Add qemu 2.4.0
[kvmfornfv.git] / qemu / target-mips / op_helper.c
diff --git a/qemu/target-mips/op_helper.c b/qemu/target-mips/op_helper.c
new file mode 100644 (file)
index 0000000..db4f6b9
--- /dev/null
@@ -0,0 +1,3706 @@
+/*
+ *  MIPS emulation helpers for qemu.
+ *
+ *  Copyright (c) 2004-2005 Jocelyn Mayer
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see <http://www.gnu.org/licenses/>.
+ */
+#include <stdlib.h>
+#include "cpu.h"
+#include "qemu/host-utils.h"
+#include "exec/helper-proto.h"
+#include "exec/cpu_ldst.h"
+#include "sysemu/kvm.h"
+
+#ifndef CONFIG_USER_ONLY
+static inline void cpu_mips_tlb_flush (CPUMIPSState *env, int flush_global);
+#endif
+
+/*****************************************************************************/
+/* Exceptions processing helpers */
+
+static inline void QEMU_NORETURN do_raise_exception_err(CPUMIPSState *env,
+                                                        uint32_t exception,
+                                                        int error_code,
+                                                        uintptr_t pc)
+{
+    CPUState *cs = CPU(mips_env_get_cpu(env));
+
+    if (exception < EXCP_SC) {
+        qemu_log("%s: %d %d\n", __func__, exception, error_code);
+    }
+    cs->exception_index = exception;
+    env->error_code = error_code;
+
+    if (pc) {
+        /* now we have a real cpu fault */
+        cpu_restore_state(cs, pc);
+    }
+
+    cpu_loop_exit(cs);
+}
+
+static inline void QEMU_NORETURN do_raise_exception(CPUMIPSState *env,
+                                                    uint32_t exception,
+                                                    uintptr_t pc)
+{
+    do_raise_exception_err(env, exception, 0, pc);
+}
+
+void helper_raise_exception_err(CPUMIPSState *env, uint32_t exception,
+                                int error_code)
+{
+    do_raise_exception_err(env, exception, error_code, 0);
+}
+
+void helper_raise_exception(CPUMIPSState *env, uint32_t exception)
+{
+    do_raise_exception(env, exception, 0);
+}
+
+#if defined(CONFIG_USER_ONLY)
+#define HELPER_LD(name, insn, type)                                     \
+static inline type do_##name(CPUMIPSState *env, target_ulong addr,      \
+                             int mem_idx)                               \
+{                                                                       \
+    return (type) cpu_##insn##_data(env, addr);                         \
+}
+#else
+#define HELPER_LD(name, insn, type)                                     \
+static inline type do_##name(CPUMIPSState *env, target_ulong addr,      \
+                             int mem_idx)                               \
+{                                                                       \
+    switch (mem_idx)                                                    \
+    {                                                                   \
+    case 0: return (type) cpu_##insn##_kernel(env, addr); break;        \
+    case 1: return (type) cpu_##insn##_super(env, addr); break;         \
+    default:                                                            \
+    case 2: return (type) cpu_##insn##_user(env, addr); break;          \
+    }                                                                   \
+}
+#endif
+HELPER_LD(lw, ldl, int32_t)
+#if defined(TARGET_MIPS64)
+HELPER_LD(ld, ldq, int64_t)
+#endif
+#undef HELPER_LD
+
+#if defined(CONFIG_USER_ONLY)
+#define HELPER_ST(name, insn, type)                                     \
+static inline void do_##name(CPUMIPSState *env, target_ulong addr,      \
+                             type val, int mem_idx)                     \
+{                                                                       \
+    cpu_##insn##_data(env, addr, val);                                  \
+}
+#else
+#define HELPER_ST(name, insn, type)                                     \
+static inline void do_##name(CPUMIPSState *env, target_ulong addr,      \
+                             type val, int mem_idx)                     \
+{                                                                       \
+    switch (mem_idx)                                                    \
+    {                                                                   \
+    case 0: cpu_##insn##_kernel(env, addr, val); break;                 \
+    case 1: cpu_##insn##_super(env, addr, val); break;                  \
+    default:                                                            \
+    case 2: cpu_##insn##_user(env, addr, val); break;                   \
+    }                                                                   \
+}
+#endif
+HELPER_ST(sb, stb, uint8_t)
+HELPER_ST(sw, stl, uint32_t)
+#if defined(TARGET_MIPS64)
+HELPER_ST(sd, stq, uint64_t)
+#endif
+#undef HELPER_ST
+
+target_ulong helper_clo (target_ulong arg1)
+{
+    return clo32(arg1);
+}
+
+target_ulong helper_clz (target_ulong arg1)
+{
+    return clz32(arg1);
+}
+
+#if defined(TARGET_MIPS64)
+target_ulong helper_dclo (target_ulong arg1)
+{
+    return clo64(arg1);
+}
+
+target_ulong helper_dclz (target_ulong arg1)
+{
+    return clz64(arg1);
+}
+#endif /* TARGET_MIPS64 */
+
+/* 64 bits arithmetic for 32 bits hosts */
+static inline uint64_t get_HILO(CPUMIPSState *env)
+{
+    return ((uint64_t)(env->active_tc.HI[0]) << 32) | (uint32_t)env->active_tc.LO[0];
+}
+
+static inline target_ulong set_HIT0_LO(CPUMIPSState *env, uint64_t HILO)
+{
+    target_ulong tmp;
+    env->active_tc.LO[0] = (int32_t)(HILO & 0xFFFFFFFF);
+    tmp = env->active_tc.HI[0] = (int32_t)(HILO >> 32);
+    return tmp;
+}
+
+static inline target_ulong set_HI_LOT0(CPUMIPSState *env, uint64_t HILO)
+{
+    target_ulong tmp = env->active_tc.LO[0] = (int32_t)(HILO & 0xFFFFFFFF);
+    env->active_tc.HI[0] = (int32_t)(HILO >> 32);
+    return tmp;
+}
+
+/* Multiplication variants of the vr54xx. */
+target_ulong helper_muls(CPUMIPSState *env, target_ulong arg1,
+                         target_ulong arg2)
+{
+    return set_HI_LOT0(env, 0 - ((int64_t)(int32_t)arg1 *
+                                 (int64_t)(int32_t)arg2));
+}
+
+target_ulong helper_mulsu(CPUMIPSState *env, target_ulong arg1,
+                          target_ulong arg2)
+{
+    return set_HI_LOT0(env, 0 - (uint64_t)(uint32_t)arg1 *
+                       (uint64_t)(uint32_t)arg2);
+}
+
+target_ulong helper_macc(CPUMIPSState *env, target_ulong arg1,
+                         target_ulong arg2)
+{
+    return set_HI_LOT0(env, (int64_t)get_HILO(env) + (int64_t)(int32_t)arg1 *
+                       (int64_t)(int32_t)arg2);
+}
+
+target_ulong helper_macchi(CPUMIPSState *env, target_ulong arg1,
+                           target_ulong arg2)
+{
+    return set_HIT0_LO(env, (int64_t)get_HILO(env) + (int64_t)(int32_t)arg1 *
+                       (int64_t)(int32_t)arg2);
+}
+
+target_ulong helper_maccu(CPUMIPSState *env, target_ulong arg1,
+                          target_ulong arg2)
+{
+    return set_HI_LOT0(env, (uint64_t)get_HILO(env) +
+                       (uint64_t)(uint32_t)arg1 * (uint64_t)(uint32_t)arg2);
+}
+
+target_ulong helper_macchiu(CPUMIPSState *env, target_ulong arg1,
+                            target_ulong arg2)
+{
+    return set_HIT0_LO(env, (uint64_t)get_HILO(env) +
+                       (uint64_t)(uint32_t)arg1 * (uint64_t)(uint32_t)arg2);
+}
+
+target_ulong helper_msac(CPUMIPSState *env, target_ulong arg1,
+                         target_ulong arg2)
+{
+    return set_HI_LOT0(env, (int64_t)get_HILO(env) - (int64_t)(int32_t)arg1 *
+                       (int64_t)(int32_t)arg2);
+}
+
+target_ulong helper_msachi(CPUMIPSState *env, target_ulong arg1,
+                           target_ulong arg2)
+{
+    return set_HIT0_LO(env, (int64_t)get_HILO(env) - (int64_t)(int32_t)arg1 *
+                       (int64_t)(int32_t)arg2);
+}
+
+target_ulong helper_msacu(CPUMIPSState *env, target_ulong arg1,
+                          target_ulong arg2)
+{
+    return set_HI_LOT0(env, (uint64_t)get_HILO(env) -
+                       (uint64_t)(uint32_t)arg1 * (uint64_t)(uint32_t)arg2);
+}
+
+target_ulong helper_msachiu(CPUMIPSState *env, target_ulong arg1,
+                            target_ulong arg2)
+{
+    return set_HIT0_LO(env, (uint64_t)get_HILO(env) -
+                       (uint64_t)(uint32_t)arg1 * (uint64_t)(uint32_t)arg2);
+}
+
+target_ulong helper_mulhi(CPUMIPSState *env, target_ulong arg1,
+                          target_ulong arg2)
+{
+    return set_HIT0_LO(env, (int64_t)(int32_t)arg1 * (int64_t)(int32_t)arg2);
+}
+
+target_ulong helper_mulhiu(CPUMIPSState *env, target_ulong arg1,
+                           target_ulong arg2)
+{
+    return set_HIT0_LO(env, (uint64_t)(uint32_t)arg1 *
+                       (uint64_t)(uint32_t)arg2);
+}
+
+target_ulong helper_mulshi(CPUMIPSState *env, target_ulong arg1,
+                           target_ulong arg2)
+{
+    return set_HIT0_LO(env, 0 - (int64_t)(int32_t)arg1 *
+                       (int64_t)(int32_t)arg2);
+}
+
+target_ulong helper_mulshiu(CPUMIPSState *env, target_ulong arg1,
+                            target_ulong arg2)
+{
+    return set_HIT0_LO(env, 0 - (uint64_t)(uint32_t)arg1 *
+                       (uint64_t)(uint32_t)arg2);
+}
+
+static inline target_ulong bitswap(target_ulong v)
+{
+    v = ((v >> 1) & (target_ulong)0x5555555555555555ULL) |
+              ((v & (target_ulong)0x5555555555555555ULL) << 1);
+    v = ((v >> 2) & (target_ulong)0x3333333333333333ULL) |
+              ((v & (target_ulong)0x3333333333333333ULL) << 2);
+    v = ((v >> 4) & (target_ulong)0x0F0F0F0F0F0F0F0FULL) |
+              ((v & (target_ulong)0x0F0F0F0F0F0F0F0FULL) << 4);
+    return v;
+}
+
+#ifdef TARGET_MIPS64
+target_ulong helper_dbitswap(target_ulong rt)
+{
+    return bitswap(rt);
+}
+#endif
+
+target_ulong helper_bitswap(target_ulong rt)
+{
+    return (int32_t)bitswap(rt);
+}
+
+#ifndef CONFIG_USER_ONLY
+
+static inline hwaddr do_translate_address(CPUMIPSState *env,
+                                                      target_ulong address,
+                                                      int rw)
+{
+    hwaddr lladdr;
+
+    lladdr = cpu_mips_translate_address(env, address, rw);
+
+    if (lladdr == -1LL) {
+        cpu_loop_exit(CPU(mips_env_get_cpu(env)));
+    } else {
+        return lladdr;
+    }
+}
+
+#define HELPER_LD_ATOMIC(name, insn, almask)                                  \
+target_ulong helper_##name(CPUMIPSState *env, target_ulong arg, int mem_idx)  \
+{                                                                             \
+    if (arg & almask) {                                                       \
+        env->CP0_BadVAddr = arg;                                              \
+        helper_raise_exception(env, EXCP_AdEL);                               \
+    }                                                                         \
+    env->lladdr = do_translate_address(env, arg, 0);                          \
+    env->llval = do_##insn(env, arg, mem_idx);                                \
+    return env->llval;                                                        \
+}
+HELPER_LD_ATOMIC(ll, lw, 0x3)
+#ifdef TARGET_MIPS64
+HELPER_LD_ATOMIC(lld, ld, 0x7)
+#endif
+#undef HELPER_LD_ATOMIC
+
+#define HELPER_ST_ATOMIC(name, ld_insn, st_insn, almask)                      \
+target_ulong helper_##name(CPUMIPSState *env, target_ulong arg1,              \
+                           target_ulong arg2, int mem_idx)                    \
+{                                                                             \
+    target_long tmp;                                                          \
+                                                                              \
+    if (arg2 & almask) {                                                      \
+        env->CP0_BadVAddr = arg2;                                             \
+        helper_raise_exception(env, EXCP_AdES);                               \
+    }                                                                         \
+    if (do_translate_address(env, arg2, 1) == env->lladdr) {                  \
+        tmp = do_##ld_insn(env, arg2, mem_idx);                               \
+        if (tmp == env->llval) {                                              \
+            do_##st_insn(env, arg2, arg1, mem_idx);                           \
+            return 1;                                                         \
+        }                                                                     \
+    }                                                                         \
+    return 0;                                                                 \
+}
+HELPER_ST_ATOMIC(sc, lw, sw, 0x3)
+#ifdef TARGET_MIPS64
+HELPER_ST_ATOMIC(scd, ld, sd, 0x7)
+#endif
+#undef HELPER_ST_ATOMIC
+#endif
+
+#ifdef TARGET_WORDS_BIGENDIAN
+#define GET_LMASK(v) ((v) & 3)
+#define GET_OFFSET(addr, offset) (addr + (offset))
+#else
+#define GET_LMASK(v) (((v) & 3) ^ 3)
+#define GET_OFFSET(addr, offset) (addr - (offset))
+#endif
+
+void helper_swl(CPUMIPSState *env, target_ulong arg1, target_ulong arg2,
+                int mem_idx)
+{
+    do_sb(env, arg2, (uint8_t)(arg1 >> 24), mem_idx);
+
+    if (GET_LMASK(arg2) <= 2)
+        do_sb(env, GET_OFFSET(arg2, 1), (uint8_t)(arg1 >> 16), mem_idx);
+
+    if (GET_LMASK(arg2) <= 1)
+        do_sb(env, GET_OFFSET(arg2, 2), (uint8_t)(arg1 >> 8), mem_idx);
+
+    if (GET_LMASK(arg2) == 0)
+        do_sb(env, GET_OFFSET(arg2, 3), (uint8_t)arg1, mem_idx);
+}
+
+void helper_swr(CPUMIPSState *env, target_ulong arg1, target_ulong arg2,
+                int mem_idx)
+{
+    do_sb(env, arg2, (uint8_t)arg1, mem_idx);
+
+    if (GET_LMASK(arg2) >= 1)
+        do_sb(env, GET_OFFSET(arg2, -1), (uint8_t)(arg1 >> 8), mem_idx);
+
+    if (GET_LMASK(arg2) >= 2)
+        do_sb(env, GET_OFFSET(arg2, -2), (uint8_t)(arg1 >> 16), mem_idx);
+
+    if (GET_LMASK(arg2) == 3)
+        do_sb(env, GET_OFFSET(arg2, -3), (uint8_t)(arg1 >> 24), mem_idx);
+}
+
+#if defined(TARGET_MIPS64)
+/* "half" load and stores.  We must do the memory access inline,
+   or fault handling won't work.  */
+
+#ifdef TARGET_WORDS_BIGENDIAN
+#define GET_LMASK64(v) ((v) & 7)
+#else
+#define GET_LMASK64(v) (((v) & 7) ^ 7)
+#endif
+
+void helper_sdl(CPUMIPSState *env, target_ulong arg1, target_ulong arg2,
+                int mem_idx)
+{
+    do_sb(env, arg2, (uint8_t)(arg1 >> 56), mem_idx);
+
+    if (GET_LMASK64(arg2) <= 6)
+        do_sb(env, GET_OFFSET(arg2, 1), (uint8_t)(arg1 >> 48), mem_idx);
+
+    if (GET_LMASK64(arg2) <= 5)
+        do_sb(env, GET_OFFSET(arg2, 2), (uint8_t)(arg1 >> 40), mem_idx);
+
+    if (GET_LMASK64(arg2) <= 4)
+        do_sb(env, GET_OFFSET(arg2, 3), (uint8_t)(arg1 >> 32), mem_idx);
+
+    if (GET_LMASK64(arg2) <= 3)
+        do_sb(env, GET_OFFSET(arg2, 4), (uint8_t)(arg1 >> 24), mem_idx);
+
+    if (GET_LMASK64(arg2) <= 2)
+        do_sb(env, GET_OFFSET(arg2, 5), (uint8_t)(arg1 >> 16), mem_idx);
+
+    if (GET_LMASK64(arg2) <= 1)
+        do_sb(env, GET_OFFSET(arg2, 6), (uint8_t)(arg1 >> 8), mem_idx);
+
+    if (GET_LMASK64(arg2) <= 0)
+        do_sb(env, GET_OFFSET(arg2, 7), (uint8_t)arg1, mem_idx);
+}
+
+void helper_sdr(CPUMIPSState *env, target_ulong arg1, target_ulong arg2,
+                int mem_idx)
+{
+    do_sb(env, arg2, (uint8_t)arg1, mem_idx);
+
+    if (GET_LMASK64(arg2) >= 1)
+        do_sb(env, GET_OFFSET(arg2, -1), (uint8_t)(arg1 >> 8), mem_idx);
+
+    if (GET_LMASK64(arg2) >= 2)
+        do_sb(env, GET_OFFSET(arg2, -2), (uint8_t)(arg1 >> 16), mem_idx);
+
+    if (GET_LMASK64(arg2) >= 3)
+        do_sb(env, GET_OFFSET(arg2, -3), (uint8_t)(arg1 >> 24), mem_idx);
+
+    if (GET_LMASK64(arg2) >= 4)
+        do_sb(env, GET_OFFSET(arg2, -4), (uint8_t)(arg1 >> 32), mem_idx);
+
+    if (GET_LMASK64(arg2) >= 5)
+        do_sb(env, GET_OFFSET(arg2, -5), (uint8_t)(arg1 >> 40), mem_idx);
+
+    if (GET_LMASK64(arg2) >= 6)
+        do_sb(env, GET_OFFSET(arg2, -6), (uint8_t)(arg1 >> 48), mem_idx);
+
+    if (GET_LMASK64(arg2) == 7)
+        do_sb(env, GET_OFFSET(arg2, -7), (uint8_t)(arg1 >> 56), mem_idx);
+}
+#endif /* TARGET_MIPS64 */
+
+static const int multiple_regs[] = { 16, 17, 18, 19, 20, 21, 22, 23, 30 };
+
+void helper_lwm(CPUMIPSState *env, target_ulong addr, target_ulong reglist,
+                uint32_t mem_idx)
+{
+    target_ulong base_reglist = reglist & 0xf;
+    target_ulong do_r31 = reglist & 0x10;
+
+    if (base_reglist > 0 && base_reglist <= ARRAY_SIZE (multiple_regs)) {
+        target_ulong i;
+
+        for (i = 0; i < base_reglist; i++) {
+            env->active_tc.gpr[multiple_regs[i]] =
+                (target_long)do_lw(env, addr, mem_idx);
+            addr += 4;
+        }
+    }
+
+    if (do_r31) {
+        env->active_tc.gpr[31] = (target_long)do_lw(env, addr, mem_idx);
+    }
+}
+
+void helper_swm(CPUMIPSState *env, target_ulong addr, target_ulong reglist,
+                uint32_t mem_idx)
+{
+    target_ulong base_reglist = reglist & 0xf;
+    target_ulong do_r31 = reglist & 0x10;
+
+    if (base_reglist > 0 && base_reglist <= ARRAY_SIZE (multiple_regs)) {
+        target_ulong i;
+
+        for (i = 0; i < base_reglist; i++) {
+            do_sw(env, addr, env->active_tc.gpr[multiple_regs[i]], mem_idx);
+            addr += 4;
+        }
+    }
+
+    if (do_r31) {
+        do_sw(env, addr, env->active_tc.gpr[31], mem_idx);
+    }
+}
+
+#if defined(TARGET_MIPS64)
+void helper_ldm(CPUMIPSState *env, target_ulong addr, target_ulong reglist,
+                uint32_t mem_idx)
+{
+    target_ulong base_reglist = reglist & 0xf;
+    target_ulong do_r31 = reglist & 0x10;
+
+    if (base_reglist > 0 && base_reglist <= ARRAY_SIZE (multiple_regs)) {
+        target_ulong i;
+
+        for (i = 0; i < base_reglist; i++) {
+            env->active_tc.gpr[multiple_regs[i]] = do_ld(env, addr, mem_idx);
+            addr += 8;
+        }
+    }
+
+    if (do_r31) {
+        env->active_tc.gpr[31] = do_ld(env, addr, mem_idx);
+    }
+}
+
+void helper_sdm(CPUMIPSState *env, target_ulong addr, target_ulong reglist,
+                uint32_t mem_idx)
+{
+    target_ulong base_reglist = reglist & 0xf;
+    target_ulong do_r31 = reglist & 0x10;
+
+    if (base_reglist > 0 && base_reglist <= ARRAY_SIZE (multiple_regs)) {
+        target_ulong i;
+
+        for (i = 0; i < base_reglist; i++) {
+            do_sd(env, addr, env->active_tc.gpr[multiple_regs[i]], mem_idx);
+            addr += 8;
+        }
+    }
+
+    if (do_r31) {
+        do_sd(env, addr, env->active_tc.gpr[31], mem_idx);
+    }
+}
+#endif
+
+#ifndef CONFIG_USER_ONLY
+/* SMP helpers.  */
+static bool mips_vpe_is_wfi(MIPSCPU *c)
+{
+    CPUState *cpu = CPU(c);
+    CPUMIPSState *env = &c->env;
+
+    /* If the VPE is halted but otherwise active, it means it's waiting for
+       an interrupt.  */
+    return cpu->halted && mips_vpe_active(env);
+}
+
+static inline void mips_vpe_wake(MIPSCPU *c)
+{
+    /* Dont set ->halted = 0 directly, let it be done via cpu_has_work
+       because there might be other conditions that state that c should
+       be sleeping.  */
+    cpu_interrupt(CPU(c), CPU_INTERRUPT_WAKE);
+}
+
+static inline void mips_vpe_sleep(MIPSCPU *cpu)
+{
+    CPUState *cs = CPU(cpu);
+
+    /* The VPE was shut off, really go to bed.
+       Reset any old _WAKE requests.  */
+    cs->halted = 1;
+    cpu_reset_interrupt(cs, CPU_INTERRUPT_WAKE);
+}
+
+static inline void mips_tc_wake(MIPSCPU *cpu, int tc)
+{
+    CPUMIPSState *c = &cpu->env;
+
+    /* FIXME: TC reschedule.  */
+    if (mips_vpe_active(c) && !mips_vpe_is_wfi(cpu)) {
+        mips_vpe_wake(cpu);
+    }
+}
+
+static inline void mips_tc_sleep(MIPSCPU *cpu, int tc)
+{
+    CPUMIPSState *c = &cpu->env;
+
+    /* FIXME: TC reschedule.  */
+    if (!mips_vpe_active(c)) {
+        mips_vpe_sleep(cpu);
+    }
+}
+
+/**
+ * mips_cpu_map_tc:
+ * @env: CPU from which mapping is performed.
+ * @tc: Should point to an int with the value of the global TC index.
+ *
+ * This function will transform @tc into a local index within the
+ * returned #CPUMIPSState.
+ */
+/* FIXME: This code assumes that all VPEs have the same number of TCs,
+          which depends on runtime setup. Can probably be fixed by
+          walking the list of CPUMIPSStates.  */
+static CPUMIPSState *mips_cpu_map_tc(CPUMIPSState *env, int *tc)
+{
+    MIPSCPU *cpu;
+    CPUState *cs;
+    CPUState *other_cs;
+    int vpe_idx;
+    int tc_idx = *tc;
+
+    if (!(env->CP0_VPEConf0 & (1 << CP0VPEC0_MVP))) {
+        /* Not allowed to address other CPUs.  */
+        *tc = env->current_tc;
+        return env;
+    }
+
+    cs = CPU(mips_env_get_cpu(env));
+    vpe_idx = tc_idx / cs->nr_threads;
+    *tc = tc_idx % cs->nr_threads;
+    other_cs = qemu_get_cpu(vpe_idx);
+    if (other_cs == NULL) {
+        return env;
+    }
+    cpu = MIPS_CPU(other_cs);
+    return &cpu->env;
+}
+
+/* The per VPE CP0_Status register shares some fields with the per TC
+   CP0_TCStatus registers. These fields are wired to the same registers,
+   so changes to either of them should be reflected on both registers.
+
+   Also, EntryHi shares the bottom 8 bit ASID with TCStauts.
+
+   These helper call synchronizes the regs for a given cpu.  */
+
+/* Called for updates to CP0_Status.  Defined in "cpu.h" for gdbstub.c.  */
+/* static inline void sync_c0_status(CPUMIPSState *env, CPUMIPSState *cpu,
+                                     int tc);  */
+
+/* Called for updates to CP0_TCStatus.  */
+static void sync_c0_tcstatus(CPUMIPSState *cpu, int tc,
+                             target_ulong v)
+{
+    uint32_t status;
+    uint32_t tcu, tmx, tasid, tksu;
+    uint32_t mask = ((1U << CP0St_CU3)
+                       | (1 << CP0St_CU2)
+                       | (1 << CP0St_CU1)
+                       | (1 << CP0St_CU0)
+                       | (1 << CP0St_MX)
+                       | (3 << CP0St_KSU));
+
+    tcu = (v >> CP0TCSt_TCU0) & 0xf;
+    tmx = (v >> CP0TCSt_TMX) & 0x1;
+    tasid = v & 0xff;
+    tksu = (v >> CP0TCSt_TKSU) & 0x3;
+
+    status = tcu << CP0St_CU0;
+    status |= tmx << CP0St_MX;
+    status |= tksu << CP0St_KSU;
+
+    cpu->CP0_Status &= ~mask;
+    cpu->CP0_Status |= status;
+
+    /* Sync the TASID with EntryHi.  */
+    cpu->CP0_EntryHi &= ~0xff;
+    cpu->CP0_EntryHi |= tasid;
+
+    compute_hflags(cpu);
+}
+
+/* Called for updates to CP0_EntryHi.  */
+static void sync_c0_entryhi(CPUMIPSState *cpu, int tc)
+{
+    int32_t *tcst;
+    uint32_t asid, v = cpu->CP0_EntryHi;
+
+    asid = v & 0xff;
+
+    if (tc == cpu->current_tc) {
+        tcst = &cpu->active_tc.CP0_TCStatus;
+    } else {
+        tcst = &cpu->tcs[tc].CP0_TCStatus;
+    }
+
+    *tcst &= ~0xff;
+    *tcst |= asid;
+}
+
+/* CP0 helpers */
+target_ulong helper_mfc0_mvpcontrol(CPUMIPSState *env)
+{
+    return env->mvp->CP0_MVPControl;
+}
+
+target_ulong helper_mfc0_mvpconf0(CPUMIPSState *env)
+{
+    return env->mvp->CP0_MVPConf0;
+}
+
+target_ulong helper_mfc0_mvpconf1(CPUMIPSState *env)
+{
+    return env->mvp->CP0_MVPConf1;
+}
+
+target_ulong helper_mfc0_random(CPUMIPSState *env)
+{
+    return (int32_t)cpu_mips_get_random(env);
+}
+
+target_ulong helper_mfc0_tcstatus(CPUMIPSState *env)
+{
+    return env->active_tc.CP0_TCStatus;
+}
+
+target_ulong helper_mftc0_tcstatus(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        return other->active_tc.CP0_TCStatus;
+    else
+        return other->tcs[other_tc].CP0_TCStatus;
+}
+
+target_ulong helper_mfc0_tcbind(CPUMIPSState *env)
+{
+    return env->active_tc.CP0_TCBind;
+}
+
+target_ulong helper_mftc0_tcbind(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        return other->active_tc.CP0_TCBind;
+    else
+        return other->tcs[other_tc].CP0_TCBind;
+}
+
+target_ulong helper_mfc0_tcrestart(CPUMIPSState *env)
+{
+    return env->active_tc.PC;
+}
+
+target_ulong helper_mftc0_tcrestart(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        return other->active_tc.PC;
+    else
+        return other->tcs[other_tc].PC;
+}
+
+target_ulong helper_mfc0_tchalt(CPUMIPSState *env)
+{
+    return env->active_tc.CP0_TCHalt;
+}
+
+target_ulong helper_mftc0_tchalt(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        return other->active_tc.CP0_TCHalt;
+    else
+        return other->tcs[other_tc].CP0_TCHalt;
+}
+
+target_ulong helper_mfc0_tccontext(CPUMIPSState *env)
+{
+    return env->active_tc.CP0_TCContext;
+}
+
+target_ulong helper_mftc0_tccontext(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        return other->active_tc.CP0_TCContext;
+    else
+        return other->tcs[other_tc].CP0_TCContext;
+}
+
+target_ulong helper_mfc0_tcschedule(CPUMIPSState *env)
+{
+    return env->active_tc.CP0_TCSchedule;
+}
+
+target_ulong helper_mftc0_tcschedule(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        return other->active_tc.CP0_TCSchedule;
+    else
+        return other->tcs[other_tc].CP0_TCSchedule;
+}
+
+target_ulong helper_mfc0_tcschefback(CPUMIPSState *env)
+{
+    return env->active_tc.CP0_TCScheFBack;
+}
+
+target_ulong helper_mftc0_tcschefback(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        return other->active_tc.CP0_TCScheFBack;
+    else
+        return other->tcs[other_tc].CP0_TCScheFBack;
+}
+
+target_ulong helper_mfc0_count(CPUMIPSState *env)
+{
+    return (int32_t)cpu_mips_get_count(env);
+}
+
+target_ulong helper_mftc0_entryhi(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    return other->CP0_EntryHi;
+}
+
+target_ulong helper_mftc0_cause(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    int32_t tccause;
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc) {
+        tccause = other->CP0_Cause;
+    } else {
+        tccause = other->CP0_Cause;
+    }
+
+    return tccause;
+}
+
+target_ulong helper_mftc0_status(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    return other->CP0_Status;
+}
+
+target_ulong helper_mfc0_lladdr(CPUMIPSState *env)
+{
+    return (int32_t)(env->lladdr >> env->CP0_LLAddr_shift);
+}
+
+target_ulong helper_mfc0_watchlo(CPUMIPSState *env, uint32_t sel)
+{
+    return (int32_t)env->CP0_WatchLo[sel];
+}
+
+target_ulong helper_mfc0_watchhi(CPUMIPSState *env, uint32_t sel)
+{
+    return env->CP0_WatchHi[sel];
+}
+
+target_ulong helper_mfc0_debug(CPUMIPSState *env)
+{
+    target_ulong t0 = env->CP0_Debug;
+    if (env->hflags & MIPS_HFLAG_DM)
+        t0 |= 1 << CP0DB_DM;
+
+    return t0;
+}
+
+target_ulong helper_mftc0_debug(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    int32_t tcstatus;
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        tcstatus = other->active_tc.CP0_Debug_tcstatus;
+    else
+        tcstatus = other->tcs[other_tc].CP0_Debug_tcstatus;
+
+    /* XXX: Might be wrong, check with EJTAG spec. */
+    return (other->CP0_Debug & ~((1 << CP0DB_SSt) | (1 << CP0DB_Halt))) |
+            (tcstatus & ((1 << CP0DB_SSt) | (1 << CP0DB_Halt)));
+}
+
+#if defined(TARGET_MIPS64)
+target_ulong helper_dmfc0_tcrestart(CPUMIPSState *env)
+{
+    return env->active_tc.PC;
+}
+
+target_ulong helper_dmfc0_tchalt(CPUMIPSState *env)
+{
+    return env->active_tc.CP0_TCHalt;
+}
+
+target_ulong helper_dmfc0_tccontext(CPUMIPSState *env)
+{
+    return env->active_tc.CP0_TCContext;
+}
+
+target_ulong helper_dmfc0_tcschedule(CPUMIPSState *env)
+{
+    return env->active_tc.CP0_TCSchedule;
+}
+
+target_ulong helper_dmfc0_tcschefback(CPUMIPSState *env)
+{
+    return env->active_tc.CP0_TCScheFBack;
+}
+
+target_ulong helper_dmfc0_lladdr(CPUMIPSState *env)
+{
+    return env->lladdr >> env->CP0_LLAddr_shift;
+}
+
+target_ulong helper_dmfc0_watchlo(CPUMIPSState *env, uint32_t sel)
+{
+    return env->CP0_WatchLo[sel];
+}
+#endif /* TARGET_MIPS64 */
+
+void helper_mtc0_index(CPUMIPSState *env, target_ulong arg1)
+{
+    uint32_t index_p = env->CP0_Index & 0x80000000;
+    uint32_t tlb_index = arg1 & 0x7fffffff;
+    if (tlb_index < env->tlb->nb_tlb) {
+        if (env->insn_flags & ISA_MIPS32R6) {
+            index_p |= arg1 & 0x80000000;
+        }
+        env->CP0_Index = index_p | tlb_index;
+    }
+}
+
+void helper_mtc0_mvpcontrol(CPUMIPSState *env, target_ulong arg1)
+{
+    uint32_t mask = 0;
+    uint32_t newval;
+
+    if (env->CP0_VPEConf0 & (1 << CP0VPEC0_MVP))
+        mask |= (1 << CP0MVPCo_CPA) | (1 << CP0MVPCo_VPC) |
+                (1 << CP0MVPCo_EVP);
+    if (env->mvp->CP0_MVPControl & (1 << CP0MVPCo_VPC))
+        mask |= (1 << CP0MVPCo_STLB);
+    newval = (env->mvp->CP0_MVPControl & ~mask) | (arg1 & mask);
+
+    // TODO: Enable/disable shared TLB, enable/disable VPEs.
+
+    env->mvp->CP0_MVPControl = newval;
+}
+
+void helper_mtc0_vpecontrol(CPUMIPSState *env, target_ulong arg1)
+{
+    uint32_t mask;
+    uint32_t newval;
+
+    mask = (1 << CP0VPECo_YSI) | (1 << CP0VPECo_GSI) |
+           (1 << CP0VPECo_TE) | (0xff << CP0VPECo_TargTC);
+    newval = (env->CP0_VPEControl & ~mask) | (arg1 & mask);
+
+    /* Yield scheduler intercept not implemented. */
+    /* Gating storage scheduler intercept not implemented. */
+
+    // TODO: Enable/disable TCs.
+
+    env->CP0_VPEControl = newval;
+}
+
+void helper_mttc0_vpecontrol(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+    uint32_t mask;
+    uint32_t newval;
+
+    mask = (1 << CP0VPECo_YSI) | (1 << CP0VPECo_GSI) |
+           (1 << CP0VPECo_TE) | (0xff << CP0VPECo_TargTC);
+    newval = (other->CP0_VPEControl & ~mask) | (arg1 & mask);
+
+    /* TODO: Enable/disable TCs.  */
+
+    other->CP0_VPEControl = newval;
+}
+
+target_ulong helper_mftc0_vpecontrol(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+    /* FIXME: Mask away return zero on read bits.  */
+    return other->CP0_VPEControl;
+}
+
+target_ulong helper_mftc0_vpeconf0(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    return other->CP0_VPEConf0;
+}
+
+void helper_mtc0_vpeconf0(CPUMIPSState *env, target_ulong arg1)
+{
+    uint32_t mask = 0;
+    uint32_t newval;
+
+    if (env->CP0_VPEConf0 & (1 << CP0VPEC0_MVP)) {
+        if (env->CP0_VPEConf0 & (1 << CP0VPEC0_VPA))
+            mask |= (0xff << CP0VPEC0_XTC);
+        mask |= (1 << CP0VPEC0_MVP) | (1 << CP0VPEC0_VPA);
+    }
+    newval = (env->CP0_VPEConf0 & ~mask) | (arg1 & mask);
+
+    // TODO: TC exclusive handling due to ERL/EXL.
+
+    env->CP0_VPEConf0 = newval;
+}
+
+void helper_mttc0_vpeconf0(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+    uint32_t mask = 0;
+    uint32_t newval;
+
+    mask |= (1 << CP0VPEC0_MVP) | (1 << CP0VPEC0_VPA);
+    newval = (other->CP0_VPEConf0 & ~mask) | (arg1 & mask);
+
+    /* TODO: TC exclusive handling due to ERL/EXL.  */
+    other->CP0_VPEConf0 = newval;
+}
+
+void helper_mtc0_vpeconf1(CPUMIPSState *env, target_ulong arg1)
+{
+    uint32_t mask = 0;
+    uint32_t newval;
+
+    if (env->mvp->CP0_MVPControl & (1 << CP0MVPCo_VPC))
+        mask |= (0xff << CP0VPEC1_NCX) | (0xff << CP0VPEC1_NCP2) |
+                (0xff << CP0VPEC1_NCP1);
+    newval = (env->CP0_VPEConf1 & ~mask) | (arg1 & mask);
+
+    /* UDI not implemented. */
+    /* CP2 not implemented. */
+
+    // TODO: Handle FPU (CP1) binding.
+
+    env->CP0_VPEConf1 = newval;
+}
+
+void helper_mtc0_yqmask(CPUMIPSState *env, target_ulong arg1)
+{
+    /* Yield qualifier inputs not implemented. */
+    env->CP0_YQMask = 0x00000000;
+}
+
+void helper_mtc0_vpeopt(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_VPEOpt = arg1 & 0x0000ffff;
+}
+
+#define MTC0_ENTRYLO_MASK(env) ((env->PAMask >> 6) & 0x3FFFFFFF)
+
+void helper_mtc0_entrylo0(CPUMIPSState *env, target_ulong arg1)
+{
+    /* 1k pages not implemented */
+    target_ulong rxi = arg1 & (env->CP0_PageGrain & (3u << CP0PG_XIE));
+    env->CP0_EntryLo0 = (arg1 & MTC0_ENTRYLO_MASK(env))
+                        | (rxi << (CP0EnLo_XI - 30));
+}
+
+#if defined(TARGET_MIPS64)
+#define DMTC0_ENTRYLO_MASK(env) (env->PAMask >> 6)
+
+void helper_dmtc0_entrylo0(CPUMIPSState *env, uint64_t arg1)
+{
+    uint64_t rxi = arg1 & ((env->CP0_PageGrain & (3ull << CP0PG_XIE)) << 32);
+    env->CP0_EntryLo0 = (arg1 & DMTC0_ENTRYLO_MASK(env)) | rxi;
+}
+#endif
+
+void helper_mtc0_tcstatus(CPUMIPSState *env, target_ulong arg1)
+{
+    uint32_t mask = env->CP0_TCStatus_rw_bitmask;
+    uint32_t newval;
+
+    newval = (env->active_tc.CP0_TCStatus & ~mask) | (arg1 & mask);
+
+    env->active_tc.CP0_TCStatus = newval;
+    sync_c0_tcstatus(env, env->current_tc, newval);
+}
+
+void helper_mttc0_tcstatus(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        other->active_tc.CP0_TCStatus = arg1;
+    else
+        other->tcs[other_tc].CP0_TCStatus = arg1;
+    sync_c0_tcstatus(other, other_tc, arg1);
+}
+
+void helper_mtc0_tcbind(CPUMIPSState *env, target_ulong arg1)
+{
+    uint32_t mask = (1 << CP0TCBd_TBE);
+    uint32_t newval;
+
+    if (env->mvp->CP0_MVPControl & (1 << CP0MVPCo_VPC))
+        mask |= (1 << CP0TCBd_CurVPE);
+    newval = (env->active_tc.CP0_TCBind & ~mask) | (arg1 & mask);
+    env->active_tc.CP0_TCBind = newval;
+}
+
+void helper_mttc0_tcbind(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    uint32_t mask = (1 << CP0TCBd_TBE);
+    uint32_t newval;
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other->mvp->CP0_MVPControl & (1 << CP0MVPCo_VPC))
+        mask |= (1 << CP0TCBd_CurVPE);
+    if (other_tc == other->current_tc) {
+        newval = (other->active_tc.CP0_TCBind & ~mask) | (arg1 & mask);
+        other->active_tc.CP0_TCBind = newval;
+    } else {
+        newval = (other->tcs[other_tc].CP0_TCBind & ~mask) | (arg1 & mask);
+        other->tcs[other_tc].CP0_TCBind = newval;
+    }
+}
+
+void helper_mtc0_tcrestart(CPUMIPSState *env, target_ulong arg1)
+{
+    env->active_tc.PC = arg1;
+    env->active_tc.CP0_TCStatus &= ~(1 << CP0TCSt_TDS);
+    env->lladdr = 0ULL;
+    /* MIPS16 not implemented. */
+}
+
+void helper_mttc0_tcrestart(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc) {
+        other->active_tc.PC = arg1;
+        other->active_tc.CP0_TCStatus &= ~(1 << CP0TCSt_TDS);
+        other->lladdr = 0ULL;
+        /* MIPS16 not implemented. */
+    } else {
+        other->tcs[other_tc].PC = arg1;
+        other->tcs[other_tc].CP0_TCStatus &= ~(1 << CP0TCSt_TDS);
+        other->lladdr = 0ULL;
+        /* MIPS16 not implemented. */
+    }
+}
+
+void helper_mtc0_tchalt(CPUMIPSState *env, target_ulong arg1)
+{
+    MIPSCPU *cpu = mips_env_get_cpu(env);
+
+    env->active_tc.CP0_TCHalt = arg1 & 0x1;
+
+    // TODO: Halt TC / Restart (if allocated+active) TC.
+    if (env->active_tc.CP0_TCHalt & 1) {
+        mips_tc_sleep(cpu, env->current_tc);
+    } else {
+        mips_tc_wake(cpu, env->current_tc);
+    }
+}
+
+void helper_mttc0_tchalt(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+    MIPSCPU *other_cpu = mips_env_get_cpu(other);
+
+    // TODO: Halt TC / Restart (if allocated+active) TC.
+
+    if (other_tc == other->current_tc)
+        other->active_tc.CP0_TCHalt = arg1;
+    else
+        other->tcs[other_tc].CP0_TCHalt = arg1;
+
+    if (arg1 & 1) {
+        mips_tc_sleep(other_cpu, other_tc);
+    } else {
+        mips_tc_wake(other_cpu, other_tc);
+    }
+}
+
+void helper_mtc0_tccontext(CPUMIPSState *env, target_ulong arg1)
+{
+    env->active_tc.CP0_TCContext = arg1;
+}
+
+void helper_mttc0_tccontext(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        other->active_tc.CP0_TCContext = arg1;
+    else
+        other->tcs[other_tc].CP0_TCContext = arg1;
+}
+
+void helper_mtc0_tcschedule(CPUMIPSState *env, target_ulong arg1)
+{
+    env->active_tc.CP0_TCSchedule = arg1;
+}
+
+void helper_mttc0_tcschedule(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        other->active_tc.CP0_TCSchedule = arg1;
+    else
+        other->tcs[other_tc].CP0_TCSchedule = arg1;
+}
+
+void helper_mtc0_tcschefback(CPUMIPSState *env, target_ulong arg1)
+{
+    env->active_tc.CP0_TCScheFBack = arg1;
+}
+
+void helper_mttc0_tcschefback(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        other->active_tc.CP0_TCScheFBack = arg1;
+    else
+        other->tcs[other_tc].CP0_TCScheFBack = arg1;
+}
+
+void helper_mtc0_entrylo1(CPUMIPSState *env, target_ulong arg1)
+{
+    /* 1k pages not implemented */
+    target_ulong rxi = arg1 & (env->CP0_PageGrain & (3u << CP0PG_XIE));
+    env->CP0_EntryLo1 = (arg1 & MTC0_ENTRYLO_MASK(env))
+                        | (rxi << (CP0EnLo_XI - 30));
+}
+
+#if defined(TARGET_MIPS64)
+void helper_dmtc0_entrylo1(CPUMIPSState *env, uint64_t arg1)
+{
+    uint64_t rxi = arg1 & ((env->CP0_PageGrain & (3ull << CP0PG_XIE)) << 32);
+    env->CP0_EntryLo1 = (arg1 & DMTC0_ENTRYLO_MASK(env)) | rxi;
+}
+#endif
+
+void helper_mtc0_context(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_Context = (env->CP0_Context & 0x007FFFFF) | (arg1 & ~0x007FFFFF);
+}
+
+void helper_mtc0_pagemask(CPUMIPSState *env, target_ulong arg1)
+{
+    uint64_t mask = arg1 >> (TARGET_PAGE_BITS + 1);
+    if (!(env->insn_flags & ISA_MIPS32R6) || (arg1 == ~0) ||
+        (mask == 0x0000 || mask == 0x0003 || mask == 0x000F ||
+         mask == 0x003F || mask == 0x00FF || mask == 0x03FF ||
+         mask == 0x0FFF || mask == 0x3FFF || mask == 0xFFFF)) {
+        env->CP0_PageMask = arg1 & (0x1FFFFFFF & (TARGET_PAGE_MASK << 1));
+    }
+}
+
+void helper_mtc0_pagegrain(CPUMIPSState *env, target_ulong arg1)
+{
+    /* SmartMIPS not implemented */
+    /* 1k pages not implemented */
+    env->CP0_PageGrain = (arg1 & env->CP0_PageGrain_rw_bitmask) |
+                         (env->CP0_PageGrain & ~env->CP0_PageGrain_rw_bitmask);
+    compute_hflags(env);
+    restore_pamask(env);
+}
+
+void helper_mtc0_wired(CPUMIPSState *env, target_ulong arg1)
+{
+    if (env->insn_flags & ISA_MIPS32R6) {
+        if (arg1 < env->tlb->nb_tlb) {
+            env->CP0_Wired = arg1;
+        }
+    } else {
+        env->CP0_Wired = arg1 % env->tlb->nb_tlb;
+    }
+}
+
+void helper_mtc0_srsconf0(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_SRSConf0 |= arg1 & env->CP0_SRSConf0_rw_bitmask;
+}
+
+void helper_mtc0_srsconf1(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_SRSConf1 |= arg1 & env->CP0_SRSConf1_rw_bitmask;
+}
+
+void helper_mtc0_srsconf2(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_SRSConf2 |= arg1 & env->CP0_SRSConf2_rw_bitmask;
+}
+
+void helper_mtc0_srsconf3(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_SRSConf3 |= arg1 & env->CP0_SRSConf3_rw_bitmask;
+}
+
+void helper_mtc0_srsconf4(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_SRSConf4 |= arg1 & env->CP0_SRSConf4_rw_bitmask;
+}
+
+void helper_mtc0_hwrena(CPUMIPSState *env, target_ulong arg1)
+{
+    uint32_t mask = 0x0000000F;
+
+    if (env->CP0_Config3 & (1 << CP0C3_ULRI)) {
+        mask |= (1 << 29);
+
+        if (arg1 & (1 << 29)) {
+            env->hflags |= MIPS_HFLAG_HWRENA_ULR;
+        } else {
+            env->hflags &= ~MIPS_HFLAG_HWRENA_ULR;
+        }
+    }
+
+    env->CP0_HWREna = arg1 & mask;
+}
+
+void helper_mtc0_count(CPUMIPSState *env, target_ulong arg1)
+{
+    cpu_mips_store_count(env, arg1);
+}
+
+void helper_mtc0_entryhi(CPUMIPSState *env, target_ulong arg1)
+{
+    target_ulong old, val, mask;
+    mask = (TARGET_PAGE_MASK << 1) | 0xFF;
+    if (((env->CP0_Config4 >> CP0C4_IE) & 0x3) >= 2) {
+        mask |= 1 << CP0EnHi_EHINV;
+    }
+
+    /* 1k pages not implemented */
+#if defined(TARGET_MIPS64)
+    if (env->insn_flags & ISA_MIPS32R6) {
+        int entryhi_r = extract64(arg1, 62, 2);
+        int config0_at = extract32(env->CP0_Config0, 13, 2);
+        bool no_supervisor = (env->CP0_Status_rw_bitmask & 0x8) == 0;
+        if ((entryhi_r == 2) ||
+            (entryhi_r == 1 && (no_supervisor || config0_at == 1))) {
+            /* skip EntryHi.R field if new value is reserved */
+            mask &= ~(0x3ull << 62);
+        }
+    }
+    mask &= env->SEGMask;
+#endif
+    old = env->CP0_EntryHi;
+    val = (arg1 & mask) | (old & ~mask);
+    env->CP0_EntryHi = val;
+    if (env->CP0_Config3 & (1 << CP0C3_MT)) {
+        sync_c0_entryhi(env, env->current_tc);
+    }
+    /* If the ASID changes, flush qemu's TLB.  */
+    if ((old & 0xFF) != (val & 0xFF))
+        cpu_mips_tlb_flush(env, 1);
+}
+
+void helper_mttc0_entryhi(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    other->CP0_EntryHi = arg1;
+    sync_c0_entryhi(other, other_tc);
+}
+
+void helper_mtc0_compare(CPUMIPSState *env, target_ulong arg1)
+{
+    cpu_mips_store_compare(env, arg1);
+}
+
+void helper_mtc0_status(CPUMIPSState *env, target_ulong arg1)
+{
+    MIPSCPU *cpu = mips_env_get_cpu(env);
+    uint32_t val, old;
+
+    old = env->CP0_Status;
+    cpu_mips_store_status(env, arg1);
+    val = env->CP0_Status;
+
+    if (qemu_loglevel_mask(CPU_LOG_EXEC)) {
+        qemu_log("Status %08x (%08x) => %08x (%08x) Cause %08x",
+                old, old & env->CP0_Cause & CP0Ca_IP_mask,
+                val, val & env->CP0_Cause & CP0Ca_IP_mask,
+                env->CP0_Cause);
+        switch (env->hflags & MIPS_HFLAG_KSU) {
+        case MIPS_HFLAG_UM: qemu_log(", UM\n"); break;
+        case MIPS_HFLAG_SM: qemu_log(", SM\n"); break;
+        case MIPS_HFLAG_KM: qemu_log("\n"); break;
+        default:
+            cpu_abort(CPU(cpu), "Invalid MMU mode!\n");
+            break;
+        }
+    }
+}
+
+void helper_mttc0_status(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    uint32_t mask = env->CP0_Status_rw_bitmask & ~0xf1000018;
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    other->CP0_Status = (other->CP0_Status & ~mask) | (arg1 & mask);
+    sync_c0_status(env, other, other_tc);
+}
+
+void helper_mtc0_intctl(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_IntCtl = (env->CP0_IntCtl & ~0x000003e0) | (arg1 & 0x000003e0);
+}
+
+void helper_mtc0_srsctl(CPUMIPSState *env, target_ulong arg1)
+{
+    uint32_t mask = (0xf << CP0SRSCtl_ESS) | (0xf << CP0SRSCtl_PSS);
+    env->CP0_SRSCtl = (env->CP0_SRSCtl & ~mask) | (arg1 & mask);
+}
+
+void helper_mtc0_cause(CPUMIPSState *env, target_ulong arg1)
+{
+    cpu_mips_store_cause(env, arg1);
+}
+
+void helper_mttc0_cause(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    cpu_mips_store_cause(other, arg1);
+}
+
+target_ulong helper_mftc0_epc(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    return other->CP0_EPC;
+}
+
+target_ulong helper_mftc0_ebase(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    return other->CP0_EBase;
+}
+
+void helper_mtc0_ebase(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_EBase = (env->CP0_EBase & ~0x3FFFF000) | (arg1 & 0x3FFFF000);
+}
+
+void helper_mttc0_ebase(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+    other->CP0_EBase = (other->CP0_EBase & ~0x3FFFF000) | (arg1 & 0x3FFFF000);
+}
+
+target_ulong helper_mftc0_configx(CPUMIPSState *env, target_ulong idx)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    switch (idx) {
+    case 0: return other->CP0_Config0;
+    case 1: return other->CP0_Config1;
+    case 2: return other->CP0_Config2;
+    case 3: return other->CP0_Config3;
+    /* 4 and 5 are reserved.  */
+    case 6: return other->CP0_Config6;
+    case 7: return other->CP0_Config7;
+    default:
+        break;
+    }
+    return 0;
+}
+
+void helper_mtc0_config0(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_Config0 = (env->CP0_Config0 & 0x81FFFFF8) | (arg1 & 0x00000007);
+}
+
+void helper_mtc0_config2(CPUMIPSState *env, target_ulong arg1)
+{
+    /* tertiary/secondary caches not implemented */
+    env->CP0_Config2 = (env->CP0_Config2 & 0x8FFF0FFF);
+}
+
+void helper_mtc0_config3(CPUMIPSState *env, target_ulong arg1)
+{
+    if (env->insn_flags & ASE_MICROMIPS) {
+        env->CP0_Config3 = (env->CP0_Config3 & ~(1 << CP0C3_ISA_ON_EXC)) |
+                           (arg1 & (1 << CP0C3_ISA_ON_EXC));
+    }
+}
+
+void helper_mtc0_config4(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_Config4 = (env->CP0_Config4 & (~env->CP0_Config4_rw_bitmask)) |
+                       (arg1 & env->CP0_Config4_rw_bitmask);
+}
+
+void helper_mtc0_config5(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_Config5 = (env->CP0_Config5 & (~env->CP0_Config5_rw_bitmask)) |
+                       (arg1 & env->CP0_Config5_rw_bitmask);
+    compute_hflags(env);
+}
+
+void helper_mtc0_lladdr(CPUMIPSState *env, target_ulong arg1)
+{
+    target_long mask = env->CP0_LLAddr_rw_bitmask;
+    arg1 = arg1 << env->CP0_LLAddr_shift;
+    env->lladdr = (env->lladdr & ~mask) | (arg1 & mask);
+}
+
+void helper_mtc0_watchlo(CPUMIPSState *env, target_ulong arg1, uint32_t sel)
+{
+    /* Watch exceptions for instructions, data loads, data stores
+       not implemented. */
+    env->CP0_WatchLo[sel] = (arg1 & ~0x7);
+}
+
+void helper_mtc0_watchhi(CPUMIPSState *env, target_ulong arg1, uint32_t sel)
+{
+    env->CP0_WatchHi[sel] = (arg1 & 0x40FF0FF8);
+    env->CP0_WatchHi[sel] &= ~(env->CP0_WatchHi[sel] & arg1 & 0x7);
+}
+
+void helper_mtc0_xcontext(CPUMIPSState *env, target_ulong arg1)
+{
+    target_ulong mask = (1ULL << (env->SEGBITS - 7)) - 1;
+    env->CP0_XContext = (env->CP0_XContext & mask) | (arg1 & ~mask);
+}
+
+void helper_mtc0_framemask(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_Framemask = arg1; /* XXX */
+}
+
+void helper_mtc0_debug(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_Debug = (env->CP0_Debug & 0x8C03FC1F) | (arg1 & 0x13300120);
+    if (arg1 & (1 << CP0DB_DM))
+        env->hflags |= MIPS_HFLAG_DM;
+    else
+        env->hflags &= ~MIPS_HFLAG_DM;
+}
+
+void helper_mttc0_debug(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    uint32_t val = arg1 & ((1 << CP0DB_SSt) | (1 << CP0DB_Halt));
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    /* XXX: Might be wrong, check with EJTAG spec. */
+    if (other_tc == other->current_tc)
+        other->active_tc.CP0_Debug_tcstatus = val;
+    else
+        other->tcs[other_tc].CP0_Debug_tcstatus = val;
+    other->CP0_Debug = (other->CP0_Debug &
+                     ((1 << CP0DB_SSt) | (1 << CP0DB_Halt))) |
+                     (arg1 & ~((1 << CP0DB_SSt) | (1 << CP0DB_Halt)));
+}
+
+void helper_mtc0_performance0(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_Performance0 = arg1 & 0x000007ff;
+}
+
+void helper_mtc0_taglo(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_TagLo = arg1 & 0xFFFFFCF6;
+}
+
+void helper_mtc0_datalo(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_DataLo = arg1; /* XXX */
+}
+
+void helper_mtc0_taghi(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_TagHi = arg1; /* XXX */
+}
+
+void helper_mtc0_datahi(CPUMIPSState *env, target_ulong arg1)
+{
+    env->CP0_DataHi = arg1; /* XXX */
+}
+
+/* MIPS MT functions */
+target_ulong helper_mftgpr(CPUMIPSState *env, uint32_t sel)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        return other->active_tc.gpr[sel];
+    else
+        return other->tcs[other_tc].gpr[sel];
+}
+
+target_ulong helper_mftlo(CPUMIPSState *env, uint32_t sel)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        return other->active_tc.LO[sel];
+    else
+        return other->tcs[other_tc].LO[sel];
+}
+
+target_ulong helper_mfthi(CPUMIPSState *env, uint32_t sel)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        return other->active_tc.HI[sel];
+    else
+        return other->tcs[other_tc].HI[sel];
+}
+
+target_ulong helper_mftacx(CPUMIPSState *env, uint32_t sel)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        return other->active_tc.ACX[sel];
+    else
+        return other->tcs[other_tc].ACX[sel];
+}
+
+target_ulong helper_mftdsp(CPUMIPSState *env)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        return other->active_tc.DSPControl;
+    else
+        return other->tcs[other_tc].DSPControl;
+}
+
+void helper_mttgpr(CPUMIPSState *env, target_ulong arg1, uint32_t sel)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        other->active_tc.gpr[sel] = arg1;
+    else
+        other->tcs[other_tc].gpr[sel] = arg1;
+}
+
+void helper_mttlo(CPUMIPSState *env, target_ulong arg1, uint32_t sel)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        other->active_tc.LO[sel] = arg1;
+    else
+        other->tcs[other_tc].LO[sel] = arg1;
+}
+
+void helper_mtthi(CPUMIPSState *env, target_ulong arg1, uint32_t sel)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        other->active_tc.HI[sel] = arg1;
+    else
+        other->tcs[other_tc].HI[sel] = arg1;
+}
+
+void helper_mttacx(CPUMIPSState *env, target_ulong arg1, uint32_t sel)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        other->active_tc.ACX[sel] = arg1;
+    else
+        other->tcs[other_tc].ACX[sel] = arg1;
+}
+
+void helper_mttdsp(CPUMIPSState *env, target_ulong arg1)
+{
+    int other_tc = env->CP0_VPEControl & (0xff << CP0VPECo_TargTC);
+    CPUMIPSState *other = mips_cpu_map_tc(env, &other_tc);
+
+    if (other_tc == other->current_tc)
+        other->active_tc.DSPControl = arg1;
+    else
+        other->tcs[other_tc].DSPControl = arg1;
+}
+
+/* MIPS MT functions */
+target_ulong helper_dmt(void)
+{
+    // TODO
+     return 0;
+}
+
+target_ulong helper_emt(void)
+{
+    // TODO
+    return 0;
+}
+
+target_ulong helper_dvpe(CPUMIPSState *env)
+{
+    CPUState *other_cs = first_cpu;
+    target_ulong prev = env->mvp->CP0_MVPControl;
+
+    CPU_FOREACH(other_cs) {
+        MIPSCPU *other_cpu = MIPS_CPU(other_cs);
+        /* Turn off all VPEs except the one executing the dvpe.  */
+        if (&other_cpu->env != env) {
+            other_cpu->env.mvp->CP0_MVPControl &= ~(1 << CP0MVPCo_EVP);
+            mips_vpe_sleep(other_cpu);
+        }
+    }
+    return prev;
+}
+
+target_ulong helper_evpe(CPUMIPSState *env)
+{
+    CPUState *other_cs = first_cpu;
+    target_ulong prev = env->mvp->CP0_MVPControl;
+
+    CPU_FOREACH(other_cs) {
+        MIPSCPU *other_cpu = MIPS_CPU(other_cs);
+
+        if (&other_cpu->env != env
+            /* If the VPE is WFI, don't disturb its sleep.  */
+            && !mips_vpe_is_wfi(other_cpu)) {
+            /* Enable the VPE.  */
+            other_cpu->env.mvp->CP0_MVPControl |= (1 << CP0MVPCo_EVP);
+            mips_vpe_wake(other_cpu); /* And wake it up.  */
+        }
+    }
+    return prev;
+}
+#endif /* !CONFIG_USER_ONLY */
+
+void helper_fork(target_ulong arg1, target_ulong arg2)
+{
+    // arg1 = rt, arg2 = rs
+    // TODO: store to TC register
+}
+
+target_ulong helper_yield(CPUMIPSState *env, target_ulong arg)
+{
+    target_long arg1 = arg;
+
+    if (arg1 < 0) {
+        /* No scheduling policy implemented. */
+        if (arg1 != -2) {
+            if (env->CP0_VPEControl & (1 << CP0VPECo_YSI) &&
+                env->active_tc.CP0_TCStatus & (1 << CP0TCSt_DT)) {
+                env->CP0_VPEControl &= ~(0x7 << CP0VPECo_EXCPT);
+                env->CP0_VPEControl |= 4 << CP0VPECo_EXCPT;
+                helper_raise_exception(env, EXCP_THREAD);
+            }
+        }
+    } else if (arg1 == 0) {
+        if (0 /* TODO: TC underflow */) {
+            env->CP0_VPEControl &= ~(0x7 << CP0VPECo_EXCPT);
+            helper_raise_exception(env, EXCP_THREAD);
+        } else {
+            // TODO: Deallocate TC
+        }
+    } else if (arg1 > 0) {
+        /* Yield qualifier inputs not implemented. */
+        env->CP0_VPEControl &= ~(0x7 << CP0VPECo_EXCPT);
+        env->CP0_VPEControl |= 2 << CP0VPECo_EXCPT;
+        helper_raise_exception(env, EXCP_THREAD);
+    }
+    return env->CP0_YQMask;
+}
+
+#ifndef CONFIG_USER_ONLY
+/* TLB management */
+static void cpu_mips_tlb_flush (CPUMIPSState *env, int flush_global)
+{
+    MIPSCPU *cpu = mips_env_get_cpu(env);
+
+    /* Flush qemu's TLB and discard all shadowed entries.  */
+    tlb_flush(CPU(cpu), flush_global);
+    env->tlb->tlb_in_use = env->tlb->nb_tlb;
+}
+
+static void r4k_mips_tlb_flush_extra (CPUMIPSState *env, int first)
+{
+    /* Discard entries from env->tlb[first] onwards.  */
+    while (env->tlb->tlb_in_use > first) {
+        r4k_invalidate_tlb(env, --env->tlb->tlb_in_use, 0);
+    }
+}
+
+static inline uint64_t get_tlb_pfn_from_entrylo(uint64_t entrylo)
+{
+#if defined(TARGET_MIPS64)
+    return extract64(entrylo, 6, 54);
+#else
+    return extract64(entrylo, 6, 24) | /* PFN */
+           (extract64(entrylo, 32, 32) << 24); /* PFNX */
+#endif
+}
+
+static void r4k_fill_tlb(CPUMIPSState *env, int idx)
+{
+    r4k_tlb_t *tlb;
+
+    /* XXX: detect conflicting TLBs and raise a MCHECK exception when needed */
+    tlb = &env->tlb->mmu.r4k.tlb[idx];
+    if (env->CP0_EntryHi & (1 << CP0EnHi_EHINV)) {
+        tlb->EHINV = 1;
+        return;
+    }
+    tlb->EHINV = 0;
+    tlb->VPN = env->CP0_EntryHi & (TARGET_PAGE_MASK << 1);
+#if defined(TARGET_MIPS64)
+    tlb->VPN &= env->SEGMask;
+#endif
+    tlb->ASID = env->CP0_EntryHi & 0xFF;
+    tlb->PageMask = env->CP0_PageMask;
+    tlb->G = env->CP0_EntryLo0 & env->CP0_EntryLo1 & 1;
+    tlb->V0 = (env->CP0_EntryLo0 & 2) != 0;
+    tlb->D0 = (env->CP0_EntryLo0 & 4) != 0;
+    tlb->C0 = (env->CP0_EntryLo0 >> 3) & 0x7;
+    tlb->XI0 = (env->CP0_EntryLo0 >> CP0EnLo_XI) & 1;
+    tlb->RI0 = (env->CP0_EntryLo0 >> CP0EnLo_RI) & 1;
+    tlb->PFN[0] = get_tlb_pfn_from_entrylo(env->CP0_EntryLo0) << 12;
+    tlb->V1 = (env->CP0_EntryLo1 & 2) != 0;
+    tlb->D1 = (env->CP0_EntryLo1 & 4) != 0;
+    tlb->C1 = (env->CP0_EntryLo1 >> 3) & 0x7;
+    tlb->XI1 = (env->CP0_EntryLo1 >> CP0EnLo_XI) & 1;
+    tlb->RI1 = (env->CP0_EntryLo1 >> CP0EnLo_RI) & 1;
+    tlb->PFN[1] = get_tlb_pfn_from_entrylo(env->CP0_EntryLo1) << 12;
+}
+
+void r4k_helper_tlbinv(CPUMIPSState *env)
+{
+    int idx;
+    r4k_tlb_t *tlb;
+    uint8_t ASID = env->CP0_EntryHi & 0xFF;
+
+    for (idx = 0; idx < env->tlb->nb_tlb; idx++) {
+        tlb = &env->tlb->mmu.r4k.tlb[idx];
+        if (!tlb->G && tlb->ASID == ASID) {
+            tlb->EHINV = 1;
+        }
+    }
+    cpu_mips_tlb_flush(env, 1);
+}
+
+void r4k_helper_tlbinvf(CPUMIPSState *env)
+{
+    int idx;
+
+    for (idx = 0; idx < env->tlb->nb_tlb; idx++) {
+        env->tlb->mmu.r4k.tlb[idx].EHINV = 1;
+    }
+    cpu_mips_tlb_flush(env, 1);
+}
+
+void r4k_helper_tlbwi(CPUMIPSState *env)
+{
+    r4k_tlb_t *tlb;
+    int idx;
+    target_ulong VPN;
+    uint8_t ASID;
+    bool G, V0, D0, V1, D1;
+
+    idx = (env->CP0_Index & ~0x80000000) % env->tlb->nb_tlb;
+    tlb = &env->tlb->mmu.r4k.tlb[idx];
+    VPN = env->CP0_EntryHi & (TARGET_PAGE_MASK << 1);
+#if defined(TARGET_MIPS64)
+    VPN &= env->SEGMask;
+#endif
+    ASID = env->CP0_EntryHi & 0xff;
+    G = env->CP0_EntryLo0 & env->CP0_EntryLo1 & 1;
+    V0 = (env->CP0_EntryLo0 & 2) != 0;
+    D0 = (env->CP0_EntryLo0 & 4) != 0;
+    V1 = (env->CP0_EntryLo1 & 2) != 0;
+    D1 = (env->CP0_EntryLo1 & 4) != 0;
+
+    /* Discard cached TLB entries, unless tlbwi is just upgrading access
+       permissions on the current entry. */
+    if (tlb->VPN != VPN || tlb->ASID != ASID || tlb->G != G ||
+        (tlb->V0 && !V0) || (tlb->D0 && !D0) ||
+        (tlb->V1 && !V1) || (tlb->D1 && !D1)) {
+        r4k_mips_tlb_flush_extra(env, env->tlb->nb_tlb);
+    }
+
+    r4k_invalidate_tlb(env, idx, 0);
+    r4k_fill_tlb(env, idx);
+}
+
+void r4k_helper_tlbwr(CPUMIPSState *env)
+{
+    int r = cpu_mips_get_random(env);
+
+    r4k_invalidate_tlb(env, r, 1);
+    r4k_fill_tlb(env, r);
+}
+
+void r4k_helper_tlbp(CPUMIPSState *env)
+{
+    r4k_tlb_t *tlb;
+    target_ulong mask;
+    target_ulong tag;
+    target_ulong VPN;
+    uint8_t ASID;
+    int i;
+
+    ASID = env->CP0_EntryHi & 0xFF;
+    for (i = 0; i < env->tlb->nb_tlb; i++) {
+        tlb = &env->tlb->mmu.r4k.tlb[i];
+        /* 1k pages are not supported. */
+        mask = tlb->PageMask | ~(TARGET_PAGE_MASK << 1);
+        tag = env->CP0_EntryHi & ~mask;
+        VPN = tlb->VPN & ~mask;
+#if defined(TARGET_MIPS64)
+        tag &= env->SEGMask;
+#endif
+        /* Check ASID, virtual page number & size */
+        if ((tlb->G == 1 || tlb->ASID == ASID) && VPN == tag && !tlb->EHINV) {
+            /* TLB match */
+            env->CP0_Index = i;
+            break;
+        }
+    }
+    if (i == env->tlb->nb_tlb) {
+        /* No match.  Discard any shadow entries, if any of them match.  */
+        for (i = env->tlb->nb_tlb; i < env->tlb->tlb_in_use; i++) {
+            tlb = &env->tlb->mmu.r4k.tlb[i];
+            /* 1k pages are not supported. */
+            mask = tlb->PageMask | ~(TARGET_PAGE_MASK << 1);
+            tag = env->CP0_EntryHi & ~mask;
+            VPN = tlb->VPN & ~mask;
+#if defined(TARGET_MIPS64)
+            tag &= env->SEGMask;
+#endif
+            /* Check ASID, virtual page number & size */
+            if ((tlb->G == 1 || tlb->ASID == ASID) && VPN == tag) {
+                r4k_mips_tlb_flush_extra (env, i);
+                break;
+            }
+        }
+
+        env->CP0_Index |= 0x80000000;
+    }
+}
+
+static inline uint64_t get_entrylo_pfn_from_tlb(uint64_t tlb_pfn)
+{
+#if defined(TARGET_MIPS64)
+    return tlb_pfn << 6;
+#else
+    return (extract64(tlb_pfn, 0, 24) << 6) | /* PFN */
+           (extract64(tlb_pfn, 24, 32) << 32); /* PFNX */
+#endif
+}
+
+void r4k_helper_tlbr(CPUMIPSState *env)
+{
+    r4k_tlb_t *tlb;
+    uint8_t ASID;
+    int idx;
+
+    ASID = env->CP0_EntryHi & 0xFF;
+    idx = (env->CP0_Index & ~0x80000000) % env->tlb->nb_tlb;
+    tlb = &env->tlb->mmu.r4k.tlb[idx];
+
+    /* If this will change the current ASID, flush qemu's TLB.  */
+    if (ASID != tlb->ASID)
+        cpu_mips_tlb_flush (env, 1);
+
+    r4k_mips_tlb_flush_extra(env, env->tlb->nb_tlb);
+
+    if (tlb->EHINV) {
+        env->CP0_EntryHi = 1 << CP0EnHi_EHINV;
+        env->CP0_PageMask = 0;
+        env->CP0_EntryLo0 = 0;
+        env->CP0_EntryLo1 = 0;
+    } else {
+        env->CP0_EntryHi = tlb->VPN | tlb->ASID;
+        env->CP0_PageMask = tlb->PageMask;
+        env->CP0_EntryLo0 = tlb->G | (tlb->V0 << 1) | (tlb->D0 << 2) |
+                        ((uint64_t)tlb->RI0 << CP0EnLo_RI) |
+                        ((uint64_t)tlb->XI0 << CP0EnLo_XI) | (tlb->C0 << 3) |
+                        get_entrylo_pfn_from_tlb(tlb->PFN[0] >> 12);
+        env->CP0_EntryLo1 = tlb->G | (tlb->V1 << 1) | (tlb->D1 << 2) |
+                        ((uint64_t)tlb->RI1 << CP0EnLo_RI) |
+                        ((uint64_t)tlb->XI1 << CP0EnLo_XI) | (tlb->C1 << 3) |
+                        get_entrylo_pfn_from_tlb(tlb->PFN[1] >> 12);
+    }
+}
+
+void helper_tlbwi(CPUMIPSState *env)
+{
+    env->tlb->helper_tlbwi(env);
+}
+
+void helper_tlbwr(CPUMIPSState *env)
+{
+    env->tlb->helper_tlbwr(env);
+}
+
+void helper_tlbp(CPUMIPSState *env)
+{
+    env->tlb->helper_tlbp(env);
+}
+
+void helper_tlbr(CPUMIPSState *env)
+{
+    env->tlb->helper_tlbr(env);
+}
+
+void helper_tlbinv(CPUMIPSState *env)
+{
+    env->tlb->helper_tlbinv(env);
+}
+
+void helper_tlbinvf(CPUMIPSState *env)
+{
+    env->tlb->helper_tlbinvf(env);
+}
+
+/* Specials */
+target_ulong helper_di(CPUMIPSState *env)
+{
+    target_ulong t0 = env->CP0_Status;
+
+    env->CP0_Status = t0 & ~(1 << CP0St_IE);
+    return t0;
+}
+
+target_ulong helper_ei(CPUMIPSState *env)
+{
+    target_ulong t0 = env->CP0_Status;
+
+    env->CP0_Status = t0 | (1 << CP0St_IE);
+    return t0;
+}
+
+static void debug_pre_eret(CPUMIPSState *env)
+{
+    if (qemu_loglevel_mask(CPU_LOG_EXEC)) {
+        qemu_log("ERET: PC " TARGET_FMT_lx " EPC " TARGET_FMT_lx,
+                env->active_tc.PC, env->CP0_EPC);
+        if (env->CP0_Status & (1 << CP0St_ERL))
+            qemu_log(" ErrorEPC " TARGET_FMT_lx, env->CP0_ErrorEPC);
+        if (env->hflags & MIPS_HFLAG_DM)
+            qemu_log(" DEPC " TARGET_FMT_lx, env->CP0_DEPC);
+        qemu_log("\n");
+    }
+}
+
+static void debug_post_eret(CPUMIPSState *env)
+{
+    MIPSCPU *cpu = mips_env_get_cpu(env);
+
+    if (qemu_loglevel_mask(CPU_LOG_EXEC)) {
+        qemu_log("  =>  PC " TARGET_FMT_lx " EPC " TARGET_FMT_lx,
+                env->active_tc.PC, env->CP0_EPC);
+        if (env->CP0_Status & (1 << CP0St_ERL))
+            qemu_log(" ErrorEPC " TARGET_FMT_lx, env->CP0_ErrorEPC);
+        if (env->hflags & MIPS_HFLAG_DM)
+            qemu_log(" DEPC " TARGET_FMT_lx, env->CP0_DEPC);
+        switch (env->hflags & MIPS_HFLAG_KSU) {
+        case MIPS_HFLAG_UM: qemu_log(", UM\n"); break;
+        case MIPS_HFLAG_SM: qemu_log(", SM\n"); break;
+        case MIPS_HFLAG_KM: qemu_log("\n"); break;
+        default:
+            cpu_abort(CPU(cpu), "Invalid MMU mode!\n");
+            break;
+        }
+    }
+}
+
+static void set_pc(CPUMIPSState *env, target_ulong error_pc)
+{
+    env->active_tc.PC = error_pc & ~(target_ulong)1;
+    if (error_pc & 1) {
+        env->hflags |= MIPS_HFLAG_M16;
+    } else {
+        env->hflags &= ~(MIPS_HFLAG_M16);
+    }
+}
+
+static inline void exception_return(CPUMIPSState *env)
+{
+    debug_pre_eret(env);
+    if (env->CP0_Status & (1 << CP0St_ERL)) {
+        set_pc(env, env->CP0_ErrorEPC);
+        env->CP0_Status &= ~(1 << CP0St_ERL);
+    } else {
+        set_pc(env, env->CP0_EPC);
+        env->CP0_Status &= ~(1 << CP0St_EXL);
+    }
+    compute_hflags(env);
+    debug_post_eret(env);
+}
+
+void helper_eret(CPUMIPSState *env)
+{
+    exception_return(env);
+    env->lladdr = 1;
+}
+
+void helper_eretnc(CPUMIPSState *env)
+{
+    exception_return(env);
+}
+
+void helper_deret(CPUMIPSState *env)
+{
+    debug_pre_eret(env);
+    set_pc(env, env->CP0_DEPC);
+
+    env->hflags &= ~MIPS_HFLAG_DM;
+    compute_hflags(env);
+    debug_post_eret(env);
+}
+#endif /* !CONFIG_USER_ONLY */
+
+target_ulong helper_rdhwr_cpunum(CPUMIPSState *env)
+{
+    if ((env->hflags & MIPS_HFLAG_CP0) ||
+        (env->CP0_HWREna & (1 << 0)))
+        return env->CP0_EBase & 0x3ff;
+    else
+        helper_raise_exception(env, EXCP_RI);
+
+    return 0;
+}
+
+target_ulong helper_rdhwr_synci_step(CPUMIPSState *env)
+{
+    if ((env->hflags & MIPS_HFLAG_CP0) ||
+        (env->CP0_HWREna & (1 << 1)))
+        return env->SYNCI_Step;
+    else
+        helper_raise_exception(env, EXCP_RI);
+
+    return 0;
+}
+
+target_ulong helper_rdhwr_cc(CPUMIPSState *env)
+{
+    if ((env->hflags & MIPS_HFLAG_CP0) ||
+        (env->CP0_HWREna & (1 << 2)))
+        return env->CP0_Count;
+    else
+        helper_raise_exception(env, EXCP_RI);
+
+    return 0;
+}
+
+target_ulong helper_rdhwr_ccres(CPUMIPSState *env)
+{
+    if ((env->hflags & MIPS_HFLAG_CP0) ||
+        (env->CP0_HWREna & (1 << 3)))
+        return env->CCRes;
+    else
+        helper_raise_exception(env, EXCP_RI);
+
+    return 0;
+}
+
+void helper_pmon(CPUMIPSState *env, int function)
+{
+    function /= 2;
+    switch (function) {
+    case 2: /* TODO: char inbyte(int waitflag); */
+        if (env->active_tc.gpr[4] == 0)
+            env->active_tc.gpr[2] = -1;
+        /* Fall through */
+    case 11: /* TODO: char inbyte (void); */
+        env->active_tc.gpr[2] = -1;
+        break;
+    case 3:
+    case 12:
+        printf("%c", (char)(env->active_tc.gpr[4] & 0xFF));
+        break;
+    case 17:
+        break;
+    case 158:
+        {
+            unsigned char *fmt = (void *)(uintptr_t)env->active_tc.gpr[4];
+            printf("%s", fmt);
+        }
+        break;
+    }
+}
+
+void helper_wait(CPUMIPSState *env)
+{
+    CPUState *cs = CPU(mips_env_get_cpu(env));
+
+    cs->halted = 1;
+    cpu_reset_interrupt(cs, CPU_INTERRUPT_WAKE);
+    helper_raise_exception(env, EXCP_HLT);
+}
+
+#if !defined(CONFIG_USER_ONLY)
+
+void mips_cpu_do_unaligned_access(CPUState *cs, vaddr addr,
+                                  int access_type, int is_user,
+                                  uintptr_t retaddr)
+{
+    MIPSCPU *cpu = MIPS_CPU(cs);
+    CPUMIPSState *env = &cpu->env;
+    int error_code = 0;
+    int excp;
+
+    env->CP0_BadVAddr = addr;
+
+    if (access_type == MMU_DATA_STORE) {
+        excp = EXCP_AdES;
+    } else {
+        excp = EXCP_AdEL;
+        if (access_type == MMU_INST_FETCH) {
+            error_code |= EXCP_INST_NOTAVAIL;
+        }
+    }
+
+    do_raise_exception_err(env, excp, error_code, retaddr);
+}
+
+void tlb_fill(CPUState *cs, target_ulong addr, int is_write, int mmu_idx,
+              uintptr_t retaddr)
+{
+    int ret;
+
+    ret = mips_cpu_handle_mmu_fault(cs, addr, is_write, mmu_idx);
+    if (ret) {
+        MIPSCPU *cpu = MIPS_CPU(cs);
+        CPUMIPSState *env = &cpu->env;
+
+        do_raise_exception_err(env, cs->exception_index,
+                               env->error_code, retaddr);
+    }
+}
+
+void mips_cpu_unassigned_access(CPUState *cs, hwaddr addr,
+                                bool is_write, bool is_exec, int unused,
+                                unsigned size)
+{
+    MIPSCPU *cpu = MIPS_CPU(cs);
+    CPUMIPSState *env = &cpu->env;
+
+    /*
+     * Raising an exception with KVM enabled will crash because it won't be from
+     * the main execution loop so the longjmp won't have a matching setjmp.
+     * Until we can trigger a bus error exception through KVM lets just ignore
+     * the access.
+     */
+    if (kvm_enabled()) {
+        return;
+    }
+
+    if (is_exec) {
+        helper_raise_exception(env, EXCP_IBE);
+    } else {
+        helper_raise_exception(env, EXCP_DBE);
+    }
+}
+#endif /* !CONFIG_USER_ONLY */
+
+/* Complex FPU operations which may need stack space. */
+
+#define FLOAT_TWO32 make_float32(1 << 30)
+#define FLOAT_TWO64 make_float64(1ULL << 62)
+#define FP_TO_INT32_OVERFLOW 0x7fffffff
+#define FP_TO_INT64_OVERFLOW 0x7fffffffffffffffULL
+
+/* convert MIPS rounding mode in FCR31 to IEEE library */
+unsigned int ieee_rm[] = {
+    float_round_nearest_even,
+    float_round_to_zero,
+    float_round_up,
+    float_round_down
+};
+
+target_ulong helper_cfc1(CPUMIPSState *env, uint32_t reg)
+{
+    target_ulong arg1 = 0;
+
+    switch (reg) {
+    case 0:
+        arg1 = (int32_t)env->active_fpu.fcr0;
+        break;
+    case 1:
+        /* UFR Support - Read Status FR */
+        if (env->active_fpu.fcr0 & (1 << FCR0_UFRP)) {
+            if (env->CP0_Config5 & (1 << CP0C5_UFR)) {
+                arg1 = (int32_t)
+                       ((env->CP0_Status & (1  << CP0St_FR)) >> CP0St_FR);
+            } else {
+                helper_raise_exception(env, EXCP_RI);
+            }
+        }
+        break;
+    case 5:
+        /* FRE Support - read Config5.FRE bit */
+        if (env->active_fpu.fcr0 & (1 << FCR0_FREP)) {
+            if (env->CP0_Config5 & (1 << CP0C5_UFE)) {
+                arg1 = (env->CP0_Config5 >> CP0C5_FRE) & 1;
+            } else {
+                helper_raise_exception(env, EXCP_RI);
+            }
+        }
+        break;
+    case 25:
+        arg1 = ((env->active_fpu.fcr31 >> 24) & 0xfe) | ((env->active_fpu.fcr31 >> 23) & 0x1);
+        break;
+    case 26:
+        arg1 = env->active_fpu.fcr31 & 0x0003f07c;
+        break;
+    case 28:
+        arg1 = (env->active_fpu.fcr31 & 0x00000f83) | ((env->active_fpu.fcr31 >> 22) & 0x4);
+        break;
+    default:
+        arg1 = (int32_t)env->active_fpu.fcr31;
+        break;
+    }
+
+    return arg1;
+}
+
+void helper_ctc1(CPUMIPSState *env, target_ulong arg1, uint32_t fs, uint32_t rt)
+{
+    switch (fs) {
+    case 1:
+        /* UFR Alias - Reset Status FR */
+        if (!((env->active_fpu.fcr0 & (1 << FCR0_UFRP)) && (rt == 0))) {
+            return;
+        }
+        if (env->CP0_Config5 & (1 << CP0C5_UFR)) {
+            env->CP0_Status &= ~(1 << CP0St_FR);
+            compute_hflags(env);
+        } else {
+            helper_raise_exception(env, EXCP_RI);
+        }
+        break;
+    case 4:
+        /* UNFR Alias - Set Status FR */
+        if (!((env->active_fpu.fcr0 & (1 << FCR0_UFRP)) && (rt == 0))) {
+            return;
+        }
+        if (env->CP0_Config5 & (1 << CP0C5_UFR)) {
+            env->CP0_Status |= (1 << CP0St_FR);
+            compute_hflags(env);
+        } else {
+            helper_raise_exception(env, EXCP_RI);
+        }
+        break;
+    case 5:
+        /* FRE Support - clear Config5.FRE bit */
+        if (!((env->active_fpu.fcr0 & (1 << FCR0_FREP)) && (rt == 0))) {
+            return;
+        }
+        if (env->CP0_Config5 & (1 << CP0C5_UFE)) {
+            env->CP0_Config5 &= ~(1 << CP0C5_FRE);
+            compute_hflags(env);
+        } else {
+            helper_raise_exception(env, EXCP_RI);
+        }
+        break;
+    case 6:
+        /* FRE Support - set Config5.FRE bit */
+        if (!((env->active_fpu.fcr0 & (1 << FCR0_FREP)) && (rt == 0))) {
+            return;
+        }
+        if (env->CP0_Config5 & (1 << CP0C5_UFE)) {
+            env->CP0_Config5 |= (1 << CP0C5_FRE);
+            compute_hflags(env);
+        } else {
+            helper_raise_exception(env, EXCP_RI);
+        }
+        break;
+    case 25:
+        if ((env->insn_flags & ISA_MIPS32R6) || (arg1 & 0xffffff00)) {
+            return;
+        }
+        env->active_fpu.fcr31 = (env->active_fpu.fcr31 & 0x017fffff) | ((arg1 & 0xfe) << 24) |
+                     ((arg1 & 0x1) << 23);
+        break;
+    case 26:
+        if (arg1 & 0x007c0000)
+            return;
+        env->active_fpu.fcr31 = (env->active_fpu.fcr31 & 0xfffc0f83) | (arg1 & 0x0003f07c);
+        break;
+    case 28:
+        if (arg1 & 0x007c0000)
+            return;
+        env->active_fpu.fcr31 = (env->active_fpu.fcr31 & 0xfefff07c) | (arg1 & 0x00000f83) |
+                     ((arg1 & 0x4) << 22);
+        break;
+    case 31:
+        if (env->insn_flags & ISA_MIPS32R6) {
+            uint32_t mask = 0xfefc0000;
+            env->active_fpu.fcr31 = (arg1 & ~mask) |
+                (env->active_fpu.fcr31 & mask);
+        } else if (!(arg1 & 0x007c0000)) {
+            env->active_fpu.fcr31 = arg1;
+        }
+        break;
+    default:
+        return;
+    }
+    /* set rounding mode */
+    restore_rounding_mode(env);
+    /* set flush-to-zero mode */
+    restore_flush_mode(env);
+    set_float_exception_flags(0, &env->active_fpu.fp_status);
+    if ((GET_FP_ENABLE(env->active_fpu.fcr31) | 0x20) & GET_FP_CAUSE(env->active_fpu.fcr31))
+        do_raise_exception(env, EXCP_FPE, GETPC());
+}
+
+int ieee_ex_to_mips(int xcpt)
+{
+    int ret = 0;
+    if (xcpt) {
+        if (xcpt & float_flag_invalid) {
+            ret |= FP_INVALID;
+        }
+        if (xcpt & float_flag_overflow) {
+            ret |= FP_OVERFLOW;
+        }
+        if (xcpt & float_flag_underflow) {
+            ret |= FP_UNDERFLOW;
+        }
+        if (xcpt & float_flag_divbyzero) {
+            ret |= FP_DIV0;
+        }
+        if (xcpt & float_flag_inexact) {
+            ret |= FP_INEXACT;
+        }
+    }
+    return ret;
+}
+
+static inline void update_fcr31(CPUMIPSState *env, uintptr_t pc)
+{
+    int tmp = ieee_ex_to_mips(get_float_exception_flags(&env->active_fpu.fp_status));
+
+    SET_FP_CAUSE(env->active_fpu.fcr31, tmp);
+
+    if (tmp) {
+        set_float_exception_flags(0, &env->active_fpu.fp_status);
+
+        if (GET_FP_ENABLE(env->active_fpu.fcr31) & tmp) {
+            do_raise_exception(env, EXCP_FPE, pc);
+        } else {
+            UPDATE_FP_FLAGS(env->active_fpu.fcr31, tmp);
+        }
+    }
+}
+
+/* Float support.
+   Single precition routines have a "s" suffix, double precision a
+   "d" suffix, 32bit integer "w", 64bit integer "l", paired single "ps",
+   paired single lower "pl", paired single upper "pu".  */
+
+/* unary operations, modifying fp status  */
+uint64_t helper_float_sqrt_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    fdt0 = float64_sqrt(fdt0, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fdt0;
+}
+
+uint32_t helper_float_sqrt_s(CPUMIPSState *env, uint32_t fst0)
+{
+    fst0 = float32_sqrt(fst0, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fst0;
+}
+
+uint64_t helper_float_cvtd_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint64_t fdt2;
+
+    fdt2 = float32_to_float64(fst0, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fdt2;
+}
+
+uint64_t helper_float_cvtd_w(CPUMIPSState *env, uint32_t wt0)
+{
+    uint64_t fdt2;
+
+    fdt2 = int32_to_float64(wt0, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fdt2;
+}
+
+uint64_t helper_float_cvtd_l(CPUMIPSState *env, uint64_t dt0)
+{
+    uint64_t fdt2;
+
+    fdt2 = int64_to_float64(dt0, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fdt2;
+}
+
+uint64_t helper_float_cvtl_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint64_t dt2;
+
+    dt2 = float64_to_int64(fdt0, &env->active_fpu.fp_status);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        dt2 = FP_TO_INT64_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return dt2;
+}
+
+uint64_t helper_float_cvtl_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint64_t dt2;
+
+    dt2 = float32_to_int64(fst0, &env->active_fpu.fp_status);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        dt2 = FP_TO_INT64_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return dt2;
+}
+
+uint64_t helper_float_cvtps_pw(CPUMIPSState *env, uint64_t dt0)
+{
+    uint32_t fst2;
+    uint32_t fsth2;
+
+    fst2 = int32_to_float32(dt0 & 0XFFFFFFFF, &env->active_fpu.fp_status);
+    fsth2 = int32_to_float32(dt0 >> 32, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return ((uint64_t)fsth2 << 32) | fst2;
+}
+
+uint64_t helper_float_cvtpw_ps(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint32_t wt2;
+    uint32_t wth2;
+    int excp, excph;
+
+    wt2 = float32_to_int32(fdt0 & 0XFFFFFFFF, &env->active_fpu.fp_status);
+    excp = get_float_exception_flags(&env->active_fpu.fp_status);
+    if (excp & (float_flag_overflow | float_flag_invalid)) {
+        wt2 = FP_TO_INT32_OVERFLOW;
+    }
+
+    set_float_exception_flags(0, &env->active_fpu.fp_status);
+    wth2 = float32_to_int32(fdt0 >> 32, &env->active_fpu.fp_status);
+    excph = get_float_exception_flags(&env->active_fpu.fp_status);
+    if (excph & (float_flag_overflow | float_flag_invalid)) {
+        wth2 = FP_TO_INT32_OVERFLOW;
+    }
+
+    set_float_exception_flags(excp | excph, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+
+    return ((uint64_t)wth2 << 32) | wt2;
+}
+
+uint32_t helper_float_cvts_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint32_t fst2;
+
+    fst2 = float64_to_float32(fdt0, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fst2;
+}
+
+uint32_t helper_float_cvts_w(CPUMIPSState *env, uint32_t wt0)
+{
+    uint32_t fst2;
+
+    fst2 = int32_to_float32(wt0, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fst2;
+}
+
+uint32_t helper_float_cvts_l(CPUMIPSState *env, uint64_t dt0)
+{
+    uint32_t fst2;
+
+    fst2 = int64_to_float32(dt0, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fst2;
+}
+
+uint32_t helper_float_cvts_pl(CPUMIPSState *env, uint32_t wt0)
+{
+    uint32_t wt2;
+
+    wt2 = wt0;
+    update_fcr31(env, GETPC());
+    return wt2;
+}
+
+uint32_t helper_float_cvts_pu(CPUMIPSState *env, uint32_t wth0)
+{
+    uint32_t wt2;
+
+    wt2 = wth0;
+    update_fcr31(env, GETPC());
+    return wt2;
+}
+
+uint32_t helper_float_cvtw_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint32_t wt2;
+
+    wt2 = float32_to_int32(fst0, &env->active_fpu.fp_status);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        wt2 = FP_TO_INT32_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return wt2;
+}
+
+uint32_t helper_float_cvtw_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint32_t wt2;
+
+    wt2 = float64_to_int32(fdt0, &env->active_fpu.fp_status);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        wt2 = FP_TO_INT32_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return wt2;
+}
+
+uint64_t helper_float_roundl_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint64_t dt2;
+
+    set_float_rounding_mode(float_round_nearest_even, &env->active_fpu.fp_status);
+    dt2 = float64_to_int64(fdt0, &env->active_fpu.fp_status);
+    restore_rounding_mode(env);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        dt2 = FP_TO_INT64_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return dt2;
+}
+
+uint64_t helper_float_roundl_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint64_t dt2;
+
+    set_float_rounding_mode(float_round_nearest_even, &env->active_fpu.fp_status);
+    dt2 = float32_to_int64(fst0, &env->active_fpu.fp_status);
+    restore_rounding_mode(env);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        dt2 = FP_TO_INT64_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return dt2;
+}
+
+uint32_t helper_float_roundw_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint32_t wt2;
+
+    set_float_rounding_mode(float_round_nearest_even, &env->active_fpu.fp_status);
+    wt2 = float64_to_int32(fdt0, &env->active_fpu.fp_status);
+    restore_rounding_mode(env);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        wt2 = FP_TO_INT32_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return wt2;
+}
+
+uint32_t helper_float_roundw_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint32_t wt2;
+
+    set_float_rounding_mode(float_round_nearest_even, &env->active_fpu.fp_status);
+    wt2 = float32_to_int32(fst0, &env->active_fpu.fp_status);
+    restore_rounding_mode(env);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        wt2 = FP_TO_INT32_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return wt2;
+}
+
+uint64_t helper_float_truncl_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint64_t dt2;
+
+    dt2 = float64_to_int64_round_to_zero(fdt0, &env->active_fpu.fp_status);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        dt2 = FP_TO_INT64_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return dt2;
+}
+
+uint64_t helper_float_truncl_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint64_t dt2;
+
+    dt2 = float32_to_int64_round_to_zero(fst0, &env->active_fpu.fp_status);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        dt2 = FP_TO_INT64_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return dt2;
+}
+
+uint32_t helper_float_truncw_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint32_t wt2;
+
+    wt2 = float64_to_int32_round_to_zero(fdt0, &env->active_fpu.fp_status);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        wt2 = FP_TO_INT32_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return wt2;
+}
+
+uint32_t helper_float_truncw_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint32_t wt2;
+
+    wt2 = float32_to_int32_round_to_zero(fst0, &env->active_fpu.fp_status);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        wt2 = FP_TO_INT32_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return wt2;
+}
+
+uint64_t helper_float_ceill_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint64_t dt2;
+
+    set_float_rounding_mode(float_round_up, &env->active_fpu.fp_status);
+    dt2 = float64_to_int64(fdt0, &env->active_fpu.fp_status);
+    restore_rounding_mode(env);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        dt2 = FP_TO_INT64_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return dt2;
+}
+
+uint64_t helper_float_ceill_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint64_t dt2;
+
+    set_float_rounding_mode(float_round_up, &env->active_fpu.fp_status);
+    dt2 = float32_to_int64(fst0, &env->active_fpu.fp_status);
+    restore_rounding_mode(env);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        dt2 = FP_TO_INT64_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return dt2;
+}
+
+uint32_t helper_float_ceilw_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint32_t wt2;
+
+    set_float_rounding_mode(float_round_up, &env->active_fpu.fp_status);
+    wt2 = float64_to_int32(fdt0, &env->active_fpu.fp_status);
+    restore_rounding_mode(env);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        wt2 = FP_TO_INT32_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return wt2;
+}
+
+uint32_t helper_float_ceilw_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint32_t wt2;
+
+    set_float_rounding_mode(float_round_up, &env->active_fpu.fp_status);
+    wt2 = float32_to_int32(fst0, &env->active_fpu.fp_status);
+    restore_rounding_mode(env);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        wt2 = FP_TO_INT32_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return wt2;
+}
+
+uint64_t helper_float_floorl_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint64_t dt2;
+
+    set_float_rounding_mode(float_round_down, &env->active_fpu.fp_status);
+    dt2 = float64_to_int64(fdt0, &env->active_fpu.fp_status);
+    restore_rounding_mode(env);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        dt2 = FP_TO_INT64_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return dt2;
+}
+
+uint64_t helper_float_floorl_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint64_t dt2;
+
+    set_float_rounding_mode(float_round_down, &env->active_fpu.fp_status);
+    dt2 = float32_to_int64(fst0, &env->active_fpu.fp_status);
+    restore_rounding_mode(env);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        dt2 = FP_TO_INT64_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return dt2;
+}
+
+uint32_t helper_float_floorw_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint32_t wt2;
+
+    set_float_rounding_mode(float_round_down, &env->active_fpu.fp_status);
+    wt2 = float64_to_int32(fdt0, &env->active_fpu.fp_status);
+    restore_rounding_mode(env);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        wt2 = FP_TO_INT32_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return wt2;
+}
+
+uint32_t helper_float_floorw_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint32_t wt2;
+
+    set_float_rounding_mode(float_round_down, &env->active_fpu.fp_status);
+    wt2 = float32_to_int32(fst0, &env->active_fpu.fp_status);
+    restore_rounding_mode(env);
+    if (get_float_exception_flags(&env->active_fpu.fp_status)
+        & (float_flag_invalid | float_flag_overflow)) {
+        wt2 = FP_TO_INT32_OVERFLOW;
+    }
+    update_fcr31(env, GETPC());
+    return wt2;
+}
+
+/* unary operations, not modifying fp status  */
+#define FLOAT_UNOP(name)                                       \
+uint64_t helper_float_ ## name ## _d(uint64_t fdt0)                \
+{                                                              \
+    return float64_ ## name(fdt0);                             \
+}                                                              \
+uint32_t helper_float_ ## name ## _s(uint32_t fst0)                \
+{                                                              \
+    return float32_ ## name(fst0);                             \
+}                                                              \
+uint64_t helper_float_ ## name ## _ps(uint64_t fdt0)               \
+{                                                              \
+    uint32_t wt0;                                              \
+    uint32_t wth0;                                             \
+                                                               \
+    wt0 = float32_ ## name(fdt0 & 0XFFFFFFFF);                 \
+    wth0 = float32_ ## name(fdt0 >> 32);                       \
+    return ((uint64_t)wth0 << 32) | wt0;                       \
+}
+FLOAT_UNOP(abs)
+FLOAT_UNOP(chs)
+#undef FLOAT_UNOP
+
+/* MIPS specific unary operations */
+uint64_t helper_float_recip_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint64_t fdt2;
+
+    fdt2 = float64_div(float64_one, fdt0, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fdt2;
+}
+
+uint32_t helper_float_recip_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint32_t fst2;
+
+    fst2 = float32_div(float32_one, fst0, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fst2;
+}
+
+uint64_t helper_float_rsqrt_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint64_t fdt2;
+
+    fdt2 = float64_sqrt(fdt0, &env->active_fpu.fp_status);
+    fdt2 = float64_div(float64_one, fdt2, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fdt2;
+}
+
+uint32_t helper_float_rsqrt_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint32_t fst2;
+
+    fst2 = float32_sqrt(fst0, &env->active_fpu.fp_status);
+    fst2 = float32_div(float32_one, fst2, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fst2;
+}
+
+uint64_t helper_float_recip1_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint64_t fdt2;
+
+    fdt2 = float64_div(float64_one, fdt0, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fdt2;
+}
+
+uint32_t helper_float_recip1_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint32_t fst2;
+
+    fst2 = float32_div(float32_one, fst0, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fst2;
+}
+
+uint64_t helper_float_recip1_ps(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint32_t fst2;
+    uint32_t fsth2;
+
+    fst2 = float32_div(float32_one, fdt0 & 0XFFFFFFFF, &env->active_fpu.fp_status);
+    fsth2 = float32_div(float32_one, fdt0 >> 32, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return ((uint64_t)fsth2 << 32) | fst2;
+}
+
+uint64_t helper_float_rsqrt1_d(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint64_t fdt2;
+
+    fdt2 = float64_sqrt(fdt0, &env->active_fpu.fp_status);
+    fdt2 = float64_div(float64_one, fdt2, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fdt2;
+}
+
+uint32_t helper_float_rsqrt1_s(CPUMIPSState *env, uint32_t fst0)
+{
+    uint32_t fst2;
+
+    fst2 = float32_sqrt(fst0, &env->active_fpu.fp_status);
+    fst2 = float32_div(float32_one, fst2, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return fst2;
+}
+
+uint64_t helper_float_rsqrt1_ps(CPUMIPSState *env, uint64_t fdt0)
+{
+    uint32_t fst2;
+    uint32_t fsth2;
+
+    fst2 = float32_sqrt(fdt0 & 0XFFFFFFFF, &env->active_fpu.fp_status);
+    fsth2 = float32_sqrt(fdt0 >> 32, &env->active_fpu.fp_status);
+    fst2 = float32_div(float32_one, fst2, &env->active_fpu.fp_status);
+    fsth2 = float32_div(float32_one, fsth2, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return ((uint64_t)fsth2 << 32) | fst2;
+}
+
+#define FLOAT_RINT(name, bits)                                              \
+uint ## bits ## _t helper_float_ ## name (CPUMIPSState *env,                \
+                                          uint ## bits ## _t fs)            \
+{                                                                           \
+    uint ## bits ## _t fdret;                                               \
+                                                                            \
+    fdret = float ## bits ## _round_to_int(fs, &env->active_fpu.fp_status); \
+    update_fcr31(env, GETPC());                                             \
+    return fdret;                                                           \
+}
+
+FLOAT_RINT(rint_s, 32)
+FLOAT_RINT(rint_d, 64)
+#undef FLOAT_RINT
+
+#define FLOAT_CLASS_SIGNALING_NAN      0x001
+#define FLOAT_CLASS_QUIET_NAN          0x002
+#define FLOAT_CLASS_NEGATIVE_INFINITY  0x004
+#define FLOAT_CLASS_NEGATIVE_NORMAL    0x008
+#define FLOAT_CLASS_NEGATIVE_SUBNORMAL 0x010
+#define FLOAT_CLASS_NEGATIVE_ZERO      0x020
+#define FLOAT_CLASS_POSITIVE_INFINITY  0x040
+#define FLOAT_CLASS_POSITIVE_NORMAL    0x080
+#define FLOAT_CLASS_POSITIVE_SUBNORMAL 0x100
+#define FLOAT_CLASS_POSITIVE_ZERO      0x200
+
+#define FLOAT_CLASS(name, bits)                                      \
+uint ## bits ## _t helper_float_ ## name (uint ## bits ## _t arg)    \
+{                                                                    \
+    if (float ## bits ## _is_signaling_nan(arg)) {                   \
+        return FLOAT_CLASS_SIGNALING_NAN;                            \
+    } else if (float ## bits ## _is_quiet_nan(arg)) {                \
+        return FLOAT_CLASS_QUIET_NAN;                                \
+    } else if (float ## bits ## _is_neg(arg)) {                      \
+        if (float ## bits ## _is_infinity(arg)) {                    \
+            return FLOAT_CLASS_NEGATIVE_INFINITY;                    \
+        } else if (float ## bits ## _is_zero(arg)) {                 \
+            return FLOAT_CLASS_NEGATIVE_ZERO;                        \
+        } else if (float ## bits ## _is_zero_or_denormal(arg)) {     \
+            return FLOAT_CLASS_NEGATIVE_SUBNORMAL;                   \
+        } else {                                                     \
+            return FLOAT_CLASS_NEGATIVE_NORMAL;                      \
+        }                                                            \
+    } else {                                                         \
+        if (float ## bits ## _is_infinity(arg)) {                    \
+            return FLOAT_CLASS_POSITIVE_INFINITY;                    \
+        } else if (float ## bits ## _is_zero(arg)) {                 \
+            return FLOAT_CLASS_POSITIVE_ZERO;                        \
+        } else if (float ## bits ## _is_zero_or_denormal(arg)) {     \
+            return FLOAT_CLASS_POSITIVE_SUBNORMAL;                   \
+        } else {                                                     \
+            return FLOAT_CLASS_POSITIVE_NORMAL;                      \
+        }                                                            \
+    }                                                                \
+}
+
+FLOAT_CLASS(class_s, 32)
+FLOAT_CLASS(class_d, 64)
+#undef FLOAT_CLASS
+
+/* binary operations */
+#define FLOAT_BINOP(name)                                          \
+uint64_t helper_float_ ## name ## _d(CPUMIPSState *env,            \
+                                     uint64_t fdt0, uint64_t fdt1) \
+{                                                                  \
+    uint64_t dt2;                                                  \
+                                                                   \
+    dt2 = float64_ ## name (fdt0, fdt1, &env->active_fpu.fp_status);     \
+    update_fcr31(env, GETPC());                                    \
+    return dt2;                                                    \
+}                                                                  \
+                                                                   \
+uint32_t helper_float_ ## name ## _s(CPUMIPSState *env,            \
+                                     uint32_t fst0, uint32_t fst1) \
+{                                                                  \
+    uint32_t wt2;                                                  \
+                                                                   \
+    wt2 = float32_ ## name (fst0, fst1, &env->active_fpu.fp_status);     \
+    update_fcr31(env, GETPC());                                    \
+    return wt2;                                                    \
+}                                                                  \
+                                                                   \
+uint64_t helper_float_ ## name ## _ps(CPUMIPSState *env,           \
+                                      uint64_t fdt0,               \
+                                      uint64_t fdt1)               \
+{                                                                  \
+    uint32_t fst0 = fdt0 & 0XFFFFFFFF;                             \
+    uint32_t fsth0 = fdt0 >> 32;                                   \
+    uint32_t fst1 = fdt1 & 0XFFFFFFFF;                             \
+    uint32_t fsth1 = fdt1 >> 32;                                   \
+    uint32_t wt2;                                                  \
+    uint32_t wth2;                                                 \
+                                                                   \
+    wt2 = float32_ ## name (fst0, fst1, &env->active_fpu.fp_status);     \
+    wth2 = float32_ ## name (fsth0, fsth1, &env->active_fpu.fp_status);  \
+    update_fcr31(env, GETPC());                                    \
+    return ((uint64_t)wth2 << 32) | wt2;                           \
+}
+
+FLOAT_BINOP(add)
+FLOAT_BINOP(sub)
+FLOAT_BINOP(mul)
+FLOAT_BINOP(div)
+#undef FLOAT_BINOP
+
+/* MIPS specific binary operations */
+uint64_t helper_float_recip2_d(CPUMIPSState *env, uint64_t fdt0, uint64_t fdt2)
+{
+    fdt2 = float64_mul(fdt0, fdt2, &env->active_fpu.fp_status);
+    fdt2 = float64_chs(float64_sub(fdt2, float64_one, &env->active_fpu.fp_status));
+    update_fcr31(env, GETPC());
+    return fdt2;
+}
+
+uint32_t helper_float_recip2_s(CPUMIPSState *env, uint32_t fst0, uint32_t fst2)
+{
+    fst2 = float32_mul(fst0, fst2, &env->active_fpu.fp_status);
+    fst2 = float32_chs(float32_sub(fst2, float32_one, &env->active_fpu.fp_status));
+    update_fcr31(env, GETPC());
+    return fst2;
+}
+
+uint64_t helper_float_recip2_ps(CPUMIPSState *env, uint64_t fdt0, uint64_t fdt2)
+{
+    uint32_t fst0 = fdt0 & 0XFFFFFFFF;
+    uint32_t fsth0 = fdt0 >> 32;
+    uint32_t fst2 = fdt2 & 0XFFFFFFFF;
+    uint32_t fsth2 = fdt2 >> 32;
+
+    fst2 = float32_mul(fst0, fst2, &env->active_fpu.fp_status);
+    fsth2 = float32_mul(fsth0, fsth2, &env->active_fpu.fp_status);
+    fst2 = float32_chs(float32_sub(fst2, float32_one, &env->active_fpu.fp_status));
+    fsth2 = float32_chs(float32_sub(fsth2, float32_one, &env->active_fpu.fp_status));
+    update_fcr31(env, GETPC());
+    return ((uint64_t)fsth2 << 32) | fst2;
+}
+
+uint64_t helper_float_rsqrt2_d(CPUMIPSState *env, uint64_t fdt0, uint64_t fdt2)
+{
+    fdt2 = float64_mul(fdt0, fdt2, &env->active_fpu.fp_status);
+    fdt2 = float64_sub(fdt2, float64_one, &env->active_fpu.fp_status);
+    fdt2 = float64_chs(float64_div(fdt2, FLOAT_TWO64, &env->active_fpu.fp_status));
+    update_fcr31(env, GETPC());
+    return fdt2;
+}
+
+uint32_t helper_float_rsqrt2_s(CPUMIPSState *env, uint32_t fst0, uint32_t fst2)
+{
+    fst2 = float32_mul(fst0, fst2, &env->active_fpu.fp_status);
+    fst2 = float32_sub(fst2, float32_one, &env->active_fpu.fp_status);
+    fst2 = float32_chs(float32_div(fst2, FLOAT_TWO32, &env->active_fpu.fp_status));
+    update_fcr31(env, GETPC());
+    return fst2;
+}
+
+uint64_t helper_float_rsqrt2_ps(CPUMIPSState *env, uint64_t fdt0, uint64_t fdt2)
+{
+    uint32_t fst0 = fdt0 & 0XFFFFFFFF;
+    uint32_t fsth0 = fdt0 >> 32;
+    uint32_t fst2 = fdt2 & 0XFFFFFFFF;
+    uint32_t fsth2 = fdt2 >> 32;
+
+    fst2 = float32_mul(fst0, fst2, &env->active_fpu.fp_status);
+    fsth2 = float32_mul(fsth0, fsth2, &env->active_fpu.fp_status);
+    fst2 = float32_sub(fst2, float32_one, &env->active_fpu.fp_status);
+    fsth2 = float32_sub(fsth2, float32_one, &env->active_fpu.fp_status);
+    fst2 = float32_chs(float32_div(fst2, FLOAT_TWO32, &env->active_fpu.fp_status));
+    fsth2 = float32_chs(float32_div(fsth2, FLOAT_TWO32, &env->active_fpu.fp_status));
+    update_fcr31(env, GETPC());
+    return ((uint64_t)fsth2 << 32) | fst2;
+}
+
+uint64_t helper_float_addr_ps(CPUMIPSState *env, uint64_t fdt0, uint64_t fdt1)
+{
+    uint32_t fst0 = fdt0 & 0XFFFFFFFF;
+    uint32_t fsth0 = fdt0 >> 32;
+    uint32_t fst1 = fdt1 & 0XFFFFFFFF;
+    uint32_t fsth1 = fdt1 >> 32;
+    uint32_t fst2;
+    uint32_t fsth2;
+
+    fst2 = float32_add (fst0, fsth0, &env->active_fpu.fp_status);
+    fsth2 = float32_add (fst1, fsth1, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return ((uint64_t)fsth2 << 32) | fst2;
+}
+
+uint64_t helper_float_mulr_ps(CPUMIPSState *env, uint64_t fdt0, uint64_t fdt1)
+{
+    uint32_t fst0 = fdt0 & 0XFFFFFFFF;
+    uint32_t fsth0 = fdt0 >> 32;
+    uint32_t fst1 = fdt1 & 0XFFFFFFFF;
+    uint32_t fsth1 = fdt1 >> 32;
+    uint32_t fst2;
+    uint32_t fsth2;
+
+    fst2 = float32_mul (fst0, fsth0, &env->active_fpu.fp_status);
+    fsth2 = float32_mul (fst1, fsth1, &env->active_fpu.fp_status);
+    update_fcr31(env, GETPC());
+    return ((uint64_t)fsth2 << 32) | fst2;
+}
+
+#define FLOAT_MINMAX(name, bits, minmaxfunc)                            \
+uint ## bits ## _t helper_float_ ## name (CPUMIPSState *env,            \
+                                          uint ## bits ## _t fs,        \
+                                          uint ## bits ## _t ft)        \
+{                                                                       \
+    uint ## bits ## _t fdret;                                           \
+                                                                        \
+    fdret = float ## bits ## _ ## minmaxfunc(fs, ft,                    \
+                                           &env->active_fpu.fp_status); \
+    update_fcr31(env, GETPC());                                         \
+    return fdret;                                                       \
+}
+
+FLOAT_MINMAX(max_s, 32, maxnum)
+FLOAT_MINMAX(max_d, 64, maxnum)
+FLOAT_MINMAX(maxa_s, 32, maxnummag)
+FLOAT_MINMAX(maxa_d, 64, maxnummag)
+
+FLOAT_MINMAX(min_s, 32, minnum)
+FLOAT_MINMAX(min_d, 64, minnum)
+FLOAT_MINMAX(mina_s, 32, minnummag)
+FLOAT_MINMAX(mina_d, 64, minnummag)
+#undef FLOAT_MINMAX
+
+/* ternary operations */
+#define UNFUSED_FMA(prefix, a, b, c, flags)                          \
+{                                                                    \
+    a = prefix##_mul(a, b, &env->active_fpu.fp_status);              \
+    if ((flags) & float_muladd_negate_c) {                           \
+        a = prefix##_sub(a, c, &env->active_fpu.fp_status);          \
+    } else {                                                         \
+        a = prefix##_add(a, c, &env->active_fpu.fp_status);          \
+    }                                                                \
+    if ((flags) & float_muladd_negate_result) {                      \
+        a = prefix##_chs(a);                                         \
+    }                                                                \
+}
+
+/* FMA based operations */
+#define FLOAT_FMA(name, type)                                        \
+uint64_t helper_float_ ## name ## _d(CPUMIPSState *env,              \
+                                     uint64_t fdt0, uint64_t fdt1,   \
+                                     uint64_t fdt2)                  \
+{                                                                    \
+    UNFUSED_FMA(float64, fdt0, fdt1, fdt2, type);                    \
+    update_fcr31(env, GETPC());                                      \
+    return fdt0;                                                     \
+}                                                                    \
+                                                                     \
+uint32_t helper_float_ ## name ## _s(CPUMIPSState *env,              \
+                                     uint32_t fst0, uint32_t fst1,   \
+                                     uint32_t fst2)                  \
+{                                                                    \
+    UNFUSED_FMA(float32, fst0, fst1, fst2, type);                    \
+    update_fcr31(env, GETPC());                                      \
+    return fst0;                                                     \
+}                                                                    \
+                                                                     \
+uint64_t helper_float_ ## name ## _ps(CPUMIPSState *env,             \
+                                      uint64_t fdt0, uint64_t fdt1,  \
+                                      uint64_t fdt2)                 \
+{                                                                    \
+    uint32_t fst0 = fdt0 & 0XFFFFFFFF;                               \
+    uint32_t fsth0 = fdt0 >> 32;                                     \
+    uint32_t fst1 = fdt1 & 0XFFFFFFFF;                               \
+    uint32_t fsth1 = fdt1 >> 32;                                     \
+    uint32_t fst2 = fdt2 & 0XFFFFFFFF;                               \
+    uint32_t fsth2 = fdt2 >> 32;                                     \
+                                                                     \
+    UNFUSED_FMA(float32, fst0, fst1, fst2, type);                    \
+    UNFUSED_FMA(float32, fsth0, fsth1, fsth2, type);                 \
+    update_fcr31(env, GETPC());                                      \
+    return ((uint64_t)fsth0 << 32) | fst0;                           \
+}
+FLOAT_FMA(madd, 0)
+FLOAT_FMA(msub, float_muladd_negate_c)
+FLOAT_FMA(nmadd, float_muladd_negate_result)
+FLOAT_FMA(nmsub, float_muladd_negate_result | float_muladd_negate_c)
+#undef FLOAT_FMA
+
+#define FLOAT_FMADDSUB(name, bits, muladd_arg)                          \
+uint ## bits ## _t helper_float_ ## name (CPUMIPSState *env,            \
+                                          uint ## bits ## _t fs,        \
+                                          uint ## bits ## _t ft,        \
+                                          uint ## bits ## _t fd)        \
+{                                                                       \
+    uint ## bits ## _t fdret;                                           \
+                                                                        \
+    fdret = float ## bits ## _muladd(fs, ft, fd, muladd_arg,            \
+                                     &env->active_fpu.fp_status);       \
+    update_fcr31(env, GETPC());                                         \
+    return fdret;                                                       \
+}
+
+FLOAT_FMADDSUB(maddf_s, 32, 0)
+FLOAT_FMADDSUB(maddf_d, 64, 0)
+FLOAT_FMADDSUB(msubf_s, 32, float_muladd_negate_product)
+FLOAT_FMADDSUB(msubf_d, 64, float_muladd_negate_product)
+#undef FLOAT_FMADDSUB
+
+/* compare operations */
+#define FOP_COND_D(op, cond)                                   \
+void helper_cmp_d_ ## op(CPUMIPSState *env, uint64_t fdt0,     \
+                         uint64_t fdt1, int cc)                \
+{                                                              \
+    int c;                                                     \
+    c = cond;                                                  \
+    update_fcr31(env, GETPC());                                \
+    if (c)                                                     \
+        SET_FP_COND(cc, env->active_fpu);                      \
+    else                                                       \
+        CLEAR_FP_COND(cc, env->active_fpu);                    \
+}                                                              \
+void helper_cmpabs_d_ ## op(CPUMIPSState *env, uint64_t fdt0,  \
+                            uint64_t fdt1, int cc)             \
+{                                                              \
+    int c;                                                     \
+    fdt0 = float64_abs(fdt0);                                  \
+    fdt1 = float64_abs(fdt1);                                  \
+    c = cond;                                                  \
+    update_fcr31(env, GETPC());                                \
+    if (c)                                                     \
+        SET_FP_COND(cc, env->active_fpu);                      \
+    else                                                       \
+        CLEAR_FP_COND(cc, env->active_fpu);                    \
+}
+
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float64_unordered_quiet() is still called. */
+FOP_COND_D(f,   (float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status), 0))
+FOP_COND_D(un,  float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status))
+FOP_COND_D(eq,  float64_eq_quiet(fdt0, fdt1, &env->active_fpu.fp_status))
+FOP_COND_D(ueq, float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status)  || float64_eq_quiet(fdt0, fdt1, &env->active_fpu.fp_status))
+FOP_COND_D(olt, float64_lt_quiet(fdt0, fdt1, &env->active_fpu.fp_status))
+FOP_COND_D(ult, float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status)  || float64_lt_quiet(fdt0, fdt1, &env->active_fpu.fp_status))
+FOP_COND_D(ole, float64_le_quiet(fdt0, fdt1, &env->active_fpu.fp_status))
+FOP_COND_D(ule, float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status)  || float64_le_quiet(fdt0, fdt1, &env->active_fpu.fp_status))
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float64_unordered() is still called. */
+FOP_COND_D(sf,  (float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status), 0))
+FOP_COND_D(ngle,float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status))
+FOP_COND_D(seq, float64_eq(fdt0, fdt1, &env->active_fpu.fp_status))
+FOP_COND_D(ngl, float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status)  || float64_eq(fdt0, fdt1, &env->active_fpu.fp_status))
+FOP_COND_D(lt,  float64_lt(fdt0, fdt1, &env->active_fpu.fp_status))
+FOP_COND_D(nge, float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status)  || float64_lt(fdt0, fdt1, &env->active_fpu.fp_status))
+FOP_COND_D(le,  float64_le(fdt0, fdt1, &env->active_fpu.fp_status))
+FOP_COND_D(ngt, float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status)  || float64_le(fdt0, fdt1, &env->active_fpu.fp_status))
+
+#define FOP_COND_S(op, cond)                                   \
+void helper_cmp_s_ ## op(CPUMIPSState *env, uint32_t fst0,     \
+                         uint32_t fst1, int cc)                \
+{                                                              \
+    int c;                                                     \
+    c = cond;                                                  \
+    update_fcr31(env, GETPC());                                \
+    if (c)                                                     \
+        SET_FP_COND(cc, env->active_fpu);                      \
+    else                                                       \
+        CLEAR_FP_COND(cc, env->active_fpu);                    \
+}                                                              \
+void helper_cmpabs_s_ ## op(CPUMIPSState *env, uint32_t fst0,  \
+                            uint32_t fst1, int cc)             \
+{                                                              \
+    int c;                                                     \
+    fst0 = float32_abs(fst0);                                  \
+    fst1 = float32_abs(fst1);                                  \
+    c = cond;                                                  \
+    update_fcr31(env, GETPC());                                \
+    if (c)                                                     \
+        SET_FP_COND(cc, env->active_fpu);                      \
+    else                                                       \
+        CLEAR_FP_COND(cc, env->active_fpu);                    \
+}
+
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float32_unordered_quiet() is still called. */
+FOP_COND_S(f,   (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status), 0))
+FOP_COND_S(un,  float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status))
+FOP_COND_S(eq,  float32_eq_quiet(fst0, fst1, &env->active_fpu.fp_status))
+FOP_COND_S(ueq, float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)  || float32_eq_quiet(fst0, fst1, &env->active_fpu.fp_status))
+FOP_COND_S(olt, float32_lt_quiet(fst0, fst1, &env->active_fpu.fp_status))
+FOP_COND_S(ult, float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)  || float32_lt_quiet(fst0, fst1, &env->active_fpu.fp_status))
+FOP_COND_S(ole, float32_le_quiet(fst0, fst1, &env->active_fpu.fp_status))
+FOP_COND_S(ule, float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)  || float32_le_quiet(fst0, fst1, &env->active_fpu.fp_status))
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float32_unordered() is still called. */
+FOP_COND_S(sf,  (float32_unordered(fst1, fst0, &env->active_fpu.fp_status), 0))
+FOP_COND_S(ngle,float32_unordered(fst1, fst0, &env->active_fpu.fp_status))
+FOP_COND_S(seq, float32_eq(fst0, fst1, &env->active_fpu.fp_status))
+FOP_COND_S(ngl, float32_unordered(fst1, fst0, &env->active_fpu.fp_status)  || float32_eq(fst0, fst1, &env->active_fpu.fp_status))
+FOP_COND_S(lt,  float32_lt(fst0, fst1, &env->active_fpu.fp_status))
+FOP_COND_S(nge, float32_unordered(fst1, fst0, &env->active_fpu.fp_status)  || float32_lt(fst0, fst1, &env->active_fpu.fp_status))
+FOP_COND_S(le,  float32_le(fst0, fst1, &env->active_fpu.fp_status))
+FOP_COND_S(ngt, float32_unordered(fst1, fst0, &env->active_fpu.fp_status)  || float32_le(fst0, fst1, &env->active_fpu.fp_status))
+
+#define FOP_COND_PS(op, condl, condh)                           \
+void helper_cmp_ps_ ## op(CPUMIPSState *env, uint64_t fdt0,     \
+                          uint64_t fdt1, int cc)                \
+{                                                               \
+    uint32_t fst0, fsth0, fst1, fsth1;                          \
+    int ch, cl;                                                 \
+    fst0 = fdt0 & 0XFFFFFFFF;                                   \
+    fsth0 = fdt0 >> 32;                                         \
+    fst1 = fdt1 & 0XFFFFFFFF;                                   \
+    fsth1 = fdt1 >> 32;                                         \
+    cl = condl;                                                 \
+    ch = condh;                                                 \
+    update_fcr31(env, GETPC());                                 \
+    if (cl)                                                     \
+        SET_FP_COND(cc, env->active_fpu);                       \
+    else                                                        \
+        CLEAR_FP_COND(cc, env->active_fpu);                     \
+    if (ch)                                                     \
+        SET_FP_COND(cc + 1, env->active_fpu);                   \
+    else                                                        \
+        CLEAR_FP_COND(cc + 1, env->active_fpu);                 \
+}                                                               \
+void helper_cmpabs_ps_ ## op(CPUMIPSState *env, uint64_t fdt0,  \
+                             uint64_t fdt1, int cc)             \
+{                                                               \
+    uint32_t fst0, fsth0, fst1, fsth1;                          \
+    int ch, cl;                                                 \
+    fst0 = float32_abs(fdt0 & 0XFFFFFFFF);                      \
+    fsth0 = float32_abs(fdt0 >> 32);                            \
+    fst1 = float32_abs(fdt1 & 0XFFFFFFFF);                      \
+    fsth1 = float32_abs(fdt1 >> 32);                            \
+    cl = condl;                                                 \
+    ch = condh;                                                 \
+    update_fcr31(env, GETPC());                                 \
+    if (cl)                                                     \
+        SET_FP_COND(cc, env->active_fpu);                       \
+    else                                                        \
+        CLEAR_FP_COND(cc, env->active_fpu);                     \
+    if (ch)                                                     \
+        SET_FP_COND(cc + 1, env->active_fpu);                   \
+    else                                                        \
+        CLEAR_FP_COND(cc + 1, env->active_fpu);                 \
+}
+
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float32_unordered_quiet() is still called. */
+FOP_COND_PS(f,   (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status), 0),
+                 (float32_unordered_quiet(fsth1, fsth0, &env->active_fpu.fp_status), 0))
+FOP_COND_PS(un,  float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status),
+                 float32_unordered_quiet(fsth1, fsth0, &env->active_fpu.fp_status))
+FOP_COND_PS(eq,  float32_eq_quiet(fst0, fst1, &env->active_fpu.fp_status),
+                 float32_eq_quiet(fsth0, fsth1, &env->active_fpu.fp_status))
+FOP_COND_PS(ueq, float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)    || float32_eq_quiet(fst0, fst1, &env->active_fpu.fp_status),
+                 float32_unordered_quiet(fsth1, fsth0, &env->active_fpu.fp_status)  || float32_eq_quiet(fsth0, fsth1, &env->active_fpu.fp_status))
+FOP_COND_PS(olt, float32_lt_quiet(fst0, fst1, &env->active_fpu.fp_status),
+                 float32_lt_quiet(fsth0, fsth1, &env->active_fpu.fp_status))
+FOP_COND_PS(ult, float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)    || float32_lt_quiet(fst0, fst1, &env->active_fpu.fp_status),
+                 float32_unordered_quiet(fsth1, fsth0, &env->active_fpu.fp_status)  || float32_lt_quiet(fsth0, fsth1, &env->active_fpu.fp_status))
+FOP_COND_PS(ole, float32_le_quiet(fst0, fst1, &env->active_fpu.fp_status),
+                 float32_le_quiet(fsth0, fsth1, &env->active_fpu.fp_status))
+FOP_COND_PS(ule, float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)    || float32_le_quiet(fst0, fst1, &env->active_fpu.fp_status),
+                 float32_unordered_quiet(fsth1, fsth0, &env->active_fpu.fp_status)  || float32_le_quiet(fsth0, fsth1, &env->active_fpu.fp_status))
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float32_unordered() is still called. */
+FOP_COND_PS(sf,  (float32_unordered(fst1, fst0, &env->active_fpu.fp_status), 0),
+                 (float32_unordered(fsth1, fsth0, &env->active_fpu.fp_status), 0))
+FOP_COND_PS(ngle,float32_unordered(fst1, fst0, &env->active_fpu.fp_status),
+                 float32_unordered(fsth1, fsth0, &env->active_fpu.fp_status))
+FOP_COND_PS(seq, float32_eq(fst0, fst1, &env->active_fpu.fp_status),
+                 float32_eq(fsth0, fsth1, &env->active_fpu.fp_status))
+FOP_COND_PS(ngl, float32_unordered(fst1, fst0, &env->active_fpu.fp_status)    || float32_eq(fst0, fst1, &env->active_fpu.fp_status),
+                 float32_unordered(fsth1, fsth0, &env->active_fpu.fp_status)  || float32_eq(fsth0, fsth1, &env->active_fpu.fp_status))
+FOP_COND_PS(lt,  float32_lt(fst0, fst1, &env->active_fpu.fp_status),
+                 float32_lt(fsth0, fsth1, &env->active_fpu.fp_status))
+FOP_COND_PS(nge, float32_unordered(fst1, fst0, &env->active_fpu.fp_status)    || float32_lt(fst0, fst1, &env->active_fpu.fp_status),
+                 float32_unordered(fsth1, fsth0, &env->active_fpu.fp_status)  || float32_lt(fsth0, fsth1, &env->active_fpu.fp_status))
+FOP_COND_PS(le,  float32_le(fst0, fst1, &env->active_fpu.fp_status),
+                 float32_le(fsth0, fsth1, &env->active_fpu.fp_status))
+FOP_COND_PS(ngt, float32_unordered(fst1, fst0, &env->active_fpu.fp_status)    || float32_le(fst0, fst1, &env->active_fpu.fp_status),
+                 float32_unordered(fsth1, fsth0, &env->active_fpu.fp_status)  || float32_le(fsth0, fsth1, &env->active_fpu.fp_status))
+
+/* R6 compare operations */
+#define FOP_CONDN_D(op, cond)                                       \
+uint64_t helper_r6_cmp_d_ ## op(CPUMIPSState * env, uint64_t fdt0,  \
+                         uint64_t fdt1)                             \
+{                                                                   \
+    uint64_t c;                                                     \
+    c = cond;                                                       \
+    update_fcr31(env, GETPC());                                     \
+    if (c) {                                                        \
+        return -1;                                                  \
+    } else {                                                        \
+        return 0;                                                   \
+    }                                                               \
+}
+
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float64_unordered_quiet() is still called. */
+FOP_CONDN_D(af,  (float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status), 0))
+FOP_CONDN_D(un,  (float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status)))
+FOP_CONDN_D(eq,  (float64_eq_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(ueq, (float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                  || float64_eq_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(lt,  (float64_lt_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(ult, (float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                  || float64_lt_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(le,  (float64_le_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(ule, (float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                  || float64_le_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float64_unordered() is still called. */
+FOP_CONDN_D(saf,  (float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status), 0))
+FOP_CONDN_D(sun,  (float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status)))
+FOP_CONDN_D(seq,  (float64_eq(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sueq, (float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_eq(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(slt,  (float64_lt(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sult, (float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sle,  (float64_le(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sule, (float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_le(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(or,   (float64_le_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_le_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(une,  (float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(ne,   (float64_lt_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sor,  (float64_le(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_le(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sune, (float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sne,  (float64_lt(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt(fdt0, fdt1, &env->active_fpu.fp_status)))
+
+#define FOP_CONDN_S(op, cond)                                       \
+uint32_t helper_r6_cmp_s_ ## op(CPUMIPSState * env, uint32_t fst0,  \
+                         uint32_t fst1)                             \
+{                                                                   \
+    uint64_t c;                                                     \
+    c = cond;                                                       \
+    update_fcr31(env, GETPC());                                     \
+    if (c) {                                                        \
+        return -1;                                                  \
+    } else {                                                        \
+        return 0;                                                   \
+    }                                                               \
+}
+
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float32_unordered_quiet() is still called. */
+FOP_CONDN_S(af,   (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status), 0))
+FOP_CONDN_S(un,   (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)))
+FOP_CONDN_S(eq,   (float32_eq_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(ueq,  (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_eq_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(lt,   (float32_lt_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(ult,  (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(le,   (float32_le_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(ule,  (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_le_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float32_unordered() is still called. */
+FOP_CONDN_S(saf,  (float32_unordered(fst1, fst0, &env->active_fpu.fp_status), 0))
+FOP_CONDN_S(sun,  (float32_unordered(fst1, fst0, &env->active_fpu.fp_status)))
+FOP_CONDN_S(seq,  (float32_eq(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sueq, (float32_unordered(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_eq(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(slt,  (float32_lt(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sult, (float32_unordered(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sle,  (float32_le(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sule, (float32_unordered(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_le(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(or,   (float32_le_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_le_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(une,  (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(ne,   (float32_lt_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sor,  (float32_le(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_le(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sune, (float32_unordered(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sne,  (float32_lt(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt(fst0, fst1, &env->active_fpu.fp_status)))
+
+/* MSA */
+/* Data format min and max values */
+#define DF_BITS(df) (1 << ((df) + 3))
+
+/* Element-by-element access macros */
+#define DF_ELEMENTS(df) (MSA_WRLEN / DF_BITS(df))
+
+#if !defined(CONFIG_USER_ONLY)
+#define MEMOP_IDX(DF)                                           \
+        TCGMemOpIdx oi = make_memop_idx(MO_TE | DF | MO_UNALN,  \
+                                        cpu_mmu_index(env));
+#else
+#define MEMOP_IDX(DF)
+#endif
+
+#define MSA_LD_DF(DF, TYPE, LD_INSN, ...)                               \
+void helper_msa_ld_ ## TYPE(CPUMIPSState *env, uint32_t wd,             \
+                            target_ulong addr)                          \
+{                                                                       \
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);                          \
+    wr_t wx;                                                            \
+    int i;                                                              \
+    MEMOP_IDX(DF)                                                       \
+    for (i = 0; i < DF_ELEMENTS(DF); i++) {                             \
+        wx.TYPE[i] = LD_INSN(env, addr + (i << DF), ##__VA_ARGS__);     \
+    }                                                                   \
+    memcpy(pwd, &wx, sizeof(wr_t));                                     \
+}
+
+#if !defined(CONFIG_USER_ONLY)
+MSA_LD_DF(DF_BYTE,   b, helper_ret_ldub_mmu, oi, GETRA())
+MSA_LD_DF(DF_HALF,   h, helper_ret_lduw_mmu, oi, GETRA())
+MSA_LD_DF(DF_WORD,   w, helper_ret_ldul_mmu, oi, GETRA())
+MSA_LD_DF(DF_DOUBLE, d, helper_ret_ldq_mmu,  oi, GETRA())
+#else
+MSA_LD_DF(DF_BYTE,   b, cpu_ldub_data)
+MSA_LD_DF(DF_HALF,   h, cpu_lduw_data)
+MSA_LD_DF(DF_WORD,   w, cpu_ldl_data)
+MSA_LD_DF(DF_DOUBLE, d, cpu_ldq_data)
+#endif
+
+#define MSA_PAGESPAN(x) \
+        ((((x) & ~TARGET_PAGE_MASK) + MSA_WRLEN/8 - 1) >= TARGET_PAGE_SIZE)
+
+static inline void ensure_writable_pages(CPUMIPSState *env,
+                                         target_ulong addr,
+                                         int mmu_idx,
+                                         uintptr_t retaddr)
+{
+#if !defined(CONFIG_USER_ONLY)
+    target_ulong page_addr;
+    if (unlikely(MSA_PAGESPAN(addr))) {
+        /* first page */
+        probe_write(env, addr, mmu_idx, retaddr);
+        /* second page */
+        page_addr = (addr & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE;
+        probe_write(env, page_addr, mmu_idx, retaddr);
+    }
+#endif
+}
+
+#define MSA_ST_DF(DF, TYPE, ST_INSN, ...)                               \
+void helper_msa_st_ ## TYPE(CPUMIPSState *env, uint32_t wd,             \
+                            target_ulong addr)                          \
+{                                                                       \
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);                          \
+    int mmu_idx = cpu_mmu_index(env);                                   \
+    int i;                                                              \
+    MEMOP_IDX(DF)                                                       \
+    ensure_writable_pages(env, addr, mmu_idx, GETRA());                 \
+    for (i = 0; i < DF_ELEMENTS(DF); i++) {                             \
+        ST_INSN(env, addr + (i << DF), pwd->TYPE[i], ##__VA_ARGS__);    \
+    }                                                                   \
+}
+
+#if !defined(CONFIG_USER_ONLY)
+MSA_ST_DF(DF_BYTE,   b, helper_ret_stb_mmu, oi, GETRA())
+MSA_ST_DF(DF_HALF,   h, helper_ret_stw_mmu, oi, GETRA())
+MSA_ST_DF(DF_WORD,   w, helper_ret_stl_mmu, oi, GETRA())
+MSA_ST_DF(DF_DOUBLE, d, helper_ret_stq_mmu, oi, GETRA())
+#else
+MSA_ST_DF(DF_BYTE,   b, cpu_stb_data)
+MSA_ST_DF(DF_HALF,   h, cpu_stw_data)
+MSA_ST_DF(DF_WORD,   w, cpu_stl_data)
+MSA_ST_DF(DF_DOUBLE, d, cpu_stq_data)
+#endif