convert of few alpha insn to TCG
[qemu] / target-mips / translate.c
index 6734390..ec11d97 100644 (file)
@@ -423,53 +423,102 @@ enum {
 };
 
 /* global register indices */
-static TCGv cpu_env, current_tc_gprs, cpu_T[2];
+static TCGv cpu_env, bcond, btarget, current_fpu;
 
-/* The code generator doesn't like lots of temporaries, so maintain our own
-   cache for reuse within a function.  */
-#define MAX_TEMPS 4
-static int num_temps;
-static TCGv temps[MAX_TEMPS];
+#include "gen-icount.h"
 
-/* Allocate a temporary variable.  */
-static TCGv new_tmp(void)
+static inline void tcg_gen_helper_0_i(void *func, TCGv arg)
 {
-    TCGv tmp;
-    if (num_temps == MAX_TEMPS)
-        abort();
+    TCGv tmp = tcg_const_i32(arg);
 
-    if (GET_TCGV(temps[num_temps]))
-      return temps[num_temps++];
+    tcg_gen_helper_0_1(func, tmp);
+    tcg_temp_free(tmp);
+}
+
+static inline void tcg_gen_helper_0_ii(void *func, TCGv arg1, TCGv arg2)
+{
+    TCGv tmp1 = tcg_const_i32(arg1);
+    TCGv tmp2 = tcg_const_i32(arg2);
 
-    tmp = tcg_temp_new(TCG_TYPE_I32);
-    temps[num_temps++] = tmp;
-    return tmp;
+    tcg_gen_helper_0_2(func, tmp1, tmp2);
+    tcg_temp_free(tmp1);
+    tcg_temp_free(tmp2);
 }
 
-/* Release a temporary variable.  */
-static void dead_tmp(TCGv tmp)
+static inline void tcg_gen_helper_0_1i(void *func, TCGv arg1, TCGv arg2)
 {
-    int i;
-    num_temps--;
-    i = num_temps;
-    if (GET_TCGV(temps[i]) == GET_TCGV(tmp))
-        return;
+    TCGv tmp = tcg_const_i32(arg2);
 
-    /* Shuffle this temp to the last slot.  */
-    while (GET_TCGV(temps[i]) != GET_TCGV(tmp))
-        i--;
-    while (i < num_temps) {
-        temps[i] = temps[i + 1];
-        i++;
-    }
-    temps[i] = tmp;
+    tcg_gen_helper_0_2(func, arg1, tmp);
+    tcg_temp_free(tmp);
+}
+
+static inline void tcg_gen_helper_0_2i(void *func, TCGv arg1, TCGv arg2, TCGv arg3)
+{
+    TCGv tmp = tcg_const_i32(arg3);
+
+    tcg_gen_helper_0_3(func, arg1, arg2, tmp);
+    tcg_temp_free(tmp);
+}
+
+static inline void tcg_gen_helper_0_1ii(void *func, TCGv arg1, TCGv arg2, TCGv arg3)
+{
+    TCGv tmp1 = tcg_const_i32(arg2);
+    TCGv tmp2 = tcg_const_i32(arg3);
+
+    tcg_gen_helper_0_3(func, arg1, tmp1, tmp2);
+    tcg_temp_free(tmp1);
+    tcg_temp_free(tmp2);
+}
+
+static inline void tcg_gen_helper_1_i(void *func, TCGv ret, TCGv arg)
+{
+    TCGv tmp = tcg_const_i32(arg);
+
+    tcg_gen_helper_1_1(func, ret, tmp);
+    tcg_temp_free(tmp);
+}
+
+static inline void tcg_gen_helper_1_1i(void *func, TCGv ret, TCGv arg1, TCGv arg2)
+{
+    TCGv tmp = tcg_const_i32(arg2);
+
+    tcg_gen_helper_1_2(func, ret, arg1, tmp);
+    tcg_temp_free(tmp);
+}
+
+static inline void tcg_gen_helper_1_1ii(void *func, TCGv ret, TCGv arg1, TCGv arg2, TCGv arg3)
+{
+    TCGv tmp1 = tcg_const_i32(arg2);
+    TCGv tmp2 = tcg_const_i32(arg3);
+
+    tcg_gen_helper_1_3(func, ret, arg1, tmp1, tmp2);
+    tcg_temp_free(tmp1);
+    tcg_temp_free(tmp2);
+}
+
+static inline void tcg_gen_helper_1_2i(void *func, TCGv ret, TCGv arg1, TCGv arg2, TCGv arg3)
+{
+    TCGv tmp = tcg_const_i32(arg3);
+
+    tcg_gen_helper_1_3(func, ret, arg1, arg2, tmp);
+    tcg_temp_free(tmp);
+}
+
+static inline void tcg_gen_helper_1_2ii(void *func, TCGv ret, TCGv arg1, TCGv arg2, TCGv arg3, TCGv arg4)
+{
+    TCGv tmp1 = tcg_const_i32(arg3);
+    TCGv tmp2 = tcg_const_i32(arg4);
+
+    tcg_gen_helper_1_4(func, ret, arg1, arg2, tmp1, tmp2);
+    tcg_temp_free(tmp1);
+    tcg_temp_free(tmp2);
 }
 
 typedef struct DisasContext {
     struct TranslationBlock *tb;
     target_ulong pc, saved_pc;
     uint32_t opcode;
-    uint32_t fp_status;
     /* Routine used to access memory */
     int mem_idx;
     uint32_t hflags, saved_hflags;
@@ -479,8 +528,7 @@ typedef struct DisasContext {
 
 enum {
     BS_NONE     = 0, /* We go out of the TB without reaching a branch or an
-                      * exception condition
-                      */
+                      * exception condition */
     BS_STOP     = 1, /* We want to stop translation for any reason */
     BS_BRANCH   = 2, /* We reached a branch condition     */
     BS_EXCP     = 3, /* We reached an exception condition */
@@ -522,147 +570,202 @@ static inline void gen_load_gpr (TCGv t, int reg)
     if (reg == 0)
         tcg_gen_movi_tl(t, 0);
     else
-        tcg_gen_ld_tl(t, current_tc_gprs, sizeof(target_ulong) * reg);
+        tcg_gen_ld_tl(t, cpu_env, offsetof(CPUState, active_tc.gpr) +
+                                  sizeof(target_ulong) * reg);
 }
 
 static inline void gen_store_gpr (TCGv t, int reg)
 {
     if (reg != 0)
-        tcg_gen_st_tl(t, current_tc_gprs, sizeof(target_ulong) * reg);
+        tcg_gen_st_tl(t, cpu_env, offsetof(CPUState, active_tc.gpr) +
+                                  sizeof(target_ulong) * reg);
+}
+
+/* Moves to/from HI and LO registers.  */
+static inline void gen_load_LO (TCGv t, int reg)
+{
+    tcg_gen_ld_tl(t, cpu_env, offsetof(CPUState, active_tc.LO) +
+                              sizeof(target_ulong) * reg);
+}
+
+static inline void gen_store_LO (TCGv t, int reg)
+{
+    tcg_gen_st_tl(t, cpu_env, offsetof(CPUState, active_tc.LO) +
+                              sizeof(target_ulong) * reg);
+}
+
+static inline void gen_load_HI (TCGv t, int reg)
+{
+    tcg_gen_ld_tl(t, cpu_env, offsetof(CPUState, active_tc.HI) +
+                              sizeof(target_ulong) * reg);
+}
+
+static inline void gen_store_HI (TCGv t, int reg)
+{
+    tcg_gen_st_tl(t, cpu_env, offsetof(CPUState, active_tc.HI) +
+                              sizeof(target_ulong) * reg);
 }
 
 /* Moves to/from shadow registers. */
-static inline void gen_load_srsgpr (TCGv t, int reg)
+static inline void gen_load_srsgpr (int from, int to)
 {
-    if (reg == 0)
-        tcg_gen_movi_tl(t, 0);
+    TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_TL);
+
+    if (from == 0)
+        tcg_gen_movi_tl(r_tmp1, 0);
     else {
-        TCGv r_tmp = new_tmp();
+        TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I32);
 
-        tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSCtl));
-        tcg_gen_shri_i32(r_tmp, r_tmp, CP0SRSCtl_PSS);
-        tcg_gen_andi_i32(r_tmp, r_tmp, 0xf);
-        tcg_gen_muli_i32(r_tmp, r_tmp, sizeof(target_ulong) * 32);
-        tcg_gen_add_i32(r_tmp, cpu_env, r_tmp);
+        tcg_gen_ld_i32(r_tmp2, cpu_env, offsetof(CPUState, CP0_SRSCtl));
+        tcg_gen_shri_i32(r_tmp2, r_tmp2, CP0SRSCtl_PSS);
+        tcg_gen_andi_i32(r_tmp2, r_tmp2, 0xf);
+        tcg_gen_muli_i32(r_tmp2, r_tmp2, sizeof(target_ulong) * 32);
+        tcg_gen_add_i32(r_tmp2, cpu_env, r_tmp2);
 
-        tcg_gen_ld_tl(t, r_tmp, sizeof(target_ulong) * reg);
-        dead_tmp(r_tmp);
+        tcg_gen_ld_tl(r_tmp1, r_tmp2, sizeof(target_ulong) * from);
+        tcg_temp_free(r_tmp2);
     }
+    gen_store_gpr(r_tmp1, to);
+    tcg_temp_free(r_tmp1);
 }
 
-static inline void gen_store_srsgpr (TCGv t, int reg)
+static inline void gen_store_srsgpr (int from, int to)
 {
-    if (reg != 0) {
-        TCGv r_tmp = new_tmp();
-
-        tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSCtl));
-        tcg_gen_shri_i32(r_tmp, r_tmp, CP0SRSCtl_PSS);
-        tcg_gen_andi_i32(r_tmp, r_tmp, 0xf);
-        tcg_gen_muli_i32(r_tmp, r_tmp, sizeof(target_ulong) * 32);
-        tcg_gen_add_i32(r_tmp, cpu_env, r_tmp);
-
-        tcg_gen_st_tl(t, r_tmp, sizeof(target_ulong) * reg);
-        dead_tmp(r_tmp);
+    if (to != 0) {
+        TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_TL);
+        TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I32);
+
+        gen_load_gpr(r_tmp1, from);
+        tcg_gen_ld_i32(r_tmp2, cpu_env, offsetof(CPUState, CP0_SRSCtl));
+        tcg_gen_shri_i32(r_tmp2, r_tmp2, CP0SRSCtl_PSS);
+        tcg_gen_andi_i32(r_tmp2, r_tmp2, 0xf);
+        tcg_gen_muli_i32(r_tmp2, r_tmp2, sizeof(target_ulong) * 32);
+        tcg_gen_add_i32(r_tmp2, cpu_env, r_tmp2);
+
+        tcg_gen_st_tl(r_tmp1, r_tmp2, sizeof(target_ulong) * to);
+        tcg_temp_free(r_tmp1);
+        tcg_temp_free(r_tmp2);
     }
 }
 
 /* Floating point register moves. */
-#define FGEN32(func, NAME)                       \
-static GenOpFunc *NAME ## _table [32] = {        \
-NAME ## 0,  NAME ## 1,  NAME ## 2,  NAME ## 3,   \
-NAME ## 4,  NAME ## 5,  NAME ## 6,  NAME ## 7,   \
-NAME ## 8,  NAME ## 9,  NAME ## 10, NAME ## 11,  \
-NAME ## 12, NAME ## 13, NAME ## 14, NAME ## 15,  \
-NAME ## 16, NAME ## 17, NAME ## 18, NAME ## 19,  \
-NAME ## 20, NAME ## 21, NAME ## 22, NAME ## 23,  \
-NAME ## 24, NAME ## 25, NAME ## 26, NAME ## 27,  \
-NAME ## 28, NAME ## 29, NAME ## 30, NAME ## 31,  \
-};                                               \
-static always_inline void func(int n)            \
-{                                                \
-    NAME ## _table[n]();                         \
+static inline void gen_load_fpr32 (TCGv t, int reg)
+{
+    tcg_gen_ld_i32(t, current_fpu, 8 * reg + 4 * FP_ENDIAN_IDX);
 }
 
-FGEN32(gen_op_load_fpr_WT0,  gen_op_load_fpr_WT0_fpr);
-FGEN32(gen_op_store_fpr_WT0, gen_op_store_fpr_WT0_fpr);
-
-FGEN32(gen_op_load_fpr_WT1,  gen_op_load_fpr_WT1_fpr);
-FGEN32(gen_op_store_fpr_WT1, gen_op_store_fpr_WT1_fpr);
-
-FGEN32(gen_op_load_fpr_WT2,  gen_op_load_fpr_WT2_fpr);
-FGEN32(gen_op_store_fpr_WT2, gen_op_store_fpr_WT2_fpr);
-
-FGEN32(gen_op_load_fpr_DT0,  gen_op_load_fpr_DT0_fpr);
-FGEN32(gen_op_store_fpr_DT0, gen_op_store_fpr_DT0_fpr);
-
-FGEN32(gen_op_load_fpr_DT1,  gen_op_load_fpr_DT1_fpr);
-FGEN32(gen_op_store_fpr_DT1, gen_op_store_fpr_DT1_fpr);
-
-FGEN32(gen_op_load_fpr_DT2,  gen_op_load_fpr_DT2_fpr);
-FGEN32(gen_op_store_fpr_DT2, gen_op_store_fpr_DT2_fpr);
+static inline void gen_store_fpr32 (TCGv t, int reg)
+{
+    tcg_gen_st_i32(t, current_fpu, 8 * reg + 4 * FP_ENDIAN_IDX);
+}
 
-FGEN32(gen_op_load_fpr_WTH0,  gen_op_load_fpr_WTH0_fpr);
-FGEN32(gen_op_store_fpr_WTH0, gen_op_store_fpr_WTH0_fpr);
+static inline void gen_load_fpr64 (DisasContext *ctx, TCGv t, int reg)
+{
+    if (ctx->hflags & MIPS_HFLAG_F64) {
+        tcg_gen_ld_i64(t, current_fpu, 8 * reg);
+    } else {
+        TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I32);
+        TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I64);
+
+        tcg_gen_ld_i32(r_tmp1, current_fpu, 8 * (reg | 1) + 4 * FP_ENDIAN_IDX);
+        tcg_gen_extu_i32_i64(t, r_tmp1);
+        tcg_gen_shli_i64(t, t, 32);
+        tcg_gen_ld_i32(r_tmp1, current_fpu, 8 * (reg & ~1) + 4 * FP_ENDIAN_IDX);
+        tcg_gen_extu_i32_i64(r_tmp2, r_tmp1);
+        tcg_gen_or_i64(t, t, r_tmp2);
+        tcg_temp_free(r_tmp1);
+        tcg_temp_free(r_tmp2);
+    }
+}
 
-FGEN32(gen_op_load_fpr_WTH1,  gen_op_load_fpr_WTH1_fpr);
-FGEN32(gen_op_store_fpr_WTH1, gen_op_store_fpr_WTH1_fpr);
+static inline void gen_store_fpr64 (DisasContext *ctx, TCGv t, int reg)
+{
+    if (ctx->hflags & MIPS_HFLAG_F64) {
+        tcg_gen_st_i64(t, current_fpu, 8 * reg);
+    } else {
+        TCGv r_tmp = tcg_temp_new(TCG_TYPE_I32);
+
+        tcg_gen_trunc_i64_i32(r_tmp, t);
+        tcg_gen_st_i32(r_tmp, current_fpu, 8 * (reg & ~1) + 4 * FP_ENDIAN_IDX);
+        tcg_gen_shri_i64(t, t, 32);
+        tcg_gen_trunc_i64_i32(r_tmp, t);
+        tcg_gen_st_i32(r_tmp, current_fpu, 8 * (reg | 1) + 4 * FP_ENDIAN_IDX);
+        tcg_temp_free(r_tmp);
+    }
+}
 
-FGEN32(gen_op_load_fpr_WTH2,  gen_op_load_fpr_WTH2_fpr);
-FGEN32(gen_op_store_fpr_WTH2, gen_op_store_fpr_WTH2_fpr);
+static inline void gen_load_fpr32h (TCGv t, int reg)
+{
+    tcg_gen_ld_i32(t, current_fpu, 8 * reg + 4 * !FP_ENDIAN_IDX);
+}
 
-#define GEN_LOAD_FREG_FTN(FTn, Fn)                                            \
-do {                                                                          \
-    glue(gen_op_load_fpr_, FTn)(Fn);                                          \
-} while (0)
+static inline void gen_store_fpr32h (TCGv t, int reg)
+{
+    tcg_gen_st_i32(t, current_fpu, 8 * reg + 4 * !FP_ENDIAN_IDX);
+}
 
-#define GEN_STORE_FTN_FREG(Fn, FTn)                                           \
-do {                                                                          \
-    glue(gen_op_store_fpr_, FTn)(Fn);                                         \
-} while (0)
+static inline void get_fp_cond (TCGv t)
+{
+    TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I32);
+    TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I32);
+
+    tcg_gen_ld_i32(r_tmp1, current_fpu, offsetof(CPUMIPSFPUContext, fcr31));
+    tcg_gen_shri_i32(r_tmp2, r_tmp1, 24);
+    tcg_gen_andi_i32(r_tmp2, r_tmp2, 0xfe);
+    tcg_gen_shri_i32(r_tmp1, r_tmp1, 23);
+    tcg_gen_andi_i32(r_tmp1, r_tmp1, 0x1);
+    tcg_gen_or_i32(t, r_tmp1, r_tmp2);
+    tcg_temp_free(r_tmp1);
+    tcg_temp_free(r_tmp2);
+}
 
-#define FOP_CONDS(type, fmt)                                            \
-static GenOpFunc1 * gen_op_cmp ## type ## _ ## fmt ## _table[16] = {    \
-    gen_op_cmp ## type ## _ ## fmt ## _f,                               \
-    gen_op_cmp ## type ## _ ## fmt ## _un,                              \
-    gen_op_cmp ## type ## _ ## fmt ## _eq,                              \
-    gen_op_cmp ## type ## _ ## fmt ## _ueq,                             \
-    gen_op_cmp ## type ## _ ## fmt ## _olt,                             \
-    gen_op_cmp ## type ## _ ## fmt ## _ult,                             \
-    gen_op_cmp ## type ## _ ## fmt ## _ole,                             \
-    gen_op_cmp ## type ## _ ## fmt ## _ule,                             \
-    gen_op_cmp ## type ## _ ## fmt ## _sf,                              \
-    gen_op_cmp ## type ## _ ## fmt ## _ngle,                            \
-    gen_op_cmp ## type ## _ ## fmt ## _seq,                             \
-    gen_op_cmp ## type ## _ ## fmt ## _ngl,                             \
-    gen_op_cmp ## type ## _ ## fmt ## _lt,                              \
-    gen_op_cmp ## type ## _ ## fmt ## _nge,                             \
-    gen_op_cmp ## type ## _ ## fmt ## _le,                              \
-    gen_op_cmp ## type ## _ ## fmt ## _ngt,                             \
-};                                                                      \
-static always_inline void gen_cmp ## type ## _ ## fmt(int n, long cc)   \
-{                                                                       \
-    gen_op_cmp ## type ## _ ## fmt ## _table[n](cc);                    \
+typedef void (fcmp_fun32)(uint32_t, uint32_t, int);
+typedef void (fcmp_fun64)(uint64_t, uint64_t, int);
+
+#define FOP_CONDS(fcmp_fun, type, arg0, arg1, fmt)                            \
+static fcmp_fun * fcmp ## type ## _ ## fmt ## _table[16] = {                  \
+    do_cmp ## type ## _ ## fmt ## _f,                                         \
+    do_cmp ## type ## _ ## fmt ## _un,                                        \
+    do_cmp ## type ## _ ## fmt ## _eq,                                        \
+    do_cmp ## type ## _ ## fmt ## _ueq,                                       \
+    do_cmp ## type ## _ ## fmt ## _olt,                                       \
+    do_cmp ## type ## _ ## fmt ## _ult,                                       \
+    do_cmp ## type ## _ ## fmt ## _ole,                                       \
+    do_cmp ## type ## _ ## fmt ## _ule,                                       \
+    do_cmp ## type ## _ ## fmt ## _sf,                                        \
+    do_cmp ## type ## _ ## fmt ## _ngle,                                      \
+    do_cmp ## type ## _ ## fmt ## _seq,                                       \
+    do_cmp ## type ## _ ## fmt ## _ngl,                                       \
+    do_cmp ## type ## _ ## fmt ## _lt,                                        \
+    do_cmp ## type ## _ ## fmt ## _nge,                                       \
+    do_cmp ## type ## _ ## fmt ## _le,                                        \
+    do_cmp ## type ## _ ## fmt ## _ngt,                                       \
+};                                                                            \
+static inline void gen_cmp ## type ## _ ## fmt(int n, arg0 a, arg1 b, int cc) \
+{                                                                             \
+    tcg_gen_helper_0_2i(fcmp ## type ## _ ## fmt ## _table[n], a, b, cc);     \
 }
 
-FOP_CONDS(, d)
-FOP_CONDS(abs, d)
-FOP_CONDS(, s)
-FOP_CONDS(abs, s)
-FOP_CONDS(, ps)
-FOP_CONDS(abs, ps)
+FOP_CONDS(fcmp_fun64, , uint64_t, uint64_t, d)
+FOP_CONDS(fcmp_fun64, abs, uint64_t, uint64_t, d)
+FOP_CONDS(fcmp_fun32, , uint32_t, uint32_t, s)
+FOP_CONDS(fcmp_fun32, abs, uint32_t, uint32_t, s)
+FOP_CONDS(fcmp_fun64, , uint64_t, uint64_t, ps)
+FOP_CONDS(fcmp_fun64, abs, uint64_t, uint64_t, ps)
+#undef FOP_CONDS
 
 /* Tests */
 #define OP_COND(name, cond)                                   \
-void glue(gen_op_, name) (void)                               \
+static inline void glue(gen_op_, name) (TCGv t0, TCGv t1)     \
 {                                                             \
     int l1 = gen_new_label();                                 \
     int l2 = gen_new_label();                                 \
                                                               \
-    tcg_gen_brcond_tl(cond, cpu_T[0], cpu_T[1], l1);          \
-    tcg_gen_movi_tl(cpu_T[0], 0);                             \
+    tcg_gen_brcond_tl(cond, t0, t1, l1);                      \
+    tcg_gen_movi_tl(t0, 0);                                   \
     tcg_gen_br(l2);                                           \
     gen_set_label(l1);                                        \
-    tcg_gen_movi_tl(cpu_T[0], 1);                             \
+    tcg_gen_movi_tl(t0, 1);                                   \
     gen_set_label(l2);                                        \
 }
 OP_COND(eq, TCG_COND_EQ);
@@ -674,16 +777,16 @@ OP_COND(ltu, TCG_COND_LTU);
 #undef OP_COND
 
 #define OP_CONDI(name, cond)                                  \
-void glue(gen_op_, name) (target_ulong val)                   \
+static inline void glue(gen_op_, name) (TCGv t, target_ulong val) \
 {                                                             \
     int l1 = gen_new_label();                                 \
     int l2 = gen_new_label();                                 \
                                                               \
-    tcg_gen_brcondi_tl(cond, cpu_T[0], val, l1);              \
-    tcg_gen_movi_tl(cpu_T[0], 0);                             \
+    tcg_gen_brcondi_tl(cond, t, val, l1);                     \
+    tcg_gen_movi_tl(t, 0);                                    \
     tcg_gen_br(l2);                                           \
     gen_set_label(l1);                                        \
-    tcg_gen_movi_tl(cpu_T[0], 1);                             \
+    tcg_gen_movi_tl(t, 1);                                    \
     gen_set_label(l2);                                        \
 }
 OP_CONDI(lti, TCG_COND_LT);
@@ -691,16 +794,16 @@ OP_CONDI(ltiu, TCG_COND_LTU);
 #undef OP_CONDI
 
 #define OP_CONDZ(name, cond)                                  \
-void glue(gen_op_, name) (void)                               \
+static inline void glue(gen_op_, name) (TCGv t)               \
 {                                                             \
     int l1 = gen_new_label();                                 \
     int l2 = gen_new_label();                                 \
                                                               \
-    tcg_gen_brcondi_tl(cond, cpu_T[0], 0, l1);                \
-    tcg_gen_movi_tl(cpu_T[0], 0);                             \
+    tcg_gen_brcondi_tl(cond, t, 0, l1);                       \
+    tcg_gen_movi_tl(t, 0);                                    \
     tcg_gen_br(l2);                                           \
     gen_set_label(l1);                                        \
-    tcg_gen_movi_tl(cpu_T[0], 1);                             \
+    tcg_gen_movi_tl(t, 1);                                    \
     gen_set_label(l2);                                        \
 }
 OP_CONDZ(gez, TCG_COND_GE);
@@ -712,52 +815,13 @@ OP_CONDZ(ltz, TCG_COND_LT);
 static inline void gen_save_pc(target_ulong pc)
 {
     TCGv r_tmp = tcg_temp_new(TCG_TYPE_TL);
-    TCGv r_tc_off = new_tmp();
-    TCGv r_tc_off_tl = tcg_temp_new(TCG_TYPE_TL);
-    TCGv r_ptr = tcg_temp_new(TCG_TYPE_PTR);
 
     tcg_gen_movi_tl(r_tmp, pc);
-    tcg_gen_ld_i32(r_tc_off, cpu_env, offsetof(CPUState, current_tc));
-    tcg_gen_muli_i32(r_tc_off, r_tc_off, sizeof(target_ulong));
-    tcg_gen_ext_i32_ptr(r_tc_off_tl, r_tc_off);
-    tcg_gen_add_ptr(r_ptr, cpu_env, r_tc_off_tl);
-    tcg_gen_st_tl(r_tmp, r_ptr, offsetof(CPUState, PC));
-    dead_tmp(r_tc_off);
-}
-
-static inline void gen_breg_pc(void)
-{
-    TCGv r_tmp = tcg_temp_new(TCG_TYPE_TL);
-    TCGv r_tc_off = new_tmp();
-    TCGv r_tc_off_tl = tcg_temp_new(TCG_TYPE_TL);
-    TCGv r_ptr = tcg_temp_new(TCG_TYPE_PTR);
-
-    tcg_gen_ld_tl(r_tmp, cpu_env, offsetof(CPUState, btarget));
-    tcg_gen_ld_i32(r_tc_off, cpu_env, offsetof(CPUState, current_tc));
-    tcg_gen_muli_i32(r_tc_off, r_tc_off, sizeof(target_ulong));
-    tcg_gen_ext_i32_ptr(r_tc_off_tl, r_tc_off);
-    tcg_gen_add_ptr(r_ptr, cpu_env, r_tc_off_tl);
-    tcg_gen_st_tl(r_tmp, r_ptr, offsetof(CPUState, PC));
-    dead_tmp(r_tc_off);
-}
-
-static inline void gen_save_btarget(target_ulong btarget)
-{
-    TCGv r_tmp = tcg_temp_new(TCG_TYPE_TL);
-
-    tcg_gen_movi_tl(r_tmp, btarget);
-    tcg_gen_st_tl(r_tmp, cpu_env, offsetof(CPUState, btarget));
-}
-
-static always_inline void gen_save_breg_target(int reg)
-{
-    TCGv r_tmp = tcg_temp_new(TCG_TYPE_TL);
-
-    gen_load_gpr(r_tmp, reg);
-    tcg_gen_st_tl(r_tmp, cpu_env, offsetof(CPUState, btarget));
+    tcg_gen_st_tl(r_tmp, cpu_env, offsetof(CPUState, active_tc.PC));
+    tcg_temp_free(r_tmp);
 }
 
-static always_inline void save_cpu_state (DisasContext *ctx, int do_save_pc)
+static inline void save_cpu_state (DisasContext *ctx, int do_save_pc)
 {
 #if defined MIPS_DEBUG_DISAS
     if (loglevel & CPU_LOG_TB_IN_ASM) {
@@ -770,7 +834,11 @@ static always_inline void save_cpu_state (DisasContext *ctx, int do_save_pc)
         ctx->saved_pc = ctx->pc;
     }
     if (ctx->hflags != ctx->saved_hflags) {
-        gen_op_save_state(ctx->hflags);
+        TCGv r_tmp = tcg_temp_new(TCG_TYPE_I32);
+
+        tcg_gen_movi_i32(r_tmp, ctx->hflags);
+        tcg_gen_st_i32(r_tmp, cpu_env, offsetof(CPUState, hflags));
+        tcg_temp_free(r_tmp);
         ctx->saved_hflags = ctx->hflags;
         switch (ctx->hflags & MIPS_HFLAG_BMASK) {
         case MIPS_HFLAG_BR:
@@ -778,13 +846,13 @@ static always_inline void save_cpu_state (DisasContext *ctx, int do_save_pc)
         case MIPS_HFLAG_BC:
         case MIPS_HFLAG_BL:
         case MIPS_HFLAG_B:
-            gen_save_btarget(ctx->btarget);
+            tcg_gen_movi_tl(btarget, ctx->btarget);
             break;
         }
     }
 }
 
-static always_inline void restore_cpu_state (CPUState *env, DisasContext *ctx)
+static inline void restore_cpu_state (CPUState *env, DisasContext *ctx)
 {
     ctx->saved_hflags = ctx->hflags;
     switch (ctx->hflags & MIPS_HFLAG_BMASK) {
@@ -798,36 +866,36 @@ static always_inline void restore_cpu_state (CPUState *env, DisasContext *ctx)
     }
 }
 
-static always_inline void
+static inline void
 generate_exception_err (DisasContext *ctx, int excp, int err)
 {
     save_cpu_state(ctx, 1);
-    tcg_gen_helper_0_2(do_raise_exception_err, tcg_const_i32(excp), tcg_const_i32(err));
+    tcg_gen_helper_0_ii(do_raise_exception_err, excp, err);
     tcg_gen_helper_0_0(do_interrupt_restart);
     tcg_gen_exit_tb(0);
 }
 
-static always_inline void
+static inline void
 generate_exception (DisasContext *ctx, int excp)
 {
     save_cpu_state(ctx, 1);
-    tcg_gen_helper_0_1(do_raise_exception, tcg_const_i32(excp));
+    tcg_gen_helper_0_i(do_raise_exception, excp);
     tcg_gen_helper_0_0(do_interrupt_restart);
     tcg_gen_exit_tb(0);
 }
 
 /* Addresses computation */
-static inline void gen_op_addr_add (void)
+static inline void gen_op_addr_add (TCGv t0, TCGv t1)
 {
-    tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
+    tcg_gen_add_tl(t0, t0, t1);
 
 #if defined(TARGET_MIPS64)
     /* For compatibility with 32-bit code, data reference in user mode
        with Status_UX = 0 should be casted to 32-bit and sign extended.
        See the MIPS64 PRA manual, section 4.10. */
     {
-        TCGv r_tmp = new_tmp();
         int l1 = gen_new_label();
+        TCGv r_tmp = tcg_temp_local_new(TCG_TYPE_I32);
 
         tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, hflags));
         tcg_gen_andi_i32(r_tmp, r_tmp, MIPS_HFLAG_KSU);
@@ -835,20 +903,20 @@ static inline void gen_op_addr_add (void)
         tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Status));
         tcg_gen_andi_i32(r_tmp, r_tmp, (1 << CP0St_UX));
         tcg_gen_brcondi_i32(TCG_COND_NE, r_tmp, 0, l1);
-        tcg_gen_ext32s_i64(cpu_T[0], cpu_T[0]);
+        tcg_temp_free(r_tmp);
+        tcg_gen_ext32s_i64(t0, t0);
         gen_set_label(l1);
-        dead_tmp(r_tmp);
     }
 #endif
 }
 
-static always_inline void check_cp0_enabled(DisasContext *ctx)
+static inline void check_cp0_enabled(DisasContext *ctx)
 {
     if (unlikely(!(ctx->hflags & MIPS_HFLAG_CP0)))
         generate_exception_err(ctx, EXCP_CpU, 1);
 }
 
-static always_inline void check_cp1_enabled(DisasContext *ctx)
+static inline void check_cp1_enabled(DisasContext *ctx)
 {
     if (unlikely(!(ctx->hflags & MIPS_HFLAG_FPU)))
         generate_exception_err(ctx, EXCP_CpU, 1);
@@ -858,7 +926,7 @@ static always_inline void check_cp1_enabled(DisasContext *ctx)
    This is associated with the nabla symbol in the MIPS32 and MIPS64
    opcode tables.  */
 
-static always_inline void check_cop1x(DisasContext *ctx)
+static inline void check_cop1x(DisasContext *ctx)
 {
     if (unlikely(!(ctx->hflags & MIPS_HFLAG_COP1X)))
         generate_exception(ctx, EXCP_RI);
@@ -867,7 +935,7 @@ static always_inline void check_cop1x(DisasContext *ctx)
 /* Verify that the processor is running with 64-bit floating-point
    operations enabled.  */
 
-static always_inline void check_cp1_64bitmode(DisasContext *ctx)
+static inline void check_cp1_64bitmode(DisasContext *ctx)
 {
     if (unlikely(~ctx->hflags & (MIPS_HFLAG_F64 | MIPS_HFLAG_COP1X)))
         generate_exception(ctx, EXCP_RI);
@@ -884,7 +952,7 @@ static always_inline void check_cp1_64bitmode(DisasContext *ctx)
  * Multiple 64 bit wide registers can be checked by calling
  * gen_op_cp1_registers(freg1 | freg2 | ... | fregN);
  */
-void check_cp1_registers(DisasContext *ctx, int regs)
+static inline void check_cp1_registers(DisasContext *ctx, int regs)
 {
     if (unlikely(!(ctx->hflags & MIPS_HFLAG_F64) && (regs & 1)))
         generate_exception(ctx, EXCP_RI);
@@ -892,7 +960,7 @@ void check_cp1_registers(DisasContext *ctx, int regs)
 
 /* This code generates a "reserved instruction" exception if the
    CPU does not support the instruction set corresponding to flags. */
-static always_inline void check_insn(CPUState *env, DisasContext *ctx, int flags)
+static inline void check_insn(CPUState *env, DisasContext *ctx, int flags)
 {
     if (unlikely(!(env->insn_flags & flags)))
         generate_exception(ctx, EXCP_RI);
@@ -900,54 +968,17 @@ static always_inline void check_insn(CPUState *env, DisasContext *ctx, int flags
 
 /* This code generates a "reserved instruction" exception if 64-bit
    instructions are not enabled. */
-static always_inline void check_mips_64(DisasContext *ctx)
+static inline void check_mips_64(DisasContext *ctx)
 {
     if (unlikely(!(ctx->hflags & MIPS_HFLAG_64)))
         generate_exception(ctx, EXCP_RI);
 }
 
 /* load/store instructions. */
-#if defined(CONFIG_USER_ONLY)
-#define op_ldst(name)        gen_op_##name##_raw()
-#define OP_LD_TABLE(width)
-#define OP_ST_TABLE(width)
-#else
-#define op_ldst(name)        (*gen_op_##name[ctx->mem_idx])()
-#define OP_LD_TABLE(width)                                                    \
-static GenOpFunc *gen_op_l##width[] = {                                       \
-    &gen_op_l##width##_kernel,                                                \
-    &gen_op_l##width##_super,                                                 \
-    &gen_op_l##width##_user,                                                  \
-}
-#define OP_ST_TABLE(width)                                                    \
-static GenOpFunc *gen_op_s##width[] = {                                       \
-    &gen_op_s##width##_kernel,                                                \
-    &gen_op_s##width##_super,                                                 \
-    &gen_op_s##width##_user,                                                  \
-}
-#endif
-
-#if defined(TARGET_MIPS64)
-OP_LD_TABLE(dl);
-OP_LD_TABLE(dr);
-OP_ST_TABLE(dl);
-OP_ST_TABLE(dr);
-#endif
-OP_LD_TABLE(wl);
-OP_LD_TABLE(wr);
-OP_ST_TABLE(wl);
-OP_ST_TABLE(wr);
-OP_LD_TABLE(wc1);
-OP_ST_TABLE(wc1);
-OP_LD_TABLE(dc1);
-OP_ST_TABLE(dc1);
-OP_LD_TABLE(uxc1);
-OP_ST_TABLE(uxc1);
-
 #define OP_LD(insn,fname)                                        \
-void inline op_ldst_##insn(DisasContext *ctx)                    \
+static inline void op_ldst_##insn(TCGv t0, DisasContext *ctx)    \
 {                                                                \
-    tcg_gen_qemu_##fname(cpu_T[0], cpu_T[0], ctx->mem_idx);      \
+    tcg_gen_qemu_##fname(t0, t0, ctx->mem_idx);                  \
 }
 OP_LD(lb,ld8s);
 OP_LD(lbu,ld8u);
@@ -961,9 +992,9 @@ OP_LD(ld,ld64);
 #undef OP_LD
 
 #define OP_ST(insn,fname)                                        \
-void inline op_ldst_##insn(DisasContext *ctx)                    \
+static inline void op_ldst_##insn(TCGv t0, TCGv t1, DisasContext *ctx) \
 {                                                                \
-    tcg_gen_qemu_##fname(cpu_T[1], cpu_T[0], ctx->mem_idx);      \
+    tcg_gen_qemu_##fname(t1, t0, ctx->mem_idx);                  \
 }
 OP_ST(sb,st8);
 OP_ST(sh,st16);
@@ -974,11 +1005,11 @@ OP_ST(sd,st64);
 #undef OP_ST
 
 #define OP_LD_ATOMIC(insn,fname)                                        \
-void inline op_ldst_##insn(DisasContext *ctx)                           \
+static inline void op_ldst_##insn(TCGv t0, TCGv t1, DisasContext *ctx)  \
 {                                                                       \
-    tcg_gen_mov_tl(cpu_T[1], cpu_T[0]);                                 \
-    tcg_gen_qemu_##fname(cpu_T[0], cpu_T[0], ctx->mem_idx);             \
-    tcg_gen_st_tl(cpu_T[1], cpu_env, offsetof(CPUState, CP0_LLAddr));   \
+    tcg_gen_mov_tl(t1, t0);                                             \
+    tcg_gen_qemu_##fname(t0, t0, ctx->mem_idx);                         \
+    tcg_gen_st_tl(t1, cpu_env, offsetof(CPUState, CP0_LLAddr));         \
 }
 OP_LD_ATOMIC(ll,ld32s);
 #if defined(TARGET_MIPS64)
@@ -987,25 +1018,26 @@ OP_LD_ATOMIC(lld,ld64);
 #undef OP_LD_ATOMIC
 
 #define OP_ST_ATOMIC(insn,fname,almask)                                 \
-void inline op_ldst_##insn(DisasContext *ctx)                           \
+static inline void op_ldst_##insn(TCGv t0, TCGv t1, DisasContext *ctx)  \
 {                                                                       \
-    TCGv r_tmp = tcg_temp_new(TCG_TYPE_TL);                             \
+    TCGv r_tmp = tcg_temp_local_new(TCG_TYPE_TL);                       \
     int l1 = gen_new_label();                                           \
     int l2 = gen_new_label();                                           \
     int l3 = gen_new_label();                                           \
                                                                         \
-    tcg_gen_andi_tl(r_tmp, cpu_T[0], almask);                           \
+    tcg_gen_andi_tl(r_tmp, t0, almask);                                 \
     tcg_gen_brcondi_tl(TCG_COND_EQ, r_tmp, 0, l1);                      \
-    tcg_gen_st_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_BadVAddr)); \
+    tcg_gen_st_tl(t0, cpu_env, offsetof(CPUState, CP0_BadVAddr));       \
     generate_exception(ctx, EXCP_AdES);                                 \
     gen_set_label(l1);                                                  \
     tcg_gen_ld_tl(r_tmp, cpu_env, offsetof(CPUState, CP0_LLAddr));      \
-    tcg_gen_brcond_tl(TCG_COND_NE, cpu_T[0], r_tmp, l2);                \
-    tcg_gen_qemu_##fname(cpu_T[1], cpu_T[0], ctx->mem_idx);             \
-    tcg_gen_movi_tl(cpu_T[0], 1);                                       \
+    tcg_gen_brcond_tl(TCG_COND_NE, t0, r_tmp, l2);                      \
+    tcg_temp_free(r_tmp);                                               \
+    tcg_gen_qemu_##fname(t1, t0, ctx->mem_idx);                         \
+    tcg_gen_movi_tl(t0, 1);                                             \
     tcg_gen_br(l3);                                                     \
     gen_set_label(l2);                                                  \
-    tcg_gen_movi_tl(cpu_T[0], 0);                                       \
+    tcg_gen_movi_tl(t0, 0);                                             \
     gen_set_label(l3);                                                  \
 }
 OP_ST_ATOMIC(sc,st32,0x3);
@@ -1014,221 +1046,240 @@ OP_ST_ATOMIC(scd,st64,0x7);
 #endif
 #undef OP_ST_ATOMIC
 
-void inline op_ldst_lwc1(DisasContext *ctx)
-{
-    op_ldst(lwc1);
-}
-
-void inline op_ldst_ldc1(DisasContext *ctx)
-{
-    op_ldst(ldc1);
-}
-
-void inline op_ldst_swc1(DisasContext *ctx)
-{
-    op_ldst(swc1);
-}
-
-void inline op_ldst_sdc1(DisasContext *ctx)
-{
-    op_ldst(sdc1);
-}
-
 /* Load and store */
 static void gen_ldst (DisasContext *ctx, uint32_t opc, int rt,
                       int base, int16_t offset)
 {
     const char *opn = "ldst";
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+    TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
 
     if (base == 0) {
-        tcg_gen_movi_tl(cpu_T[0], offset);
+        tcg_gen_movi_tl(t0, offset);
     } else if (offset == 0) {
-        gen_load_gpr(cpu_T[0], base);
+        gen_load_gpr(t0, base);
     } else {
-        gen_load_gpr(cpu_T[0], base);
-        tcg_gen_movi_tl(cpu_T[1], offset);
-        gen_op_addr_add();
+        gen_load_gpr(t0, base);
+        tcg_gen_movi_tl(t1, offset);
+        gen_op_addr_add(t0, t1);
     }
     /* Don't do NOP if destination is zero: we must perform the actual
        memory access. */
     switch (opc) {
 #if defined(TARGET_MIPS64)
     case OPC_LWU:
-        op_ldst_lwu(ctx);
-        gen_store_gpr(cpu_T[0], rt);
+        op_ldst_lwu(t0, ctx);
+        gen_store_gpr(t0, rt);
         opn = "lwu";
         break;
     case OPC_LD:
-        op_ldst_ld(ctx);
-        gen_store_gpr(cpu_T[0], rt);
+        op_ldst_ld(t0, ctx);
+        gen_store_gpr(t0, rt);
         opn = "ld";
         break;
     case OPC_LLD:
-        op_ldst_lld(ctx);
-        gen_store_gpr(cpu_T[0], rt);
+        op_ldst_lld(t0, t1, ctx);
+        gen_store_gpr(t0, rt);
         opn = "lld";
         break;
     case OPC_SD:
-        gen_load_gpr(cpu_T[1], rt);
-        op_ldst_sd(ctx);
+        gen_load_gpr(t1, rt);
+        op_ldst_sd(t0, t1, ctx);
         opn = "sd";
         break;
     case OPC_SCD:
         save_cpu_state(ctx, 1);
-        gen_load_gpr(cpu_T[1], rt);
-        op_ldst_scd(ctx);
-        gen_store_gpr(cpu_T[0], rt);
+        gen_load_gpr(t1, rt);
+        op_ldst_scd(t0, t1, ctx);
+        gen_store_gpr(t0, rt);
         opn = "scd";
         break;
     case OPC_LDL:
-        gen_load_gpr(cpu_T[1], rt);
-        op_ldst(ldl);
-        gen_store_gpr(cpu_T[1], rt);
+        save_cpu_state(ctx, 1);
+        gen_load_gpr(t1, rt);
+        tcg_gen_helper_1_2i(do_ldl, t1, t0, t1, ctx->mem_idx);
+        gen_store_gpr(t1, rt);
         opn = "ldl";
         break;
     case OPC_SDL:
-        gen_load_gpr(cpu_T[1], rt);
-        op_ldst(sdl);
+        save_cpu_state(ctx, 1);
+        gen_load_gpr(t1, rt);
+        tcg_gen_helper_0_2i(do_sdl, t0, t1, ctx->mem_idx);
         opn = "sdl";
         break;
     case OPC_LDR:
-        gen_load_gpr(cpu_T[1], rt);
-        op_ldst(ldr);
-        gen_store_gpr(cpu_T[1], rt);
+        save_cpu_state(ctx, 1);
+        gen_load_gpr(t1, rt);
+        tcg_gen_helper_1_2i(do_ldr, t1, t0, t1, ctx->mem_idx);
+        gen_store_gpr(t1, rt);
         opn = "ldr";
         break;
     case OPC_SDR:
-        gen_load_gpr(cpu_T[1], rt);
-        op_ldst(sdr);
+        save_cpu_state(ctx, 1);
+        gen_load_gpr(t1, rt);
+        tcg_gen_helper_0_2i(do_sdr, t0, t1, ctx->mem_idx);
         opn = "sdr";
         break;
 #endif
     case OPC_LW:
-        op_ldst_lw(ctx);
-        gen_store_gpr(cpu_T[0], rt);
+        op_ldst_lw(t0, ctx);
+        gen_store_gpr(t0, rt);
         opn = "lw";
         break;
     case OPC_SW:
-        gen_load_gpr(cpu_T[1], rt);
-        op_ldst_sw(ctx);
+        gen_load_gpr(t1, rt);
+        op_ldst_sw(t0, t1, ctx);
         opn = "sw";
         break;
     case OPC_LH:
-        op_ldst_lh(ctx);
-        gen_store_gpr(cpu_T[0], rt);
+        op_ldst_lh(t0, ctx);
+        gen_store_gpr(t0, rt);
         opn = "lh";
         break;
     case OPC_SH:
-        gen_load_gpr(cpu_T[1], rt);
-        op_ldst_sh(ctx);
+        gen_load_gpr(t1, rt);
+        op_ldst_sh(t0, t1, ctx);
         opn = "sh";
         break;
     case OPC_LHU:
-        op_ldst_lhu(ctx);
-        gen_store_gpr(cpu_T[0], rt);
+        op_ldst_lhu(t0, ctx);
+        gen_store_gpr(t0, rt);
         opn = "lhu";
         break;
     case OPC_LB:
-        op_ldst_lb(ctx);
-        gen_store_gpr(cpu_T[0], rt);
+        op_ldst_lb(t0, ctx);
+        gen_store_gpr(t0, rt);
         opn = "lb";
         break;
     case OPC_SB:
-        gen_load_gpr(cpu_T[1], rt);
-        op_ldst_sb(ctx);
+        gen_load_gpr(t1, rt);
+        op_ldst_sb(t0, t1, ctx);
         opn = "sb";
         break;
     case OPC_LBU:
-        op_ldst_lbu(ctx);
-        gen_store_gpr(cpu_T[0], rt);
+        op_ldst_lbu(t0, ctx);
+        gen_store_gpr(t0, rt);
         opn = "lbu";
         break;
     case OPC_LWL:
-       gen_load_gpr(cpu_T[1], rt);
-        op_ldst(lwl);
-        gen_store_gpr(cpu_T[1], rt);
+        save_cpu_state(ctx, 1);
+       gen_load_gpr(t1, rt);
+        tcg_gen_helper_1_2i(do_lwl, t1, t0, t1, ctx->mem_idx);
+        gen_store_gpr(t1, rt);
         opn = "lwl";
         break;
     case OPC_SWL:
-        gen_load_gpr(cpu_T[1], rt);
-        op_ldst(swl);
+        save_cpu_state(ctx, 1);
+        gen_load_gpr(t1, rt);
+        tcg_gen_helper_0_2i(do_swl, t0, t1, ctx->mem_idx);
         opn = "swr";
         break;
     case OPC_LWR:
-       gen_load_gpr(cpu_T[1], rt);
-        op_ldst(lwr);
-        gen_store_gpr(cpu_T[1], rt);
+        save_cpu_state(ctx, 1);
+       gen_load_gpr(t1, rt);
+        tcg_gen_helper_1_2i(do_lwr, t1, t0, t1, ctx->mem_idx);
+        gen_store_gpr(t1, rt);
         opn = "lwr";
         break;
     case OPC_SWR:
-        gen_load_gpr(cpu_T[1], rt);
-        op_ldst(swr);
+        save_cpu_state(ctx, 1);
+        gen_load_gpr(t1, rt);
+        tcg_gen_helper_0_2i(do_swr, t0, t1, ctx->mem_idx);
         opn = "swr";
         break;
     case OPC_LL:
-        op_ldst_ll(ctx);
-        gen_store_gpr(cpu_T[0], rt);
+        op_ldst_ll(t0, t1, ctx);
+        gen_store_gpr(t0, rt);
         opn = "ll";
         break;
     case OPC_SC:
         save_cpu_state(ctx, 1);
-        gen_load_gpr(cpu_T[1], rt);
-        op_ldst_sc(ctx);
-        gen_store_gpr(cpu_T[0], rt);
+        gen_load_gpr(t1, rt);
+        op_ldst_sc(t0, t1, ctx);
+        gen_store_gpr(t0, rt);
         opn = "sc";
         break;
     default:
         MIPS_INVAL(opn);
         generate_exception(ctx, EXCP_RI);
-        return;
+        goto out;
     }
     MIPS_DEBUG("%s %s, %d(%s)", opn, regnames[rt], offset, regnames[base]);
+ out:
+    tcg_temp_free(t0);
+    tcg_temp_free(t1);
 }
 
 /* Load and store */
 static void gen_flt_ldst (DisasContext *ctx, uint32_t opc, int ft,
-                      int base, int16_t offset)
+                          int base, int16_t offset)
 {
     const char *opn = "flt_ldst";
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
 
     if (base == 0) {
-        tcg_gen_movi_tl(cpu_T[0], offset);
+        tcg_gen_movi_tl(t0, offset);
     } else if (offset == 0) {
-        gen_load_gpr(cpu_T[0], base);
+        gen_load_gpr(t0, base);
     } else {
-        gen_load_gpr(cpu_T[0], base);
-        tcg_gen_movi_tl(cpu_T[1], offset);
-        gen_op_addr_add();
+        TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
+
+        gen_load_gpr(t0, base);
+        tcg_gen_movi_tl(t1, offset);
+        gen_op_addr_add(t0, t1);
+        tcg_temp_free(t1);
     }
     /* Don't do NOP if destination is zero: we must perform the actual
        memory access. */
     switch (opc) {
     case OPC_LWC1:
-        op_ldst_lwc1(ctx);
-        GEN_STORE_FTN_FREG(ft, WT0);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            tcg_gen_qemu_ld32s(fp0, t0, ctx->mem_idx);
+            gen_store_fpr32(fp0, ft);
+            tcg_temp_free(fp0);
+        }
         opn = "lwc1";
         break;
     case OPC_SWC1:
-        GEN_LOAD_FREG_FTN(WT0, ft);
-        op_ldst_swc1(ctx);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, ft);
+            tcg_gen_qemu_st32(fp0, t0, ctx->mem_idx);
+            tcg_temp_free(fp0);
+        }
         opn = "swc1";
         break;
     case OPC_LDC1:
-        op_ldst_ldc1(ctx);
-        GEN_STORE_FTN_FREG(ft, DT0);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            tcg_gen_qemu_ld64(fp0, t0, ctx->mem_idx);
+            gen_store_fpr64(ctx, fp0, ft);
+            tcg_temp_free(fp0);
+        }
         opn = "ldc1";
         break;
     case OPC_SDC1:
-        GEN_LOAD_FREG_FTN(DT0, ft);
-        op_ldst_sdc1(ctx);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, ft);
+            tcg_gen_qemu_st64(fp0, t0, ctx->mem_idx);
+            tcg_temp_free(fp0);
+        }
         opn = "sdc1";
         break;
     default:
         MIPS_INVAL(opn);
         generate_exception(ctx, EXCP_RI);
-        return;
+        goto out;
     }
     MIPS_DEBUG("%s %s, %d(%s)", opn, fregnames[ft], offset, regnames[base]);
+ out:
+    tcg_temp_free(t0);
 }
 
 /* Arithmetic with immediate operand */
@@ -1237,12 +1288,13 @@ static void gen_arith_imm (CPUState *env, DisasContext *ctx, uint32_t opc,
 {
     target_ulong uimm;
     const char *opn = "imm arith";
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
 
     if (rt == 0 && opc != OPC_ADDI && opc != OPC_DADDI) {
         /* If no destination, treat it as a NOP.
            For addi, we must generate the overflow exception when needed. */
         MIPS_DEBUG("NOP");
-        return;
+        goto out;
     }
     uimm = (uint16_t)imm;
     switch (opc) {
@@ -1255,15 +1307,14 @@ static void gen_arith_imm (CPUState *env, DisasContext *ctx, uint32_t opc,
     case OPC_SLTI:
     case OPC_SLTIU:
         uimm = (target_long)imm; /* Sign extend to 32/64 bits */
-        tcg_gen_movi_tl(cpu_T[1], uimm);
         /* Fall through. */
     case OPC_ANDI:
     case OPC_ORI:
     case OPC_XORI:
-        gen_load_gpr(cpu_T[0], rs);
+        gen_load_gpr(t0, rs);
         break;
     case OPC_LUI:
-        tcg_gen_movi_tl(cpu_T[0], imm << 16);
+        tcg_gen_movi_tl(t0, imm << 16);
         break;
     case OPC_SLL:
     case OPC_SRA:
@@ -1277,57 +1328,61 @@ static void gen_arith_imm (CPUState *env, DisasContext *ctx, uint32_t opc,
     case OPC_DSRL32:
 #endif
         uimm &= 0x1f;
-        gen_load_gpr(cpu_T[0], rs);
+        gen_load_gpr(t0, rs);
         break;
     }
     switch (opc) {
     case OPC_ADDI:
         {
-            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_TL);
+            TCGv r_tmp1 = tcg_temp_local_new(TCG_TYPE_TL);
             TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_TL);
             int l1 = gen_new_label();
 
             save_cpu_state(ctx, 1);
-            tcg_gen_ext32s_tl(r_tmp1, cpu_T[0]);
-            tcg_gen_addi_tl(cpu_T[0], r_tmp1, uimm);
+            tcg_gen_ext32s_tl(r_tmp1, t0);
+            tcg_gen_addi_tl(t0, r_tmp1, uimm);
 
             tcg_gen_xori_tl(r_tmp1, r_tmp1, uimm);
             tcg_gen_xori_tl(r_tmp1, r_tmp1, -1);
-            tcg_gen_xori_tl(r_tmp2, cpu_T[0], uimm);
+            tcg_gen_xori_tl(r_tmp2, t0, uimm);
             tcg_gen_and_tl(r_tmp1, r_tmp1, r_tmp2);
+            tcg_temp_free(r_tmp2);
             tcg_gen_shri_tl(r_tmp1, r_tmp1, 31);
             tcg_gen_brcondi_tl(TCG_COND_EQ, r_tmp1, 0, l1);
+            tcg_temp_free(r_tmp1);
             /* operands of same sign, result different sign */
             generate_exception(ctx, EXCP_OVERFLOW);
             gen_set_label(l1);
 
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ext32s_tl(t0, t0);
         }
         opn = "addi";
         break;
     case OPC_ADDIU:
-        tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
-        tcg_gen_addi_tl(cpu_T[0], cpu_T[0], uimm);
-        tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+        tcg_gen_ext32s_tl(t0, t0);
+        tcg_gen_addi_tl(t0, t0, uimm);
+        tcg_gen_ext32s_tl(t0, t0);
         opn = "addiu";
         break;
 #if defined(TARGET_MIPS64)
     case OPC_DADDI:
         {
-            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_TL);
+            TCGv r_tmp1 = tcg_temp_local_new(TCG_TYPE_TL);
             TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_TL);
             int l1 = gen_new_label();
 
             save_cpu_state(ctx, 1);
-            tcg_gen_mov_tl(r_tmp1, cpu_T[0]);
-            tcg_gen_addi_tl(cpu_T[0], cpu_T[0], uimm);
+            tcg_gen_mov_tl(r_tmp1, t0);
+            tcg_gen_addi_tl(t0, t0, uimm);
 
             tcg_gen_xori_tl(r_tmp1, r_tmp1, uimm);
             tcg_gen_xori_tl(r_tmp1, r_tmp1, -1);
-            tcg_gen_xori_tl(r_tmp2, cpu_T[0], uimm);
+            tcg_gen_xori_tl(r_tmp2, t0, uimm);
             tcg_gen_and_tl(r_tmp1, r_tmp1, r_tmp2);
+            tcg_temp_free(r_tmp2);
             tcg_gen_shri_tl(r_tmp1, r_tmp1, 63);
             tcg_gen_brcondi_tl(TCG_COND_EQ, r_tmp1, 0, l1);
+            tcg_temp_free(r_tmp1);
             /* operands of same sign, result different sign */
             generate_exception(ctx, EXCP_OVERFLOW);
             gen_set_label(l1);
@@ -1335,75 +1390,75 @@ static void gen_arith_imm (CPUState *env, DisasContext *ctx, uint32_t opc,
         opn = "daddi";
         break;
     case OPC_DADDIU:
-        tcg_gen_addi_tl(cpu_T[0], cpu_T[0], uimm);
+        tcg_gen_addi_tl(t0, t0, uimm);
         opn = "daddiu";
         break;
 #endif
     case OPC_SLTI:
-        gen_op_lti(uimm);
+        gen_op_lti(t0, uimm);
         opn = "slti";
         break;
     case OPC_SLTIU:
-        gen_op_ltiu(uimm);
+        gen_op_ltiu(t0, uimm);
         opn = "sltiu";
         break;
     case OPC_ANDI:
-        tcg_gen_andi_tl(cpu_T[0], cpu_T[0], uimm);
+        tcg_gen_andi_tl(t0, t0, uimm);
         opn = "andi";
         break;
     case OPC_ORI:
-        tcg_gen_ori_tl(cpu_T[0], cpu_T[0], uimm);
+        tcg_gen_ori_tl(t0, t0, uimm);
         opn = "ori";
         break;
     case OPC_XORI:
-        tcg_gen_xori_tl(cpu_T[0], cpu_T[0], uimm);
+        tcg_gen_xori_tl(t0, t0, uimm);
         opn = "xori";
         break;
     case OPC_LUI:
         opn = "lui";
         break;
     case OPC_SLL:
-        tcg_gen_ext32u_tl(cpu_T[0], cpu_T[0]);
-        tcg_gen_shli_tl(cpu_T[0], cpu_T[0], uimm);
-        tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+        tcg_gen_ext32u_tl(t0, t0);
+        tcg_gen_shli_tl(t0, t0, uimm);
+        tcg_gen_ext32s_tl(t0, t0);
         opn = "sll";
         break;
     case OPC_SRA:
-        tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
-        tcg_gen_sari_tl(cpu_T[0], cpu_T[0], uimm);
-        tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+        tcg_gen_ext32s_tl(t0, t0);
+        tcg_gen_sari_tl(t0, t0, uimm);
+        tcg_gen_ext32s_tl(t0, t0);
         opn = "sra";
         break;
     case OPC_SRL:
         switch ((ctx->opcode >> 21) & 0x1f) {
         case 0:
-            tcg_gen_ext32u_tl(cpu_T[0], cpu_T[0]);
-            tcg_gen_shri_tl(cpu_T[0], cpu_T[0], uimm);
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ext32u_tl(t0, t0);
+            tcg_gen_shri_tl(t0, t0, uimm);
+            tcg_gen_ext32s_tl(t0, t0);
             opn = "srl";
             break;
         case 1:
             /* rotr is decoded as srl on non-R2 CPUs */
             if (env->insn_flags & ISA_MIPS32R2) {
                 if (uimm != 0) {
-                    TCGv r_tmp1 = new_tmp();
-                    TCGv r_tmp2 = new_tmp();
+                    TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I32);
+                    TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I32);
 
-                    tcg_gen_trunc_tl_i32(r_tmp1, cpu_T[0]);
+                    tcg_gen_trunc_tl_i32(r_tmp1, t0);
                     tcg_gen_movi_i32(r_tmp2, 0x20);
                     tcg_gen_subi_i32(r_tmp2, r_tmp2, uimm);
                     tcg_gen_shl_i32(r_tmp2, r_tmp1, r_tmp2);
                     tcg_gen_shri_i32(r_tmp1, r_tmp1, uimm);
                     tcg_gen_or_i32(r_tmp1, r_tmp1, r_tmp2);
-                    tcg_gen_ext_i32_tl(r_tmp1, cpu_T[0]);
-                    dead_tmp(r_tmp1);
-                    dead_tmp(r_tmp2);
+                    tcg_gen_ext_i32_tl(t0, r_tmp1);
+                    tcg_temp_free(r_tmp1);
+                    tcg_temp_free(r_tmp2);
                 }
                 opn = "rotr";
             } else {
-                tcg_gen_ext32u_tl(cpu_T[0], cpu_T[0]);
-                tcg_gen_shri_tl(cpu_T[0], cpu_T[0], uimm);
-                tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+                tcg_gen_ext32u_tl(t0, t0);
+                tcg_gen_shri_tl(t0, t0, uimm);
+                tcg_gen_ext32s_tl(t0, t0);
                 opn = "srl";
             }
             break;
@@ -1415,17 +1470,17 @@ static void gen_arith_imm (CPUState *env, DisasContext *ctx, uint32_t opc,
         break;
 #if defined(TARGET_MIPS64)
     case OPC_DSLL:
-        tcg_gen_shli_tl(cpu_T[0], cpu_T[0], uimm);
+        tcg_gen_shli_tl(t0, t0, uimm);
         opn = "dsll";
         break;
     case OPC_DSRA:
-        tcg_gen_sari_tl(cpu_T[0], cpu_T[0], uimm);
+        tcg_gen_sari_tl(t0, t0, uimm);
         opn = "dsra";
         break;
     case OPC_DSRL:
         switch ((ctx->opcode >> 21) & 0x1f) {
         case 0:
-            tcg_gen_shri_tl(cpu_T[0], cpu_T[0], uimm);
+            tcg_gen_shri_tl(t0, t0, uimm);
             opn = "dsrl";
             break;
         case 1:
@@ -1436,13 +1491,14 @@ static void gen_arith_imm (CPUState *env, DisasContext *ctx, uint32_t opc,
 
                     tcg_gen_movi_tl(r_tmp1, 0x40);
                     tcg_gen_subi_tl(r_tmp1, r_tmp1, uimm);
-                    tcg_gen_shl_tl(r_tmp1, cpu_T[0], r_tmp1);
-                    tcg_gen_shri_tl(cpu_T[0], cpu_T[0], uimm);
-                    tcg_gen_or_tl(cpu_T[0], cpu_T[0], r_tmp1);
+                    tcg_gen_shl_tl(r_tmp1, t0, r_tmp1);
+                    tcg_gen_shri_tl(t0, t0, uimm);
+                    tcg_gen_or_tl(t0, t0, r_tmp1);
+                    tcg_temp_free(r_tmp1);
                 }
                 opn = "drotr";
             } else {
-                tcg_gen_shri_tl(cpu_T[0], cpu_T[0], uimm);
+                tcg_gen_shri_tl(t0, t0, uimm);
                 opn = "dsrl";
             }
             break;
@@ -1453,17 +1509,17 @@ static void gen_arith_imm (CPUState *env, DisasContext *ctx, uint32_t opc,
         }
         break;
     case OPC_DSLL32:
-        tcg_gen_shli_tl(cpu_T[0], cpu_T[0], uimm + 32);
+        tcg_gen_shli_tl(t0, t0, uimm + 32);
         opn = "dsll32";
         break;
     case OPC_DSRA32:
-        tcg_gen_sari_tl(cpu_T[0], cpu_T[0], uimm + 32);
+        tcg_gen_sari_tl(t0, t0, uimm + 32);
         opn = "dsra32";
         break;
     case OPC_DSRL32:
         switch ((ctx->opcode >> 21) & 0x1f) {
         case 0:
-            tcg_gen_shri_tl(cpu_T[0], cpu_T[0], uimm + 32);
+            tcg_gen_shri_tl(t0, t0, uimm + 32);
             opn = "dsrl32";
             break;
         case 1:
@@ -1476,12 +1532,14 @@ static void gen_arith_imm (CPUState *env, DisasContext *ctx, uint32_t opc,
                 tcg_gen_movi_tl(r_tmp2, 32);
                 tcg_gen_addi_tl(r_tmp2, r_tmp2, uimm);
                 tcg_gen_sub_tl(r_tmp1, r_tmp1, r_tmp2);
-                tcg_gen_shl_tl(r_tmp1, cpu_T[0], r_tmp1);
-                tcg_gen_shr_tl(cpu_T[0], cpu_T[0], r_tmp2);
-                tcg_gen_or_tl(cpu_T[0], cpu_T[0], r_tmp1);
+                tcg_gen_shl_tl(r_tmp1, t0, r_tmp1);
+                tcg_gen_shr_tl(t0, t0, r_tmp2);
+                tcg_gen_or_tl(t0, t0, r_tmp1);
+                tcg_temp_free(r_tmp1);
+                tcg_temp_free(r_tmp2);
                 opn = "drotr32";
             } else {
-                tcg_gen_shri_tl(cpu_T[0], cpu_T[0], uimm + 32);
+                tcg_gen_shri_tl(t0, t0, uimm + 32);
                 opn = "dsrl32";
             }
             break;
@@ -1495,10 +1553,12 @@ static void gen_arith_imm (CPUState *env, DisasContext *ctx, uint32_t opc,
     default:
         MIPS_INVAL(opn);
         generate_exception(ctx, EXCP_RI);
-        return;
+        goto out;
     }
-    gen_store_gpr(cpu_T[0], rt);
+    gen_store_gpr(t0, rt);
     MIPS_DEBUG("%s %s, %s, " TARGET_FMT_lx, opn, regnames[rt], regnames[rs], uimm);
+ out:
+    tcg_temp_free(t0);
 }
 
 /* Arithmetic */
@@ -1506,103 +1566,111 @@ static void gen_arith (CPUState *env, DisasContext *ctx, uint32_t opc,
                        int rd, int rs, int rt)
 {
     const char *opn = "arith";
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+    TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
 
     if (rd == 0 && opc != OPC_ADD && opc != OPC_SUB
        && opc != OPC_DADD && opc != OPC_DSUB) {
         /* If no destination, treat it as a NOP.
            For add & sub, we must generate the overflow exception when needed. */
         MIPS_DEBUG("NOP");
-        return;
+        goto out;
     }
-    gen_load_gpr(cpu_T[0], rs);
+    gen_load_gpr(t0, rs);
     /* Specialcase the conventional move operation. */
     if (rt == 0 && (opc == OPC_ADDU || opc == OPC_DADDU
                     || opc == OPC_SUBU || opc == OPC_DSUBU)) {
-        gen_store_gpr(cpu_T[0], rd);
-        return;
+        gen_store_gpr(t0, rd);
+        goto out;
     }
-    gen_load_gpr(cpu_T[1], rt);
+    gen_load_gpr(t1, rt);
     switch (opc) {
     case OPC_ADD:
         {
-            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_TL);
+            TCGv r_tmp1 = tcg_temp_local_new(TCG_TYPE_TL);
             TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_TL);
             int l1 = gen_new_label();
 
             save_cpu_state(ctx, 1);
-            tcg_gen_ext32s_tl(r_tmp1, cpu_T[0]);
-            tcg_gen_ext32s_tl(r_tmp2, cpu_T[1]);
-            tcg_gen_add_tl(cpu_T[0], r_tmp1, r_tmp2);
+            tcg_gen_ext32s_tl(r_tmp1, t0);
+            tcg_gen_ext32s_tl(r_tmp2, t1);
+            tcg_gen_add_tl(t0, r_tmp1, r_tmp2);
 
-            tcg_gen_xor_tl(r_tmp1, r_tmp1, cpu_T[1]);
+            tcg_gen_xor_tl(r_tmp1, r_tmp1, t1);
             tcg_gen_xori_tl(r_tmp1, r_tmp1, -1);
-            tcg_gen_xor_tl(r_tmp2, cpu_T[0], cpu_T[1]);
+            tcg_gen_xor_tl(r_tmp2, t0, t1);
             tcg_gen_and_tl(r_tmp1, r_tmp1, r_tmp2);
+            tcg_temp_free(r_tmp2);
             tcg_gen_shri_tl(r_tmp1, r_tmp1, 31);
             tcg_gen_brcondi_tl(TCG_COND_EQ, r_tmp1, 0, l1);
+            tcg_temp_free(r_tmp1);
             /* operands of same sign, result different sign */
             generate_exception(ctx, EXCP_OVERFLOW);
             gen_set_label(l1);
 
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ext32s_tl(t0, t0);
         }
         opn = "add";
         break;
     case OPC_ADDU:
-        tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
-        tcg_gen_ext32s_tl(cpu_T[1], cpu_T[1]);
-        tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
-        tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+        tcg_gen_ext32s_tl(t0, t0);
+        tcg_gen_ext32s_tl(t1, t1);
+        tcg_gen_add_tl(t0, t0, t1);
+        tcg_gen_ext32s_tl(t0, t0);
         opn = "addu";
         break;
     case OPC_SUB:
         {
-            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_TL);
+            TCGv r_tmp1 = tcg_temp_local_new(TCG_TYPE_TL);
             TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_TL);
             int l1 = gen_new_label();
 
             save_cpu_state(ctx, 1);
-            tcg_gen_ext32s_tl(r_tmp1, cpu_T[0]);
-            tcg_gen_ext32s_tl(r_tmp2, cpu_T[1]);
-            tcg_gen_sub_tl(cpu_T[0], r_tmp1, r_tmp2);
+            tcg_gen_ext32s_tl(r_tmp1, t0);
+            tcg_gen_ext32s_tl(r_tmp2, t1);
+            tcg_gen_sub_tl(t0, r_tmp1, r_tmp2);
 
-            tcg_gen_xor_tl(r_tmp2, r_tmp1, cpu_T[1]);
-            tcg_gen_xor_tl(r_tmp1, r_tmp1, cpu_T[0]);
+            tcg_gen_xor_tl(r_tmp2, r_tmp1, t1);
+            tcg_gen_xor_tl(r_tmp1, r_tmp1, t0);
             tcg_gen_and_tl(r_tmp1, r_tmp1, r_tmp2);
+            tcg_temp_free(r_tmp2);
             tcg_gen_shri_tl(r_tmp1, r_tmp1, 31);
             tcg_gen_brcondi_tl(TCG_COND_EQ, r_tmp1, 0, l1);
+            tcg_temp_free(r_tmp1);
             /* operands of different sign, first operand and result different sign */
             generate_exception(ctx, EXCP_OVERFLOW);
             gen_set_label(l1);
 
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ext32s_tl(t0, t0);
         }
         opn = "sub";
         break;
     case OPC_SUBU:
-        tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
-        tcg_gen_ext32s_tl(cpu_T[1], cpu_T[1]);
-        tcg_gen_sub_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
-        tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+        tcg_gen_ext32s_tl(t0, t0);
+        tcg_gen_ext32s_tl(t1, t1);
+        tcg_gen_sub_tl(t0, t0, t1);
+        tcg_gen_ext32s_tl(t0, t0);
         opn = "subu";
         break;
 #if defined(TARGET_MIPS64)
     case OPC_DADD:
         {
-            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_TL);
+            TCGv r_tmp1 = tcg_temp_local_new(TCG_TYPE_TL);
             TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_TL);
             int l1 = gen_new_label();
 
             save_cpu_state(ctx, 1);
-            tcg_gen_mov_tl(r_tmp1, cpu_T[0]);
-            tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
+            tcg_gen_mov_tl(r_tmp1, t0);
+            tcg_gen_add_tl(t0, t0, t1);
 
-            tcg_gen_xor_tl(r_tmp1, r_tmp1, cpu_T[1]);
+            tcg_gen_xor_tl(r_tmp1, r_tmp1, t1);
             tcg_gen_xori_tl(r_tmp1, r_tmp1, -1);
-            tcg_gen_xor_tl(r_tmp2, cpu_T[0], cpu_T[1]);
+            tcg_gen_xor_tl(r_tmp2, t0, t1);
             tcg_gen_and_tl(r_tmp1, r_tmp1, r_tmp2);
+            tcg_temp_free(r_tmp2);
             tcg_gen_shri_tl(r_tmp1, r_tmp1, 63);
             tcg_gen_brcondi_tl(TCG_COND_EQ, r_tmp1, 0, l1);
+            tcg_temp_free(r_tmp1);
             /* operands of same sign, result different sign */
             generate_exception(ctx, EXCP_OVERFLOW);
             gen_set_label(l1);
@@ -1610,24 +1678,26 @@ static void gen_arith (CPUState *env, DisasContext *ctx, uint32_t opc,
         opn = "dadd";
         break;
     case OPC_DADDU:
-        tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
+        tcg_gen_add_tl(t0, t0, t1);
         opn = "daddu";
         break;
     case OPC_DSUB:
         {
-            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_TL);
+            TCGv r_tmp1 = tcg_temp_local_new(TCG_TYPE_TL);
             TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_TL);
             int l1 = gen_new_label();
 
             save_cpu_state(ctx, 1);
-            tcg_gen_mov_tl(r_tmp1, cpu_T[0]);
-            tcg_gen_sub_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
+            tcg_gen_mov_tl(r_tmp1, t0);
+            tcg_gen_sub_tl(t0, t0, t1);
 
-            tcg_gen_xor_tl(r_tmp2, r_tmp1, cpu_T[1]);
-            tcg_gen_xor_tl(r_tmp1, r_tmp1, cpu_T[0]);
+            tcg_gen_xor_tl(r_tmp2, r_tmp1, t1);
+            tcg_gen_xor_tl(r_tmp1, r_tmp1, t0);
             tcg_gen_and_tl(r_tmp1, r_tmp1, r_tmp2);
+            tcg_temp_free(r_tmp2);
             tcg_gen_shri_tl(r_tmp1, r_tmp1, 63);
             tcg_gen_brcondi_tl(TCG_COND_EQ, r_tmp1, 0, l1);
+            tcg_temp_free(r_tmp1);
             /* operands of different sign, first operand and result different sign */
             generate_exception(ctx, EXCP_OVERFLOW);
             gen_set_label(l1);
@@ -1635,48 +1705,48 @@ static void gen_arith (CPUState *env, DisasContext *ctx, uint32_t opc,
         opn = "dsub";
         break;
     case OPC_DSUBU:
-        tcg_gen_sub_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
+        tcg_gen_sub_tl(t0, t0, t1);
         opn = "dsubu";
         break;
 #endif
     case OPC_SLT:
-        gen_op_lt();
+        gen_op_lt(t0, t1);
         opn = "slt";
         break;
     case OPC_SLTU:
-        gen_op_ltu();
+        gen_op_ltu(t0, t1);
         opn = "sltu";
         break;
     case OPC_AND:
-        tcg_gen_and_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
+        tcg_gen_and_tl(t0, t0, t1);
         opn = "and";
         break;
     case OPC_NOR:
-        tcg_gen_or_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
-        tcg_gen_not_tl(cpu_T[0], cpu_T[0]);
+        tcg_gen_or_tl(t0, t0, t1);
+        tcg_gen_not_tl(t0, t0);
         opn = "nor";
         break;
     case OPC_OR:
-        tcg_gen_or_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
+        tcg_gen_or_tl(t0, t0, t1);
         opn = "or";
         break;
     case OPC_XOR:
-        tcg_gen_xor_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
+        tcg_gen_xor_tl(t0, t0, t1);
         opn = "xor";
         break;
     case OPC_MUL:
-        tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
-        tcg_gen_ext32s_tl(cpu_T[1], cpu_T[1]);
-        tcg_gen_mul_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
-        tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+        tcg_gen_ext32s_tl(t0, t0);
+        tcg_gen_ext32s_tl(t1, t1);
+        tcg_gen_mul_tl(t0, t0, t1);
+        tcg_gen_ext32s_tl(t0, t0);
         opn = "mul";
         break;
     case OPC_MOVN:
         {
             int l1 = gen_new_label();
 
-            tcg_gen_brcondi_tl(TCG_COND_EQ, cpu_T[1], 0, l1);
-            gen_store_gpr(cpu_T[0], rd);
+            tcg_gen_brcondi_tl(TCG_COND_EQ, t1, 0, l1);
+            gen_store_gpr(t0, rd);
             gen_set_label(l1);
         }
         opn = "movn";
@@ -1685,34 +1755,34 @@ static void gen_arith (CPUState *env, DisasContext *ctx, uint32_t opc,
         {
             int l1 = gen_new_label();
 
-            tcg_gen_brcondi_tl(TCG_COND_NE, cpu_T[1], 0, l1);
-            gen_store_gpr(cpu_T[0], rd);
+            tcg_gen_brcondi_tl(TCG_COND_NE, t1, 0, l1);
+            gen_store_gpr(t0, rd);
             gen_set_label(l1);
         }
         opn = "movz";
         goto print;
     case OPC_SLLV:
-        tcg_gen_ext32u_tl(cpu_T[0], cpu_T[0]);
-        tcg_gen_ext32u_tl(cpu_T[1], cpu_T[1]);
-        tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0x1f);
-        tcg_gen_shl_tl(cpu_T[0], cpu_T[1], cpu_T[0]);
-        tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+        tcg_gen_ext32u_tl(t0, t0);
+        tcg_gen_ext32u_tl(t1, t1);
+        tcg_gen_andi_tl(t0, t0, 0x1f);
+        tcg_gen_shl_tl(t0, t1, t0);
+        tcg_gen_ext32s_tl(t0, t0);
         opn = "sllv";
         break;
     case OPC_SRAV:
-        tcg_gen_ext32s_tl(cpu_T[1], cpu_T[1]);
-        tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0x1f);
-        tcg_gen_sar_tl(cpu_T[0], cpu_T[1], cpu_T[0]);
-        tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+        tcg_gen_ext32s_tl(t1, t1);
+        tcg_gen_andi_tl(t0, t0, 0x1f);
+        tcg_gen_sar_tl(t0, t1, t0);
+        tcg_gen_ext32s_tl(t0, t0);
         opn = "srav";
         break;
     case OPC_SRLV:
         switch ((ctx->opcode >> 6) & 0x1f) {
         case 0:
-            tcg_gen_ext32u_tl(cpu_T[1], cpu_T[1]);
-            tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0x1f);
-            tcg_gen_shr_tl(cpu_T[0], cpu_T[1], cpu_T[0]);
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ext32u_tl(t1, t1);
+            tcg_gen_andi_tl(t0, t0, 0x1f);
+            tcg_gen_shr_tl(t0, t1, t0);
+            tcg_gen_ext32s_tl(t0, t0);
             opn = "srlv";
             break;
         case 1:
@@ -1721,35 +1791,35 @@ static void gen_arith (CPUState *env, DisasContext *ctx, uint32_t opc,
                 int l1 = gen_new_label();
                 int l2 = gen_new_label();
 
-                tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0x1f);
-                tcg_gen_brcondi_tl(TCG_COND_EQ, cpu_T[0], 0, l1);
+                tcg_gen_andi_tl(t0, t0, 0x1f);
+                tcg_gen_brcondi_tl(TCG_COND_EQ, t0, 0, l1);
                 {
-                    TCGv r_tmp1 = new_tmp();
-                    TCGv r_tmp2 = new_tmp();
-                    TCGv r_tmp3 = new_tmp();
+                    TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I32);
+                    TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I32);
+                    TCGv r_tmp3 = tcg_temp_new(TCG_TYPE_I32);
 
-                    tcg_gen_trunc_tl_i32(r_tmp1, cpu_T[0]);
-                    tcg_gen_trunc_tl_i32(r_tmp2, cpu_T[1]);
+                    tcg_gen_trunc_tl_i32(r_tmp1, t0);
+                    tcg_gen_trunc_tl_i32(r_tmp2, t1);
                     tcg_gen_movi_i32(r_tmp3, 0x20);
                     tcg_gen_sub_i32(r_tmp3, r_tmp3, r_tmp1);
                     tcg_gen_shl_i32(r_tmp3, r_tmp2, r_tmp3);
                     tcg_gen_shr_i32(r_tmp1, r_tmp2, r_tmp1);
                     tcg_gen_or_i32(r_tmp1, r_tmp1, r_tmp3);
-                    tcg_gen_ext_i32_tl(r_tmp1, cpu_T[0]);
-                    dead_tmp(r_tmp1);
-                    dead_tmp(r_tmp2);
-                    dead_tmp(r_tmp3);
+                    tcg_gen_ext_i32_tl(t0, r_tmp1);
+                    tcg_temp_free(r_tmp1);
+                    tcg_temp_free(r_tmp2);
+                    tcg_temp_free(r_tmp3);
                     tcg_gen_br(l2);
                 }
                 gen_set_label(l1);
-                tcg_gen_mov_tl(cpu_T[0], cpu_T[1]);
+                tcg_gen_mov_tl(t0, t1);
                 gen_set_label(l2);
                 opn = "rotrv";
             } else {
-                tcg_gen_ext32u_tl(cpu_T[1], cpu_T[1]);
-                tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0x1f);
-                tcg_gen_shr_tl(cpu_T[0], cpu_T[1], cpu_T[0]);
-                tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+                tcg_gen_ext32u_tl(t1, t1);
+                tcg_gen_andi_tl(t0, t0, 0x1f);
+                tcg_gen_shr_tl(t0, t1, t0);
+                tcg_gen_ext32s_tl(t0, t0);
                 opn = "srlv";
             }
             break;
@@ -1761,20 +1831,20 @@ static void gen_arith (CPUState *env, DisasContext *ctx, uint32_t opc,
         break;
 #if defined(TARGET_MIPS64)
     case OPC_DSLLV:
-        tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0x3f);
-        tcg_gen_shl_tl(cpu_T[0], cpu_T[1], cpu_T[0]);
+        tcg_gen_andi_tl(t0, t0, 0x3f);
+        tcg_gen_shl_tl(t0, t1, t0);
         opn = "dsllv";
         break;
     case OPC_DSRAV:
-        tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0x3f);
-        tcg_gen_sar_tl(cpu_T[0], cpu_T[1], cpu_T[0]);
+        tcg_gen_andi_tl(t0, t0, 0x3f);
+        tcg_gen_sar_tl(t0, t1, t0);
         opn = "dsrav";
         break;
     case OPC_DSRLV:
         switch ((ctx->opcode >> 6) & 0x1f) {
         case 0:
-            tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0x3f);
-            tcg_gen_shr_tl(cpu_T[0], cpu_T[1], cpu_T[0]);
+            tcg_gen_andi_tl(t0, t0, 0x3f);
+            tcg_gen_shr_tl(t0, t1, t0);
             opn = "dsrlv";
             break;
         case 1:
@@ -1783,25 +1853,26 @@ static void gen_arith (CPUState *env, DisasContext *ctx, uint32_t opc,
                 int l1 = gen_new_label();
                 int l2 = gen_new_label();
 
-                tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0x3f);
-                tcg_gen_brcondi_tl(TCG_COND_EQ, cpu_T[0], 0, l1);
+                tcg_gen_andi_tl(t0, t0, 0x3f);
+                tcg_gen_brcondi_tl(TCG_COND_EQ, t0, 0, l1);
                 {
                     TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_TL);
 
                     tcg_gen_movi_tl(r_tmp1, 0x40);
-                    tcg_gen_sub_tl(r_tmp1, r_tmp1, cpu_T[0]);
-                    tcg_gen_shl_tl(r_tmp1, cpu_T[1], r_tmp1);
-                    tcg_gen_shr_tl(cpu_T[0], cpu_T[1], cpu_T[0]);
-                    tcg_gen_or_tl(cpu_T[0], cpu_T[0], r_tmp1);
+                    tcg_gen_sub_tl(r_tmp1, r_tmp1, t0);
+                    tcg_gen_shl_tl(r_tmp1, t1, r_tmp1);
+                    tcg_gen_shr_tl(t0, t1, t0);
+                    tcg_gen_or_tl(t0, t0, r_tmp1);
+                    tcg_temp_free(r_tmp1);
                     tcg_gen_br(l2);
                 }
                 gen_set_label(l1);
-                tcg_gen_mov_tl(cpu_T[0], cpu_T[1]);
+                tcg_gen_mov_tl(t0, t1);
                 gen_set_label(l2);
                 opn = "drotrv";
             } else {
-                tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0x3f);
-                tcg_gen_shr_tl(cpu_T[0], cpu_T[1], cpu_T[0]);
+                tcg_gen_andi_tl(t0, t0, 0x3f);
+                tcg_gen_shr_tl(t0, t1, t0);
                 opn = "dsrlv";
             }
             break;
@@ -1815,89 +1886,93 @@ static void gen_arith (CPUState *env, DisasContext *ctx, uint32_t opc,
     default:
         MIPS_INVAL(opn);
         generate_exception(ctx, EXCP_RI);
-        return;
+        goto out;
     }
-    gen_store_gpr(cpu_T[0], rd);
+    gen_store_gpr(t0, rd);
  print:
     MIPS_DEBUG("%s %s, %s, %s", opn, regnames[rd], regnames[rs], regnames[rt]);
+ out:
+    tcg_temp_free(t0);
+    tcg_temp_free(t1);
 }
 
 /* Arithmetic on HI/LO registers */
 static void gen_HILO (DisasContext *ctx, uint32_t opc, int reg)
 {
     const char *opn = "hilo";
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
 
     if (reg == 0 && (opc == OPC_MFHI || opc == OPC_MFLO)) {
         /* Treat as NOP. */
         MIPS_DEBUG("NOP");
-        return;
+        goto out;
     }
     switch (opc) {
     case OPC_MFHI:
-        tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, HI[0]));
-        gen_store_gpr(cpu_T[0], reg);
+        gen_load_HI(t0, 0);
+        gen_store_gpr(t0, reg);
         opn = "mfhi";
         break;
     case OPC_MFLO:
-        tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, LO[0]));
-        gen_store_gpr(cpu_T[0], reg);
+        gen_load_LO(t0, 0);
+        gen_store_gpr(t0, reg);
         opn = "mflo";
         break;
     case OPC_MTHI:
-        gen_load_gpr(cpu_T[0], reg);
-        tcg_gen_st_tl(cpu_T[0], cpu_env, offsetof(CPUState, HI[0]));
+        gen_load_gpr(t0, reg);
+        gen_store_HI(t0, 0);
         opn = "mthi";
         break;
     case OPC_MTLO:
-        gen_load_gpr(cpu_T[0], reg);
-        tcg_gen_st_tl(cpu_T[0], cpu_env, offsetof(CPUState, LO[0]));
+        gen_load_gpr(t0, reg);
+        gen_store_LO(t0, 0);
         opn = "mtlo";
         break;
     default:
         MIPS_INVAL(opn);
         generate_exception(ctx, EXCP_RI);
-        return;
+        goto out;
     }
     MIPS_DEBUG("%s %s", opn, regnames[reg]);
+ out:
+    tcg_temp_free(t0);
 }
 
 static void gen_muldiv (DisasContext *ctx, uint32_t opc,
                         int rs, int rt)
 {
     const char *opn = "mul/div";
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+    TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
 
-    gen_load_gpr(cpu_T[0], rs);
-    gen_load_gpr(cpu_T[1], rt);
+    gen_load_gpr(t0, rs);
+    gen_load_gpr(t1, rt);
     switch (opc) {
     case OPC_DIV:
         {
             int l1 = gen_new_label();
 
-            tcg_gen_brcondi_tl(TCG_COND_EQ, cpu_T[1], 0, l1);
+            tcg_gen_ext32s_tl(t0, t0);
+            tcg_gen_ext32s_tl(t1, t1);
+            tcg_gen_brcondi_tl(TCG_COND_EQ, t1, 0, l1);
             {
-                TCGv r_tmp1 = new_tmp();
-                TCGv r_tmp2 = new_tmp();
-                TCGv r_tmp3 = new_tmp();
-                TCGv r_tc_off = new_tmp();
-                TCGv r_tc_off_tl = tcg_temp_new(TCG_TYPE_TL);
-                TCGv r_ptr = tcg_temp_new(TCG_TYPE_PTR);
-
-                tcg_gen_ext_i32_tl(r_tmp1, cpu_T[0]);
-                tcg_gen_ext_i32_tl(r_tmp2, cpu_T[1]);
-                tcg_gen_div_i32(r_tmp3, r_tmp1, r_tmp2);
-                tcg_gen_rem_i32(r_tmp1, r_tmp1, r_tmp2);
-                tcg_gen_trunc_tl_i32(cpu_T[0], r_tmp3);
-                tcg_gen_trunc_tl_i32(cpu_T[1], r_tmp1);
-                dead_tmp(r_tmp1);
-                dead_tmp(r_tmp2);
-                dead_tmp(r_tmp3);
-                tcg_gen_ld_i32(r_tc_off, cpu_env, offsetof(CPUState, current_tc));
-                tcg_gen_muli_i32(r_tc_off, r_tc_off, sizeof(target_ulong));
-                tcg_gen_ext_i32_ptr(r_tc_off_tl, r_tc_off);
-                tcg_gen_add_ptr(r_ptr, cpu_env, r_tc_off_tl);
-                tcg_gen_st_tl(cpu_T[0], r_ptr, offsetof(CPUState, LO));
-                tcg_gen_st_tl(cpu_T[1], r_ptr, offsetof(CPUState, HI));
-                dead_tmp(r_tc_off);
+                TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I64);
+                TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I64);
+                TCGv r_tmp3 = tcg_temp_new(TCG_TYPE_I64);
+
+                tcg_gen_ext_tl_i64(r_tmp1, t0);
+                tcg_gen_ext_tl_i64(r_tmp2, t1);
+                tcg_gen_div_i64(r_tmp3, r_tmp1, r_tmp2);
+                tcg_gen_rem_i64(r_tmp2, r_tmp1, r_tmp2);
+                tcg_gen_trunc_i64_tl(t0, r_tmp3);
+                tcg_gen_trunc_i64_tl(t1, r_tmp2);
+                tcg_temp_free(r_tmp1);
+                tcg_temp_free(r_tmp2);
+                tcg_temp_free(r_tmp3);
+                tcg_gen_ext32s_tl(t0, t0);
+                tcg_gen_ext32s_tl(t1, t1);
+                gen_store_LO(t0, 0);
+                gen_store_HI(t1, 0);
             }
             gen_set_label(l1);
         }
@@ -1907,42 +1982,71 @@ static void gen_muldiv (DisasContext *ctx, uint32_t opc,
         {
             int l1 = gen_new_label();
 
-            tcg_gen_brcondi_tl(TCG_COND_EQ, cpu_T[1], 0, l1);
+            tcg_gen_ext32s_tl(t1, t1);
+            tcg_gen_brcondi_tl(TCG_COND_EQ, t1, 0, l1);
             {
-                TCGv r_tmp1 = new_tmp();
-                TCGv r_tmp2 = new_tmp();
-                TCGv r_tmp3 = new_tmp();
-                TCGv r_tc_off = new_tmp();
-                TCGv r_tc_off_tl = tcg_temp_new(TCG_TYPE_TL);
-                TCGv r_ptr = tcg_temp_new(TCG_TYPE_PTR);
-
-                tcg_gen_ext_i32_tl(r_tmp1, cpu_T[0]);
-                tcg_gen_ext_i32_tl(r_tmp2, cpu_T[1]);
+                TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I32);
+                TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I32);
+                TCGv r_tmp3 = tcg_temp_new(TCG_TYPE_I32);
+
+                tcg_gen_trunc_tl_i32(r_tmp1, t0);
+                tcg_gen_trunc_tl_i32(r_tmp2, t1);
                 tcg_gen_divu_i32(r_tmp3, r_tmp1, r_tmp2);
                 tcg_gen_remu_i32(r_tmp1, r_tmp1, r_tmp2);
-                tcg_gen_trunc_tl_i32(cpu_T[0], r_tmp3);
-                tcg_gen_trunc_tl_i32(cpu_T[1], r_tmp1);
-                dead_tmp(r_tmp1);
-                dead_tmp(r_tmp2);
-                dead_tmp(r_tmp3);
-                tcg_gen_ld_i32(r_tc_off, cpu_env, offsetof(CPUState, current_tc));
-                tcg_gen_muli_i32(r_tc_off, r_tc_off, sizeof(target_ulong));
-                tcg_gen_ext_i32_ptr(r_tc_off_tl, r_tc_off);
-                tcg_gen_add_ptr(r_ptr, cpu_env, r_tc_off_tl);
-                tcg_gen_st_tl(cpu_T[0], r_ptr, offsetof(CPUState, LO));
-                tcg_gen_st_tl(cpu_T[1], r_ptr, offsetof(CPUState, HI));
-                dead_tmp(r_tc_off);
+                tcg_gen_ext_i32_tl(t0, r_tmp3);
+                tcg_gen_ext_i32_tl(t1, r_tmp1);
+                tcg_temp_free(r_tmp1);
+                tcg_temp_free(r_tmp2);
+                tcg_temp_free(r_tmp3);
+                gen_store_LO(t0, 0);
+                gen_store_HI(t1, 0);
             }
             gen_set_label(l1);
         }
         opn = "divu";
         break;
     case OPC_MULT:
-        gen_op_mult();
+        {
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I64);
+
+            tcg_gen_ext32s_tl(t0, t0);
+            tcg_gen_ext32s_tl(t1, t1);
+            tcg_gen_ext_tl_i64(r_tmp1, t0);
+            tcg_gen_ext_tl_i64(r_tmp2, t1);
+            tcg_gen_mul_i64(r_tmp1, r_tmp1, r_tmp2);
+            tcg_temp_free(r_tmp2);
+            tcg_gen_trunc_i64_tl(t0, r_tmp1);
+            tcg_gen_shri_i64(r_tmp1, r_tmp1, 32);
+            tcg_gen_trunc_i64_tl(t1, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_ext32s_tl(t0, t0);
+            tcg_gen_ext32s_tl(t1, t1);
+            gen_store_LO(t0, 0);
+            gen_store_HI(t1, 0);
+        }
         opn = "mult";
         break;
     case OPC_MULTU:
-        gen_op_multu();
+        {
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I64);
+
+            tcg_gen_ext32u_tl(t0, t0);
+            tcg_gen_ext32u_tl(t1, t1);
+            tcg_gen_extu_tl_i64(r_tmp1, t0);
+            tcg_gen_extu_tl_i64(r_tmp2, t1);
+            tcg_gen_mul_i64(r_tmp1, r_tmp1, r_tmp2);
+            tcg_temp_free(r_tmp2);
+            tcg_gen_trunc_i64_tl(t0, r_tmp1);
+            tcg_gen_shri_i64(r_tmp1, r_tmp1, 32);
+            tcg_gen_trunc_i64_tl(t1, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_ext32s_tl(t0, t0);
+            tcg_gen_ext32s_tl(t1, t1);
+            gen_store_LO(t0, 0);
+            gen_store_HI(t1, 0);
+        }
         opn = "multu";
         break;
 #if defined(TARGET_MIPS64)
@@ -1950,31 +2054,30 @@ static void gen_muldiv (DisasContext *ctx, uint32_t opc,
         {
             int l1 = gen_new_label();
 
-            tcg_gen_brcondi_tl(TCG_COND_EQ, cpu_T[1], 0, l1);
+            tcg_gen_brcondi_tl(TCG_COND_EQ, t1, 0, l1);
             {
-                TCGv r_tc_off = new_tmp();
-                TCGv r_tc_off_tl = tcg_temp_new(TCG_TYPE_TL);
-                TCGv r_ptr = tcg_temp_new(TCG_TYPE_PTR);
                 int l2 = gen_new_label();
-                int l3 = gen_new_label();
 
-                tcg_gen_brcondi_tl(TCG_COND_NE, cpu_T[0], 1ULL << 63, l2);
-                tcg_gen_brcondi_tl(TCG_COND_NE, cpu_T[1], -1ULL, l2);
-                tcg_gen_div_i64(cpu_T[0], cpu_T[0], cpu_T[1]);
-                tcg_gen_movi_tl(cpu_T[1], 0);
-                tcg_gen_br(l3);
+                tcg_gen_brcondi_tl(TCG_COND_NE, t0, -1LL << 63, l2);
+                tcg_gen_brcondi_tl(TCG_COND_NE, t1, -1LL, l2);
+                {
+                    tcg_gen_movi_tl(t1, 0);
+                    gen_store_LO(t0, 0);
+                    gen_store_HI(t1, 0);
+                    tcg_gen_br(l1);
+                }
                 gen_set_label(l2);
-                tcg_gen_div_i64(cpu_T[0], cpu_T[0], cpu_T[1]);
-                tcg_gen_rem_i64(cpu_T[1], cpu_T[0], cpu_T[1]);
-                gen_set_label(l3);
-
-                tcg_gen_ld_i32(r_tc_off, cpu_env, offsetof(CPUState, current_tc));
-                tcg_gen_muli_i32(r_tc_off, r_tc_off, sizeof(target_ulong));
-                tcg_gen_ext_i32_ptr(r_tc_off_tl, r_tc_off);
-                tcg_gen_add_ptr(r_ptr, cpu_env, r_tc_off_tl);
-                tcg_gen_st_tl(cpu_T[0], r_ptr, offsetof(CPUState, LO));
-                tcg_gen_st_tl(cpu_T[1], r_ptr, offsetof(CPUState, HI));
-                dead_tmp(r_tc_off);
+                {
+                    TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I64);
+                    TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I64);
+
+                    tcg_gen_div_i64(r_tmp1, t0, t1);
+                    tcg_gen_rem_i64(r_tmp2, t0, t1);
+                    gen_store_LO(r_tmp1, 0);
+                    gen_store_HI(r_tmp2, 0);
+                    tcg_temp_free(r_tmp1);
+                    tcg_temp_free(r_tmp2);
+                }
             }
             gen_set_label(l1);
         }
@@ -1984,171 +2087,287 @@ static void gen_muldiv (DisasContext *ctx, uint32_t opc,
         {
             int l1 = gen_new_label();
 
-            tcg_gen_brcondi_tl(TCG_COND_EQ, cpu_T[1], 0, l1);
+            tcg_gen_brcondi_tl(TCG_COND_EQ, t1, 0, l1);
             {
                 TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I64);
                 TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I64);
-                TCGv r_tc_off = new_tmp();
-                TCGv r_tc_off_tl = tcg_temp_new(TCG_TYPE_TL);
-                TCGv r_ptr = tcg_temp_new(TCG_TYPE_PTR);
-
-                tcg_gen_divu_i64(r_tmp1, cpu_T[0], cpu_T[1]);
-                tcg_gen_remu_i64(r_tmp2, cpu_T[0], cpu_T[1]);
-                tcg_gen_ld_i32(r_tc_off, cpu_env, offsetof(CPUState, current_tc));
-                tcg_gen_muli_i32(r_tc_off, r_tc_off, sizeof(target_ulong));
-                tcg_gen_ext_i32_ptr(r_tc_off_tl, r_tc_off);
-                tcg_gen_add_ptr(r_ptr, cpu_env, r_tc_off_tl);
-                tcg_gen_st_tl(r_tmp1, r_ptr, offsetof(CPUState, LO));
-                tcg_gen_st_tl(r_tmp2, r_ptr, offsetof(CPUState, HI));
-                dead_tmp(r_tc_off);
+
+                tcg_gen_divu_i64(r_tmp1, t0, t1);
+                tcg_gen_remu_i64(r_tmp2, t0, t1);
+                tcg_temp_free(r_tmp1);
+                tcg_temp_free(r_tmp2);
+                gen_store_LO(r_tmp1, 0);
+                gen_store_HI(r_tmp2, 0);
             }
             gen_set_label(l1);
         }
         opn = "ddivu";
         break;
     case OPC_DMULT:
-        gen_op_dmult();
+        tcg_gen_helper_0_2(do_dmult, t0, t1);
         opn = "dmult";
         break;
     case OPC_DMULTU:
-        gen_op_dmultu();
+        tcg_gen_helper_0_2(do_dmultu, t0, t1);
         opn = "dmultu";
         break;
 #endif
     case OPC_MADD:
-        gen_op_madd();
+        {
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv r_tmp3 = tcg_temp_new(TCG_TYPE_I64);
+
+            tcg_gen_ext32s_tl(t0, t0);
+            tcg_gen_ext32s_tl(t1, t1);
+            tcg_gen_ext_tl_i64(r_tmp1, t0);
+            tcg_gen_ext_tl_i64(r_tmp2, t1);
+            tcg_gen_mul_i64(r_tmp1, r_tmp1, r_tmp2);
+            gen_load_LO(t0, 0);
+            gen_load_HI(t1, 0);
+            tcg_gen_extu_tl_i64(r_tmp2, t0);
+            tcg_gen_extu_tl_i64(r_tmp3, t1);
+            tcg_gen_shli_i64(r_tmp3, r_tmp3, 32);
+            tcg_gen_or_i64(r_tmp2, r_tmp2, r_tmp3);
+            tcg_temp_free(r_tmp3);
+            tcg_gen_add_i64(r_tmp1, r_tmp1, r_tmp2);
+            tcg_temp_free(r_tmp2);
+            tcg_gen_trunc_i64_tl(t0, r_tmp1);
+            tcg_gen_shri_i64(r_tmp1, r_tmp1, 32);
+            tcg_gen_trunc_i64_tl(t1, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_ext32s_tl(t0, t0);
+            tcg_gen_ext32s_tl(t1, t1);
+            gen_store_LO(t0, 0);
+            gen_store_HI(t1, 0);
+        }
         opn = "madd";
         break;
     case OPC_MADDU:
-        gen_op_maddu();
+       {
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv r_tmp3 = tcg_temp_new(TCG_TYPE_I64);
+
+            tcg_gen_ext32u_tl(t0, t0);
+            tcg_gen_ext32u_tl(t1, t1);
+            tcg_gen_extu_tl_i64(r_tmp1, t0);
+            tcg_gen_extu_tl_i64(r_tmp2, t1);
+            tcg_gen_mul_i64(r_tmp1, r_tmp1, r_tmp2);
+            gen_load_LO(t0, 0);
+            gen_load_HI(t1, 0);
+            tcg_gen_extu_tl_i64(r_tmp2, t0);
+            tcg_gen_extu_tl_i64(r_tmp3, t1);
+            tcg_gen_shli_i64(r_tmp3, r_tmp3, 32);
+            tcg_gen_or_i64(r_tmp2, r_tmp2, r_tmp3);
+            tcg_temp_free(r_tmp3);
+            tcg_gen_add_i64(r_tmp1, r_tmp1, r_tmp2);
+            tcg_temp_free(r_tmp2);
+            tcg_gen_trunc_i64_tl(t0, r_tmp1);
+            tcg_gen_shri_i64(r_tmp1, r_tmp1, 32);
+            tcg_gen_trunc_i64_tl(t1, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_ext32s_tl(t0, t0);
+            tcg_gen_ext32s_tl(t1, t1);
+            gen_store_LO(t0, 0);
+            gen_store_HI(t1, 0);
+        }
         opn = "maddu";
         break;
     case OPC_MSUB:
-        gen_op_msub();
+        {
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv r_tmp3 = tcg_temp_new(TCG_TYPE_I64);
+
+            tcg_gen_ext32s_tl(t0, t0);
+            tcg_gen_ext32s_tl(t1, t1);
+            tcg_gen_ext_tl_i64(r_tmp1, t0);
+            tcg_gen_ext_tl_i64(r_tmp2, t1);
+            tcg_gen_mul_i64(r_tmp1, r_tmp1, r_tmp2);
+            gen_load_LO(t0, 0);
+            gen_load_HI(t1, 0);
+            tcg_gen_extu_tl_i64(r_tmp2, t0);
+            tcg_gen_extu_tl_i64(r_tmp3, t1);
+            tcg_gen_shli_i64(r_tmp3, r_tmp3, 32);
+            tcg_gen_or_i64(r_tmp2, r_tmp2, r_tmp3);
+            tcg_temp_free(r_tmp3);
+            tcg_gen_sub_i64(r_tmp1, r_tmp1, r_tmp2);
+            tcg_temp_free(r_tmp2);
+            tcg_gen_trunc_i64_tl(t0, r_tmp1);
+            tcg_gen_shri_i64(r_tmp1, r_tmp1, 32);
+            tcg_gen_trunc_i64_tl(t1, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_ext32s_tl(t0, t0);
+            tcg_gen_ext32s_tl(t1, t1);
+            gen_store_LO(t0, 0);
+            gen_store_HI(t1, 0);
+        }
         opn = "msub";
         break;
     case OPC_MSUBU:
-        gen_op_msubu();
+        {
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv r_tmp2 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv r_tmp3 = tcg_temp_new(TCG_TYPE_I64);
+
+            tcg_gen_ext32u_tl(t0, t0);
+            tcg_gen_ext32u_tl(t1, t1);
+            tcg_gen_extu_tl_i64(r_tmp1, t0);
+            tcg_gen_extu_tl_i64(r_tmp2, t1);
+            tcg_gen_mul_i64(r_tmp1, r_tmp1, r_tmp2);
+            gen_load_LO(t0, 0);
+            gen_load_HI(t1, 0);
+            tcg_gen_extu_tl_i64(r_tmp2, t0);
+            tcg_gen_extu_tl_i64(r_tmp3, t1);
+            tcg_gen_shli_i64(r_tmp3, r_tmp3, 32);
+            tcg_gen_or_i64(r_tmp2, r_tmp2, r_tmp3);
+            tcg_temp_free(r_tmp3);
+            tcg_gen_sub_i64(r_tmp1, r_tmp1, r_tmp2);
+            tcg_temp_free(r_tmp2);
+            tcg_gen_trunc_i64_tl(t0, r_tmp1);
+            tcg_gen_shri_i64(r_tmp1, r_tmp1, 32);
+            tcg_gen_trunc_i64_tl(t1, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_ext32s_tl(t0, t0);
+            tcg_gen_ext32s_tl(t1, t1);
+            gen_store_LO(t0, 0);
+            gen_store_HI(t1, 0);
+        }
         opn = "msubu";
         break;
     default:
         MIPS_INVAL(opn);
         generate_exception(ctx, EXCP_RI);
-        return;
+        goto out;
     }
     MIPS_DEBUG("%s %s %s", opn, regnames[rs], regnames[rt]);
+ out:
+    tcg_temp_free(t0);
+    tcg_temp_free(t1);
 }
 
 static void gen_mul_vr54xx (DisasContext *ctx, uint32_t opc,
                             int rd, int rs, int rt)
 {
     const char *opn = "mul vr54xx";
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+    TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
 
-    gen_load_gpr(cpu_T[0], rs);
-    gen_load_gpr(cpu_T[1], rt);
+    gen_load_gpr(t0, rs);
+    gen_load_gpr(t1, rt);
 
     switch (opc) {
     case OPC_VR54XX_MULS:
-        gen_op_muls();
+        tcg_gen_helper_1_2(do_muls, t0, t0, t1);
         opn = "muls";
        break;
     case OPC_VR54XX_MULSU:
-        gen_op_mulsu();
+        tcg_gen_helper_1_2(do_mulsu, t0, t0, t1);
         opn = "mulsu";
        break;
     case OPC_VR54XX_MACC:
-        gen_op_macc();
+        tcg_gen_helper_1_2(do_macc, t0, t0, t1);
         opn = "macc";
        break;
     case OPC_VR54XX_MACCU:
-        gen_op_maccu();
+        tcg_gen_helper_1_2(do_maccu, t0, t0, t1);
         opn = "maccu";
        break;
     case OPC_VR54XX_MSAC:
-        gen_op_msac();
+        tcg_gen_helper_1_2(do_msac, t0, t0, t1);
         opn = "msac";
        break;
     case OPC_VR54XX_MSACU:
-        gen_op_msacu();
+        tcg_gen_helper_1_2(do_msacu, t0, t0, t1);
         opn = "msacu";
        break;
     case OPC_VR54XX_MULHI:
-        gen_op_mulhi();
+        tcg_gen_helper_1_2(do_mulhi, t0, t0, t1);
         opn = "mulhi";
        break;
     case OPC_VR54XX_MULHIU:
-        gen_op_mulhiu();
+        tcg_gen_helper_1_2(do_mulhiu, t0, t0, t1);
         opn = "mulhiu";
        break;
     case OPC_VR54XX_MULSHI:
-        gen_op_mulshi();
+        tcg_gen_helper_1_2(do_mulshi, t0, t0, t1);
         opn = "mulshi";
        break;
     case OPC_VR54XX_MULSHIU:
-        gen_op_mulshiu();
+        tcg_gen_helper_1_2(do_mulshiu, t0, t0, t1);
         opn = "mulshiu";
        break;
     case OPC_VR54XX_MACCHI:
-        gen_op_macchi();
+        tcg_gen_helper_1_2(do_macchi, t0, t0, t1);
         opn = "macchi";
        break;
     case OPC_VR54XX_MACCHIU:
-        gen_op_macchiu();
+        tcg_gen_helper_1_2(do_macchiu, t0, t0, t1);
         opn = "macchiu";
        break;
     case OPC_VR54XX_MSACHI:
-        gen_op_msachi();
+        tcg_gen_helper_1_2(do_msachi, t0, t0, t1);
         opn = "msachi";
        break;
     case OPC_VR54XX_MSACHIU:
-        gen_op_msachiu();
+        tcg_gen_helper_1_2(do_msachiu, t0, t0, t1);
         opn = "msachiu";
        break;
     default:
         MIPS_INVAL("mul vr54xx");
         generate_exception(ctx, EXCP_RI);
-        return;
+        goto out;
     }
-    gen_store_gpr(cpu_T[0], rd);
+    gen_store_gpr(t0, rd);
     MIPS_DEBUG("%s %s, %s, %s", opn, regnames[rd], regnames[rs], regnames[rt]);
+
+ out:
+    tcg_temp_free(t0);
+    tcg_temp_free(t1);
 }
 
 static void gen_cl (DisasContext *ctx, uint32_t opc,
                     int rd, int rs)
 {
     const char *opn = "CLx";
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+
     if (rd == 0) {
         /* Treat as NOP. */
         MIPS_DEBUG("NOP");
-        return;
+        goto out;
     }
-    gen_load_gpr(cpu_T[0], rs);
+    gen_load_gpr(t0, rs);
     switch (opc) {
     case OPC_CLO:
-        tcg_gen_helper_0_0(do_clo);
+        tcg_gen_helper_1_1(do_clo, t0, t0);
         opn = "clo";
         break;
     case OPC_CLZ:
-        tcg_gen_helper_0_0(do_clz);
+        tcg_gen_helper_1_1(do_clz, t0, t0);
         opn = "clz";
         break;
 #if defined(TARGET_MIPS64)
     case OPC_DCLO:
-        tcg_gen_helper_0_0(do_dclo);
+        tcg_gen_helper_1_1(do_dclo, t0, t0);
         opn = "dclo";
         break;
     case OPC_DCLZ:
-        tcg_gen_helper_0_0(do_dclz);
+        tcg_gen_helper_1_1(do_dclz, t0, t0);
         opn = "dclz";
         break;
 #endif
     default:
         MIPS_INVAL(opn);
         generate_exception(ctx, EXCP_RI);
-        return;
+        goto out;
     }
-    gen_store_gpr(cpu_T[0], rd);
+    gen_store_gpr(t0, rd);
     MIPS_DEBUG("%s %s, %s", opn, regnames[rd], regnames[rs]);
+
+ out:
+    tcg_temp_free(t0);
 }
 
 /* Traps */
@@ -2156,6 +2375,8 @@ static void gen_trap (DisasContext *ctx, uint32_t opc,
                       int rs, int rt, int16_t imm)
 {
     int cond;
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+    TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
 
     cond = 0;
     /* Load needed operands */
@@ -2168,8 +2389,8 @@ static void gen_trap (DisasContext *ctx, uint32_t opc,
     case OPC_TNE:
         /* Compare two registers */
         if (rs != rt) {
-            gen_load_gpr(cpu_T[0], rs);
-            gen_load_gpr(cpu_T[1], rt);
+            gen_load_gpr(t0, rs);
+            gen_load_gpr(t1, rt);
             cond = 1;
         }
         break;
@@ -2181,8 +2402,8 @@ static void gen_trap (DisasContext *ctx, uint32_t opc,
     case OPC_TNEI:
         /* Compare register to immediate */
         if (rs != 0 || imm != 0) {
-            gen_load_gpr(cpu_T[0], rs);
-            tcg_gen_movi_tl(cpu_T[1], (int32_t)imm);
+            gen_load_gpr(t0, rs);
+            tcg_gen_movi_tl(t1, (int32_t)imm);
             cond = 1;
         }
         break;
@@ -2196,7 +2417,7 @@ static void gen_trap (DisasContext *ctx, uint32_t opc,
         case OPC_TGEU:  /* rs >= rs unsigned */
         case OPC_TGEIU: /* r0 >= 0  unsigned */
             /* Always trap */
-            tcg_gen_movi_tl(cpu_T[0], 1);
+            tcg_gen_movi_tl(t0, 1);
             break;
         case OPC_TLT:   /* rs < rs           */
         case OPC_TLTI:  /* r0 < 0            */
@@ -2205,50 +2426,59 @@ static void gen_trap (DisasContext *ctx, uint32_t opc,
         case OPC_TNE:   /* rs != rs          */
         case OPC_TNEI:  /* r0 != 0           */
             /* Never trap: treat as NOP. */
-            return;
+            goto out;
         default:
             MIPS_INVAL("trap");
             generate_exception(ctx, EXCP_RI);
-            return;
+            goto out;
         }
     } else {
         switch (opc) {
         case OPC_TEQ:
         case OPC_TEQI:
-            gen_op_eq();
+            gen_op_eq(t0, t1);
             break;
         case OPC_TGE:
         case OPC_TGEI:
-            gen_op_ge();
+            gen_op_ge(t0, t1);
             break;
         case OPC_TGEU:
         case OPC_TGEIU:
-            gen_op_geu();
+            gen_op_geu(t0, t1);
             break;
         case OPC_TLT:
         case OPC_TLTI:
-            gen_op_lt();
+            gen_op_lt(t0, t1);
             break;
         case OPC_TLTU:
         case OPC_TLTIU:
-            gen_op_ltu();
+            gen_op_ltu(t0, t1);
             break;
         case OPC_TNE:
         case OPC_TNEI:
-            gen_op_ne();
+            gen_op_ne(t0, t1);
             break;
         default:
             MIPS_INVAL("trap");
             generate_exception(ctx, EXCP_RI);
-            return;
+            goto out;
         }
     }
     save_cpu_state(ctx, 1);
-    gen_op_trap();
+    {
+        int l1 = gen_new_label();
+
+        tcg_gen_brcondi_tl(TCG_COND_EQ, t0, 0, l1);
+        tcg_gen_helper_0_i(do_raise_exception, EXCP_TRAP);
+        gen_set_label(l1);
+    }
     ctx->bstate = BS_STOP;
+ out:
+    tcg_temp_free(t0);
+    tcg_temp_free(t1);
 }
 
-static always_inline void gen_goto_tb(DisasContext *ctx, int n, target_ulong dest)
+static inline void gen_goto_tb(DisasContext *ctx, int n, target_ulong dest)
 {
     TranslationBlock *tb;
     tb = ctx->tb;
@@ -2266,9 +2496,11 @@ static always_inline void gen_goto_tb(DisasContext *ctx, int n, target_ulong des
 static void gen_compute_branch (DisasContext *ctx, uint32_t opc,
                                 int rs, int rt, int32_t offset)
 {
-    target_ulong btarget = -1;
+    target_ulong btgt = -1;
     int blink = 0;
     int bcond = 0;
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+    TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
 
     if (ctx->hflags & MIPS_HFLAG_BMASK) {
 #ifdef MIPS_DEBUG_DISAS
@@ -2279,7 +2511,7 @@ static void gen_compute_branch (DisasContext *ctx, uint32_t opc,
        }
 #endif
         generate_exception(ctx, EXCP_RI);
-        return;
+        goto out;
     }
 
     /* Load needed operands */
@@ -2290,11 +2522,11 @@ static void gen_compute_branch (DisasContext *ctx, uint32_t opc,
     case OPC_BNEL:
         /* Compare two registers */
         if (rs != rt) {
-            gen_load_gpr(cpu_T[0], rs);
-            gen_load_gpr(cpu_T[1], rt);
+            gen_load_gpr(t0, rs);
+            gen_load_gpr(t1, rt);
             bcond = 1;
         }
-        btarget = ctx->pc + 4 + offset;
+        btgt = ctx->pc + 4 + offset;
         break;
     case OPC_BGEZ:
     case OPC_BGEZAL:
@@ -2310,15 +2542,15 @@ static void gen_compute_branch (DisasContext *ctx, uint32_t opc,
     case OPC_BLTZL:
         /* Compare to zero */
         if (rs != 0) {
-            gen_load_gpr(cpu_T[0], rs);
+            gen_load_gpr(t0, rs);
             bcond = 1;
         }
-        btarget = ctx->pc + 4 + offset;
+        btgt = ctx->pc + 4 + offset;
         break;
     case OPC_J:
     case OPC_JAL:
         /* Jump to immediate */
-        btarget = ((ctx->pc + 4) & (int32_t)0xF0000000) | (uint32_t)offset;
+        btgt = ((ctx->pc + 4) & (int32_t)0xF0000000) | (uint32_t)offset;
         break;
     case OPC_JR:
     case OPC_JALR:
@@ -2328,14 +2560,14 @@ static void gen_compute_branch (DisasContext *ctx, uint32_t opc,
                others are reserved. */
             MIPS_INVAL("jump hint");
             generate_exception(ctx, EXCP_RI);
-            return;
+            goto out;
         }
-        gen_save_breg_target(rs);
+        gen_load_gpr(btarget, rs);
         break;
     default:
         MIPS_INVAL("branch/jump");
         generate_exception(ctx, EXCP_RI);
-        return;
+        goto out;
     }
     if (bcond == 0) {
         /* No condition to be computed */
@@ -2362,34 +2594,34 @@ static void gen_compute_branch (DisasContext *ctx, uint32_t opc,
         case OPC_BLTZ:    /* 0 < 0           */
             /* Treat as NOP. */
             MIPS_DEBUG("bnever (NOP)");
-            return;
+            goto out;
         case OPC_BLTZAL:  /* 0 < 0           */
-            tcg_gen_movi_tl(cpu_T[0], ctx->pc + 8);
-            gen_store_gpr(cpu_T[0], 31);
+            tcg_gen_movi_tl(t0, ctx->pc + 8);
+            gen_store_gpr(t0, 31);
             MIPS_DEBUG("bnever and link");
-            return;
+            goto out;
         case OPC_BLTZALL: /* 0 < 0 likely */
-            tcg_gen_movi_tl(cpu_T[0], ctx->pc + 8);
-            gen_store_gpr(cpu_T[0], 31);
+            tcg_gen_movi_tl(t0, ctx->pc + 8);
+            gen_store_gpr(t0, 31);
             /* Skip the instruction in the delay slot */
             MIPS_DEBUG("bnever, link and skip");
             ctx->pc += 4;
-            return;
+            goto out;
         case OPC_BNEL:    /* rx != rx likely */
         case OPC_BGTZL:   /* 0 > 0 likely */
         case OPC_BLTZL:   /* 0 < 0 likely */
             /* Skip the instruction in the delay slot */
             MIPS_DEBUG("bnever and skip");
             ctx->pc += 4;
-            return;
+            goto out;
         case OPC_J:
             ctx->hflags |= MIPS_HFLAG_B;
-            MIPS_DEBUG("j " TARGET_FMT_lx, btarget);
+            MIPS_DEBUG("j " TARGET_FMT_lx, btgt);
             break;
         case OPC_JAL:
             blink = 31;
             ctx->hflags |= MIPS_HFLAG_B;
-            MIPS_DEBUG("jal " TARGET_FMT_lx, btarget);
+            MIPS_DEBUG("jal " TARGET_FMT_lx, btgt);
             break;
         case OPC_JR:
             ctx->hflags |= MIPS_HFLAG_BR;
@@ -2403,173 +2635,213 @@ static void gen_compute_branch (DisasContext *ctx, uint32_t opc,
         default:
             MIPS_INVAL("branch/jump");
             generate_exception(ctx, EXCP_RI);
-            return;
+            goto out;
         }
     } else {
         switch (opc) {
         case OPC_BEQ:
-            gen_op_eq();
+            gen_op_eq(t0, t1);
             MIPS_DEBUG("beq %s, %s, " TARGET_FMT_lx,
-                       regnames[rs], regnames[rt], btarget);
+                       regnames[rs], regnames[rt], btgt);
             goto not_likely;
         case OPC_BEQL:
-            gen_op_eq();
+            gen_op_eq(t0, t1);
             MIPS_DEBUG("beql %s, %s, " TARGET_FMT_lx,
-                       regnames[rs], regnames[rt], btarget);
+                       regnames[rs], regnames[rt], btgt);
             goto likely;
         case OPC_BNE:
-            gen_op_ne();
+            gen_op_ne(t0, t1);
             MIPS_DEBUG("bne %s, %s, " TARGET_FMT_lx,
-                       regnames[rs], regnames[rt], btarget);
+                       regnames[rs], regnames[rt], btgt);
             goto not_likely;
         case OPC_BNEL:
-            gen_op_ne();
+            gen_op_ne(t0, t1);
             MIPS_DEBUG("bnel %s, %s, " TARGET_FMT_lx,
-                       regnames[rs], regnames[rt], btarget);
+                       regnames[rs], regnames[rt], btgt);
             goto likely;
         case OPC_BGEZ:
-            gen_op_gez();
-            MIPS_DEBUG("bgez %s, " TARGET_FMT_lx, regnames[rs], btarget);
+            gen_op_gez(t0);
+            MIPS_DEBUG("bgez %s, " TARGET_FMT_lx, regnames[rs], btgt);
             goto not_likely;
         case OPC_BGEZL:
-            gen_op_gez();
-            MIPS_DEBUG("bgezl %s, " TARGET_FMT_lx, regnames[rs], btarget);
+            gen_op_gez(t0);
+            MIPS_DEBUG("bgezl %s, " TARGET_FMT_lx, regnames[rs], btgt);
             goto likely;
         case OPC_BGEZAL:
-            gen_op_gez();
-            MIPS_DEBUG("bgezal %s, " TARGET_FMT_lx, regnames[rs], btarget);
+            gen_op_gez(t0);
+            MIPS_DEBUG("bgezal %s, " TARGET_FMT_lx, regnames[rs], btgt);
             blink = 31;
             goto not_likely;
         case OPC_BGEZALL:
-            gen_op_gez();
+            gen_op_gez(t0);
             blink = 31;
-            MIPS_DEBUG("bgezall %s, " TARGET_FMT_lx, regnames[rs], btarget);
+            MIPS_DEBUG("bgezall %s, " TARGET_FMT_lx, regnames[rs], btgt);
             goto likely;
         case OPC_BGTZ:
-            gen_op_gtz();
-            MIPS_DEBUG("bgtz %s, " TARGET_FMT_lx, regnames[rs], btarget);
+            gen_op_gtz(t0);
+            MIPS_DEBUG("bgtz %s, " TARGET_FMT_lx, regnames[rs], btgt);
             goto not_likely;
         case OPC_BGTZL:
-            gen_op_gtz();
-            MIPS_DEBUG("bgtzl %s, " TARGET_FMT_lx, regnames[rs], btarget);
+            gen_op_gtz(t0);
+            MIPS_DEBUG("bgtzl %s, " TARGET_FMT_lx, regnames[rs], btgt);
             goto likely;
         case OPC_BLEZ:
-            gen_op_lez();
-            MIPS_DEBUG("blez %s, " TARGET_FMT_lx, regnames[rs], btarget);
+            gen_op_lez(t0);
+            MIPS_DEBUG("blez %s, " TARGET_FMT_lx, regnames[rs], btgt);
             goto not_likely;
         case OPC_BLEZL:
-            gen_op_lez();
-            MIPS_DEBUG("blezl %s, " TARGET_FMT_lx, regnames[rs], btarget);
+            gen_op_lez(t0);
+            MIPS_DEBUG("blezl %s, " TARGET_FMT_lx, regnames[rs], btgt);
             goto likely;
         case OPC_BLTZ:
-            gen_op_ltz();
-            MIPS_DEBUG("bltz %s, " TARGET_FMT_lx, regnames[rs], btarget);
+            gen_op_ltz(t0);
+            MIPS_DEBUG("bltz %s, " TARGET_FMT_lx, regnames[rs], btgt);
             goto not_likely;
         case OPC_BLTZL:
-            gen_op_ltz();
-            MIPS_DEBUG("bltzl %s, " TARGET_FMT_lx, regnames[rs], btarget);
+            gen_op_ltz(t0);
+            MIPS_DEBUG("bltzl %s, " TARGET_FMT_lx, regnames[rs], btgt);
             goto likely;
         case OPC_BLTZAL:
-            gen_op_ltz();
+            gen_op_ltz(t0);
             blink = 31;
-            MIPS_DEBUG("bltzal %s, " TARGET_FMT_lx, regnames[rs], btarget);
+            MIPS_DEBUG("bltzal %s, " TARGET_FMT_lx, regnames[rs], btgt);
         not_likely:
             ctx->hflags |= MIPS_HFLAG_BC;
-            tcg_gen_st_tl(cpu_T[0], cpu_env, offsetof(CPUState, bcond));
+            tcg_gen_trunc_tl_i32(bcond, t0);
             break;
         case OPC_BLTZALL:
-            gen_op_ltz();
+            gen_op_ltz(t0);
             blink = 31;
-            MIPS_DEBUG("bltzall %s, " TARGET_FMT_lx, regnames[rs], btarget);
+            MIPS_DEBUG("bltzall %s, " TARGET_FMT_lx, regnames[rs], btgt);
         likely:
             ctx->hflags |= MIPS_HFLAG_BL;
-            tcg_gen_st_tl(cpu_T[0], cpu_env, offsetof(CPUState, bcond));
+            tcg_gen_trunc_tl_i32(bcond, t0);
             break;
         default:
             MIPS_INVAL("conditional branch/jump");
             generate_exception(ctx, EXCP_RI);
-            return;
+            goto out;
         }
     }
     MIPS_DEBUG("enter ds: link %d cond %02x target " TARGET_FMT_lx,
-               blink, ctx->hflags, btarget);
+               blink, ctx->hflags, btgt);
 
-    ctx->btarget = btarget;
+    ctx->btarget = btgt;
     if (blink > 0) {
-        tcg_gen_movi_tl(cpu_T[0], ctx->pc + 8);
-        gen_store_gpr(cpu_T[0], blink);
+        tcg_gen_movi_tl(t0, ctx->pc + 8);
+        gen_store_gpr(t0, blink);
     }
+
+ out:
+    tcg_temp_free(t0);
+    tcg_temp_free(t1);
 }
 
 /* special3 bitfield operations */
 static void gen_bitops (DisasContext *ctx, uint32_t opc, int rt,
-                       int rs, int lsb, int msb)
+                        int rs, int lsb, int msb)
 {
-    gen_load_gpr(cpu_T[1], rs);
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+    TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
+
+    gen_load_gpr(t1, rs);
     switch (opc) {
     case OPC_EXT:
         if (lsb + msb > 31)
             goto fail;
-        gen_op_ext(lsb, msb + 1);
+        tcg_gen_helper_1_1ii(do_ext, t0, t1, lsb, msb + 1);
         break;
 #if defined(TARGET_MIPS64)
     case OPC_DEXTM:
         if (lsb + msb > 63)
             goto fail;
-        gen_op_dext(lsb, msb + 1 + 32);
+        tcg_gen_helper_1_1ii(do_dext, t0, t1, lsb, msb + 1 + 32);
         break;
     case OPC_DEXTU:
         if (lsb + msb > 63)
             goto fail;
-        gen_op_dext(lsb + 32, msb + 1);
+        tcg_gen_helper_1_1ii(do_dext, t0, t1, lsb + 32, msb + 1);
         break;
     case OPC_DEXT:
         if (lsb + msb > 63)
             goto fail;
-        gen_op_dext(lsb, msb + 1);
+        tcg_gen_helper_1_1ii(do_dext, t0, t1, lsb, msb + 1);
         break;
 #endif
     case OPC_INS:
         if (lsb > msb)
             goto fail;
-        gen_load_gpr(cpu_T[0], rt);
-        gen_op_ins(lsb, msb - lsb + 1);
+        gen_load_gpr(t0, rt);
+        tcg_gen_helper_1_2ii(do_ins, t0, t0, t1, lsb, msb - lsb + 1);
         break;
 #if defined(TARGET_MIPS64)
     case OPC_DINSM:
         if (lsb > msb)
             goto fail;
-        gen_load_gpr(cpu_T[0], rt);
-        gen_op_dins(lsb, msb - lsb + 1 + 32);
+        gen_load_gpr(t0, rt);
+        tcg_gen_helper_1_2ii(do_dins, t0, t0, t1, lsb, msb - lsb + 1 + 32);
         break;
     case OPC_DINSU:
         if (lsb > msb)
             goto fail;
-        gen_load_gpr(cpu_T[0], rt);
-        gen_op_dins(lsb + 32, msb - lsb + 1);
+        gen_load_gpr(t0, rt);
+        tcg_gen_helper_1_2ii(do_dins, t0, t0, t1, lsb + 32, msb - lsb + 1);
         break;
     case OPC_DINS:
         if (lsb > msb)
             goto fail;
-        gen_load_gpr(cpu_T[0], rt);
-        gen_op_dins(lsb, msb - lsb + 1);
+        gen_load_gpr(t0, rt);
+        tcg_gen_helper_1_2ii(do_dins, t0, t0, t1, lsb, msb - lsb + 1);
         break;
 #endif
     default:
 fail:
         MIPS_INVAL("bitops");
         generate_exception(ctx, EXCP_RI);
+        tcg_temp_free(t0);
+        tcg_temp_free(t1);
         return;
     }
-    gen_store_gpr(cpu_T[0], rt);
+    gen_store_gpr(t0, rt);
+    tcg_temp_free(t0);
+    tcg_temp_free(t1);
 }
 
+#ifndef CONFIG_USER_ONLY
 /* CP0 (MMU and control) */
-static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
+static inline void gen_mfc0_load32 (TCGv t, target_ulong off)
 {
-    const char *rn = "invalid";
     TCGv r_tmp = tcg_temp_new(TCG_TYPE_I32);
-    TCGv r_tmp64 = tcg_temp_new(TCG_TYPE_I64);
+
+    tcg_gen_ld_i32(r_tmp, cpu_env, off);
+    tcg_gen_ext_i32_tl(t, r_tmp);
+    tcg_temp_free(r_tmp);
+}
+
+static inline void gen_mfc0_load64 (TCGv t, target_ulong off)
+{
+    tcg_gen_ld_tl(t, cpu_env, off);
+    tcg_gen_ext32s_tl(t, t);
+}
+
+static inline void gen_mtc0_store32 (TCGv t, target_ulong off)
+{
+    TCGv r_tmp = tcg_temp_new(TCG_TYPE_I32);
+
+    tcg_gen_trunc_tl_i32(r_tmp, t);
+    tcg_gen_st_i32(r_tmp, cpu_env, off);
+    tcg_temp_free(r_tmp);
+}
+
+static inline void gen_mtc0_store64 (TCGv t, target_ulong off)
+{
+    tcg_gen_ext32s_tl(t, t);
+    tcg_gen_st_tl(t, cpu_env, off);
+}
+
+static void gen_mfc0 (CPUState *env, DisasContext *ctx, TCGv t0, int reg, int sel)
+{
+    const char *rn = "invalid";
 
     if (sel != 0)
         check_insn(env, ctx, ISA_MIPS32);
@@ -2578,23 +2850,22 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 0:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Index));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Index));
             rn = "Index";
             break;
         case 1:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_mvpcontrol();
+            tcg_gen_helper_1_0(do_mfc0_mvpcontrol, t0);
             rn = "MVPControl";
             break;
         case 2:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_mvpconf0();
+            tcg_gen_helper_1_0(do_mfc0_mvpconf0, t0);
             rn = "MVPConf0";
             break;
         case 3:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_mvpconf1();
+            tcg_gen_helper_1_0(do_mfc0_mvpconf1, t0);
             rn = "MVPConf1";
             break;
         default:
@@ -2604,49 +2875,42 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 1:
         switch (sel) {
         case 0:
-            gen_op_mfc0_random();
+            tcg_gen_helper_1_0(do_mfc0_random, t0);
             rn = "Random";
             break;
         case 1:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_VPEControl));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_VPEControl));
             rn = "VPEControl";
             break;
         case 2:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_VPEConf0));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_VPEConf0));
             rn = "VPEConf0";
             break;
         case 3:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_VPEConf1));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_VPEConf1));
             rn = "VPEConf1";
             break;
         case 4:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_i64(r_tmp64, cpu_env, offsetof(CPUState, CP0_YQMask));
-            tcg_gen_trunc_i64_tl(cpu_T[0], r_tmp64);
+            gen_mfc0_load64(t0, offsetof(CPUState, CP0_YQMask));
             rn = "YQMask";
             break;
         case 5:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_tl(r_tmp64, cpu_env, offsetof(CPUState, CP0_VPESchedule));
-            tcg_gen_trunc_i64_tl(cpu_T[0], r_tmp64);
+            gen_mfc0_load64(t0, offsetof(CPUState, CP0_VPESchedule));
             rn = "VPESchedule";
             break;
         case 6:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_tl(r_tmp64, cpu_env, offsetof(CPUState, CP0_VPEScheFBack));
-            tcg_gen_trunc_i64_tl(cpu_T[0], r_tmp64);
+            gen_mfc0_load64(t0, offsetof(CPUState, CP0_VPEScheFBack));
             rn = "VPEScheFBack";
             break;
         case 7:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_VPEOpt));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_VPEOpt));
             rn = "VPEOpt";
             break;
         default:
@@ -2656,43 +2920,43 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 2:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_EntryLo0));
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_EntryLo0));
+            tcg_gen_ext32s_tl(t0, t0);
             rn = "EntryLo0";
             break;
         case 1:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_tcstatus();
+            tcg_gen_helper_1_0(do_mfc0_tcstatus, t0);
             rn = "TCStatus";
             break;
         case 2:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_tcbind();
+            tcg_gen_helper_1_0(do_mfc0_tcbind, t0);
             rn = "TCBind";
             break;
         case 3:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_tcrestart();
+            tcg_gen_helper_1_0(do_mfc0_tcrestart, t0);
             rn = "TCRestart";
             break;
         case 4:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_tchalt();
+            tcg_gen_helper_1_0(do_mfc0_tchalt, t0);
             rn = "TCHalt";
             break;
         case 5:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_tccontext();
+            tcg_gen_helper_1_0(do_mfc0_tccontext, t0);
             rn = "TCContext";
             break;
         case 6:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_tcschedule();
+            tcg_gen_helper_1_0(do_mfc0_tcschedule, t0);
             rn = "TCSchedule";
             break;
         case 7:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_tcschefback();
+            tcg_gen_helper_1_0(do_mfc0_tcschefback, t0);
             rn = "TCScheFBack";
             break;
         default:
@@ -2702,8 +2966,8 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 3:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_EntryLo1));
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_EntryLo1));
+            tcg_gen_ext32s_tl(t0, t0);
             rn = "EntryLo1";
             break;
         default:
@@ -2713,12 +2977,12 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 4:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_Context));
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_Context));
+            tcg_gen_ext32s_tl(t0, t0);
             rn = "Context";
             break;
         case 1:
-//            gen_op_mfc0_contextconfig(); /* SmartMIPS ASE */
+//            tcg_gen_helper_1_0(do_mfc0_contextconfig, t0); /* SmartMIPS ASE */
             rn = "ContextConfig";
 //            break;
         default:
@@ -2728,14 +2992,12 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 5:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_PageMask));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_PageMask));
             rn = "PageMask";
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_PageGrain));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_PageGrain));
             rn = "PageGrain";
             break;
         default:
@@ -2745,38 +3007,32 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 6:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Wired));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Wired));
             rn = "Wired";
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSConf0));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSConf0));
             rn = "SRSConf0";
             break;
         case 2:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSConf1));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSConf1));
             rn = "SRSConf1";
             break;
         case 3:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSConf2));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSConf2));
             rn = "SRSConf2";
             break;
         case 4:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSConf3));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSConf3));
             rn = "SRSConf3";
             break;
         case 5:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSConf4));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSConf4));
             rn = "SRSConf4";
             break;
         default:
@@ -2787,8 +3043,7 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         switch (sel) {
         case 0:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_HWREna));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_HWREna));
             rn = "HWREna";
             break;
         default:
@@ -2798,8 +3053,8 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 8:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_BadVAddr));
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_BadVAddr));
+            tcg_gen_ext32s_tl(t0, t0);
             rn = "BadVAddr";
             break;
         default:
@@ -2809,7 +3064,14 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 9:
         switch (sel) {
         case 0:
-            gen_op_mfc0_count();
+            /* Mark as an IO operation because we read the time.  */
+            if (use_icount)
+                gen_io_start();
+            tcg_gen_helper_1_0(do_mfc0_count, t0);
+            if (use_icount) {
+                gen_io_end();
+                ctx->bstate = BS_STOP;
+            }
             rn = "Count";
             break;
         /* 6,7 are implementation dependent */
@@ -2820,8 +3082,8 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 10:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_EntryHi));
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_EntryHi));
+            tcg_gen_ext32s_tl(t0, t0);
             rn = "EntryHi";
             break;
         default:
@@ -2831,8 +3093,7 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 11:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Compare));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Compare));
             rn = "Compare";
             break;
         /* 6,7 are implementation dependent */
@@ -2843,26 +3104,22 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 12:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Status));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Status));
             rn = "Status";
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_IntCtl));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_IntCtl));
             rn = "IntCtl";
             break;
         case 2:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSCtl));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSCtl));
             rn = "SRSCtl";
             break;
         case 3:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSMap));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSMap));
             rn = "SRSMap";
             break;
         default:
@@ -2872,8 +3129,7 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 13:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Cause));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Cause));
             rn = "Cause";
             break;
         default:
@@ -2883,8 +3139,8 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 14:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_EPC));
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_EPC));
+            tcg_gen_ext32s_tl(t0, t0);
             rn = "EPC";
             break;
         default:
@@ -2894,14 +3150,12 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 15:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_PRid));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_PRid));
             rn = "PRid";
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_EBase));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_EBase));
             rn = "EBase";
             break;
         default:
@@ -2911,35 +3165,29 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 16:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Config0));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Config0));
             rn = "Config";
             break;
         case 1:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Config1));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Config1));
             rn = "Config1";
             break;
         case 2:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Config2));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Config2));
             rn = "Config2";
             break;
         case 3:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Config3));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Config3));
             rn = "Config3";
             break;
         /* 4,5 are reserved */
         /* 6,7 are implementation dependent */
         case 6:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Config6));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Config6));
             rn = "Config6";
             break;
         case 7:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Config7));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Config7));
             rn = "Config7";
             break;
         default:
@@ -2949,7 +3197,7 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 17:
         switch (sel) {
         case 0:
-            gen_op_mfc0_lladdr();
+            tcg_gen_helper_1_0(do_mfc0_lladdr, t0);
             rn = "LLAddr";
             break;
         default:
@@ -2959,7 +3207,7 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 18:
         switch (sel) {
         case 0 ... 7:
-            gen_op_mfc0_watchlo(sel);
+            tcg_gen_helper_1_i(do_mfc0_watchlo, t0, sel);
             rn = "WatchLo";
             break;
         default:
@@ -2969,7 +3217,7 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 19:
         switch (sel) {
         case 0 ...7:
-            gen_op_mfc0_watchhi(sel);
+            tcg_gen_helper_1_i(do_mfc0_watchhi, t0, sel);
             rn = "WatchHi";
             break;
         default:
@@ -2981,8 +3229,8 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         case 0:
 #if defined(TARGET_MIPS64)
             check_insn(env, ctx, ISA_MIPS3);
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_XContext));
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_XContext));
+            tcg_gen_ext32s_tl(t0, t0);
             rn = "XContext";
             break;
 #endif
@@ -2994,8 +3242,7 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
        /* Officially reserved, but sel 0 is used for R1x000 framemask */
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Framemask));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Framemask));
             rn = "Framemask";
             break;
         default:
@@ -3009,23 +3256,23 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 23:
         switch (sel) {
         case 0:
-            gen_op_mfc0_debug(); /* EJTAG support */
+            tcg_gen_helper_1_0(do_mfc0_debug, t0); /* EJTAG support */
             rn = "Debug";
             break;
         case 1:
-//            gen_op_mfc0_tracecontrol(); /* PDtrace support */
+//            tcg_gen_helper_1_0(do_mfc0_tracecontrol, t0); /* PDtrace support */
             rn = "TraceControl";
 //            break;
         case 2:
-//            gen_op_mfc0_tracecontrol2(); /* PDtrace support */
+//            tcg_gen_helper_1_0(do_mfc0_tracecontrol2, t0); /* PDtrace support */
             rn = "TraceControl2";
 //            break;
         case 3:
-//            gen_op_mfc0_usertracedata(); /* PDtrace support */
+//            tcg_gen_helper_1_0(do_mfc0_usertracedata, t0); /* PDtrace support */
             rn = "UserTraceData";
 //            break;
         case 4:
-//            gen_op_mfc0_debug(); /* PDtrace support */
+//            tcg_gen_helper_1_0(do_mfc0_tracebpc, t0); /* PDtrace support */
             rn = "TraceBPC";
 //            break;
         default:
@@ -3036,8 +3283,8 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         switch (sel) {
         case 0:
             /* EJTAG support */
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_DEPC));
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_DEPC));
+            tcg_gen_ext32s_tl(t0, t0);
             rn = "DEPC";
             break;
         default:
@@ -3047,36 +3294,35 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 25:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Performance0));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Performance0));
             rn = "Performance0";
             break;
         case 1:
-//            gen_op_mfc0_performance1();
+//            tcg_gen_helper_1_0(do_mfc0_performance1, t0);
             rn = "Performance1";
 //            break;
         case 2:
-//            gen_op_mfc0_performance2();
+//            tcg_gen_helper_1_0(do_mfc0_performance2, t0);
             rn = "Performance2";
 //            break;
         case 3:
-//            gen_op_mfc0_performance3();
+//            tcg_gen_helper_1_0(do_mfc0_performance3, t0);
             rn = "Performance3";
 //            break;
         case 4:
-//            gen_op_mfc0_performance4();
+//            tcg_gen_helper_1_0(do_mfc0_performance4, t0);
             rn = "Performance4";
 //            break;
         case 5:
-//            gen_op_mfc0_performance5();
+//            tcg_gen_helper_1_0(do_mfc0_performance5, t0);
             rn = "Performance5";
 //            break;
         case 6:
-//            gen_op_mfc0_performance6();
+//            tcg_gen_helper_1_0(do_mfc0_performance6, t0);
             rn = "Performance6";
 //            break;
         case 7:
-//            gen_op_mfc0_performance7();
+//            tcg_gen_helper_1_0(do_mfc0_performance7, t0);
             rn = "Performance7";
 //            break;
         default:
@@ -3102,16 +3348,14 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         case 2:
         case 4:
         case 6:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_TagLo));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_TagLo));
             rn = "TagLo";
             break;
         case 1:
         case 3:
         case 5:
         case 7:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_DataLo));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_DataLo));
             rn = "DataLo";
             break;
         default:
@@ -3124,16 +3368,14 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         case 2:
         case 4:
         case 6:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_TagHi));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_TagHi));
             rn = "TagHi";
             break;
         case 1:
         case 3:
         case 5:
         case 7:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_DataHi));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_DataHi));
             rn = "DataHi";
             break;
         default:
@@ -3143,8 +3385,8 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 30:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_ErrorEPC));
-            tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_ErrorEPC));
+            tcg_gen_ext32s_tl(t0, t0);
             rn = "ErrorEPC";
             break;
         default:
@@ -3155,8 +3397,7 @@ static void gen_mfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         switch (sel) {
         case 0:
             /* EJTAG support */
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_DESAVE));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_DESAVE));
             rn = "DESAVE";
             break;
         default:
@@ -3184,23 +3425,26 @@ die:
     generate_exception(ctx, EXCP_RI);
 }
 
-static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
+static void gen_mtc0 (CPUState *env, DisasContext *ctx, TCGv t0, int reg, int sel)
 {
     const char *rn = "invalid";
 
     if (sel != 0)
         check_insn(env, ctx, ISA_MIPS32);
 
+    if (use_icount)
+        gen_io_start();
+
     switch (reg) {
     case 0:
         switch (sel) {
         case 0:
-            gen_op_mtc0_index();
+            tcg_gen_helper_0_1(do_mtc0_index, t0);
             rn = "Index";
             break;
         case 1:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_mvpcontrol();
+            tcg_gen_helper_0_1(do_mtc0_mvpcontrol, t0);
             rn = "MVPControl";
             break;
         case 2:
@@ -3225,37 +3469,37 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
             break;
         case 1:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_vpecontrol();
+            tcg_gen_helper_0_1(do_mtc0_vpecontrol, t0);
             rn = "VPEControl";
             break;
         case 2:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_vpeconf0();
+            tcg_gen_helper_0_1(do_mtc0_vpeconf0, t0);
             rn = "VPEConf0";
             break;
         case 3:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_vpeconf1();
+            tcg_gen_helper_0_1(do_mtc0_vpeconf1, t0);
             rn = "VPEConf1";
             break;
         case 4:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_yqmask();
+            tcg_gen_helper_0_1(do_mtc0_yqmask, t0);
             rn = "YQMask";
             break;
         case 5:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_vpeschedule();
+            gen_mtc0_store64(t0, offsetof(CPUState, CP0_VPESchedule));
             rn = "VPESchedule";
             break;
         case 6:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_vpeschefback();
+            gen_mtc0_store64(t0, offsetof(CPUState, CP0_VPEScheFBack));
             rn = "VPEScheFBack";
             break;
         case 7:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_vpeopt();
+            tcg_gen_helper_0_1(do_mtc0_vpeopt, t0);
             rn = "VPEOpt";
             break;
         default:
@@ -3265,42 +3509,42 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 2:
         switch (sel) {
         case 0:
-            gen_op_mtc0_entrylo0();
+            tcg_gen_helper_0_1(do_mtc0_entrylo0, t0);
             rn = "EntryLo0";
             break;
         case 1:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tcstatus();
+            tcg_gen_helper_0_1(do_mtc0_tcstatus, t0);
             rn = "TCStatus";
             break;
         case 2:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tcbind();
+            tcg_gen_helper_0_1(do_mtc0_tcbind, t0);
             rn = "TCBind";
             break;
         case 3:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tcrestart();
+            tcg_gen_helper_0_1(do_mtc0_tcrestart, t0);
             rn = "TCRestart";
             break;
         case 4:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tchalt();
+            tcg_gen_helper_0_1(do_mtc0_tchalt, t0);
             rn = "TCHalt";
             break;
         case 5:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tccontext();
+            tcg_gen_helper_0_1(do_mtc0_tccontext, t0);
             rn = "TCContext";
             break;
         case 6:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tcschedule();
+            tcg_gen_helper_0_1(do_mtc0_tcschedule, t0);
             rn = "TCSchedule";
             break;
         case 7:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tcschefback();
+            tcg_gen_helper_0_1(do_mtc0_tcschefback, t0);
             rn = "TCScheFBack";
             break;
         default:
@@ -3310,7 +3554,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 3:
         switch (sel) {
         case 0:
-            gen_op_mtc0_entrylo1();
+            tcg_gen_helper_0_1(do_mtc0_entrylo1, t0);
             rn = "EntryLo1";
             break;
         default:
@@ -3320,11 +3564,11 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 4:
         switch (sel) {
         case 0:
-            gen_op_mtc0_context();
+            tcg_gen_helper_0_1(do_mtc0_context, t0);
             rn = "Context";
             break;
         case 1:
-//            gen_op_mtc0_contextconfig(); /* SmartMIPS ASE */
+//            tcg_gen_helper_0_1(do_mtc0_contextconfig, t0); /* SmartMIPS ASE */
             rn = "ContextConfig";
 //            break;
         default:
@@ -3334,12 +3578,12 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 5:
         switch (sel) {
         case 0:
-            gen_op_mtc0_pagemask();
+            tcg_gen_helper_0_1(do_mtc0_pagemask, t0);
             rn = "PageMask";
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_pagegrain();
+            tcg_gen_helper_0_1(do_mtc0_pagegrain, t0);
             rn = "PageGrain";
             break;
         default:
@@ -3349,32 +3593,32 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 6:
         switch (sel) {
         case 0:
-            gen_op_mtc0_wired();
+            tcg_gen_helper_0_1(do_mtc0_wired, t0);
             rn = "Wired";
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsconf0();
+            tcg_gen_helper_0_1(do_mtc0_srsconf0, t0);
             rn = "SRSConf0";
             break;
         case 2:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsconf1();
+            tcg_gen_helper_0_1(do_mtc0_srsconf1, t0);
             rn = "SRSConf1";
             break;
         case 3:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsconf2();
+            tcg_gen_helper_0_1(do_mtc0_srsconf2, t0);
             rn = "SRSConf2";
             break;
         case 4:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsconf3();
+            tcg_gen_helper_0_1(do_mtc0_srsconf3, t0);
             rn = "SRSConf3";
             break;
         case 5:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsconf4();
+            tcg_gen_helper_0_1(do_mtc0_srsconf4, t0);
             rn = "SRSConf4";
             break;
         default:
@@ -3385,7 +3629,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         switch (sel) {
         case 0:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_hwrena();
+            tcg_gen_helper_0_1(do_mtc0_hwrena, t0);
             rn = "HWREna";
             break;
         default:
@@ -3399,7 +3643,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 9:
         switch (sel) {
         case 0:
-            gen_op_mtc0_count();
+            tcg_gen_helper_0_1(do_mtc0_count, t0);
             rn = "Count";
             break;
         /* 6,7 are implementation dependent */
@@ -3412,7 +3656,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 10:
         switch (sel) {
         case 0:
-            gen_op_mtc0_entryhi();
+            tcg_gen_helper_0_1(do_mtc0_entryhi, t0);
             rn = "EntryHi";
             break;
         default:
@@ -3422,7 +3666,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 11:
         switch (sel) {
         case 0:
-            gen_op_mtc0_compare();
+            tcg_gen_helper_0_1(do_mtc0_compare, t0);
             rn = "Compare";
             break;
         /* 6,7 are implementation dependent */
@@ -3435,7 +3679,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 12:
         switch (sel) {
         case 0:
-            gen_op_mtc0_status();
+            tcg_gen_helper_0_1(do_mtc0_status, t0);
             /* BS_STOP isn't good enough here, hflags may have changed. */
             gen_save_pc(ctx->pc + 4);
             ctx->bstate = BS_EXCP;
@@ -3443,21 +3687,21 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_intctl();
+            tcg_gen_helper_0_1(do_mtc0_intctl, t0);
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
             rn = "IntCtl";
             break;
         case 2:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsctl();
+            tcg_gen_helper_0_1(do_mtc0_srsctl, t0);
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
             rn = "SRSCtl";
             break;
         case 3:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsmap();
+            gen_mtc0_store32(t0, offsetof(CPUState, CP0_SRSMap));
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
             rn = "SRSMap";
@@ -3469,7 +3713,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 13:
         switch (sel) {
         case 0:
-            gen_op_mtc0_cause();
+            tcg_gen_helper_0_1(do_mtc0_cause, t0);
             rn = "Cause";
             break;
         default:
@@ -3481,7 +3725,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 14:
         switch (sel) {
         case 0:
-            gen_op_mtc0_epc();
+            gen_mtc0_store64(t0, offsetof(CPUState, CP0_EPC));
             rn = "EPC";
             break;
         default:
@@ -3496,7 +3740,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_ebase();
+            tcg_gen_helper_0_1(do_mtc0_ebase, t0);
             rn = "EBase";
             break;
         default:
@@ -3506,7 +3750,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 16:
         switch (sel) {
         case 0:
-            gen_op_mtc0_config0();
+            tcg_gen_helper_0_1(do_mtc0_config0, t0);
             rn = "Config";
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
@@ -3516,7 +3760,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
             rn = "Config1";
             break;
         case 2:
-            gen_op_mtc0_config2();
+            tcg_gen_helper_0_1(do_mtc0_config2, t0);
             rn = "Config2";
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
@@ -3553,7 +3797,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 18:
         switch (sel) {
         case 0 ... 7:
-            gen_op_mtc0_watchlo(sel);
+            tcg_gen_helper_0_1i(do_mtc0_watchlo, t0, sel);
             rn = "WatchLo";
             break;
         default:
@@ -3563,7 +3807,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 19:
         switch (sel) {
         case 0 ... 7:
-            gen_op_mtc0_watchhi(sel);
+            tcg_gen_helper_0_1i(do_mtc0_watchhi, t0, sel);
             rn = "WatchHi";
             break;
         default:
@@ -3575,7 +3819,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         case 0:
 #if defined(TARGET_MIPS64)
             check_insn(env, ctx, ISA_MIPS3);
-            gen_op_mtc0_xcontext();
+            tcg_gen_helper_0_1(do_mtc0_xcontext, t0);
             rn = "XContext";
             break;
 #endif
@@ -3587,7 +3831,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
        /* Officially reserved, but sel 0 is used for R1x000 framemask */
         switch (sel) {
         case 0:
-            gen_op_mtc0_framemask();
+            tcg_gen_helper_0_1(do_mtc0_framemask, t0);
             rn = "Framemask";
             break;
         default:
@@ -3601,20 +3845,20 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 23:
         switch (sel) {
         case 0:
-            gen_op_mtc0_debug(); /* EJTAG support */
+            tcg_gen_helper_0_1(do_mtc0_debug, t0); /* EJTAG support */
             /* BS_STOP isn't good enough here, hflags may have changed. */
             gen_save_pc(ctx->pc + 4);
             ctx->bstate = BS_EXCP;
             rn = "Debug";
             break;
         case 1:
-//            gen_op_mtc0_tracecontrol(); /* PDtrace support */
+//            tcg_gen_helper_0_1(do_mtc0_tracecontrol, t0); /* PDtrace support */
             rn = "TraceControl";
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
 //            break;
         case 2:
-//            gen_op_mtc0_tracecontrol2(); /* PDtrace support */
+//            tcg_gen_helper_0_1(do_mtc0_tracecontrol2, t0); /* PDtrace support */
             rn = "TraceControl2";
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
@@ -3622,13 +3866,13 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         case 3:
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
-//            gen_op_mtc0_usertracedata(); /* PDtrace support */
+//            tcg_gen_helper_0_1(do_mtc0_usertracedata, t0); /* PDtrace support */
             rn = "UserTraceData";
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
 //            break;
         case 4:
-//            gen_op_mtc0_debug(); /* PDtrace support */
+//            tcg_gen_helper_0_1(do_mtc0_tracebpc, t0); /* PDtrace support */
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
             rn = "TraceBPC";
@@ -3640,7 +3884,8 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 24:
         switch (sel) {
         case 0:
-            gen_op_mtc0_depc(); /* EJTAG support */
+            /* EJTAG support */
+            gen_mtc0_store64(t0, offsetof(CPUState, CP0_DEPC));
             rn = "DEPC";
             break;
         default:
@@ -3650,35 +3895,35 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 25:
         switch (sel) {
         case 0:
-            gen_op_mtc0_performance0();
+            tcg_gen_helper_0_1(do_mtc0_performance0, t0);
             rn = "Performance0";
             break;
         case 1:
-//            gen_op_mtc0_performance1();
+//            tcg_gen_helper_0_1(do_mtc0_performance1, t0);
             rn = "Performance1";
 //            break;
         case 2:
-//            gen_op_mtc0_performance2();
+//            tcg_gen_helper_0_1(do_mtc0_performance2, t0);
             rn = "Performance2";
 //            break;
         case 3:
-//            gen_op_mtc0_performance3();
+//            tcg_gen_helper_0_1(do_mtc0_performance3, t0);
             rn = "Performance3";
 //            break;
         case 4:
-//            gen_op_mtc0_performance4();
+//            tcg_gen_helper_0_1(do_mtc0_performance4, t0);
             rn = "Performance4";
 //            break;
         case 5:
-//            gen_op_mtc0_performance5();
+//            tcg_gen_helper_0_1(do_mtc0_performance5, t0);
             rn = "Performance5";
 //            break;
         case 6:
-//            gen_op_mtc0_performance6();
+//            tcg_gen_helper_0_1(do_mtc0_performance6, t0);
             rn = "Performance6";
 //            break;
         case 7:
-//            gen_op_mtc0_performance7();
+//            tcg_gen_helper_0_1(do_mtc0_performance7, t0);
             rn = "Performance7";
 //            break;
         default:
@@ -3705,14 +3950,14 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         case 2:
         case 4:
         case 6:
-            gen_op_mtc0_taglo();
+            tcg_gen_helper_0_1(do_mtc0_taglo, t0);
             rn = "TagLo";
             break;
         case 1:
         case 3:
         case 5:
         case 7:
-            gen_op_mtc0_datalo();
+            tcg_gen_helper_0_1(do_mtc0_datalo, t0);
             rn = "DataLo";
             break;
         default:
@@ -3725,14 +3970,14 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         case 2:
         case 4:
         case 6:
-            gen_op_mtc0_taghi();
+            tcg_gen_helper_0_1(do_mtc0_taghi, t0);
             rn = "TagHi";
             break;
         case 1:
         case 3:
         case 5:
         case 7:
-            gen_op_mtc0_datahi();
+            tcg_gen_helper_0_1(do_mtc0_datahi, t0);
             rn = "DataHi";
             break;
         default:
@@ -3743,7 +3988,7 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 30:
         switch (sel) {
         case 0:
-            gen_op_mtc0_errorepc();
+            gen_mtc0_store64(t0, offsetof(CPUState, CP0_ErrorEPC));
             rn = "ErrorEPC";
             break;
         default:
@@ -3753,7 +3998,8 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 31:
         switch (sel) {
         case 0:
-            gen_op_mtc0_desave(); /* EJTAG support */
+            /* EJTAG support */
+            gen_mtc0_store32(t0, offsetof(CPUState, CP0_DESAVE));
             rn = "DESAVE";
             break;
         default:
@@ -3771,6 +4017,11 @@ static void gen_mtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
                 rn, reg, sel);
     }
 #endif
+    /* For simplicity assume that all writes can cause interrupts.  */
+    if (use_icount) {
+        gen_io_end();
+        ctx->bstate = BS_STOP;
+    }
     return;
 
 die:
@@ -3784,10 +4035,9 @@ die:
 }
 
 #if defined(TARGET_MIPS64)
-static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
+static void gen_dmfc0 (CPUState *env, DisasContext *ctx, TCGv t0, int reg, int sel)
 {
     const char *rn = "invalid";
-    TCGv r_tmp = tcg_temp_new(TCG_TYPE_I32);
 
     if (sel != 0)
         check_insn(env, ctx, ISA_MIPS64);
@@ -3796,23 +4046,22 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 0:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Index));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Index));
             rn = "Index";
             break;
         case 1:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_mvpcontrol();
+            tcg_gen_helper_1_0(do_mfc0_mvpcontrol, t0);
             rn = "MVPControl";
             break;
         case 2:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_mvpconf0();
+            tcg_gen_helper_1_0(do_mfc0_mvpconf0, t0);
             rn = "MVPConf0";
             break;
         case 3:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_mvpconf1();
+            tcg_gen_helper_1_0(do_mfc0_mvpconf1, t0);
             rn = "MVPConf1";
             break;
         default:
@@ -3822,46 +4071,42 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 1:
         switch (sel) {
         case 0:
-            gen_op_mfc0_random();
+            tcg_gen_helper_1_0(do_mfc0_random, t0);
             rn = "Random";
             break;
         case 1:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_VPEControl));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_VPEControl));
             rn = "VPEControl";
             break;
         case 2:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_VPEConf0));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_VPEConf0));
             rn = "VPEConf0";
             break;
         case 3:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_VPEConf1));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_VPEConf1));
             rn = "VPEConf1";
             break;
         case 4:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_YQMask));
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_YQMask));
             rn = "YQMask";
             break;
         case 5:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_VPESchedule));
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_VPESchedule));
             rn = "VPESchedule";
             break;
         case 6:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_VPEScheFBack));
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_VPEScheFBack));
             rn = "VPEScheFBack";
             break;
         case 7:
             check_insn(env, ctx, ASE_MT);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_VPEOpt));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_VPEOpt));
             rn = "VPEOpt";
             break;
         default:
@@ -3871,42 +4116,42 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 2:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_EntryLo0));
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_EntryLo0));
             rn = "EntryLo0";
             break;
         case 1:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_tcstatus();
+            tcg_gen_helper_1_0(do_mfc0_tcstatus, t0);
             rn = "TCStatus";
             break;
         case 2:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mfc0_tcbind();
+            tcg_gen_helper_1_0(do_mfc0_tcbind, t0);
             rn = "TCBind";
             break;
         case 3:
             check_insn(env, ctx, ASE_MT);
-            gen_op_dmfc0_tcrestart();
+            tcg_gen_helper_1_0(do_dmfc0_tcrestart, t0);
             rn = "TCRestart";
             break;
         case 4:
             check_insn(env, ctx, ASE_MT);
-            gen_op_dmfc0_tchalt();
+            tcg_gen_helper_1_0(do_dmfc0_tchalt, t0);
             rn = "TCHalt";
             break;
         case 5:
             check_insn(env, ctx, ASE_MT);
-            gen_op_dmfc0_tccontext();
+            tcg_gen_helper_1_0(do_dmfc0_tccontext, t0);
             rn = "TCContext";
             break;
         case 6:
             check_insn(env, ctx, ASE_MT);
-            gen_op_dmfc0_tcschedule();
+            tcg_gen_helper_1_0(do_dmfc0_tcschedule, t0);
             rn = "TCSchedule";
             break;
         case 7:
             check_insn(env, ctx, ASE_MT);
-            gen_op_dmfc0_tcschefback();
+            tcg_gen_helper_1_0(do_dmfc0_tcschefback, t0);
             rn = "TCScheFBack";
             break;
         default:
@@ -3916,7 +4161,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 3:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_EntryLo1));
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_EntryLo1));
             rn = "EntryLo1";
             break;
         default:
@@ -3926,11 +4171,11 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 4:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_Context));
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_Context));
             rn = "Context";
             break;
         case 1:
-//            gen_op_dmfc0_contextconfig(); /* SmartMIPS ASE */
+//            tcg_gen_helper_1_0(do_dmfc0_contextconfig, t0); /* SmartMIPS ASE */
             rn = "ContextConfig";
 //            break;
         default:
@@ -3940,14 +4185,12 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 5:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_PageMask));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_PageMask));
             rn = "PageMask";
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_PageGrain));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_PageGrain));
             rn = "PageGrain";
             break;
         default:
@@ -3957,38 +4200,32 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 6:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Wired));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Wired));
             rn = "Wired";
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSConf0));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSConf0));
             rn = "SRSConf0";
             break;
         case 2:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSConf1));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSConf1));
             rn = "SRSConf1";
             break;
         case 3:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSConf2));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSConf2));
             rn = "SRSConf2";
             break;
         case 4:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSConf3));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSConf3));
             rn = "SRSConf3";
             break;
         case 5:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSConf4));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSConf4));
             rn = "SRSConf4";
             break;
         default:
@@ -3999,8 +4236,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         switch (sel) {
         case 0:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_HWREna));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_HWREna));
             rn = "HWREna";
             break;
         default:
@@ -4010,7 +4246,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 8:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_BadVAddr));
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_BadVAddr));
             rn = "BadVAddr";
             break;
         default:
@@ -4020,7 +4256,14 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 9:
         switch (sel) {
         case 0:
-            gen_op_mfc0_count();
+            /* Mark as an IO operation because we read the time.  */
+            if (use_icount)
+                gen_io_start();
+            tcg_gen_helper_1_0(do_mfc0_count, t0);
+            if (use_icount) {
+                gen_io_end();
+                ctx->bstate = BS_STOP;
+            }
             rn = "Count";
             break;
         /* 6,7 are implementation dependent */
@@ -4031,7 +4274,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 10:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_EntryHi));
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_EntryHi));
             rn = "EntryHi";
             break;
         default:
@@ -4041,8 +4284,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 11:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Compare));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Compare));
             rn = "Compare";
             break;
         /* 6,7 are implementation dependent */
@@ -4053,26 +4295,22 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 12:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Status));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Status));
             rn = "Status";
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_IntCtl));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_IntCtl));
             rn = "IntCtl";
             break;
         case 2:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSCtl));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSCtl));
             rn = "SRSCtl";
             break;
         case 3:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_SRSMap));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_SRSMap));
             rn = "SRSMap";
             break;
         default:
@@ -4082,8 +4320,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 13:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Cause));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Cause));
             rn = "Cause";
             break;
         default:
@@ -4093,7 +4330,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 14:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_EPC));
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_EPC));
             rn = "EPC";
             break;
         default:
@@ -4103,14 +4340,12 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 15:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_PRid));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_PRid));
             rn = "PRid";
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_EBase));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_EBase));
             rn = "EBase";
             break;
         default:
@@ -4120,34 +4355,28 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 16:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Config0));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Config0));
             rn = "Config";
             break;
         case 1:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Config1));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Config1));
             rn = "Config1";
             break;
         case 2:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Config2));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Config2));
             rn = "Config2";
             break;
         case 3:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Config3));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Config3));
             rn = "Config3";
             break;
        /* 6,7 are implementation dependent */
         case 6:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Config6));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Config6));
             rn = "Config6";
             break;
         case 7:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Config7));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Config7));
             rn = "Config7";
             break;
         default:
@@ -4157,7 +4386,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 17:
         switch (sel) {
         case 0:
-            gen_op_dmfc0_lladdr();
+            tcg_gen_helper_1_0(do_dmfc0_lladdr, t0);
             rn = "LLAddr";
             break;
         default:
@@ -4167,7 +4396,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 18:
         switch (sel) {
         case 0 ... 7:
-            gen_op_dmfc0_watchlo(sel);
+            tcg_gen_helper_1_i(do_dmfc0_watchlo, t0, sel);
             rn = "WatchLo";
             break;
         default:
@@ -4177,7 +4406,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 19:
         switch (sel) {
         case 0 ... 7:
-            gen_op_mfc0_watchhi(sel);
+            tcg_gen_helper_1_i(do_mfc0_watchhi, t0, sel);
             rn = "WatchHi";
             break;
         default:
@@ -4188,7 +4417,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         switch (sel) {
         case 0:
             check_insn(env, ctx, ISA_MIPS3);
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_XContext));
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_XContext));
             rn = "XContext";
             break;
         default:
@@ -4199,8 +4428,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
        /* Officially reserved, but sel 0 is used for R1x000 framemask */
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Framemask));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Framemask));
             rn = "Framemask";
             break;
         default:
@@ -4214,23 +4442,23 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 23:
         switch (sel) {
         case 0:
-            gen_op_mfc0_debug(); /* EJTAG support */
+            tcg_gen_helper_1_0(do_mfc0_debug, t0); /* EJTAG support */
             rn = "Debug";
             break;
         case 1:
-//            gen_op_dmfc0_tracecontrol(); /* PDtrace support */
+//            tcg_gen_helper_1_0(do_dmfc0_tracecontrol, t0); /* PDtrace support */
             rn = "TraceControl";
 //            break;
         case 2:
-//            gen_op_dmfc0_tracecontrol2(); /* PDtrace support */
+//            tcg_gen_helper_1_0(do_dmfc0_tracecontrol2, t0); /* PDtrace support */
             rn = "TraceControl2";
 //            break;
         case 3:
-//            gen_op_dmfc0_usertracedata(); /* PDtrace support */
+//            tcg_gen_helper_1_0(do_dmfc0_usertracedata, t0); /* PDtrace support */
             rn = "UserTraceData";
 //            break;
         case 4:
-//            gen_op_dmfc0_debug(); /* PDtrace support */
+//            tcg_gen_helper_1_0(do_dmfc0_tracebpc, t0); /* PDtrace support */
             rn = "TraceBPC";
 //            break;
         default:
@@ -4241,7 +4469,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         switch (sel) {
         case 0:
             /* EJTAG support */
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_DEPC));
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_DEPC));
             rn = "DEPC";
             break;
         default:
@@ -4251,36 +4479,35 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 25:
         switch (sel) {
         case 0:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_Performance0));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_Performance0));
             rn = "Performance0";
             break;
         case 1:
-//            gen_op_dmfc0_performance1();
+//            tcg_gen_helper_1_0(do_dmfc0_performance1, t0);
             rn = "Performance1";
 //            break;
         case 2:
-//            gen_op_dmfc0_performance2();
+//            tcg_gen_helper_1_0(do_dmfc0_performance2, t0);
             rn = "Performance2";
 //            break;
         case 3:
-//            gen_op_dmfc0_performance3();
+//            tcg_gen_helper_1_0(do_dmfc0_performance3, t0);
             rn = "Performance3";
 //            break;
         case 4:
-//            gen_op_dmfc0_performance4();
+//            tcg_gen_helper_1_0(do_dmfc0_performance4, t0);
             rn = "Performance4";
 //            break;
         case 5:
-//            gen_op_dmfc0_performance5();
+//            tcg_gen_helper_1_0(do_dmfc0_performance5, t0);
             rn = "Performance5";
 //            break;
         case 6:
-//            gen_op_dmfc0_performance6();
+//            tcg_gen_helper_1_0(do_dmfc0_performance6, t0);
             rn = "Performance6";
 //            break;
         case 7:
-//            gen_op_dmfc0_performance7();
+//            tcg_gen_helper_1_0(do_dmfc0_performance7, t0);
             rn = "Performance7";
 //            break;
         default:
@@ -4306,16 +4533,14 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         case 2:
         case 4:
         case 6:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_TagLo));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_TagLo));
             rn = "TagLo";
             break;
         case 1:
         case 3:
         case 5:
         case 7:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_DataLo));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_DataLo));
             rn = "DataLo";
             break;
         default:
@@ -4328,16 +4553,14 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         case 2:
         case 4:
         case 6:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_TagHi));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_TagHi));
             rn = "TagHi";
             break;
         case 1:
         case 3:
         case 5:
         case 7:
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_DataHi));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_DataHi));
             rn = "DataHi";
             break;
         default:
@@ -4347,7 +4570,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 30:
         switch (sel) {
         case 0:
-            tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUState, CP0_ErrorEPC));
+            tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, CP0_ErrorEPC));
             rn = "ErrorEPC";
             break;
         default:
@@ -4358,8 +4581,7 @@ static void gen_dmfc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         switch (sel) {
         case 0:
             /* EJTAG support */
-            tcg_gen_ld_i32(r_tmp, cpu_env, offsetof(CPUState, CP0_DESAVE));
-            tcg_gen_ext_i32_tl(cpu_T[0], r_tmp);
+            gen_mfc0_load32(t0, offsetof(CPUState, CP0_DESAVE));
             rn = "DESAVE";
             break;
         default:
@@ -4387,23 +4609,26 @@ die:
     generate_exception(ctx, EXCP_RI);
 }
 
-static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
+static void gen_dmtc0 (CPUState *env, DisasContext *ctx, TCGv t0, int reg, int sel)
 {
     const char *rn = "invalid";
 
     if (sel != 0)
         check_insn(env, ctx, ISA_MIPS64);
 
+    if (use_icount)
+        gen_io_start();
+
     switch (reg) {
     case 0:
         switch (sel) {
         case 0:
-            gen_op_mtc0_index();
+            tcg_gen_helper_0_1(do_mtc0_index, t0);
             rn = "Index";
             break;
         case 1:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_mvpcontrol();
+            tcg_gen_helper_0_1(do_mtc0_mvpcontrol, t0);
             rn = "MVPControl";
             break;
         case 2:
@@ -4428,37 +4653,37 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
             break;
         case 1:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_vpecontrol();
+            tcg_gen_helper_0_1(do_mtc0_vpecontrol, t0);
             rn = "VPEControl";
             break;
         case 2:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_vpeconf0();
+            tcg_gen_helper_0_1(do_mtc0_vpeconf0, t0);
             rn = "VPEConf0";
             break;
         case 3:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_vpeconf1();
+            tcg_gen_helper_0_1(do_mtc0_vpeconf1, t0);
             rn = "VPEConf1";
             break;
         case 4:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_yqmask();
+            tcg_gen_helper_0_1(do_mtc0_yqmask, t0);
             rn = "YQMask";
             break;
         case 5:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_vpeschedule();
+            tcg_gen_st_tl(t0, cpu_env, offsetof(CPUState, CP0_VPESchedule));
             rn = "VPESchedule";
             break;
         case 6:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_vpeschefback();
+            tcg_gen_st_tl(t0, cpu_env, offsetof(CPUState, CP0_VPEScheFBack));
             rn = "VPEScheFBack";
             break;
         case 7:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_vpeopt();
+            tcg_gen_helper_0_1(do_mtc0_vpeopt, t0);
             rn = "VPEOpt";
             break;
         default:
@@ -4468,42 +4693,42 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 2:
         switch (sel) {
         case 0:
-            gen_op_mtc0_entrylo0();
+            tcg_gen_helper_0_1(do_mtc0_entrylo0, t0);
             rn = "EntryLo0";
             break;
         case 1:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tcstatus();
+            tcg_gen_helper_0_1(do_mtc0_tcstatus, t0);
             rn = "TCStatus";
             break;
         case 2:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tcbind();
+            tcg_gen_helper_0_1(do_mtc0_tcbind, t0);
             rn = "TCBind";
             break;
         case 3:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tcrestart();
+            tcg_gen_helper_0_1(do_mtc0_tcrestart, t0);
             rn = "TCRestart";
             break;
         case 4:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tchalt();
+            tcg_gen_helper_0_1(do_mtc0_tchalt, t0);
             rn = "TCHalt";
             break;
         case 5:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tccontext();
+            tcg_gen_helper_0_1(do_mtc0_tccontext, t0);
             rn = "TCContext";
             break;
         case 6:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tcschedule();
+            tcg_gen_helper_0_1(do_mtc0_tcschedule, t0);
             rn = "TCSchedule";
             break;
         case 7:
             check_insn(env, ctx, ASE_MT);
-            gen_op_mtc0_tcschefback();
+            tcg_gen_helper_0_1(do_mtc0_tcschefback, t0);
             rn = "TCScheFBack";
             break;
         default:
@@ -4513,7 +4738,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 3:
         switch (sel) {
         case 0:
-            gen_op_mtc0_entrylo1();
+            tcg_gen_helper_0_1(do_mtc0_entrylo1, t0);
             rn = "EntryLo1";
             break;
         default:
@@ -4523,11 +4748,11 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 4:
         switch (sel) {
         case 0:
-            gen_op_mtc0_context();
+            tcg_gen_helper_0_1(do_mtc0_context, t0);
             rn = "Context";
             break;
         case 1:
-//           gen_op_mtc0_contextconfig(); /* SmartMIPS ASE */
+//           tcg_gen_helper_0_1(do_mtc0_contextconfig, t0); /* SmartMIPS ASE */
             rn = "ContextConfig";
 //           break;
         default:
@@ -4537,12 +4762,12 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 5:
         switch (sel) {
         case 0:
-            gen_op_mtc0_pagemask();
+            tcg_gen_helper_0_1(do_mtc0_pagemask, t0);
             rn = "PageMask";
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_pagegrain();
+            tcg_gen_helper_0_1(do_mtc0_pagegrain, t0);
             rn = "PageGrain";
             break;
         default:
@@ -4552,32 +4777,32 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 6:
         switch (sel) {
         case 0:
-            gen_op_mtc0_wired();
+            tcg_gen_helper_0_1(do_mtc0_wired, t0);
             rn = "Wired";
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsconf0();
+            tcg_gen_helper_0_1(do_mtc0_srsconf0, t0);
             rn = "SRSConf0";
             break;
         case 2:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsconf1();
+            tcg_gen_helper_0_1(do_mtc0_srsconf1, t0);
             rn = "SRSConf1";
             break;
         case 3:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsconf2();
+            tcg_gen_helper_0_1(do_mtc0_srsconf2, t0);
             rn = "SRSConf2";
             break;
         case 4:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsconf3();
+            tcg_gen_helper_0_1(do_mtc0_srsconf3, t0);
             rn = "SRSConf3";
             break;
         case 5:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsconf4();
+            tcg_gen_helper_0_1(do_mtc0_srsconf4, t0);
             rn = "SRSConf4";
             break;
         default:
@@ -4588,7 +4813,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         switch (sel) {
         case 0:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_hwrena();
+            tcg_gen_helper_0_1(do_mtc0_hwrena, t0);
             rn = "HWREna";
             break;
         default:
@@ -4602,7 +4827,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 9:
         switch (sel) {
         case 0:
-            gen_op_mtc0_count();
+            tcg_gen_helper_0_1(do_mtc0_count, t0);
             rn = "Count";
             break;
         /* 6,7 are implementation dependent */
@@ -4615,7 +4840,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 10:
         switch (sel) {
         case 0:
-            gen_op_mtc0_entryhi();
+            tcg_gen_helper_0_1(do_mtc0_entryhi, t0);
             rn = "EntryHi";
             break;
         default:
@@ -4625,7 +4850,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 11:
         switch (sel) {
         case 0:
-            gen_op_mtc0_compare();
+            tcg_gen_helper_0_1(do_mtc0_compare, t0);
             rn = "Compare";
             break;
         /* 6,7 are implementation dependent */
@@ -4638,7 +4863,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 12:
         switch (sel) {
         case 0:
-            gen_op_mtc0_status();
+            tcg_gen_helper_0_1(do_mtc0_status, t0);
             /* BS_STOP isn't good enough here, hflags may have changed. */
             gen_save_pc(ctx->pc + 4);
             ctx->bstate = BS_EXCP;
@@ -4646,21 +4871,21 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_intctl();
+            tcg_gen_helper_0_1(do_mtc0_intctl, t0);
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
             rn = "IntCtl";
             break;
         case 2:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsctl();
+            tcg_gen_helper_0_1(do_mtc0_srsctl, t0);
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
             rn = "SRSCtl";
             break;
         case 3:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_srsmap();
+            gen_mtc0_store32(t0, offsetof(CPUState, CP0_SRSMap));
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
             rn = "SRSMap";
@@ -4672,7 +4897,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 13:
         switch (sel) {
         case 0:
-            gen_op_mtc0_cause();
+            tcg_gen_helper_0_1(do_mtc0_cause, t0);
             rn = "Cause";
             break;
         default:
@@ -4684,7 +4909,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 14:
         switch (sel) {
         case 0:
-            gen_op_mtc0_epc();
+            tcg_gen_st_tl(t0, cpu_env, offsetof(CPUState, CP0_EPC));
             rn = "EPC";
             break;
         default:
@@ -4699,7 +4924,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
             break;
         case 1:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_op_mtc0_ebase();
+            tcg_gen_helper_0_1(do_mtc0_ebase, t0);
             rn = "EBase";
             break;
         default:
@@ -4709,7 +4934,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 16:
         switch (sel) {
         case 0:
-            gen_op_mtc0_config0();
+            tcg_gen_helper_0_1(do_mtc0_config0, t0);
             rn = "Config";
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
@@ -4719,7 +4944,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
             rn = "Config1";
             break;
         case 2:
-            gen_op_mtc0_config2();
+            tcg_gen_helper_0_1(do_mtc0_config2, t0);
             rn = "Config2";
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
@@ -4747,7 +4972,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 18:
         switch (sel) {
         case 0 ... 7:
-            gen_op_mtc0_watchlo(sel);
+            tcg_gen_helper_0_1i(do_mtc0_watchlo, t0, sel);
             rn = "WatchLo";
             break;
         default:
@@ -4757,7 +4982,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 19:
         switch (sel) {
         case 0 ... 7:
-            gen_op_mtc0_watchhi(sel);
+            tcg_gen_helper_0_1i(do_mtc0_watchhi, t0, sel);
             rn = "WatchHi";
             break;
         default:
@@ -4768,7 +4993,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         switch (sel) {
         case 0:
             check_insn(env, ctx, ISA_MIPS3);
-            gen_op_mtc0_xcontext();
+            tcg_gen_helper_0_1(do_mtc0_xcontext, t0);
             rn = "XContext";
             break;
         default:
@@ -4779,7 +5004,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
        /* Officially reserved, but sel 0 is used for R1x000 framemask */
         switch (sel) {
         case 0:
-            gen_op_mtc0_framemask();
+            tcg_gen_helper_0_1(do_mtc0_framemask, t0);
             rn = "Framemask";
             break;
         default:
@@ -4793,32 +5018,32 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 23:
         switch (sel) {
         case 0:
-            gen_op_mtc0_debug(); /* EJTAG support */
+            tcg_gen_helper_0_1(do_mtc0_debug, t0); /* EJTAG support */
             /* BS_STOP isn't good enough here, hflags may have changed. */
             gen_save_pc(ctx->pc + 4);
             ctx->bstate = BS_EXCP;
             rn = "Debug";
             break;
         case 1:
-//            gen_op_mtc0_tracecontrol(); /* PDtrace support */
+//            tcg_gen_helper_0_1(do_mtc0_tracecontrol, t0); /* PDtrace support */
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
             rn = "TraceControl";
 //            break;
         case 2:
-//            gen_op_mtc0_tracecontrol2(); /* PDtrace support */
+//            tcg_gen_helper_0_1(do_mtc0_tracecontrol2, t0); /* PDtrace support */
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
             rn = "TraceControl2";
 //            break;
         case 3:
-//            gen_op_mtc0_usertracedata(); /* PDtrace support */
+//            tcg_gen_helper_0_1(do_mtc0_usertracedata, t0); /* PDtrace support */
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
             rn = "UserTraceData";
 //            break;
         case 4:
-//            gen_op_mtc0_debug(); /* PDtrace support */
+//            tcg_gen_helper_0_1(do_mtc0_tracebpc, t0); /* PDtrace support */
             /* Stop translation as we may have switched the execution mode */
             ctx->bstate = BS_STOP;
             rn = "TraceBPC";
@@ -4830,7 +5055,8 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 24:
         switch (sel) {
         case 0:
-            gen_op_mtc0_depc(); /* EJTAG support */
+            /* EJTAG support */
+            tcg_gen_st_tl(t0, cpu_env, offsetof(CPUState, CP0_DEPC));
             rn = "DEPC";
             break;
         default:
@@ -4840,35 +5066,35 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 25:
         switch (sel) {
         case 0:
-            gen_op_mtc0_performance0();
+            tcg_gen_helper_0_1(do_mtc0_performance0, t0);
             rn = "Performance0";
             break;
         case 1:
-//            gen_op_mtc0_performance1();
+//            tcg_gen_helper_0_1(do_mtc0_performance1, t0);
             rn = "Performance1";
 //            break;
         case 2:
-//            gen_op_mtc0_performance2();
+//            tcg_gen_helper_0_1(do_mtc0_performance2, t0);
             rn = "Performance2";
 //            break;
         case 3:
-//            gen_op_mtc0_performance3();
+//            tcg_gen_helper_0_1(do_mtc0_performance3, t0);
             rn = "Performance3";
 //            break;
         case 4:
-//            gen_op_mtc0_performance4();
+//            tcg_gen_helper_0_1(do_mtc0_performance4, t0);
             rn = "Performance4";
 //            break;
         case 5:
-//            gen_op_mtc0_performance5();
+//            tcg_gen_helper_0_1(do_mtc0_performance5, t0);
             rn = "Performance5";
 //            break;
         case 6:
-//            gen_op_mtc0_performance6();
+//            tcg_gen_helper_0_1(do_mtc0_performance6, t0);
             rn = "Performance6";
 //            break;
         case 7:
-//            gen_op_mtc0_performance7();
+//            tcg_gen_helper_0_1(do_mtc0_performance7, t0);
             rn = "Performance7";
 //            break;
         default:
@@ -4895,14 +5121,14 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         case 2:
         case 4:
         case 6:
-            gen_op_mtc0_taglo();
+            tcg_gen_helper_0_1(do_mtc0_taglo, t0);
             rn = "TagLo";
             break;
         case 1:
         case 3:
         case 5:
         case 7:
-            gen_op_mtc0_datalo();
+            tcg_gen_helper_0_1(do_mtc0_datalo, t0);
             rn = "DataLo";
             break;
         default:
@@ -4915,14 +5141,14 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
         case 2:
         case 4:
         case 6:
-            gen_op_mtc0_taghi();
+            tcg_gen_helper_0_1(do_mtc0_taghi, t0);
             rn = "TagHi";
             break;
         case 1:
         case 3:
         case 5:
         case 7:
-            gen_op_mtc0_datahi();
+            tcg_gen_helper_0_1(do_mtc0_datahi, t0);
             rn = "DataHi";
             break;
         default:
@@ -4933,7 +5159,7 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 30:
         switch (sel) {
         case 0:
-            gen_op_mtc0_errorepc();
+            tcg_gen_st_tl(t0, cpu_env, offsetof(CPUState, CP0_ErrorEPC));
             rn = "ErrorEPC";
             break;
         default:
@@ -4943,7 +5169,8 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
     case 31:
         switch (sel) {
         case 0:
-            gen_op_mtc0_desave(); /* EJTAG support */
+            /* EJTAG support */
+            gen_mtc0_store32(t0, offsetof(CPUState, CP0_DESAVE));
             rn = "DESAVE";
             break;
         default:
@@ -4961,6 +5188,11 @@ static void gen_dmtc0 (CPUState *env, DisasContext *ctx, int reg, int sel)
                 rn, reg, sel);
     }
 #endif
+    /* For simplicity assume that all writes can cause interrupts.  */
+    if (use_icount) {
+        gen_io_end();
+        ctx->bstate = BS_STOP;
+    }
     return;
 
 die:
@@ -4974,125 +5206,126 @@ die:
 }
 #endif /* TARGET_MIPS64 */
 
-static void gen_mftr(CPUState *env, DisasContext *ctx, int rt,
+static void gen_mftr(CPUState *env, DisasContext *ctx, int rt, int rd,
                      int u, int sel, int h)
 {
     int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
 
     if ((env->CP0_VPEConf0 & (1 << CP0VPEC0_MVP)) == 0 &&
-        ((env->CP0_TCBind[other_tc] & (0xf << CP0TCBd_CurVPE)) !=
-         (env->CP0_TCBind[env->current_tc] & (0xf << CP0TCBd_CurVPE))))
-        tcg_gen_movi_tl(cpu_T[0], -1);
+        ((env->tcs[other_tc].CP0_TCBind & (0xf << CP0TCBd_CurVPE)) !=
+         (env->active_tc.CP0_TCBind & (0xf << CP0TCBd_CurVPE))))
+        tcg_gen_movi_tl(t0, -1);
     else if ((env->CP0_VPEControl & (0xff << CP0VPECo_TargTC)) >
              (env->mvp->CP0_MVPConf0 & (0xff << CP0MVPC0_PTC)))
-        tcg_gen_movi_tl(cpu_T[0], -1);
+        tcg_gen_movi_tl(t0, -1);
     else if (u == 0) {
         switch (rt) {
         case 2:
             switch (sel) {
             case 1:
-                gen_op_mftc0_tcstatus();
+                tcg_gen_helper_1_1(do_mftc0_tcstatus, t0, t0);
                 break;
             case 2:
-                gen_op_mftc0_tcbind();
+                tcg_gen_helper_1_1(do_mftc0_tcbind, t0, t0);
                 break;
             case 3:
-                gen_op_mftc0_tcrestart();
+                tcg_gen_helper_1_1(do_mftc0_tcrestart, t0, t0);
                 break;
             case 4:
-                gen_op_mftc0_tchalt();
+                tcg_gen_helper_1_1(do_mftc0_tchalt, t0, t0);
                 break;
             case 5:
-                gen_op_mftc0_tccontext();
+                tcg_gen_helper_1_1(do_mftc0_tccontext, t0, t0);
                 break;
             case 6:
-                gen_op_mftc0_tcschedule();
+                tcg_gen_helper_1_1(do_mftc0_tcschedule, t0, t0);
                 break;
             case 7:
-                gen_op_mftc0_tcschefback();
+                tcg_gen_helper_1_1(do_mftc0_tcschefback, t0, t0);
                 break;
             default:
-                gen_mfc0(env, ctx, rt, sel);
+                gen_mfc0(env, ctx, t0, rt, sel);
                 break;
             }
             break;
         case 10:
             switch (sel) {
             case 0:
-                gen_op_mftc0_entryhi();
+                tcg_gen_helper_1_1(do_mftc0_entryhi, t0, t0);
                 break;
             default:
-                gen_mfc0(env, ctx, rt, sel);
+                gen_mfc0(env, ctx, t0, rt, sel);
                 break;
             }
         case 12:
             switch (sel) {
             case 0:
-                gen_op_mftc0_status();
+                tcg_gen_helper_1_1(do_mftc0_status, t0, t0);
                 break;
             default:
-                gen_mfc0(env, ctx, rt, sel);
+                gen_mfc0(env, ctx, t0, rt, sel);
                 break;
             }
         case 23:
             switch (sel) {
             case 0:
-                gen_op_mftc0_debug();
+                tcg_gen_helper_1_1(do_mftc0_debug, t0, t0);
                 break;
             default:
-                gen_mfc0(env, ctx, rt, sel);
+                gen_mfc0(env, ctx, t0, rt, sel);
                 break;
             }
             break;
         default:
-            gen_mfc0(env, ctx, rt, sel);
+            gen_mfc0(env, ctx, t0, rt, sel);
         }
     } else switch (sel) {
     /* GPR registers. */
     case 0:
-        gen_op_mftgpr(rt);
+        tcg_gen_helper_1_1i(do_mftgpr, t0, t0, rt);
         break;
     /* Auxiliary CPU registers */
     case 1:
         switch (rt) {
         case 0:
-            gen_op_mftlo(0);
+            tcg_gen_helper_1_1i(do_mftlo, t0, t0, 0);
             break;
         case 1:
-            gen_op_mfthi(0);
+            tcg_gen_helper_1_1i(do_mfthi, t0, t0, 0);
             break;
         case 2:
-            gen_op_mftacx(0);
+            tcg_gen_helper_1_1i(do_mftacx, t0, t0, 0);
             break;
         case 4:
-            gen_op_mftlo(1);
+            tcg_gen_helper_1_1i(do_mftlo, t0, t0, 1);
             break;
         case 5:
-            gen_op_mfthi(1);
+            tcg_gen_helper_1_1i(do_mfthi, t0, t0, 1);
             break;
         case 6:
-            gen_op_mftacx(1);
+            tcg_gen_helper_1_1i(do_mftacx, t0, t0, 1);
             break;
         case 8:
-            gen_op_mftlo(2);
+            tcg_gen_helper_1_1i(do_mftlo, t0, t0, 2);
             break;
         case 9:
-            gen_op_mfthi(2);
+            tcg_gen_helper_1_1i(do_mfthi, t0, t0, 2);
             break;
         case 10:
-            gen_op_mftacx(2);
+            tcg_gen_helper_1_1i(do_mftacx, t0, t0, 2);
             break;
         case 12:
-            gen_op_mftlo(3);
+            tcg_gen_helper_1_1i(do_mftlo, t0, t0, 3);
             break;
         case 13:
-            gen_op_mfthi(3);
+            tcg_gen_helper_1_1i(do_mfthi, t0, t0, 3);
             break;
         case 14:
-            gen_op_mftacx(3);
+            tcg_gen_helper_1_1i(do_mftacx, t0, t0, 3);
             break;
         case 16:
-            gen_op_mftdsp();
+            tcg_gen_helper_1_1(do_mftdsp, t0, t0);
             break;
         default:
             goto die;
@@ -5102,16 +5335,22 @@ static void gen_mftr(CPUState *env, DisasContext *ctx, int rt,
     case 2:
         /* XXX: For now we support only a single FPU context. */
         if (h == 0) {
-            GEN_LOAD_FREG_FTN(WT0, rt);
-            gen_op_mfc1();
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, rt);
+            tcg_gen_ext_i32_tl(t0, fp0);
+            tcg_temp_free(fp0);
         } else {
-            GEN_LOAD_FREG_FTN(WTH0, rt);
-            gen_op_mfhc1();
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32h(fp0, rt);
+            tcg_gen_ext_i32_tl(t0, fp0);
+            tcg_temp_free(fp0);
         }
         break;
     case 3:
         /* XXX: For now we support only a single FPU context. */
-        gen_op_cfc1(rt);
+        tcg_gen_helper_1_1i(do_cfc1, t0, t0, rt);
         break;
     /* COP2: Not implemented. */
     case 4:
@@ -5126,9 +5365,12 @@ static void gen_mftr(CPUState *env, DisasContext *ctx, int rt,
                 rt, u, sel, h);
     }
 #endif
+    gen_store_gpr(t0, rd);
+    tcg_temp_free(t0);
     return;
 
 die:
+    tcg_temp_free(t0);
 #if defined MIPS_DEBUG_DISAS
     if (loglevel & CPU_LOG_TB_IN_ASM) {
         fprintf(logfile, "mftr (reg %d u %d sel %d h %d)\n",
@@ -5138,14 +5380,16 @@ die:
     generate_exception(ctx, EXCP_RI);
 }
 
-static void gen_mttr(CPUState *env, DisasContext *ctx, int rd,
+static void gen_mttr(CPUState *env, DisasContext *ctx, int rd, int rt,
                      int u, int sel, int h)
 {
     int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
 
+    gen_load_gpr(t0, rt);
     if ((env->CP0_VPEConf0 & (1 << CP0VPEC0_MVP)) == 0 &&
-        ((env->CP0_TCBind[other_tc] & (0xf << CP0TCBd_CurVPE)) !=
-         (env->CP0_TCBind[env->current_tc] & (0xf << CP0TCBd_CurVPE))))
+        ((env->tcs[other_tc].CP0_TCBind & (0xf << CP0TCBd_CurVPE)) !=
+         (env->active_tc.CP0_TCBind & (0xf << CP0TCBd_CurVPE))))
         /* NOP */ ;
     else if ((env->CP0_VPEControl & (0xff << CP0VPECo_TargTC)) >
              (env->mvp->CP0_MVPConf0 & (0xff << CP0MVPC0_PTC)))
@@ -5155,108 +5399,108 @@ static void gen_mttr(CPUState *env, DisasContext *ctx, int rd,
         case 2:
             switch (sel) {
             case 1:
-                gen_op_mttc0_tcstatus();
+                tcg_gen_helper_0_1(do_mttc0_tcstatus, t0);
                 break;
             case 2:
-                gen_op_mttc0_tcbind();
+                tcg_gen_helper_0_1(do_mttc0_tcbind, t0);
                 break;
             case 3:
-                gen_op_mttc0_tcrestart();
+                tcg_gen_helper_0_1(do_mttc0_tcrestart, t0);
                 break;
             case 4:
-                gen_op_mttc0_tchalt();
+                tcg_gen_helper_0_1(do_mttc0_tchalt, t0);
                 break;
             case 5:
-                gen_op_mttc0_tccontext();
+                tcg_gen_helper_0_1(do_mttc0_tccontext, t0);
                 break;
             case 6:
-                gen_op_mttc0_tcschedule();
+                tcg_gen_helper_0_1(do_mttc0_tcschedule, t0);
                 break;
             case 7:
-                gen_op_mttc0_tcschefback();
+                tcg_gen_helper_0_1(do_mttc0_tcschefback, t0);
                 break;
             default:
-                gen_mtc0(env, ctx, rd, sel);
+                gen_mtc0(env, ctx, t0, rd, sel);
                 break;
             }
             break;
         case 10:
             switch (sel) {
             case 0:
-                gen_op_mttc0_entryhi();
+                tcg_gen_helper_0_1(do_mttc0_entryhi, t0);
                 break;
             default:
-                gen_mtc0(env, ctx, rd, sel);
+                gen_mtc0(env, ctx, t0, rd, sel);
                 break;
             }
         case 12:
             switch (sel) {
             case 0:
-                gen_op_mttc0_status();
+                tcg_gen_helper_0_1(do_mttc0_status, t0);
                 break;
             default:
-                gen_mtc0(env, ctx, rd, sel);
+                gen_mtc0(env, ctx, t0, rd, sel);
                 break;
             }
         case 23:
             switch (sel) {
             case 0:
-                gen_op_mttc0_debug();
+                tcg_gen_helper_0_1(do_mttc0_debug, t0);
                 break;
             default:
-                gen_mtc0(env, ctx, rd, sel);
+                gen_mtc0(env, ctx, t0, rd, sel);
                 break;
             }
             break;
         default:
-            gen_mtc0(env, ctx, rd, sel);
+            gen_mtc0(env, ctx, t0, rd, sel);
         }
     } else switch (sel) {
     /* GPR registers. */
     case 0:
-        gen_op_mttgpr(rd);
+        tcg_gen_helper_0_1i(do_mttgpr, t0, rd);
         break;
     /* Auxiliary CPU registers */
     case 1:
         switch (rd) {
         case 0:
-            gen_op_mttlo(0);
+            tcg_gen_helper_0_1i(do_mttlo, t0, 0);
             break;
         case 1:
-            gen_op_mtthi(0);
+            tcg_gen_helper_0_1i(do_mtthi, t0, 0);
             break;
         case 2:
-            gen_op_mttacx(0);
+            tcg_gen_helper_0_1i(do_mttacx, t0, 0);
             break;
         case 4:
-            gen_op_mttlo(1);
+            tcg_gen_helper_0_1i(do_mttlo, t0, 1);
             break;
         case 5:
-            gen_op_mtthi(1);
+            tcg_gen_helper_0_1i(do_mtthi, t0, 1);
             break;
         case 6:
-            gen_op_mttacx(1);
+            tcg_gen_helper_0_1i(do_mttacx, t0, 1);
             break;
         case 8:
-            gen_op_mttlo(2);
+            tcg_gen_helper_0_1i(do_mttlo, t0, 2);
             break;
         case 9:
-            gen_op_mtthi(2);
+            tcg_gen_helper_0_1i(do_mtthi, t0, 2);
             break;
         case 10:
-            gen_op_mttacx(2);
+            tcg_gen_helper_0_1i(do_mttacx, t0, 2);
             break;
         case 12:
-            gen_op_mttlo(3);
+            tcg_gen_helper_0_1i(do_mttlo, t0, 3);
             break;
         case 13:
-            gen_op_mtthi(3);
+            tcg_gen_helper_0_1i(do_mtthi, t0, 3);
             break;
         case 14:
-            gen_op_mttacx(3);
+            tcg_gen_helper_0_1i(do_mttacx, t0, 3);
             break;
         case 16:
-            gen_op_mttdsp();
+            tcg_gen_helper_0_1(do_mttdsp, t0);
             break;
         default:
             goto die;
@@ -5266,16 +5510,22 @@ static void gen_mttr(CPUState *env, DisasContext *ctx, int rd,
     case 2:
         /* XXX: For now we support only a single FPU context. */
         if (h == 0) {
-            gen_op_mtc1();
-            GEN_STORE_FTN_FREG(rd, WT0);
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            tcg_gen_trunc_tl_i32(fp0, t0);
+            gen_store_fpr32(fp0, rd);
+            tcg_temp_free(fp0);
         } else {
-            gen_op_mthc1();
-            GEN_STORE_FTN_FREG(rd, WTH0);
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            tcg_gen_trunc_tl_i32(fp0, t0);
+            gen_store_fpr32h(fp0, rd);
+            tcg_temp_free(fp0);
         }
         break;
     case 3:
         /* XXX: For now we support only a single FPU context. */
-        gen_op_ctc1(rd);
+        tcg_gen_helper_0_1i(do_ctc1, t0, rd);
         break;
     /* COP2: Not implemented. */
     case 4:
@@ -5290,9 +5540,11 @@ static void gen_mttr(CPUState *env, DisasContext *ctx, int rd,
                 rd, u, sel, h);
     }
 #endif
+    tcg_temp_free(t0);
     return;
 
 die:
+    tcg_temp_free(t0);
 #if defined MIPS_DEBUG_DISAS
     if (loglevel & CPU_LOG_TB_IN_ASM) {
         fprintf(logfile, "mttr (reg %d u %d sel %d h %d)\n",
@@ -5312,14 +5564,24 @@ static void gen_cp0 (CPUState *env, DisasContext *ctx, uint32_t opc, int rt, int
             /* Treat as NOP. */
             return;
         }
-        gen_mfc0(env, ctx, rd, ctx->opcode & 0x7);
-        gen_store_gpr(cpu_T[0], rt);
+        {
+            TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+
+            gen_mfc0(env, ctx, t0, rd, ctx->opcode & 0x7);
+            gen_store_gpr(t0, rt);
+            tcg_temp_free(t0);
+        }
         opn = "mfc0";
         break;
     case OPC_MTC0:
-        gen_load_gpr(cpu_T[0], rt);
-        save_cpu_state(ctx, 1);
-        gen_mtc0(env, ctx, rd, ctx->opcode & 0x7);
+        {
+            TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+
+            gen_load_gpr(t0, rt);
+            save_cpu_state(ctx, 1);
+            gen_mtc0(env, ctx, t0, rd, ctx->opcode & 0x7);
+            tcg_temp_free(t0);
+        }
         opn = "mtc0";
         break;
 #if defined(TARGET_MIPS64)
@@ -5329,15 +5591,25 @@ static void gen_cp0 (CPUState *env, DisasContext *ctx, uint32_t opc, int rt, int
             /* Treat as NOP. */
             return;
         }
-        gen_dmfc0(env, ctx, rd, ctx->opcode & 0x7);
-        gen_store_gpr(cpu_T[0], rt);
+        {
+            TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+
+            gen_dmfc0(env, ctx, t0, rd, ctx->opcode & 0x7);
+            gen_store_gpr(t0, rt);
+            tcg_temp_free(t0);
+        }
         opn = "dmfc0";
         break;
     case OPC_DMTC0:
         check_insn(env, ctx, ISA_MIPS3);
-        gen_load_gpr(cpu_T[0], rt);
-        save_cpu_state(ctx, 1);
-        gen_dmtc0(env, ctx, rd, ctx->opcode & 0x7);
+        {
+            TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+
+            gen_load_gpr(t0, rt);
+            save_cpu_state(ctx, 1);
+            gen_dmtc0(env, ctx, t0, rd, ctx->opcode & 0x7);
+            tcg_temp_free(t0);
+        }
         opn = "dmtc0";
         break;
 #endif
@@ -5347,15 +5619,13 @@ static void gen_cp0 (CPUState *env, DisasContext *ctx, uint32_t opc, int rt, int
             /* Treat as NOP. */
             return;
         }
-        gen_mftr(env, ctx, rt, (ctx->opcode >> 5) & 1,
+        gen_mftr(env, ctx, rt, rd, (ctx->opcode >> 5) & 1,
                  ctx->opcode & 0x7, (ctx->opcode >> 4) & 1);
-        gen_store_gpr(cpu_T[0], rd);
         opn = "mftr";
         break;
     case OPC_MTTR:
         check_insn(env, ctx, ASE_MT);
-        gen_load_gpr(cpu_T[0], rt);
-        gen_mttr(env, ctx, rd, (ctx->opcode >> 5) & 1,
+        gen_mttr(env, ctx, rd, rt, (ctx->opcode >> 5) & 1,
                  ctx->opcode & 0x7, (ctx->opcode >> 4) & 1);
         opn = "mttr";
         break;
@@ -5363,31 +5633,31 @@ static void gen_cp0 (CPUState *env, DisasContext *ctx, uint32_t opc, int rt, int
         opn = "tlbwi";
         if (!env->tlb->do_tlbwi)
             goto die;
-        gen_op_tlbwi();
+        tcg_gen_helper_0_0(env->tlb->do_tlbwi);
         break;
     case OPC_TLBWR:
         opn = "tlbwr";
         if (!env->tlb->do_tlbwr)
             goto die;
-        gen_op_tlbwr();
+        tcg_gen_helper_0_0(env->tlb->do_tlbwr);
         break;
     case OPC_TLBP:
         opn = "tlbp";
         if (!env->tlb->do_tlbp)
             goto die;
-        gen_op_tlbp();
+        tcg_gen_helper_0_0(env->tlb->do_tlbp);
         break;
     case OPC_TLBR:
         opn = "tlbr";
         if (!env->tlb->do_tlbr)
             goto die;
-        gen_op_tlbr();
+        tcg_gen_helper_0_0(env->tlb->do_tlbr);
         break;
     case OPC_ERET:
         opn = "eret";
         check_insn(env, ctx, ISA_MIPS2);
         save_cpu_state(ctx, 1);
-        gen_op_eret();
+        tcg_gen_helper_0_0(do_eret);
         ctx->bstate = BS_EXCP;
         break;
     case OPC_DERET:
@@ -5398,7 +5668,7 @@ static void gen_cp0 (CPUState *env, DisasContext *ctx, uint32_t opc, int rt, int
             generate_exception(ctx, EXCP_RI);
         } else {
             save_cpu_state(ctx, 1);
-            gen_op_deret();
+            tcg_gen_helper_0_0(do_deret);
             ctx->bstate = BS_EXCP;
         }
         break;
@@ -5409,7 +5679,7 @@ static void gen_cp0 (CPUState *env, DisasContext *ctx, uint32_t opc, int rt, int
         ctx->pc += 4;
         save_cpu_state(ctx, 1);
         ctx->pc -= 4;
-        gen_op_wait();
+        tcg_gen_helper_0_0(do_wait);
         ctx->bstate = BS_EXCP;
         break;
     default:
@@ -5420,6 +5690,7 @@ static void gen_cp0 (CPUState *env, DisasContext *ctx, uint32_t opc, int rt, int
     }
     MIPS_DEBUG("%s %s %d", opn, regnames[rt], rd);
 }
+#endif /* !CONFIG_USER_ONLY */
 
 /* CP1 Branches (before delay slot) */
 static void gen_compute_branch1 (CPUState *env, DisasContext *ctx, uint32_t op,
@@ -5427,6 +5698,8 @@ static void gen_compute_branch1 (CPUState *env, DisasContext *ctx, uint32_t op,
 {
     target_ulong btarget;
     const char *opn = "cp1 cond branch";
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+    TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
 
     if (cc != 0)
         check_insn(env, ctx, ISA_MIPS4 | ISA_MIPS32);
@@ -5435,51 +5708,187 @@ static void gen_compute_branch1 (CPUState *env, DisasContext *ctx, uint32_t op,
 
     switch (op) {
     case OPC_BC1F:
-        gen_op_bc1f(cc);
+        {
+            int l1 = gen_new_label();
+            int l2 = gen_new_label();
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            get_fp_cond(r_tmp1);
+            tcg_gen_ext_i32_tl(t0, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_not_tl(t0, t0);
+            tcg_gen_movi_tl(t1, 0x1 << cc);
+            tcg_gen_and_tl(t0, t0, t1);
+            tcg_gen_brcondi_tl(TCG_COND_NE, t0, 0, l1);
+            tcg_gen_movi_tl(t0, 0);
+            tcg_gen_br(l2);
+            gen_set_label(l1);
+            tcg_gen_movi_tl(t0, 1);
+            gen_set_label(l2);
+        }
         opn = "bc1f";
         goto not_likely;
     case OPC_BC1FL:
-        gen_op_bc1f(cc);
+        {
+            int l1 = gen_new_label();
+            int l2 = gen_new_label();
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            get_fp_cond(r_tmp1);
+            tcg_gen_ext_i32_tl(t0, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_not_tl(t0, t0);
+            tcg_gen_movi_tl(t1, 0x1 << cc);
+            tcg_gen_and_tl(t0, t0, t1);
+            tcg_gen_brcondi_tl(TCG_COND_NE, t0, 0, l1);
+            tcg_gen_movi_tl(t0, 0);
+            tcg_gen_br(l2);
+            gen_set_label(l1);
+            tcg_gen_movi_tl(t0, 1);
+            gen_set_label(l2);
+        }
         opn = "bc1fl";
         goto likely;
     case OPC_BC1T:
-        gen_op_bc1t(cc);
+        {
+            int l1 = gen_new_label();
+            int l2 = gen_new_label();
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            get_fp_cond(r_tmp1);
+            tcg_gen_ext_i32_tl(t0, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_movi_tl(t1, 0x1 << cc);
+            tcg_gen_and_tl(t0, t0, t1);
+            tcg_gen_brcondi_tl(TCG_COND_NE, t0, 0, l1);
+            tcg_gen_movi_tl(t0, 0);
+            tcg_gen_br(l2);
+            gen_set_label(l1);
+            tcg_gen_movi_tl(t0, 1);
+            gen_set_label(l2);
+        }
         opn = "bc1t";
         goto not_likely;
     case OPC_BC1TL:
-        gen_op_bc1t(cc);
-        opn = "bc1tl";
+        {
+            int l1 = gen_new_label();
+            int l2 = gen_new_label();
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            get_fp_cond(r_tmp1);
+            tcg_gen_ext_i32_tl(t0, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_movi_tl(t1, 0x1 << cc);
+            tcg_gen_and_tl(t0, t0, t1);
+            tcg_gen_brcondi_tl(TCG_COND_NE, t0, 0, l1);
+            tcg_gen_movi_tl(t0, 0);
+            tcg_gen_br(l2);
+            gen_set_label(l1);
+            tcg_gen_movi_tl(t0, 1);
+            gen_set_label(l2);
+        }
+        opn = "bc1tl";
     likely:
         ctx->hflags |= MIPS_HFLAG_BL;
-        tcg_gen_st_tl(cpu_T[0], cpu_env, offsetof(CPUState, bcond));
+        tcg_gen_trunc_tl_i32(bcond, t0);
         break;
     case OPC_BC1FANY2:
-        gen_op_bc1any2f(cc);
+        {
+            int l1 = gen_new_label();
+            int l2 = gen_new_label();
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            get_fp_cond(r_tmp1);
+            tcg_gen_ext_i32_tl(t0, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_not_tl(t0, t0);
+            tcg_gen_movi_tl(t1, 0x3 << cc);
+            tcg_gen_and_tl(t0, t0, t1);
+            tcg_gen_brcondi_tl(TCG_COND_NE, t0, 0, l1);
+            tcg_gen_movi_tl(t0, 0);
+            tcg_gen_br(l2);
+            gen_set_label(l1);
+            tcg_gen_movi_tl(t0, 1);
+            gen_set_label(l2);
+        }
         opn = "bc1any2f";
         goto not_likely;
     case OPC_BC1TANY2:
-        gen_op_bc1any2t(cc);
+        {
+            int l1 = gen_new_label();
+            int l2 = gen_new_label();
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            get_fp_cond(r_tmp1);
+            tcg_gen_ext_i32_tl(t0, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_movi_tl(t1, 0x3 << cc);
+            tcg_gen_and_tl(t0, t0, t1);
+            tcg_gen_brcondi_tl(TCG_COND_NE, t0, 0, l1);
+            tcg_gen_movi_tl(t0, 0);
+            tcg_gen_br(l2);
+            gen_set_label(l1);
+            tcg_gen_movi_tl(t0, 1);
+            gen_set_label(l2);
+        }
         opn = "bc1any2t";
         goto not_likely;
     case OPC_BC1FANY4:
-        gen_op_bc1any4f(cc);
+        {
+            int l1 = gen_new_label();
+            int l2 = gen_new_label();
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            get_fp_cond(r_tmp1);
+            tcg_gen_ext_i32_tl(t0, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_not_tl(t0, t0);
+            tcg_gen_movi_tl(t1, 0xf << cc);
+            tcg_gen_and_tl(t0, t0, t1);
+            tcg_gen_brcondi_tl(TCG_COND_NE, t0, 0, l1);
+            tcg_gen_movi_tl(t0, 0);
+            tcg_gen_br(l2);
+            gen_set_label(l1);
+            tcg_gen_movi_tl(t0, 1);
+            gen_set_label(l2);
+        }
         opn = "bc1any4f";
         goto not_likely;
     case OPC_BC1TANY4:
-        gen_op_bc1any4t(cc);
+        {
+            int l1 = gen_new_label();
+            int l2 = gen_new_label();
+            TCGv r_tmp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            get_fp_cond(r_tmp1);
+            tcg_gen_ext_i32_tl(t0, r_tmp1);
+            tcg_temp_free(r_tmp1);
+            tcg_gen_movi_tl(t1, 0xf << cc);
+            tcg_gen_and_tl(t0, t0, t1);
+            tcg_gen_brcondi_tl(TCG_COND_NE, t0, 0, l1);
+            tcg_gen_movi_tl(t0, 0);
+            tcg_gen_br(l2);
+            gen_set_label(l1);
+            tcg_gen_movi_tl(t0, 1);
+            gen_set_label(l2);
+        }
         opn = "bc1any4t";
     not_likely:
         ctx->hflags |= MIPS_HFLAG_BC;
-        tcg_gen_st_tl(cpu_T[0], cpu_env, offsetof(CPUState, bcond));
+        tcg_gen_trunc_tl_i32(bcond, t0);
         break;
     default:
         MIPS_INVAL(opn);
         generate_exception (ctx, EXCP_RI);
-        return;
+        goto out;
     }
     MIPS_DEBUG("%s: cond %02x target " TARGET_FMT_lx, opn,
                ctx->hflags, btarget);
     ctx->btarget = btarget;
+
+ out:
+    tcg_temp_free(t0);
+    tcg_temp_free(t1);
 }
 
 /* Coprocessor 1 (FPU) */
@@ -5489,60 +5898,94 @@ static void gen_compute_branch1 (CPUState *env, DisasContext *ctx, uint32_t op,
 static void gen_cp1 (DisasContext *ctx, uint32_t opc, int rt, int fs)
 {
     const char *opn = "cp1 move";
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
 
     switch (opc) {
     case OPC_MFC1:
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_mfc1();
-        gen_store_gpr(cpu_T[0], rt);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_ext_i32_tl(t0, fp0);
+            tcg_temp_free(fp0);
+       }
+        gen_store_gpr(t0, rt);
         opn = "mfc1";
         break;
     case OPC_MTC1:
-        gen_load_gpr(cpu_T[0], rt);
-        gen_op_mtc1();
-        GEN_STORE_FTN_FREG(fs, WT0);
+        gen_load_gpr(t0, rt);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            tcg_gen_trunc_tl_i32(fp0, t0);
+            gen_store_fpr32(fp0, fs);
+            tcg_temp_free(fp0);
+       }
         opn = "mtc1";
         break;
     case OPC_CFC1:
-        gen_op_cfc1(fs);
-        gen_store_gpr(cpu_T[0], rt);
+        tcg_gen_helper_1_i(do_cfc1, t0, fs);
+        gen_store_gpr(t0, rt);
         opn = "cfc1";
         break;
     case OPC_CTC1:
-        gen_load_gpr(cpu_T[0], rt);
-        gen_op_ctc1(fs);
+        gen_load_gpr(t0, rt);
+        tcg_gen_helper_0_1i(do_ctc1, t0, fs);
         opn = "ctc1";
         break;
     case OPC_DMFC1:
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_dmfc1();
-        gen_store_gpr(cpu_T[0], rt);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_mov_tl(t0, fp0);
+            tcg_temp_free(fp0);
+       }
+        gen_store_gpr(t0, rt);
         opn = "dmfc1";
         break;
     case OPC_DMTC1:
-        gen_load_gpr(cpu_T[0], rt);
-        gen_op_dmtc1();
-        GEN_STORE_FTN_FREG(fs, DT0);
+        gen_load_gpr(t0, rt);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            tcg_gen_mov_tl(fp0, t0);
+            gen_store_fpr64(ctx, fp0, fs);
+            tcg_temp_free(fp0);
+       }
         opn = "dmtc1";
         break;
     case OPC_MFHC1:
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        gen_op_mfhc1();
-        gen_store_gpr(cpu_T[0], rt);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32h(fp0, fs);
+            tcg_gen_ext_i32_tl(t0, fp0);
+            tcg_temp_free(fp0);
+       }
+        gen_store_gpr(t0, rt);
         opn = "mfhc1";
         break;
     case OPC_MTHC1:
-        gen_load_gpr(cpu_T[0], rt);
-        gen_op_mthc1();
-        GEN_STORE_FTN_FREG(fs, WTH0);
+        gen_load_gpr(t0, rt);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            tcg_gen_trunc_tl_i32(fp0, t0);
+            gen_store_fpr32h(fp0, fs);
+            tcg_temp_free(fp0);
+       }
         opn = "mthc1";
         break;
     default:
         MIPS_INVAL(opn);
         generate_exception (ctx, EXCP_RI);
-        return;
+        goto out;
     }
     MIPS_DEBUG("%s %s %s", opn, regnames[rt], fregnames[fs]);
+
+ out:
+    tcg_temp_free(t0);
 }
 
 static void gen_movci (DisasContext *ctx, int rd, int rs, int cc, int tf)
@@ -5550,6 +5993,9 @@ static void gen_movci (DisasContext *ctx, int rd, int rs, int cc, int tf)
     int l1 = gen_new_label();
     uint32_t ccbit;
     TCGCond cond;
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+    TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
+    TCGv r_tmp = tcg_temp_local_new(TCG_TYPE_I32);
 
     if (cc)
         ccbit = 1 << (24 + cc);
@@ -5560,41 +6006,126 @@ static void gen_movci (DisasContext *ctx, int rd, int rs, int cc, int tf)
     else
         cond = TCG_COND_NE;
 
-    gen_load_gpr(cpu_T[0], rd);
-    gen_load_gpr(cpu_T[1], rs);
-    {
-        TCGv r_ptr = tcg_temp_new(TCG_TYPE_PTR);
-        TCGv r_tmp = new_tmp();
-
-        tcg_gen_ld_ptr(r_ptr, cpu_env, offsetof(CPUState, fpu));
-        tcg_gen_ld_i32(r_tmp, r_ptr, offsetof(CPUMIPSFPUContext, fcr31));
-        tcg_gen_andi_i32(r_tmp, r_tmp, ccbit);
-        tcg_gen_brcondi_i32(cond, r_tmp, 0, l1);
-        dead_tmp(r_tmp);
-    }
-    tcg_gen_mov_tl(cpu_T[0], cpu_T[1]);
+    gen_load_gpr(t0, rd);
+    gen_load_gpr(t1, rs);
+    tcg_gen_ld_i32(r_tmp, current_fpu, offsetof(CPUMIPSFPUContext, fcr31));
+    tcg_gen_andi_i32(r_tmp, r_tmp, ccbit);
+    tcg_gen_brcondi_i32(cond, r_tmp, 0, l1);
+    tcg_temp_free(r_tmp);
+
+    tcg_gen_mov_tl(t0, t1);
+    tcg_temp_free(t1);
 
     gen_set_label(l1);
-    gen_store_gpr(cpu_T[0], rd);
+    gen_store_gpr(t0, rd);
+    tcg_temp_free(t0);
 }
 
-#define GEN_MOVCF(fmt)                                                \
-static void glue(gen_movcf_, fmt) (DisasContext *ctx, int cc, int tf) \
-{                                                                     \
-    uint32_t ccbit;                                                   \
-                                                                      \
-    if (cc) {                                                         \
-        ccbit = 1 << (24 + cc);                                       \
-    } else                                                            \
-        ccbit = 1 << 23;                                              \
-    if (!tf)                                                          \
-        glue(gen_op_float_movf_, fmt)(ccbit);                         \
-    else                                                              \
-        glue(gen_op_float_movt_, fmt)(ccbit);                         \
+static inline void gen_movcf_s (int fs, int fd, int cc, int tf)
+{
+    uint32_t ccbit;
+    int cond;
+    TCGv r_tmp1 = tcg_temp_local_new(TCG_TYPE_I32);
+    TCGv fp0 = tcg_temp_local_new(TCG_TYPE_I32);
+    TCGv fp1 = tcg_temp_local_new(TCG_TYPE_I32);
+    int l1 = gen_new_label();
+
+    if (cc)
+        ccbit = 1 << (24 + cc);
+    else
+        ccbit = 1 << 23;
+
+    if (tf)
+        cond = TCG_COND_EQ;
+    else
+        cond = TCG_COND_NE;
+
+    gen_load_fpr32(fp0, fs);
+    gen_load_fpr32(fp1, fd);
+    tcg_gen_ld_i32(r_tmp1, current_fpu, offsetof(CPUMIPSFPUContext, fcr31));
+    tcg_gen_andi_i32(r_tmp1, r_tmp1, ccbit);
+    tcg_gen_brcondi_i32(cond, r_tmp1, 0, l1);
+    tcg_gen_movi_i32(fp1, fp0);
+    tcg_temp_free(fp0);
+    gen_set_label(l1);
+    tcg_temp_free(r_tmp1);
+    gen_store_fpr32(fp1, fd);
+    tcg_temp_free(fp1);
 }
-GEN_MOVCF(d);
-GEN_MOVCF(s);
-#undef GEN_MOVCF
+
+static inline void gen_movcf_d (DisasContext *ctx, int fs, int fd, int cc, int tf)
+{
+    uint32_t ccbit;
+    int cond;
+    TCGv r_tmp1 = tcg_temp_local_new(TCG_TYPE_I32);
+    TCGv fp0 = tcg_temp_local_new(TCG_TYPE_I64);
+    TCGv fp1 = tcg_temp_local_new(TCG_TYPE_I64);
+    int l1 = gen_new_label();
+
+    if (cc)
+        ccbit = 1 << (24 + cc);
+    else
+        ccbit = 1 << 23;
+
+    if (tf)
+        cond = TCG_COND_EQ;
+    else
+        cond = TCG_COND_NE;
+
+    gen_load_fpr64(ctx, fp0, fs);
+    gen_load_fpr64(ctx, fp1, fd);
+    tcg_gen_ld_i32(r_tmp1, current_fpu, offsetof(CPUMIPSFPUContext, fcr31));
+    tcg_gen_andi_i32(r_tmp1, r_tmp1, ccbit);
+    tcg_gen_brcondi_i32(cond, r_tmp1, 0, l1);
+    tcg_gen_movi_i64(fp1, fp0);
+    tcg_temp_free(fp0);
+    gen_set_label(l1);
+    tcg_temp_free(r_tmp1);
+    gen_store_fpr64(ctx, fp1, fd);
+    tcg_temp_free(fp1);
+}
+
+static inline void gen_movcf_ps (int fs, int fd, int cc, int tf)
+{
+    int cond;
+    TCGv r_tmp1 = tcg_temp_local_new(TCG_TYPE_I32);
+    TCGv r_tmp2 = tcg_temp_local_new(TCG_TYPE_I32);
+    TCGv fp0 = tcg_temp_local_new(TCG_TYPE_I32);
+    TCGv fph0 = tcg_temp_local_new(TCG_TYPE_I32);
+    TCGv fp1 = tcg_temp_local_new(TCG_TYPE_I32);
+    TCGv fph1 = tcg_temp_local_new(TCG_TYPE_I32);
+    int l1 = gen_new_label();
+    int l2 = gen_new_label();
+
+    if (tf)
+        cond = TCG_COND_EQ;
+    else
+        cond = TCG_COND_NE;
+
+    gen_load_fpr32(fp0, fs);
+    gen_load_fpr32h(fph0, fs);
+    gen_load_fpr32(fp1, fd);
+    gen_load_fpr32h(fph1, fd);
+    get_fp_cond(r_tmp1);
+    tcg_gen_shri_i32(r_tmp1, r_tmp1, cc);
+    tcg_gen_andi_i32(r_tmp2, r_tmp1, 0x1);
+    tcg_gen_brcondi_i32(cond, r_tmp2, 0, l1);
+    tcg_gen_movi_i32(fp1, fp0);
+    tcg_temp_free(fp0);
+    gen_set_label(l1);
+    tcg_gen_andi_i32(r_tmp2, r_tmp1, 0x2);
+    tcg_gen_brcondi_i32(cond, r_tmp2, 0, l2);
+    tcg_gen_movi_i32(fph1, fph0);
+    tcg_temp_free(fph0);
+    gen_set_label(l2);
+    tcg_temp_free(r_tmp1);
+    tcg_temp_free(r_tmp2);
+    gen_store_fpr32(fp1, fd);
+    gen_store_fpr32h(fph1, fd);
+    tcg_temp_free(fp1);
+    tcg_temp_free(fph1);
+}
+
 
 static void gen_farith (DisasContext *ctx, uint32_t op1,
                         int ft, int fs, int fd, int cc)
@@ -5641,207 +6172,381 @@ static void gen_farith (DisasContext *ctx, uint32_t op1,
 
     switch (ctx->opcode & FOP(0x3f, 0x1f)) {
     case FOP(0, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        gen_op_float_add_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32(fp1, ft);
+            tcg_gen_helper_1_2(do_float_add_s, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "add.s";
         optype = BINOP;
         break;
     case FOP(1, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        gen_op_float_sub_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32(fp1, ft);
+            tcg_gen_helper_1_2(do_float_sub_s, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "sub.s";
         optype = BINOP;
         break;
     case FOP(2, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        gen_op_float_mul_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32(fp1, ft);
+            tcg_gen_helper_1_2(do_float_mul_s, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "mul.s";
         optype = BINOP;
         break;
     case FOP(3, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        gen_op_float_div_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32(fp1, ft);
+            tcg_gen_helper_1_2(do_float_div_s, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "div.s";
         optype = BINOP;
         break;
     case FOP(4, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_sqrt_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_sqrt_s, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "sqrt.s";
         break;
     case FOP(5, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_abs_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_abs_s, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "abs.s";
         break;
     case FOP(6, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_mov_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "mov.s";
         break;
     case FOP(7, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_chs_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_chs_s, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "neg.s";
         break;
     case FOP(8, 16):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_roundl_s();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr32(fp32, fs);
+            tcg_gen_helper_1_1(do_float_roundl_s, fp64, fp32);
+            tcg_temp_free(fp32);
+            gen_store_fpr64(ctx, fp64, fd);
+            tcg_temp_free(fp64);
+        }
         opn = "round.l.s";
         break;
     case FOP(9, 16):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_truncl_s();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr32(fp32, fs);
+            tcg_gen_helper_1_1(do_float_truncl_s, fp64, fp32);
+            tcg_temp_free(fp32);
+            gen_store_fpr64(ctx, fp64, fd);
+            tcg_temp_free(fp64);
+        }
         opn = "trunc.l.s";
         break;
     case FOP(10, 16):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_ceill_s();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr32(fp32, fs);
+            tcg_gen_helper_1_1(do_float_ceill_s, fp64, fp32);
+            tcg_temp_free(fp32);
+            gen_store_fpr64(ctx, fp64, fd);
+            tcg_temp_free(fp64);
+        }
         opn = "ceil.l.s";
         break;
     case FOP(11, 16):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_floorl_s();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr32(fp32, fs);
+            tcg_gen_helper_1_1(do_float_floorl_s, fp64, fp32);
+            tcg_temp_free(fp32);
+            gen_store_fpr64(ctx, fp64, fd);
+            tcg_temp_free(fp64);
+        }
         opn = "floor.l.s";
         break;
     case FOP(12, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_roundw_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_roundw_s, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "round.w.s";
         break;
     case FOP(13, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_truncw_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_truncw_s, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "trunc.w.s";
         break;
     case FOP(14, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_ceilw_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_ceilw_s, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "ceil.w.s";
         break;
     case FOP(15, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_floorw_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_floorw_s, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "floor.w.s";
         break;
     case FOP(17, 16):
-        gen_load_gpr(cpu_T[0], ft);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT2, fd);
-        gen_movcf_s(ctx, (ft >> 2) & 0x7, ft & 0x1);
-        GEN_STORE_FTN_FREG(fd, WT2);
+        gen_movcf_s(fs, fd, (ft >> 2) & 0x7, ft & 0x1);
         opn = "movcf.s";
         break;
     case FOP(18, 16):
-        gen_load_gpr(cpu_T[0], ft);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT2, fd);
-        gen_op_float_movz_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            int l1 = gen_new_label();
+            TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+            TCGv fp0 = tcg_temp_local_new(TCG_TYPE_I32);
+
+            gen_load_gpr(t0, ft);
+            tcg_gen_brcondi_tl(TCG_COND_NE, t0, 0, l1);
+            tcg_temp_free(t0);
+            gen_load_fpr32(fp0, fs);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+            gen_set_label(l1);
+        }
         opn = "movz.s";
         break;
     case FOP(19, 16):
-        gen_load_gpr(cpu_T[0], ft);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT2, fd);
-        gen_op_float_movn_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            int l1 = gen_new_label();
+            TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+            TCGv fp0 = tcg_temp_local_new(TCG_TYPE_I32);
+
+            gen_load_gpr(t0, ft);
+            tcg_gen_brcondi_tl(TCG_COND_EQ, t0, 0, l1);
+            tcg_temp_free(t0);
+            gen_load_fpr32(fp0, fs);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+            gen_set_label(l1);
+        }
         opn = "movn.s";
         break;
     case FOP(21, 16):
         check_cop1x(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_recip_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_recip_s, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "recip.s";
         break;
     case FOP(22, 16):
         check_cop1x(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_rsqrt_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_rsqrt_s, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "rsqrt.s";
         break;
     case FOP(28, 16):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT2, fd);
-        gen_op_float_recip2_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32(fp1, fd);
+            tcg_gen_helper_1_2(do_float_recip2_s, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "recip2.s";
         break;
     case FOP(29, 16):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_recip1_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_recip1_s, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "recip1.s";
         break;
     case FOP(30, 16):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_rsqrt1_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_rsqrt1_s, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "rsqrt1.s";
         break;
     case FOP(31, 16):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT2, ft);
-        gen_op_float_rsqrt2_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32(fp1, ft);
+            tcg_gen_helper_1_2(do_float_rsqrt2_s, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "rsqrt2.s";
         break;
     case FOP(33, 16):
         check_cp1_registers(ctx, fd);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_cvtd_s();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr32(fp32, fs);
+            tcg_gen_helper_1_1(do_float_cvtd_s, fp64, fp32);
+            tcg_temp_free(fp32);
+            gen_store_fpr64(ctx, fp64, fd);
+            tcg_temp_free(fp64);
+        }
         opn = "cvt.d.s";
         break;
     case FOP(36, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_cvtw_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_cvtw_s, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "cvt.w.s";
         break;
     case FOP(37, 16):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_cvtl_s();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr32(fp32, fs);
+            tcg_gen_helper_1_1(do_float_cvtl_s, fp64, fp32);
+            tcg_temp_free(fp32);
+            gen_store_fpr64(ctx, fp64, fd);
+            tcg_temp_free(fp64);
+        }
         opn = "cvt.l.s";
         break;
     case FOP(38, 16):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT1, fs);
-        GEN_LOAD_FREG_FTN(WT0, ft);
-        gen_op_float_cvtps_s();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp64_0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp64_1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp32_0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp32_1 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp32_0, fs);
+            gen_load_fpr32(fp32_1, ft);
+            tcg_gen_extu_i32_i64(fp64_0, fp32_0);
+            tcg_gen_extu_i32_i64(fp64_1, fp32_1);
+            tcg_temp_free(fp32_0);
+            tcg_temp_free(fp32_1);
+            tcg_gen_shli_i64(fp64_1, fp64_1, 32);
+            tcg_gen_or_i64(fp64_0, fp64_0, fp64_1);
+            tcg_temp_free(fp64_1);
+            gen_store_fpr64(ctx, fp64_0, fd);
+            tcg_temp_free(fp64_0);
+        }
         opn = "cvt.ps.s";
         break;
     case FOP(48, 16):
@@ -5860,203 +6565,351 @@ static void gen_farith (DisasContext *ctx, uint32_t op1,
     case FOP(61, 16):
     case FOP(62, 16):
     case FOP(63, 16):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        if (ctx->opcode & (1 << 6)) {
-            check_cop1x(ctx);
-            gen_cmpabs_s(func-48, cc);
-            opn = condnames_abs[func-48];
-        } else {
-            gen_cmp_s(func-48, cc);
-            opn = condnames[func-48];
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32(fp1, ft);
+            if (ctx->opcode & (1 << 6)) {
+                check_cop1x(ctx);
+                gen_cmpabs_s(func-48, fp0, fp1, cc);
+                opn = condnames_abs[func-48];
+            } else {
+                gen_cmp_s(func-48, fp0, fp1, cc);
+                opn = condnames[func-48];
+            }
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
         }
         break;
     case FOP(0, 17):
         check_cp1_registers(ctx, fs | ft | fd);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT1, ft);
-        gen_op_float_add_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            tcg_gen_helper_1_2(do_float_add_d, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "add.d";
         optype = BINOP;
         break;
     case FOP(1, 17):
         check_cp1_registers(ctx, fs | ft | fd);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT1, ft);
-        gen_op_float_sub_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            tcg_gen_helper_1_2(do_float_sub_d, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "sub.d";
         optype = BINOP;
         break;
     case FOP(2, 17):
         check_cp1_registers(ctx, fs | ft | fd);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT1, ft);
-        gen_op_float_mul_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            tcg_gen_helper_1_2(do_float_mul_d, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "mul.d";
         optype = BINOP;
         break;
     case FOP(3, 17):
         check_cp1_registers(ctx, fs | ft | fd);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT1, ft);
-        gen_op_float_div_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            tcg_gen_helper_1_2(do_float_div_d, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "div.d";
         optype = BINOP;
         break;
     case FOP(4, 17):
         check_cp1_registers(ctx, fs | fd);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_sqrt_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_sqrt_d, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "sqrt.d";
         break;
     case FOP(5, 17):
         check_cp1_registers(ctx, fs | fd);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_abs_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_abs_d, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "abs.d";
         break;
     case FOP(6, 17):
         check_cp1_registers(ctx, fs | fd);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_mov_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "mov.d";
         break;
     case FOP(7, 17):
         check_cp1_registers(ctx, fs | fd);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_chs_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_chs_d, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "neg.d";
         break;
     case FOP(8, 17):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_roundl_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_roundl_d, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "round.l.d";
         break;
     case FOP(9, 17):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_truncl_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_truncl_d, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "trunc.l.d";
         break;
     case FOP(10, 17):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_ceill_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_ceill_d, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "ceil.l.d";
         break;
     case FOP(11, 17):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_floorl_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_floorl_d, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "floor.l.d";
         break;
     case FOP(12, 17):
         check_cp1_registers(ctx, fs);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_roundw_d();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp64, fs);
+            tcg_gen_helper_1_1(do_float_roundw_d, fp32, fp64);
+            tcg_temp_free(fp64);
+            gen_store_fpr32(fp32, fd);
+            tcg_temp_free(fp32);
+        }
         opn = "round.w.d";
         break;
     case FOP(13, 17):
         check_cp1_registers(ctx, fs);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_truncw_d();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp64, fs);
+            tcg_gen_helper_1_1(do_float_truncw_d, fp32, fp64);
+            tcg_temp_free(fp64);
+            gen_store_fpr32(fp32, fd);
+            tcg_temp_free(fp32);
+        }
         opn = "trunc.w.d";
         break;
     case FOP(14, 17):
         check_cp1_registers(ctx, fs);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_ceilw_d();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp64, fs);
+            tcg_gen_helper_1_1(do_float_ceilw_d, fp32, fp64);
+            tcg_temp_free(fp64);
+            gen_store_fpr32(fp32, fd);
+            tcg_temp_free(fp32);
+        }
         opn = "ceil.w.d";
         break;
     case FOP(15, 17):
         check_cp1_registers(ctx, fs);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_floorw_d();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp64, fs);
+            tcg_gen_helper_1_1(do_float_floorw_d, fp32, fp64);
+            tcg_temp_free(fp64);
+            gen_store_fpr32(fp32, fd);
+            tcg_temp_free(fp32);
+        }
         opn = "floor.w.d";
         break;
     case FOP(17, 17):
-        gen_load_gpr(cpu_T[0], ft);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT2, fd);
-        gen_movcf_d(ctx, (ft >> 2) & 0x7, ft & 0x1);
-        GEN_STORE_FTN_FREG(fd, DT2);
+        gen_movcf_d(ctx, fs, fd, (ft >> 2) & 0x7, ft & 0x1);
         opn = "movcf.d";
         break;
     case FOP(18, 17):
-        gen_load_gpr(cpu_T[0], ft);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT2, fd);
-        gen_op_float_movz_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            int l1 = gen_new_label();
+            TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+            TCGv fp0 = tcg_temp_local_new(TCG_TYPE_I64);
+
+            gen_load_gpr(t0, ft);
+            tcg_gen_brcondi_tl(TCG_COND_NE, t0, 0, l1);
+            tcg_temp_free(t0);
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+            gen_set_label(l1);
+        }
         opn = "movz.d";
         break;
     case FOP(19, 17):
-        gen_load_gpr(cpu_T[0], ft);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT2, fd);
-        gen_op_float_movn_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            int l1 = gen_new_label();
+            TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+            TCGv fp0 = tcg_temp_local_new(TCG_TYPE_I64);
+
+            gen_load_gpr(t0, ft);
+            tcg_gen_brcondi_tl(TCG_COND_EQ, t0, 0, l1);
+            tcg_temp_free(t0);
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+            gen_set_label(l1);
+        }
         opn = "movn.d";
         break;
     case FOP(21, 17):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_recip_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_recip_d, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "recip.d";
         break;
     case FOP(22, 17):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_rsqrt_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_rsqrt_d, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "rsqrt.d";
         break;
     case FOP(28, 17):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT2, ft);
-        gen_op_float_recip2_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            tcg_gen_helper_1_2(do_float_recip2_d, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "recip2.d";
         break;
     case FOP(29, 17):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_recip1_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_recip1_d, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "recip1.d";
         break;
     case FOP(30, 17):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_rsqrt1_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_rsqrt1_d, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "rsqrt1.d";
         break;
     case FOP(31, 17):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT2, ft);
-        gen_op_float_rsqrt2_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            tcg_gen_helper_1_2(do_float_rsqrt2_d, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "rsqrt2.d";
         break;
     case FOP(48, 17):
@@ -6075,290 +6928,434 @@ static void gen_farith (DisasContext *ctx, uint32_t op1,
     case FOP(61, 17):
     case FOP(62, 17):
     case FOP(63, 17):
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT1, ft);
-        if (ctx->opcode & (1 << 6)) {
-            check_cop1x(ctx);
-            check_cp1_registers(ctx, fs | ft);
-            gen_cmpabs_d(func-48, cc);
-            opn = condnames_abs[func-48];
-        } else {
-            check_cp1_registers(ctx, fs | ft);
-            gen_cmp_d(func-48, cc);
-            opn = condnames[func-48];
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            if (ctx->opcode & (1 << 6)) {
+                check_cop1x(ctx);
+                check_cp1_registers(ctx, fs | ft);
+                gen_cmpabs_d(func-48, fp0, fp1, cc);
+                opn = condnames_abs[func-48];
+            } else {
+                check_cp1_registers(ctx, fs | ft);
+                gen_cmp_d(func-48, fp0, fp1, cc);
+                opn = condnames[func-48];
+            }
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
         }
         break;
     case FOP(32, 17):
         check_cp1_registers(ctx, fs);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_cvts_d();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp64, fs);
+            tcg_gen_helper_1_1(do_float_cvts_d, fp32, fp64);
+            tcg_temp_free(fp64);
+            gen_store_fpr32(fp32, fd);
+            tcg_temp_free(fp32);
+        }
         opn = "cvt.s.d";
         break;
     case FOP(36, 17):
         check_cp1_registers(ctx, fs);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_cvtw_d();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp64, fs);
+            tcg_gen_helper_1_1(do_float_cvtw_d, fp32, fp64);
+            tcg_temp_free(fp64);
+            gen_store_fpr32(fp32, fd);
+            tcg_temp_free(fp32);
+        }
         opn = "cvt.w.d";
         break;
     case FOP(37, 17):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_cvtl_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_cvtl_d, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "cvt.l.d";
         break;
     case FOP(32, 20):
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_cvts_w();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_cvts_w, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "cvt.s.w";
         break;
     case FOP(33, 20):
         check_cp1_registers(ctx, fd);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_cvtd_w();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr32(fp32, fs);
+            tcg_gen_helper_1_1(do_float_cvtd_w, fp64, fp32);
+            tcg_temp_free(fp32);
+            gen_store_fpr64(ctx, fp64, fd);
+            tcg_temp_free(fp64);
+        }
         opn = "cvt.d.w";
         break;
     case FOP(32, 21):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_cvts_l();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp32 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp64 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp64, fs);
+            tcg_gen_helper_1_1(do_float_cvts_l, fp32, fp64);
+            tcg_temp_free(fp64);
+            gen_store_fpr32(fp32, fd);
+            tcg_temp_free(fp32);
+        }
         opn = "cvt.s.l";
         break;
     case FOP(33, 21):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        gen_op_float_cvtd_l();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_cvtd_l, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "cvt.d.l";
         break;
     case FOP(38, 20):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        gen_op_float_cvtps_pw();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_cvtps_pw, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "cvt.ps.pw";
         break;
     case FOP(0, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        GEN_LOAD_FREG_FTN(WTH1, ft);
-        gen_op_float_add_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            tcg_gen_helper_1_2(do_float_add_ps, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "add.ps";
         break;
     case FOP(1, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        GEN_LOAD_FREG_FTN(WTH1, ft);
-        gen_op_float_sub_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            tcg_gen_helper_1_2(do_float_sub_ps, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "sub.ps";
         break;
     case FOP(2, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        GEN_LOAD_FREG_FTN(WTH1, ft);
-        gen_op_float_mul_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            tcg_gen_helper_1_2(do_float_mul_ps, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "mul.ps";
         break;
     case FOP(5, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        gen_op_float_abs_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_abs_ps, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "abs.ps";
         break;
     case FOP(6, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        gen_op_float_mov_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "mov.ps";
         break;
     case FOP(7, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        gen_op_float_chs_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_chs_ps, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "neg.ps";
         break;
     case FOP(17, 22):
         check_cp1_64bitmode(ctx);
-        gen_load_gpr(cpu_T[0], ft);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT2, fd);
-        GEN_LOAD_FREG_FTN(WTH2, fd);
-        if (ft & 0x1)
-            gen_op_float_movt_ps ((ft >> 2) & 0x7);
-        else
-            gen_op_float_movf_ps ((ft >> 2) & 0x7);
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        gen_movcf_ps(fs, fd, (ft >> 2) & 0x7, ft & 0x1);
         opn = "movcf.ps";
         break;
     case FOP(18, 22):
         check_cp1_64bitmode(ctx);
-        gen_load_gpr(cpu_T[0], ft);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT2, fd);
-        GEN_LOAD_FREG_FTN(WTH2, fd);
-        gen_op_float_movz_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            int l1 = gen_new_label();
+            TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+            TCGv fp0 = tcg_temp_local_new(TCG_TYPE_I32);
+            TCGv fph0 = tcg_temp_local_new(TCG_TYPE_I32);
+
+            gen_load_gpr(t0, ft);
+            tcg_gen_brcondi_tl(TCG_COND_NE, t0, 0, l1);
+            tcg_temp_free(t0);
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32h(fph0, fs);
+            gen_store_fpr32(fp0, fd);
+            gen_store_fpr32h(fph0, fd);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fph0);
+            gen_set_label(l1);
+        }
         opn = "movz.ps";
         break;
     case FOP(19, 22):
         check_cp1_64bitmode(ctx);
-        gen_load_gpr(cpu_T[0], ft);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT2, fd);
-        GEN_LOAD_FREG_FTN(WTH2, fd);
-        gen_op_float_movn_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            int l1 = gen_new_label();
+            TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+            TCGv fp0 = tcg_temp_local_new(TCG_TYPE_I32);
+            TCGv fph0 = tcg_temp_local_new(TCG_TYPE_I32);
+
+            gen_load_gpr(t0, ft);
+            tcg_gen_brcondi_tl(TCG_COND_EQ, t0, 0, l1);
+            tcg_temp_free(t0);
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32h(fph0, fs);
+            gen_store_fpr32(fp0, fd);
+            gen_store_fpr32h(fph0, fd);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fph0);
+            gen_set_label(l1);
+        }
         opn = "movn.ps";
         break;
     case FOP(24, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, ft);
-        GEN_LOAD_FREG_FTN(WTH0, ft);
-        GEN_LOAD_FREG_FTN(WT1, fs);
-        GEN_LOAD_FREG_FTN(WTH1, fs);
-        gen_op_float_addr_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, ft);
+            gen_load_fpr64(ctx, fp1, fs);
+            tcg_gen_helper_1_2(do_float_addr_ps, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "addr.ps";
         break;
     case FOP(26, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, ft);
-        GEN_LOAD_FREG_FTN(WTH0, ft);
-        GEN_LOAD_FREG_FTN(WT1, fs);
-        GEN_LOAD_FREG_FTN(WTH1, fs);
-        gen_op_float_mulr_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, ft);
+            gen_load_fpr64(ctx, fp1, fs);
+            tcg_gen_helper_1_2(do_float_mulr_ps, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "mulr.ps";
         break;
     case FOP(28, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT2, fd);
-        GEN_LOAD_FREG_FTN(WTH2, fd);
-        gen_op_float_recip2_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, fd);
+            tcg_gen_helper_1_2(do_float_recip2_ps, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "recip2.ps";
         break;
     case FOP(29, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        gen_op_float_recip1_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_recip1_ps, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "recip1.ps";
         break;
     case FOP(30, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        gen_op_float_rsqrt1_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_rsqrt1_ps, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "rsqrt1.ps";
         break;
     case FOP(31, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT2, ft);
-        GEN_LOAD_FREG_FTN(WTH2, ft);
-        gen_op_float_rsqrt2_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            tcg_gen_helper_1_2(do_float_rsqrt2_ps, fp0, fp0, fp1);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "rsqrt2.ps";
         break;
     case FOP(32, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        gen_op_float_cvts_pu();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32h(fp0, fs);
+            tcg_gen_helper_1_1(do_float_cvts_pu, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "cvt.s.pu";
         break;
     case FOP(36, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        gen_op_float_cvtpw_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_helper_1_1(do_float_cvtpw_ps, fp0, fp0);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "cvt.pw.ps";
         break;
     case FOP(40, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        gen_op_float_cvts_pl();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_helper_1_1(do_float_cvts_pl, fp0, fp0);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "cvt.s.pl";
         break;
     case FOP(44, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        gen_op_float_pll_ps();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32(fp1, ft);
+            gen_store_fpr32h(fp0, fd);
+            gen_store_fpr32(fp1, fd);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+        }
         opn = "pll.ps";
         break;
     case FOP(45, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH1, ft);
-        gen_op_float_plu_ps();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32h(fp1, ft);
+            gen_store_fpr32(fp1, fd);
+            gen_store_fpr32h(fp0, fd);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+        }
         opn = "plu.ps";
         break;
     case FOP(46, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        gen_op_float_pul_ps();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32h(fp0, fs);
+            gen_load_fpr32(fp1, ft);
+            gen_store_fpr32(fp1, fd);
+            gen_store_fpr32h(fp0, fd);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+        }
         opn = "pul.ps";
         break;
     case FOP(47, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WTH1, ft);
-        gen_op_float_puu_ps();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32h(fp0, fs);
+            gen_load_fpr32h(fp1, ft);
+            gen_store_fpr32(fp1, fd);
+            gen_store_fpr32h(fp0, fd);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+        }
         opn = "puu.ps";
         break;
     case FOP(48, 22):
@@ -6378,16 +7375,21 @@ static void gen_farith (DisasContext *ctx, uint32_t op1,
     case FOP(62, 22):
     case FOP(63, 22):
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        GEN_LOAD_FREG_FTN(WTH1, ft);
-        if (ctx->opcode & (1 << 6)) {
-            gen_cmpabs_ps(func-48, cc);
-            opn = condnames_abs[func-48];
-        } else {
-            gen_cmp_ps(func-48, cc);
-            opn = condnames[func-48];
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            if (ctx->opcode & (1 << 6)) {
+                gen_cmpabs_ps(func-48, fp0, fp1, cc);
+                opn = condnames_abs[func-48];
+            } else {
+                gen_cmp_ps(func-48, fp0, fp1, cc);
+                opn = condnames[func-48];
+            }
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
         }
         break;
     default:
@@ -6414,65 +7416,103 @@ static void gen_flt3_ldst (DisasContext *ctx, uint32_t opc,
 {
     const char *opn = "extended float load/store";
     int store = 0;
+    TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+    TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
 
     if (base == 0) {
-        gen_load_gpr(cpu_T[0], index);
+        gen_load_gpr(t0, index);
     } else if (index == 0) {
-        gen_load_gpr(cpu_T[0], base);
+        gen_load_gpr(t0, base);
     } else {
-        gen_load_gpr(cpu_T[0], base);
-        gen_load_gpr(cpu_T[1], index);
-        gen_op_addr_add();
+        gen_load_gpr(t0, base);
+        gen_load_gpr(t1, index);
+        gen_op_addr_add(t0, t1);
     }
     /* Don't do NOP if destination is zero: we must perform the actual
        memory access. */
     switch (opc) {
     case OPC_LWXC1:
         check_cop1x(ctx);
-        op_ldst_lwc1(ctx);
-        GEN_STORE_FTN_FREG(fd, WT0);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            tcg_gen_qemu_ld32s(fp0, t0, ctx->mem_idx);
+            gen_store_fpr32(fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "lwxc1";
         break;
     case OPC_LDXC1:
         check_cop1x(ctx);
         check_cp1_registers(ctx, fd);
-        op_ldst_ldc1(ctx);
-        GEN_STORE_FTN_FREG(fd, DT0);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            tcg_gen_qemu_ld64(fp0, t0, ctx->mem_idx);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "ldxc1";
         break;
     case OPC_LUXC1:
         check_cp1_64bitmode(ctx);
-        op_ldst(luxc1);
-        GEN_STORE_FTN_FREG(fd, DT0);
+        tcg_gen_andi_tl(t0, t0, ~0x7);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            tcg_gen_qemu_ld64(fp0, t0, ctx->mem_idx);
+            gen_store_fpr64(ctx, fp0, fd);
+            tcg_temp_free(fp0);
+        }
         opn = "luxc1";
         break;
     case OPC_SWXC1:
         check_cop1x(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        op_ldst_swc1(ctx);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            tcg_gen_qemu_st32(fp0, t0, ctx->mem_idx);
+            tcg_temp_free(fp0);
+        }
         opn = "swxc1";
         store = 1;
         break;
     case OPC_SDXC1:
         check_cop1x(ctx);
         check_cp1_registers(ctx, fs);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        op_ldst_sdc1(ctx);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_qemu_st64(fp0, t0, ctx->mem_idx);
+            tcg_temp_free(fp0);
+        }
         opn = "sdxc1";
         store = 1;
         break;
     case OPC_SUXC1:
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        op_ldst(suxc1);
+        tcg_gen_andi_tl(t0, t0, ~0x7);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            tcg_gen_qemu_st64(fp0, t0, ctx->mem_idx);
+            tcg_temp_free(fp0);
+        }
         opn = "suxc1";
         store = 1;
         break;
     default:
         MIPS_INVAL(opn);
         generate_exception(ctx, EXCP_RI);
+        tcg_temp_free(t0);
+        tcg_temp_free(t1);
         return;
     }
+    tcg_temp_free(t0);
+    tcg_temp_free(t1);
     MIPS_DEBUG("%s %s, %s(%s)", opn, fregnames[store ? fs : fd],
                regnames[index], regnames[base]);
 }
@@ -6485,139 +7525,262 @@ static void gen_flt3_arith (DisasContext *ctx, uint32_t opc,
     switch (opc) {
     case OPC_ALNV_PS:
         check_cp1_64bitmode(ctx);
-        gen_load_gpr(cpu_T[0], fr);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT1, ft);
-        gen_op_float_alnv_ps();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+            TCGv fp0 = tcg_temp_local_new(TCG_TYPE_I32);
+            TCGv fph0 = tcg_temp_local_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_local_new(TCG_TYPE_I32);
+            TCGv fph1 = tcg_temp_local_new(TCG_TYPE_I32);
+            int l1 = gen_new_label();
+            int l2 = gen_new_label();
+
+            gen_load_gpr(t0, fr);
+            tcg_gen_andi_tl(t0, t0, 0x7);
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32h(fph0, fs);
+            gen_load_fpr32(fp1, ft);
+            gen_load_fpr32h(fph1, ft);
+
+            tcg_gen_brcondi_tl(TCG_COND_NE, t0, 0, l1);
+            gen_store_fpr32(fp0, fd);
+            gen_store_fpr32h(fph0, fd);
+            tcg_gen_br(l2);
+            gen_set_label(l1);
+            tcg_gen_brcondi_tl(TCG_COND_NE, t0, 4, l2);
+            tcg_temp_free(t0);
+#ifdef TARGET_WORDS_BIGENDIAN
+            gen_store_fpr32(fph1, fd);
+            gen_store_fpr32h(fp0, fd);
+#else
+            gen_store_fpr32(fph0, fd);
+            gen_store_fpr32h(fp1, fd);
+#endif
+            gen_set_label(l2);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fph0);
+            tcg_temp_free(fp1);
+            tcg_temp_free(fph1);
+        }
         opn = "alnv.ps";
         break;
     case OPC_MADD_S:
         check_cop1x(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        GEN_LOAD_FREG_FTN(WT2, fr);
-        gen_op_float_muladd_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp2 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32(fp1, ft);
+            gen_load_fpr32(fp2, fr);
+            tcg_gen_helper_1_3(do_float_muladd_s, fp2, fp0, fp1, fp2);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+            gen_store_fpr32(fp2, fd);
+            tcg_temp_free(fp2);
+        }
         opn = "madd.s";
         break;
     case OPC_MADD_D:
         check_cop1x(ctx);
         check_cp1_registers(ctx, fd | fs | ft | fr);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT1, ft);
-        GEN_LOAD_FREG_FTN(DT2, fr);
-        gen_op_float_muladd_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp2 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            gen_load_fpr64(ctx, fp2, fr);
+            tcg_gen_helper_1_3(do_float_muladd_d, fp2, fp0, fp1, fp2);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp2, fd);
+            tcg_temp_free(fp2);
+        }
         opn = "madd.d";
         break;
     case OPC_MADD_PS:
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        GEN_LOAD_FREG_FTN(WTH1, ft);
-        GEN_LOAD_FREG_FTN(WT2, fr);
-        GEN_LOAD_FREG_FTN(WTH2, fr);
-        gen_op_float_muladd_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp2 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            gen_load_fpr64(ctx, fp2, fr);
+            tcg_gen_helper_1_3(do_float_muladd_ps, fp2, fp0, fp1, fp2);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp2, fd);
+            tcg_temp_free(fp2);
+        }
         opn = "madd.ps";
         break;
     case OPC_MSUB_S:
         check_cop1x(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        GEN_LOAD_FREG_FTN(WT2, fr);
-        gen_op_float_mulsub_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp2 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32(fp1, ft);
+            gen_load_fpr32(fp2, fr);
+            tcg_gen_helper_1_3(do_float_mulsub_s, fp2, fp0, fp1, fp2);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+            gen_store_fpr32(fp2, fd);
+            tcg_temp_free(fp2);
+        }
         opn = "msub.s";
         break;
     case OPC_MSUB_D:
         check_cop1x(ctx);
         check_cp1_registers(ctx, fd | fs | ft | fr);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT1, ft);
-        GEN_LOAD_FREG_FTN(DT2, fr);
-        gen_op_float_mulsub_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp2 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            gen_load_fpr64(ctx, fp2, fr);
+            tcg_gen_helper_1_3(do_float_mulsub_d, fp2, fp0, fp1, fp2);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp2, fd);
+            tcg_temp_free(fp2);
+        }
         opn = "msub.d";
         break;
     case OPC_MSUB_PS:
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        GEN_LOAD_FREG_FTN(WTH1, ft);
-        GEN_LOAD_FREG_FTN(WT2, fr);
-        GEN_LOAD_FREG_FTN(WTH2, fr);
-        gen_op_float_mulsub_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp2 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            gen_load_fpr64(ctx, fp2, fr);
+            tcg_gen_helper_1_3(do_float_mulsub_ps, fp2, fp0, fp1, fp2);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp2, fd);
+            tcg_temp_free(fp2);
+        }
         opn = "msub.ps";
         break;
     case OPC_NMADD_S:
         check_cop1x(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        GEN_LOAD_FREG_FTN(WT2, fr);
-        gen_op_float_nmuladd_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp2 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32(fp1, ft);
+            gen_load_fpr32(fp2, fr);
+            tcg_gen_helper_1_3(do_float_nmuladd_s, fp2, fp0, fp1, fp2);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+            gen_store_fpr32(fp2, fd);
+            tcg_temp_free(fp2);
+        }
         opn = "nmadd.s";
         break;
     case OPC_NMADD_D:
         check_cop1x(ctx);
         check_cp1_registers(ctx, fd | fs | ft | fr);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT1, ft);
-        GEN_LOAD_FREG_FTN(DT2, fr);
-        gen_op_float_nmuladd_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp2 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            gen_load_fpr64(ctx, fp2, fr);
+            tcg_gen_helper_1_3(do_float_nmuladd_d, fp2, fp0, fp1, fp2);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp2, fd);
+            tcg_temp_free(fp2);
+        }
         opn = "nmadd.d";
         break;
     case OPC_NMADD_PS:
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        GEN_LOAD_FREG_FTN(WTH1, ft);
-        GEN_LOAD_FREG_FTN(WT2, fr);
-        GEN_LOAD_FREG_FTN(WTH2, fr);
-        gen_op_float_nmuladd_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp2 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            gen_load_fpr64(ctx, fp2, fr);
+            tcg_gen_helper_1_3(do_float_nmuladd_ps, fp2, fp0, fp1, fp2);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp2, fd);
+            tcg_temp_free(fp2);
+        }
         opn = "nmadd.ps";
         break;
     case OPC_NMSUB_S:
         check_cop1x(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        GEN_LOAD_FREG_FTN(WT2, fr);
-        gen_op_float_nmulsub_s();
-        GEN_STORE_FTN_FREG(fd, WT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I32);
+            TCGv fp2 = tcg_temp_new(TCG_TYPE_I32);
+
+            gen_load_fpr32(fp0, fs);
+            gen_load_fpr32(fp1, ft);
+            gen_load_fpr32(fp2, fr);
+            tcg_gen_helper_1_3(do_float_nmulsub_s, fp2, fp0, fp1, fp2);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+            gen_store_fpr32(fp2, fd);
+            tcg_temp_free(fp2);
+        }
         opn = "nmsub.s";
         break;
     case OPC_NMSUB_D:
         check_cop1x(ctx);
         check_cp1_registers(ctx, fd | fs | ft | fr);
-        GEN_LOAD_FREG_FTN(DT0, fs);
-        GEN_LOAD_FREG_FTN(DT1, ft);
-        GEN_LOAD_FREG_FTN(DT2, fr);
-        gen_op_float_nmulsub_d();
-        GEN_STORE_FTN_FREG(fd, DT2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp2 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            gen_load_fpr64(ctx, fp2, fr);
+            tcg_gen_helper_1_3(do_float_nmulsub_d, fp2, fp0, fp1, fp2);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp2, fd);
+            tcg_temp_free(fp2);
+        }
         opn = "nmsub.d";
         break;
     case OPC_NMSUB_PS:
         check_cp1_64bitmode(ctx);
-        GEN_LOAD_FREG_FTN(WT0, fs);
-        GEN_LOAD_FREG_FTN(WTH0, fs);
-        GEN_LOAD_FREG_FTN(WT1, ft);
-        GEN_LOAD_FREG_FTN(WTH1, ft);
-        GEN_LOAD_FREG_FTN(WT2, fr);
-        GEN_LOAD_FREG_FTN(WTH2, fr);
-        gen_op_float_nmulsub_ps();
-        GEN_STORE_FTN_FREG(fd, WT2);
-        GEN_STORE_FTN_FREG(fd, WTH2);
+        {
+            TCGv fp0 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp1 = tcg_temp_new(TCG_TYPE_I64);
+            TCGv fp2 = tcg_temp_new(TCG_TYPE_I64);
+
+            gen_load_fpr64(ctx, fp0, fs);
+            gen_load_fpr64(ctx, fp1, ft);
+            gen_load_fpr64(ctx, fp2, fr);
+            tcg_gen_helper_1_3(do_float_nmulsub_ps, fp2, fp0, fp1, fp2);
+            tcg_temp_free(fp0);
+            tcg_temp_free(fp1);
+            gen_store_fpr64(ctx, fp2, fd);
+            tcg_temp_free(fp2);
+        }
         opn = "nmsub.ps";
         break;
     default:
@@ -6655,13 +7818,17 @@ static void decode_opc (CPUState *env, DisasContext *ctx)
 
     /* Handle blikely not taken case */
     if ((ctx->hflags & MIPS_HFLAG_BMASK) == MIPS_HFLAG_BL) {
-        TCGv r_tmp = tcg_temp_new(TCG_TYPE_TL);
         int l1 = gen_new_label();
 
         MIPS_DEBUG("blikely condition (" TARGET_FMT_lx ")", ctx->pc + 4);
-        tcg_gen_ld_tl(r_tmp, cpu_env, offsetof(CPUState, bcond));
-        tcg_gen_brcondi_tl(TCG_COND_NE, r_tmp, 0, l1);
-        gen_op_save_state(ctx->hflags & ~MIPS_HFLAG_BMASK);
+        tcg_gen_brcondi_i32(TCG_COND_NE, bcond, 0, l1);
+        {
+            TCGv r_tmp = tcg_temp_new(TCG_TYPE_I32);
+
+            tcg_gen_movi_i32(r_tmp, ctx->hflags & ~MIPS_HFLAG_BMASK);
+            tcg_gen_st_i32(r_tmp, cpu_env, offsetof(CPUState, hflags));
+            tcg_temp_free(r_tmp);
+        }
         gen_goto_tb(ctx, 1, ctx->pc + 4);
         gen_set_label(l1);
     }
@@ -6715,7 +7882,7 @@ static void decode_opc (CPUState *env, DisasContext *ctx)
             MIPS_INVAL("PMON / selsl");
             generate_exception(ctx, EXCP_RI);
 #else
-            gen_op_pmon(sa);
+            tcg_gen_helper_0_i(do_pmon, sa);
 #endif
             break;
         case OPC_SYSCALL:
@@ -6820,78 +7987,107 @@ static void decode_opc (CPUState *env, DisasContext *ctx)
         }
         break;
     case OPC_SPECIAL3:
-         op1 = MASK_SPECIAL3(ctx->opcode);
-         switch (op1) {
-         case OPC_EXT:
-         case OPC_INS:
-             check_insn(env, ctx, ISA_MIPS32R2);
-             gen_bitops(ctx, op1, rt, rs, sa, rd);
-             break;
-         case OPC_BSHFL:
-             check_insn(env, ctx, ISA_MIPS32R2);
-             op2 = MASK_BSHFL(ctx->opcode);
-             switch (op2) {
-             case OPC_WSBH:
-                 gen_load_gpr(cpu_T[1], rt);
-                 gen_op_wsbh();
-                 break;
-             case OPC_SEB:
-                 gen_load_gpr(cpu_T[1], rt);
-                 tcg_gen_ext8s_tl(cpu_T[0], cpu_T[1]);
-                 break;
-             case OPC_SEH:
-                 gen_load_gpr(cpu_T[1], rt);
-                 tcg_gen_ext16s_tl(cpu_T[0], cpu_T[1]);
-                 break;
-             default:            /* Invalid */
-                 MIPS_INVAL("bshfl");
-                 generate_exception(ctx, EXCP_RI);
-                 break;
+        op1 = MASK_SPECIAL3(ctx->opcode);
+        switch (op1) {
+        case OPC_EXT:
+        case OPC_INS:
+            check_insn(env, ctx, ISA_MIPS32R2);
+            gen_bitops(ctx, op1, rt, rs, sa, rd);
+            break;
+        case OPC_BSHFL:
+            check_insn(env, ctx, ISA_MIPS32R2);
+            op2 = MASK_BSHFL(ctx->opcode);
+            {
+                TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+                TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
+
+                switch (op2) {
+                case OPC_WSBH:
+                    gen_load_gpr(t1, rt);
+                    tcg_gen_helper_1_1(do_wsbh, t0, t1);
+                    gen_store_gpr(t0, rd);
+                    break;
+                case OPC_SEB:
+                    gen_load_gpr(t1, rt);
+                    tcg_gen_ext8s_tl(t0, t1);
+                    gen_store_gpr(t0, rd);
+                    break;
+                case OPC_SEH:
+                    gen_load_gpr(t1, rt);
+                    tcg_gen_ext16s_tl(t0, t1);
+                    gen_store_gpr(t0, rd);
+                    break;
+                default:            /* Invalid */
+                    MIPS_INVAL("bshfl");
+                    generate_exception(ctx, EXCP_RI);
+                    break;
+                }
+                tcg_temp_free(t0);
+                tcg_temp_free(t1);
             }
-            gen_store_gpr(cpu_T[0], rd);
             break;
         case OPC_RDHWR:
             check_insn(env, ctx, ISA_MIPS32R2);
-            switch (rd) {
-            case 0:
-                save_cpu_state(ctx, 1);
-                gen_op_rdhwr_cpunum();
-                break;
-            case 1:
-                save_cpu_state(ctx, 1);
-                gen_op_rdhwr_synci_step();
-                break;
-            case 2:
-                save_cpu_state(ctx, 1);
-                gen_op_rdhwr_cc();
-                break;
-            case 3:
-                save_cpu_state(ctx, 1);
-                gen_op_rdhwr_ccres();
-                break;
-            case 29:
-#if defined (CONFIG_USER_ONLY)
-                gen_op_tls_value();
-                break;
-#endif
-            default:            /* Invalid */
-                MIPS_INVAL("rdhwr");
-                generate_exception(ctx, EXCP_RI);
-                break;
+            {
+                TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+
+                switch (rd) {
+                case 0:
+                    save_cpu_state(ctx, 1);
+                    tcg_gen_helper_1_0(do_rdhwr_cpunum, t0);
+                    break;
+                case 1:
+                    save_cpu_state(ctx, 1);
+                    tcg_gen_helper_1_0(do_rdhwr_synci_step, t0);
+                    break;
+                case 2:
+                    save_cpu_state(ctx, 1);
+                    tcg_gen_helper_1_0(do_rdhwr_cc, t0);
+                    break;
+                case 3:
+                    save_cpu_state(ctx, 1);
+                    tcg_gen_helper_1_0(do_rdhwr_ccres, t0);
+                    break;
+                case 29:
+                    if (env->user_mode_only) {
+                        tcg_gen_ld_tl(t0, cpu_env, offsetof(CPUState, tls_value));
+                        break;
+                    } else {
+                        /* XXX: Some CPUs implement this in hardware.
+                           Not supported yet. */
+                    }
+                default:            /* Invalid */
+                    MIPS_INVAL("rdhwr");
+                    generate_exception(ctx, EXCP_RI);
+                    break;
+                }
+                gen_store_gpr(t0, rt);
+                tcg_temp_free(t0);
             }
-            gen_store_gpr(cpu_T[0], rt);
             break;
         case OPC_FORK:
             check_insn(env, ctx, ASE_MT);
-            gen_load_gpr(cpu_T[0], rt);
-            gen_load_gpr(cpu_T[1], rs);
-            gen_op_fork();
+            {
+                TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+                TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
+
+                gen_load_gpr(t0, rt);
+                gen_load_gpr(t1, rs);
+                tcg_gen_helper_0_2(do_fork, t0, t1);
+                tcg_temp_free(t0);
+                tcg_temp_free(t1);
+            }
             break;
         case OPC_YIELD:
             check_insn(env, ctx, ASE_MT);
-            gen_load_gpr(cpu_T[0], rs);
-            gen_op_yield();
-            gen_store_gpr(cpu_T[0], rd);
+            {
+                TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+
+                gen_load_gpr(t0, rs);
+                tcg_gen_helper_1_1(do_yield, t0, t0);
+                gen_store_gpr(t0, rd);
+                tcg_temp_free(t0);
+            }
             break;
 #if defined(TARGET_MIPS64)
         case OPC_DEXTM ... OPC_DEXT:
@@ -6904,21 +8100,28 @@ static void decode_opc (CPUState *env, DisasContext *ctx)
             check_insn(env, ctx, ISA_MIPS64R2);
             check_mips_64(ctx);
             op2 = MASK_DBSHFL(ctx->opcode);
-            switch (op2) {
-            case OPC_DSBH:
-                gen_load_gpr(cpu_T[1], rt);
-                gen_op_dsbh();
-                break;
-            case OPC_DSHD:
-                gen_load_gpr(cpu_T[1], rt);
-                gen_op_dshd();
-                break;
-            default:            /* Invalid */
-                MIPS_INVAL("dbshfl");
-                generate_exception(ctx, EXCP_RI);
-                break;
+            {
+                TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+                TCGv t1 = tcg_temp_local_new(TCG_TYPE_TL);
+
+                switch (op2) {
+                case OPC_DSBH:
+                    gen_load_gpr(t1, rt);
+                    tcg_gen_helper_1_1(do_dsbh, t0, t1);
+                    break;
+                case OPC_DSHD:
+                    gen_load_gpr(t1, rt);
+                    tcg_gen_helper_1_1(do_dshd, t0, t1);
+                    break;
+                default:            /* Invalid */
+                    MIPS_INVAL("dbshfl");
+                    generate_exception(ctx, EXCP_RI);
+                    break;
+                }
+                gen_store_gpr(t0, rd);
+                tcg_temp_free(t0);
+                tcg_temp_free(t1);
             }
-            gen_store_gpr(cpu_T[0], rd);
             break;
 #endif
         default:            /* Invalid */
@@ -6960,60 +8163,71 @@ static void decode_opc (CPUState *env, DisasContext *ctx)
         case OPC_DMFC0:
         case OPC_DMTC0:
 #endif
-            gen_cp0(env, ctx, op1, rt, rd);
+#ifndef CONFIG_USER_ONLY
+            if (!env->user_mode_only)
+                gen_cp0(env, ctx, op1, rt, rd);
+#endif /* !CONFIG_USER_ONLY */
             break;
         case OPC_C0_FIRST ... OPC_C0_LAST:
-            gen_cp0(env, ctx, MASK_C0(ctx->opcode), rt, rd);
+#ifndef CONFIG_USER_ONLY
+            if (!env->user_mode_only)
+                gen_cp0(env, ctx, MASK_C0(ctx->opcode), rt, rd);
+#endif /* !CONFIG_USER_ONLY */
             break;
         case OPC_MFMC0:
-            op2 = MASK_MFMC0(ctx->opcode);
-            switch (op2) {
-            case OPC_DMT:
-                check_insn(env, ctx, ASE_MT);
-                gen_op_dmt();
-                break;
-            case OPC_EMT:
-                check_insn(env, ctx, ASE_MT);
-                gen_op_emt();
-                break;
-            case OPC_DVPE:
-                check_insn(env, ctx, ASE_MT);
-                gen_op_dvpe();
-                break;
-            case OPC_EVPE:
-                check_insn(env, ctx, ASE_MT);
-                gen_op_evpe();
-                break;
-            case OPC_DI:
-                check_insn(env, ctx, ISA_MIPS32R2);
-                save_cpu_state(ctx, 1);
-                gen_op_di();
-                /* Stop translation as we may have switched the execution mode */
-                ctx->bstate = BS_STOP;
-                break;
-            case OPC_EI:
-                check_insn(env, ctx, ISA_MIPS32R2);
-                save_cpu_state(ctx, 1);
-                gen_op_ei();
-                /* Stop translation as we may have switched the execution mode */
-                ctx->bstate = BS_STOP;
-                break;
-            default:            /* Invalid */
-                MIPS_INVAL("mfmc0");
-                generate_exception(ctx, EXCP_RI);
-                break;
+#ifndef CONFIG_USER_ONLY
+            if (!env->user_mode_only) {
+                TCGv t0 = tcg_temp_local_new(TCG_TYPE_TL);
+
+                op2 = MASK_MFMC0(ctx->opcode);
+                switch (op2) {
+                case OPC_DMT:
+                    check_insn(env, ctx, ASE_MT);
+                    tcg_gen_helper_1_1(do_dmt, t0, t0);
+                    break;
+                case OPC_EMT:
+                    check_insn(env, ctx, ASE_MT);
+                    tcg_gen_helper_1_1(do_emt, t0, t0);
+                     break;
+                case OPC_DVPE:
+                    check_insn(env, ctx, ASE_MT);
+                    tcg_gen_helper_1_1(do_dvpe, t0, t0);
+                    break;
+                case OPC_EVPE:
+                    check_insn(env, ctx, ASE_MT);
+                    tcg_gen_helper_1_1(do_evpe, t0, t0);
+                    break;
+                case OPC_DI:
+                    check_insn(env, ctx, ISA_MIPS32R2);
+                    save_cpu_state(ctx, 1);
+                    tcg_gen_helper_1_0(do_di, t0);
+                    /* Stop translation as we may have switched the execution mode */
+                    ctx->bstate = BS_STOP;
+                    break;
+                case OPC_EI:
+                    check_insn(env, ctx, ISA_MIPS32R2);
+                    save_cpu_state(ctx, 1);
+                    tcg_gen_helper_1_0(do_ei, t0);
+                    /* Stop translation as we may have switched the execution mode */
+                    ctx->bstate = BS_STOP;
+                    break;
+                default:            /* Invalid */
+                    MIPS_INVAL("mfmc0");
+                    generate_exception(ctx, EXCP_RI);
+                    break;
+                }
+                gen_store_gpr(t0, rt);
+                tcg_temp_free(t0);
             }
-            gen_store_gpr(cpu_T[0], rt);
+#endif /* !CONFIG_USER_ONLY */
             break;
         case OPC_RDPGPR:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_load_srsgpr(cpu_T[0], rt);
-            gen_store_gpr(cpu_T[0], rd);
+            gen_load_srsgpr(rt, rd);
             break;
         case OPC_WRPGPR:
             check_insn(env, ctx, ISA_MIPS32R2);
-            gen_load_gpr(cpu_T[0], rt);
-            gen_store_srsgpr(cpu_T[0], rd);
+            gen_store_srsgpr(rt, rd);
             break;
         default:
             MIPS_INVAL("cp0");
@@ -7199,6 +8413,7 @@ static void decode_opc (CPUState *env, DisasContext *ctx)
         ctx->hflags &= ~MIPS_HFLAG_BMASK;
         ctx->bstate = BS_BRANCH;
         save_cpu_state(ctx, 0);
+        /* FIXME: Need to clear can_do_io.  */
         switch (hflags) {
         case MIPS_HFLAG_B:
             /* unconditional branch */
@@ -7214,11 +8429,9 @@ static void decode_opc (CPUState *env, DisasContext *ctx)
             /* Conditional branch */
             MIPS_DEBUG("conditional branch");
             {
-                TCGv r_tmp = tcg_temp_new(TCG_TYPE_TL);
                 int l1 = gen_new_label();
 
-                tcg_gen_ld_tl(r_tmp, cpu_env, offsetof(CPUState, bcond));
-                tcg_gen_brcondi_tl(TCG_COND_NE, r_tmp, 0, l1);
+                tcg_gen_brcondi_i32(TCG_COND_NE, bcond, 0, l1);
                 gen_goto_tb(ctx, 1, ctx->pc + 4);
                 gen_set_label(l1);
                 gen_goto_tb(ctx, 0, ctx->btarget);
@@ -7227,7 +8440,7 @@ static void decode_opc (CPUState *env, DisasContext *ctx)
         case MIPS_HFLAG_BR:
             /* unconditional branch to register */
             MIPS_DEBUG("branch to register");
-            gen_breg_pc();
+            tcg_gen_st_tl(btarget, cpu_env, offsetof(CPUState, active_tc.PC));
             tcg_gen_exit_tb(0);
             break;
         default:
@@ -7237,7 +8450,7 @@ static void decode_opc (CPUState *env, DisasContext *ctx)
     }
 }
 
-static always_inline int
+static inline void
 gen_intermediate_code_internal (CPUState *env, TranslationBlock *tb,
                                 int search_pc)
 {
@@ -7245,18 +8458,15 @@ gen_intermediate_code_internal (CPUState *env, TranslationBlock *tb,
     target_ulong pc_start;
     uint16_t *gen_opc_end;
     int j, lj = -1;
+    int num_insns;
+    int max_insns;
 
     if (search_pc && loglevel)
         fprintf (logfile, "search pc %d\n", search_pc);
 
-    num_temps = 0;
-    memset(temps, 0, sizeof(temps));
-
-    num_temps = 0;
-    memset(temps, 0, sizeof(temps));
-
     pc_start = tb->pc;
-    gen_opc_end = gen_opc_buf + OPC_MAX_SIZE;
+    /* Leave some spare opc slots for branch handling. */
+    gen_opc_end = gen_opc_buf + OPC_MAX_SIZE - 16;
     ctx.pc = pc_start;
     ctx.saved_pc = -1;
     ctx.tb = tb;
@@ -7264,11 +8474,14 @@ gen_intermediate_code_internal (CPUState *env, TranslationBlock *tb,
     /* Restore delay slot state from the tb context.  */
     ctx.hflags = (uint32_t)tb->flags; /* FIXME: maybe use 64 bits here? */
     restore_cpu_state(env, &ctx);
-#if defined(CONFIG_USER_ONLY)
-    ctx.mem_idx = MIPS_HFLAG_UM;
-#else
-    ctx.mem_idx = ctx.hflags & MIPS_HFLAG_KSU;
-#endif
+    if (env->user_mode_only)
+        ctx.mem_idx = MIPS_HFLAG_UM;
+    else
+        ctx.mem_idx = ctx.hflags & MIPS_HFLAG_KSU;
+    num_insns = 0;
+    max_insns = tb->cflags & CF_COUNT_MASK;
+    if (max_insns == 0)
+        max_insns = CF_COUNT_MASK;
 #ifdef DEBUG_DISAS
     if (loglevel & CPU_LOG_TB_CPU) {
         fprintf(logfile, "------------------------------------------------\n");
@@ -7281,13 +8494,14 @@ gen_intermediate_code_internal (CPUState *env, TranslationBlock *tb,
         fprintf(logfile, "\ntb %p idx %d hflags %04x\n",
                 tb, ctx.mem_idx, ctx.hflags);
 #endif
-    while (ctx.bstate == BS_NONE && gen_opc_ptr < gen_opc_end) {
+    gen_icount_start();
+    while (ctx.bstate == BS_NONE) {
         if (env->nb_breakpoints > 0) {
             for(j = 0; j < env->nb_breakpoints; j++) {
                 if (env->breakpoints[j] == ctx.pc) {
                     save_cpu_state(&ctx, 1);
                     ctx.bstate = BS_BRANCH;
-                    gen_op_debug();
+                    tcg_gen_helper_0_i(do_raise_exception, EXCP_DEBUG);
                     /* Include the breakpoint location or the tb won't
                      * be flushed when it must be.  */
                     ctx.pc += 4;
@@ -7306,16 +8520,14 @@ gen_intermediate_code_internal (CPUState *env, TranslationBlock *tb,
             gen_opc_pc[lj] = ctx.pc;
             gen_opc_hflags[lj] = ctx.hflags & MIPS_HFLAG_BMASK;
             gen_opc_instr_start[lj] = 1;
+            gen_opc_icount[lj] = num_insns;
         }
+        if (num_insns + 1 == max_insns && (tb->cflags & CF_LAST_IO))
+            gen_io_start();
         ctx.opcode = ldl_code(ctx.pc);
         decode_opc(env, &ctx);
-        if (num_temps) {
-            fprintf(stderr,
-                    "Internal resource leak before " TARGET_FMT_lx "\n",
-                    ctx.pc);
-            num_temps = 0;
-        }
         ctx.pc += 4;
+        num_insns++;
 
         if (env->singlestep_enabled)
             break;
@@ -7323,13 +8535,20 @@ gen_intermediate_code_internal (CPUState *env, TranslationBlock *tb,
         if ((ctx.pc & (TARGET_PAGE_SIZE - 1)) == 0)
             break;
 
+        if (gen_opc_ptr >= gen_opc_end)
+            break;
+
+        if (num_insns >= max_insns)
+            break;
 #if defined (MIPS_SINGLE_STEP)
         break;
 #endif
     }
+    if (tb->cflags & CF_LAST_IO)
+        gen_io_end();
     if (env->singlestep_enabled) {
         save_cpu_state(&ctx, ctx.bstate == BS_NONE);
-        gen_op_debug();
+        tcg_gen_helper_0_i(do_raise_exception, EXCP_DEBUG);
     } else {
        switch (ctx.bstate) {
         case BS_STOP:
@@ -7350,6 +8569,7 @@ gen_intermediate_code_internal (CPUState *env, TranslationBlock *tb,
        }
     }
 done_generating:
+    gen_icount_end(tb, num_insns);
     *gen_opc_ptr = INDEX_op_end;
     if (search_pc) {
         j = gen_opc_ptr - gen_opc_buf;
@@ -7358,6 +8578,7 @@ done_generating:
             gen_opc_instr_start[lj++] = 0;
     } else {
         tb->size = ctx.pc - pc_start;
+        tb->icount = num_insns;
     }
 #ifdef DEBUG_DISAS
 #if defined MIPS_DEBUG_DISAS
@@ -7373,23 +8594,21 @@ done_generating:
         fprintf(logfile, "---------------- %d %08x\n", ctx.bstate, ctx.hflags);
     }
 #endif
-
-    return 0;
 }
 
-int gen_intermediate_code (CPUState *env, struct TranslationBlock *tb)
+void gen_intermediate_code (CPUState *env, struct TranslationBlock *tb)
 {
-    return gen_intermediate_code_internal(env, tb, 0);
+    gen_intermediate_code_internal(env, tb, 0);
 }
 
-int gen_intermediate_code_pc (CPUState *env, struct TranslationBlock *tb)
+void gen_intermediate_code_pc (CPUState *env, struct TranslationBlock *tb)
 {
-    return gen_intermediate_code_internal(env, tb, 1);
+    gen_intermediate_code_internal(env, tb, 1);
 }
 
-void fpu_dump_state(CPUState *env, FILE *f,
-                    int (*fpu_fprintf)(FILE *f, const char *fmt, ...),
-                    int flags)
+static void fpu_dump_state(CPUState *env, FILE *f,
+                           int (*fpu_fprintf)(FILE *f, const char *fmt, ...),
+                           int flags)
 {
     int i;
     int is_fpu64 = !!(env->hflags & MIPS_HFLAG_F64);
@@ -7414,9 +8633,6 @@ void fpu_dump_state(CPUState *env, FILE *f,
     fpu_fprintf(f, "CP1 FCR0 0x%08x  FCR31 0x%08x  SR.FR %d  fp_status 0x%08x(0x%02x)\n",
                 env->fpu->fcr0, env->fpu->fcr31, is_fpu64, env->fpu->fp_status,
                 get_float_exception_flags(&env->fpu->fp_status));
-    fpu_fprintf(f, "FT0: "); printfpr(&env->fpu->ft0);
-    fpu_fprintf(f, "FT1: "); printfpr(&env->fpu->ft1);
-    fpu_fprintf(f, "FT2: "); printfpr(&env->fpu->ft2);
     for (i = 0; i < 32; (is_fpu64) ? i++ : (i += 2)) {
         fpu_fprintf(f, "%3s: ", fregnames[i]);
         printfpr(&env->fpu->fpr[i]);
@@ -7425,44 +8641,31 @@ void fpu_dump_state(CPUState *env, FILE *f,
 #undef printfpr
 }
 
-void dump_fpu (CPUState *env)
-{
-    if (loglevel) {
-        fprintf(logfile,
-                "pc=0x" TARGET_FMT_lx " HI=0x" TARGET_FMT_lx
-                " LO=0x" TARGET_FMT_lx " ds %04x " TARGET_FMT_lx
-                " %04x\n",
-                env->PC[env->current_tc], env->HI[env->current_tc][0],
-                env->LO[env->current_tc][0], env->hflags, env->btarget,
-                env->bcond);
-       fpu_dump_state(env, logfile, fprintf, 0);
-    }
-}
-
 #if defined(TARGET_MIPS64) && defined(MIPS_DEBUG_SIGN_EXTENSIONS)
 /* Debug help: The architecture requires 32bit code to maintain proper
-   sign-extened values on 64bit machines.  */
+   sign-extended values on 64bit machines.  */
 
 #define SIGN_EXT_P(val) ((((val) & ~0x7fffffff) == 0) || (((val) & ~0x7fffffff) == ~0x7fffffff))
 
-void cpu_mips_check_sign_extensions (CPUState *env, FILE *f,
-                     int (*cpu_fprintf)(FILE *f, const char *fmt, ...),
-                     int flags)
+static void
+cpu_mips_check_sign_extensions (CPUState *env, FILE *f,
+                                int (*cpu_fprintf)(FILE *f, const char *fmt, ...),
+                                int flags)
 {
     int i;
 
-    if (!SIGN_EXT_P(env->PC[env->current_tc]))
-        cpu_fprintf(f, "BROKEN: pc=0x" TARGET_FMT_lx "\n", env->PC[env->current_tc]);
-    if (!SIGN_EXT_P(env->HI[env->current_tc][0]))
-        cpu_fprintf(f, "BROKEN: HI=0x" TARGET_FMT_lx "\n", env->HI[env->current_tc][0]);
-    if (!SIGN_EXT_P(env->LO[env->current_tc][0]))
-        cpu_fprintf(f, "BROKEN: LO=0x" TARGET_FMT_lx "\n", env->LO[env->current_tc][0]);
+    if (!SIGN_EXT_P(env->active_tc.PC))
+        cpu_fprintf(f, "BROKEN: pc=0x" TARGET_FMT_lx "\n", env->active_tc.PC);
+    if (!SIGN_EXT_P(env->active_tc.HI[0]))
+        cpu_fprintf(f, "BROKEN: HI=0x" TARGET_FMT_lx "\n", env->active_tc.HI[0]);
+    if (!SIGN_EXT_P(env->active_tc.LO[0]))
+        cpu_fprintf(f, "BROKEN: LO=0x" TARGET_FMT_lx "\n", env->active_tc.LO[0]);
     if (!SIGN_EXT_P(env->btarget))
         cpu_fprintf(f, "BROKEN: btarget=0x" TARGET_FMT_lx "\n", env->btarget);
 
     for (i = 0; i < 32; i++) {
-        if (!SIGN_EXT_P(env->gpr[env->current_tc][i]))
-            cpu_fprintf(f, "BROKEN: %s=0x" TARGET_FMT_lx "\n", regnames[i], env->gpr[env->current_tc][i]);
+        if (!SIGN_EXT_P(env->active_tc.gpr[i]))
+            cpu_fprintf(f, "BROKEN: %s=0x" TARGET_FMT_lx "\n", regnames[i], env->active_tc.gpr[i]);
     }
 
     if (!SIGN_EXT_P(env->CP0_EPC))
@@ -7479,11 +8682,12 @@ void cpu_dump_state (CPUState *env, FILE *f,
     int i;
 
     cpu_fprintf(f, "pc=0x" TARGET_FMT_lx " HI=0x" TARGET_FMT_lx " LO=0x" TARGET_FMT_lx " ds %04x " TARGET_FMT_lx " %d\n",
-                env->PC[env->current_tc], env->HI[env->current_tc], env->LO[env->current_tc], env->hflags, env->btarget, env->bcond);
+                env->active_tc.PC, env->active_tc.HI[0], env->active_tc.LO[0],
+                env->hflags, env->btarget, env->bcond);
     for (i = 0; i < 32; i++) {
         if ((i & 3) == 0)
             cpu_fprintf(f, "GPR%02d:", i);
-        cpu_fprintf(f, " %s " TARGET_FMT_lx, regnames[i], env->gpr[env->current_tc][i]);
+        cpu_fprintf(f, " %s " TARGET_FMT_lx, regnames[i], env->active_tc.gpr[i]);
         if ((i & 3) == 3)
             cpu_fprintf(f, "\n");
     }
@@ -7508,19 +8712,19 @@ static void mips_tcg_init(void)
        return;
 
     cpu_env = tcg_global_reg_new(TCG_TYPE_PTR, TCG_AREG0, "env");
-    current_tc_gprs = tcg_global_mem_new(TCG_TYPE_PTR,
-                                         TCG_AREG0,
-                                         offsetof(CPUState, current_tc_gprs),
-                                         "current_tc_gprs");
-#if TARGET_LONG_BITS > HOST_LONG_BITS
-    cpu_T[0] = tcg_global_mem_new(TCG_TYPE_TL,
-                                  TCG_AREG0, offsetof(CPUState, t0), "T0");
-    cpu_T[1] = tcg_global_mem_new(TCG_TYPE_TL,
-                                  TCG_AREG0, offsetof(CPUState, t1), "T1");
-#else
-    cpu_T[0] = tcg_global_reg_new(TCG_TYPE_TL, TCG_AREG1, "T0");
-    cpu_T[1] = tcg_global_reg_new(TCG_TYPE_TL, TCG_AREG2, "T1");
-#endif
+    bcond = tcg_global_mem_new(TCG_TYPE_I32, TCG_AREG0,
+                               offsetof(CPUState, bcond), "bcond");
+    btarget = tcg_global_mem_new(TCG_TYPE_TL, TCG_AREG0,
+                                 offsetof(CPUState, btarget), "btarget");
+    current_fpu = tcg_global_mem_new(TCG_TYPE_PTR,
+                                     TCG_AREG0,
+                                     offsetof(CPUState, fpu),
+                                     "current_fpu");
+
+    /* register helpers */
+#undef DEF_HELPER
+#define DEF_HELPER(ret, name, params) tcg_register_helper(name, #name);
+#include "helper.h"
 
     inited = 1;
 }
@@ -7554,49 +8758,49 @@ void cpu_reset (CPUMIPSState *env)
     tlb_flush(env, 1);
 
     /* Minimal init */
-#if !defined(CONFIG_USER_ONLY)
-    if (env->hflags & MIPS_HFLAG_BMASK) {
-        /* If the exception was raised from a delay slot,
-         * come back to the jump.  */
-        env->CP0_ErrorEPC = env->PC[env->current_tc] - 4;
+#if defined(CONFIG_USER_ONLY)
+    env->user_mode_only = 1;
+#endif
+    if (env->user_mode_only) {
+        env->hflags = MIPS_HFLAG_UM;
     } else {
-        env->CP0_ErrorEPC = env->PC[env->current_tc];
-    }
-    env->PC[env->current_tc] = (int32_t)0xBFC00000;
-    env->CP0_Wired = 0;
-    /* SMP not implemented */
-    env->CP0_EBase = 0x80000000;
-    env->CP0_Status = (1 << CP0St_BEV) | (1 << CP0St_ERL);
-    /* vectored interrupts not implemented, timer on int 7,
-       no performance counters. */
-    env->CP0_IntCtl = 0xe0000000;
-    {
-        int i;
+        if (env->hflags & MIPS_HFLAG_BMASK) {
+            /* If the exception was raised from a delay slot,
+               come back to the jump.  */
+            env->CP0_ErrorEPC = env->active_tc.PC - 4;
+        } else {
+            env->CP0_ErrorEPC = env->active_tc.PC;
+        }
+        env->active_tc.PC = (int32_t)0xBFC00000;
+        env->CP0_Wired = 0;
+        /* SMP not implemented */
+        env->CP0_EBase = 0x80000000;
+        env->CP0_Status = (1 << CP0St_BEV) | (1 << CP0St_ERL);
+        /* vectored interrupts not implemented, timer on int 7,
+           no performance counters. */
+        env->CP0_IntCtl = 0xe0000000;
+        {
+            int i;
 
-        for (i = 0; i < 7; i++) {
-            env->CP0_WatchLo[i] = 0;
-            env->CP0_WatchHi[i] = 0x80000000;
+            for (i = 0; i < 7; i++) {
+                env->CP0_WatchLo[i] = 0;
+                env->CP0_WatchHi[i] = 0x80000000;
+            }
+            env->CP0_WatchLo[7] = 0;
+            env->CP0_WatchHi[7] = 0;
         }
-        env->CP0_WatchLo[7] = 0;
-        env->CP0_WatchHi[7] = 0;
+        /* Count register increments in debug mode, EJTAG version 1 */
+        env->CP0_Debug = (1 << CP0DB_CNT) | (0x1 << CP0DB_VER);
+        env->hflags = MIPS_HFLAG_CP0;
     }
-    /* Count register increments in debug mode, EJTAG version 1 */
-    env->CP0_Debug = (1 << CP0DB_CNT) | (0x1 << CP0DB_VER);
-#endif
     env->exception_index = EXCP_NONE;
-#if defined(CONFIG_USER_ONLY)
-    env->hflags = MIPS_HFLAG_UM;
-    env->user_mode_only = 1;
-#else
-    env->hflags = MIPS_HFLAG_CP0;
-#endif
     cpu_mips_register(env, env->cpu_model);
 }
 
 void gen_pc_load(CPUState *env, TranslationBlock *tb,
                 unsigned long searched_pc, int pc_pos, void *puc)
 {
-    env->PC[env->current_tc] = gen_opc_pc[pc_pos];
+    env->active_tc.PC = gen_opc_pc[pc_pos];
     env->hflags &= ~MIPS_HFLAG_BMASK;
     env->hflags |= gen_opc_hflags[pc_pos];
 }