ARM TCG conversion 14/16.

git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@4151 c046a42c-6fe2-441c-8c8c-71466251a162

ARM TCG conversion 14/16.
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@4151 c046a42c-6fe2-441c-8c8c-71466251a162
pbrook
1 parent 8f8e3aa4
Showing 7 changed files with 1779 additions and 1718 deletions
Makefile.target
target-arm/helper.c
target-arm/helpers.h
target-arm/neon_helper.c
target-arm/op.c
target-arm/op_helper.c
target-arm/op_neon.h
@@ -211,7 +211,7 @@ LIBOBJS+= op_helper.o helper.o
 endif
 ifeq ($(TARGET_BASE_ARCH), arm)
-LIBOBJS+= op_helper.o helper.o
+LIBOBJS+= op_helper.o helper.o neon_helper.o
 endif
 ifeq ($(TARGET_BASE_ARCH), sh4)
@@ -256,30 +256,6 @@ void cpu_arm_close(CPUARMState *env)
     free(env);
 }
-/* Polynomial multiplication is like integer multiplcation except the
-   partial products are XORed, not added.  */
-uint32_t helper_neon_mul_p8(uint32_t op1, uint32_t op2)
-{
-    uint32_t mask;
-    uint32_t result;
-    result = 0;
-    while (op1) {
-        mask = 0;
-        if (op1 & 1)
-            mask |= 0xff;
-        if (op1 & (1 << 8))
-            mask |= (0xff << 8);
-        if (op1 & (1 << 16))
-            mask |= (0xff << 16);
-        if (op1 & (1 << 24))
-            mask |= (0xff << 24);
-        result ^= op2 & mask;
-        op1 = (op1 >> 1) & 0x7f7f7f7f;
-        op2 = (op2 << 1) & 0xfefefefe;
-    }
-    return result;
-}
-
 uint32_t cpsr_read(CPUARMState *env)
 {
     int ZF;
@@ -376,6 +352,11 @@ uint32_t HELPER(rbit)(uint32_t x)
     return x;
 }
+uint32_t HELPER(abs)(uint32_t x)
+{
+    return ((int32_t)x < 0) ? -x : x;
+}
+
 #if defined(CONFIG_USER_ONLY)
 void do_interrupt (CPUState *env)
@@ -84,6 +84,7 @@ DEF_HELPER_1_1(double_saturate, uint32_t, (int32_t))
 DEF_HELPER_1_2(sdiv, int32_t, (int32_t, int32_t))
 DEF_HELPER_1_2(udiv, uint32_t, (uint32_t, uint32_t))
 DEF_HELPER_1_1(rbit, uint32_t, (uint32_t))
+DEF_HELPER_1_1(abs, uint32_t, (uint32_t))
 #define PAS_OP(pfx)  \
     DEF_HELPER_1_3(pfx ## add8, uint32_t, (uint32_t, uint32_t, uint32_t *)) \
@@ -208,6 +209,10 @@ 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 *))
 DEF_HELPER_1_4(neon_tbl, uint32_t, (uint32_t, uint32_t, uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_add_saturate_u64, uint64_t, (uint64_t, uint64_t))
+DEF_HELPER_1_2(neon_add_saturate_s64, uint64_t, (uint64_t, uint64_t))
+DEF_HELPER_1_2(neon_sub_saturate_u64, uint64_t, (uint64_t, uint64_t))
+DEF_HELPER_1_2(neon_sub_saturate_s64, uint64_t, (uint64_t, uint64_t))
 DEF_HELPER_1_2(add_cc, uint32_t, (uint32_t, uint32_t))
 DEF_HELPER_1_2(adc_cc, uint32_t, (uint32_t, uint32_t))
@@ -223,6 +228,209 @@ DEF_HELPER_1_2(shr_cc, uint32_t, (uint32_t, uint32_t))
 DEF_HELPER_1_2(sar_cc, uint32_t, (uint32_t, uint32_t))
 DEF_HELPER_1_2(ror_cc, uint32_t, (uint32_t, uint32_t))
+/* neon_helper.c */
+DEF_HELPER_1_3(neon_qadd_u8, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qadd_s8, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qadd_u16, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qadd_s16, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qsub_u8, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qsub_s8, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qsub_u16, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qsub_s16, uint32_t, (CPUState *, uint32_t, uint32_t))
+
+DEF_HELPER_1_2(neon_hadd_s8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_hadd_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_hadd_s16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_hadd_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_hadd_s32, int32_t, (int32_t, int32_t))
+DEF_HELPER_1_2(neon_hadd_u32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_rhadd_s8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_rhadd_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_rhadd_s16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_rhadd_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_rhadd_s32, int32_t, (int32_t, int32_t))
+DEF_HELPER_1_2(neon_rhadd_u32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_hsub_s8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_hsub_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_hsub_s16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_hsub_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_hsub_s32, int32_t, (int32_t, int32_t))
+DEF_HELPER_1_2(neon_hsub_u32, uint32_t, (uint32_t, uint32_t))
+
+DEF_HELPER_1_2(neon_cgt_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_cgt_s8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_cgt_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_cgt_s16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_cgt_u32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_cgt_s32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_cge_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_cge_s8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_cge_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_cge_s16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_cge_u32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_cge_s32, uint32_t, (uint32_t, uint32_t))
+
+DEF_HELPER_1_2(neon_min_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_min_s8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_min_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_min_s16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_min_u32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_min_s32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_max_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_max_s8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_max_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_max_s16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_max_u32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_max_s32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_pmin_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_pmin_s8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_pmin_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_pmin_s16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_pmin_u32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_pmin_s32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_pmax_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_pmax_s8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_pmax_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_pmax_s16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_pmax_u32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_pmax_s32, uint32_t, (uint32_t, uint32_t))
+
+DEF_HELPER_1_2(neon_abd_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_abd_s8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_abd_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_abd_s16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_abd_u32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_abd_s32, uint32_t, (uint32_t, uint32_t))
+
+DEF_HELPER_1_2(neon_shl_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_shl_s8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_shl_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_shl_s16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_shl_u32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_shl_s32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_shl_u64, uint64_t, (uint64_t, uint64_t))
+DEF_HELPER_1_2(neon_shl_s64, uint64_t, (uint64_t, uint64_t))
+DEF_HELPER_1_2(neon_rshl_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_rshl_s8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_rshl_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_rshl_s16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_rshl_u32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_rshl_s32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_rshl_u64, uint64_t, (uint64_t, uint64_t))
+DEF_HELPER_1_2(neon_rshl_s64, uint64_t, (uint64_t, uint64_t))
+DEF_HELPER_1_3(neon_qshl_u8, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qshl_s8, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qshl_u16, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qshl_s16, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qshl_u32, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qshl_s32, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qshl_u64, uint64_t, (CPUState *, uint64_t, uint64_t))
+DEF_HELPER_1_3(neon_qshl_s64, uint64_t, (CPUState *, uint64_t, uint64_t))
+DEF_HELPER_1_3(neon_qrshl_u8, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qrshl_s8, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qrshl_u16, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qrshl_s16, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qrshl_u32, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qrshl_s32, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qrshl_u64, uint64_t, (CPUState *, uint64_t, uint64_t))
+DEF_HELPER_1_3(neon_qrshl_s64, uint64_t, (CPUState *, uint64_t, uint64_t))
+
+DEF_HELPER_1_2(neon_add_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_add_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_padd_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_padd_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_sub_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_sub_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_mul_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_mul_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_mul_p8, uint32_t, (uint32_t, uint32_t))
+
+DEF_HELPER_1_2(neon_tst_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_tst_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_tst_u32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_ceq_u8, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_ceq_u16, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_ceq_u32, uint32_t, (uint32_t, uint32_t))
+
+DEF_HELPER_1_1(neon_abs_s8, uint32_t, (uint32_t))
+DEF_HELPER_1_1(neon_abs_s16, uint32_t, (uint32_t))
+DEF_HELPER_1_1(neon_clz_u8, uint32_t, (uint32_t))
+DEF_HELPER_1_1(neon_clz_u16, uint32_t, (uint32_t))
+DEF_HELPER_1_1(neon_cls_s8, uint32_t, (uint32_t))
+DEF_HELPER_1_1(neon_cls_s16, uint32_t, (uint32_t))
+DEF_HELPER_1_1(neon_cls_s32, uint32_t, (uint32_t))
+DEF_HELPER_1_1(neon_cnt_u8, uint32_t, (uint32_t))
+
+DEF_HELPER_1_3(neon_qdmulh_s16, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qrdmulh_s16, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qdmulh_s32, uint32_t, (CPUState *, uint32_t, uint32_t))
+DEF_HELPER_1_3(neon_qrdmulh_s32, uint32_t, (CPUState *, uint32_t, uint32_t))
+
+DEF_HELPER_1_1(neon_narrow_u8, uint32_t, (uint64_t))
+DEF_HELPER_1_1(neon_narrow_u16, uint32_t, (uint64_t))
+DEF_HELPER_1_2(neon_narrow_sat_u8, uint32_t, (CPUState *, uint64_t))
+DEF_HELPER_1_2(neon_narrow_sat_s8, uint32_t, (CPUState *, uint64_t))
+DEF_HELPER_1_2(neon_narrow_sat_u16, uint32_t, (CPUState *, uint64_t))
+DEF_HELPER_1_2(neon_narrow_sat_s16, uint32_t, (CPUState *, uint64_t))
+DEF_HELPER_1_2(neon_narrow_sat_u32, uint32_t, (CPUState *, uint64_t))
+DEF_HELPER_1_2(neon_narrow_sat_s32, uint32_t, (CPUState *, uint64_t))
+DEF_HELPER_1_1(neon_narrow_high_u8, uint32_t, (uint64_t))
+DEF_HELPER_1_1(neon_narrow_high_u16, uint32_t, (uint64_t))
+DEF_HELPER_1_1(neon_narrow_round_high_u8, uint32_t, (uint64_t))
+DEF_HELPER_1_1(neon_narrow_round_high_u16, uint32_t, (uint64_t))
+DEF_HELPER_1_1(neon_widen_u8, uint64_t, (uint32_t))
+DEF_HELPER_1_1(neon_widen_s8, uint64_t, (uint32_t))
+DEF_HELPER_1_1(neon_widen_u16, uint64_t, (uint32_t))
+DEF_HELPER_1_1(neon_widen_s16, uint64_t, (uint32_t))
+
+DEF_HELPER_1_2(neon_addl_u16, uint64_t, (uint64_t, uint64_t))
+DEF_HELPER_1_2(neon_addl_u32, uint64_t, (uint64_t, uint64_t))
+DEF_HELPER_1_2(neon_paddl_u16, uint64_t, (uint64_t, uint64_t))
+DEF_HELPER_1_2(neon_paddl_u32, uint64_t, (uint64_t, uint64_t))
+DEF_HELPER_1_2(neon_subl_u16, uint64_t, (uint64_t, uint64_t))
+DEF_HELPER_1_2(neon_subl_u32, uint64_t, (uint64_t, uint64_t))
+DEF_HELPER_1_3(neon_addl_saturate_s32, uint64_t, (CPUState *, uint64_t, uint64_t))
+DEF_HELPER_1_3(neon_addl_saturate_s64, uint64_t, (CPUState *, uint64_t, uint64_t))
+DEF_HELPER_1_2(neon_abdl_u16, uint64_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_abdl_s16, uint64_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_abdl_u32, uint64_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_abdl_s32, uint64_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_abdl_u64, uint64_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_abdl_s64, uint64_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_mull_u8, uint64_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_mull_s8, uint64_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_mull_u16, uint64_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_mull_s16, uint64_t, (uint32_t, uint32_t))
+
+DEF_HELPER_1_1(neon_negl_u16, uint64_t, (uint64_t))
+DEF_HELPER_1_1(neon_negl_u32, uint64_t, (uint64_t))
+DEF_HELPER_1_1(neon_negl_u64, uint64_t, (uint64_t))
+
+DEF_HELPER_1_2(neon_qabs_s8, uint32_t, (CPUState *, uint32_t))
+DEF_HELPER_1_2(neon_qabs_s16, uint32_t, (CPUState *, uint32_t))
+DEF_HELPER_1_2(neon_qabs_s32, uint32_t, (CPUState *, uint32_t))
+DEF_HELPER_1_2(neon_qneg_s8, uint32_t, (CPUState *, uint32_t))
+DEF_HELPER_1_2(neon_qneg_s16, uint32_t, (CPUState *, uint32_t))
+DEF_HELPER_1_2(neon_qneg_s32, uint32_t, (CPUState *, uint32_t))
+
+DEF_HELPER_0_0(neon_trn_u8, void, (void))
+DEF_HELPER_0_0(neon_trn_u16, void, (void))
+DEF_HELPER_0_0(neon_unzip_u8, void, (void))
+DEF_HELPER_0_0(neon_zip_u8, void, (void))
+DEF_HELPER_0_0(neon_zip_u16, void, (void))
+
+DEF_HELPER_1_2(neon_min_f32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_max_f32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_abd_f32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_add_f32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_sub_f32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_mul_f32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_ceq_f32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_cge_f32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_cgt_f32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_acge_f32, uint32_t, (uint32_t, uint32_t))
+DEF_HELPER_1_2(neon_acgt_f32, uint32_t, (uint32_t, uint32_t))
+
 #undef DEF_HELPER
 #undef DEF_HELPER_0_0
 #undef DEF_HELPER_0_1
+#include <stdlib.h>
+#include <stdio.h>
+
+#include "cpu.h"
+#include "exec-all.h"
+#include "helpers.h"
+
+#define SIGNBIT (uint32_t)0x80000000
+#define SIGNBIT64 ((uint64_t)1 << 63)
+
+#define SET_QC() env->vfp.xregs[ARM_VFP_FPSCR] = CPSR_Q
+
+static float_status neon_float_status;
+#define NFS &neon_float_status
+
+/* 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;
+}
+
+#define NEON_TYPE1(name, type) \
+typedef struct \
+{ \
+    type v1; \
+} neon_##name;
+#ifdef WORDS_BIGENDIAN
+#define NEON_TYPE2(name, type) \
+typedef struct \
+{ \
+    type v2; \
+    type v1; \
+} neon_##name;
+#define NEON_TYPE4(name, type) \
+typedef struct \
+{ \
+    type v4; \
+    type v3; \
+    type v2; \
+    type v1; \
+} neon_##name;
+#else
+#define NEON_TYPE2(name, type) \
+typedef struct \
+{ \
+    type v1; \
+    type v2; \
+} neon_##name;
+#define NEON_TYPE4(name, type) \
+typedef struct \
+{ \
+    type v1; \
+    type v2; \
+    type v3; \
+    type v4; \
+} neon_##name;
+#endif
+
+NEON_TYPE4(s8, int8_t)
+NEON_TYPE4(u8, uint8_t)
+NEON_TYPE2(s16, int16_t)
+NEON_TYPE2(u16, uint16_t)
+NEON_TYPE1(s32, int32_t)
+NEON_TYPE1(u32, uint32_t)
+#undef NEON_TYPE4
+#undef NEON_TYPE2
+#undef NEON_TYPE1
+
+/* Copy from a uint32_t to a vector structure type.  */
+#define NEON_UNPACK(vtype, dest, val) do { \
+    union { \
+        vtype v; \
+        uint32_t i; \
+    } conv_u; \
+    conv_u.i = (val); \
+    dest = conv_u.v; \
+    } while(0)
+
+/* Copy from a vector structure type to a uint32_t.  */
+#define NEON_PACK(vtype, dest, val) do { \
+    union { \
+        vtype v; \
+        uint32_t i; \
+    } conv_u; \
+    conv_u.v = (val); \
+    dest = conv_u.i; \
+    } while(0)
+
+#define NEON_DO1 \
+    NEON_FN(vdest.v1, vsrc1.v1, vsrc2.v1);
+#define NEON_DO2 \
+    NEON_FN(vdest.v1, vsrc1.v1, vsrc2.v1); \
+    NEON_FN(vdest.v2, vsrc1.v2, vsrc2.v2);
+#define NEON_DO4 \
+    NEON_FN(vdest.v1, vsrc1.v1, vsrc2.v1); \
+    NEON_FN(vdest.v2, vsrc1.v2, vsrc2.v2); \
+    NEON_FN(vdest.v3, vsrc1.v3, vsrc2.v3); \
+    NEON_FN(vdest.v4, vsrc1.v4, vsrc2.v4);
+
+#define NEON_VOP_BODY(vtype, n) \
+{ \
+    uint32_t res; \
+    vtype vsrc1; \
+    vtype vsrc2; \
+    vtype vdest; \
+    NEON_UNPACK(vtype, vsrc1, arg1); \
+    NEON_UNPACK(vtype, vsrc2, arg2); \
+    NEON_DO##n; \
+    NEON_PACK(vtype, res, vdest); \
+    return res; \
+}
+
+#define NEON_VOP(name, vtype, n) \
+uint32_t HELPER(glue(neon_,name))(uint32_t arg1, uint32_t arg2) \
+NEON_VOP_BODY(vtype, n)
+
+#define NEON_VOP_ENV(name, vtype, n) \
+uint32_t HELPER(glue(neon_,name))(CPUState *env, uint32_t arg1, uint32_t arg2) \
+NEON_VOP_BODY(vtype, n)
+
+/* Pairwise operations.  */
+/* For 32-bit elements each segment only contains a single element, so
+   the elementwise and pairwise operations are the same.  */
+#define NEON_PDO2 \
+    NEON_FN(vdest.v1, vsrc1.v1, vsrc1.v2); \
+    NEON_FN(vdest.v2, vsrc2.v1, vsrc2.v2);
+#define NEON_PDO4 \
+    NEON_FN(vdest.v1, vsrc1.v1, vsrc1.v2); \
+    NEON_FN(vdest.v2, vsrc1.v3, vsrc1.v4); \
+    NEON_FN(vdest.v3, vsrc2.v1, vsrc2.v2); \
+    NEON_FN(vdest.v4, vsrc2.v3, vsrc2.v4); \
+
+#define NEON_POP(name, vtype, n) \
+uint32_t HELPER(glue(neon_,name))(uint32_t arg1, uint32_t arg2) \
+{ \
+    uint32_t res; \
+    vtype vsrc1; \
+    vtype vsrc2; \
+    vtype vdest; \
+    NEON_UNPACK(vtype, vsrc1, arg1); \
+    NEON_UNPACK(vtype, vsrc2, arg2); \
+    NEON_PDO##n; \
+    NEON_PACK(vtype, res, vdest); \
+    return res; \
+}
+
+/* Unary operators.  */
+#define NEON_VOP1(name, vtype, n) \
+uint32_t HELPER(glue(neon_,name))(uint32_t arg) \
+{ \
+    vtype vsrc1; \
+    vtype vdest; \
+    NEON_UNPACK(vtype, vsrc1, arg); \
+    NEON_DO##n; \
+    NEON_PACK(vtype, arg, vdest); \
+    return arg; \
+}
+
+
+#define NEON_USAT(dest, src1, src2, type) do { \
+    uint32_t tmp = (uint32_t)src1 + (uint32_t)src2; \
+    if (tmp != (type)tmp) { \
+        SET_QC(); \
+        dest = ~0; \
+    } else { \
+        dest = tmp; \
+    }} while(0)
+#define NEON_FN(dest, src1, src2) NEON_USAT(dest, src1, src2, uint8_t)
+NEON_VOP_ENV(qadd_u8, neon_u8, 4)
+#undef NEON_FN
+#define NEON_FN(dest, src1, src2) NEON_USAT(dest, src1, src2, uint16_t)
+NEON_VOP_ENV(qadd_u16, neon_u16, 2)
+#undef NEON_FN
+#undef NEON_USAT
+
+#define NEON_SSAT(dest, src1, src2, type) do { \
+    int32_t tmp = (uint32_t)src1 + (uint32_t)src2; \
+    if (tmp != (type)tmp) { \
+        SET_QC(); \
+        if (src2 > 0) { \
+            tmp = (1 << (sizeof(type) * 8 - 1)) - 1; \
+        } else { \
+            tmp = 1 << (sizeof(type) * 8 - 1); \
+        } \
+    } \
+    dest = tmp; \
+    } while(0)
+#define NEON_FN(dest, src1, src2) NEON_SSAT(dest, src1, src2, int8_t)
+NEON_VOP_ENV(qadd_s8, neon_s8, 4)
+#undef NEON_FN
+#define NEON_FN(dest, src1, src2) NEON_SSAT(dest, src1, src2, int16_t)
+NEON_VOP_ENV(qadd_s16, neon_s16, 2)
+#undef NEON_FN
+#undef NEON_SSAT
+
+#define NEON_USAT(dest, src1, src2, type) do { \
+    uint32_t tmp = (uint32_t)src1 - (uint32_t)src2; \
+    if (tmp != (type)tmp) { \
+        SET_QC(); \
+        dest = 0; \
+    } else { \
+        dest = tmp; \
+    }} while(0)
+#define NEON_FN(dest, src1, src2) NEON_USAT(dest, src1, src2, uint8_t)
+NEON_VOP_ENV(qsub_u8, neon_u8, 4)
+#undef NEON_FN
+#define NEON_FN(dest, src1, src2) NEON_USAT(dest, src1, src2, uint16_t)
+NEON_VOP_ENV(qsub_u16, neon_u16, 2)
+#undef NEON_FN
+#undef NEON_USAT
+
+#define NEON_SSAT(dest, src1, src2, type) do { \
+    int32_t tmp = (uint32_t)src1 - (uint32_t)src2; \
+    if (tmp != (type)tmp) { \
+        SET_QC(); \
+        if (src2 < 0) { \
+            tmp = (1 << (sizeof(type) * 8 - 1)) - 1; \
+        } else { \
+            tmp = 1 << (sizeof(type) * 8 - 1); \
+        } \
+    } \
+    dest = tmp; \
+    } while(0)
+#define NEON_FN(dest, src1, src2) NEON_SSAT(dest, src1, src2, int8_t)
+NEON_VOP_ENV(qsub_s8, neon_s8, 4)
+#undef NEON_FN
+#define NEON_FN(dest, src1, src2) NEON_SSAT(dest, src1, src2, int16_t)
+NEON_VOP_ENV(qsub_s16, neon_s16, 2)
+#undef NEON_FN
+#undef NEON_SSAT
+
+#define NEON_FN(dest, src1, src2) dest = (src1 + src2) >> 1
+NEON_VOP(hadd_s8, neon_s8, 4)
+NEON_VOP(hadd_u8, neon_u8, 4)
+NEON_VOP(hadd_s16, neon_s16, 2)
+NEON_VOP(hadd_u16, neon_u16, 2)
+#undef NEON_FN
+
+int32_t HELPER(neon_hadd_s32)(int32_t src1, int32_t src2)
+{
+    int32_t dest;
+
+    dest = (src1 >> 1) + (src2 >> 1);
+    if (src1 & src2 & 1)
+        dest++;
+    return dest;
+}
+
+uint32_t HELPER(neon_hadd_u32)(uint32_t src1, uint32_t src2)
+{
+    uint32_t dest;
+
+    dest = (src1 >> 1) + (src2 >> 1);
+    if (src1 & src2 & 1)
+        dest++;
+    return dest;
+}
+
+#define NEON_FN(dest, src1, src2) dest = (src1 + src2 + 1) >> 1
+NEON_VOP(rhadd_s8, neon_s8, 4)
+NEON_VOP(rhadd_u8, neon_u8, 4)
+NEON_VOP(rhadd_s16, neon_s16, 2)
+NEON_VOP(rhadd_u16, neon_u16, 2)
+#undef NEON_FN
+
+int32_t HELPER(neon_rhadd_s32)(int32_t src1, int32_t src2)
+{
+    int32_t dest;
+
+    dest = (src1 >> 1) + (src2 >> 1);
+    if ((src1 | src2) & 1)
+        dest++;
+    return dest;
+}
+
+uint32_t HELPER(neon_rhadd_u32)(uint32_t src1, uint32_t src2)
+{
+    uint32_t dest;
+
+    dest = (src1 >> 1) + (src2 >> 1);
+    if ((src1 | src2) & 1)
+        dest++;
+    return dest;
+}
+
+#define NEON_FN(dest, src1, src2) dest = (src1 - src2) >> 1
+NEON_VOP(hsub_s8, neon_s8, 4)
+NEON_VOP(hsub_u8, neon_u8, 4)
+NEON_VOP(hsub_s16, neon_s16, 2)
+NEON_VOP(hsub_u16, neon_u16, 2)
+#undef NEON_FN
+
+int32_t HELPER(neon_hsub_s32)(int32_t src1, int32_t src2)
+{
+    int32_t dest;
+
+    dest = (src1 >> 1) - (src2 >> 1);
+    if ((~src1) & src2 & 1)
+        dest--;
+    return dest;
+}
+
+uint32_t HELPER(neon_hsub_u32)(uint32_t src1, uint32_t src2)
+{
+    uint32_t dest;
+
+    dest = (src1 >> 1) - (src2 >> 1);
+    if ((~src1) & src2 & 1)
+        dest--;
+    return dest;
+}
+
+#define NEON_FN(dest, src1, src2) dest = (src1 > src2) ? ~0 : 0
+NEON_VOP(cgt_s8, neon_s8, 4)
+NEON_VOP(cgt_u8, neon_u8, 4)
+NEON_VOP(cgt_s16, neon_s16, 2)
+NEON_VOP(cgt_u16, neon_u16, 2)
+NEON_VOP(cgt_s32, neon_s32, 1)
+NEON_VOP(cgt_u32, neon_u32, 1)
+#undef NEON_FN
+
+#define NEON_FN(dest, src1, src2) dest = (src1 >= src2) ? ~0 : 0
+NEON_VOP(cge_s8, neon_s8, 4)
+NEON_VOP(cge_u8, neon_u8, 4)
+NEON_VOP(cge_s16, neon_s16, 2)
+NEON_VOP(cge_u16, neon_u16, 2)
+NEON_VOP(cge_s32, neon_s32, 1)
+NEON_VOP(cge_u32, neon_u32, 1)
+#undef NEON_FN
+
+#define NEON_FN(dest, src1, src2) dest = (src1 < src2) ? src1 : src2
+NEON_VOP(min_s8, neon_s8, 4)
+NEON_VOP(min_u8, neon_u8, 4)
+NEON_VOP(min_s16, neon_s16, 2)
+NEON_VOP(min_u16, neon_u16, 2)
+NEON_VOP(min_s32, neon_s32, 1)
+NEON_VOP(min_u32, neon_u32, 1)
+NEON_POP(pmin_s8, neon_s8, 4)
+NEON_POP(pmin_u8, neon_u8, 4)
+NEON_POP(pmin_s16, neon_s16, 2)
+NEON_POP(pmin_u16, neon_u16, 2)
+#undef NEON_FN
+
+#define NEON_FN(dest, src1, src2) dest = (src1 > src2) ? src1 : src2
+NEON_VOP(max_s8, neon_s8, 4)
+NEON_VOP(max_u8, neon_u8, 4)
+NEON_VOP(max_s16, neon_s16, 2)
+NEON_VOP(max_u16, neon_u16, 2)
+NEON_VOP(max_s32, neon_s32, 1)
+NEON_VOP(max_u32, neon_u32, 1)
+NEON_POP(pmax_s8, neon_s8, 4)
+NEON_POP(pmax_u8, neon_u8, 4)
+NEON_POP(pmax_s16, neon_s16, 2)
+NEON_POP(pmax_u16, neon_u16, 2)
+#undef NEON_FN
+
+#define NEON_FN(dest, src1, src2) \
+    dest = (src1 > src2) ? (src1 - src2) : (src2 - src1)
+NEON_VOP(abd_s8, neon_s8, 4)
+NEON_VOP(abd_u8, neon_u8, 4)
+NEON_VOP(abd_s16, neon_s16, 2)
+NEON_VOP(abd_u16, neon_u16, 2)
+NEON_VOP(abd_s32, neon_s32, 1)
+NEON_VOP(abd_u32, neon_u32, 1)
+#undef NEON_FN
+
+#define NEON_FN(dest, src1, src2) do { \
+    int8_t tmp; \
+    tmp = (int8_t)src2; \
+    if (tmp >= sizeof(src1) * 8 || tmp <= -sizeof(src1) * 8) { \
+        dest = 0; \
+    } else if (tmp < 0) { \
+        dest = src1 >> -tmp; \
+    } else { \
+        dest = src1 << tmp; \
+    }} while (0)
+NEON_VOP(shl_u8, neon_u8, 4)
+NEON_VOP(shl_u16, neon_u16, 2)
+NEON_VOP(shl_u32, neon_u32, 1)
+#undef NEON_FN
+
+uint64_t HELPER(neon_shl_u64)(uint64_t val, uint64_t shiftop)
+{
+    int8_t shift = (int8_t)shiftop;
+    if (shift >= 64 || shift <= -64) {
+        val = 0;
+    } else if (shift < 0) {
+        val >>= -shift;
+    } else {
+        val <<= shift;
+    }
+    return val;
+}
+
+#define NEON_FN(dest, src1, src2) do { \
+    int8_t tmp; \
+    tmp = (int8_t)src2; \
+    if (tmp >= sizeof(src1) * 8) { \
+        dest = 0; \
+    } else if (tmp <= -sizeof(src1) * 8) { \
+        dest = src1 >> (sizeof(src1) * 8 - 1); \
+    } else if (tmp < 0) { \
+        dest = src1 >> -tmp; \
+    } else { \
+        dest = src1 << tmp; \
+    }} while (0)
+NEON_VOP(shl_s8, neon_s8, 4)
+NEON_VOP(shl_s16, neon_s16, 2)
+NEON_VOP(shl_s32, neon_s32, 1)
+#undef NEON_FN
+
+uint64_t HELPER(neon_shl_s64)(uint64_t valop, uint64_t shiftop)
+{
+    int8_t shift = (int8_t)shiftop;
+    int64_t val = valop;
+    if (shift >= 64) {
+        val = 0;
+    } else if (shift <= -64) {
+        val >>= 63;
+    } else if (shift < 0) {
+        val >>= -shift;
+    } else {
+        val <<= shift;
+    }
+    return val;
+}
+
+#define NEON_FN(dest, src1, src2) do { \
+    int8_t tmp; \
+    tmp = (int8_t)src2; \
+    if (tmp >= sizeof(src1) * 8) { \
+        dest = 0; \
+    } else if (tmp < -sizeof(src1) * 8) { \
+        dest >>= sizeof(src1) * 8 - 1; \
+    } else if (tmp == -sizeof(src1) * 8) { \
+        dest = src1 >> (tmp - 1); \
+        dest++; \
+        src2 >>= 1; \
+    } else if (tmp < 0) { \
+        dest = (src1 + (1 << (-1 - tmp))) >> -tmp; \
+    } else { \
+        dest = src1 << tmp; \
+    }} while (0)
+NEON_VOP(rshl_s8, neon_s8, 4)
+NEON_VOP(rshl_s16, neon_s16, 2)
+NEON_VOP(rshl_s32, neon_s32, 1)
+#undef NEON_FN
+
+uint64_t HELPER(neon_rshl_s64)(uint64_t valop, uint64_t shiftop)
+{
+    int8_t shift = (int8_t)shiftop;
+    int64_t val = valop;
+    if (shift >= 64) {
+        val = 0;
+    } else if (shift < -64) {
+        val >>= 63;
+    } else if (shift == -63) {
+        val >>= 63;
+        val++;
+        val >>= 1;
+    } else if (shift < 0) {
+        val = (val + ((int64_t)1 << (-1 - shift))) >> -shift;
+    } else {
+        val <<= shift;
+    }
+    return val;
+}
+
+#define NEON_FN(dest, src1, src2) do { \
+    int8_t tmp; \
+    tmp = (int8_t)src2; \
+    if (tmp >= sizeof(src1) * 8 || tmp < -sizeof(src1) * 8) { \
+        dest = 0; \
+    } else if (tmp == -sizeof(src1) * 8) { \
+        dest = src1 >> (tmp - 1); \
+    } else if (tmp < 0) { \
+        dest = (src1 + (1 << (-1 - tmp))) >> -tmp; \
+    } else { \
+        dest = src1 << tmp; \
+    }} while (0)
+NEON_VOP(rshl_u8, neon_u8, 4)
+NEON_VOP(rshl_u16, neon_u16, 2)
+NEON_VOP(rshl_u32, neon_u32, 1)
+#undef NEON_FN
+
+uint64_t HELPER(neon_rshl_u64)(uint64_t val, uint64_t shiftop)
+{
+    int8_t shift = (uint8_t)shiftop;
+    if (shift >= 64 || shift < 64) {
+        val = 0;
+    } else if (shift == -64) {
+        /* Rounding a 1-bit result just preserves that bit.  */
+        val >>= 63;
+    } if (shift < 0) {
+        val = (val + ((uint64_t)1 << (-1 - shift))) >> -shift;
+        val >>= -shift;
+    } else {
+        val <<= shift;
+    }
+    return val;
+}
+
+#define NEON_FN(dest, src1, src2) do { \
+    int8_t tmp; \
+    tmp = (int8_t)src2; \
+    if (tmp >= sizeof(src1) * 8) { \
+        if (src1) { \
+            SET_QC(); \
+            dest = ~0; \
+        } else { \
+            dest = 0; \
+        } \
+    } else if (tmp <= -sizeof(src1) * 8) { \
+        dest = 0; \
+    } else if (tmp < 0) { \
+        dest = src1 >> -tmp; \
+    } else { \
+        dest = src1 << tmp; \
+        if ((dest >> tmp) != src1) { \
+            SET_QC(); \
+            dest = ~0; \
+        } \
+    }} while (0)
+NEON_VOP_ENV(qshl_u8, neon_u8, 4)
+NEON_VOP_ENV(qshl_u16, neon_u16, 2)
+NEON_VOP_ENV(qshl_u32, neon_u32, 1)
+#undef NEON_FN
+
+uint64_t HELPER(neon_qshl_u64)(CPUState *env, uint64_t val, uint64_t shiftop)
+{
+    int8_t shift = (int8_t)shiftop;
+    if (shift >= 64) {
+        if (val) {
+            val = ~(uint64_t)0;
+            SET_QC();
+        } else {
+            val = 0;
+        }
+    } else if (shift <= -64) {
+        val = 0;
+    } else if (shift < 0) {
+        val >>= -shift;
+    } else {
+        uint64_t tmp = val;
+        val <<= shift;
+        if ((val >> shift) != tmp) {
+            SET_QC();
+            val = ~(uint64_t)0;
+        }
+    }
+    return val;
+}
+
+#define NEON_FN(dest, src1, src2) do { \
+    int8_t tmp; \
+    tmp = (int8_t)src2; \
+    if (tmp >= sizeof(src1) * 8) { \
+        if (src1) \
+            SET_QC(); \
+        dest = src1 >> 31; \
+    } else if (tmp <= -sizeof(src1) * 8) { \
+        dest = src1 >> 31; \
+    } else if (tmp < 0) { \
+        dest = src1 >> -tmp; \
+    } else { \
+        dest = src1 << tmp; \
+        if ((dest >> tmp) != src1) { \
+            SET_QC(); \
+            dest = src2 >> 31; \
+        } \
+    }} while (0)
+NEON_VOP_ENV(qshl_s8, neon_s8, 4)
+NEON_VOP_ENV(qshl_s16, neon_s16, 2)
+NEON_VOP_ENV(qshl_s32, neon_s32, 1)
+#undef NEON_FN
+
+uint64_t HELPER(neon_qshl_s64)(CPUState *env, uint64_t valop, uint64_t shiftop)
+{
+    int8_t shift = (uint8_t)shiftop;
+    int64_t val = valop;
+    if (shift >= 64) {
+        if (val) {
+            SET_QC();
+            val = (val >> 63) & ~SIGNBIT64;
+        }
+    } else if (shift <= 64) {
+        val >>= 63;
+    } else if (shift < 0) {
+        val >>= -shift;
+    } else {
+        int64_t tmp = val;
+        val <<= shift;
+        if ((val >> shift) != tmp) {
+            SET_QC();
+            val = (tmp >> 63) ^ ~SIGNBIT64;
+        }
+    }
+    return val;
+}
+
+
+/* FIXME: This is wrong.  */
+#define NEON_FN(dest, src1, src2) do { \
+    int8_t tmp; \
+    tmp = (int8_t)src2; \
+    if (tmp < 0) { \
+        dest = (src1 + (1 << (-1 - tmp))) >> -tmp; \
+    } else { \
+        dest = src1 << tmp; \
+        if ((dest >> tmp) != src1) { \
+            SET_QC(); \
+            dest = ~0; \
+        } \
+    }} while (0)
+NEON_VOP_ENV(qrshl_u8, neon_u8, 4)
+NEON_VOP_ENV(qrshl_u16, neon_u16, 2)
+NEON_VOP_ENV(qrshl_u32, neon_u32, 1)
+#undef NEON_FN
+
+uint64_t HELPER(neon_qrshl_u64)(CPUState *env, uint64_t val, uint64_t shiftop)
+{
+    int8_t shift = (int8_t)shiftop;
+    if (shift < 0) {
+        val = (val + (1 << (-1 - shift))) >> -shift;
+    } else { \
+        uint64_t tmp = val;
+        val <<= shift;
+        if ((val >> shift) != tmp) {
+            SET_QC();
+            val = ~0;
+        }
+    }
+    return val;
+}
+
+#define NEON_FN(dest, src1, src2) do { \
+    int8_t tmp; \
+    tmp = (int8_t)src2; \
+    if (tmp < 0) { \
+        dest = (src1 + (1 << (-1 - tmp))) >> -tmp; \
+    } else { \
+        dest = src1 << tmp; \
+        if ((dest >> tmp) != src1) { \
+            SET_QC(); \
+            dest = src1 >> 31; \
+        } \
+    }} while (0)
+NEON_VOP_ENV(qrshl_s8, neon_s8, 4)
+NEON_VOP_ENV(qrshl_s16, neon_s16, 2)
+NEON_VOP_ENV(qrshl_s32, neon_s32, 1)
+#undef NEON_FN
+
+uint64_t HELPER(neon_qrshl_s64)(CPUState *env, uint64_t valop, uint64_t shiftop)
+{
+    int8_t shift = (uint8_t)shiftop;
+    int64_t val = valop;
+
+    if (shift < 0) {
+        val = (val + (1 << (-1 - shift))) >> -shift;
+    } else {
+        int64_t tmp = val;;
+        val <<= shift;
+        if ((val >> shift) != tmp) {
+            SET_QC();
+            val = tmp >> 31;
+        }
+    }
+    return val;
+}
+
+uint32_t HELPER(neon_add_u8)(uint32_t a, uint32_t b)
+{
+    uint32_t mask;
+    mask = (a ^ b) & 0x80808080u;
+    a &= ~0x80808080u;
+    b &= ~0x80808080u;
+    return (a + b) ^ mask;
+}
+
+uint32_t HELPER(neon_add_u16)(uint32_t a, uint32_t b)
+{
+    uint32_t mask;
+    mask = (a ^ b) & 0x80008000u;
+    a &= ~0x80008000u;
+    b &= ~0x80008000u;
+    return (a + b) ^ mask;
+}
+
+#define NEON_FN(dest, src1, src2) dest = src1 + src2
+NEON_POP(padd_u8, neon_u8, 4)
+NEON_POP(padd_u16, neon_u16, 2)
+#undef NEON_FN
+
+#define NEON_FN(dest, src1, src2) dest = src1 - src2
+NEON_VOP(sub_u8, neon_u8, 4)
+NEON_VOP(sub_u16, neon_u16, 2)
+#undef NEON_FN
+
+#define NEON_FN(dest, src1, src2) dest = src1 * src2
+NEON_VOP(mul_u8, neon_u8, 4)
+NEON_VOP(mul_u16, neon_u16, 2)
+#undef NEON_FN
+
+/* Polynomial multiplication is like integer multiplcation except the
+   partial products are XORed, not added.  */
+uint32_t HELPER(neon_mul_p8)(uint32_t op1, uint32_t op2)
+{
+    uint32_t mask;
+    uint32_t result;
+    result = 0;
+    while (op1) {
+        mask = 0;
+        if (op1 & 1)
+            mask |= 0xff;
+        if (op1 & (1 << 8))
+            mask |= (0xff << 8);
+        if (op1 & (1 << 16))
+            mask |= (0xff << 16);
+        if (op1 & (1 << 24))
+            mask |= (0xff << 24);
+        result ^= op2 & mask;
+        op1 = (op1 >> 1) & 0x7f7f7f7f;
+        op2 = (op2 << 1) & 0xfefefefe;
+    }
+    return result;
+}
+
+#define NEON_FN(dest, src1, src2) dest = (src1 & src2) ? -1 : 0
+NEON_VOP(tst_u8, neon_u8, 4)
+NEON_VOP(tst_u16, neon_u16, 2)
+NEON_VOP(tst_u32, neon_u32, 1)
+#undef NEON_FN
+
+#define NEON_FN(dest, src1, src2) dest = (src1 == src2) ? -1 : 0
+NEON_VOP(ceq_u8, neon_u8, 4)
+NEON_VOP(ceq_u16, neon_u16, 2)
+NEON_VOP(ceq_u32, neon_u32, 1)
+#undef NEON_FN
+
+#define NEON_FN(dest, src, dummy) dest = (src < 0) ? -src : src
+NEON_VOP1(abs_s8, neon_s8, 4)
+NEON_VOP1(abs_s16, neon_s16, 2)
+#undef NEON_FN
+
+/* Count Leading Sign/Zero Bits.  */
+static inline int do_clz8(uint8_t x)
+{
+    int n;
+    for (n = 8; x; n--)
+        x >>= 1;
+    return n;
+}
+
+static inline int do_clz16(uint16_t x)
+{
+    int n;
+    for (n = 16; x; n--)
+        x >>= 1;
+    return n;
+}
+
+#define NEON_FN(dest, src, dummy) dest = do_clz8(src)
+NEON_VOP1(clz_u8, neon_u8, 4)
+#undef NEON_FN
+
+#define NEON_FN(dest, src, dummy) dest = do_clz16(src)
+NEON_VOP1(clz_u16, neon_u16, 2)
+#undef NEON_FN
+
+#define NEON_FN(dest, src, dummy) dest = do_clz8((src < 0) ? ~src : src) - 1
+NEON_VOP1(cls_s8, neon_s8, 4)
+#undef NEON_FN
+
+#define NEON_FN(dest, src, dummy) dest = do_clz16((src < 0) ? ~src : src) - 1
+NEON_VOP1(cls_s16, neon_s16, 2)
+#undef NEON_FN
+
+uint32_t HELPER(neon_cls_s32)(uint32_t x)
+{
+    int count;
+    if ((int32_t)x < 0)
+        x = ~x;
+    for (count = 32; x; count--)
+        x = x >> 1;
+    return count - 1;
+}
+
+/* Bit count.  */
+uint32_t HELPER(neon_cnt_u8)(uint32_t x)
+{
+    x = (x & 0x55555555) + ((x >>  1) & 0x55555555);
+    x = (x & 0x33333333) + ((x >>  2) & 0x33333333);
+    x = (x & 0x0f0f0f0f) + ((x >>  4) & 0x0f0f0f0f);
+    return x;
+}
+
+#define NEON_QDMULH16(dest, src1, src2, round) do { \
+    uint32_t tmp = (int32_t)(int16_t) src1 * (int16_t) src2; \
+    if ((tmp ^ (tmp << 1)) & SIGNBIT) { \
+        SET_QC(); \
+        tmp = (tmp >> 31) ^ ~SIGNBIT; \
+    } \
+    tmp <<= 1; \
+    if (round) { \
+        int32_t old = tmp; \
+        tmp += 1 << 15; \
+        if ((int32_t)tmp < old) { \
+            SET_QC(); \
+            tmp = SIGNBIT - 1; \
+        } \
+    } \
+    dest = tmp >> 16; \
+    } while(0)
+#define NEON_FN(dest, src1, src2) NEON_QDMULH16(dest, src1, src2, 0)
+NEON_VOP_ENV(qdmulh_s16, neon_s16, 2)
+#undef NEON_FN
+#define NEON_FN(dest, src1, src2) NEON_QDMULH16(dest, src1, src2, 1)
+NEON_VOP_ENV(qrdmulh_s16, neon_s16, 2)
+#undef NEON_FN
+#undef NEON_QDMULH16
+
+#define NEON_QDMULH32(dest, src1, src2, round) do { \
+    uint64_t tmp = (int64_t)(int32_t) src1 * (int32_t) src2; \
+    if ((tmp ^ (tmp << 1)) & SIGNBIT64) { \
+        SET_QC(); \
+        tmp = (tmp >> 63) ^ ~SIGNBIT64; \
+    } else { \
+        tmp <<= 1; \
+    } \
+    if (round) { \
+        int64_t old = tmp; \
+        tmp += (int64_t)1 << 31; \
+        if ((int64_t)tmp < old) { \
+            SET_QC(); \
+            tmp = SIGNBIT64 - 1; \
+        } \
+    } \
+    dest = tmp >> 32; \
+    } while(0)
+#define NEON_FN(dest, src1, src2) NEON_QDMULH32(dest, src1, src2, 0)
+NEON_VOP_ENV(qdmulh_s32, neon_s32, 1)
+#undef NEON_FN
+#define NEON_FN(dest, src1, src2) NEON_QDMULH32(dest, src1, src2, 1)
+NEON_VOP_ENV(qrdmulh_s32, neon_s32, 1)
+#undef NEON_FN
+#undef NEON_QDMULH32
+
+uint32_t HELPER(neon_narrow_u8)(uint64_t x)
+{
+    return (x & 0xffu) | ((x >> 8) & 0xff00u) | ((x >> 16) & 0xff0000u)
+           | ((x >> 24) & 0xff000000u);
+}
+
+uint32_t HELPER(neon_narrow_u16)(uint64_t x)
+{
+    return (x & 0xffffu) | ((x >> 16) & 0xffff0000u);
+}
+
+uint32_t HELPER(neon_narrow_high_u8)(uint64_t x)
+{
+    return ((x >> 8) & 0xff) | ((x >> 16) & 0xff00)
+            | ((x >> 24) & 0xff0000) | ((x >> 32) & 0xff000000);
+}
+
+uint32_t HELPER(neon_narrow_high_u16)(uint64_t x)
+{
+    return ((x >> 16) & 0xffff) | ((x >> 32) & 0xffff0000);
+}
+
+uint32_t HELPER(neon_narrow_round_high_u8)(uint64_t x)
+{
+    x &= 0xff80ff80ff80ff80ull;
+    x += 0x0080008000800080ull;
+    return ((x >> 8) & 0xff) | ((x >> 16) & 0xff00)
+            | ((x >> 24) & 0xff0000) | ((x >> 32) & 0xff000000);
+}
+
+uint32_t HELPER(neon_narrow_round_high_u16)(uint64_t x)
+{
+    x &= 0xffff8000ffff8000ull;
+    x += 0x0000800000008000ull;
+    return ((x >> 16) & 0xffff) | ((x >> 32) & 0xffff0000);
+}
+
+uint32_t HELPER(neon_narrow_sat_u8)(CPUState *env, uint64_t x)
+{
+    uint16_t s;
+    uint8_t d;
+    uint32_t res = 0;
+#define SAT8(n) \
+    s = x >> n; \
+    if (s > 0xff) { \
+        d = 0xff; \
+        SET_QC(); \
+    } else  { \
+        d = s; \
+    } \
+    res |= (uint32_t)d << (n / 2);
+
+    SAT8(0);
+    SAT8(16);
+    SAT8(32);
+    SAT8(48);
+#undef SAT8
+    return res;
+}
+
+uint32_t HELPER(neon_narrow_sat_s8)(CPUState *env, uint64_t x)
+{
+    int16_t s;
+    uint8_t d;
+    uint32_t res = 0;
+#define SAT8(n) \
+    s = x >> n; \
+    if (s != (int8_t)s) { \
+        d = (s >> 15) ^ 0x7f; \
+        SET_QC(); \
+    } else  { \
+        d = s; \
+    } \
+    res |= (uint32_t)d << (n / 2);
+
+    SAT8(0);
+    SAT8(16);
+    SAT8(32);
+    SAT8(48);
+#undef SAT8
+    return res;
+}
+
+uint32_t HELPER(neon_narrow_sat_u16)(CPUState *env, uint64_t x)
+{
+    uint32_t high;
+    uint32_t low;
+    low = x;
+    if (low > 0xffff) {
+        low = 0xffff;
+        SET_QC();
+    }
+    high = x >> 32;
+    if (high > 0xffff) {
+        high = 0xffff;
+        SET_QC();
+    }
+    return low | (high << 16);
+}
+
+uint32_t HELPER(neon_narrow_sat_s16)(CPUState *env, uint64_t x)
+{
+    int32_t low;
+    int32_t high;
+    low = x;
+    if (low != (int16_t)low) {
+        low = (low >> 31) ^ 0x7fff;
+        SET_QC();
+    }
+    high = x >> 32;
+    if (high != (int16_t)high) {
+        high = (high >> 31) ^ 0x7fff;
+        SET_QC();
+    }
+    return (uint16_t)low | (high << 16);
+}
+
+uint32_t HELPER(neon_narrow_sat_u32)(CPUState *env, uint64_t x)
+{
+    if (x > 0xffffffffu) {
+        SET_QC();
+        return 0xffffffffu;
+    }
+    return x;
+}
+
+uint32_t HELPER(neon_narrow_sat_s32)(CPUState *env, uint64_t x)
+{
+    if ((int64_t)x != (int32_t)x) {
+        SET_QC();
+        return (x >> 63) ^ 0x7fffffff;
+    }
+    return x;
+}
+
+uint64_t HELPER(neon_widen_u8)(uint32_t x)
+{
+    uint64_t tmp;
+    uint64_t ret;
+    ret = (uint8_t)x;
+    tmp = (uint8_t)(x >> 8);
+    ret |= tmp << 16;
+    tmp = (uint8_t)(x >> 16);
+    ret |= tmp << 32;
+    tmp = (uint8_t)(x >> 24);
+    ret |= tmp << 48;
+    return ret;
+}
+
+uint64_t HELPER(neon_widen_s8)(uint32_t x)
+{
+    uint64_t tmp;
+    uint64_t ret;
+    ret = (uint16_t)(int8_t)x;
+    tmp = (uint16_t)(int8_t)(x >> 8);
+    ret |= tmp << 16;
+    tmp = (uint16_t)(int8_t)(x >> 16);
+    ret |= tmp << 32;
+    tmp = (uint16_t)(int8_t)(x >> 24);
+    ret |= tmp << 48;
+    return ret;
+}
+
+uint64_t HELPER(neon_widen_u16)(uint32_t x)
+{
+    uint64_t high = (uint16_t)(x >> 16);
+    return ((uint16_t)x) | (high << 32);
+}
+
+uint64_t HELPER(neon_widen_s16)(uint32_t x)
+{
+    uint64_t high = (int16_t)(x >> 16);
+    return ((uint32_t)(int16_t)x) | (high << 32);
+}
+
+uint64_t HELPER(neon_addl_u16)(uint64_t a, uint64_t b)
+{
+    uint64_t mask;
+    mask = (a ^ b) & 0x8000800080008000ull;
+    a &= ~0x8000800080008000ull;
+    b &= ~0x8000800080008000ull;
+    return (a + b) ^ mask;
+}
+
+uint64_t HELPER(neon_addl_u32)(uint64_t a, uint64_t b)
+{
+    uint64_t mask;
+    mask = (a ^ b) & 0x8000000080000000ull;
+    a &= ~0x8000000080000000ull;
+    b &= ~0x8000000080000000ull;
+    return (a + b) ^ mask;
+}
+
+uint64_t HELPER(neon_paddl_u16)(uint64_t a, uint64_t b)
+{
+    uint64_t tmp;
+    uint64_t tmp2;
+
+    tmp = a & 0x0000ffff0000ffffull;
+    tmp += (a >> 16) & 0x0000ffff0000ffffull;
+    tmp2 = b & 0xffff0000ffff0000ull;
+    tmp2 += (b << 16) & 0xffff0000ffff0000ull;
+    return    ( tmp         & 0xffff)
+            | ((tmp  >> 16) & 0xffff0000ull)
+            | ((tmp2 << 16) & 0xffff00000000ull)
+            | ( tmp2        & 0xffff000000000000ull);
+}
+
+uint64_t HELPER(neon_paddl_u32)(uint64_t a, uint64_t b)
+{
+    uint32_t low = a + (a >> 32);
+    uint32_t high = b + (b >> 32);
+    return low + ((uint64_t)high << 32);
+}
+
+uint64_t HELPER(neon_subl_u16)(uint64_t a, uint64_t b)
+{
+    uint64_t mask;
+    mask = (a ^ ~b) & 0x8000800080008000ull;
+    a |= 0x8000800080008000ull;
+    b &= ~0x8000800080008000ull;
+    return (a - b) ^ mask;
+}
+
+uint64_t HELPER(neon_subl_u32)(uint64_t a, uint64_t b)
+{
+    uint64_t mask;
+    mask = (a ^ ~b) & 0x8000000080000000ull;
+    a |= 0x8000000080000000ull;
+    b &= ~0x8000000080000000ull;
+    return (a - b) ^ mask;
+}
+
+uint64_t HELPER(neon_addl_saturate_s32)(CPUState *env, uint64_t a, uint64_t b)
+{
+    uint32_t x, y;
+    uint32_t low, high;
+
+    x = a;
+    y = b;
+    low = x + y;
+    if (((low ^ x) & SIGNBIT) && !((x ^ y) & SIGNBIT)) {
+        SET_QC();
+        low = ((int32_t)x >> 31) ^ ~SIGNBIT;
+    }
+    x = a >> 32;
+    y = b >> 32;
+    high = x + y;
+    if (((high ^ x) & SIGNBIT) && !((x ^ y) & SIGNBIT)) {
+        SET_QC();
+        high = ((int32_t)x >> 31) ^ ~SIGNBIT;
+    }
+    return low | ((uint64_t)high << 32);
+}
+
+uint64_t HELPER(neon_addl_saturate_s64)(CPUState *env, uint64_t a, uint64_t b)
+{
+    uint64_t result;
+
+    result = a + b;
+    if (((result ^ a) & SIGNBIT64) && !((a ^ b) & SIGNBIT64)) {
+        SET_QC();
+        result = ((int64_t)a >> 63) ^ ~SIGNBIT64;
+    }
+    return result;
+}
+
+#define DO_ABD(dest, x, y, type) do { \
+    type tmp_x = x; \
+    type tmp_y = y; \
+    dest = ((tmp_x > tmp_y) ? tmp_x - tmp_y : tmp_y - tmp_x); \
+    } while(0)
+
+uint64_t HELPER(neon_abdl_u16)(uint32_t a, uint32_t b)
+{
+    uint64_t tmp;
+    uint64_t result;
+    DO_ABD(result, a, b, uint8_t);
+    DO_ABD(tmp, a >> 8, b >> 8, uint8_t);
+    result |= tmp << 16;
+    DO_ABD(tmp, a >> 16, b >> 16, uint8_t);
+    result |= tmp << 32;
+    DO_ABD(tmp, a >> 24, b >> 24, uint8_t);
+    result |= tmp << 48;
+    return result;
+}
+
+uint64_t HELPER(neon_abdl_s16)(uint32_t a, uint32_t b)
+{
+    uint64_t tmp;
+    uint64_t result;
+    DO_ABD(result, a, b, int8_t);
+    DO_ABD(tmp, a >> 8, b >> 8, int8_t);
+    result |= tmp << 16;
+    DO_ABD(tmp, a >> 16, b >> 16, int8_t);
+    result |= tmp << 32;
+    DO_ABD(tmp, a >> 24, b >> 24, int8_t);
+    result |= tmp << 48;
+    return result;
+}
+
+uint64_t HELPER(neon_abdl_u32)(uint32_t a, uint32_t b)
+{
+    uint64_t tmp;
+    uint64_t result;
+    DO_ABD(result, a, b, uint16_t);
+    DO_ABD(tmp, a >> 16, b >> 16, uint16_t);
+    return result | (tmp << 32);
+}
+
+uint64_t HELPER(neon_abdl_s32)(uint32_t a, uint32_t b)
+{
+    uint64_t tmp;
+    uint64_t result;
+    DO_ABD(result, a, b, int16_t);
+    DO_ABD(tmp, a >> 16, b >> 16, int16_t);
+    return result | (tmp << 32);
+}
+
+uint64_t HELPER(neon_abdl_u64)(uint32_t a, uint32_t b)
+{
+    uint64_t result;
+    DO_ABD(result, a, b, uint32_t);
+    return result;
+}
+
+uint64_t HELPER(neon_abdl_s64)(uint32_t a, uint32_t b)
+{
+    uint64_t result;
+    DO_ABD(result, a, b, int32_t);
+    return result;
+}
+#undef DO_ABD
+
+/* Widening multiply. Named type is the source type.  */
+#define DO_MULL(dest, x, y, type1, type2) do { \
+    type1 tmp_x = x; \
+    type1 tmp_y = y; \
+    dest = (type2)((type2)tmp_x * (type2)tmp_y); \
+    } while(0)
+
+uint64_t HELPER(neon_mull_u8)(uint32_t a, uint32_t b)
+{
+    uint64_t tmp;
+    uint64_t result;
+
+    DO_MULL(result, a, b, uint8_t, uint16_t);
+    DO_MULL(tmp, a >> 8, b >> 8, uint8_t, uint16_t);
+    result |= tmp << 16;
+    DO_MULL(tmp, a >> 16, b >> 16, uint8_t, uint16_t);
+    result |= tmp << 32;
+    DO_MULL(tmp, a >> 24, b >> 24, uint8_t, uint16_t);
+    result |= tmp << 48;
+    return result;
+}
+
+uint64_t HELPER(neon_mull_s8)(uint32_t a, uint32_t b)
+{
+    uint64_t tmp;
+    uint64_t result;
+
+    DO_MULL(result, a, b, int8_t, uint16_t);
+    DO_MULL(tmp, a >> 8, b >> 8, int8_t, uint16_t);
+    result |= tmp << 16;
+    DO_MULL(tmp, a >> 16, b >> 16, int8_t, uint16_t);
+    result |= tmp << 32;
+    DO_MULL(tmp, a >> 24, b >> 24, int8_t, uint16_t);
+    result |= tmp << 48;
+    return result;
+}
+
+uint64_t HELPER(neon_mull_u16)(uint32_t a, uint32_t b)
+{
+    uint64_t tmp;
+    uint64_t result;
+
+    DO_MULL(result, a, b, uint16_t, uint32_t);
+    DO_MULL(tmp, a >> 16, b >> 16, uint16_t, uint32_t);
+    return result | (tmp << 32);
+}
+
+uint64_t HELPER(neon_mull_s16)(uint32_t a, uint32_t b)
+{
+    uint64_t tmp;
+    uint64_t result;
+
+    DO_MULL(result, a, b, int16_t, uint32_t);
+    DO_MULL(tmp, a >> 16, b >> 16, int16_t, uint32_t);
+    return result | (tmp << 32);
+}
+
+uint64_t HELPER(neon_negl_u16)(uint64_t x)
+{
+    uint16_t tmp;
+    uint64_t result;
+    result = (uint16_t)-x;
+    tmp = -(x >> 16);
+    result |= (uint64_t)tmp << 16;
+    tmp = -(x >> 32);
+    result |= (uint64_t)tmp << 32;
+    tmp = -(x >> 48);
+    result |= (uint64_t)tmp << 48;
+    return result;
+}
+
+#include <stdio.h>
+uint64_t HELPER(neon_negl_u32)(uint64_t x)
+{
+    uint32_t low = -x;
+    uint32_t high = -(x >> 32);
+    return low | ((uint64_t)high << 32);
+}
+
+/* FIXME:  There should be a native op for this.  */
+uint64_t HELPER(neon_negl_u64)(uint64_t x)
+{
+    return -x;
+}
+
+/* Saturnating sign manuipulation.  */
+/* ??? Make these use NEON_VOP1 */
+#define DO_QABS8(x) do { \
+    if (x == (int8_t)0x80) { \
+        x = 0x7f; \
+        SET_QC(); \
+    } else if (x < 0) { \
+        x = -x; \
+    }} while (0)
+uint32_t HELPER(neon_qabs_s8)(CPUState *env, uint32_t x)
+{
+    neon_s8 vec;
+    NEON_UNPACK(neon_s8, vec, x);
+    DO_QABS8(vec.v1);
+    DO_QABS8(vec.v2);
+    DO_QABS8(vec.v3);
+    DO_QABS8(vec.v4);
+    NEON_PACK(neon_s8, x, vec);
+    return x;
+}
+#undef DO_QABS8
+
+#define DO_QNEG8(x) do { \
+    if (x == (int8_t)0x80) { \
+        x = 0x7f; \
+        SET_QC(); \
+    } else { \
+        x = -x; \
+    }} while (0)
+uint32_t HELPER(neon_qneg_s8)(CPUState *env, uint32_t x)
+{
+    neon_s8 vec;
+    NEON_UNPACK(neon_s8, vec, x);
+    DO_QNEG8(vec.v1);
+    DO_QNEG8(vec.v2);
+    DO_QNEG8(vec.v3);
+    DO_QNEG8(vec.v4);
+    NEON_PACK(neon_s8, x, vec);
+    return x;
+}
+#undef DO_QNEG8
+
+#define DO_QABS16(x) do { \
+    if (x == (int16_t)0x8000) { \
+        x = 0x7fff; \
+        SET_QC(); \
+    } else if (x < 0) { \
+        x = -x; \
+    }} while (0)
+uint32_t HELPER(neon_qabs_s16)(CPUState *env, uint32_t x)
+{
+    neon_s16 vec;
+    NEON_UNPACK(neon_s16, vec, x);
+    DO_QABS16(vec.v1);
+    DO_QABS16(vec.v2);
+    NEON_PACK(neon_s16, x, vec);
+    return x;
+}
+#undef DO_QABS16
+
+#define DO_QNEG16(x) do { \
+    if (x == (int16_t)0x8000) { \
+        x = 0x7fff; \
+        SET_QC(); \
+    } else { \
+        x = -x; \
+    }} while (0)
+uint32_t HELPER(neon_qneg_s16)(CPUState *env, uint32_t x)
+{
+    neon_s16 vec;
+    NEON_UNPACK(neon_s16, vec, x);
+    DO_QNEG16(vec.v1);
+    DO_QNEG16(vec.v2);
+    NEON_PACK(neon_s16, x, vec);
+    return x;
+}
+#undef DO_QNEG16
+
+uint32_t HELPER(neon_qabs_s32)(CPUState *env, uint32_t x)
+{
+    if (x == SIGNBIT) {
+        SET_QC();
+        x = ~SIGNBIT;
+    } else if ((int32_t)x < 0) {
+        x = -x;
+    }
+    return x;
+}
+
+uint32_t HELPER(neon_qneg_s32)(CPUState *env, uint32_t x)
+{
+    if (x == SIGNBIT) {
+        SET_QC();
+        x = ~SIGNBIT;
+    } else {
+        x = -x;
+    }
+    return x;
+}
+
+/* NEON Float helpers.  */
+uint32_t HELPER(neon_min_f32)(uint32_t a, uint32_t b)
+{
+    float32 f0 = vfp_itos(a);
+    float32 f1 = vfp_itos(b);
+    return (float32_compare_quiet(f0, f1, NFS) == -1) ? a : b;
+}
+
+uint32_t HELPER(neon_max_f32)(uint32_t a, uint32_t b)
+{
+    float32 f0 = vfp_itos(a);
+    float32 f1 = vfp_itos(b);
+    return (float32_compare_quiet(f0, f1, NFS) == 1) ? a : b;
+}
+
+uint32_t HELPER(neon_abd_f32)(uint32_t a, uint32_t b)
+{
+    float32 f0 = vfp_itos(a);
+    float32 f1 = vfp_itos(b);
+    return vfp_stoi((float32_compare_quiet(f0, f1, NFS) == 1)
+                    ? float32_sub(f0, f1, NFS)
+                    : float32_sub(f1, f0, NFS));
+}
+
+uint32_t HELPER(neon_add_f32)(uint32_t a, uint32_t b)
+{
+    return vfp_stoi(float32_add(vfp_itos(a), vfp_itos(b), NFS));
+}
+
+uint32_t HELPER(neon_sub_f32)(uint32_t a, uint32_t b)
+{
+    return vfp_stoi(float32_sub(vfp_itos(a), vfp_itos(b), NFS));
+}
+
+uint32_t HELPER(neon_mul_f32)(uint32_t a, uint32_t b)
+{
+    return vfp_stoi(float32_mul(vfp_itos(a), vfp_itos(b), NFS));
+}
+
+/* Floating point comparisons produce an integer result.  */
+#define NEON_VOP_FCMP(name, cmp) \
+uint32_t HELPER(neon_##name)(uint32_t a, uint32_t b) \
+{ \
+    if (float32_compare_quiet(vfp_itos(a), vfp_itos(b), NFS) cmp 0) \
+        return ~0; \
+    else \
+        return 0; \
+}
+
+NEON_VOP_FCMP(ceq_f32, ==)
+NEON_VOP_FCMP(cge_f32, >=)
+NEON_VOP_FCMP(cgt_f32, >)
+
+uint32_t HELPER(neon_acge_f32)(uint32_t a, uint32_t b)
+{
+    float32 f0 = float32_abs(vfp_itos(a));
+    float32 f1 = float32_abs(vfp_itos(b));
+    return (float32_compare_quiet(f0, f1,NFS) >= 0) ? ~0 : 0;
+}
+
+uint32_t HELPER(neon_acgt_f32)(uint32_t a, uint32_t b)
+{
+    float32 f0 = float32_abs(vfp_itos(a));
+    float32 f1 = float32_abs(vfp_itos(b));
+    return (float32_compare_quiet(f0, f1, NFS) > 0) ? ~0 : 0;
+}
@@ -32,7 +32,5 @@
 #include "op_mem.h"
 #endif
-#include "op_neon.h"
-
 /* iwMMXt support */
 #include "op_iwmmxt.c"
@@ -20,6 +20,9 @@
 #include "exec.h"
 #include "helpers.h"
+#define SIGNBIT (uint32_t)0x80000000
+#define SIGNBIT64 ((uint64_t)1 << 63)
+
 void raise_exception(int tt)
 {
     env->exception_index = tt;
@@ -116,7 +119,8 @@ void tlb_fill (target_ulong addr, int is_write, int mmu_idx, void *retaddr)
 }
 #endif
-#define SIGNBIT (uint32_t)0x80000000
+/* FIXME: Pass an axplicit pointer to QF to CPUState, and move saturating
+   instructions into helper.c  */
 uint32_t HELPER(add_setq)(uint32_t a, uint32_t b)
 {
     uint32_t res = a + b;
@@ -451,3 +455,114 @@ uint32_t HELPER(ror_cc)(uint32_t x, uint32_t i)
     }
 }
+uint64_t HELPER(neon_add_saturate_s64)(uint64_t src1, uint64_t src2)
+{
+    uint64_t res;
+
+    res = src1 + src2;
+    if (((res ^ src1) & SIGNBIT64) && !((src1 ^ src2) & SIGNBIT64)) {
+        env->QF = 1;
+        res = ((int64_t)src1 >> 63) ^ ~SIGNBIT64;
+    }
+    return res;
+}
+
+uint64_t HELPER(neon_add_saturate_u64)(uint64_t src1, uint64_t src2)
+{
+    uint64_t res;
+
+    res = src1 + src2;
+    if (res < src1) {
+        env->QF = 1;
+        res = ~(uint64_t)0;
+    }
+    return res;
+}
+
+uint64_t HELPER(neon_sub_saturate_s64)(uint64_t src1, uint64_t src2)
+{
+    uint64_t res;
+
+    res = src1 - src2;
+    if (((res ^ src1) & SIGNBIT64) && ((src1 ^ src2) & SIGNBIT64)) {
+        env->QF = 1;
+        res = ((int64_t)src1 >> 63) ^ ~SIGNBIT64;
+    }
+    return res;
+}
+
+uint64_t HELPER(neon_sub_saturate_u64)(uint64_t src1, uint64_t src2)
+{
+    uint64_t res;
+
+    if (src1 < src2) {
+        env->QF = 1;
+        res = 0;
+    } else {
+        res = src1 - src2;
+    }
+    return res;
+}
+
+/* These need to return a pair of value, so still use T0/T1.  */
+/* Transpose.  Argument order is rather strange to avoid special casing
+   the tranlation code.
+   On input T0 = rm, T1 = rd.  On output T0 = rd, T1 = rm  */
+void HELPER(neon_trn_u8)(void)
+{
+    uint32_t rd;
+    uint32_t rm;
+    rd = ((T0 & 0x00ff00ff) << 8) | (T1 & 0x00ff00ff);
+    rm = ((T1 & 0xff00ff00) >> 8) | (T0 & 0xff00ff00);
+    T0 = rd;
+    T1 = rm;
+    FORCE_RET();
+}
+
+void HELPER(neon_trn_u16)(void)
+{
+    uint32_t rd;
+    uint32_t rm;
+    rd = (T0 << 16) | (T1 & 0xffff);
+    rm = (T1 >> 16) | (T0 & 0xffff0000);
+    T0 = rd;
+    T1 = rm;
+    FORCE_RET();
+}
+
+/* Worker routines for zip and unzip.  */
+void HELPER(neon_unzip_u8)(void)
+{
+    uint32_t rd;
+    uint32_t rm;
+    rd = (T0 & 0xff) | ((T0 >> 8) & 0xff00)
+         | ((T1 << 16) & 0xff0000) | ((T1 << 8) & 0xff000000);
+    rm = ((T0 >> 8) & 0xff) | ((T0 >> 16) & 0xff00)
+         | ((T1 << 8) & 0xff0000) | (T1 & 0xff000000);
+    T0 = rd;
+    T1 = rm;
+    FORCE_RET();
+}
+
+void HELPER(neon_zip_u8)(void)
+{
+    uint32_t rd;
+    uint32_t rm;
+    rd = (T0 & 0xff) | ((T1 << 8) & 0xff00)
+         | ((T0 << 16) & 0xff0000) | ((T1 << 24) & 0xff000000);
+    rm = ((T0 >> 16) & 0xff) | ((T1 >> 8) & 0xff00)
+         | ((T0 >> 8) & 0xff0000) | (T1 & 0xff000000);
+    T0 = rd;
+    T1 = rm;
+    FORCE_RET();
+}
+
+void HELPER(neon_zip_u16)(void)
+{
+    uint32_t tmp;
+
+    tmp = (T0 & 0xffff) | (T1 << 16);
+    T1 = (T1 & 0xffff0000) | (T0 >> 16);
+    T0 = tmp;
+    FORCE_RET();
+}
-/*
- * ARM NEON vector operations.
- *
- * Copyright (c) 2007 CodeSourcery.
- * Written by Paul Brook
- *
- * This code is licenced under the GPL.
- */
-/* Note that for NEON an "l" prefix means it is a wide operation, unlike
-   scalar arm ops where it means a word size operation.  */
-
-#define SIGNBIT (uint32_t)0x80000000
-/* ??? NEON ops should probably have their own float status.  */
-#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);
-}
-
-NEON_OP(getreg_T1)
-{
-    T1 = *(uint32_t *)((char *) env + PARAM1);
-}
-
-NEON_OP(setreg_T0)
-{
-    *(uint32_t *)((char *) env + PARAM1) = T0;
-}
-
-NEON_OP(setreg_T1)
-{
-    *(uint32_t *)((char *) env + PARAM1) = T1;
-}
-
-#define NEON_TYPE1(name, type) \
-typedef struct \
-{ \
-    type v1; \
-} neon_##name;
-#ifdef WORDS_BIGENDIAN
-#define NEON_TYPE2(name, type) \
-typedef struct \
-{ \
-    type v2; \
-    type v1; \
-} neon_##name;
-#define NEON_TYPE4(name, type) \
-typedef struct \
-{ \
-    type v4; \
-    type v3; \
-    type v2; \
-    type v1; \
-} neon_##name;
-#else
-#define NEON_TYPE2(name, type) \
-typedef struct \
-{ \
-    type v1; \
-    type v2; \
-} neon_##name;
-#define NEON_TYPE4(name, type) \
-typedef struct \
-{ \
-    type v1; \
-    type v2; \
-    type v3; \
-    type v4; \
-} neon_##name;
-#endif
-
-NEON_TYPE4(s8, int8_t)
-NEON_TYPE4(u8, uint8_t)
-NEON_TYPE2(s16, int16_t)
-NEON_TYPE2(u16, uint16_t)
-NEON_TYPE1(s32, int32_t)
-NEON_TYPE1(u32, uint32_t)
-#undef NEON_TYPE4
-#undef NEON_TYPE2
-#undef NEON_TYPE1
-
-/* Copy from a uint32_t to a vector structure type.  */
-#define NEON_UNPACK(vtype, dest, val) do { \
-    union { \
-        vtype v; \
-        uint32_t i; \
-    } conv_u; \
-    conv_u.i = (val); \
-    dest = conv_u.v; \
-    } while(0)
-
-/* Copy from a vector structure type to a uint32_t.  */
-#define NEON_PACK(vtype, dest, val) do { \
-    union { \
-        vtype v; \
-        uint32_t i; \
-    } conv_u; \
-    conv_u.v = (val); \
-    dest = conv_u.i; \
-    } while(0)
-
-#define NEON_DO1 \
-    NEON_FN(vdest.v1, vsrc1.v1, vsrc2.v1);
-#define NEON_DO2 \
-    NEON_FN(vdest.v1, vsrc1.v1, vsrc2.v1); \
-    NEON_FN(vdest.v2, vsrc1.v2, vsrc2.v2);
-#define NEON_DO4 \
-    NEON_FN(vdest.v1, vsrc1.v1, vsrc2.v1); \
-    NEON_FN(vdest.v2, vsrc1.v2, vsrc2.v2); \
-    NEON_FN(vdest.v3, vsrc1.v3, vsrc2.v3); \
-    NEON_FN(vdest.v4, vsrc1.v4, vsrc2.v4);
-
-#define NEON_VOP(name, vtype, n) \
-NEON_OP(name) \
-{ \
-    vtype vsrc1; \
-    vtype vsrc2; \
-    vtype vdest; \
-    NEON_UNPACK(vtype, vsrc1, T0); \
-    NEON_UNPACK(vtype, vsrc2, T1); \
-    NEON_DO##n; \
-    NEON_PACK(vtype, T0, vdest); \
-    FORCE_RET(); \
-}
-
-#define NEON_VOP1(name, vtype, n) \
-NEON_OP(name) \
-{ \
-    vtype vsrc1; \
-    vtype vdest; \
-    NEON_UNPACK(vtype, vsrc1, T0); \
-    NEON_DO##n; \
-    NEON_PACK(vtype, T0, vdest); \
-    FORCE_RET(); \
-}
-
-/* Pairwise operations.  */
-/* For 32-bit elements each segment only contains a single element, so
-   the elementwise and pairwise operations are the same.  */
-#define NEON_PDO2 \
-    NEON_FN(vdest.v1, vsrc1.v1, vsrc1.v2); \
-    NEON_FN(vdest.v2, vsrc2.v1, vsrc2.v2);
-#define NEON_PDO4 \
-    NEON_FN(vdest.v1, vsrc1.v1, vsrc1.v2); \
-    NEON_FN(vdest.v2, vsrc1.v3, vsrc1.v4); \
-    NEON_FN(vdest.v3, vsrc2.v1, vsrc2.v2); \
-    NEON_FN(vdest.v4, vsrc2.v3, vsrc2.v4); \
-
-#define NEON_POP(name, vtype, n) \
-NEON_OP(name) \
-{ \
-    vtype vsrc1; \
-    vtype vsrc2; \
-    vtype vdest; \
-    NEON_UNPACK(vtype, vsrc1, T0); \
-    NEON_UNPACK(vtype, vsrc2, T1); \
-    NEON_PDO##n; \
-    NEON_PACK(vtype, T0, vdest); \
-    FORCE_RET(); \
-}
-
-#define NEON_FN(dest, src1, src2) dest = (src1 + src2) >> 1
-NEON_VOP(hadd_s8, neon_s8, 4)
-NEON_VOP(hadd_u8, neon_u8, 4)
-NEON_VOP(hadd_s16, neon_s16, 2)
-NEON_VOP(hadd_u16, neon_u16, 2)
-#undef NEON_FN
-
-NEON_OP(hadd_s32)
-{
-    int32_t src1 = T0;
-    int32_t src2 = T1;
-    int32_t dest;
-
-    dest = (src1 >> 1) + (src2 >> 1);
-    if (src1 & src2 & 1)
-        dest++;
-    T0 = dest;
-    FORCE_RET();
-}
-
-NEON_OP(hadd_u32)
-{
-    uint32_t src1 = T0;
-    uint32_t src2 = T1;
-    uint32_t dest;
-
-    dest = (src1 >> 1) + (src2 >> 1);
-    if (src1 & src2 & 1)
-        dest++;
-    T0 = dest;
-    FORCE_RET();
-}
-
-#define NEON_FN(dest, src1, src2) dest = (src1 + src2 + 1) >> 1
-NEON_VOP(rhadd_s8, neon_s8, 4)
-NEON_VOP(rhadd_u8, neon_u8, 4)
-NEON_VOP(rhadd_s16, neon_s16, 2)
-NEON_VOP(rhadd_u16, neon_u16, 2)
-#undef NEON_FN
-
-NEON_OP(rhadd_s32)
-{
-    int32_t src1 = T0;
-    int32_t src2 = T1;
-    int32_t dest;
-
-    dest = (src1 >> 1) + (src2 >> 1);
-    if ((src1 | src2) & 1)
-        dest++;
-    T0 = dest;
-    FORCE_RET();
-}
-
-NEON_OP(rhadd_u32)
-{
-    uint32_t src1 = T0;
-    uint32_t src2 = T1;
-    uint32_t dest;
-
-    dest = (src1 >> 1) + (src2 >> 1);
-    if ((src1 | src2) & 1)
-        dest++;
-    T0 = dest;
-    FORCE_RET();
-}
-
-#define NEON_FN(dest, src1, src2) dest = (src1 - src2) >> 1
-NEON_VOP(hsub_s8, neon_s8, 4)
-NEON_VOP(hsub_u8, neon_u8, 4)
-NEON_VOP(hsub_s16, neon_s16, 2)
-NEON_VOP(hsub_u16, neon_u16, 2)
-#undef NEON_FN
-
-NEON_OP(hsub_s32)
-{
-    int32_t src1 = T0;
-    int32_t src2 = T1;
-    int32_t dest;
-
-    dest = (src1 >> 1) - (src2 >> 1);
-    if ((~src1) & src2 & 1)
-        dest--;
-    T0 = dest;
-    FORCE_RET();
-}
-
-NEON_OP(hsub_u32)
-{
-    uint32_t src1 = T0;
-    uint32_t src2 = T1;
-    uint32_t dest;
-
-    dest = (src1 >> 1) - (src2 >> 1);
-    if ((~src1) & src2 & 1)
-        dest--;
-    T0 = dest;
-    FORCE_RET();
-}
-
-#define NEON_USAT(dest, src1, src2, type) do { \
-    uint32_t tmp = (uint32_t)src1 + (uint32_t)src2; \
-    if (tmp != (type)tmp) { \
-        env->QF = 1; \
-        dest = ~0; \
-    } else { \
-        dest = tmp; \
-    }} while(0)
-#define NEON_FN(dest, src1, src2) NEON_USAT(dest, src1, src2, uint8_t)
-NEON_VOP(qadd_u8, neon_u8, 4)
-#undef NEON_FN
-#define NEON_FN(dest, src1, src2) NEON_USAT(dest, src1, src2, uint16_t)
-NEON_VOP(qadd_u16, neon_u16, 2)
-#undef NEON_FN
-#undef NEON_USAT
-
-#define NEON_SSAT(dest, src1, src2, type) do { \
-    int32_t tmp = (uint32_t)src1 + (uint32_t)src2; \
-    if (tmp != (type)tmp) { \
-        env->QF = 1; \
-        if (src2 > 0) { \
-            tmp = (1 << (sizeof(type) * 8 - 1)) - 1; \
-        } else { \
-            tmp = 1 << (sizeof(type) * 8 - 1); \
-        } \
-    } \
-    dest = tmp; \
-    } while(0)
-#define NEON_FN(dest, src1, src2) NEON_SSAT(dest, src1, src2, int8_t)
-NEON_VOP(qadd_s8, neon_s8, 4)
-#undef NEON_FN
-#define NEON_FN(dest, src1, src2) NEON_SSAT(dest, src1, src2, int16_t)
-NEON_VOP(qadd_s16, neon_s16, 2)
-#undef NEON_FN
-#undef NEON_SSAT
-
-#define NEON_USAT(dest, src1, src2, type) do { \
-    uint32_t tmp = (uint32_t)src1 - (uint32_t)src2; \
-    if (tmp != (type)tmp) { \
-        env->QF = 1; \
-        dest = 0; \
-    } else { \
-        dest = tmp; \
-    }} while(0)
-#define NEON_FN(dest, src1, src2) NEON_USAT(dest, src1, src2, uint8_t)
-NEON_VOP(qsub_u8, neon_u8, 4)
-#undef NEON_FN
-#define NEON_FN(dest, src1, src2) NEON_USAT(dest, src1, src2, uint16_t)
-NEON_VOP(qsub_u16, neon_u16, 2)
-#undef NEON_FN
-#undef NEON_USAT
-
-#define NEON_SSAT(dest, src1, src2, type) do { \
-    int32_t tmp = (uint32_t)src1 - (uint32_t)src2; \
-    if (tmp != (type)tmp) { \
-        env->QF = 1; \
-        if (src2 < 0) { \
-            tmp = (1 << (sizeof(type) * 8 - 1)) - 1; \
-        } else { \
-            tmp = 1 << (sizeof(type) * 8 - 1); \
-        } \
-    } \
-    dest = tmp; \
-    } while(0)
-#define NEON_FN(dest, src1, src2) NEON_SSAT(dest, src1, src2, int8_t)
-NEON_VOP(qsub_s8, neon_s8, 4)
-#undef NEON_FN
-#define NEON_FN(dest, src1, src2) NEON_SSAT(dest, src1, src2, int16_t)
-NEON_VOP(qsub_s16, neon_s16, 2)
-#undef NEON_FN
-#undef NEON_SSAT
-
-#define NEON_FN(dest, src1, src2) dest = (src1 > src2) ? ~0 : 0
-NEON_VOP(cgt_s8, neon_s8, 4)
-NEON_VOP(cgt_u8, neon_u8, 4)
-NEON_VOP(cgt_s16, neon_s16, 2)
-NEON_VOP(cgt_u16, neon_u16, 2)
-NEON_VOP(cgt_s32, neon_s32, 1)
-NEON_VOP(cgt_u32, neon_u32, 1)
-#undef NEON_FN
-
-#define NEON_FN(dest, src1, src2) dest = (src1 >= src2) ? ~0 : 0
-NEON_VOP(cge_s8, neon_s8, 4)
-NEON_VOP(cge_u8, neon_u8, 4)
-NEON_VOP(cge_s16, neon_s16, 2)
-NEON_VOP(cge_u16, neon_u16, 2)
-NEON_VOP(cge_s32, neon_s32, 1)
-NEON_VOP(cge_u32, neon_u32, 1)
-#undef NEON_FN
-
-#define NEON_FN(dest, src1, src2) do { \
-    int8_t tmp; \
-    tmp = (int8_t)src2; \
-    if (tmp < 0) { \
-        dest = src1 >> -tmp; \
-    } else { \
-        dest = src1 << tmp; \
-    }} while (0)
-NEON_VOP(shl_s8, neon_s8, 4)
-NEON_VOP(shl_u8, neon_u8, 4)
-NEON_VOP(shl_s16, neon_s16, 2)
-NEON_VOP(shl_u16, neon_u16, 2)
-NEON_VOP(shl_s32, neon_s32, 1)
-NEON_VOP(shl_u32, neon_u32, 1)
-#undef NEON_FN
-
-NEON_OP(shl_u64)
-{
-    int8_t shift = env->vfp.scratch[0];
-    uint64_t val = T0 | ((uint64_t)T1 << 32);
-    if (shift < 0) {
-        val >>= -shift;
-    } else {
-        val <<= shift;
-    }
-    T0 = val;
-    T1 = val >> 32;
-    FORCE_RET();
-}
-
-NEON_OP(shl_s64)
-{
-    int8_t shift = env->vfp.scratch[0];
-    int64_t val = T0 | ((uint64_t)T1 << 32);
-    if (shift < 0) {
-        val >>= -shift;
-    } else {
-        val <<= shift;
-    }
-    T0 = val;
-    T1 = val >> 32;
-    FORCE_RET();
-}
-
-#define NEON_FN(dest, src1, src2) do { \
-    int8_t tmp; \
-    tmp = (int8_t)src1; \
-    if (tmp < 0) { \
-        dest = (src2 + (1 << (-1 - tmp))) >> -tmp; \
-    } else { \
-        dest = src2 << tmp; \
-    }} while (0)
-
-NEON_VOP(rshl_s8, neon_s8, 4)
-NEON_VOP(rshl_u8, neon_u8, 4)
-NEON_VOP(rshl_s16, neon_s16, 2)
-NEON_VOP(rshl_u16, neon_u16, 2)
-NEON_VOP(rshl_s32, neon_s32, 1)
-NEON_VOP(rshl_u32, neon_u32, 1)
-#undef NEON_FN
-
-NEON_OP(rshl_u64)
-{
-    int8_t shift = env->vfp.scratch[0];
-    uint64_t val = T0 | ((uint64_t)T1 << 32);
-    if (shift < 0) {
-        val = (val + ((uint64_t)1 << (-1 - shift))) >> -shift;
-        val >>= -shift;
-    } else {
-        val <<= shift;
-    }
-    T0 = val;
-    T1 = val >> 32;
-    FORCE_RET();
-}
-
-NEON_OP(rshl_s64)
-{
-    int8_t shift = env->vfp.scratch[0];
-    int64_t val = T0 | ((uint64_t)T1 << 32);
-    if (shift < 0) {
-        val = (val + ((int64_t)1 << (-1 - shift))) >> -shift;
-    } else {
-        val <<= shift;
-    }
-    T0 = val;
-    T1 = val >> 32;
-    FORCE_RET();
-}
-
-#define NEON_FN(dest, src1, src2) do { \
-    int8_t tmp; \
-    tmp = (int8_t)src1; \
-    if (tmp < 0) { \
-        dest = src2 >> -tmp; \
-    } else { \
-        dest = src2 << tmp; \
-        if ((dest >> tmp) != src2) { \
-            env->QF = 1; \
-            dest = ~0; \
-        } \
-    }} while (0)
-NEON_VOP(qshl_s8, neon_s8, 4)
-NEON_VOP(qshl_s16, neon_s16, 2)
-NEON_VOP(qshl_s32, neon_s32, 1)
-#undef NEON_FN
-
-NEON_OP(qshl_s64)
-{
-    int8_t shift = env->vfp.scratch[0];
-    int64_t val = T0 | ((uint64_t)T1 << 32);
-    if (shift < 0) {
-        val >>= -shift;
-    } else {
-        int64_t tmp = val;
-        val <<= shift;
-        if ((val >> shift) != tmp) {
-            env->QF = 1;
-            val = (tmp >> 63) ^ 0x7fffffffffffffffULL;
-        }
-    }
-    T0 = val;
-    T1 = val >> 32;
-    FORCE_RET();
-}
-
-#define NEON_FN(dest, src1, src2) do { \
-    int8_t tmp; \
-    tmp = (int8_t)src1; \
-    if (tmp < 0) { \
-        dest = src2 >> -tmp; \
-    } else { \
-        dest = src2 << tmp; \
-        if ((dest >> tmp) != src2) { \
-            env->QF = 1; \
-            dest = src2 >> 31; \
-        } \
-    }} while (0)
-NEON_VOP(qshl_u8, neon_u8, 4)
-NEON_VOP(qshl_u16, neon_u16, 2)
-NEON_VOP(qshl_u32, neon_u32, 1)
-#undef NEON_FN
-
-NEON_OP(qshl_u64)
-{
-    int8_t shift = env->vfp.scratch[0];
-    uint64_t val = T0 | ((uint64_t)T1 << 32);
-    if (shift < 0) {
-        val >>= -shift;
-    } else {
-        uint64_t tmp = val;
-        val <<= shift;
-        if ((val >> shift) != tmp) {
-            env->QF = 1;
-            val = ~(uint64_t)0;
-        }
-    }
-    T0 = val;
-    T1 = val >> 32;
-    FORCE_RET();
-}
-
-#define NEON_FN(dest, src1, src2) do { \
-    int8_t tmp; \
-    tmp = (int8_t)src1; \
-    if (tmp < 0) { \
-        dest = (src2 + (1 << (-1 - tmp))) >> -tmp; \
-    } else { \
-        dest = src2 << tmp; \
-        if ((dest >> tmp) != src2) { \
-            dest = ~0; \
-        } \
-    }} while (0)
-NEON_VOP(qrshl_s8, neon_s8, 4)
-NEON_VOP(qrshl_s16, neon_s16, 2)
-NEON_VOP(qrshl_s32, neon_s32, 1)
-#undef NEON_FN
-
-#define NEON_FN(dest, src1, src2) do { \
-    int8_t tmp; \
-    tmp = (int8_t)src1; \
-    if (tmp < 0) { \
-        dest = (src2 + (1 << (-1 - tmp))) >> -tmp; \
-    } else { \
-        dest = src2 << tmp; \
-        if ((dest >> tmp) != src2) { \
-            env->QF = 1; \
-            dest = src2 >> 31; \
-        } \
-    }} while (0)
-NEON_VOP(qrshl_u8, neon_u8, 4)
-NEON_VOP(qrshl_u16, neon_u16, 2)
-NEON_VOP(qrshl_u32, neon_u32, 1)
-#undef NEON_FN
-
-#define NEON_FN(dest, src1, src2) dest = (src1 > src2) ? src1 : src2
-NEON_VOP(max_s8, neon_s8, 4)
-NEON_VOP(max_u8, neon_u8, 4)
-NEON_VOP(max_s16, neon_s16, 2)
-NEON_VOP(max_u16, neon_u16, 2)
-NEON_VOP(max_s32, neon_s32, 1)
-NEON_VOP(max_u32, neon_u32, 1)
-NEON_POP(pmax_s8, neon_s8, 4)
-NEON_POP(pmax_u8, neon_u8, 4)
-NEON_POP(pmax_s16, neon_s16, 2)
-NEON_POP(pmax_u16, neon_u16, 2)
-#undef NEON_FN
-
-NEON_OP(max_f32)
-{
-    float32 f0 = vfp_itos(T0);
-    float32 f1 = vfp_itos(T1);
-    T0 = (float32_compare_quiet(f0, f1, NFS) == 1) ? T0 : T1;
-    FORCE_RET();
-}
-
-#define NEON_FN(dest, src1, src2) dest = (src1 < src2) ? src1 : src2
-NEON_VOP(min_s8, neon_s8, 4)
-NEON_VOP(min_u8, neon_u8, 4)
-NEON_VOP(min_s16, neon_s16, 2)
-NEON_VOP(min_u16, neon_u16, 2)
-NEON_VOP(min_s32, neon_s32, 1)
-NEON_VOP(min_u32, neon_u32, 1)
-NEON_POP(pmin_s8, neon_s8, 4)
-NEON_POP(pmin_u8, neon_u8, 4)
-NEON_POP(pmin_s16, neon_s16, 2)
-NEON_POP(pmin_u16, neon_u16, 2)
-#undef NEON_FN
-
-NEON_OP(min_f32)
-{
-    float32 f0 = vfp_itos(T0);
-    float32 f1 = vfp_itos(T1);
-    T0 = (float32_compare_quiet(f0, f1, NFS) == -1) ? T0 : T1;
-    FORCE_RET();
-}
-
-#define NEON_FN(dest, src1, src2) \
-    dest = (src1 > src2) ? (src1 - src2) : (src2 - src1)
-NEON_VOP(abd_s8, neon_s8, 4)
-NEON_VOP(abd_u8, neon_u8, 4)
-NEON_VOP(abd_s16, neon_s16, 2)
-NEON_VOP(abd_u16, neon_u16, 2)
-NEON_VOP(abd_s32, neon_s32, 1)
-NEON_VOP(abd_u32, neon_u32, 1)
-#undef NEON_FN
-
-NEON_OP(abd_f32)
-{
-    float32 f0 = vfp_itos(T0);
-    float32 f1 = vfp_itos(T1);
-    T0 = vfp_stoi((float32_compare_quiet(f0, f1, NFS) == 1)
-                  ? float32_sub(f0, f1, NFS)
-                  : float32_sub(f1, f0, NFS));
-    FORCE_RET();
-}
-
-#define NEON_FN(dest, src1, src2) dest = src1 + src2
-NEON_VOP(add_u8, neon_u8, 4)
-NEON_VOP(add_u16, neon_u16, 2)
-NEON_POP(padd_u8, neon_u8, 4)
-NEON_POP(padd_u16, neon_u16, 2)
-#undef NEON_FN
-
-NEON_OP(add_f32)
-{
-    T0 = vfp_stoi(float32_add(vfp_itos(T0), vfp_itos(T1), NFS));
-    FORCE_RET();
-}
-
-#define NEON_FN(dest, src1, src2) dest = src1 - src2
-NEON_VOP(sub_u8, neon_u8, 4)
-NEON_VOP(sub_u16, neon_u16, 2)
-#undef NEON_FN
-
-NEON_OP(sub_f32)
-{
-    T0 = vfp_stoi(float32_sub(vfp_itos(T0), vfp_itos(T1), NFS));
-    FORCE_RET();
-}
-
-#define NEON_FN(dest, src1, src2) dest = src2 - src1
-NEON_VOP(rsb_u8, neon_u8, 4)
-NEON_VOP(rsb_u16, neon_u16, 2)
-#undef NEON_FN
-
-NEON_OP(rsb_f32)
-{
-    T0 = vfp_stoi(float32_sub(vfp_itos(T1), vfp_itos(T0), NFS));
-    FORCE_RET();
-}
-
-#define NEON_FN(dest, src1, src2) dest = src1 * src2
-NEON_VOP(mul_u8, neon_u8, 4)
-NEON_VOP(mul_u16, neon_u16, 2)
-#undef NEON_FN
-
-NEON_OP(mul_f32)
-{
-    T0 = vfp_stoi(float32_mul(vfp_itos(T0), vfp_itos(T1), NFS));
-    FORCE_RET();
-}
-
-NEON_OP(mul_p8)
-{
-    T0 = helper_neon_mul_p8(T0, T1);
-}
-
-#define NEON_FN(dest, src1, src2) dest = (src1 & src2) ? -1 : 0
-NEON_VOP(tst_u8, neon_u8, 4)
-NEON_VOP(tst_u16, neon_u16, 2)
-NEON_VOP(tst_u32, neon_u32, 1)
-#undef NEON_FN
-
-#define NEON_FN(dest, src1, src2) dest = (src1 == src2) ? -1 : 0
-NEON_VOP(ceq_u8, neon_u8, 4)
-NEON_VOP(ceq_u16, neon_u16, 2)
-NEON_VOP(ceq_u32, neon_u32, 1)
-#undef NEON_FN
-
-#define NEON_QDMULH16(dest, src1, src2, round) do { \
-    uint32_t tmp = (int32_t)(int16_t) src1 * (int16_t) src2; \
-    if ((tmp ^ (tmp << 1)) & SIGNBIT) { \
-        env->QF = 1; \
-        tmp = (tmp >> 31) ^ ~SIGNBIT; \
-    } \
-    tmp <<= 1; \
-    if (round) { \
-        int32_t old = tmp; \
-        tmp += 1 << 15; \
-        if ((int32_t)tmp < old) { \
-            env->QF = 1; \
-            tmp = SIGNBIT - 1; \
-        } \
-    } \
-    dest = tmp >> 16; \
-    } while(0)
-#define NEON_FN(dest, src1, src2) NEON_QDMULH16(dest, src1, src2, 0)
-NEON_VOP(qdmulh_s16, neon_s16, 2)
-#undef NEON_FN
-#define NEON_FN(dest, src1, src2) NEON_QDMULH16(dest, src1, src2, 1)
-NEON_VOP(qrdmulh_s16, neon_s16, 2)
-#undef NEON_FN
-#undef NEON_QDMULH16
-
-#define SIGNBIT64 ((uint64_t)1 << 63)
-#define NEON_QDMULH32(dest, src1, src2, round) do { \
-    uint64_t tmp = (int64_t)(int32_t) src1 * (int32_t) src2; \
-    if ((tmp ^ (tmp << 1)) & SIGNBIT64) { \
-        env->QF = 1; \
-        tmp = (tmp >> 63) ^ ~SIGNBIT64; \
-    } else { \
-        tmp <<= 1; \
-    } \
-    if (round) { \
-        int64_t old = tmp; \
-        tmp += (int64_t)1 << 31; \
-        if ((int64_t)tmp < old) { \
-            env->QF = 1; \
-            tmp = SIGNBIT64 - 1; \
-        } \
-    } \
-    dest = tmp >> 32; \
-    } while(0)
-#define NEON_FN(dest, src1, src2) NEON_QDMULH32(dest, src1, src2, 0)
-NEON_VOP(qdmulh_s32, neon_s32, 1)
-#undef NEON_FN
-#define NEON_FN(dest, src1, src2) NEON_QDMULH32(dest, src1, src2, 1)
-NEON_VOP(qrdmulh_s32, neon_s32, 1)
-#undef NEON_FN
-#undef NEON_QDMULH32
-
-/* Floating point comparisons produce an integer result.  */
-#define NEON_VOP_FCMP(name, cmp) \
-NEON_OP(name) \
-{ \
-    if (float32_compare_quiet(vfp_itos(T0), vfp_itos(T1), NFS) cmp 0) \
-        T0 = -1; \
-    else \
-        T0 = 0; \
-    FORCE_RET(); \
-}
-
-NEON_VOP_FCMP(ceq_f32, ==)
-NEON_VOP_FCMP(cge_f32, >=)
-NEON_VOP_FCMP(cgt_f32, >)
-
-NEON_OP(acge_f32)
-{
-    float32 f0 = float32_abs(vfp_itos(T0));
-    float32 f1 = float32_abs(vfp_itos(T1));
-    T0 = (float32_compare_quiet(f0, f1,NFS) >= 0) ? -1 : 0;
-    FORCE_RET();
-}
-
-NEON_OP(acgt_f32)
-{
-    float32 f0 = float32_abs(vfp_itos(T0));
-    float32 f1 = float32_abs(vfp_itos(T1));
-    T0 = (float32_compare_quiet(f0, f1, NFS) > 0) ? -1 : 0;
-    FORCE_RET();
-}
-
-/* Narrowing instructions.  The named type is the destination type.  */
-NEON_OP(narrow_u8)
-{
-    T0 = (T0 & 0xff) | ((T0 >> 8) & 0xff00)
-         | ((T1 << 16) & 0xff0000) | (T1 << 24);
-    FORCE_RET();
-}
-
-NEON_OP(narrow_sat_u8)
-{
-    neon_u16 src;
-    neon_u8 dest;
-#define SAT8(d, s) \
-    if (s > 0xff) { \
-        d = 0xff; \
-        env->QF = 1; \
-    } else  { \
-        d = s; \
-    }
-
-    NEON_UNPACK(neon_u16, src, T0);
-    SAT8(dest.v1, src.v1);
-    SAT8(dest.v2, src.v2);
-    NEON_UNPACK(neon_u16, src, T1);
-    SAT8(dest.v3, src.v1);
-    SAT8(dest.v4, src.v2);
-    NEON_PACK(neon_u8, T0, dest);
-    FORCE_RET();
-#undef SAT8
-}
-
-NEON_OP(narrow_sat_s8)
-{
-    neon_s16 src;
-    neon_s8 dest;
-#define SAT8(d, s) \
-    if (s != (uint8_t)s) { \
-        d = (s >> 15) ^ 0x7f; \
-        env->QF = 1; \
-    } else  { \
-        d = s; \
-    }
-
-    NEON_UNPACK(neon_s16, src, T0);
-    SAT8(dest.v1, src.v1);
-    SAT8(dest.v2, src.v2);
-    NEON_UNPACK(neon_s16, src, T1);
-    SAT8(dest.v3, src.v1);
-    SAT8(dest.v4, src.v2);
-    NEON_PACK(neon_s8, T0, dest);
-    FORCE_RET();
-#undef SAT8
-}
-
-NEON_OP(narrow_u16)
-{
-    T0 = (T0 & 0xffff) | (T1 << 16);
-}
-
-NEON_OP(narrow_sat_u16)
-{
-    if (T0 > 0xffff) {
-        T0 = 0xffff;
-        env->QF = 1;
-    }
-    if (T1 > 0xffff) {
-        T1 = 0xffff;
-        env->QF = 1;
-    }
-    T0 |= T1 << 16;
-    FORCE_RET();
-}
-
-NEON_OP(narrow_sat_s16)
-{
-    if ((int32_t)T0 != (int16_t)T0) {
-        T0 = ((int32_t)T0 >> 31) ^ 0x7fff;
-        env->QF = 1;
-    }
-    if ((int32_t)T1 != (int16_t) T1) {
-        T1 = ((int32_t)T1 >> 31) ^ 0x7fff;
-        env->QF = 1;
-    }
-    T0 = (uint16_t)T0 | (T1 << 16);
-    FORCE_RET();
-}
-
-NEON_OP(narrow_sat_u32)
-{
-    if (T1) {
-        T0 = 0xffffffffu;
-        env->QF = 1;
-    }
-    FORCE_RET();
-}
-
-NEON_OP(narrow_sat_s32)
-{
-    int32_t sign = (int32_t)T1 >> 31;
-
-    if ((int32_t)T1 != sign) {
-        T0 = sign ^ 0x7fffffff;
-        env->QF = 1;
-    }
-    FORCE_RET();
-}
-
-/* Narrowing instructions.  Named type is the narrow type.  */
-NEON_OP(narrow_high_u8)
-{
-    T0 = ((T0 >> 8) & 0xff) | ((T0 >> 16) & 0xff00)
-        | ((T1 << 8) & 0xff0000) | (T1 & 0xff000000);
-    FORCE_RET();
-}
-
-NEON_OP(narrow_high_u16)
-{
-    T0 = (T0 >> 16) | (T1 & 0xffff0000);
-    FORCE_RET();
-}
-
-NEON_OP(narrow_high_round_u8)
-{
-    T0 = (((T0 + 0x80) >> 8) & 0xff) | (((T0 + 0x800000) >> 16) & 0xff00)
-        | (((T1 + 0x80) << 8) & 0xff0000) | ((T1 + 0x800000) & 0xff000000);
-    FORCE_RET();
-}
-
-NEON_OP(narrow_high_round_u16)
-{
-    T0 = ((T0 + 0x8000) >> 16) | ((T1 + 0x8000) & 0xffff0000);
-    FORCE_RET();
-}
-
-NEON_OP(narrow_high_round_u32)
-{
-    if (T0 >= 0x80000000u)
-        T0 = T1 + 1;
-    else
-        T0 = T1;
-    FORCE_RET();
-}
-
-/* Widening instructions.  Named type is source type.  */
-NEON_OP(widen_s8)
-{
-    uint32_t src;
-
-    src = T0;
-    T0 = (uint16_t)(int8_t)src | ((int8_t)(src >> 8) << 16);
-    T1 = (uint16_t)(int8_t)(src >> 16) | ((int8_t)(src >> 24) << 16);
-}
-
-NEON_OP(widen_u8)
-{
-    T1 = ((T0 >> 8) & 0xff0000) | ((T0 >> 16) & 0xff);
-    T0 = ((T0 << 8) & 0xff0000) | (T0 & 0xff);
-}
-
-NEON_OP(widen_s16)
-{
-    int32_t src;
-
-    src = T0;
-    T0 = (int16_t)src;
-    T1 = src >> 16;
-}
-
-NEON_OP(widen_u16)
-{
-    T1 = T0 >> 16;
-    T0 &= 0xffff;
-}
-
-NEON_OP(widen_s32)
-{
-    T1 = (int32_t)T0 >> 31;
-    FORCE_RET();
-}
-
-NEON_OP(widen_high_u8)
-{
-    T1 = (T0 & 0xff000000) | ((T0 >> 8) & 0xff00);
-    T0 = ((T0 << 16) & 0xff000000) | ((T0 << 8) & 0xff00);
-}
-
-NEON_OP(widen_high_u16)
-{
-    T1 = T0 & 0xffff0000;
-    T0 <<= 16;
-}
-
-/* Long operations.  The type is the wide type.  */
-NEON_OP(shll_u16)
-{
-    int shift = PARAM1;
-    uint32_t mask;
-
-    mask = 0xffff >> (16 - shift);
-    mask |= mask << 16;
-    mask = ~mask;
-
-    T0 = (T0 << shift) & mask;
-    T1 = (T1 << shift) & mask;
-    FORCE_RET();
-}
-
-NEON_OP(shll_u64)
-{
-    int shift = PARAM1;
-
-    T1 <<= shift;
-    T1 |= T0 >> (32 - shift);
-    T0 <<= shift;
-    FORCE_RET();
-}
-
-NEON_OP(addl_u16)
-{
-    uint32_t tmp;
-    uint32_t high;
-
-    tmp = env->vfp.scratch[0];
-    high = (T0 >> 16) + (tmp >> 16);
-    T0 = (uint16_t)(T0 + tmp);
-    T0 |= (high << 16);
-    tmp = env->vfp.scratch[1];
-    high = (T1 >> 16) + (tmp >> 16);
-    T1 = (uint16_t)(T1 + tmp);
-    T1 |= (high << 16);
-    FORCE_RET();
-}
-
-NEON_OP(addl_u32)
-{
-    T0 += env->vfp.scratch[0];
-    T1 += env->vfp.scratch[1];
-    FORCE_RET();
-}
-
-NEON_OP(addl_u64)
-{
-    uint64_t tmp;
-    tmp = T0 | ((uint64_t)T1 << 32);
-    tmp += env->vfp.scratch[0];
-    tmp += (uint64_t)env->vfp.scratch[1] << 32;
-    T0 = tmp;
-    T1 = tmp >> 32;
-    FORCE_RET();
-}
-
-NEON_OP(subl_u16)
-{
-    uint32_t tmp;
-    uint32_t high;
-
-    tmp = env->vfp.scratch[0];
-    high = (T0 >> 16) - (tmp >> 16);
-    T0 = (uint16_t)(T0 - tmp);
-    T0 |= (high << 16);
-    tmp = env->vfp.scratch[1];
-    high = (T1 >> 16) - (tmp >> 16);
-    T1 = (uint16_t)(T1 - tmp);
-    T1 |= (high << 16);
-    FORCE_RET();
-}
-
-NEON_OP(subl_u32)
-{
-    T0 -= env->vfp.scratch[0];
-    T1 -= env->vfp.scratch[1];
-    FORCE_RET();
-}
-
-NEON_OP(subl_u64)
-{
-    uint64_t tmp;
-    tmp = T0 | ((uint64_t)T1 << 32);
-    tmp -= env->vfp.scratch[0];
-    tmp -= (uint64_t)env->vfp.scratch[1] << 32;
-    T0 = tmp;
-    T1 = tmp >> 32;
-    FORCE_RET();
-}
-
-#define DO_ABD(dest, x, y, type) do { \
-    type tmp_x = x; \
-    type tmp_y = y; \
-    dest = ((tmp_x > tmp_y) ? tmp_x - tmp_y : tmp_y - tmp_x); \
-    } while(0)
-
-NEON_OP(abdl_u16)
-{
-    uint32_t tmp;
-    uint32_t low;
-    uint32_t high;
-
-    DO_ABD(low, T0, T1, uint8_t);
-    DO_ABD(tmp, T0 >> 8, T1 >> 8, uint8_t);
-    low |= tmp << 16;
-    DO_ABD(high, T0 >> 16, T1 >> 16, uint8_t);
-    DO_ABD(tmp, T0 >> 24, T1 >> 24, uint8_t);
-    high |= tmp << 16;
-    T0 = low;
-    T1 = high;
-    FORCE_RET();
-}
-
-NEON_OP(abdl_s16)
-{
-    uint32_t tmp;
-    uint32_t low;
-    uint32_t high;
-
-    DO_ABD(low, T0, T1, int8_t);
-    DO_ABD(tmp, T0 >> 8, T1 >> 8, int8_t);
-    low |= tmp << 16;
-    DO_ABD(high, T0 >> 16, T1 >> 16, int8_t);
-    DO_ABD(tmp, T0 >> 24, T1 >> 24, int8_t);
-    high |= tmp << 16;
-    T0 = low;
-    T1 = high;
-    FORCE_RET();
-}
-
-NEON_OP(abdl_u32)
-{
-    uint32_t low;
-    uint32_t high;
-
-    DO_ABD(low, T0, T1, uint16_t);
-    DO_ABD(high, T0 >> 16, T1 >> 16, uint16_t);
-    T0 = low;
-    T1 = high;
-    FORCE_RET();
-}
-
-NEON_OP(abdl_s32)
-{
-    uint32_t low;
-    uint32_t high;
-
-    DO_ABD(low, T0, T1, int16_t);
-    DO_ABD(high, T0 >> 16, T1 >> 16, int16_t);
-    T0 = low;
-    T1 = high;
-    FORCE_RET();
-}
-
-NEON_OP(abdl_u64)
-{
-    DO_ABD(T0, T0, T1, uint32_t);
-    T1 = 0;
-}
-
-NEON_OP(abdl_s64)
-{
-    DO_ABD(T0, T0, T1, int32_t);
-    T1 = 0;
-}
-#undef DO_ABD
-
-/* Widening multiple. Named type is the source type.  */
-#define DO_MULL(dest, x, y, type1, type2) do { \
-    type1 tmp_x = x; \
-    type1 tmp_y = y; \
-    dest = (type2)((type2)tmp_x * (type2)tmp_y); \
-    } while(0)
-
-NEON_OP(mull_u8)
-{
-    uint32_t tmp;
-    uint32_t low;
-    uint32_t high;
-
-    DO_MULL(low, T0, T1, uint8_t, uint16_t);
-    DO_MULL(tmp, T0 >> 8, T1 >> 8, uint8_t, uint16_t);
-    low |= tmp << 16;
-    DO_MULL(high, T0 >> 16, T1 >> 16, uint8_t, uint16_t);
-    DO_MULL(tmp, T0 >> 24, T1 >> 24, uint8_t, uint16_t);
-    high |= tmp << 16;
-    T0 = low;
-    T1 = high;
-    FORCE_RET();
-}
-
-NEON_OP(mull_s8)
-{
-    uint32_t tmp;
-    uint32_t low;
-    uint32_t high;
-
-    DO_MULL(low, T0, T1, int8_t, uint16_t);
-    DO_MULL(tmp, T0 >> 8, T1 >> 8, int8_t, uint16_t);
-    low |= tmp << 16;
-    DO_MULL(high, T0 >> 16, T1 >> 16, int8_t, uint16_t);
-    DO_MULL(tmp, T0 >> 24, T1 >> 24, int8_t, uint16_t);
-    high |= tmp << 16;
-    T0 = low;
-    T1 = high;
-    FORCE_RET();
-}
-
-NEON_OP(mull_u16)
-{
-    uint32_t low;
-    uint32_t high;
-
-    DO_MULL(low, T0, T1, uint16_t, uint32_t);
-    DO_MULL(high, T0 >> 16, T1 >> 16, uint16_t, uint32_t);
-    T0 = low;
-    T1 = high;
-    FORCE_RET();
-}
-
-NEON_OP(mull_s16)
-{
-    uint32_t low;
-    uint32_t high;
-
-    DO_MULL(low, T0, T1, int16_t, uint32_t);
-    DO_MULL(high, T0 >> 16, T1 >> 16, int16_t, uint32_t);
-    T0 = low;
-    T1 = high;
-    FORCE_RET();
-}
-
-NEON_OP(addl_saturate_s32)
-{
-    uint32_t tmp;
-    uint32_t res;
-
-    tmp = env->vfp.scratch[0];
-    res = T0 + tmp;
-    if (((res ^ T0) & SIGNBIT) && !((T0 ^ tmp) & SIGNBIT)) {
-        env->QF = 1;
-        T0 = (T0 >> 31) ^ 0x7fffffff;
-    } else {
-      T0 = res;
-    }
-    tmp = env->vfp.scratch[1];
-    res = T1 + tmp;
-    if (((res ^ T1) & SIGNBIT) && !((T1 ^ tmp) & SIGNBIT)) {
-        env->QF = 1;
-        T1 = (T1 >> 31) ^ 0x7fffffff;
-    } else {
-      T1 = res;
-    }
-    FORCE_RET();
-}
-
-NEON_OP(addl_saturate_s64)
-{
-    uint64_t src1;
-    uint64_t src2;
-    uint64_t res;
-
-    src1 = T0 + ((uint64_t)T1 << 32);
-    src2 = env->vfp.scratch[0] + ((uint64_t)env->vfp.scratch[1] << 32);
-    res = src1 + src2;
-    if (((res ^ src1) & SIGNBIT64) && !((src1 ^ src2) & SIGNBIT64)) {
-        env->QF = 1;
-        T0 = ~(int64_t)src1 >> 63;
-        T1 = T0 ^ 0x80000000;
-    } else {
-      T0 = res;
-      T1 = res >> 32;
-    }
-    FORCE_RET();
-}
-
-NEON_OP(addl_saturate_u64)
-{
-    uint64_t src1;
-    uint64_t src2;
-    uint64_t res;
-
-    src1 = T0 + ((uint64_t)T1 << 32);
-    src2 = env->vfp.scratch[0] + ((uint64_t)env->vfp.scratch[1] << 32);
-    res = src1 + src2;
-    if (res < src1) {
-        env->QF = 1;
-        T0 = 0xffffffff;
-        T1 = 0xffffffff;
-    } else {
-      T0 = res;
-      T1 = res >> 32;
-    }
-    FORCE_RET();
-}
-
-NEON_OP(subl_saturate_s64)
-{
-    uint64_t src1;
-    uint64_t src2;
-    uint64_t res;
-
-    src1 = T0 + ((uint64_t)T1 << 32);
-    src2 = env->vfp.scratch[0] + ((uint64_t)env->vfp.scratch[1] << 32);
-    res = src1 - src2;
-    if (((res ^ src1) & SIGNBIT64) && ((src1 ^ src2) & SIGNBIT64)) {
-        env->QF = 1;
-        T0 = ~(int64_t)src1 >> 63;
-        T1 = T0 ^ 0x80000000;
-    } else {
-      T0 = res;
-      T1 = res >> 32;
-    }
-    FORCE_RET();
-}
-
-NEON_OP(subl_saturate_u64)
-{
-    uint64_t src1;
-    uint64_t src2;
-    uint64_t res;
-
-    src1 = T0 + ((uint64_t)T1 << 32);
-    src2 = env->vfp.scratch[0] + ((uint64_t)env->vfp.scratch[1] << 32);
-    if (src1 < src2) {
-        env->QF = 1;
-        T0 = 0;
-        T1 = 0;
-    } else {
-      res = src1 - src2;
-      T0 = res;
-      T1 = res >> 32;
-    }
-    FORCE_RET();
-}
-
-NEON_OP(negl_u16)
-{
-    uint32_t tmp;
-    tmp = T0 >> 16;
-    tmp = -tmp;
-    T0 = (-T0 & 0xffff) | (tmp << 16);
-    tmp = T1 >> 16;
-    tmp = -tmp;
-    T1 = (-T1 & 0xffff) | (tmp << 16);
-    FORCE_RET();
-}
-
-NEON_OP(negl_u32)
-{
-    T0 = -T0;
-    T1 = -T1;
-    FORCE_RET();
-}
-
-NEON_OP(negl_u64)
-{
-    uint64_t val;
-
-    val = T0 | ((uint64_t)T1 << 32);
-    val = -val;
-    T0 = val;
-    T1 = val >> 32;
-    FORCE_RET();
-}
-
-/* Scalar operations.  */
-NEON_OP(dup_low16)
-{
-    T0 = (T0 & 0xffff) | (T0 << 16);
-    FORCE_RET();
-}
-
-NEON_OP(dup_high16)
-{
-    T0 = (T0 >> 16) | (T0 & 0xffff0000);
-    FORCE_RET();
-}
-
-/* Helper for VEXT */
-NEON_OP(extract)
-{
-    int shift = PARAM1;
-    T0 = (T0 >> shift) | (T1 << (32 - shift));
-    FORCE_RET();
-}
-
-/* Pairwise add long.  Named type is source type.  */
-NEON_OP(paddl_s8)
-{
-    int8_t src1;
-    int8_t src2;
-    uint16_t result;
-    src1 = T0 >> 24;
-    src2 = T0 >> 16;
-    result = (uint16_t)src1 + src2;
-    src1 = T0 >> 8;
-    src2 = T0;
-    T0 = (uint16_t)((uint16_t)src1 + src2) | ((uint32_t)result << 16);
-    FORCE_RET();
-}
-
-NEON_OP(paddl_u8)
-{
-    uint8_t src1;
-    uint8_t src2;
-    uint16_t result;
-    src1 = T0 >> 24;
-    src2 = T0 >> 16;
-    result = (uint16_t)src1 + src2;
-    src1 = T0 >> 8;
-    src2 = T0;
-    T0 = (uint16_t)((uint16_t)src1 + src2) | ((uint32_t)result << 16);
-    FORCE_RET();
-}
-
-NEON_OP(paddl_s16)
-{
-    T0 = (uint32_t)(int16_t)T0 + (uint32_t)(int16_t)(T0 >> 16);
-    FORCE_RET();
-}
-
-NEON_OP(paddl_u16)
-{
-    T0 = (uint32_t)(uint16_t)T0 + (uint32_t)(uint16_t)(T0 >> 16);
-    FORCE_RET();
-}
-
-NEON_OP(paddl_s32)
-{
-    int64_t tmp;
-    tmp = (int64_t)(int32_t)T0 + (int64_t)(int32_t)T1;
-    T0 = tmp;
-    T1 = tmp >> 32;
-    FORCE_RET();
-}
-
-NEON_OP(paddl_u32)
-{
-    uint64_t tmp;
-    tmp = (uint64_t)T0 + (uint64_t)T1;
-    T0 = tmp;
-    T1 = tmp >> 32;
-    FORCE_RET();
-}
-
-/* Count Leading Sign/Zero Bits.  */
-static inline int do_clz8(uint8_t x)
-{
-    int n;
-    for (n = 8; x; n--)
-        x >>= 1;
-    return n;
-}
-
-static inline int do_clz16(uint16_t x)
-{
-    int n;
-    for (n = 16; x; n--)
-        x >>= 1;
-    return n;
-}
-
-NEON_OP(clz_u8)
-{
-    uint32_t result;
-    uint32_t tmp;
-
-    tmp = T0;
-    result = do_clz8(tmp);
-    result |= do_clz8(tmp >> 8) << 8;
-    result |= do_clz8(tmp >> 16) << 16;
-    result |= do_clz8(tmp >> 24) << 24;
-    T0 = result;
-    FORCE_RET();
-}
-
-NEON_OP(clz_u16)
-{
-    uint32_t result;
-    uint32_t tmp;
-    tmp = T0;
-    result = do_clz16(tmp);
-    result |= do_clz16(tmp >> 16) << 16;
-    T0 = result;
-    FORCE_RET();
-}
-
-NEON_OP(cls_s8)
-{
-    uint32_t result;
-    int8_t tmp;
-    tmp = T0;
-    result = do_clz8((tmp < 0) ? ~tmp : tmp) - 1;
-    tmp = T0 >> 8;
-    result |= (do_clz8((tmp < 0) ? ~tmp : tmp) - 1) << 8;
-    tmp = T0 >> 16;
-    result |= (do_clz8((tmp < 0) ? ~tmp : tmp) - 1) << 16;
-    tmp = T0 >> 24;
-    result |= (do_clz8((tmp < 0) ? ~tmp : tmp) - 1) << 24;
-    T0 = result;
-    FORCE_RET();
-}
-
-NEON_OP(cls_s16)
-{
-    uint32_t result;
-    int16_t tmp;
-    tmp = T0;
-    result = do_clz16((tmp < 0) ? ~tmp : tmp) - 1;
-    tmp = T0 >> 16;
-    result |= (do_clz16((tmp < 0) ? ~tmp : tmp) - 1) << 16;
-    T0 = result;
-    FORCE_RET();
-}
-
-NEON_OP(cls_s32)
-{
-    int count;
-    if ((int32_t)T0 < 0)
-        T0 = ~T0;
-    for (count = 32; T0 > 0; count--)
-        T0 = T0 >> 1;
-    T0 = count - 1;
-    FORCE_RET();
-}
-
-/* Bit count.  */
-NEON_OP(cnt_u8)
-{
-    T0 = (T0 & 0x55555555) + ((T0 >>  1) & 0x55555555);
-    T0 = (T0 & 0x33333333) + ((T0 >>  2) & 0x33333333);
-    T0 = (T0 & 0x0f0f0f0f) + ((T0 >>  4) & 0x0f0f0f0f);
-    FORCE_RET();
-}
-
-/* Saturnating negation.  */
-/* ??? Make these use NEON_VOP1 */
-#define DO_QABS8(x) do { \
-    if (x == (int8_t)0x80) { \
-        x = 0x7f; \
-        env->QF = 1; \
-    } else if (x < 0) { \
-        x = -x; \
-    }} while (0)
-NEON_OP(qabs_s8)
-{
-    neon_s8 vec;
-    NEON_UNPACK(neon_s8, vec, T0);
-    DO_QABS8(vec.v1);
-    DO_QABS8(vec.v2);
-    DO_QABS8(vec.v3);
-    DO_QABS8(vec.v4);
-    NEON_PACK(neon_s8, T0, vec);
-    FORCE_RET();
-}
-#undef DO_QABS8
-
-#define DO_QNEG8(x) do { \
-    if (x == (int8_t)0x80) { \
-        x = 0x7f; \
-        env->QF = 1; \
-    } else { \
-        x = -x; \
-    }} while (0)
-NEON_OP(qneg_s8)
-{
-    neon_s8 vec;
-    NEON_UNPACK(neon_s8, vec, T0);
-    DO_QNEG8(vec.v1);
-    DO_QNEG8(vec.v2);
-    DO_QNEG8(vec.v3);
-    DO_QNEG8(vec.v4);
-    NEON_PACK(neon_s8, T0, vec);
-    FORCE_RET();
-}
-#undef DO_QNEG8
-
-#define DO_QABS16(x) do { \
-    if (x == (int16_t)0x8000) { \
-        x = 0x7fff; \
-        env->QF = 1; \
-    } else if (x < 0) { \
-        x = -x; \
-    }} while (0)
-NEON_OP(qabs_s16)
-{
-    neon_s16 vec;
-    NEON_UNPACK(neon_s16, vec, T0);
-    DO_QABS16(vec.v1);
-    DO_QABS16(vec.v2);
-    NEON_PACK(neon_s16, T0, vec);
-    FORCE_RET();
-}
-#undef DO_QABS16
-
-#define DO_QNEG16(x) do { \
-    if (x == (int16_t)0x8000) { \
-        x = 0x7fff; \
-        env->QF = 1; \
-    } else { \
-        x = -x; \
-    }} while (0)
-NEON_OP(qneg_s16)
-{
-    neon_s16 vec;
-    NEON_UNPACK(neon_s16, vec, T0);
-    DO_QNEG16(vec.v1);
-    DO_QNEG16(vec.v2);
-    NEON_PACK(neon_s16, T0, vec);
-    FORCE_RET();
-}
-#undef DO_QNEG16
-
-NEON_OP(qabs_s32)
-{
-    if (T0 == 0x80000000) {
-        T0 = 0x7fffffff;
-        env->QF = 1;
-    } else if ((int32_t)T0 < 0) {
-        T0 = -T0;
-    }
-    FORCE_RET();
-}
-
-NEON_OP(qneg_s32)
-{
-    if (T0 == 0x80000000) {
-        T0 = 0x7fffffff;
-        env->QF = 1;
-    } else {
-        T0 = -T0;
-    }
-    FORCE_RET();
-}
-
-/* Unary opperations */
-#define NEON_FN(dest, src, dummy) dest = (src < 0) ? -src : src
-NEON_VOP1(abs_s8, neon_s8, 4)
-NEON_VOP1(abs_s16, neon_s16, 2)
-NEON_OP(abs_s32)
-{
-    if ((int32_t)T0 < 0)
-        T0 = -T0;
-    FORCE_RET();
-}
-#undef NEON_FN
-
-/* Transpose.  Argument order is rather strange to avoid special casing
-   the tranlation code.
-   On input T0 = rm, T1 = rd.  On output T0 = rd, T1 = rm  */
-NEON_OP(trn_u8)
-{
-    uint32_t rd;
-    uint32_t rm;
-    rd = ((T0 & 0x00ff00ff) << 8) | (T1 & 0x00ff00ff);
-    rm = ((T1 & 0xff00ff00) >> 8) | (T0 & 0xff00ff00);
-    T0 = rd;
-    T1 = rm;
-    FORCE_RET();
-}
-
-NEON_OP(trn_u16)
-{
-    uint32_t rd;
-    uint32_t rm;
-    rd = (T0 << 16) | (T1 & 0xffff);
-    rm = (T1 >> 16) | (T0 & 0xffff0000);
-    T0 = rd;
-    T1 = rm;
-    FORCE_RET();
-}
-
-/* Worker routines for zip and unzip.  */
-NEON_OP(unzip_u8)
-{
-    uint32_t rd;
-    uint32_t rm;
-    rd = (T0 & 0xff) | ((T0 >> 8) & 0xff00)
-         | ((T1 << 16) & 0xff0000) | ((T1 << 8) & 0xff000000);
-    rm = ((T0 >> 8) & 0xff) | ((T0 >> 16) & 0xff00)
-         | ((T1 << 8) & 0xff0000) | (T1 & 0xff000000);
-    T0 = rd;
-    T1 = rm;
-    FORCE_RET();
-}
-
-NEON_OP(zip_u8)
-{
-    uint32_t rd;
-    uint32_t rm;
-    rd = (T0 & 0xff) | ((T1 << 8) & 0xff00)
-         | ((T0 << 16) & 0xff0000) | ((T1 << 24) & 0xff000000);
-    rm = ((T0 >> 16) & 0xff) | ((T1 >> 8) & 0xff00)
-         | ((T0 >> 8) & 0xff0000) | (T1 & 0xff000000);
-    T0 = rd;
-    T1 = rm;
-    FORCE_RET();
-}
-
-NEON_OP(zip_u16)
-{
-    uint32_t tmp;
-
-    tmp = (T0 & 0xffff) | (T1 << 16);
-    T1 = (T1 & 0xffff0000) | (T0 >> 16);
-    T0 = tmp;
-    FORCE_RET();
-}
-
-NEON_OP(dup_u8)
-{
-    T0 = (T0 >> PARAM1) & 0xff;
-    T0 |= T0 << 8;
-    T0 |= T0 << 16;
-    FORCE_RET();
-}