *d, Reg *s)
 {
-    Reg *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-
     d->Q(0) = (uint64_t)s->L(0) * (uint64_t)d->L(0);
 #if SHIFT == 1
     d->Q(1) = (uint64_t)s->L(2) * (uint64_t)d->L(2);
 #endif
 }
 
-void OPPROTO glue(op_pmaddwd, SUFFIX) (void)
+void glue(helper_pmaddwd, SUFFIX) (Reg *d, Reg *s)
 {
     int i;
-    Reg *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
 
     for(i = 0; i < (2 << SHIFT); i++) {
         d->L(i) = (int16_t)s->W(2*i) * (int16_t)d->W(2*i) +
@@ -500,12 +443,9 @@ static inline int abs1(int a)
         return a;
 }
 #endif
-void OPPROTO glue(op_psadbw, SUFFIX) (void)
+void glue(helper_psadbw, SUFFIX) (Reg *d, Reg *s)
 {
     unsigned int val;
-    Reg *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
 
     val = 0;
     val += abs1(d->B(0) - s->B(0));
@@ -531,12 +471,9 @@ void OPPROTO glue(op_psadbw, SUFFIX) (void)
 #endif
 }
 
-void OPPROTO glue(op_maskmov, SUFFIX) (void)
+void glue(helper_maskmov, SUFFIX) (Reg *d, Reg *s)
 {
     int i;
-    Reg *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
     for(i = 0; i < (8 << SHIFT); i++) {
         if (s->B(i) & 0x80)
             stb(A0 + i, d->B(i));
@@ -544,51 +481,29 @@ void OPPROTO glue(op_maskmov, SUFFIX) (void)
     FORCE_RET();
 }
 
-void OPPROTO glue(op_movl_mm_T0, SUFFIX) (void)
+void glue(helper_movl_mm_T0, SUFFIX) (Reg *d, uint32_t val)
 {
-    Reg *d;
-    d = (Reg *)((char *)env + PARAM1);
-    d->L(0) = T0;
+    d->L(0) = val;
     d->L(1) = 0;
 #if SHIFT == 1
     d->Q(1) = 0;
 #endif
 }
 
-void OPPROTO glue(op_movl_T0_mm, SUFFIX) (void)
-{
-    Reg *s;
-    s = (Reg *)((char *)env + PARAM1);
-    T0 = s->L(0);
-}
-
 #ifdef TARGET_X86_64
-void OPPROTO glue(op_movq_mm_T0, SUFFIX) (void)
+void glue(helper_movq_mm_T0, SUFFIX) (Reg *d, uint64_t val)
 {
-    Reg *d;
-    d = (Reg *)((char *)env + PARAM1);
-    d->Q(0) = T0;
+    d->Q(0) = val;
 #if SHIFT == 1
     d->Q(1) = 0;
 #endif
 }
-
-void OPPROTO glue(op_movq_T0_mm, SUFFIX) (void)
-{
-    Reg *s;
-    s = (Reg *)((char *)env + PARAM1);
-    T0 = s->Q(0);
-}
 #endif
 
 #if SHIFT == 0
-void OPPROTO glue(op_pshufw, SUFFIX) (void)
+void glue(helper_pshufw, SUFFIX) (Reg *d, Reg *s, int order)
 {
-    Reg r, *d, *s;
-    int order;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-    order = PARAM3;
+    Reg r;
     r.W(0) = s->W(order & 3);
     r.W(1) = s->W((order >> 2) & 3);
     r.W(2) = s->W((order >> 4) & 3);
@@ -596,13 +511,9 @@ void OPPROTO glue(op_pshufw, SUFFIX) (void)
     *d = r;
 }
 #else
-void OPPROTO op_shufps(void)
+void helper_shufps(Reg *d, Reg *s, int order)
 {
-    Reg r, *d, *s;
-    int order;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-    order = PARAM3;
+    Reg r;
     r.L(0) = d->L(order & 3);
     r.L(1) = d->L((order >> 2) & 3);
     r.L(2) = s->L((order >> 4) & 3);
@@ -610,25 +521,17 @@ void OPPROTO op_shufps(void)
     *d = r;
 }
 
-void OPPROTO op_shufpd(void)
+void helper_shufpd(Reg *d, Reg *s, int order)
 {
-    Reg r, *d, *s;
-    int order;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-    order = PARAM3;
+    Reg r;
     r.Q(0) = d->Q(order & 1);
     r.Q(1) = s->Q((order >> 1) & 1);
     *d = r;
 }
 
-void OPPROTO glue(op_pshufd, SUFFIX) (void)
+void glue(helper_pshufd, SUFFIX) (Reg *d, Reg *s, int order)
 {
-    Reg r, *d, *s;
-    int order;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-    order = PARAM3;
+    Reg r;
     r.L(0) = s->L(order & 3);
     r.L(1) = s->L((order >> 2) & 3);
     r.L(2) = s->L((order >> 4) & 3);
@@ -636,13 +539,9 @@ void OPPROTO glue(op_pshufd, SUFFIX) (void)
     *d = r;
 }
 
-void OPPROTO glue(op_pshuflw, SUFFIX) (void)
+void glue(helper_pshuflw, SUFFIX) (Reg *d, Reg *s, int order)
 {
-    Reg r, *d, *s;
-    int order;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-    order = PARAM3;
+    Reg r;
     r.W(0) = s->W(order & 3);
     r.W(1) = s->W((order >> 2) & 3);
     r.W(2) = s->W((order >> 4) & 3);
@@ -651,13 +550,9 @@ void OPPROTO glue(op_pshuflw, SUFFIX) (void)
     *d = r;
 }
 
-void OPPROTO glue(op_pshufhw, SUFFIX) (void)
+void glue(helper_pshufhw, SUFFIX) (Reg *d, Reg *s, int order)
 {
-    Reg r, *d, *s;
-    int order;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
-    order = PARAM3;
+    Reg r;
     r.Q(0) = s->Q(0);
     r.W(4) = s->W(4 + (order & 3));
     r.W(5) = s->W(4 + ((order >> 2) & 3));
@@ -671,39 +566,27 @@ void OPPROTO glue(op_pshufhw, SUFFIX) (void)
 /* FPU ops */
 /* XXX: not accurate */
 
-#define SSE_OP_S(name, F)\
-void OPPROTO op_ ## name ## ps (void)\
+#define SSE_HELPER_S(name, F)\
+void helper_ ## name ## ps (Reg *d, Reg *s)\
 {\
-    Reg *d, *s;\
-    d = (Reg *)((char *)env + PARAM1);\
-    s = (Reg *)((char *)env + PARAM2);\
     d->XMM_S(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
     d->XMM_S(1) = F(32, d->XMM_S(1), s->XMM_S(1));\
     d->XMM_S(2) = F(32, d->XMM_S(2), s->XMM_S(2));\
     d->XMM_S(3) = F(32, d->XMM_S(3), s->XMM_S(3));\
 }\
 \
-void OPPROTO op_ ## name ## ss (void)\
+void helper_ ## name ## ss (Reg *d, Reg *s)\
 {\
-    Reg *d, *s;\
-    d = (Reg *)((char *)env + PARAM1);\
-    s = (Reg *)((char *)env + PARAM2);\
     d->XMM_S(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
 }\
-void OPPROTO op_ ## name ## pd (void)\
+void helper_ ## name ## pd (Reg *d, Reg *s)\
 {\
-    Reg *d, *s;\
-    d = (Reg *)((char *)env + PARAM1);\
-    s = (Reg *)((char *)env + PARAM2);\
     d->XMM_D(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
     d->XMM_D(1) = F(64, d->XMM_D(1), s->XMM_D(1));\
 }\
 \
-void OPPROTO op_ ## name ## sd (void)\
+void helper_ ## name ## sd (Reg *d, Reg *s)\
 {\
-    Reg *d, *s;\
-    d = (Reg *)((char *)env + PARAM1);\
-    s = (Reg *)((char *)env + PARAM2);\
     d->XMM_D(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
 }
 
@@ -715,69 +598,53 @@ void OPPROTO op_ ## name ## sd (void)\
 #define FPU_MAX(size, a, b) (a) > (b) ? (a) : (b)
 #define FPU_SQRT(size, a, b) float ## size ## _sqrt(b, &env->sse_status)
 
-SSE_OP_S(add, FPU_ADD)
-SSE_OP_S(sub, FPU_SUB)
-SSE_OP_S(mul, FPU_MUL)
-SSE_OP_S(div, FPU_DIV)
-SSE_OP_S(min, FPU_MIN)
-SSE_OP_S(max, FPU_MAX)
-SSE_OP_S(sqrt, FPU_SQRT)
+SSE_HELPER_S(add, FPU_ADD)
+SSE_HELPER_S(sub, FPU_SUB)
+SSE_HELPER_S(mul, FPU_MUL)
+SSE_HELPER_S(div, FPU_DIV)
+SSE_HELPER_S(min, FPU_MIN)
+SSE_HELPER_S(max, FPU_MAX)
+SSE_HELPER_S(sqrt, FPU_SQRT)
 
 
 /* float to float conversions */
-void OPPROTO op_cvtps2pd(void)
+void helper_cvtps2pd(Reg *d, Reg *s)
 {
     float32 s0, s1;
-    Reg *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
     s0 = s->XMM_S(0);
     s1 = s->XMM_S(1);
     d->XMM_D(0) = float32_to_float64(s0, &env->sse_status);
     d->XMM_D(1) = float32_to_float64(s1, &env->sse_status);
 }
 
-void OPPROTO op_cvtpd2ps(void)
+void helper_cvtpd2ps(Reg *d, Reg *s)
 {
-    Reg *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
     d->XMM_S(0) = float64_to_float32(s->XMM_D(0), &env->sse_status);
     d->XMM_S(1) = float64_to_float32(s->XMM_D(1), &env->sse_status);
     d->Q(1) = 0;
 }
 
-void OPPROTO op_cvtss2sd(void)
+void helper_cvtss2sd(Reg *d, Reg *s)
 {
-    Reg *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
     d->XMM_D(0) = float32_to_float64(s->XMM_S(0), &env->sse_status);
 }
 
-void OPPROTO op_cvtsd2ss(void)
+void helper_cvtsd2ss(Reg *d, Reg *s)
 {
-    Reg *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
     d->XMM_S(0) = float64_to_float32(s->XMM_D(0), &env->sse_status);
 }
 
 /* integer to float */
-void OPPROTO op_cvtdq2ps(void)
+void helper_cvtdq2ps(Reg *d, Reg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->XMM_S(0) = int32_to_float32(s->XMM_L(0), &env->sse_status);
     d->XMM_S(1) = int32_to_float32(s->XMM_L(1), &env->sse_status);
     d->XMM_S(2) = int32_to_float32(s->XMM_L(2), &env->sse_status);
     d->XMM_S(3) = int32_to_float32(s->XMM_L(3), &env->sse_status);
 }
 
-void OPPROTO op_cvtdq2pd(void)
+void helper_cvtdq2pd(Reg *d, Reg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     int32_t l0, l1;
     l0 = (int32_t)s->XMM_L(0);
     l1 = (int32_t)s->XMM_L(1);
@@ -785,210 +652,168 @@ void OPPROTO op_cvtdq2pd(void)
     d->XMM_D(1) = int32_to_float64(l1, &env->sse_status);
 }
 
-void OPPROTO op_cvtpi2ps(void)
+void helper_cvtpi2ps(XMMReg *d, MMXReg *s)
 {
-    XMMReg *d = (Reg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->XMM_S(0) = int32_to_float32(s->MMX_L(0), &env->sse_status);
     d->XMM_S(1) = int32_to_float32(s->MMX_L(1), &env->sse_status);
 }
 
-void OPPROTO op_cvtpi2pd(void)
+void helper_cvtpi2pd(XMMReg *d, MMXReg *s)
 {
-    XMMReg *d = (Reg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->XMM_D(0) = int32_to_float64(s->MMX_L(0), &env->sse_status);
     d->XMM_D(1) = int32_to_float64(s->MMX_L(1), &env->sse_status);
 }
 
-void OPPROTO op_cvtsi2ss(void)
+void helper_cvtsi2ss(XMMReg *d, uint32_t val)
 {
-    XMMReg *d = (Reg *)((char *)env + PARAM1);
-    d->XMM_S(0) = int32_to_float32(T0, &env->sse_status);
+    d->XMM_S(0) = int32_to_float32(val, &env->sse_status);
 }
 
-void OPPROTO op_cvtsi2sd(void)
+void helper_cvtsi2sd(XMMReg *d, uint32_t val)
 {
-    XMMReg *d = (Reg *)((char *)env + PARAM1);
-    d->XMM_D(0) = int32_to_float64(T0, &env->sse_status);
+    d->XMM_D(0) = int32_to_float64(val, &env->sse_status);
 }
 
 #ifdef TARGET_X86_64
-void OPPROTO op_cvtsq2ss(void)
+void helper_cvtsq2ss(XMMReg *d, uint64_t val)
 {
-    XMMReg *d = (Reg *)((char *)env + PARAM1);
-    d->XMM_S(0) = int64_to_float32(T0, &env->sse_status);
+    d->XMM_S(0) = int64_to_float32(val, &env->sse_status);
 }
 
-void OPPROTO op_cvtsq2sd(void)
+void helper_cvtsq2sd(XMMReg *d, uint64_t val)
 {
-    XMMReg *d = (Reg *)((char *)env + PARAM1);
-    d->XMM_D(0) = int64_to_float64(T0, &env->sse_status);
+    d->XMM_D(0) = int64_to_float64(val, &env->sse_status);
 }
 #endif
 
 /* float to integer */
-void OPPROTO op_cvtps2dq(void)
+void helper_cvtps2dq(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->XMM_L(0) = float32_to_int32(s->XMM_S(0), &env->sse_status);
     d->XMM_L(1) = float32_to_int32(s->XMM_S(1), &env->sse_status);
     d->XMM_L(2) = float32_to_int32(s->XMM_S(2), &env->sse_status);
     d->XMM_L(3) = float32_to_int32(s->XMM_S(3), &env->sse_status);
 }
 
-void OPPROTO op_cvtpd2dq(void)
+void helper_cvtpd2dq(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->XMM_L(0) = float64_to_int32(s->XMM_D(0), &env->sse_status);
     d->XMM_L(1) = float64_to_int32(s->XMM_D(1), &env->sse_status);
     d->XMM_Q(1) = 0;
 }
 
-void OPPROTO op_cvtps2pi(void)
+void helper_cvtps2pi(MMXReg *d, XMMReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->MMX_L(0) = float32_to_int32(s->XMM_S(0), &env->sse_status);
     d->MMX_L(1) = float32_to_int32(s->XMM_S(1), &env->sse_status);
 }
 
-void OPPROTO op_cvtpd2pi(void)
+void helper_cvtpd2pi(MMXReg *d, XMMReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->MMX_L(0) = float64_to_int32(s->XMM_D(0), &env->sse_status);
     d->MMX_L(1) = float64_to_int32(s->XMM_D(1), &env->sse_status);
 }
 
-void OPPROTO op_cvtss2si(void)
+int32_t helper_cvtss2si(XMMReg *s)
 {
-    XMMReg *s = (XMMReg *)((char *)env + PARAM1);
-    T0 = float32_to_int32(s->XMM_S(0), &env->sse_status);
+    return float32_to_int32(s->XMM_S(0), &env->sse_status);
 }
 
-void OPPROTO op_cvtsd2si(void)
+int32_t helper_cvtsd2si(XMMReg *s)
 {
-    XMMReg *s = (XMMReg *)((char *)env + PARAM1);
-    T0 = float64_to_int32(s->XMM_D(0), &env->sse_status);
+    return float64_to_int32(s->XMM_D(0), &env->sse_status);
 }
 
 #ifdef TARGET_X86_64
-void OPPROTO op_cvtss2sq(void)
+int64_t helper_cvtss2sq(XMMReg *s)
 {
-    XMMReg *s = (XMMReg *)((char *)env + PARAM1);
-    T0 = float32_to_int64(s->XMM_S(0), &env->sse_status);
+    return float32_to_int64(s->XMM_S(0), &env->sse_status);
 }
 
-void OPPROTO op_cvtsd2sq(void)
+int64_t helper_cvtsd2sq(XMMReg *s)
 {
-    XMMReg *s = (XMMReg *)((char *)env + PARAM1);
-    T0 = float64_to_int64(s->XMM_D(0), &env->sse_status);
+    return float64_to_int64(s->XMM_D(0), &env->sse_status);
 }
 #endif
 
 /* float to integer truncated */
-void OPPROTO op_cvttps2dq(void)
+void helper_cvttps2dq(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->XMM_L(0) = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
     d->XMM_L(1) = float32_to_int32_round_to_zero(s->XMM_S(1), &env->sse_status);
     d->XMM_L(2) = float32_to_int32_round_to_zero(s->XMM_S(2), &env->sse_status);
     d->XMM_L(3) = float32_to_int32_round_to_zero(s->XMM_S(3), &env->sse_status);
 }
 
-void OPPROTO op_cvttpd2dq(void)
+void helper_cvttpd2dq(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->XMM_L(0) = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
     d->XMM_L(1) = float64_to_int32_round_to_zero(s->XMM_D(1), &env->sse_status);
     d->XMM_Q(1) = 0;
 }
 
-void OPPROTO op_cvttps2pi(void)
+void helper_cvttps2pi(MMXReg *d, XMMReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->MMX_L(0) = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
     d->MMX_L(1) = float32_to_int32_round_to_zero(s->XMM_S(1), &env->sse_status);
 }
 
-void OPPROTO op_cvttpd2pi(void)
+void helper_cvttpd2pi(MMXReg *d, XMMReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->MMX_L(0) = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
     d->MMX_L(1) = float64_to_int32_round_to_zero(s->XMM_D(1), &env->sse_status);
 }
 
-void OPPROTO op_cvttss2si(void)
+int32_t helper_cvttss2si(XMMReg *s)
 {
-    XMMReg *s = (XMMReg *)((char *)env + PARAM1);
-    T0 = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
+    return float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
 }
 
-void OPPROTO op_cvttsd2si(void)
+int32_t helper_cvttsd2si(XMMReg *s)
 {
-    XMMReg *s = (XMMReg *)((char *)env + PARAM1);
-    T0 = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
+    return float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
 }
 
 #ifdef TARGET_X86_64
-void OPPROTO op_cvttss2sq(void)
+int64_t helper_cvttss2sq(XMMReg *s)
 {
-    XMMReg *s = (XMMReg *)((char *)env + PARAM1);
-    T0 = float32_to_int64_round_to_zero(s->XMM_S(0), &env->sse_status);
+    return float32_to_int64_round_to_zero(s->XMM_S(0), &env->sse_status);
 }
 
-void OPPROTO op_cvttsd2sq(void)
+int64_t helper_cvttsd2sq(XMMReg *s)
 {
-    XMMReg *s = (XMMReg *)((char *)env + PARAM1);
-    T0 = float64_to_int64_round_to_zero(s->XMM_D(0), &env->sse_status);
+    return float64_to_int64_round_to_zero(s->XMM_D(0), &env->sse_status);
 }
 #endif
 
-void OPPROTO op_rsqrtps(void)
+void helper_rsqrtps(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->XMM_S(0) = approx_rsqrt(s->XMM_S(0));
     d->XMM_S(1) = approx_rsqrt(s->XMM_S(1));
     d->XMM_S(2) = approx_rsqrt(s->XMM_S(2));
     d->XMM_S(3) = approx_rsqrt(s->XMM_S(3));
 }
 
-void OPPROTO op_rsqrtss(void)
+void helper_rsqrtss(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->XMM_S(0) = approx_rsqrt(s->XMM_S(0));
 }
 
-void OPPROTO op_rcpps(void)
+void helper_rcpps(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->XMM_S(0) = approx_rcp(s->XMM_S(0));
     d->XMM_S(1) = approx_rcp(s->XMM_S(1));
     d->XMM_S(2) = approx_rcp(s->XMM_S(2));
     d->XMM_S(3) = approx_rcp(s->XMM_S(3));
 }
 
-void OPPROTO op_rcpss(void)
+void helper_rcpss(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->XMM_S(0) = approx_rcp(s->XMM_S(0));
 }
 
-void OPPROTO op_haddps(void)
+void helper_haddps(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     XMMReg r;
     r.XMM_S(0) = d->XMM_S(0) + d->XMM_S(1);
     r.XMM_S(1) = d->XMM_S(2) + d->XMM_S(3);
@@ -997,20 +822,16 @@ void OPPROTO op_haddps(void)
     *d = r;
 }
 
-void OPPROTO op_haddpd(void)
+void helper_haddpd(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     XMMReg r;
     r.XMM_D(0) = d->XMM_D(0) + d->XMM_D(1);
     r.XMM_D(1) = s->XMM_D(0) + s->XMM_D(1);
     *d = r;
 }
 
-void OPPROTO op_hsubps(void)
+void helper_hsubps(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     XMMReg r;
     r.XMM_S(0) = d->XMM_S(0) - d->XMM_S(1);
     r.XMM_S(1) = d->XMM_S(2) - d->XMM_S(3);
@@ -1019,68 +840,50 @@ void OPPROTO op_hsubps(void)
     *d = r;
 }
 
-void OPPROTO op_hsubpd(void)
+void helper_hsubpd(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     XMMReg r;
     r.XMM_D(0) = d->XMM_D(0) - d->XMM_D(1);
     r.XMM_D(1) = s->XMM_D(0) - s->XMM_D(1);
     *d = r;
 }
 
-void OPPROTO op_addsubps(void)
+void helper_addsubps(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->XMM_S(0) = d->XMM_S(0) - s->XMM_S(0);
     d->XMM_S(1) = d->XMM_S(1) + s->XMM_S(1);
     d->XMM_S(2) = d->XMM_S(2) - s->XMM_S(2);
     d->XMM_S(3) = d->XMM_S(3) + s->XMM_S(3);
 }
 
-void OPPROTO op_addsubpd(void)
+void helper_addsubpd(XMMReg *d, XMMReg *s)
 {
-    XMMReg *d = (XMMReg *)((char *)env + PARAM1);
-    XMMReg *s = (XMMReg *)((char *)env + PARAM2);
     d->XMM_D(0) = d->XMM_D(0) - s->XMM_D(0);
     d->XMM_D(1) = d->XMM_D(1) + s->XMM_D(1);
 }
 
 /* XXX: unordered */
-#define SSE_OP_CMP(name, F)\
-void OPPROTO op_ ## name ## ps (void)\
+#define SSE_HELPER_CMP(name, F)\
+void helper_ ## name ## ps (Reg *d, Reg *s)\
 {\
-    Reg *d, *s;\
-    d = (Reg *)((char *)env + PARAM1);\
-    s = (Reg *)((char *)env + PARAM2);\
     d->XMM_L(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
     d->XMM_L(1) = F(32, d->XMM_S(1), s->XMM_S(1));\
     d->XMM_L(2) = F(32, d->XMM_S(2), s->XMM_S(2));\
     d->XMM_L(3) = F(32, d->XMM_S(3), s->XMM_S(3));\
 }\
 \
-void OPPROTO op_ ## name ## ss (void)\
+void helper_ ## name ## ss (Reg *d, Reg *s)\
 {\
-    Reg *d, *s;\
-    d = (Reg *)((char *)env + PARAM1);\
-    s = (Reg *)((char *)env + PARAM2);\
     d->XMM_L(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
 }\
-void OPPROTO op_ ## name ## pd (void)\
+void helper_ ## name ## pd (Reg *d, Reg *s)\
 {\
-    Reg *d, *s;\
-    d = (Reg *)((char *)env + PARAM1);\
-    s = (Reg *)((char *)env + PARAM2);\
     d->XMM_Q(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
     d->XMM_Q(1) = F(64, d->XMM_D(1), s->XMM_D(1));\
 }\
 \
-void OPPROTO op_ ## name ## sd (void)\
+void helper_ ## name ## sd (Reg *d, Reg *s)\
 {\
-    Reg *d, *s;\
-    d = (Reg *)((char *)env + PARAM1);\
-    s = (Reg *)((char *)env + PARAM2);\
     d->XMM_Q(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
 }
 
@@ -1093,24 +896,21 @@ void OPPROTO op_ ## name ## sd (void)\
 #define FPU_CMPNLE(size, a, b) float ## size ## _le(a, b, &env->sse_status) ? 0 : -1
 #define FPU_CMPORD(size, a, b) float ## size ## _unordered(a, b, &env->sse_status) ? 0 : -1
 
-SSE_OP_CMP(cmpeq, FPU_CMPEQ)
-SSE_OP_CMP(cmplt, FPU_CMPLT)
-SSE_OP_CMP(cmple, FPU_CMPLE)
-SSE_OP_CMP(cmpunord, FPU_CMPUNORD)
-SSE_OP_CMP(cmpneq, FPU_CMPNEQ)
-SSE_OP_CMP(cmpnlt, FPU_CMPNLT)
-SSE_OP_CMP(cmpnle, FPU_CMPNLE)
-SSE_OP_CMP(cmpord, FPU_CMPORD)
+SSE_HELPER_CMP(cmpeq, FPU_CMPEQ)
+SSE_HELPER_CMP(cmplt, FPU_CMPLT)
+SSE_HELPER_CMP(cmple, FPU_CMPLE)
+SSE_HELPER_CMP(cmpunord, FPU_CMPUNORD)
+SSE_HELPER_CMP(cmpneq, FPU_CMPNEQ)
+SSE_HELPER_CMP(cmpnlt, FPU_CMPNLT)
+SSE_HELPER_CMP(cmpnle, FPU_CMPNLE)
+SSE_HELPER_CMP(cmpord, FPU_CMPORD)
 
 const int comis_eflags[4] = {CC_C, CC_Z, 0, CC_Z | CC_P | CC_C};
 
-void OPPROTO op_ucomiss(void)
+void helper_ucomiss(Reg *d, Reg *s)
 {
     int ret;
     float32 s0, s1;
-    Reg *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
 
     s0 = d->XMM_S(0);
     s1 = s->XMM_S(0);
@@ -1119,13 +919,10 @@ void OPPROTO op_ucomiss(void)
     FORCE_RET();
 }
 
-void OPPROTO op_comiss(void)
+void helper_comiss(Reg *d, Reg *s)
 {
     int ret;
     float32 s0, s1;
-    Reg *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
 
     s0 = d->XMM_S(0);
     s1 = s->XMM_S(0);
@@ -1134,13 +931,10 @@ void OPPROTO op_comiss(void)
     FORCE_RET();
 }
 
-void OPPROTO op_ucomisd(void)
+void helper_ucomisd(Reg *d, Reg *s)
 {
     int ret;
     float64 d0, d1;
-    Reg *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
 
     d0 = d->XMM_D(0);
     d1 = s->XMM_D(0);
@@ -1149,13 +943,10 @@ void OPPROTO op_ucomisd(void)
     FORCE_RET();
 }
 
-void OPPROTO op_comisd(void)
+void helper_comisd(Reg *d, Reg *s)
 {
     int ret;
     float64 d0, d1;
-    Reg *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
 
     d0 = d->XMM_D(0);
     d1 = s->XMM_D(0);
@@ -1164,76 +955,54 @@ void OPPROTO op_comisd(void)
     FORCE_RET();
 }
 
-void OPPROTO op_movmskps(void)
+uint32_t helper_movmskps(Reg *s)
 {
     int b0, b1, b2, b3;
-    Reg *s;
-    s = (Reg *)((char *)env + PARAM1);
     b0 = s->XMM_L(0) >> 31;
     b1 = s->XMM_L(1) >> 31;
     b2 = s->XMM_L(2) >> 31;
     b3 = s->XMM_L(3) >> 31;
-    T0 = b0 | (b1 << 1) | (b2 << 2) | (b3 << 3);
+    return b0 | (b1 << 1) | (b2 << 2) | (b3 << 3);
 }
 
-void OPPROTO op_movmskpd(void)
+uint32_t helper_movmskpd(Reg *s)
 {
     int b0, b1;
-    Reg *s;
-    s = (Reg *)((char *)env + PARAM1);
     b0 = s->XMM_L(1) >> 31;
     b1 = s->XMM_L(3) >> 31;
-    T0 = b0 | (b1 << 1);
+    return b0 | (b1 << 1);
 }
 
 #endif
 
-void OPPROTO glue(op_pmovmskb, SUFFIX)(void)
-{
-    Reg *s;
-    s = (Reg *)((char *)env + PARAM1);
-    T0 = 0;
-    T0 |= (s->XMM_B(0) >> 7);
-    T0 |= (s->XMM_B(1) >> 6) & 0x02;
-    T0 |= (s->XMM_B(2) >> 5) & 0x04;
-    T0 |= (s->XMM_B(3) >> 4) & 0x08;
-    T0 |= (s->XMM_B(4) >> 3) & 0x10;
-    T0 |= (s->XMM_B(5) >> 2) & 0x20;
-    T0 |= (s->XMM_B(6) >> 1) & 0x40;
-    T0 |= (s->XMM_B(7)) & 0x80;
+uint32_t glue(helper_pmovmskb, SUFFIX)(Reg *s)
+{
+    uint32_t val;
+    val = 0;
+    val |= (s->XMM_B(0) >> 7);
+    val |= (s->XMM_B(1) >> 6) & 0x02;
+    val |= (s->XMM_B(2) >> 5) & 0x04;
+    val |= (s->XMM_B(3) >> 4) & 0x08;
+    val |= (s->XMM_B(4) >> 3) & 0x10;
+    val |= (s->XMM_B(5) >> 2) & 0x20;
+    val |= (s->XMM_B(6) >> 1) & 0x40;
+    val |= (s->XMM_B(7)) & 0x80;
 #if SHIFT == 1
-    T0 |= (s->XMM_B(8) << 1) & 0x0100;
-    T0 |= (s->XMM_B(9) << 2) & 0x0200;
-    T0 |= (s->XMM_B(10) << 3) & 0x0400;
-    T0 |= (s->XMM_B(11) << 4) & 0x0800;
-    T0 |= (s->XMM_B(12) << 5) & 0x1000;
-    T0 |= (s->XMM_B(13) << 6) & 0x2000;
-    T0 |= (s->XMM_B(14) << 7) & 0x4000;
-    T0 |= (s->XMM_B(15) << 8) & 0x8000;
+    val |= (s->XMM_B(8) << 1) & 0x0100;
+    val |= (s->XMM_B(9) << 2) & 0x0200;
+    val |= (s->XMM_B(10) << 3) & 0x0400;
+    val |= (s->XMM_B(11) << 4) & 0x0800;
+    val |= (s->XMM_B(12) << 5) & 0x1000;
+    val |= (s->XMM_B(13) << 6) & 0x2000;
+    val |= (s->XMM_B(14) << 7) & 0x4000;
+    val |= (s->XMM_B(15) << 8) & 0x8000;
 #endif
+    return val;
 }
 
-void OPPROTO glue(op_pinsrw, SUFFIX) (void)
-{
-    Reg *d = (Reg *)((char *)env + PARAM1);
-    int pos = PARAM2;
-
-    d->W(pos) = T0;
-}
-
-void OPPROTO glue(op_pextrw, SUFFIX) (void)
-{
-    Reg *s = (Reg *)((char *)env + PARAM1);
-    int pos = PARAM2;
-
-    T0 = s->W(pos);
-}
-
-void OPPROTO glue(op_packsswb, SUFFIX) (void)
+void glue(helper_packsswb, SUFFIX) (Reg *d, Reg *s)
 {
-    Reg r, *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
+    Reg r;
 
     r.B(0) = satsb((int16_t)d->W(0));
     r.B(1) = satsb((int16_t)d->W(1));
@@ -1258,11 +1027,9 @@ void OPPROTO glue(op_packsswb, SUFFIX) (void)
     *d = r;
 }
 
-void OPPROTO glue(op_packuswb, SUFFIX) (void)
+void glue(helper_packuswb, SUFFIX) (Reg *d, Reg *s)
 {
-    Reg r, *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
+    Reg r;
 
     r.B(0) = satub((int16_t)d->W(0));
     r.B(1) = satub((int16_t)d->W(1));
@@ -1287,11 +1054,9 @@ void OPPROTO glue(op_packuswb, SUFFIX) (void)
     *d = r;
 }
 
-void OPPROTO glue(op_packssdw, SUFFIX) (void)
+void glue(helper_packssdw, SUFFIX) (Reg *d, Reg *s)
 {
-    Reg r, *d, *s;
-    d = (Reg *)((char *)env + PARAM1);
-    s = (Reg *)((char *)env + PARAM2);
+    Reg r;
 
     r.W(0) = satsw(d->L(0));
     r.W(1) = satsw(d->L(1));
@@ -1310,11 +1075,9 @@ void OPPROTO glue(op_packssdw, SUFFIX) (void)
 
 #define UNPCK_OP(base_name, base)                               \
                                                                 \
-void OPPROTO glue(op_punpck ## base_name ## bw, SUFFIX) (void)   \
+void glue(helper_punpck ## base_name ## bw, SUFFIX) (Reg *d, Reg *s)   \
 {                                                               \
-    Reg r, *d, *s;                                              \
-    d = (Reg *)((char *)env + PARAM1);                          \
-    s = (Reg *)((char *)env + PARAM2);                          \
+    Reg r;                                              \
                                                                 \
     r.B(0) = d->B((base << (SHIFT + 2)) + 0);                   \
     r.B(1) = s->B((base << (SHIFT + 2)) + 0);                   \
@@ -1337,11 +1100,9 @@ XMM_ONLY(                                                       \
     *d = r;                                                     \
 }                                                               \
                                                                 \
-void OPPROTO glue(op_punpck ## base_name ## wd, SUFFIX) (void)   \
+void glue(helper_punpck ## base_name ## wd, SUFFIX) (Reg *d, Reg *s)   \
 {                                                               \
-    Reg r, *d, *s;                                              \
-    d = (Reg *)((char *)env + PARAM1);                          \
-    s = (Reg *)((char *)env + PARAM2);                          \
+    Reg r;                                              \
                                                                 \
     r.W(0) = d->W((base << (SHIFT + 1)) + 0);                   \
     r.W(1) = s->W((base << (SHIFT + 1)) + 0);                   \
@@ -1356,11 +1117,9 @@ XMM_ONLY(                                                       \
     *d = r;                                                     \
 }                                                               \
                                                                 \
-void OPPROTO glue(op_punpck ## base_name ## dq, SUFFIX) (void)   \
+void glue(helper_punpck ## base_name ## dq, SUFFIX) (Reg *d, Reg *s)   \
 {                                                               \
-    Reg r, *d, *s;                                              \
-    d = (Reg *)((char *)env + PARAM1);                          \
-    s = (Reg *)((char *)env + PARAM2);                          \
+    Reg r;                                              \
                                                                 \
     r.L(0) = d->L((base << SHIFT) + 0);                         \
     r.L(1) = s->L((base << SHIFT) + 0);                         \
@@ -1372,11 +1131,9 @@ XMM_ONLY(                                                       \
 }                                                               \
                                                                 \
 XMM_ONLY(                                                       \
-void OPPROTO glue(op_punpck ## base_name ## qdq, SUFFIX) (void)  \
+void glue(helper_punpck ## base_name ## qdq, SUFFIX) (Reg *d, Reg *s)  \
 {                                                               \
-    Reg r, *d, *s;                                              \
-    d = (Reg *)((char *)env + PARAM1);                          \
-    s = (Reg *)((char *)env + PARAM2);                          \
+    Reg r;                                              \
                                                                 \
     r.Q(0) = d->Q(base);                                        \
     r.Q(1) = s->Q(base);                                        \
@@ -1389,166 +1146,128 @@ UNPCK_OP(h, 1)
 
 /* 3DNow! float ops */
 #if SHIFT == 0
-void OPPROTO op_pi2fd(void)
+void helper_pi2fd(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->MMX_S(0) = int32_to_float32(s->MMX_L(0), &env->mmx_status);
     d->MMX_S(1) = int32_to_float32(s->MMX_L(1), &env->mmx_status);
 }
 
-void OPPROTO op_pi2fw(void)
+void helper_pi2fw(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->MMX_S(0) = int32_to_float32((int16_t)s->MMX_W(0), &env->mmx_status);
     d->MMX_S(1) = int32_to_float32((int16_t)s->MMX_W(2), &env->mmx_status);
 }
 
-void OPPROTO op_pf2id(void)
+void helper_pf2id(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->MMX_L(0) = float32_to_int32_round_to_zero(s->MMX_S(0), &env->mmx_status);
     d->MMX_L(1) = float32_to_int32_round_to_zero(s->MMX_S(1), &env->mmx_status);
 }
 
-void OPPROTO op_pf2iw(void)
+void helper_pf2iw(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->MMX_L(0) = satsw(float32_to_int32_round_to_zero(s->MMX_S(0), &env->mmx_status));
     d->MMX_L(1) = satsw(float32_to_int32_round_to_zero(s->MMX_S(1), &env->mmx_status));
 }
 
-void OPPROTO op_pfacc(void)
+void helper_pfacc(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     MMXReg r;
     r.MMX_S(0) = float32_add(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
     r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
     *d = r;
 }
 
-void OPPROTO op_pfadd(void)
+void helper_pfadd(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->MMX_S(0) = float32_add(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
     d->MMX_S(1) = float32_add(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
 }
 
-void OPPROTO op_pfcmpeq(void)
+void helper_pfcmpeq(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->MMX_L(0) = float32_eq(d->MMX_S(0), s->MMX_S(0), &env->mmx_status) ? -1 : 0;
     d->MMX_L(1) = float32_eq(d->MMX_S(1), s->MMX_S(1), &env->mmx_status) ? -1 : 0;
 }
 
-void OPPROTO op_pfcmpge(void)
+void helper_pfcmpge(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->MMX_L(0) = float32_le(s->MMX_S(0), d->MMX_S(0), &env->mmx_status) ? -1 : 0;
     d->MMX_L(1) = float32_le(s->MMX_S(1), d->MMX_S(1), &env->mmx_status) ? -1 : 0;
 }
 
-void OPPROTO op_pfcmpgt(void)
+void helper_pfcmpgt(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->MMX_L(0) = float32_lt(s->MMX_S(0), d->MMX_S(0), &env->mmx_status) ? -1 : 0;
     d->MMX_L(1) = float32_lt(s->MMX_S(1), d->MMX_S(1), &env->mmx_status) ? -1 : 0;
 }
 
-void OPPROTO op_pfmax(void)
+void helper_pfmax(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     if (float32_lt(d->MMX_S(0), s->MMX_S(0), &env->mmx_status))
         d->MMX_S(0) = s->MMX_S(0);
     if (float32_lt(d->MMX_S(1), s->MMX_S(1), &env->mmx_status))
         d->MMX_S(1) = s->MMX_S(1);
 }
 
-void OPPROTO op_pfmin(void)
+void helper_pfmin(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     if (float32_lt(s->MMX_S(0), d->MMX_S(0), &env->mmx_status))
         d->MMX_S(0) = s->MMX_S(0);
     if (float32_lt(s->MMX_S(1), d->MMX_S(1), &env->mmx_status))
         d->MMX_S(1) = s->MMX_S(1);
 }
 
-void OPPROTO op_pfmul(void)
+void helper_pfmul(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->MMX_S(0) = float32_mul(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
     d->MMX_S(1) = float32_mul(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
 }
 
-void OPPROTO op_pfnacc(void)
+void helper_pfnacc(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     MMXReg r;
     r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
     r.MMX_S(1) = float32_sub(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
     *d = r;
 }
 
-void OPPROTO op_pfpnacc(void)
+void helper_pfpnacc(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     MMXReg r;
     r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
     r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
     *d = r;
 }
 
-void OPPROTO op_pfrcp(void)
+void helper_pfrcp(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->MMX_S(0) = approx_rcp(s->MMX_S(0));
     d->MMX_S(1) = d->MMX_S(0);
 }
 
-void OPPROTO op_pfrsqrt(void)
+void helper_pfrsqrt(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->MMX_L(1) = s->MMX_L(0) & 0x7fffffff;
     d->MMX_S(1) = approx_rsqrt(d->MMX_S(1));
     d->MMX_L(1) |= s->MMX_L(0) & 0x80000000;
     d->MMX_L(0) = d->MMX_L(1);
 }
 
-void OPPROTO op_pfsub(void)
+void helper_pfsub(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->MMX_S(0) = float32_sub(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
     d->MMX_S(1) = float32_sub(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
 }
 
-void OPPROTO op_pfsubr(void)
+void helper_pfsubr(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     d->MMX_S(0) = float32_sub(s->MMX_S(0), d->MMX_S(0), &env->mmx_status);
     d->MMX_S(1) = float32_sub(s->MMX_S(1), d->MMX_S(1), &env->mmx_status);
 }
 
-void OPPROTO op_pswapd(void)
+void helper_pswapd(MMXReg *d, MMXReg *s)
 {
-    MMXReg *d = (MMXReg *)((char *)env + PARAM1);
-    MMXReg *s = (MMXReg *)((char *)env + PARAM2);
     MMXReg r;
     r.MMX_L(0) = s->MMX_L(1);
     r.MMX_L(1) = s->MMX_L(0);
diff --git a/target-i386/ops_sse_header.h b/target-i386/ops_sse_header.h
new file mode 100644
index 0000000000..07cfea40b7
--- /dev/null
+++ b/target-i386/ops_sse_header.h
@@ -0,0 +1,264 @@
+/*
+ *  MMX/3DNow!/SSE/SSE2/SSE3/PNI support
+ *
+ *  Copyright (c) 2005 Fabrice Bellard
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+#if SHIFT == 0
+#define Reg MMXReg
+#define SUFFIX _mmx
+#else
+#define Reg XMMReg
+#define SUFFIX _xmm
+#endif
+
+void glue(helper_psrlw, SUFFIX)(Reg *d, Reg *s);
+void glue(helper_psraw, SUFFIX)(Reg *d, Reg *s);
+void glue(helper_psllw, SUFFIX)(Reg *d, Reg *s);
+void glue(helper_psrld, SUFFIX)(Reg *d, Reg *s);
+void glue(helper_psrad, SUFFIX)(Reg *d, Reg *s);
+void glue(helper_pslld, SUFFIX)(Reg *d, Reg *s);
+void glue(helper_psrlq, SUFFIX)(Reg *d, Reg *s);
+void glue(helper_psllq, SUFFIX)(Reg *d, Reg *s);
+
+#if SHIFT == 1
+void glue(helper_psrldq, SUFFIX)(Reg *d, Reg *s);
+void glue(helper_pslldq, SUFFIX)(Reg *d, Reg *s);
+#endif
+
+#define SSE_HELPER_B(name, F)\
+    void glue(name, SUFFIX) (Reg *d, Reg *s);
+
+#define SSE_HELPER_W(name, F)\
+    void glue(name, SUFFIX) (Reg *d, Reg *s);
+
+#define SSE_HELPER_L(name, F)\
+    void glue(name, SUFFIX) (Reg *d, Reg *s);
+
+#define SSE_HELPER_Q(name, F)\
+    void glue(name, SUFFIX) (Reg *d, Reg *s);
+
+SSE_HELPER_B(helper_paddb, FADD);
+SSE_HELPER_W(helper_paddw, FADD);
+SSE_HELPER_L(helper_paddl, FADD);
+SSE_HELPER_Q(helper_paddq, FADD);
+
+SSE_HELPER_B(helper_psubb, FSUB);
+SSE_HELPER_W(helper_psubw, FSUB);
+SSE_HELPER_L(helper_psubl, FSUB);
+SSE_HELPER_Q(helper_psubq, FSUB);
+
+SSE_HELPER_B(helper_paddusb, FADDUB);
+SSE_HELPER_B(helper_paddsb, FADDSB);
+SSE_HELPER_B(helper_psubusb, FSUBUB);
+SSE_HELPER_B(helper_psubsb, FSUBSB);
+
+SSE_HELPER_W(helper_paddusw, FADDUW);
+SSE_HELPER_W(helper_paddsw, FADDSW);
+SSE_HELPER_W(helper_psubusw, FSUBUW);
+SSE_HELPER_W(helper_psubsw, FSUBSW);
+
+SSE_HELPER_B(helper_pminub, FMINUB);
+SSE_HELPER_B(helper_pmaxub, FMAXUB);
+
+SSE_HELPER_W(helper_pminsw, FMINSW);
+SSE_HELPER_W(helper_pmaxsw, FMAXSW);
+
+SSE_HELPER_Q(helper_pand, FAND);
+SSE_HELPER_Q(helper_pandn, FANDN);
+SSE_HELPER_Q(helper_por, FOR);
+SSE_HELPER_Q(helper_pxor, FXOR);
+
+SSE_HELPER_B(helper_pcmpgtb, FCMPGTB);
+SSE_HELPER_W(helper_pcmpgtw, FCMPGTW);
+SSE_HELPER_L(helper_pcmpgtl, FCMPGTL);
+
+SSE_HELPER_B(helper_pcmpeqb, FCMPEQ);
+SSE_HELPER_W(helper_pcmpeqw, FCMPEQ);
+SSE_HELPER_L(helper_pcmpeql, FCMPEQ);
+
+SSE_HELPER_W(helper_pmullw, FMULLW);
+#if SHIFT == 0
+SSE_HELPER_W(helper_pmulhrw, FMULHRW);
+#endif
+SSE_HELPER_W(helper_pmulhuw, FMULHUW);
+SSE_HELPER_W(helper_pmulhw, FMULHW);
+
+SSE_HELPER_B(helper_pavgb, FAVG);
+SSE_HELPER_W(helper_pavgw, FAVG);
+
+void glue(helper_pmuludq, SUFFIX) (Reg *d, Reg *s);
+void glue(helper_pmaddwd, SUFFIX) (Reg *d, Reg *s);
+
+void glue(helper_psadbw, SUFFIX) (Reg *d, Reg *s);
+void glue(helper_maskmov, SUFFIX) (Reg *d, Reg *s);
+void glue(helper_movl_mm_T0, SUFFIX) (Reg *d, uint32_t val);
+#ifdef TARGET_X86_64
+void glue(helper_movq_mm_T0, SUFFIX) (Reg *d, uint64_t val);
+#endif
+
+#if SHIFT == 0
+void glue(helper_pshufw, SUFFIX) (Reg *d, Reg *s, int order);
+#else
+void helper_shufps(Reg *d, Reg *s, int order);
+void helper_shufpd(Reg *d, Reg *s, int order);
+void glue(helper_pshufd, SUFFIX) (Reg *d, Reg *s, int order);
+void glue(helper_pshuflw, SUFFIX) (Reg *d, Reg *s, int order);
+void glue(helper_pshufhw, SUFFIX) (Reg *d, Reg *s, int order);
+#endif
+
+#if SHIFT == 1
+/* FPU ops */
+/* XXX: not accurate */
+
+#define SSE_HELPER_S(name, F)\
+    void helper_ ## name ## ps (Reg *d, Reg *s);        \
+    void helper_ ## name ## ss (Reg *d, Reg *s);        \
+    void helper_ ## name ## pd (Reg *d, Reg *s);        \
+    void helper_ ## name ## sd (Reg *d, Reg *s);
+
+SSE_HELPER_S(add, FPU_ADD);
+SSE_HELPER_S(sub, FPU_SUB);
+SSE_HELPER_S(mul, FPU_MUL);
+SSE_HELPER_S(div, FPU_DIV);
+SSE_HELPER_S(min, FPU_MIN);
+SSE_HELPER_S(max, FPU_MAX);
+SSE_HELPER_S(sqrt, FPU_SQRT);
+
+
+void helper_cvtps2pd(Reg *d, Reg *s);
+void helper_cvtpd2ps(Reg *d, Reg *s);
+void helper_cvtss2sd(Reg *d, Reg *s);
+void helper_cvtsd2ss(Reg *d, Reg *s);
+void helper_cvtdq2ps(Reg *d, Reg *s);
+void helper_cvtdq2pd(Reg *d, Reg *s);
+void helper_cvtpi2ps(XMMReg *d, MMXReg *s);
+void helper_cvtpi2pd(XMMReg *d, MMXReg *s);
+void helper_cvtsi2ss(XMMReg *d, uint32_t val);
+void helper_cvtsi2sd(XMMReg *d, uint32_t val);
+
+#ifdef TARGET_X86_64
+void helper_cvtsq2ss(XMMReg *d, uint64_t val);
+void helper_cvtsq2sd(XMMReg *d, uint64_t val);
+#endif
+
+void helper_cvtps2dq(XMMReg *d, XMMReg *s);
+void helper_cvtpd2dq(XMMReg *d, XMMReg *s);
+void helper_cvtps2pi(MMXReg *d, XMMReg *s);
+void helper_cvtpd2pi(MMXReg *d, XMMReg *s);
+int32_t helper_cvtss2si(XMMReg *s);
+int32_t helper_cvtsd2si(XMMReg *s);
+#ifdef TARGET_X86_64
+int64_t helper_cvtss2sq(XMMReg *s);
+int64_t helper_cvtsd2sq(XMMReg *s);
+#endif
+
+void helper_cvttps2dq(XMMReg *d, XMMReg *s);
+void helper_cvttpd2dq(XMMReg *d, XMMReg *s);
+void helper_cvttps2pi(MMXReg *d, XMMReg *s);
+void helper_cvttpd2pi(MMXReg *d, XMMReg *s);
+int32_t helper_cvttss2si(XMMReg *s);
+int32_t helper_cvttsd2si(XMMReg *s);
+#ifdef TARGET_X86_64
+int64_t helper_cvttss2sq(XMMReg *s);
+int64_t helper_cvttsd2sq(XMMReg *s);
+#endif
+
+void helper_rsqrtps(XMMReg *d, XMMReg *s);
+void helper_rsqrtss(XMMReg *d, XMMReg *s);
+void helper_rcpps(XMMReg *d, XMMReg *s);
+void helper_rcpss(XMMReg *d, XMMReg *s);
+void helper_haddps(XMMReg *d, XMMReg *s);
+void helper_haddpd(XMMReg *d, XMMReg *s);
+void helper_hsubps(XMMReg *d, XMMReg *s);
+void helper_hsubpd(XMMReg *d, XMMReg *s);
+void helper_addsubps(XMMReg *d, XMMReg *s);
+void helper_addsubpd(XMMReg *d, XMMReg *s);
+
+#define SSE_HELPER_CMP(name, F)\
+    void helper_ ## name ## ps (Reg *d, Reg *s);        \
+    void helper_ ## name ## ss (Reg *d, Reg *s);        \
+    void helper_ ## name ## pd (Reg *d, Reg *s);        \
+    void helper_ ## name ## sd (Reg *d, Reg *s);
+
+SSE_HELPER_CMP(cmpeq, FPU_CMPEQ);
+SSE_HELPER_CMP(cmplt, FPU_CMPLT);
+SSE_HELPER_CMP(cmple, FPU_CMPLE);
+SSE_HELPER_CMP(cmpunord, FPU_CMPUNORD);
+SSE_HELPER_CMP(cmpneq, FPU_CMPNEQ);
+SSE_HELPER_CMP(cmpnlt, FPU_CMPNLT);
+SSE_HELPER_CMP(cmpnle, FPU_CMPNLE);
+SSE_HELPER_CMP(cmpord, FPU_CMPORD);
+
+void helper_ucomiss(Reg *d, Reg *s);
+void helper_comiss(Reg *d, Reg *s);
+void helper_ucomisd(Reg *d, Reg *s);
+void helper_comisd(Reg *d, Reg *s);
+uint32_t helper_movmskps(Reg *s);
+uint32_t helper_movmskpd(Reg *s);
+#endif
+
+uint32_t glue(helper_pmovmskb, SUFFIX)(Reg *s);
+void glue(helper_packsswb, SUFFIX) (Reg *d, Reg *s);
+void glue(helper_packuswb, SUFFIX) (Reg *d, Reg *s);
+void glue(helper_packssdw, SUFFIX) (Reg *d, Reg *s);
+#define UNPCK_OP(base_name, base)                               \
+    void glue(helper_punpck ## base_name ## bw, SUFFIX) (Reg *d, Reg *s); \
+    void glue(helper_punpck ## base_name ## wd, SUFFIX) (Reg *d, Reg *s); \
+    void glue(helper_punpck ## base_name ## dq, SUFFIX) (Reg *d, Reg *s);
+
+UNPCK_OP(l, 0);
+UNPCK_OP(h, 1);
+
+#if SHIFT == 1
+void glue(helper_punpcklqdq, SUFFIX) (Reg *d, Reg *s);
+void glue(helper_punpckhqdq, SUFFIX) (Reg *d, Reg *s);
+#endif
+
+/* 3DNow! float ops */
+#if SHIFT == 0
+void helper_pi2fd(MMXReg *d, MMXReg *s);
+void helper_pi2fw(MMXReg *d, MMXReg *s);
+void helper_pf2id(MMXReg *d, MMXReg *s);
+void helper_pf2iw(MMXReg *d, MMXReg *s);
+void helper_pfacc(MMXReg *d, MMXReg *s);
+void helper_pfadd(MMXReg *d, MMXReg *s);
+void helper_pfcmpeq(MMXReg *d, MMXReg *s);
+void helper_pfcmpge(MMXReg *d, MMXReg *s);
+void helper_pfcmpgt(MMXReg *d, MMXReg *s);
+void helper_pfmax(MMXReg *d, MMXReg *s);
+void helper_pfmin(MMXReg *d, MMXReg *s);
+void helper_pfmul(MMXReg *d, MMXReg *s);
+void helper_pfnacc(MMXReg *d, MMXReg *s);
+void helper_pfpnacc(MMXReg *d, MMXReg *s);
+void helper_pfrcp(MMXReg *d, MMXReg *s);
+void helper_pfrsqrt(MMXReg *d, MMXReg *s);
+void helper_pfsub(MMXReg *d, MMXReg *s);
+void helper_pfsubr(MMXReg *d, MMXReg *s);
+void helper_pswapd(MMXReg *d, MMXReg *s);
+#endif
+
+#undef SHIFT
+#undef Reg
+#undef SUFFIX
+
+#undef SSE_HELPER_B
+#undef SSE_HELPER_W
+#undef SSE_HELPER_L
+#undef SSE_HELPER_Q
+#undef SSE_HELPER_S
+#undef SSE_HELPER_CMP
+#undef UNPCK_OP
diff --git a/target-i386/translate.c b/target-i386/translate.c
index 44755949b1..846628b7a3 100644
--- a/target-i386/translate.c
+++ b/target-i386/translate.c
@@ -60,7 +60,7 @@
 /* global register indexes */
 static TCGv cpu_env, cpu_T[2], cpu_A0;
 /* local register indexes (only used inside old micro ops) */
-static TCGv cpu_tmp0, cpu_tmp1;
+static TCGv cpu_tmp0, cpu_tmp1, cpu_tmp2, cpu_ptr0, cpu_ptr1;
 
 #ifdef TARGET_X86_64
 static int x86_64_hregs;
@@ -2410,14 +2410,40 @@ static inline void gen_sto_env_A0(int idx, int offset)
     tcg_gen_qemu_st64(cpu_tmp1, cpu_tmp0, mem_index);
 }
 
-#define SSE_SPECIAL ((GenOpFunc2 *)1)
-#define SSE_DUMMY ((GenOpFunc2 *)2)
+static inline void gen_op_movo(int d_offset, int s_offset)
+{
+    tcg_gen_ld_i64(cpu_tmp1, cpu_env, s_offset);
+    tcg_gen_st_i64(cpu_tmp1, cpu_env, d_offset);
+    tcg_gen_ld_i64(cpu_tmp1, cpu_env, s_offset + 8);
+    tcg_gen_st_i64(cpu_tmp1, cpu_env, d_offset + 8);
+}
+
+static inline void gen_op_movq(int d_offset, int s_offset)
+{
+    tcg_gen_ld_i64(cpu_tmp1, cpu_env, s_offset);
+    tcg_gen_st_i64(cpu_tmp1, cpu_env, d_offset);
+}
+
+static inline void gen_op_movl(int d_offset, int s_offset)
+{
+    tcg_gen_ld_i32(cpu_tmp2, cpu_env, s_offset);
+    tcg_gen_st_i32(cpu_tmp2, cpu_env, d_offset);
+}
+
+static inline void gen_op_movq_env_0(int d_offset)
+{
+    tcg_gen_movi_i64(cpu_tmp1, 0);
+    tcg_gen_st_i64(cpu_tmp1, cpu_env, d_offset);
+}
 
-#define MMX_OP2(x) { gen_op_ ## x ## _mmx, gen_op_ ## x ## _xmm }
-#define SSE_FOP(x) { gen_op_ ## x ## ps, gen_op_ ## x ## pd, \
-                     gen_op_ ## x ## ss, gen_op_ ## x ## sd, }
+#define SSE_SPECIAL ((void *)1)
+#define SSE_DUMMY ((void *)2)
 
-static GenOpFunc2 *sse_op_table1[256][4] = {
+#define MMX_OP2(x) { helper_ ## x ## _mmx, helper_ ## x ## _xmm }
+#define SSE_FOP(x) { helper_ ## x ## ps, helper_ ## x ## pd, \
+                     helper_ ## x ## ss, helper_ ## x ## sd, }
+
+static void *sse_op_table1[256][4] = {
     /* 3DNow! extensions */
     [0x0e] = { SSE_DUMMY }, /* femms */
     [0x0f] = { SSE_DUMMY }, /* pf... */
@@ -2426,8 +2452,8 @@ static GenOpFunc2 *sse_op_table1[256][4] = {
     [0x11] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* movups, movupd, movss, movsd */
     [0x12] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* movlps, movlpd, movsldup, movddup */
     [0x13] = { SSE_SPECIAL, SSE_SPECIAL },  /* movlps, movlpd */
-    [0x14] = { gen_op_punpckldq_xmm, gen_op_punpcklqdq_xmm },
-    [0x15] = { gen_op_punpckhdq_xmm, gen_op_punpckhqdq_xmm },
+    [0x14] = { helper_punpckldq_xmm, helper_punpcklqdq_xmm },
+    [0x15] = { helper_punpckhdq_xmm, helper_punpckhqdq_xmm },
     [0x16] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL },  /* movhps, movhpd, movshdup */
     [0x17] = { SSE_SPECIAL, SSE_SPECIAL },  /* movhps, movhpd */
 
@@ -2437,28 +2463,28 @@ static GenOpFunc2 *sse_op_table1[256][4] = {
     [0x2b] = { SSE_SPECIAL, SSE_SPECIAL },  /* movntps, movntpd */
     [0x2c] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* cvttps2pi, cvttpd2pi, cvttsd2si, cvttss2si */
     [0x2d] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* cvtps2pi, cvtpd2pi, cvtsd2si, cvtss2si */
-    [0x2e] = { gen_op_ucomiss, gen_op_ucomisd },
-    [0x2f] = { gen_op_comiss, gen_op_comisd },
+    [0x2e] = { helper_ucomiss, helper_ucomisd },
+    [0x2f] = { helper_comiss, helper_comisd },
     [0x50] = { SSE_SPECIAL, SSE_SPECIAL }, /* movmskps, movmskpd */
     [0x51] = SSE_FOP(sqrt),
-    [0x52] = { gen_op_rsqrtps, NULL, gen_op_rsqrtss, NULL },
-    [0x53] = { gen_op_rcpps, NULL, gen_op_rcpss, NULL },
-    [0x54] = { gen_op_pand_xmm, gen_op_pand_xmm }, /* andps, andpd */
-    [0x55] = { gen_op_pandn_xmm, gen_op_pandn_xmm }, /* andnps, andnpd */
-    [0x56] = { gen_op_por_xmm, gen_op_por_xmm }, /* orps, orpd */
-    [0x57] = { gen_op_pxor_xmm, gen_op_pxor_xmm }, /* xorps, xorpd */
+    [0x52] = { helper_rsqrtps, NULL, helper_rsqrtss, NULL },
+    [0x53] = { helper_rcpps, NULL, helper_rcpss, NULL },
+    [0x54] = { helper_pand_xmm, helper_pand_xmm }, /* andps, andpd */
+    [0x55] = { helper_pandn_xmm, helper_pandn_xmm }, /* andnps, andnpd */
+    [0x56] = { helper_por_xmm, helper_por_xmm }, /* orps, orpd */
+    [0x57] = { helper_pxor_xmm, helper_pxor_xmm }, /* xorps, xorpd */
     [0x58] = SSE_FOP(add),
     [0x59] = SSE_FOP(mul),
-    [0x5a] = { gen_op_cvtps2pd, gen_op_cvtpd2ps,
-               gen_op_cvtss2sd, gen_op_cvtsd2ss },
-    [0x5b] = { gen_op_cvtdq2ps, gen_op_cvtps2dq, gen_op_cvttps2dq },
+    [0x5a] = { helper_cvtps2pd, helper_cvtpd2ps,
+               helper_cvtss2sd, helper_cvtsd2ss },
+    [0x5b] = { helper_cvtdq2ps, helper_cvtps2dq, helper_cvttps2dq },
     [0x5c] = SSE_FOP(sub),
     [0x5d] = SSE_FOP(min),
     [0x5e] = SSE_FOP(div),
     [0x5f] = SSE_FOP(max),
 
     [0xc2] = SSE_FOP(cmpeq),
-    [0xc6] = { (GenOpFunc2 *)gen_op_shufps, (GenOpFunc2 *)gen_op_shufpd },
+    [0xc6] = { helper_shufps, helper_shufpd },
 
     /* MMX ops and their SSE extensions */
     [0x60] = MMX_OP2(punpcklbw),
@@ -2473,14 +2499,14 @@ static GenOpFunc2 *sse_op_table1[256][4] = {
     [0x69] = MMX_OP2(punpckhwd),
     [0x6a] = MMX_OP2(punpckhdq),
     [0x6b] = MMX_OP2(packssdw),
-    [0x6c] = { NULL, gen_op_punpcklqdq_xmm },
-    [0x6d] = { NULL, gen_op_punpckhqdq_xmm },
+    [0x6c] = { NULL, helper_punpcklqdq_xmm },
+    [0x6d] = { NULL, helper_punpckhqdq_xmm },
     [0x6e] = { SSE_SPECIAL, SSE_SPECIAL }, /* movd mm, ea */
     [0x6f] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* movq, movdqa, , movqdu */
-    [0x70] = { (GenOpFunc2 *)gen_op_pshufw_mmx,
-               (GenOpFunc2 *)gen_op_pshufd_xmm,
-               (GenOpFunc2 *)gen_op_pshufhw_xmm,
-               (GenOpFunc2 *)gen_op_pshuflw_xmm },
+    [0x70] = { helper_pshufw_mmx,
+               helper_pshufd_xmm,
+               helper_pshufhw_xmm,
+               helper_pshuflw_xmm },
     [0x71] = { SSE_SPECIAL, SSE_SPECIAL }, /* shiftw */
     [0x72] = { SSE_SPECIAL, SSE_SPECIAL }, /* shiftd */
     [0x73] = { SSE_SPECIAL, SSE_SPECIAL }, /* shiftq */
@@ -2488,13 +2514,13 @@ static GenOpFunc2 *sse_op_table1[256][4] = {
     [0x75] = MMX_OP2(pcmpeqw),
     [0x76] = MMX_OP2(pcmpeql),
     [0x77] = { SSE_DUMMY }, /* emms */
-    [0x7c] = { NULL, gen_op_haddpd, NULL, gen_op_haddps },
-    [0x7d] = { NULL, gen_op_hsubpd, NULL, gen_op_hsubps },
+    [0x7c] = { NULL, helper_haddpd, NULL, helper_haddps },
+    [0x7d] = { NULL, helper_hsubpd, NULL, helper_hsubps },
     [0x7e] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* movd, movd, , movq */
     [0x7f] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* movq, movdqa, movdqu */
     [0xc4] = { SSE_SPECIAL, SSE_SPECIAL }, /* pinsrw */
     [0xc5] = { SSE_SPECIAL, SSE_SPECIAL }, /* pextrw */
-    [0xd0] = { NULL, gen_op_addsubpd, NULL, gen_op_addsubps },
+    [0xd0] = { NULL, helper_addsubpd, NULL, helper_addsubps },
     [0xd1] = MMX_OP2(psrlw),
     [0xd2] = MMX_OP2(psrld),
     [0xd3] = MMX_OP2(psrlq),
@@ -2516,7 +2542,7 @@ static GenOpFunc2 *sse_op_table1[256][4] = {
     [0xe3] = MMX_OP2(pavgw),
     [0xe4] = MMX_OP2(pmulhuw),
     [0xe5] = MMX_OP2(pmulhw),
-    [0xe6] = { NULL, gen_op_cvttpd2dq, gen_op_cvtdq2pd, gen_op_cvtpd2dq },
+    [0xe6] = { NULL, helper_cvttpd2dq, helper_cvtdq2pd, helper_cvtpd2dq },
     [0xe7] = { SSE_SPECIAL , SSE_SPECIAL },  /* movntq, movntq */
     [0xe8] = MMX_OP2(psubsb),
     [0xe9] = MMX_OP2(psubsw),
@@ -2543,7 +2569,7 @@ static GenOpFunc2 *sse_op_table1[256][4] = {
     [0xfe] = MMX_OP2(paddl),
 };
 
-static GenOpFunc2 *sse_op_table2[3 * 8][2] = {
+static void *sse_op_table2[3 * 8][2] = {
     [0 + 2] = MMX_OP2(psrlw),
     [0 + 4] = MMX_OP2(psraw),
     [0 + 6] = MMX_OP2(psllw),
@@ -2551,29 +2577,29 @@ static GenOpFunc2 *sse_op_table2[3 * 8][2] = {
     [8 + 4] = MMX_OP2(psrad),
     [8 + 6] = MMX_OP2(pslld),
     [16 + 2] = MMX_OP2(psrlq),
-    [16 + 3] = { NULL, gen_op_psrldq_xmm },
+    [16 + 3] = { NULL, helper_psrldq_xmm },
     [16 + 6] = MMX_OP2(psllq),
-    [16 + 7] = { NULL, gen_op_pslldq_xmm },
+    [16 + 7] = { NULL, helper_pslldq_xmm },
 };
 
-static GenOpFunc1 *sse_op_table3[4 * 3] = {
-    gen_op_cvtsi2ss,
-    gen_op_cvtsi2sd,
-    X86_64_ONLY(gen_op_cvtsq2ss),
-    X86_64_ONLY(gen_op_cvtsq2sd),
-
-    gen_op_cvttss2si,
-    gen_op_cvttsd2si,
-    X86_64_ONLY(gen_op_cvttss2sq),
-    X86_64_ONLY(gen_op_cvttsd2sq),
-
-    gen_op_cvtss2si,
-    gen_op_cvtsd2si,
-    X86_64_ONLY(gen_op_cvtss2sq),
-    X86_64_ONLY(gen_op_cvtsd2sq),
+static void *sse_op_table3[4 * 3] = {
+    helper_cvtsi2ss,
+    helper_cvtsi2sd,
+    X86_64_ONLY(helper_cvtsq2ss),
+    X86_64_ONLY(helper_cvtsq2sd),
+
+    helper_cvttss2si,
+    helper_cvttsd2si,
+    X86_64_ONLY(helper_cvttss2sq),
+    X86_64_ONLY(helper_cvttsd2sq),
+
+    helper_cvtss2si,
+    helper_cvtsd2si,
+    X86_64_ONLY(helper_cvtss2sq),
+    X86_64_ONLY(helper_cvtsd2sq),
 };
 
-static GenOpFunc2 *sse_op_table4[8][4] = {
+static void *sse_op_table4[8][4] = {
     SSE_FOP(cmpeq),
     SSE_FOP(cmplt),
     SSE_FOP(cmple),
@@ -2584,39 +2610,38 @@ static GenOpFunc2 *sse_op_table4[8][4] = {
     SSE_FOP(cmpord),
 };
 
-static GenOpFunc2 *sse_op_table5[256] = {
-    [0x0c] = gen_op_pi2fw,
-    [0x0d] = gen_op_pi2fd,
-    [0x1c] = gen_op_pf2iw,
-    [0x1d] = gen_op_pf2id,
-    [0x8a] = gen_op_pfnacc,
-    [0x8e] = gen_op_pfpnacc,
-    [0x90] = gen_op_pfcmpge,
-    [0x94] = gen_op_pfmin,
-    [0x96] = gen_op_pfrcp,
-    [0x97] = gen_op_pfrsqrt,
-    [0x9a] = gen_op_pfsub,
-    [0x9e] = gen_op_pfadd,
-    [0xa0] = gen_op_pfcmpgt,
-    [0xa4] = gen_op_pfmax,
-    [0xa6] = gen_op_movq, /* pfrcpit1; no need to actually increase precision */
-    [0xa7] = gen_op_movq, /* pfrsqit1 */
-    [0xaa] = gen_op_pfsubr,
-    [0xae] = gen_op_pfacc,
-    [0xb0] = gen_op_pfcmpeq,
-    [0xb4] = gen_op_pfmul,
-    [0xb6] = gen_op_movq, /* pfrcpit2 */
-    [0xb7] = gen_op_pmulhrw_mmx,
-    [0xbb] = gen_op_pswapd,
-    [0xbf] = gen_op_pavgb_mmx /* pavgusb */
+static void *sse_op_table5[256] = {
+    [0x0c] = helper_pi2fw,
+    [0x0d] = helper_pi2fd,
+    [0x1c] = helper_pf2iw,
+    [0x1d] = helper_pf2id,
+    [0x8a] = helper_pfnacc,
+    [0x8e] = helper_pfpnacc,
+    [0x90] = helper_pfcmpge,
+    [0x94] = helper_pfmin,
+    [0x96] = helper_pfrcp,
+    [0x97] = helper_pfrsqrt,
+    [0x9a] = helper_pfsub,
+    [0x9e] = helper_pfadd,
+    [0xa0] = helper_pfcmpgt,
+    [0xa4] = helper_pfmax,
+    [0xa6] = helper_movq, /* pfrcpit1; no need to actually increase precision */
+    [0xa7] = helper_movq, /* pfrsqit1 */
+    [0xaa] = helper_pfsubr,
+    [0xae] = helper_pfacc,
+    [0xb0] = helper_pfcmpeq,
+    [0xb4] = helper_pfmul,
+    [0xb6] = helper_movq, /* pfrcpit2 */
+    [0xb7] = helper_pmulhrw_mmx,
+    [0xbb] = helper_pswapd,
+    [0xbf] = helper_pavgb_mmx /* pavgusb */
 };
 
 static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
 {
     int b1, op1_offset, op2_offset, is_xmm, val, ot;
     int modrm, mod, rm, reg, reg_addr, offset_addr;
-    GenOpFunc2 *sse_op2;
-    GenOpFunc3 *sse_op3;
+    void *sse_op2;
 
     b &= 0xff;
     if (s->prefix & PREFIX_DATA)
@@ -2656,18 +2681,18 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
         if (!(s->cpuid_ext2_features & CPUID_EXT2_3DNOW))
             goto illegal_op;
         /* femms */
-        gen_op_emms();
+        tcg_gen_helper_0_0(helper_emms);
         return;
     }
     if (b == 0x77) {
         /* emms */
-        gen_op_emms();
+        tcg_gen_helper_0_0(helper_emms);
         return;
     }
     /* prepare MMX state (XXX: optimize by storing fptt and fptags in
        the static cpu state) */
     if (!is_xmm) {
-        gen_op_enter_mmx();
+        tcg_gen_helper_0_0(helper_enter_mmx);
     }
 
     modrm = ldub_code(s->pc++);
@@ -2697,24 +2722,31 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
 #ifdef TARGET_X86_64
             if (s->dflag == 2) {
                 gen_ldst_modrm(s, modrm, OT_QUAD, OR_TMP0, 0);
-                gen_op_movq_mm_T0_mmx(offsetof(CPUX86State,fpregs[reg].mmx));
+                tcg_gen_st_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,fpregs[reg].mmx));
             } else
 #endif
             {
                 gen_ldst_modrm(s, modrm, OT_LONG, OR_TMP0, 0);
-                gen_op_movl_mm_T0_mmx(offsetof(CPUX86State,fpregs[reg].mmx));
+                tcg_gen_addi_ptr(cpu_ptr0, cpu_env, 
+                                 offsetof(CPUX86State,fpregs[reg].mmx));
+                tcg_gen_helper_0_2(helper_movl_mm_T0_mmx, cpu_ptr0, cpu_T[0]);
             }
             break;
         case 0x16e: /* movd xmm, ea */
 #ifdef TARGET_X86_64
             if (s->dflag == 2) {
                 gen_ldst_modrm(s, modrm, OT_QUAD, OR_TMP0, 0);
-                gen_op_movq_mm_T0_xmm(offsetof(CPUX86State,xmm_regs[reg]));
+                tcg_gen_addi_ptr(cpu_ptr0, cpu_env, 
+                                 offsetof(CPUX86State,xmm_regs[reg]));
+                tcg_gen_helper_0_2(helper_movq_mm_T0_xmm, cpu_ptr0, cpu_T[0]);
             } else
 #endif
             {
                 gen_ldst_modrm(s, modrm, OT_LONG, OR_TMP0, 0);
-                gen_op_movl_mm_T0_xmm(offsetof(CPUX86State,xmm_regs[reg]));
+                tcg_gen_addi_ptr(cpu_ptr0, cpu_env, 
+                                 offsetof(CPUX86State,xmm_regs[reg]));
+                tcg_gen_trunc_tl_i32(cpu_tmp2, cpu_T[0]);
+                tcg_gen_helper_0_2(helper_movl_mm_T0_xmm, cpu_ptr0, cpu_tmp2);
             }
             break;
         case 0x6f: /* movq mm, ea */
@@ -2723,8 +2755,10 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
                 gen_ldq_env_A0(s->mem_index, offsetof(CPUX86State,fpregs[reg].mmx));
             } else {
                 rm = (modrm & 7);
-                gen_op_movq(offsetof(CPUX86State,fpregs[reg].mmx),
-                            offsetof(CPUX86State,fpregs[rm].mmx));
+                tcg_gen_ld_i64(cpu_tmp1, cpu_env,
+                               offsetof(CPUX86State,fpregs[rm].mmx));
+                tcg_gen_st_i64(cpu_tmp1, cpu_env,
+                               offsetof(CPUX86State,fpregs[reg].mmx));
             }
             break;
         case 0x010: /* movups */
@@ -2841,24 +2875,28 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
         case 0x7e: /* movd ea, mm */
 #ifdef TARGET_X86_64
             if (s->dflag == 2) {
-                gen_op_movq_T0_mm_mmx(offsetof(CPUX86State,fpregs[reg].mmx));
+                tcg_gen_ld_i64(cpu_T[0], cpu_env, 
+                               offsetof(CPUX86State,fpregs[reg].mmx));
                 gen_ldst_modrm(s, modrm, OT_QUAD, OR_TMP0, 1);
             } else
 #endif
             {
-                gen_op_movl_T0_mm_mmx(offsetof(CPUX86State,fpregs[reg].mmx));
+                tcg_gen_ld32u_tl(cpu_T[0], cpu_env, 
+                                 offsetof(CPUX86State,fpregs[reg].mmx.MMX_L(0)));
                 gen_ldst_modrm(s, modrm, OT_LONG, OR_TMP0, 1);
             }
             break;
         case 0x17e: /* movd ea, xmm */
 #ifdef TARGET_X86_64
             if (s->dflag == 2) {
-                gen_op_movq_T0_mm_xmm(offsetof(CPUX86State,xmm_regs[reg]));
+                tcg_gen_ld_i64(cpu_T[0], cpu_env, 
+                               offsetof(CPUX86State,xmm_regs[reg].XMM_Q(0)));
                 gen_ldst_modrm(s, modrm, OT_QUAD, OR_TMP0, 1);
             } else
 #endif
             {
-                gen_op_movl_T0_mm_xmm(offsetof(CPUX86State,xmm_regs[reg]));
+                tcg_gen_ld32u_tl(cpu_T[0], cpu_env, 
+                                 offsetof(CPUX86State,xmm_regs[reg].XMM_L(0)));
                 gen_ldst_modrm(s, modrm, OT_LONG, OR_TMP0, 1);
             }
             break;
@@ -2967,21 +3005,29 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
                 rm = (modrm & 7);
                 op2_offset = offsetof(CPUX86State,fpregs[rm].mmx);
             }
-            sse_op2(op2_offset, op1_offset);
+            tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op2_offset);
+            tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op1_offset);
+            tcg_gen_helper_0_2(sse_op2, cpu_ptr0, cpu_ptr1);
             break;
         case 0x050: /* movmskps */
             rm = (modrm & 7) | REX_B(s);
-            gen_op_movmskps(offsetof(CPUX86State,xmm_regs[rm]));
+            tcg_gen_addi_ptr(cpu_ptr0, cpu_env, 
+                             offsetof(CPUX86State,xmm_regs[rm]));
+            tcg_gen_helper_1_1(helper_movmskps, cpu_tmp2, cpu_ptr0);
+            tcg_gen_extu_i32_i64(cpu_T[0], cpu_tmp2);
             gen_op_mov_reg_T0(OT_LONG, reg);
             break;
         case 0x150: /* movmskpd */
             rm = (modrm & 7) | REX_B(s);
-            gen_op_movmskpd(offsetof(CPUX86State,xmm_regs[rm]));
+            tcg_gen_addi_ptr(cpu_ptr0, cpu_env, 
+                             offsetof(CPUX86State,xmm_regs[rm]));
+            tcg_gen_helper_1_1(helper_movmskpd, cpu_tmp2, cpu_ptr0);
+            tcg_gen_extu_i32_i64(cpu_T[0], cpu_tmp2);
             gen_op_mov_reg_T0(OT_LONG, reg);
             break;
         case 0x02a: /* cvtpi2ps */
         case 0x12a: /* cvtpi2pd */
-            gen_op_enter_mmx();
+            tcg_gen_helper_0_0(helper_enter_mmx);
             if (mod != 3) {
                 gen_lea_modrm(s, modrm, &reg_addr, &offset_addr);
                 op2_offset = offsetof(CPUX86State,mmx_t0);
@@ -2991,13 +3037,15 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
                 op2_offset = offsetof(CPUX86State,fpregs[rm].mmx);
             }
             op1_offset = offsetof(CPUX86State,xmm_regs[reg]);
+            tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
+            tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
             switch(b >> 8) {
             case 0x0:
-                gen_op_cvtpi2ps(op1_offset, op2_offset);
+                tcg_gen_helper_0_2(helper_cvtpi2ps, cpu_ptr0, cpu_ptr1);
                 break;
             default:
             case 0x1:
-                gen_op_cvtpi2pd(op1_offset, op2_offset);
+                tcg_gen_helper_0_2(helper_cvtpi2pd, cpu_ptr0, cpu_ptr1);
                 break;
             }
             break;
@@ -3006,13 +3054,16 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
             ot = (s->dflag == 2) ? OT_QUAD : OT_LONG;
             gen_ldst_modrm(s, modrm, ot, OR_TMP0, 0);
             op1_offset = offsetof(CPUX86State,xmm_regs[reg]);
-            sse_op_table3[(s->dflag == 2) * 2 + ((b >> 8) - 2)](op1_offset);
+            tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
+            sse_op2 = sse_op_table3[(s->dflag == 2) * 2 + ((b >> 8) - 2)];
+            tcg_gen_trunc_tl_i32(cpu_tmp2, cpu_T[0]);
+            tcg_gen_helper_0_2(sse_op2, cpu_ptr0, cpu_tmp2);
             break;
         case 0x02c: /* cvttps2pi */
         case 0x12c: /* cvttpd2pi */
         case 0x02d: /* cvtps2pi */
         case 0x12d: /* cvtpd2pi */
-            gen_op_enter_mmx();
+            tcg_gen_helper_0_0(helper_enter_mmx);
             if (mod != 3) {
                 gen_lea_modrm(s, modrm, &reg_addr, &offset_addr);
                 op2_offset = offsetof(CPUX86State,xmm_t0);
@@ -3022,18 +3073,20 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
                 op2_offset = offsetof(CPUX86State,xmm_regs[rm]);
             }
             op1_offset = offsetof(CPUX86State,fpregs[reg & 7].mmx);
+            tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
+            tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
             switch(b) {
             case 0x02c:
-                gen_op_cvttps2pi(op1_offset, op2_offset);
+                tcg_gen_helper_0_2(helper_cvttps2pi, cpu_ptr0, cpu_ptr1);
                 break;
             case 0x12c:
-                gen_op_cvttpd2pi(op1_offset, op2_offset);
+                tcg_gen_helper_0_2(helper_cvttpd2pi, cpu_ptr0, cpu_ptr1);
                 break;
             case 0x02d:
-                gen_op_cvtps2pi(op1_offset, op2_offset);
+                tcg_gen_helper_0_2(helper_cvtps2pi, cpu_ptr0, cpu_ptr1);
                 break;
             case 0x12d:
-                gen_op_cvtpd2pi(op1_offset, op2_offset);
+                tcg_gen_helper_0_2(helper_cvtpd2pi, cpu_ptr0, cpu_ptr1);
                 break;
             }
             break;
@@ -3055,8 +3108,15 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
                 rm = (modrm & 7) | REX_B(s);
                 op2_offset = offsetof(CPUX86State,xmm_regs[rm]);
             }
-            sse_op_table3[(s->dflag == 2) * 2 + ((b >> 8) - 2) + 4 +
-                          (b & 1) * 4](op2_offset);
+            sse_op2 = sse_op_table3[(s->dflag == 2) * 2 + ((b >> 8) - 2) + 4 +
+                                    (b & 1) * 4];
+            tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op2_offset);
+            if (ot == OT_LONG) {
+                tcg_gen_helper_1_1(sse_op2, cpu_tmp2, cpu_ptr0);
+                tcg_gen_extu_i32_i64(cpu_T[0], cpu_tmp2);
+            } else {
+                tcg_gen_helper_1_1(sse_op2, cpu_T[0], cpu_ptr0);
+            }
             gen_op_mov_reg_T0(ot, reg);
             break;
         case 0xc4: /* pinsrw */
@@ -3066,10 +3126,12 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
             val = ldub_code(s->pc++);
             if (b1) {
                 val &= 7;
-                gen_op_pinsrw_xmm(offsetof(CPUX86State,xmm_regs[reg]), val);
+                tcg_gen_st16_tl(cpu_T[0], cpu_env,
+                                offsetof(CPUX86State,xmm_regs[reg].XMM_W(val)));
             } else {
                 val &= 3;
-                gen_op_pinsrw_mmx(offsetof(CPUX86State,fpregs[reg].mmx), val);
+                tcg_gen_st16_tl(cpu_T[0], cpu_env,
+                                offsetof(CPUX86State,fpregs[reg].mmx.MMX_W(val)));
             }
             break;
         case 0xc5: /* pextrw */
@@ -3080,11 +3142,13 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
             if (b1) {
                 val &= 7;
                 rm = (modrm & 7) | REX_B(s);
-                gen_op_pextrw_xmm(offsetof(CPUX86State,xmm_regs[rm]), val);
+                tcg_gen_ld16u_tl(cpu_T[0], cpu_env,
+                                 offsetof(CPUX86State,xmm_regs[rm].XMM_W(val)));
             } else {
                 val &= 3;
                 rm = (modrm & 7);
-                gen_op_pextrw_mmx(offsetof(CPUX86State,fpregs[rm].mmx), val);
+                tcg_gen_ld16u_tl(cpu_T[0], cpu_env,
+                                offsetof(CPUX86State,fpregs[rm].mmx.MMX_W(val)));
             }
             reg = ((modrm >> 3) & 7) | rex_r;
             gen_op_mov_reg_T0(OT_LONG, reg);
@@ -3101,14 +3165,14 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
             }
             break;
         case 0x2d6: /* movq2dq */
-            gen_op_enter_mmx();
+            tcg_gen_helper_0_0(helper_enter_mmx);
             rm = (modrm & 7);
             gen_op_movq(offsetof(CPUX86State,xmm_regs[reg].XMM_Q(0)),
                         offsetof(CPUX86State,fpregs[rm].mmx));
             gen_op_movq_env_0(offsetof(CPUX86State,xmm_regs[reg].XMM_Q(1)));
             break;
         case 0x3d6: /* movdq2q */
-            gen_op_enter_mmx();
+            tcg_gen_helper_0_0(helper_enter_mmx);
             rm = (modrm & 7) | REX_B(s);
             gen_op_movq(offsetof(CPUX86State,fpregs[reg & 7].mmx),
                         offsetof(CPUX86State,xmm_regs[rm].XMM_Q(0)));
@@ -3119,11 +3183,14 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
                 goto illegal_op;
             if (b1) {
                 rm = (modrm & 7) | REX_B(s);
-                gen_op_pmovmskb_xmm(offsetof(CPUX86State,xmm_regs[rm]));
+                tcg_gen_addi_ptr(cpu_ptr0, cpu_env, offsetof(CPUX86State,xmm_regs[rm]));
+                tcg_gen_helper_1_1(helper_pmovmskb_xmm, cpu_tmp2, cpu_ptr0);
             } else {
                 rm = (modrm & 7);
-                gen_op_pmovmskb_mmx(offsetof(CPUX86State,fpregs[rm].mmx));
+                tcg_gen_addi_ptr(cpu_ptr0, cpu_env, offsetof(CPUX86State,fpregs[rm].mmx));
+                tcg_gen_helper_1_1(helper_pmovmskb_mmx, cpu_tmp2, cpu_ptr0);
             }
+            tcg_gen_extu_i32_tl(cpu_T[0], cpu_tmp2);
             reg = ((modrm >> 3) & 7) | rex_r;
             gen_op_mov_reg_T0(OT_LONG, reg);
             break;
@@ -3199,13 +3266,16 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
             sse_op2 = sse_op_table5[val];
             if (!sse_op2)
                 goto illegal_op;
-            sse_op2(op1_offset, op2_offset);
+            tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
+            tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
+            tcg_gen_helper_0_2(sse_op2, cpu_ptr0, cpu_ptr1);
             break;
         case 0x70: /* pshufx insn */
         case 0xc6: /* pshufx insn */
             val = ldub_code(s->pc++);
-            sse_op3 = (GenOpFunc3 *)sse_op2;
-            sse_op3(op1_offset, op2_offset, val);
+            tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
+            tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
+            tcg_gen_helper_0_3(sse_op2, cpu_ptr0, cpu_ptr1, tcg_const_i32(val));
             break;
         case 0xc2:
             /* compare insns */
@@ -3213,13 +3283,19 @@ static void gen_sse(DisasContext *s, int b, target_ulong pc_start, int rex_r)
             if (val >= 8)
                 goto illegal_op;
             sse_op2 = sse_op_table4[val][b1];
-            sse_op2(op1_offset, op2_offset);
+            tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
+            tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
+            tcg_gen_helper_0_2(sse_op2, cpu_ptr0, cpu_ptr1);
             break;
         default:
-            sse_op2(op1_offset, op2_offset);
+            tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
+            tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
+            tcg_gen_helper_0_2(sse_op2, cpu_ptr0, cpu_ptr1);
             break;
         }
         if (b == 0x2e || b == 0x2f) {
+            /* just to keep the EFLAGS optimization correct */
+            gen_op_com_dummy();
             s->cc_op = CC_OP_EFLAGS;
         }
     }
@@ -6485,10 +6561,10 @@ static uint16_t opc_write_flags[NB_OPS] = {
     X86_64_DEF([INDEX_op_imulq_T0_T1] = CC_OSZAPC,)
 
     /* sse */
-    [INDEX_op_ucomiss] = CC_OSZAPC,
-    [INDEX_op_ucomisd] = CC_OSZAPC,
-    [INDEX_op_comiss] = CC_OSZAPC,
-    [INDEX_op_comisd] = CC_OSZAPC,
+    [INDEX_op_com_dummy] = CC_OSZAPC,
+    [INDEX_op_com_dummy] = CC_OSZAPC,
+    [INDEX_op_com_dummy] = CC_OSZAPC,
+    [INDEX_op_com_dummy] = CC_OSZAPC,
 
     /* bcd */
     [INDEX_op_aam] = CC_OSZAPC,
@@ -6792,6 +6868,9 @@ static inline int gen_intermediate_code_internal(CPUState *env,
 #if TARGET_LONG_BITS > HOST_LONG_BITS
     cpu_tmp1 = tcg_temp_new(TCG_TYPE_I64);
 #endif
+    cpu_tmp2 = tcg_temp_new(TCG_TYPE_I32);
+    cpu_ptr0 = tcg_temp_new(TCG_TYPE_PTR);
+    cpu_ptr1 = tcg_temp_new(TCG_TYPE_PTR);
 
     gen_opc_end = gen_opc_buf + OPC_MAX_SIZE;