diff --git a/target-ppc/op_helper.c b/target-ppc/op_helper.c
index e57fbe4..06fa6d4 100644
--- a/target-ppc/op_helper.c
+++ b/target-ppc/op_helper.c
@@ -29,6 +29,14 @@
 #include "op_helper_mem.h"
 #endif
 
+//#define DEBUG_OP
+//#define DEBUG_EXCEPTIONS
+//#define FLUSH_ALL_TLBS
+
+#define Ts0 (long)((target_long)T0)
+#define Ts1 (long)((target_long)T1)
+#define Ts2 (long)((target_long)T2)
+
 /*****************************************************************************/
 /* Exceptions processing helpers */
 void cpu_loop_exit(void)
@@ -60,29 +68,260 @@
 }
 
 /*****************************************************************************/
-/* Helpers for "fat" micro operations */
+/* Fixed point operations helpers */
+void do_addo (void)
+{
+    T2 = T0;
+    T0 += T1;
+    if (likely(!((T2 ^ T1 ^ (-1)) & (T2 ^ T0) & (1 << 31)))) {
+        xer_ov = 0;
+    } else {
+        xer_so = 1;
+        xer_ov = 1;
+    }
+}
+
+void do_addco (void)
+{
+    T2 = T0;
+    T0 += T1;
+    if (likely(T0 >= T2)) {
+        xer_ca = 0;
+    } else {
+        xer_ca = 1;
+    }
+    if (likely(!((T2 ^ T1 ^ (-1)) & (T2 ^ T0) & (1 << 31)))) {
+        xer_ov = 0;
+    } else {
+        xer_so = 1;
+        xer_ov = 1;
+    }
+}
+
+void do_adde (void)
+{
+    T2 = T0;
+    T0 += T1 + xer_ca;
+    if (likely(!(T0 < T2 || (xer_ca == 1 && T0 == T2)))) {
+        xer_ca = 0;
+    } else {
+        xer_ca = 1;
+    }
+}
+
+void do_addeo (void)
+{
+    T2 = T0;
+    T0 += T1 + xer_ca;
+    if (likely(!(T0 < T2 || (xer_ca == 1 && T0 == T2)))) {
+        xer_ca = 0;
+    } else {
+        xer_ca = 1;
+    }
+    if (likely(!((T2 ^ T1 ^ (-1)) & (T2 ^ T0) & (1 << 31)))) {
+        xer_ov = 0;
+    } else {
+        xer_so = 1;
+        xer_ov = 1;
+    }
+}
+
+void do_addmeo (void)
+{
+    T1 = T0;
+    T0 += xer_ca + (-1);
+    if (likely(!(T1 & (T1 ^ T0) & (1 << 31)))) {
+        xer_ov = 0;
+    } else {
+        xer_so = 1;
+        xer_ov = 1;
+    }
+    if (likely(T1 != 0))
+        xer_ca = 1;
+}
+
+void do_addzeo (void)
+{
+    T1 = T0;
+    T0 += xer_ca;
+    if (likely(!((T1 ^ (-1)) & (T1 ^ T0) & (1 << 31)))) {
+        xer_ov = 0;
+    } else {
+        xer_so = 1;
+        xer_ov = 1;
+    }
+    if (likely(T0 >= T1)) {
+        xer_ca = 0;
+    } else {
+        xer_ca = 1;
+    }
+}
+
+void do_divwo (void)
+{
+    if (likely(!((Ts0 == INT32_MIN && Ts1 == -1) || Ts1 == 0))) {
+        xer_ov = 0;
+        T0 = (Ts0 / Ts1);
+    } else {
+        xer_so = 1;
+        xer_ov = 1;
+        T0 = (-1) * ((uint32_t)T0 >> 31);
+    }
+}
+
+void do_divwuo (void)
+{
+    if (likely((uint32_t)T1 != 0)) {
+        xer_ov = 0;
+        T0 = (uint32_t)T0 / (uint32_t)T1;
+    } else {
+        xer_so = 1;
+        xer_ov = 1;
+        T0 = 0;
+    }
+}
+
+void do_mullwo (void)
+{
+    int64_t res = (int64_t)Ts0 * (int64_t)Ts1;
+
+    if (likely((int32_t)res == res)) {
+        xer_ov = 0;
+    } else {
+        xer_ov = 1;
+        xer_so = 1;
+    }
+    T0 = (int32_t)res;
+}
+
+void do_nego (void)
+{
+    if (likely(T0 != INT32_MIN)) {
+        xer_ov = 0;
+        T0 = -Ts0;
+    } else {
+        xer_ov = 1;
+        xer_so = 1;
+    }
+}
+
+void do_subfo (void)
+{
+    T2 = T0;
+    T0 = T1 - T0;
+    if (likely(!(((~T2) ^ T1 ^ (-1)) & ((~T2) ^ T0) & (1 << 31)))) {
+        xer_ov = 0;
+    } else {
+        xer_so = 1;
+        xer_ov = 1;
+    }
+    RETURN();
+}
+
+void do_subfco (void)
+{
+    T2 = T0;
+    T0 = T1 - T0;
+    if (likely(T0 > T1)) {
+        xer_ca = 0;
+    } else {
+        xer_ca = 1;
+    }
+    if (likely(!(((~T2) ^ T1 ^ (-1)) & ((~T2) ^ T0) & (1 << 31)))) {
+        xer_ov = 0;
+    } else {
+        xer_so = 1;
+        xer_ov = 1;
+    }
+}
+
+void do_subfe (void)
+{
+    T0 = T1 + ~T0 + xer_ca;
+    if (likely(T0 >= T1 && (xer_ca == 0 || T0 != T1))) {
+        xer_ca = 0;
+    } else {
+        xer_ca = 1;
+    }
+}
+
+void do_subfeo (void)
+{
+    T2 = T0;
+    T0 = T1 + ~T0 + xer_ca;
+    if (likely(!((~T2 ^ T1 ^ (-1)) & (~T2 ^ T0) & (1 << 31)))) {
+        xer_ov = 0;
+    } else {
+        xer_so = 1;
+        xer_ov = 1;
+    }
+    if (likely(T0 >= T1 && (xer_ca == 0 || T0 != T1))) {
+        xer_ca = 0;
+    } else {
+        xer_ca = 1;
+    }
+}
+
+void do_subfmeo (void)
+{
+    T1 = T0;
+    T0 = ~T0 + xer_ca - 1;
+    if (likely(!(~T1 & (~T1 ^ T0) & (1 << 31)))) {
+        xer_ov = 0;
+    } else {
+        xer_so = 1;
+        xer_ov = 1;
+    }
+    if (likely(T1 != -1))
+        xer_ca = 1;
+}
+
+void do_subfzeo (void)
+{
+    T1 = T0;
+    T0 = ~T0 + xer_ca;
+    if (likely(!((~T1 ^ (-1)) & ((~T1) ^ T0) & (1 << 31)))) {
+        xer_ov = 0;
+    } else {
+        xer_ov = 1;
+        xer_so = 1;
+    }
+    if (likely(T0 >= ~T1)) {
+        xer_ca = 0;
+    } else {
+        xer_ca = 1;
+    }
+}
+
 /* shift right arithmetic helper */
 void do_sraw (void)
 {
     int32_t ret;
 
+    if (likely(!(T1 & 0x20UL))) {
+        if (likely(T1 != 0)) {
+            ret = (int32_t)T0 >> (T1 & 0x1fUL);
+            if (likely(ret >= 0 || ((int32_t)T0 & ((1 << T1) - 1)) == 0)) {
     xer_ca = 0;
-    if (T1 & 0x20) {
-        ret = (-1) * (T0 >> 31);
-        if (ret < 0 && (T0 & ~0x80000000) != 0)
+            } else {
             xer_ca = 1;
-#if 1 // TRY
-    } else if (T1 == 0) {
+            }
+        } else {
         ret = T0;
-#endif
+            xer_ca = 0;
+        }
     } else {
-        ret = (int32_t)T0 >> (T1 & 0x1f);
-        if (ret < 0 && ((int32_t)T0 & ((1 << T1) - 1)) != 0)
+        ret = (-1) * ((uint32_t)T0 >> 31);
+        if (likely(ret >= 0 || ((uint32_t)T0 & ~0x80000000UL) == 0)) {
+            xer_ca = 0;
+    } else {
             xer_ca = 1;
     }
+    }
     T0 = ret;
 }
 
+/*****************************************************************************/
 /* Floating point operations helpers */
 void do_fctiw (void)
 {
@@ -116,29 +355,23 @@
 
 void do_fnmadd (void)
 {
-    FT0 = (FT0 * FT1) + FT2;
-    if (!isnan(FT0))
-        FT0 = -FT0;
+    FT0 = float64_mul(FT0, FT1, &env->fp_status);
+    FT0 = float64_add(FT0, FT2, &env->fp_status);
+    if (likely(!isnan(FT0)))
+        FT0 = float64_chs(FT0);
 }
 
 void do_fnmsub (void)
 {
-    FT0 = (FT0 * FT1) - FT2;
-    if (!isnan(FT0))
-        FT0 = -FT0;
-}
-
-void do_fdiv (void)
-{
-    if (FT0 == -0.0 && FT1 == -0.0)
-        FT0 = 0.0 / 0.0;
-    else
-        FT0 /= FT1;
+    FT0 = float64_mul(FT0, FT1, &env->fp_status);
+    FT0 = float64_sub(FT0, FT2, &env->fp_status);
+    if (likely(!isnan(FT0)))
+        FT0 = float64_chs(FT0);
 }
 
 void do_fsqrt (void)
 {
-    FT0 = sqrt(FT0);
+    FT0 = float64_sqrt(FT0, &env->fp_status);
 }
 
 void do_fres (void)
@@ -148,7 +381,7 @@
         uint64_t i;
     } p;
 
-    if (isnormal(FT0)) {
+    if (likely(isnormal(FT0))) {
         FT0 = (float)(1.0 / FT0);
     } else {
         p.d = FT0;
@@ -174,8 +407,9 @@
         uint64_t i;
     } p;
 
-    if (isnormal(FT0) && FT0 > 0.0) {
-        FT0 = (float)(1.0 / sqrt(FT0));
+    if (likely(isnormal(FT0) && FT0 > 0.0)) {
+        FT0 = float64_sqrt(FT0, &env->fp_status);
+        FT0 = float32_div(1.0, FT0, &env->fp_status);
     } else {
         p.d = FT0;
         if (p.i == 0x8000000000000000ULL) {
@@ -204,16 +438,18 @@
 
 void do_fcmpu (void)
 {
-    if (isnan(FT0) || isnan(FT1)) {
-        T0 = 0x01;
+    if (likely(!isnan(FT0) && !isnan(FT1))) {
+        if (float64_lt(FT0, FT1, &env->fp_status)) {
+            T0 = 0x08UL;
+        } else if (!float64_le(FT0, FT1, &env->fp_status)) {
+            T0 = 0x04UL;
+        } else {
+            T0 = 0x02UL;
+        }
+    } else {
+        T0 = 0x01UL;
         env->fpscr[4] |= 0x1;
         env->fpscr[6] |= 0x1;
-    } else if (FT0 < FT1) {
-        T0 = 0x08;
-    } else if (FT0 > FT1) {
-        T0 = 0x04;
-    } else {
-        T0 = 0x02;
     }
     env->fpscr[3] = T0;
 }
@@ -221,8 +457,16 @@
 void do_fcmpo (void)
 {
     env->fpscr[4] &= ~0x1;
-    if (isnan(FT0) || isnan(FT1)) {
-        T0 = 0x01;
+    if (likely(!isnan(FT0) && !isnan(FT1))) {
+        if (float64_lt(FT0, FT1, &env->fp_status)) {
+            T0 = 0x08UL;
+        } else if (!float64_le(FT0, FT1, &env->fp_status)) {
+            T0 = 0x04UL;
+        } else {
+            T0 = 0x02UL;
+        }
+    } else {
+        T0 = 0x01UL;
         env->fpscr[4] |= 0x1;
         /* I don't know how to test "quiet" nan... */
         if (0 /* || ! quiet_nan(...) */) {
@@ -232,56 +476,51 @@
         } else {
             env->fpscr[4] |= 0x8;
         }
-    } else if (FT0 < FT1) {
-        T0 = 0x08;
-    } else if (FT0 > FT1) {
-        T0 = 0x04;
-    } else {
-        T0 = 0x02;
     }
     env->fpscr[3] = T0;
 }
 
-void do_fabs (void)
+void do_rfi (void)
 {
-    union {
-        double d;
-        uint64_t i;
-    } p;
-
-    p.d = FT0;
-    p.i &= ~0x8000000000000000ULL;
-    FT0 = p.d;
+    env->nip = env->spr[SPR_SRR0] & ~0x00000003;
+    T0 = env->spr[SPR_SRR1] & ~0xFFFF0000UL;
+    do_store_msr(env, T0);
+#if defined (DEBUG_OP)
+    dump_rfi();
+#endif
+    env->interrupt_request |= CPU_INTERRUPT_EXITTB;
 }
 
-void do_fnabs (void)
+void do_tw (uint32_t cmp, int flags)
 {
-    union {
-        double d;
-        uint64_t i;
-    } p;
-
-    p.d = FT0;
-    p.i |= 0x8000000000000000ULL;
-    FT0 = p.d;
+    if (!likely(!((Ts0 < (int32_t)cmp && (flags & 0x10)) ||
+                  (Ts0 > (int32_t)cmp && (flags & 0x08)) ||
+                  (Ts0 == (int32_t)cmp && (flags & 0x04)) ||
+                  (T0 < cmp && (flags & 0x02)) ||
+                  (T0 > cmp && (flags & 0x01)))))
+        do_raise_exception_err(EXCP_PROGRAM, EXCP_TRAP);
 }
 
 /* Instruction cache invalidation helper */
-#define ICACHE_LINE_SIZE 32
-
-void do_check_reservation (void)
-{
-    if ((env->reserve & ~0x03) == T0)
-        env->reserve = -1;
-}
-
 void do_icbi (void)
 {
-    /* Invalidate one cache line */
+    uint32_t tmp;
+    /* Invalidate one cache line :
+     * PowerPC specification says this is to be treated like a load
+     * (not a fetch) by the MMU. To be sure it will be so,
+     * do the load "by hand".
+     */
+#if defined(TARGET_PPC64)
+    if (!msr_sf)
+        T0 &= 0xFFFFFFFFULL;
+#endif
+    tmp = ldl_kernel(T0);
     T0 &= ~(ICACHE_LINE_SIZE - 1);
     tb_invalidate_page_range(T0, T0 + ICACHE_LINE_SIZE);
 }
 
+/*****************************************************************************/
+/* MMU related helpers */
 /* TLB invalidation helpers */
 void do_tlbia (void)
 {
@@ -290,6 +529,62 @@
 
 void do_tlbie (void)
 {
+#if !defined(FLUSH_ALL_TLBS)
     tlb_flush_page(env, T0);
+#else
+    do_tlbia();
+#endif
 }
 
+/*****************************************************************************/
+/* Softmmu support */
+#if !defined (CONFIG_USER_ONLY)
+
+#define MMUSUFFIX _mmu
+#define GETPC() (__builtin_return_address(0))
+
+#define SHIFT 0
+#include "softmmu_template.h"
+
+#define SHIFT 1
+#include "softmmu_template.h"
+
+#define SHIFT 2
+#include "softmmu_template.h"
+
+#define SHIFT 3
+#include "softmmu_template.h"
+
+/* try to fill the TLB and return an exception if error. If retaddr is
+   NULL, it means that the function was called in C code (i.e. not
+   from generated code or from helper.c) */
+/* XXX: fix it to restore all registers */
+void tlb_fill (target_ulong addr, int is_write, int is_user, void *retaddr)
+{
+    TranslationBlock *tb;
+    CPUState *saved_env;
+    target_phys_addr_t pc;
+    int ret;
+
+    /* XXX: hack to restore env in all cases, even if not called from
+       generated code */
+    saved_env = env;
+    env = cpu_single_env;
+    ret = cpu_ppc_handle_mmu_fault(env, addr, is_write, is_user, 1);
+    if (!likely(ret == 0)) {
+        if (likely(retaddr)) {
+            /* now we have a real cpu fault */
+            pc = (target_phys_addr_t)retaddr;
+            tb = tb_find_pc(pc);
+            if (likely(tb)) {
+                /* the PC is inside the translated code. It means that we have
+                   a virtual CPU fault */
+                cpu_restore_state(tb, env, pc, NULL);
+}
+        }
+        do_raise_exception_err(env->exception_index, env->error_code);
+    }
+    env = saved_env;
+}
+#endif /* !CONFIG_USER_ONLY */
+