diff --git a/target-arm/cpu.h b/target-arm/cpu.h
index 275733e..1feec83 100644
--- a/target-arm/cpu.h
+++ b/target-arm/cpu.h
@@ -168,9 +168,6 @@
         int vec_len;
         int vec_stride;
 
-        /* Temporary variables if we don't have spare fp regs.  */
-        float32 tmp0s, tmp1s;
-        float64 tmp0d, tmp1d;
         /* scratch space when Tn are not sufficient.  */
         uint32_t scratch[8];
 
diff --git a/target-arm/exec.h b/target-arm/exec.h
index 9b284f2..d953098 100644
--- a/target-arm/exec.h
+++ b/target-arm/exec.h
@@ -25,13 +25,6 @@
 register uint32_t T1 asm(AREG2);
 register uint32_t T2 asm(AREG3);
 
-/* TODO: Put these in FP regs on targets that have such things.  */
-/* It is ok for FT0s and FT0d to overlap.  Likewise FT1s and FT1d.  */
-#define FT0s env->vfp.tmp0s
-#define FT1s env->vfp.tmp1s
-#define FT0d env->vfp.tmp0d
-#define FT1d env->vfp.tmp1d
-
 #define M0   env->iwmmxt.val
 
 #include "cpu.h"
@@ -83,23 +76,5 @@
 
 void raise_exception(int);
 
-void do_vfp_abss(void);
-void do_vfp_absd(void);
-void do_vfp_negs(void);
-void do_vfp_negd(void);
-void do_vfp_sqrts(void);
-void do_vfp_sqrtd(void);
-void do_vfp_cmps(void);
-void do_vfp_cmpd(void);
-void do_vfp_cmpes(void);
-void do_vfp_cmped(void);
-void do_vfp_set_fpscr(void);
-void do_vfp_get_fpscr(void);
-float32 helper_recps_f32(float32, float32);
-float32 helper_rsqrts_f32(float32, float32);
-uint32_t helper_recpe_u32(uint32_t);
-uint32_t helper_rsqrte_u32(uint32_t);
-float32 helper_recpe_f32(float32);
-float32 helper_rsqrte_f32(float32);
 void helper_neon_tbl(int rn, int maxindex);
 uint32_t helper_neon_mul_p8(uint32_t op1, uint32_t op2);
diff --git a/target-arm/helper.c b/target-arm/helper.c
index 8985c88..8e737a8 100644
--- a/target-arm/helper.c
+++ b/target-arm/helper.c
@@ -2167,3 +2167,366 @@
     return (a & mask) | (b & ~mask);
 }
 
+
+/* VFP support.  We follow the convention used for VFP instrunctions:
+   Single precition routines have a "s" suffix, double precision a
+   "d" suffix.  */
+
+/* Convert host exception flags to vfp form.  */
+static inline int vfp_exceptbits_from_host(int host_bits)
+{
+    int target_bits = 0;
+
+    if (host_bits & float_flag_invalid)
+        target_bits |= 1;
+    if (host_bits & float_flag_divbyzero)
+        target_bits |= 2;
+    if (host_bits & float_flag_overflow)
+        target_bits |= 4;
+    if (host_bits & float_flag_underflow)
+        target_bits |= 8;
+    if (host_bits & float_flag_inexact)
+        target_bits |= 0x10;
+    return target_bits;
+}
+
+uint32_t HELPER(vfp_get_fpscr)(CPUState *env)
+{
+    int i;
+    uint32_t fpscr;
+
+    fpscr = (env->vfp.xregs[ARM_VFP_FPSCR] & 0xffc8ffff)
+            | (env->vfp.vec_len << 16)
+            | (env->vfp.vec_stride << 20);
+    i = get_float_exception_flags(&env->vfp.fp_status);
+    fpscr |= vfp_exceptbits_from_host(i);
+    return fpscr;
+}
+
+/* Convert vfp exception flags to target form.  */
+static inline int vfp_exceptbits_to_host(int target_bits)
+{
+    int host_bits = 0;
+
+    if (target_bits & 1)
+        host_bits |= float_flag_invalid;
+    if (target_bits & 2)
+        host_bits |= float_flag_divbyzero;
+    if (target_bits & 4)
+        host_bits |= float_flag_overflow;
+    if (target_bits & 8)
+        host_bits |= float_flag_underflow;
+    if (target_bits & 0x10)
+        host_bits |= float_flag_inexact;
+    return host_bits;
+}
+
+void HELPER(vfp_set_fpscr)(CPUState *env, uint32_t val)
+{
+    int i;
+    uint32_t changed;
+
+    changed = env->vfp.xregs[ARM_VFP_FPSCR];
+    env->vfp.xregs[ARM_VFP_FPSCR] = (val & 0xffc8ffff);
+    env->vfp.vec_len = (val >> 16) & 7;
+    env->vfp.vec_stride = (val >> 20) & 3;
+
+    changed ^= val;
+    if (changed & (3 << 22)) {
+        i = (val >> 22) & 3;
+        switch (i) {
+        case 0:
+            i = float_round_nearest_even;
+            break;
+        case 1:
+            i = float_round_up;
+            break;
+        case 2:
+            i = float_round_down;
+            break;
+        case 3:
+            i = float_round_to_zero;
+            break;
+        }
+        set_float_rounding_mode(i, &env->vfp.fp_status);
+    }
+
+    i = vfp_exceptbits_to_host((val >> 8) & 0x1f);
+    set_float_exception_flags(i, &env->vfp.fp_status);
+    /* XXX: FZ and DN are not implemented.  */
+}
+
+#define VFP_HELPER(name, p) HELPER(glue(glue(vfp_,name),p))
+
+#define VFP_BINOP(name) \
+float32 VFP_HELPER(name, s)(float32 a, float32 b, CPUState *env) \
+{ \
+    return float32_ ## name (a, b, &env->vfp.fp_status); \
+} \
+float64 VFP_HELPER(name, d)(float64 a, float64 b, CPUState *env) \
+{ \
+    return float64_ ## name (a, b, &env->vfp.fp_status); \
+}
+VFP_BINOP(add)
+VFP_BINOP(sub)
+VFP_BINOP(mul)
+VFP_BINOP(div)
+#undef VFP_BINOP
+
+float32 VFP_HELPER(neg, s)(float32 a)
+{
+    return float32_chs(a);
+}
+
+float64 VFP_HELPER(neg, d)(float64 a)
+{
+    return float32_chs(a);
+}
+
+float32 VFP_HELPER(abs, s)(float32 a)
+{
+    return float32_abs(a);
+}
+
+float64 VFP_HELPER(abs, d)(float64 a)
+{
+    return float32_abs(a);
+}
+
+float32 VFP_HELPER(sqrt, s)(float32 a, CPUState *env)
+{
+    return float32_sqrt(a, &env->vfp.fp_status);
+}
+
+float64 VFP_HELPER(sqrt, d)(float64 a, CPUState *env)
+{
+    return float64_sqrt(a, &env->vfp.fp_status);
+}
+
+/* XXX: check quiet/signaling case */
+#define DO_VFP_cmp(p, type) \
+void VFP_HELPER(cmp, p)(type a, type b, CPUState *env)  \
+{ \
+    uint32_t flags; \
+    switch(type ## _compare_quiet(a, b, &env->vfp.fp_status)) { \
+    case 0: flags = 0x6; break; \
+    case -1: flags = 0x8; break; \
+    case 1: flags = 0x2; break; \
+    default: case 2: flags = 0x3; break; \
+    } \
+    env->vfp.xregs[ARM_VFP_FPSCR] = (flags << 28) \
+        | (env->vfp.xregs[ARM_VFP_FPSCR] & 0x0fffffff); \
+} \
+void VFP_HELPER(cmpe, p)(type a, type b, CPUState *env) \
+{ \
+    uint32_t flags; \
+    switch(type ## _compare(a, b, &env->vfp.fp_status)) { \
+    case 0: flags = 0x6; break; \
+    case -1: flags = 0x8; break; \
+    case 1: flags = 0x2; break; \
+    default: case 2: flags = 0x3; break; \
+    } \
+    env->vfp.xregs[ARM_VFP_FPSCR] = (flags << 28) \
+        | (env->vfp.xregs[ARM_VFP_FPSCR] & 0x0fffffff); \
+}
+DO_VFP_cmp(s, float32)
+DO_VFP_cmp(d, float64)
+#undef DO_VFP_cmp
+
+/* Helper routines to perform bitwise copies between float and int.  */
+static inline float32 vfp_itos(uint32_t i)
+{
+    union {
+        uint32_t i;
+        float32 s;
+    } v;
+
+    v.i = i;
+    return v.s;
+}
+
+static inline uint32_t vfp_stoi(float32 s)
+{
+    union {
+        uint32_t i;
+        float32 s;
+    } v;
+
+    v.s = s;
+    return v.i;
+}
+
+static inline float64 vfp_itod(uint64_t i)
+{
+    union {
+        uint64_t i;
+        float64 d;
+    } v;
+
+    v.i = i;
+    return v.d;
+}
+
+static inline uint64_t vfp_dtoi(float64 d)
+{
+    union {
+        uint64_t i;
+        float64 d;
+    } v;
+
+    v.d = d;
+    return v.i;
+}
+
+/* Integer to float conversion.  */
+float32 VFP_HELPER(uito, s)(float32 x, CPUState *env)
+{
+    return uint32_to_float32(vfp_stoi(x), &env->vfp.fp_status);
+}
+
+float64 VFP_HELPER(uito, d)(float32 x, CPUState *env)
+{
+    return uint32_to_float64(vfp_stoi(x), &env->vfp.fp_status);
+}
+
+float32 VFP_HELPER(sito, s)(float32 x, CPUState *env)
+{
+    return int32_to_float32(vfp_stoi(x), &env->vfp.fp_status);
+}
+
+float64 VFP_HELPER(sito, d)(float32 x, CPUState *env)
+{
+    return int32_to_float64(vfp_stoi(x), &env->vfp.fp_status);
+}
+
+/* Float to integer conversion.  */
+float32 VFP_HELPER(toui, s)(float32 x, CPUState *env)
+{
+    return vfp_itos(float32_to_uint32(x, &env->vfp.fp_status));
+}
+
+float32 VFP_HELPER(toui, d)(float64 x, CPUState *env)
+{
+    return vfp_itos(float64_to_uint32(x, &env->vfp.fp_status));
+}
+
+float32 VFP_HELPER(tosi, s)(float32 x, CPUState *env)
+{
+    return vfp_itos(float32_to_int32(x, &env->vfp.fp_status));
+}
+
+float32 VFP_HELPER(tosi, d)(float64 x, CPUState *env)
+{
+    return vfp_itos(float64_to_int32(x, &env->vfp.fp_status));
+}
+
+float32 VFP_HELPER(touiz, s)(float32 x, CPUState *env)
+{
+    return vfp_itos(float32_to_uint32_round_to_zero(x, &env->vfp.fp_status));
+}
+
+float32 VFP_HELPER(touiz, d)(float64 x, CPUState *env)
+{
+    return vfp_itos(float64_to_uint32_round_to_zero(x, &env->vfp.fp_status));
+}
+
+float32 VFP_HELPER(tosiz, s)(float32 x, CPUState *env)
+{
+    return vfp_itos(float32_to_int32_round_to_zero(x, &env->vfp.fp_status));
+}
+
+float32 VFP_HELPER(tosiz, d)(float64 x, CPUState *env)
+{
+    return vfp_itos(float64_to_int32_round_to_zero(x, &env->vfp.fp_status));
+}
+
+/* floating point conversion */
+float64 VFP_HELPER(fcvtd, s)(float32 x, CPUState *env)
+{
+    return float32_to_float64(x, &env->vfp.fp_status);
+}
+
+float32 VFP_HELPER(fcvts, d)(float64 x, CPUState *env)
+{
+    return float64_to_float32(x, &env->vfp.fp_status);
+}
+
+/* VFP3 fixed point conversion.  */
+#define VFP_CONV_FIX(name, p, ftype, itype, sign) \
+ftype VFP_HELPER(name##to, p)(ftype x, uint32_t shift, CPUState *env) \
+{ \
+    ftype tmp; \
+    tmp = sign##int32_to_##ftype ((itype)vfp_##p##toi(x), \
+                                  &env->vfp.fp_status); \
+    return ftype##_scalbn(tmp, shift, &env->vfp.fp_status); \
+} \
+ftype VFP_HELPER(to##name, p)(ftype x, uint32_t shift, CPUState *env) \
+{ \
+    ftype tmp; \
+    tmp = ftype##_scalbn(x, shift, &env->vfp.fp_status); \
+    return vfp_ito##p((itype)ftype##_to_##sign##int32_round_to_zero(tmp, \
+        &env->vfp.fp_status)); \
+}
+
+VFP_CONV_FIX(sh, d, float64, int16, )
+VFP_CONV_FIX(sl, d, float64, int32, )
+VFP_CONV_FIX(uh, d, float64, uint16, u)
+VFP_CONV_FIX(ul, d, float64, uint32, u)
+VFP_CONV_FIX(sh, s, float32, int16, )
+VFP_CONV_FIX(sl, s, float32, int32, )
+VFP_CONV_FIX(uh, s, float32, uint16, u)
+VFP_CONV_FIX(ul, s, float32, uint32, u)
+#undef VFP_CONV_FIX
+
+float32 HELPER(recps_f32)(float32 a, float32 b, CPUState *env)
+{
+    float_status *s = &env->vfp.fp_status;
+    float32 two = int32_to_float32(2, s);
+    return float32_sub(two, float32_mul(a, b, s), s);
+}
+
+float32 HELPER(rsqrts_f32)(float32 a, float32 b, CPUState *env)
+{
+    float_status *s = &env->vfp.fp_status;
+    float32 three = int32_to_float32(3, s);
+    return float32_sub(three, float32_mul(a, b, s), s);
+}
+
+/* TODO: The architecture specifies the value that the estimate functions
+   should return.  We return the exact reciprocal/root instead.  */
+float32 HELPER(recpe_f32)(float32 a, CPUState *env)
+{
+    float_status *s = &env->vfp.fp_status;
+    float32 one = int32_to_float32(1, s);
+    return float32_div(one, a, s);
+}
+
+float32 HELPER(rsqrte_f32)(float32 a, CPUState *env)
+{
+    float_status *s = &env->vfp.fp_status;
+    float32 one = int32_to_float32(1, s);
+    return float32_div(one, float32_sqrt(a, s), s);
+}
+
+uint32_t HELPER(recpe_u32)(uint32_t a, CPUState *env)
+{
+    float_status *s = &env->vfp.fp_status;
+    float32 tmp;
+    tmp = int32_to_float32(a, s);
+    tmp = float32_scalbn(tmp, -32, s);
+    tmp = helper_recpe_f32(tmp, env);
+    tmp = float32_scalbn(tmp, 31, s);
+    return float32_to_int32(tmp, s);
+}
+
+uint32_t HELPER(rsqrte_u32)(uint32_t a, CPUState *env)
+{
+    float_status *s = &env->vfp.fp_status;
+    float32 tmp;
+    tmp = int32_to_float32(a, s);
+    tmp = float32_scalbn(tmp, -32, s);
+    tmp = helper_rsqrte_f32(tmp, env);
+    tmp = float32_scalbn(tmp, 31, s);
+    return float32_to_int32(tmp, s);
+}
+
diff --git a/target-arm/helpers.h b/target-arm/helpers.h
index 12c9343..e4c50c6 100644
--- a/target-arm/helpers.h
+++ b/target-arm/helpers.h
@@ -122,6 +122,69 @@
 DEF_HELPER_1_1(get_user_reg, uint32_t, (uint32_t))
 DEF_HELPER_0_2(set_user_reg, void, (uint32_t, uint32_t))
 
+DEF_HELPER_1_1(vfp_get_fpscr, uint32_t, (CPUState *))
+DEF_HELPER_0_2(vfp_set_fpscr, void, (CPUState *, uint32_t))
+
+DEF_HELPER_1_3(vfp_adds, float32, (float32, float32, CPUState *))
+DEF_HELPER_1_3(vfp_addd, float64, (float64, float64, CPUState *))
+DEF_HELPER_1_3(vfp_subs, float32, (float32, float32, CPUState *))
+DEF_HELPER_1_3(vfp_subd, float64, (float64, float64, CPUState *))
+DEF_HELPER_1_3(vfp_muls, float32, (float32, float32, CPUState *))
+DEF_HELPER_1_3(vfp_muld, float64, (float64, float64, CPUState *))
+DEF_HELPER_1_3(vfp_divs, float32, (float32, float32, CPUState *))
+DEF_HELPER_1_3(vfp_divd, float64, (float64, float64, CPUState *))
+DEF_HELPER_1_1(vfp_negs, float32, (float32))
+DEF_HELPER_1_1(vfp_negd, float64, (float64))
+DEF_HELPER_1_1(vfp_abss, float32, (float32))
+DEF_HELPER_1_1(vfp_absd, float64, (float64))
+DEF_HELPER_1_2(vfp_sqrts, float32, (float32, CPUState *))
+DEF_HELPER_1_2(vfp_sqrtd, float64, (float64, CPUState *))
+DEF_HELPER_0_3(vfp_cmps, void, (float32, float32, CPUState *))
+DEF_HELPER_0_3(vfp_cmpd, void, (float64, float64, CPUState *))
+DEF_HELPER_0_3(vfp_cmpes, void, (float32, float32, CPUState *))
+DEF_HELPER_0_3(vfp_cmped, void, (float64, float64, CPUState *))
+
+DEF_HELPER_1_2(vfp_fcvtds, float64, (float32, CPUState *))
+DEF_HELPER_1_2(vfp_fcvtsd, float32, (float64, CPUState *))
+
+DEF_HELPER_1_2(vfp_uitos, float32, (float32, CPUState *))
+DEF_HELPER_1_2(vfp_uitod, float64, (float32, CPUState *))
+DEF_HELPER_1_2(vfp_sitos, float32, (float32, CPUState *))
+DEF_HELPER_1_2(vfp_sitod, float64, (float32, CPUState *))
+
+DEF_HELPER_1_2(vfp_touis, float32, (float32, CPUState *))
+DEF_HELPER_1_2(vfp_touid, float32, (float64, CPUState *))
+DEF_HELPER_1_2(vfp_touizs, float32, (float32, CPUState *))
+DEF_HELPER_1_2(vfp_touizd, float32, (float64, CPUState *))
+DEF_HELPER_1_2(vfp_tosis, float32, (float32, CPUState *))
+DEF_HELPER_1_2(vfp_tosid, float32, (float64, CPUState *))
+DEF_HELPER_1_2(vfp_tosizs, float32, (float32, CPUState *))
+DEF_HELPER_1_2(vfp_tosizd, float32, (float64, CPUState *))
+
+DEF_HELPER_1_3(vfp_toshs, float32, (float32, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_tosls, float32, (float32, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_touhs, float32, (float32, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_touls, float32, (float32, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_toshd, float64, (float64, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_tosld, float64, (float64, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_touhd, float64, (float64, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_tould, float64, (float64, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_shtos, float32, (float32, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_sltos, float32, (float32, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_uhtos, float32, (float32, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_ultos, float32, (float32, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_shtod, float64, (float64, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_sltod, float64, (float64, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_uhtod, float64, (float64, uint32_t, CPUState *))
+DEF_HELPER_1_3(vfp_ultod, float64, (float64, uint32_t, CPUState *))
+
+DEF_HELPER_1_3(recps_f32, float32, (float32, float32, CPUState *))
+DEF_HELPER_1_3(rsqrts_f32, float32, (float32, float32, CPUState *))
+DEF_HELPER_1_2(recpe_f32, float32, (float32, CPUState *))
+DEF_HELPER_1_2(rsqrte_f32, float32, (float32, CPUState *))
+DEF_HELPER_1_2(recpe_u32, uint32_t, (uint32_t, CPUState *))
+DEF_HELPER_1_2(rsqrte_u32, uint32_t, (uint32_t, CPUState *))
+
 #undef DEF_HELPER
 #undef DEF_HELPER_0_0
 #undef DEF_HELPER_0_1
diff --git a/target-arm/op.c b/target-arm/op.c
index 74bf62b..66dc737 100644
--- a/target-arm/op.c
+++ b/target-arm/op.c
@@ -252,319 +252,6 @@
     FORCE_RET();
 }
 
-/* VFP support.  We follow the convention used for VFP instrunctions:
-   Single precition routines have a "s" suffix, double precision a
-   "d" suffix.  */
-
-#define VFP_OP(name, p) void OPPROTO op_vfp_##name##p(void)
-
-#define VFP_BINOP(name) \
-VFP_OP(name, s)             \
-{                           \
-    FT0s = float32_ ## name (FT0s, FT1s, &env->vfp.fp_status);    \
-}                           \
-VFP_OP(name, d)             \
-{                           \
-    FT0d = float64_ ## name (FT0d, FT1d, &env->vfp.fp_status);    \
-}
-VFP_BINOP(add)
-VFP_BINOP(sub)
-VFP_BINOP(mul)
-VFP_BINOP(div)
-#undef VFP_BINOP
-
-#define VFP_HELPER(name)  \
-VFP_OP(name, s)           \
-{                         \
-    do_vfp_##name##s();    \
-}                         \
-VFP_OP(name, d)           \
-{                         \
-    do_vfp_##name##d();    \
-}
-VFP_HELPER(abs)
-VFP_HELPER(sqrt)
-VFP_HELPER(cmp)
-VFP_HELPER(cmpe)
-#undef VFP_HELPER
-
-/* XXX: Will this do the right thing for NANs.  Should invert the signbit
-   without looking at the rest of the value.  */
-VFP_OP(neg, s)
-{
-    FT0s = float32_chs(FT0s);
-}
-
-VFP_OP(neg, d)
-{
-    FT0d = float64_chs(FT0d);
-}
-
-VFP_OP(F1_ld0, s)
-{
-    union {
-        uint32_t i;
-        float32 s;
-    } v;
-    v.i = 0;
-    FT1s = v.s;
-}
-
-VFP_OP(F1_ld0, d)
-{
-    union {
-        uint64_t i;
-        float64 d;
-    } v;
-    v.i = 0;
-    FT1d = v.d;
-}
-
-/* Helper routines to perform bitwise copies between float and int.  */
-static inline float32 vfp_itos(uint32_t i)
-{
-    union {
-        uint32_t i;
-        float32 s;
-    } v;
-
-    v.i = i;
-    return v.s;
-}
-
-static inline uint32_t vfp_stoi(float32 s)
-{
-    union {
-        uint32_t i;
-        float32 s;
-    } v;
-
-    v.s = s;
-    return v.i;
-}
-
-static inline float64 vfp_itod(uint64_t i)
-{
-    union {
-        uint64_t i;
-        float64 d;
-    } v;
-
-    v.i = i;
-    return v.d;
-}
-
-static inline uint64_t vfp_dtoi(float64 d)
-{
-    union {
-        uint64_t i;
-        float64 d;
-    } v;
-
-    v.d = d;
-    return v.i;
-}
-
-/* Integer to float conversion.  */
-VFP_OP(uito, s)
-{
-    FT0s = uint32_to_float32(vfp_stoi(FT0s), &env->vfp.fp_status);
-}
-
-VFP_OP(uito, d)
-{
-    FT0d = uint32_to_float64(vfp_stoi(FT0s), &env->vfp.fp_status);
-}
-
-VFP_OP(sito, s)
-{
-    FT0s = int32_to_float32(vfp_stoi(FT0s), &env->vfp.fp_status);
-}
-
-VFP_OP(sito, d)
-{
-    FT0d = int32_to_float64(vfp_stoi(FT0s), &env->vfp.fp_status);
-}
-
-/* Float to integer conversion.  */
-VFP_OP(toui, s)
-{
-    FT0s = vfp_itos(float32_to_uint32(FT0s, &env->vfp.fp_status));
-}
-
-VFP_OP(toui, d)
-{
-    FT0s = vfp_itos(float64_to_uint32(FT0d, &env->vfp.fp_status));
-}
-
-VFP_OP(tosi, s)
-{
-    FT0s = vfp_itos(float32_to_int32(FT0s, &env->vfp.fp_status));
-}
-
-VFP_OP(tosi, d)
-{
-    FT0s = vfp_itos(float64_to_int32(FT0d, &env->vfp.fp_status));
-}
-
-/* TODO: Set rounding mode properly.  */
-VFP_OP(touiz, s)
-{
-    FT0s = vfp_itos(float32_to_uint32_round_to_zero(FT0s, &env->vfp.fp_status));
-}
-
-VFP_OP(touiz, d)
-{
-    FT0s = vfp_itos(float64_to_uint32_round_to_zero(FT0d, &env->vfp.fp_status));
-}
-
-VFP_OP(tosiz, s)
-{
-    FT0s = vfp_itos(float32_to_int32_round_to_zero(FT0s, &env->vfp.fp_status));
-}
-
-VFP_OP(tosiz, d)
-{
-    FT0s = vfp_itos(float64_to_int32_round_to_zero(FT0d, &env->vfp.fp_status));
-}
-
-/* floating point conversion */
-VFP_OP(fcvtd, s)
-{
-    FT0d = float32_to_float64(FT0s, &env->vfp.fp_status);
-}
-
-VFP_OP(fcvts, d)
-{
-    FT0s = float64_to_float32(FT0d, &env->vfp.fp_status);
-}
-
-/* VFP3 fixed point conversion.  */
-#define VFP_CONV_FIX(name, p, ftype, itype, sign) \
-VFP_OP(name##to, p) \
-{ \
-    ftype tmp; \
-    tmp = sign##int32_to_##ftype ((itype)vfp_##p##toi(FT0##p), \
-                                  &env->vfp.fp_status); \
-    FT0##p = ftype##_scalbn(tmp, PARAM1, &env->vfp.fp_status); \
-} \
-VFP_OP(to##name, p) \
-{ \
-    ftype tmp; \
-    tmp = ftype##_scalbn(FT0##p, PARAM1, &env->vfp.fp_status); \
-    FT0##p = vfp_ito##p((itype)ftype##_to_##sign##int32_round_to_zero(tmp, \
-            &env->vfp.fp_status)); \
-}
-
-VFP_CONV_FIX(sh, d, float64, int16, )
-VFP_CONV_FIX(sl, d, float64, int32, )
-VFP_CONV_FIX(uh, d, float64, uint16, u)
-VFP_CONV_FIX(ul, d, float64, uint32, u)
-VFP_CONV_FIX(sh, s, float32, int16, )
-VFP_CONV_FIX(sl, s, float32, int32, )
-VFP_CONV_FIX(uh, s, float32, uint16, u)
-VFP_CONV_FIX(ul, s, float32, uint32, u)
-
-/* Get and Put values from registers.  */
-VFP_OP(getreg_F0, d)
-{
-  FT0d = *(float64 *)((char *) env + PARAM1);
-}
-
-VFP_OP(getreg_F0, s)
-{
-  FT0s = *(float32 *)((char *) env + PARAM1);
-}
-
-VFP_OP(getreg_F1, d)
-{
-  FT1d = *(float64 *)((char *) env + PARAM1);
-}
-
-VFP_OP(getreg_F1, s)
-{
-  FT1s = *(float32 *)((char *) env + PARAM1);
-}
-
-VFP_OP(setreg_F0, d)
-{
-  *(float64 *)((char *) env + PARAM1) = FT0d;
-}
-
-VFP_OP(setreg_F0, s)
-{
-  *(float32 *)((char *) env + PARAM1) = FT0s;
-}
-
-void OPPROTO op_vfp_movl_T0_fpscr(void)
-{
-    do_vfp_get_fpscr ();
-}
-
-void OPPROTO op_vfp_movl_T0_fpscr_flags(void)
-{
-    T0 = env->vfp.xregs[ARM_VFP_FPSCR] & (0xf << 28);
-}
-
-void OPPROTO op_vfp_movl_fpscr_T0(void)
-{
-    do_vfp_set_fpscr();
-}
-
-void OPPROTO op_vfp_movl_T0_xreg(void)
-{
-    T0 = env->vfp.xregs[PARAM1];
-}
-
-void OPPROTO op_vfp_movl_xreg_T0(void)
-{
-    env->vfp.xregs[PARAM1] = T0;
-}
-
-/* Move between FT0s to T0  */
-void OPPROTO op_vfp_mrs(void)
-{
-    T0 = vfp_stoi(FT0s);
-}
-
-void OPPROTO op_vfp_msr(void)
-{
-    FT0s = vfp_itos(T0);
-}
-
-/* Move between FT0d and {T0,T1} */
-void OPPROTO op_vfp_mrrd(void)
-{
-    CPU_DoubleU u;
-
-    u.d = FT0d;
-    T0 = u.l.lower;
-    T1 = u.l.upper;
-}
-
-void OPPROTO op_vfp_mdrr(void)
-{
-    CPU_DoubleU u;
-
-    u.l.lower = T0;
-    u.l.upper = T1;
-    FT0d = u.d;
-}
-
-/* Load immediate.  PARAM1 is the 32 most significant bits of the value.  */
-void OPPROTO op_vfp_fconstd(void)
-{
-    CPU_DoubleU u;
-    u.l.upper = PARAM1;
-    u.l.lower = 0;
-    FT0d = u.d;
-}
-
-void OPPROTO op_vfp_fconsts(void)
-{
-    FT0s = vfp_itos(PARAM1);
-}
-
 void OPPROTO op_movl_cp_T0(void)
 {
     helper_set_cp(env, PARAM1, T0);
diff --git a/target-arm/op_helper.c b/target-arm/op_helper.c
index 939aebf..6748b06 100644
--- a/target-arm/op_helper.c
+++ b/target-arm/op_helper.c
@@ -40,194 +40,6 @@
     spin_unlock(&global_cpu_lock);
 }
 
-/* VFP support.  */
-
-void do_vfp_abss(void)
-{
-    FT0s = float32_abs(FT0s);
-}
-
-void do_vfp_absd(void)
-{
-    FT0d = float64_abs(FT0d);
-}
-
-void do_vfp_sqrts(void)
-{
-    FT0s = float32_sqrt(FT0s, &env->vfp.fp_status);
-}
-
-void do_vfp_sqrtd(void)
-{
-    FT0d = float64_sqrt(FT0d, &env->vfp.fp_status);
-}
-
-/* XXX: check quiet/signaling case */
-#define DO_VFP_cmp(p, size)               \
-void do_vfp_cmp##p(void)                  \
-{                                         \
-    uint32_t flags;                       \
-    switch(float ## size ## _compare_quiet(FT0##p, FT1##p, &env->vfp.fp_status)) {\
-    case 0: flags = 0x6; break;\
-    case -1: flags = 0x8; break;\
-    case 1: flags = 0x2; break;\
-    default: case 2: flags = 0x3; break;\
-    }\
-    env->vfp.xregs[ARM_VFP_FPSCR] = (flags << 28)\
-        | (env->vfp.xregs[ARM_VFP_FPSCR] & 0x0fffffff); \
-    FORCE_RET();                          \
-}\
-\
-void do_vfp_cmpe##p(void)                   \
-{                                           \
-    uint32_t flags;                       \
-    switch(float ## size ## _compare(FT0##p, FT1##p, &env->vfp.fp_status)) {\
-    case 0: flags = 0x6; break;\
-    case -1: flags = 0x8; break;\
-    case 1: flags = 0x2; break;\
-    default: case 2: flags = 0x3; break;\
-    }\
-    env->vfp.xregs[ARM_VFP_FPSCR] = (flags << 28)\
-        | (env->vfp.xregs[ARM_VFP_FPSCR] & 0x0fffffff); \
-    FORCE_RET();                          \
-}
-DO_VFP_cmp(s, 32)
-DO_VFP_cmp(d, 64)
-#undef DO_VFP_cmp
-
-/* Convert host exception flags to vfp form.  */
-static inline int vfp_exceptbits_from_host(int host_bits)
-{
-    int target_bits = 0;
-
-    if (host_bits & float_flag_invalid)
-        target_bits |= 1;
-    if (host_bits & float_flag_divbyzero)
-        target_bits |= 2;
-    if (host_bits & float_flag_overflow)
-        target_bits |= 4;
-    if (host_bits & float_flag_underflow)
-        target_bits |= 8;
-    if (host_bits & float_flag_inexact)
-        target_bits |= 0x10;
-    return target_bits;
-}
-
-/* Convert vfp exception flags to target form.  */
-static inline int vfp_exceptbits_to_host(int target_bits)
-{
-    int host_bits = 0;
-
-    if (target_bits & 1)
-        host_bits |= float_flag_invalid;
-    if (target_bits & 2)
-        host_bits |= float_flag_divbyzero;
-    if (target_bits & 4)
-        host_bits |= float_flag_overflow;
-    if (target_bits & 8)
-        host_bits |= float_flag_underflow;
-    if (target_bits & 0x10)
-        host_bits |= float_flag_inexact;
-    return host_bits;
-}
-
-void do_vfp_set_fpscr(void)
-{
-    int i;
-    uint32_t changed;
-
-    changed = env->vfp.xregs[ARM_VFP_FPSCR];
-    env->vfp.xregs[ARM_VFP_FPSCR] = (T0 & 0xffc8ffff);
-    env->vfp.vec_len = (T0 >> 16) & 7;
-    env->vfp.vec_stride = (T0 >> 20) & 3;
-
-    changed ^= T0;
-    if (changed & (3 << 22)) {
-        i = (T0 >> 22) & 3;
-        switch (i) {
-        case 0:
-            i = float_round_nearest_even;
-            break;
-        case 1:
-            i = float_round_up;
-            break;
-        case 2:
-            i = float_round_down;
-            break;
-        case 3:
-            i = float_round_to_zero;
-            break;
-        }
-        set_float_rounding_mode(i, &env->vfp.fp_status);
-    }
-
-    i = vfp_exceptbits_to_host((T0 >> 8) & 0x1f);
-    set_float_exception_flags(i, &env->vfp.fp_status);
-    /* XXX: FZ and DN are not implemented.  */
-}
-
-void do_vfp_get_fpscr(void)
-{
-    int i;
-
-    T0 = (env->vfp.xregs[ARM_VFP_FPSCR] & 0xffc8ffff) | (env->vfp.vec_len << 16)
-          | (env->vfp.vec_stride << 20);
-    i = get_float_exception_flags(&env->vfp.fp_status);
-    T0 |= vfp_exceptbits_from_host(i);
-}
-
-float32 helper_recps_f32(float32 a, float32 b)
-{
-    float_status *s = &env->vfp.fp_status;
-    float32 two = int32_to_float32(2, s);
-    return float32_sub(two, float32_mul(a, b, s), s);
-}
-
-float32 helper_rsqrts_f32(float32 a, float32 b)
-{
-    float_status *s = &env->vfp.fp_status;
-    float32 three = int32_to_float32(3, s);
-    return float32_sub(three, float32_mul(a, b, s), s);
-}
-
-/* TODO: The architecture specifies the value that the estimate functions
-   should return.  We return the exact reciprocal/root instead.  */
-float32 helper_recpe_f32(float32 a)
-{
-    float_status *s = &env->vfp.fp_status;
-    float32 one = int32_to_float32(1, s);
-    return float32_div(one, a, s);
-}
-
-float32 helper_rsqrte_f32(float32 a)
-{
-    float_status *s = &env->vfp.fp_status;
-    float32 one = int32_to_float32(1, s);
-    return float32_div(one, float32_sqrt(a, s), s);
-}
-
-uint32_t helper_recpe_u32(uint32_t a)
-{
-    float_status *s = &env->vfp.fp_status;
-    float32 tmp;
-    tmp = int32_to_float32(a, s);
-    tmp = float32_scalbn(tmp, -32, s);
-    tmp = helper_recpe_f32(tmp);
-    tmp = float32_scalbn(tmp, 31, s);
-    return float32_to_int32(tmp, s);
-}
-
-uint32_t helper_rsqrte_u32(uint32_t a)
-{
-    float_status *s = &env->vfp.fp_status;
-    float32 tmp;
-    tmp = int32_to_float32(a, s);
-    tmp = float32_scalbn(tmp, -32, s);
-    tmp = helper_rsqrte_f32(tmp);
-    tmp = float32_scalbn(tmp, 31, s);
-    return float32_to_int32(tmp, s);
-}
-
 void helper_neon_tbl(int rn, int maxindex)
 {
     uint32_t val;
diff --git a/target-arm/op_mem.h b/target-arm/op_mem.h
index f11c498..0e2e7b6 100644
--- a/target-arm/op_mem.h
+++ b/target-arm/op_mem.h
@@ -77,24 +77,6 @@
     FORCE_RET();
 }
 
-/* Floating point load/store.  Address is in T1 */
-#define VFP_MEM_OP(p, w) \
-void OPPROTO glue(op_vfp_ld##p,MEMSUFFIX)(void) \
-{ \
-    FT0##p = glue(ldf##w,MEMSUFFIX)(T1); \
-    FORCE_RET(); \
-} \
-void OPPROTO glue(op_vfp_st##p,MEMSUFFIX)(void) \
-{ \
-    glue(stf##w,MEMSUFFIX)(T1, FT0##p); \
-    FORCE_RET(); \
-}
-
-VFP_MEM_OP(s,l)
-VFP_MEM_OP(d,q)
-
-#undef VFP_MEM_OP
-
 /* iwMMXt load/store.  Address is in T1 */
 #define MMX_MEM_OP(name, ldname) \
 void OPPROTO glue(op_iwmmxt_ld##name,MEMSUFFIX)(void) \
diff --git a/target-arm/op_neon.h b/target-arm/op_neon.h
index e91dcc5..095165e 100644
--- a/target-arm/op_neon.h
+++ b/target-arm/op_neon.h
@@ -14,6 +14,29 @@
 #define NFS &env->vfp.fp_status
 #define NEON_OP(name) void OPPROTO op_neon_##name (void)
 
+/* Helper routines to perform bitwise copies between float and int.  */
+static inline float32 vfp_itos(uint32_t i)
+{
+    union {
+        uint32_t i;
+        float32 s;
+    } v;
+
+    v.i = i;
+    return v.s;
+}
+
+static inline uint32_t vfp_stoi(float32 s)
+{
+    union {
+        uint32_t i;
+        float32 s;
+    } v;
+
+    v.s = s;
+    return v.i;
+}
+
 NEON_OP(getreg_T0)
 {
     T0 = *(uint32_t *)((char *) env + PARAM1);
@@ -754,18 +777,6 @@
 #undef NEON_FN
 #undef NEON_QDMULH32
 
-NEON_OP(recps_f32)
-{
-    T0 = vfp_stoi(helper_recps_f32(vfp_itos(T0), vfp_itos(T1)));
-    FORCE_RET();
-}
-
-NEON_OP(rsqrts_f32)
-{
-    T0 = vfp_stoi(helper_rsqrts_f32(vfp_itos(T0), vfp_itos(T1)));
-    FORCE_RET();
-}
-
 /* Floating point comparisons produce an integer result.  */
 #define NEON_VOP_FCMP(name, cmp) \
 NEON_OP(name) \
@@ -1702,27 +1713,6 @@
     FORCE_RET();
 }
 
-/* Reciprocal/root estimate.  */
-NEON_OP(recpe_u32)
-{
-    T0 = helper_recpe_u32(T0);
-}
-
-NEON_OP(rsqrte_u32)
-{
-    T0 = helper_rsqrte_u32(T0);
-}
-
-NEON_OP(recpe_f32)
-{
-    FT0s = helper_recpe_f32(FT0s);
-}
-
-NEON_OP(rsqrte_f32)
-{
-    FT0s = helper_rsqrte_f32(FT0s);
-}
-
 /* Table lookup.  This accessed the register file directly.  */
 NEON_OP(tbl)
 {
diff --git a/target-arm/translate.c b/target-arm/translate.c
index 89515cc..3e7cbe8 100644
--- a/target-arm/translate.c
+++ b/target-arm/translate.c
@@ -79,6 +79,7 @@
 static TCGv cpu_env;
 /* FIXME:  These should be removed.  */
 static TCGv cpu_T[3];
+static TCGv cpu_F0s, cpu_F1s, cpu_F0d, cpu_F1d;
 
 /* initialize TCG globals.  */
 void arm_translate_init(void)
@@ -959,16 +960,16 @@
     }
 }
 
-#define VFP_OP(name)                      \
-static inline void gen_vfp_##name(int dp) \
-{                                         \
-    if (dp)                               \
-        gen_op_vfp_##name##d();           \
-    else                                  \
-        gen_op_vfp_##name##s();           \
+#define VFP_OP2(name)                                                 \
+static inline void gen_vfp_##name(int dp)                             \
+{                                                                     \
+    if (dp)                                                           \
+        gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, cpu_F1d, cpu_env); \
+    else                                                              \
+        gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, cpu_F1s, cpu_env); \
 }
 
-#define VFP_OP1(name)                               \
+#define VFP_OP1i(name)                               \
 static inline void gen_vfp_##name(int dp, int arg)  \
 {                                                   \
     if (dp)                                         \
@@ -977,55 +978,141 @@
         gen_op_vfp_##name##s(arg);                  \
 }
 
-VFP_OP(add)
-VFP_OP(sub)
-VFP_OP(mul)
-VFP_OP(div)
-VFP_OP(neg)
-VFP_OP(abs)
-VFP_OP(sqrt)
-VFP_OP(cmp)
-VFP_OP(cmpe)
-VFP_OP(F1_ld0)
-VFP_OP(uito)
-VFP_OP(sito)
-VFP_OP(toui)
-VFP_OP(touiz)
-VFP_OP(tosi)
-VFP_OP(tosiz)
-VFP_OP1(tosh)
-VFP_OP1(tosl)
-VFP_OP1(touh)
-VFP_OP1(toul)
-VFP_OP1(shto)
-VFP_OP1(slto)
-VFP_OP1(uhto)
-VFP_OP1(ulto)
+VFP_OP2(add)
+VFP_OP2(sub)
+VFP_OP2(mul)
+VFP_OP2(div)
 
-#undef VFP_OP
+#undef VFP_OP2
 
-static inline void gen_vfp_fconst(int dp, uint32_t val)
+static inline void gen_vfp_abs(int dp)
 {
     if (dp)
-        gen_op_vfp_fconstd(val);
+        gen_helper_vfp_absd(cpu_F0d, cpu_F0d);
     else
-        gen_op_vfp_fconsts(val);
+        gen_helper_vfp_abss(cpu_F0s, cpu_F0s);
 }
 
+static inline void gen_vfp_neg(int dp)
+{
+    if (dp)
+        gen_helper_vfp_negd(cpu_F0d, cpu_F0d);
+    else
+        gen_helper_vfp_negs(cpu_F0s, cpu_F0s);
+}
+
+static inline void gen_vfp_sqrt(int dp)
+{
+    if (dp)
+        gen_helper_vfp_sqrtd(cpu_F0d, cpu_F0d, cpu_env);
+    else
+        gen_helper_vfp_sqrts(cpu_F0s, cpu_F0s, cpu_env);
+}
+
+static inline void gen_vfp_cmp(int dp)
+{
+    if (dp)
+        gen_helper_vfp_cmpd(cpu_F0d, cpu_F1d, cpu_env);
+    else
+        gen_helper_vfp_cmps(cpu_F0s, cpu_F1s, cpu_env);
+}
+
+static inline void gen_vfp_cmpe(int dp)
+{
+    if (dp)
+        gen_helper_vfp_cmped(cpu_F0d, cpu_F1d, cpu_env);
+    else
+        gen_helper_vfp_cmpes(cpu_F0s, cpu_F1s, cpu_env);
+}
+
+static inline void gen_vfp_F1_ld0(int dp)
+{
+    if (dp)
+        tcg_gen_movi_i64(cpu_F0d, 0);
+    else
+        tcg_gen_movi_i32(cpu_F0s, 0);
+}
+
+static inline void gen_vfp_uito(int dp)
+{
+    if (dp)
+        gen_helper_vfp_uitod(cpu_F0d, cpu_F0s, cpu_env);
+    else
+        gen_helper_vfp_uitos(cpu_F0s, cpu_F0s, cpu_env);
+}
+
+static inline void gen_vfp_sito(int dp)
+{
+    if (dp)
+        gen_helper_vfp_uitod(cpu_F0d, cpu_F0s, cpu_env);
+    else
+        gen_helper_vfp_uitos(cpu_F0s, cpu_F0s, cpu_env);
+}
+
+static inline void gen_vfp_toui(int dp)
+{
+    if (dp)
+        gen_helper_vfp_touid(cpu_F0s, cpu_F0d, cpu_env);
+    else
+        gen_helper_vfp_touis(cpu_F0s, cpu_F0s, cpu_env);
+}
+
+static inline void gen_vfp_touiz(int dp)
+{
+    if (dp)
+        gen_helper_vfp_touizd(cpu_F0s, cpu_F0d, cpu_env);
+    else
+        gen_helper_vfp_touizs(cpu_F0s, cpu_F0s, cpu_env);
+}
+
+static inline void gen_vfp_tosi(int dp)
+{
+    if (dp)
+        gen_helper_vfp_tosid(cpu_F0s, cpu_F0d, cpu_env);
+    else
+        gen_helper_vfp_tosis(cpu_F0s, cpu_F0s, cpu_env);
+}
+
+static inline void gen_vfp_tosiz(int dp)
+{
+    if (dp)
+        gen_helper_vfp_tosizd(cpu_F0s, cpu_F0d, cpu_env);
+    else
+        gen_helper_vfp_tosizs(cpu_F0s, cpu_F0s, cpu_env);
+}
+
+#define VFP_GEN_FIX(name) \
+static inline void gen_vfp_##name(int dp, int shift) \
+{ \
+    if (dp) \
+        gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, tcg_const_i32(shift), cpu_env);\
+    else \
+        gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, tcg_const_i32(shift), cpu_env);\
+}
+VFP_GEN_FIX(tosh)
+VFP_GEN_FIX(tosl)
+VFP_GEN_FIX(touh)
+VFP_GEN_FIX(toul)
+VFP_GEN_FIX(shto)
+VFP_GEN_FIX(slto)
+VFP_GEN_FIX(uhto)
+VFP_GEN_FIX(ulto)
+#undef VFP_GEN_FIX
+
 static inline void gen_vfp_ld(DisasContext *s, int dp)
 {
     if (dp)
-        gen_ldst(vfp_ldd, s);
+        tcg_gen_qemu_ld64(cpu_F0d, cpu_T[1], IS_USER(s));
     else
-        gen_ldst(vfp_lds, s);
+        tcg_gen_qemu_ld32u(cpu_F0s, cpu_T[1], IS_USER(s));
 }
 
 static inline void gen_vfp_st(DisasContext *s, int dp)
 {
     if (dp)
-        gen_ldst(vfp_std, s);
+        tcg_gen_qemu_st64(cpu_F0d, cpu_T[1], IS_USER(s));
     else
-        gen_ldst(vfp_sts, s);
+        tcg_gen_qemu_st32(cpu_F0s, cpu_T[1], IS_USER(s));
 }
 
 static inline long
@@ -1055,28 +1142,33 @@
 #define NEON_GET_REG(T, reg, n) gen_op_neon_getreg_##T(neon_reg_offset(reg, n))
 #define NEON_SET_REG(T, reg, n) gen_op_neon_setreg_##T(neon_reg_offset(reg, n))
 
+#define tcg_gen_ld_f32 tcg_gen_ld_i32
+#define tcg_gen_ld_f64 tcg_gen_ld_i64
+#define tcg_gen_st_f32 tcg_gen_st_i32
+#define tcg_gen_st_f64 tcg_gen_st_i64
+
 static inline void gen_mov_F0_vreg(int dp, int reg)
 {
     if (dp)
-        gen_op_vfp_getreg_F0d(vfp_reg_offset(dp, reg));
+        tcg_gen_ld_f64(cpu_F0d, cpu_env, vfp_reg_offset(dp, reg));
     else
-        gen_op_vfp_getreg_F0s(vfp_reg_offset(dp, reg));
+        tcg_gen_ld_f32(cpu_F0s, cpu_env, vfp_reg_offset(dp, reg));
 }
 
 static inline void gen_mov_F1_vreg(int dp, int reg)
 {
     if (dp)
-        gen_op_vfp_getreg_F1d(vfp_reg_offset(dp, reg));
+        tcg_gen_ld_f64(cpu_F1d, cpu_env, vfp_reg_offset(dp, reg));
     else
-        gen_op_vfp_getreg_F1s(vfp_reg_offset(dp, reg));
+        tcg_gen_ld_f32(cpu_F1s, cpu_env, vfp_reg_offset(dp, reg));
 }
 
 static inline void gen_mov_vreg_F0(int dp, int reg)
 {
     if (dp)
-        gen_op_vfp_setreg_F0d(vfp_reg_offset(dp, reg));
+        tcg_gen_st_f64(cpu_F0d, cpu_env, vfp_reg_offset(dp, reg));
     else
-        gen_op_vfp_setreg_F0s(vfp_reg_offset(dp, reg));
+        tcg_gen_st_f32(cpu_F0s, cpu_env, vfp_reg_offset(dp, reg));
 }
 
 #define ARM_CP_RW_BIT	(1 << 20)
@@ -2262,6 +2354,20 @@
 #define VFP_SREG_M(insn) VFP_SREG(insn,  0,  5)
 #define VFP_DREG_M(reg, insn) VFP_DREG(reg, insn,  0,  5)
 
+/* Move between integer and VFP cores.  */
+static TCGv gen_vfp_mrs(void)
+{
+    TCGv tmp = new_tmp();
+    tcg_gen_mov_i32(tmp, cpu_F0s);
+    return tmp;
+}
+
+static void gen_vfp_msr(TCGv tmp)
+{
+    tcg_gen_mov_i32(cpu_F0s, tmp);
+    dead_tmp(tmp);
+}
+
 static inline int
 vfp_enabled(CPUState * env)
 {
@@ -2274,6 +2380,7 @@
 {
     uint32_t rd, rn, rm, op, i, n, offset, delta_d, delta_m, bank_mask;
     int dp, veclen;
+    TCGv tmp;
 
     if (!arm_feature(env, ARM_FEATURE_VFP))
         return 1;
@@ -2396,18 +2503,18 @@
 
                         switch (rn) {
                         case ARM_VFP_FPSID:
-                            /* VFP2 allows access for FSID from userspace.
+                            /* VFP2 allows access to FSID from userspace.
                                VFP3 restricts all id registers to privileged
                                accesses.  */
                             if (IS_USER(s)
                                 && arm_feature(env, ARM_FEATURE_VFP3))
                                 return 1;
-                            gen_op_vfp_movl_T0_xreg(rn);
+                            tmp = load_cpu_field(vfp.xregs[rn]);
                             break;
                         case ARM_VFP_FPEXC:
                             if (IS_USER(s))
                                 return 1;
-                            gen_op_vfp_movl_T0_xreg(rn);
+                            tmp = load_cpu_field(vfp.xregs[rn]);
                             break;
                         case ARM_VFP_FPINST:
                         case ARM_VFP_FPINST2:
@@ -2415,36 +2522,41 @@
                             if (IS_USER(s)
                                 || arm_feature(env, ARM_FEATURE_VFP3))
                                 return 1;
-                            gen_op_vfp_movl_T0_xreg(rn);
+                            tmp = load_cpu_field(vfp.xregs[rn]);
                             break;
                         case ARM_VFP_FPSCR:
-			    if (rd == 15)
-				gen_op_vfp_movl_T0_fpscr_flags();
-			    else
-				gen_op_vfp_movl_T0_fpscr();
+			    if (rd == 15) {
+                                tmp = load_cpu_field(vfp.xregs[ARM_VFP_FPSCR]);
+                                tcg_gen_andi_i32(tmp, tmp, 0xf0000000);
+                            } else {
+                                tmp = new_tmp();
+                                gen_helper_vfp_get_fpscr(tmp, cpu_env);
+                            }
                             break;
                         case ARM_VFP_MVFR0:
                         case ARM_VFP_MVFR1:
                             if (IS_USER(s)
                                 || !arm_feature(env, ARM_FEATURE_VFP3))
                                 return 1;
-                            gen_op_vfp_movl_T0_xreg(rn);
+                            tmp = load_cpu_field(vfp.xregs[rn]);
                             break;
                         default:
                             return 1;
                         }
                     } else {
                         gen_mov_F0_vreg(0, rn);
-                        gen_op_vfp_mrs();
+                        tmp = gen_vfp_mrs();
                     }
                     if (rd == 15) {
                         /* Set the 4 flag bits in the CPSR.  */
-                        gen_set_nzcv(cpu_T[0]);
-                    } else
-                        gen_movl_reg_T0(s, rd);
+                        gen_set_nzcv(tmp);
+                        dead_tmp(tmp);
+                    } else {
+                        store_reg(s, rd, tmp);
+                    }
                 } else {
                     /* arm->vfp */
-                    gen_movl_T0_reg(s, rd);
+                    tmp = load_reg(s, rd);
                     if (insn & (1 << 21)) {
                         rn >>= 1;
                         /* system register */
@@ -2455,24 +2567,25 @@
                             /* Writes are ignored.  */
                             break;
                         case ARM_VFP_FPSCR:
-                            gen_op_vfp_movl_fpscr_T0();
+                            gen_helper_vfp_set_fpscr(cpu_env, tmp);
+                            dead_tmp(tmp);
                             gen_lookup_tb(s);
                             break;
                         case ARM_VFP_FPEXC:
                             if (IS_USER(s))
                                 return 1;
-                            gen_op_vfp_movl_xreg_T0(rn);
+                            store_cpu_field(tmp, vfp.xregs[rn]);
                             gen_lookup_tb(s);
                             break;
                         case ARM_VFP_FPINST:
                         case ARM_VFP_FPINST2:
-                            gen_op_vfp_movl_xreg_T0(rn);
+                            store_cpu_field(tmp, vfp.xregs[rn]);
                             break;
                         default:
                             return 1;
                         }
                     } else {
-                        gen_op_vfp_msr();
+                        gen_vfp_msr(tmp);
                         gen_mov_vreg_F0(0, rn);
                     }
                 }
@@ -2640,14 +2753,15 @@
                         else
                             i |= 0x4000;
                         n |= i << 16;
+                        tcg_gen_movi_i64(cpu_F0d, ((uint64_t)n) << 32);
                     } else {
                         if (i & 0x40)
                             i |= 0x780;
                         else
                             i |= 0x800;
                         n |= i << 19;
+                        tcg_gen_movi_i32(cpu_F0d, ((uint64_t)n) << 32);
                     }
-                    gen_vfp_fconst(dp, n);
                     break;
                 case 15: /* extension space */
                     switch (rn) {
@@ -2678,9 +2792,9 @@
                         break;
                     case 15: /* single<->double conversion */
                         if (dp)
-                            gen_op_vfp_fcvtsd();
+                            gen_helper_vfp_fcvtsd(cpu_F0s, cpu_F0d, cpu_env);
                         else
-                            gen_op_vfp_fcvtds();
+                            gen_helper_vfp_fcvtds(cpu_F0d, cpu_F0s, cpu_env);
                         break;
                     case 16: /* fuito */
                         gen_vfp_uito(dp);
@@ -2814,31 +2928,35 @@
             if (insn & ARM_CP_RW_BIT) {
                 /* vfp->arm */
                 if (dp) {
-                    gen_mov_F0_vreg(1, rm);
-                    gen_op_vfp_mrrd();
-                    gen_movl_reg_T0(s, rd);
-                    gen_movl_reg_T1(s, rn);
+                    gen_mov_F0_vreg(0, rm * 2);
+                    tmp = gen_vfp_mrs();
+                    store_reg(s, rd, tmp);
+                    gen_mov_F0_vreg(0, rm * 2 + 1);
+                    tmp = gen_vfp_mrs();
+                    store_reg(s, rn, tmp);
                 } else {
                     gen_mov_F0_vreg(0, rm);
-                    gen_op_vfp_mrs();
-                    gen_movl_reg_T0(s, rn);
+                    tmp = gen_vfp_mrs();
+                    store_reg(s, rn, tmp);
                     gen_mov_F0_vreg(0, rm + 1);
-                    gen_op_vfp_mrs();
-                    gen_movl_reg_T0(s, rd);
+                    tmp = gen_vfp_mrs();
+                    store_reg(s, rd, tmp);
                 }
             } else {
                 /* arm->vfp */
                 if (dp) {
-                    gen_movl_T0_reg(s, rd);
-                    gen_movl_T1_reg(s, rn);
-                    gen_op_vfp_mdrr();
-                    gen_mov_vreg_F0(1, rm);
+                    tmp = load_reg(s, rd);
+                    gen_vfp_msr(tmp);
+                    gen_mov_vreg_F0(0, rm * 2);
+                    tmp = load_reg(s, rn);
+                    gen_vfp_msr(tmp);
+                    gen_mov_vreg_F0(0, rm * 2 + 1);
                 } else {
-                    gen_movl_T0_reg(s, rn);
-                    gen_op_vfp_msr();
+                    tmp = load_reg(s, rn);
+                    gen_vfp_msr(tmp);
                     gen_mov_vreg_F0(0, rm);
-                    gen_movl_T0_reg(s, rd);
-                    gen_op_vfp_msr();
+                    tmp = load_reg(s, rd);
+                    gen_vfp_msr(tmp);
                     gen_mov_vreg_F0(0, rm + 1);
                 }
             }
@@ -3993,9 +4111,9 @@
             break;
         case 31:
             if (size == 0)
-                gen_op_neon_recps_f32();
+                gen_helper_recps_f32(cpu_T[0], cpu_T[0], cpu_T[1], cpu_env);
             else
-                gen_op_neon_rsqrts_f32();
+                gen_helper_rsqrts_f32(cpu_T[0], cpu_T[0], cpu_T[1], cpu_env);
             break;
         default:
             abort();
@@ -4242,19 +4360,19 @@
             } else if (op == 15 || op == 16) {
                 /* VCVT fixed-point.  */
                 for (pass = 0; pass < (q ? 4 : 2); pass++) {
-                    gen_op_vfp_getreg_F0s(neon_reg_offset(rm, pass));
+                    tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, pass));
                     if (op & 1) {
                         if (u)
-                            gen_op_vfp_ultos(shift);
+                            gen_vfp_ulto(0, shift);
                         else
-                            gen_op_vfp_sltos(shift);
+                            gen_vfp_slto(0, shift);
                     } else {
                         if (u)
-                            gen_op_vfp_touls(shift);
+                            gen_vfp_toul(0, shift);
                         else
-                            gen_op_vfp_tosls(shift);
+                            gen_vfp_tosl(0, shift);
                     }
-                    gen_op_vfp_setreg_F0s(neon_reg_offset(rd, pass));
+                    tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, pass));
                 }
             } else {
                 return 1;
@@ -4898,7 +5016,8 @@
                 elementwise:
                     for (pass = 0; pass < (q ? 4 : 2); pass++) {
                         if (op == 30 || op == 31 || op >= 58) {
-                            gen_op_vfp_getreg_F0s(neon_reg_offset(rm, pass));
+                            tcg_gen_ld_f32(cpu_F0s, cpu_env,
+                                           neon_reg_offset(rm, pass));
                         } else {
                             NEON_GET_REG(T0, rm, pass);
                         }
@@ -5041,10 +5160,10 @@
                             gen_op_neon_ceq_f32();
                             break;
                         case 30: /* Float VABS */
-                            gen_op_vfp_abss();
+                            gen_vfp_abs(0);
                             break;
                         case 31: /* Float VNEG */
-                            gen_op_vfp_negs();
+                            gen_vfp_neg(0);
                             break;
                         case 32: /* VSWP */
                             NEON_GET_REG(T1, rd, pass);
@@ -5061,35 +5180,36 @@
                             NEON_SET_REG(T1, rm, pass);
                             break;
                         case 56: /* Integer VRECPE */
-                            gen_op_neon_recpe_u32();
+                            gen_helper_recpe_u32(cpu_T[0], cpu_T[0], cpu_env);
                             break;
                         case 57: /* Integer VRSQRTE */
-                            gen_op_neon_rsqrte_u32();
+                            gen_helper_rsqrte_u32(cpu_T[0], cpu_T[0], cpu_env);
                             break;
                         case 58: /* Float VRECPE */
-                            gen_op_neon_recpe_f32();
+                            gen_helper_recpe_f32(cpu_F0s, cpu_F0s, cpu_env);
                             break;
                         case 59: /* Float VRSQRTE */
-                            gen_op_neon_rsqrte_f32();
+                            gen_helper_rsqrte_f32(cpu_F0s, cpu_F0s, cpu_env);
                             break;
                         case 60: /* VCVT.F32.S32 */
-                            gen_op_vfp_tosizs();
+                            gen_vfp_tosiz(0);
                             break;
                         case 61: /* VCVT.F32.U32 */
-                            gen_op_vfp_touizs();
+                            gen_vfp_touiz(0);
                             break;
                         case 62: /* VCVT.S32.F32 */
-                            gen_op_vfp_sitos();
+                            gen_vfp_sito(0);
                             break;
                         case 63: /* VCVT.U32.F32 */
-                            gen_op_vfp_uitos();
+                            gen_vfp_uito(0);
                             break;
                         default:
                             /* Reserved: 21, 29, 39-56 */
                             return 1;
                         }
                         if (op == 30 || op == 31 || op >= 58) {
-                            gen_op_vfp_setreg_F0s(neon_reg_offset(rm, pass));
+                            tcg_gen_st_f32(cpu_F0s, cpu_env,
+                                           neon_reg_offset(rd, pass));
                         } else {
                             NEON_SET_REG(T0, rd, pass);
                         }
@@ -8062,6 +8182,10 @@
         dc->user = (env->uncached_cpsr & 0x1f) == ARM_CPU_MODE_USR;
     }
 #endif
+    cpu_F0s = tcg_temp_new(TCG_TYPE_I32);
+    cpu_F1s = tcg_temp_new(TCG_TYPE_I32);
+    cpu_F0d = tcg_temp_new(TCG_TYPE_I64);
+    cpu_F1d = tcg_temp_new(TCG_TYPE_I64);
     next_page_start = (pc_start & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE;
     lj = -1;
     /* Reset the conditional execution bits immediately. This avoids