ARMv7 support.

git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@3572 c046a42c-6fe2-441c-8c8c-71466251a162

9 files changed, 8041 insertions, 439 deletions

diff --git a/target-arm/cpu.h b/target-arm/cpu.h
index fb22b90..59b9696 100644
--- a/target-arm/cpu.h
+++ b/target-arm/cpu.h
@@ -37,6 +37,18 @@
 #define EXCP_IRQ             5
 #define EXCP_FIQ             6
 #define EXCP_BKPT            7
+#define EXCP_EXCEPTION_EXIT  8   /* Return from v7M exception.  */
+
+#define ARMV7M_EXCP_RESET   1
+#define ARMV7M_EXCP_NMI     2
+#define ARMV7M_EXCP_HARD    3
+#define ARMV7M_EXCP_MEM     4
+#define ARMV7M_EXCP_BUS     5
+#define ARMV7M_EXCP_USAGE   6
+#define ARMV7M_EXCP_SVC     11
+#define ARMV7M_EXCP_DEBUG   12
+#define ARMV7M_EXCP_PENDSV  14
+#define ARMV7M_EXCP_SYSTICK 15
 
 typedef void ARMWriteCPFunc(void *opaque, int cp_info,
                             int srcreg, int operand, uint32_t value);
@@ -76,17 +88,22 @@ typedef struct CPUARMState {
     uint32_t VF; /* V is the bit 31. All other bits are undefined */
     uint32_t NZF; /* N is bit 31. Z is computed from NZF */
     uint32_t QF; /* 0 or 1 */
-
-    int thumb; /* 0 = arm mode, 1 = thumb mode */
+    uint32_t GE; /* cpsr[19:16] */
+    int thumb; /* cprs[5]. 0 = arm mode, 1 = thumb mode. */
+    uint32_t condexec_bits; /* IT bits.  cpsr[15:10,26:25].  */
 
     /* System control coprocessor (cp15) */
     struct {
         uint32_t c0_cpuid;
         uint32_t c0_cachetype;
+        uint32_t c0_c1[8]; /* Feature registers.  */
+        uint32_t c0_c2[8]; /* Instruction set registers.  */
         uint32_t c1_sys; /* System control register.  */
         uint32_t c1_coproc; /* Coprocessor access register.  */
         uint32_t c1_xscaleauxcr; /* XScale auxiliary control register.  */
-        uint32_t c2_base; /* MMU translation table base.  */
+        uint32_t c2_base0; /* MMU translation table base 0.  */
+        uint32_t c2_base1; /* MMU translation table base 1.  */
+        uint32_t c2_mask; /* MMU translation table base mask.  */
         uint32_t c2_data; /* MPU data cachable bits.  */
         uint32_t c2_insn; /* MPU instruction cachable bits.  */
         uint32_t c3; /* MMU domain access control register
@@ -100,6 +117,9 @@ typedef struct CPUARMState {
         uint32_t c9_data;
         uint32_t c13_fcse; /* FCSE PID.  */
         uint32_t c13_context; /* Context ID.  */
+        uint32_t c13_tls1; /* User RW Thread register.  */
+        uint32_t c13_tls2; /* User RO Thread register.  */
+        uint32_t c13_tls3; /* Privileged Thread register.  */
         uint32_t c15_cpar; /* XScale Coprocessor Access Register */
         uint32_t c15_ticonfig; /* TI925T configuration byte.  */
         uint32_t c15_i_max; /* Maximum D-cache dirty line index.  */
@@ -107,6 +127,17 @@ typedef struct CPUARMState {
         uint32_t c15_threadid; /* TI debugger thread-ID.  */
     } cp15;
 
+    struct {
+        uint32_t other_sp;
+        uint32_t vecbase;
+        uint32_t basepri;
+        uint32_t control;
+        int current_sp;
+        int exception;
+        int pending_exception;
+        void *nvic;
+    } v7m;
+
     /* Coprocessor IO used by peripherals */
     struct {
         ARMReadCPFunc *cp_read;
@@ -117,6 +148,10 @@ typedef struct CPUARMState {
     /* Internal CPU feature flags.  */
     uint32_t features;
 
+    /* Callback for vectored interrupt controller.  */
+    int (*get_irq_vector)(struct CPUARMState *);
+    void *irq_opaque;
+
     /* exception/interrupt handling */
     jmp_buf jmp_env;
     int exception_index;
@@ -126,7 +161,7 @@ typedef struct CPUARMState {
 
     /* VFP coprocessor state.  */
     struct {
-        float64 regs[16];
+        float64 regs[32];
 
         uint32_t xregs[16];
         /* We store these fpcsr fields separately for convenience.  */
@@ -136,9 +171,16 @@ typedef struct CPUARMState {
         /* 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];
 
         float_status fp_status;
     } vfp;
+#if defined(CONFIG_USER_ONLY)
+    struct mmon_state *mmon_entry;
+#else
+    uint32_t mmon_addr;
+#endif
 
     /* iwMMXt coprocessor state.  */
     struct {
@@ -169,6 +211,7 @@ int cpu_arm_exec(CPUARMState *s);
 void cpu_arm_close(CPUARMState *s);
 void do_interrupt(CPUARMState *);
 void switch_mode(CPUARMState *, int);
+uint32_t do_arm_semihosting(CPUARMState *env);
 
 /* you can call this signal handler from your SIGBUS and SIGSEGV
    signal handlers to inform the virtual CPU of exceptions. non zero
@@ -176,6 +219,9 @@ void switch_mode(CPUARMState *, int);
 int cpu_arm_signal_handler(int host_signum, void *pinfo,
                            void *puc);
 
+void cpu_lock(void);
+void cpu_unlock(void);
+
 #define CPSR_M (0x1f)
 #define CPSR_T (1 << 5)
 #define CPSR_F (1 << 6)
@@ -183,13 +229,24 @@ int cpu_arm_signal_handler(int host_signum, void *pinfo,
 #define CPSR_A (1 << 8)
 #define CPSR_E (1 << 9)
 #define CPSR_IT_2_7 (0xfc00)
-/* Bits 20-23 reserved.  */
+#define CPSR_GE (0xf << 16)
+#define CPSR_RESERVED (0xf << 20)
 #define CPSR_J (1 << 24)
 #define CPSR_IT_0_1 (3 << 25)
 #define CPSR_Q (1 << 27)
-#define CPSR_NZCV (0xf << 28)
+#define CPSR_V (1 << 28)
+#define CPSR_C (1 << 29)
+#define CPSR_Z (1 << 30)
+#define CPSR_N (1 << 31)
+#define CPSR_NZCV (CPSR_N | CPSR_Z | CPSR_C | CPSR_V)
+
+#define CPSR_IT (CPSR_IT_0_1 | CPSR_IT_2_7)
+#define CACHED_CPSR_BITS (CPSR_T | CPSR_GE | CPSR_IT | CPSR_Q | CPSR_NZCV)
+/* Bits writable in user mode.  */
+#define CPSR_USER (CPSR_NZCV | CPSR_Q | CPSR_GE)
+/* Execution state bits.  MRS read as zero, MSR writes ignored.  */
+#define CPSR_EXEC (CPSR_T | CPSR_IT | CPSR_J)
 
-#define CACHED_CPSR_BITS (CPSR_T | CPSR_Q | CPSR_NZCV)
 /* Return the current CPSR value.  */
 static inline uint32_t cpsr_read(CPUARMState *env)
 {
@@ -197,7 +254,21 @@ static inline uint32_t cpsr_read(CPUARMState *env)
     ZF = (env->NZF == 0);
     return env->uncached_cpsr | (env->NZF & 0x80000000) | (ZF << 30) |
         (env->CF << 29) | ((env->VF & 0x80000000) >> 3) | (env->QF << 27)
-        | (env->thumb << 5);
+        | (env->thumb << 5) | ((env->condexec_bits & 3) << 25)
+        | ((env->condexec_bits & 0xfc) << 8)
+        | (env->GE << 16);
+}
+
+/* Return the current xPSR value.  */
+static inline uint32_t xpsr_read(CPUARMState *env)
+{
+    int ZF;
+    ZF = (env->NZF == 0);
+    return (env->NZF & 0x80000000) | (ZF << 30)
+        | (env->CF << 29) | ((env->VF & 0x80000000) >> 3) | (env->QF << 27)
+        | (env->thumb << 24) | ((env->condexec_bits & 3) << 25)
+        | ((env->condexec_bits & 0xfc) << 8)
+        | env->v7m.exception;
 }
 
 /* Set the CPSR.  Note that some bits of mask must be all-set or all-clear.  */
@@ -213,6 +284,17 @@ static inline void cpsr_write(CPUARMState *env, uint32_t val, uint32_t mask)
         env->QF = ((val & CPSR_Q) != 0);
     if (mask & CPSR_T)
         env->thumb = ((val & CPSR_T) != 0);
+    if (mask & CPSR_IT_0_1) {
+        env->condexec_bits &= ~3;
+        env->condexec_bits |= (val >> 25) & 3;
+    }
+    if (mask & CPSR_IT_2_7) {
+        env->condexec_bits &= 3;
+        env->condexec_bits |= (val >> 8) & 0xfc;
+    }
+    if (mask & CPSR_GE) {
+        env->GE = (val >> 16) & 0xf;
+    }
 
     if ((env->uncached_cpsr ^ val) & mask & CPSR_M) {
         switch_mode(env, val & CPSR_M);
@@ -221,6 +303,32 @@ static inline void cpsr_write(CPUARMState *env, uint32_t val, uint32_t mask)
     env->uncached_cpsr = (env->uncached_cpsr & ~mask) | (val & mask);
 }
 
+/* Set the xPSR.  Note that some bits of mask must be all-set or all-clear.  */
+static inline void xpsr_write(CPUARMState *env, uint32_t val, uint32_t mask)
+{
+    /* NOTE: N = 1 and Z = 1 cannot be stored currently */
+    if (mask & CPSR_NZCV) {
+        env->NZF = (val & 0xc0000000) ^ 0x40000000;
+        env->CF = (val >> 29) & 1;
+        env->VF = (val << 3) & 0x80000000;
+    }
+    if (mask & CPSR_Q)
+        env->QF = ((val & CPSR_Q) != 0);
+    if (mask & (1 << 24))
+        env->thumb = ((val & (1 << 24)) != 0);
+    if (mask & CPSR_IT_0_1) {
+        env->condexec_bits &= ~3;
+        env->condexec_bits |= (val >> 25) & 3;
+    }
+    if (mask & CPSR_IT_2_7) {
+        env->condexec_bits &= 3;
+        env->condexec_bits |= (val >> 8) & 0xfc;
+    }
+    if (mask & 0x1ff) {
+        env->v7m.exception = val & 0x1ff;
+    }
+}
+
 enum arm_cpu_mode {
   ARM_CPU_MODE_USR = 0x10,
   ARM_CPU_MODE_FIQ = 0x11,
@@ -234,6 +342,8 @@ enum arm_cpu_mode {
 /* VFP system registers.  */
 #define ARM_VFP_FPSID   0
 #define ARM_VFP_FPSCR   1
+#define ARM_VFP_MVFR1   6
+#define ARM_VFP_MVFR0   7
 #define ARM_VFP_FPEXC   8
 #define ARM_VFP_FPINST  9
 #define ARM_VFP_FPINST2 10
@@ -253,7 +363,15 @@ enum arm_features {
     ARM_FEATURE_AUXCR,  /* ARM1026 Auxiliary control register.  */
     ARM_FEATURE_XSCALE, /* Intel XScale extensions.  */
     ARM_FEATURE_IWMMXT, /* Intel iwMMXt extension.  */
+    ARM_FEATURE_V6,
+    ARM_FEATURE_V6K,
+    ARM_FEATURE_V7,
+    ARM_FEATURE_THUMB2,
     ARM_FEATURE_MPU,    /* Only has Memory Protection Unit, not full MMU.  */
+    ARM_FEATURE_VFP3,
+    ARM_FEATURE_NEON,
+    ARM_FEATURE_DIV,
+    ARM_FEATURE_M, /* Microcontroller profile.  */
     ARM_FEATURE_OMAPCP  /* OMAP specific CP15 ops handling.  */
 };
 
@@ -264,27 +382,44 @@ static inline int arm_feature(CPUARMState *env, int feature)
 
 void arm_cpu_list(FILE *f, int (*cpu_fprintf)(FILE *f, const char *fmt, ...));
 
+/* Interface between CPU and Interrupt controller.  */
+void armv7m_nvic_set_pending(void *opaque, int irq);
+int armv7m_nvic_acknowledge_irq(void *opaque);
+void armv7m_nvic_complete_irq(void *opaque, int irq);
+
 void cpu_arm_set_cp_io(CPUARMState *env, int cpnum,
                        ARMReadCPFunc *cp_read, ARMWriteCPFunc *cp_write,
                        void *opaque);
 
-#define ARM_CPUID_ARM1026   0x4106a262
-#define ARM_CPUID_ARM926    0x41069265
-#define ARM_CPUID_ARM946    0x41059461
-#define ARM_CPUID_TI915T    0x54029152
-#define ARM_CPUID_TI925T    0x54029252
-#define ARM_CPUID_PXA250    0x69052100
-#define ARM_CPUID_PXA255    0x69052d00
-#define ARM_CPUID_PXA260    0x69052903
-#define ARM_CPUID_PXA261    0x69052d05
-#define ARM_CPUID_PXA262    0x69052d06
-#define ARM_CPUID_PXA270    0x69054110
-#define ARM_CPUID_PXA270_A0 0x69054110
-#define ARM_CPUID_PXA270_A1 0x69054111
-#define ARM_CPUID_PXA270_B0 0x69054112
-#define ARM_CPUID_PXA270_B1 0x69054113
-#define ARM_CPUID_PXA270_C0 0x69054114
-#define ARM_CPUID_PXA270_C5 0x69054117
+/* Does the core conform to the the "MicroController" profile. e.g. Cortex-M3.
+   Note the M in older cores (eg. ARM7TDMI) stands for Multiply. These are
+   conventional cores (ie. Application or Realtime profile).  */
+
+#define IS_M(env) arm_feature(env, ARM_FEATURE_M)
+#define ARM_CPUID(env) (env->cp15.c0_cpuid)
+
+#define ARM_CPUID_ARM1026     0x4106a262
+#define ARM_CPUID_ARM926      0x41069265
+#define ARM_CPUID_ARM946      0x41059461
+#define ARM_CPUID_TI915T      0x54029152
+#define ARM_CPUID_TI925T      0x54029252
+#define ARM_CPUID_PXA250      0x69052100
+#define ARM_CPUID_PXA255      0x69052d00
+#define ARM_CPUID_PXA260      0x69052903
+#define ARM_CPUID_PXA261      0x69052d05
+#define ARM_CPUID_PXA262      0x69052d06
+#define ARM_CPUID_PXA270      0x69054110
+#define ARM_CPUID_PXA270_A0   0x69054110
+#define ARM_CPUID_PXA270_A1   0x69054111
+#define ARM_CPUID_PXA270_B0   0x69054112
+#define ARM_CPUID_PXA270_B1   0x69054113
+#define ARM_CPUID_PXA270_C0   0x69054114
+#define ARM_CPUID_PXA270_C5   0x69054117
+#define ARM_CPUID_ARM1136     0x4117b363
+#define ARM_CPUID_ARM11MPCORE 0x410fb022
+#define ARM_CPUID_CORTEXA8    0x410fc080
+#define ARM_CPUID_CORTEXM3    0x410fc231
+#define ARM_CPUID_ANY         0xffffffff
 
 #if defined(CONFIG_USER_ONLY)
 #define TARGET_PAGE_BITS 12
@@ -302,6 +437,8 @@ void cpu_arm_set_cp_io(CPUARMState *env, int cpnum,
 #define cpu_signal_handler cpu_arm_signal_handler
 #define cpu_list arm_cpu_list
 
+#define ARM_CPU_SAVE_VERSION 1
+
 /* MMU modes definitions */
 #define MMU_MODE0_SUFFIX _kernel
 #define MMU_MODE1_SUFFIX _user
diff --git a/target-arm/exec.h b/target-arm/exec.h
index 0c53b8d..cb20206 100644
--- a/target-arm/exec.h
+++ b/target-arm/exec.h
@@ -68,12 +68,18 @@ static inline int cpu_halted(CPUState *env) {
 
 /* In op_helper.c */
 
-void cpu_lock(void);
-void cpu_unlock(void);
 void helper_set_cp(CPUState *, uint32_t, uint32_t);
 uint32_t helper_get_cp(CPUState *, uint32_t);
 void helper_set_cp15(CPUState *, uint32_t, uint32_t);
 uint32_t helper_get_cp15(CPUState *, uint32_t);
+void helper_set_r13_banked(CPUState *env, int mode, uint32_t val);
+uint32_t helper_get_r13_banked(CPUState *env, int mode);
+uint32_t helper_v7m_mrs(CPUState *env, int reg);
+void helper_v7m_msr(CPUState *env, int reg, uint32_t val);
+
+void helper_mark_exclusive(CPUARMState *, uint32_t addr);
+int helper_test_exclusive(CPUARMState *, uint32_t addr);
+void helper_clrex(CPUARMState *env);
 
 void cpu_loop_exit(void);
 
@@ -91,4 +97,11 @@ 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 06eac66..6975256 100644
--- a/target-arm/helper.c
+++ b/target-arm/helper.c
@@ -4,6 +4,25 @@
 
 #include "cpu.h"
 #include "exec-all.h"
+#include "gdbstub.h"
+
+static uint32_t cortexa8_cp15_c0_c1[8] =
+{ 0x1031, 0x11, 0x400, 0, 0x31100003, 0x20000000, 0x01202000, 0x11 };
+
+static uint32_t cortexa8_cp15_c0_c2[8] =
+{ 0x00101111, 0x12112111, 0x21232031, 0x11112131, 0x00111142, 0, 0, 0 };
+
+static uint32_t mpcore_cp15_c0_c1[8] =
+{ 0x111, 0x1, 0, 0x2, 0x01100103, 0x10020302, 0x01222000, 0 };
+
+static uint32_t mpcore_cp15_c0_c2[8] =
+{ 0x00100011, 0x12002111, 0x11221011, 0x01102131, 0x141, 0, 0, 0 };
+
+static uint32_t arm1136_cp15_c0_c1[8] =
+{ 0x111, 0x1, 0x2, 0x3, 0x01130003, 0x10030302, 0x01222110, 0 };
+
+static uint32_t arm1136_cp15_c0_c2[8] =
+{ 0x00140011, 0x12002111, 0x11231111, 0x01102131, 0x141, 0, 0, 0 };
 
 static uint32_t cpu_arm_find_by_name(const char *name);
 
@@ -34,6 +53,62 @@ static void cpu_reset_model_id(CPUARMState *env, uint32_t id)
         env->cp15.c0_cachetype = 0x1dd20d2;
         env->cp15.c1_sys = 0x00090078;
         break;
+    case ARM_CPUID_ARM1136:
+        set_feature(env, ARM_FEATURE_V6);
+        set_feature(env, ARM_FEATURE_VFP);
+        set_feature(env, ARM_FEATURE_AUXCR);
+        env->vfp.xregs[ARM_VFP_FPSID] = 0x410120b4;
+        env->vfp.xregs[ARM_VFP_MVFR0] = 0x11111111;
+        env->vfp.xregs[ARM_VFP_MVFR1] = 0x00000000;
+        memcpy(env->cp15.c0_c1, arm1136_cp15_c0_c1, 8 * sizeof(uint32_t));
+        memcpy(env->cp15.c0_c1, arm1136_cp15_c0_c2, 8 * sizeof(uint32_t));
+        env->cp15.c0_cachetype = 0x1dd20d2;
+        break;
+    case ARM_CPUID_ARM11MPCORE:
+        set_feature(env, ARM_FEATURE_V6);
+        set_feature(env, ARM_FEATURE_V6K);
+        set_feature(env, ARM_FEATURE_VFP);
+        set_feature(env, ARM_FEATURE_AUXCR);
+        env->vfp.xregs[ARM_VFP_FPSID] = 0x410120b4;
+        env->vfp.xregs[ARM_VFP_MVFR0] = 0x11111111;
+        env->vfp.xregs[ARM_VFP_MVFR1] = 0x00000000;
+        memcpy(env->cp15.c0_c1, mpcore_cp15_c0_c1, 8 * sizeof(uint32_t));
+        memcpy(env->cp15.c0_c1, mpcore_cp15_c0_c2, 8 * sizeof(uint32_t));
+        env->cp15.c0_cachetype = 0x1dd20d2;
+        break;
+    case ARM_CPUID_CORTEXA8:
+        set_feature(env, ARM_FEATURE_V6);
+        set_feature(env, ARM_FEATURE_V6K);
+        set_feature(env, ARM_FEATURE_V7);
+        set_feature(env, ARM_FEATURE_AUXCR);
+        set_feature(env, ARM_FEATURE_THUMB2);
+        set_feature(env, ARM_FEATURE_VFP);
+        set_feature(env, ARM_FEATURE_VFP3);
+        set_feature(env, ARM_FEATURE_NEON);
+        env->vfp.xregs[ARM_VFP_FPSID] = 0x410330c0;
+        env->vfp.xregs[ARM_VFP_MVFR0] = 0x11110222;
+        env->vfp.xregs[ARM_VFP_MVFR1] = 0x00011100;
+        memcpy(env->cp15.c0_c1, cortexa8_cp15_c0_c1, 8 * sizeof(uint32_t));
+        memcpy(env->cp15.c0_c1, cortexa8_cp15_c0_c2, 8 * sizeof(uint32_t));
+        env->cp15.c0_cachetype = 0x1dd20d2;
+        break;
+    case ARM_CPUID_CORTEXM3:
+        set_feature(env, ARM_FEATURE_V6);
+        set_feature(env, ARM_FEATURE_THUMB2);
+        set_feature(env, ARM_FEATURE_V7);
+        set_feature(env, ARM_FEATURE_M);
+        set_feature(env, ARM_FEATURE_DIV);
+        break;
+    case ARM_CPUID_ANY: /* For userspace emulation.  */
+        set_feature(env, ARM_FEATURE_V6);
+        set_feature(env, ARM_FEATURE_V6K);
+        set_feature(env, ARM_FEATURE_V7);
+        set_feature(env, ARM_FEATURE_THUMB2);
+        set_feature(env, ARM_FEATURE_VFP);
+        set_feature(env, ARM_FEATURE_VFP3);
+        set_feature(env, ARM_FEATURE_NEON);
+        set_feature(env, ARM_FEATURE_DIV);
+        break;
     case ARM_CPUID_TI915T:
     case ARM_CPUID_TI925T:
         set_feature(env, ARM_FEATURE_OMAPCP);
@@ -85,6 +160,10 @@ void cpu_reset(CPUARMState *env)
 #else
     /* SVC mode with interrupts disabled.  */
     env->uncached_cpsr = ARM_CPU_MODE_SVC | CPSR_A | CPSR_F | CPSR_I;
+    /* On ARMv7-M the CPSR_I is the value of the PRIMASK register, and is
+       clear at reset.  */
+    if (IS_M(env))
+        env->uncached_cpsr &= ~CPSR_I;
     env->vfp.xregs[ARM_VFP_FPEXC] = 0;
 #endif
     env->regs[15] = 0;
@@ -117,6 +196,10 @@ static const struct arm_cpu_t arm_cpu_names[] = {
     { ARM_CPUID_ARM926, "arm926"},
     { ARM_CPUID_ARM946, "arm946"},
     { ARM_CPUID_ARM1026, "arm1026"},
+    { ARM_CPUID_ARM1136, "arm1136"},
+    { ARM_CPUID_ARM11MPCORE, "arm11mpcore"},
+    { ARM_CPUID_CORTEXM3, "cortex-m3"},
+    { ARM_CPUID_CORTEXA8, "cortex-a8"},
     { ARM_CPUID_TI925T, "ti925t" },
     { ARM_CPUID_PXA250, "pxa250" },
     { ARM_CPUID_PXA255, "pxa255" },
@@ -130,6 +213,7 @@ static const struct arm_cpu_t arm_cpu_names[] = {
     { ARM_CPUID_PXA270_B1, "pxa270-b1" },
     { ARM_CPUID_PXA270_C0, "pxa270-c0" },
     { ARM_CPUID_PXA270_C5, "pxa270-c5" },
+    { ARM_CPUID_ANY, "any"},
     { 0, NULL}
 };
 
@@ -164,6 +248,30 @@ 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;
+}
+
 #if defined(CONFIG_USER_ONLY)
 
 void do_interrupt (CPUState *env)
@@ -171,6 +279,16 @@ void do_interrupt (CPUState *env)
     env->exception_index = -1;
 }
 
+/* Structure used to record exclusive memory locations.  */
+typedef struct mmon_state {
+    struct mmon_state *next;
+    CPUARMState *cpu_env;
+    uint32_t addr;
+} mmon_state;
+
+/* Chain of current locks.  */
+static mmon_state* mmon_head = NULL;
+
 int cpu_arm_handle_mmu_fault (CPUState *env, target_ulong address, int rw,
                               int mmu_idx, int is_softmmu)
 {
@@ -184,6 +302,64 @@ int cpu_arm_handle_mmu_fault (CPUState *env, target_ulong address, int rw,
     return 1;
 }
 
+static void allocate_mmon_state(CPUState *env)
+{
+    env->mmon_entry = malloc(sizeof (mmon_state));
+    if (!env->mmon_entry)
+        abort();
+    memset (env->mmon_entry, 0, sizeof (mmon_state));
+    env->mmon_entry->cpu_env = env;
+    mmon_head = env->mmon_entry;
+}
+
+/* Flush any monitor locks for the specified address.  */
+static void flush_mmon(uint32_t addr)
+{
+    mmon_state *mon;
+
+    for (mon = mmon_head; mon; mon = mon->next)
+      {
+        if (mon->addr != addr)
+          continue;
+
+        mon->addr = 0;
+        break;
+      }
+}
+
+/* Mark an address for exclusive access.  */
+void helper_mark_exclusive(CPUState *env, uint32_t addr)
+{
+    if (!env->mmon_entry)
+        allocate_mmon_state(env);
+    /* Clear any previous locks.  */
+    flush_mmon(addr);
+    env->mmon_entry->addr = addr;
+}
+
+/* Test if an exclusive address is still exclusive.  Returns zero
+   if the address is still exclusive.   */
+int helper_test_exclusive(CPUState *env, uint32_t addr)
+{
+    int res;
+
+    if (!env->mmon_entry)
+        return 1;
+    if (env->mmon_entry->addr == addr)
+        res = 0;
+    else
+        res = 1;
+    flush_mmon(addr);
+    return res;
+}
+
+void helper_clrex(CPUState *env)
+{
+    if (!(env->mmon_entry && env->mmon_entry->addr))
+        return;
+    flush_mmon(env->mmon_entry->addr);
+}
+
 target_phys_addr_t cpu_get_phys_page_debug(CPUState *env, target_ulong addr)
 {
     return addr;
@@ -215,12 +391,35 @@ uint32_t helper_get_cp15(CPUState *env, uint32_t insn)
     return 0;
 }
 
+/* These should probably raise undefined insn exceptions.  */
+void helper_v7m_msr(CPUState *env, int reg, uint32_t val)
+{
+    cpu_abort(env, "v7m_mrs %d\n", reg);
+}
+
+uint32_t helper_v7m_mrs(CPUState *env, int reg)
+{
+    cpu_abort(env, "v7m_mrs %d\n", reg);
+    return 0;
+}
+
 void switch_mode(CPUState *env, int mode)
 {
     if (mode != ARM_CPU_MODE_USR)
         cpu_abort(env, "Tried to switch out of user mode\n");
 }
 
+void helper_set_r13_banked(CPUState *env, int mode, uint32_t val)
+{
+    cpu_abort(env, "banked r13 write\n");
+}
+
+uint32_t helper_get_r13_banked(CPUState *env, int mode)
+{
+    cpu_abort(env, "banked r13 read\n");
+    return 0;
+}
+
 #else
 
 extern int semihosting_enabled;
@@ -275,6 +474,129 @@ void switch_mode(CPUState *env, int mode)
     env->spsr = env->banked_spsr[i];
 }
 
+static void v7m_push(CPUARMState *env, uint32_t val)
+{
+    env->regs[13] -= 4;
+    stl_phys(env->regs[13], val);
+}
+
+static uint32_t v7m_pop(CPUARMState *env)
+{
+    uint32_t val;
+    val = ldl_phys(env->regs[13]);
+    env->regs[13] += 4;
+    return val;
+}
+
+/* Switch to V7M main or process stack pointer.  */
+static void switch_v7m_sp(CPUARMState *env, int process)
+{
+    uint32_t tmp;
+    if (env->v7m.current_sp != process) {
+        tmp = env->v7m.other_sp;
+        env->v7m.other_sp = env->regs[13];
+        env->regs[13] = tmp;
+        env->v7m.current_sp = process;
+    }
+}
+
+static void do_v7m_exception_exit(CPUARMState *env)
+{
+    uint32_t type;
+    uint32_t xpsr;
+
+    type = env->regs[15];
+    if (env->v7m.exception != 0)
+        armv7m_nvic_complete_irq(env->v7m.nvic, env->v7m.exception);
+
+    /* Switch to the target stack.  */
+    switch_v7m_sp(env, (type & 4) != 0);
+    /* Pop registers.  */
+    env->regs[0] = v7m_pop(env);
+    env->regs[1] = v7m_pop(env);
+    env->regs[2] = v7m_pop(env);
+    env->regs[3] = v7m_pop(env);
+    env->regs[12] = v7m_pop(env);
+    env->regs[14] = v7m_pop(env);
+    env->regs[15] = v7m_pop(env);
+    xpsr = v7m_pop(env);
+    xpsr_write(env, xpsr, 0xfffffdff);
+    /* Undo stack alignment.  */
+    if (xpsr & 0x200)
+        env->regs[13] |= 4;
+    /* ??? The exception return type specifies Thread/Handler mode.  However
+       this is also implied by the xPSR value. Not sure what to do
+       if there is a mismatch.  */
+    /* ??? Likewise for mismatches between the CONTROL register and the stack
+       pointer.  */
+}
+
+void do_interrupt_v7m(CPUARMState *env)
+{
+    uint32_t xpsr = xpsr_read(env);
+    uint32_t lr;
+    uint32_t addr;
+
+    lr = 0xfffffff1;
+    if (env->v7m.current_sp)
+        lr |= 4;
+    if (env->v7m.exception == 0)
+        lr |= 8;
+
+    /* For exceptions we just mark as pending on the NVIC, and let that
+       handle it.  */
+    /* TODO: Need to escalate if the current priority is higher than the
+       one we're raising.  */
+    switch (env->exception_index) {
+    case EXCP_UDEF:
+        armv7m_nvic_set_pending(env->v7m.nvic, ARMV7M_EXCP_USAGE);
+        return;
+    case EXCP_SWI:
+        env->regs[15] += 2;
+        armv7m_nvic_set_pending(env->v7m.nvic, ARMV7M_EXCP_SVC);
+        return;
+    case EXCP_PREFETCH_ABORT:
+    case EXCP_DATA_ABORT:
+        armv7m_nvic_set_pending(env->v7m.nvic, ARMV7M_EXCP_MEM);
+        return;
+    case EXCP_BKPT:
+        armv7m_nvic_set_pending(env->v7m.nvic, ARMV7M_EXCP_DEBUG);
+        return;
+    case EXCP_IRQ:
+        env->v7m.exception = armv7m_nvic_acknowledge_irq(env->v7m.nvic);
+        break;
+    case EXCP_EXCEPTION_EXIT:
+        do_v7m_exception_exit(env);
+        return;
+    default:
+        cpu_abort(env, "Unhandled exception 0x%x\n", env->exception_index);
+        return; /* Never happens.  Keep compiler happy.  */
+    }
+
+    /* Align stack pointer.  */
+    /* ??? Should only do this if Configuration Control Register
+       STACKALIGN bit is set.  */
+    if (env->regs[13] & 4) {
+        env->regs[13] += 4;
+        xpsr |= 0x200;
+    }
+    /* Switch to the hander mode.  */
+    v7m_push(env, xpsr);
+    v7m_push(env, env->regs[15]);
+    v7m_push(env, env->regs[14]);
+    v7m_push(env, env->regs[12]);
+    v7m_push(env, env->regs[3]);
+    v7m_push(env, env->regs[2]);
+    v7m_push(env, env->regs[1]);
+    v7m_push(env, env->regs[0]);
+    switch_v7m_sp(env, 0);
+    env->uncached_cpsr &= ~CPSR_IT;
+    env->regs[14] = lr;
+    addr = ldl_phys(env->v7m.vecbase + env->v7m.exception * 4);
+    env->regs[15] = addr & 0xfffffffe;
+    env->thumb = addr & 1;
+}
+
 /* Handle a CPU exception.  */
 void do_interrupt(CPUARMState *env)
 {
@@ -283,6 +605,10 @@ void do_interrupt(CPUARMState *env)
     int new_mode;
     uint32_t offset;
 
+    if (IS_M(env)) {
+        do_interrupt_v7m(env);
+        return;
+    }
     /* TODO: Vectored interrupt controller.  */
     switch (env->exception_index) {
     case EXCP_UDEF:
@@ -317,8 +643,19 @@ void do_interrupt(CPUARMState *env)
         /* The PC already points to the next instructon.  */
         offset = 0;
         break;
-    case EXCP_PREFETCH_ABORT:
     case EXCP_BKPT:
+        /* See if this is a semihosting syscall.  */
+        if (env->thumb) {
+            mask = lduw_code(env->regs[15]) & 0xff;
+            if (mask == 0xab
+                  && (env->uncached_cpsr & CPSR_M) != ARM_CPU_MODE_USR) {
+                env->regs[15] += 2;
+                env->regs[0] = do_arm_semihosting(env);
+                return;
+            }
+        }
+        /* Fall through to prefetch abort.  */
+    case EXCP_PREFETCH_ABORT:
         new_mode = ARM_CPU_MODE_ABT;
         addr = 0x0c;
         mask = CPSR_A | CPSR_I;
@@ -354,6 +691,8 @@ void do_interrupt(CPUARMState *env)
     }
     switch_mode (env, new_mode);
     env->spsr = cpsr_read(env);
+    /* Clear IT bits.  */
+    env->condexec_bits = 0;
     /* Switch to the new mode, and switch to Arm mode.  */
     /* ??? Thumb interrupt handlers not implemented.  */
     env->uncached_cpsr = (env->uncached_cpsr & ~CPSR_M) | new_mode;
@@ -370,9 +709,16 @@ void do_interrupt(CPUARMState *env)
 static inline int check_ap(CPUState *env, int ap, int domain, int access_type,
                            int is_user)
 {
+  int prot_ro;
+
   if (domain == 3)
     return PAGE_READ | PAGE_WRITE;
 
+  if (access_type == 1)
+      prot_ro = 0;
+  else
+      prot_ro = PAGE_READ;
+
   switch (ap) {
   case 0:
       if (access_type == 1)
@@ -389,18 +735,24 @@ static inline int check_ap(CPUState *env, int ap, int domain, int access_type,
       return is_user ? 0 : PAGE_READ | PAGE_WRITE;
   case 2:
       if (is_user)
-          return (access_type == 1) ? 0 : PAGE_READ;
+          return prot_ro;
       else
           return PAGE_READ | PAGE_WRITE;
   case 3:
       return PAGE_READ | PAGE_WRITE;
+  case 4: case 7: /* Reserved.  */
+      return 0;
+  case 5:
+      return is_user ? 0 : prot_ro;
+  case 6:
+      return prot_ro;
   default:
       abort();
   }
 }
 
-static int get_phys_addr(CPUState *env, uint32_t address, int access_type,
-                         int is_user, uint32_t *phys_ptr, int *prot)
+static int get_phys_addr_v5(CPUState *env, uint32_t address, int access_type,
+			    int is_user, uint32_t *phys_ptr, int *prot)
 {
     int code;
     uint32_t table;
@@ -410,145 +762,259 @@ static int get_phys_addr(CPUState *env, uint32_t address, int access_type,
     int domain;
     uint32_t phys_addr;
 
-    /* Fast Context Switch Extension.  */
-    if (address < 0x02000000)
-        address += env->cp15.c13_fcse;
-
-    if ((env->cp15.c1_sys & 1) == 0) {
-        /* MMU/MPU disabled.  */
-        *phys_ptr = address;
-        *prot = PAGE_READ | PAGE_WRITE;
-    } else if (arm_feature(env, ARM_FEATURE_MPU)) {
-        int n;
-        uint32_t mask;
-        uint32_t base;
-
-        *phys_ptr = address;
-        for (n = 7; n >= 0; n--) {
-            base = env->cp15.c6_region[n];
-            if ((base & 1) == 0)
-                continue;
-            mask = 1 << ((base >> 1) & 0x1f);
-            /* Keep this shift separate from the above to avoid an
-               (undefined) << 32.  */
-            mask = (mask << 1) - 1;
-            if (((base ^ address) & ~mask) == 0)
-                break;
-        }
-        if (n < 0)
-            return 2;
-
-        if (access_type == 2) {
-            mask = env->cp15.c5_insn;
-        } else {
-            mask = env->cp15.c5_data;
-        }
-        mask = (mask >> (n * 4)) & 0xf;
-        switch (mask) {
-        case 0:
-            return 1;
-        case 1:
-            if (is_user)
-              return 1;
-            *prot = PAGE_READ | PAGE_WRITE;
-            break;
-        case 2:
-            *prot = PAGE_READ;
-            if (!is_user)
-                *prot |= PAGE_WRITE;
+    /* Pagetable walk.  */
+    /* Lookup l1 descriptor.  */
+    if (address & env->cp15.c2_mask)
+        table = env->cp15.c2_base1;
+    else
+        table = env->cp15.c2_base0;
+    table = (table & 0xffffc000) | ((address >> 18) & 0x3ffc);
+    desc = ldl_phys(table);
+    type = (desc & 3);
+    domain = (env->cp15.c3 >> ((desc >> 4) & 0x1e)) & 3;
+    if (type == 0) {
+        /* Secton translation fault.  */
+        code = 5;
+        goto do_fault;
+    }
+    if (domain == 0 || domain == 2) {
+        if (type == 2)
+            code = 9; /* Section domain fault.  */
+        else
+            code = 11; /* Page domain fault.  */
+        goto do_fault;
+    }
+    if (type == 2) {
+        /* 1Mb section.  */
+        phys_addr = (desc & 0xfff00000) | (address & 0x000fffff);
+        ap = (desc >> 10) & 3;
+        code = 13;
+    } else {
+        /* Lookup l2 entry.  */
+	if (type == 1) {
+	    /* Coarse pagetable.  */
+	    table = (desc & 0xfffffc00) | ((address >> 10) & 0x3fc);
+	} else {
+	    /* Fine pagetable.  */
+	    table = (desc & 0xfffff000) | ((address >> 8) & 0xffc);
+	}
+        desc = ldl_phys(table);
+        switch (desc & 3) {
+        case 0: /* Page translation fault.  */
+            code = 7;
+            goto do_fault;
+        case 1: /* 64k page.  */
+            phys_addr = (desc & 0xffff0000) | (address & 0xffff);
+            ap = (desc >> (4 + ((address >> 13) & 6))) & 3;
             break;
-        case 3:
-            *prot = PAGE_READ | PAGE_WRITE;
+        case 2: /* 4k page.  */
+            phys_addr = (desc & 0xfffff000) | (address & 0xfff);
+            ap = (desc >> (4 + ((address >> 13) & 6))) & 3;
             break;
-        case 5:
-            if (is_user)
-                return 1;
-            *prot = PAGE_READ;
-            break;
-        case 6:
-            *prot = PAGE_READ;
+        case 3: /* 1k page.  */
+	    if (type == 1) {
+		if (arm_feature(env, ARM_FEATURE_XSCALE)) {
+		    phys_addr = (desc & 0xfffff000) | (address & 0xfff);
+		} else {
+		    /* Page translation fault.  */
+		    code = 7;
+		    goto do_fault;
+		}
+	    } else {
+		phys_addr = (desc & 0xfffffc00) | (address & 0x3ff);
+	    }
+            ap = (desc >> 4) & 3;
             break;
         default:
-            /* Bad permission.  */
-            return 1;
+            /* Never happens, but compiler isn't smart enough to tell.  */
+            abort();
         }
+        code = 15;
+    }
+    *prot = check_ap(env, ap, domain, access_type, is_user);
+    if (!*prot) {
+        /* Access permission fault.  */
+        goto do_fault;
+    }
+    *phys_ptr = phys_addr;
+    return 0;
+do_fault:
+    return code | (domain << 4);
+}
+
+static int get_phys_addr_v6(CPUState *env, uint32_t address, int access_type,
+			    int is_user, uint32_t *phys_ptr, int *prot)
+{
+    int code;
+    uint32_t table;
+    uint32_t desc;
+    uint32_t xn;
+    int type;
+    int ap;
+    int domain;
+    uint32_t phys_addr;
+
+    /* Pagetable walk.  */
+    /* Lookup l1 descriptor.  */
+    if (address & env->cp15.c2_mask)
+        table = env->cp15.c2_base1;
+    else
+        table = env->cp15.c2_base0;
+    table = (table & 0xffffc000) | ((address >> 18) & 0x3ffc);
+    desc = ldl_phys(table);
+    type = (desc & 3);
+    if (type == 0) {
+        /* Secton translation fault.  */
+        code = 5;
+        domain = 0;
+        goto do_fault;
+    } else if (type == 2 && (desc & (1 << 18))) {
+        /* Supersection.  */
+        domain = 0;
     } else {
-        /* Pagetable walk.  */
-        /* Lookup l1 descriptor.  */
-        table = (env->cp15.c2_base & 0xffffc000) | ((address >> 18) & 0x3ffc);
-        desc = ldl_phys(table);
-        type = (desc & 3);
-        domain = (env->cp15.c3 >> ((desc >> 4) & 0x1e)) & 3;
-        if (type == 0) {
-            /* Secton translation fault.  */
-            code = 5;
-            goto do_fault;
-        }
-        if (domain == 0 || domain == 2) {
-            if (type == 2)
-                code = 9; /* Section domain fault.  */
-            else
-                code = 11; /* Page domain fault.  */
-            goto do_fault;
-        }
-        if (type == 2) {
-            /* 1Mb section.  */
-            phys_addr = (desc & 0xfff00000) | (address & 0x000fffff);
-            ap = (desc >> 10) & 3;
-            code = 13;
+        /* Section or page.  */
+        domain = (desc >> 4) & 0x1e;
+    }
+    domain = (env->cp15.c3 >> domain) & 3;
+    if (domain == 0 || domain == 2) {
+        if (type == 2)
+            code = 9; /* Section domain fault.  */
+        else
+            code = 11; /* Page domain fault.  */
+        goto do_fault;
+    }
+    if (type == 2) {
+        if (desc & (1 << 18)) {
+            /* Supersection.  */
+            phys_addr = (desc & 0xff000000) | (address & 0x00ffffff);
         } else {
-            /* Lookup l2 entry.  */
-            if (type == 1) {
-                /* Coarse pagetable.  */
-                table = (desc & 0xfffffc00) | ((address >> 10) & 0x3fc);
-            } else {
-                /* Fine pagetable.  */
-                table = (desc & 0xfffff000) | ((address >> 8) & 0xffc);
-            }
-            desc = ldl_phys(table);
-            switch (desc & 3) {
-            case 0: /* Page translation fault.  */
-                code = 7;
-                goto do_fault;
-            case 1: /* 64k page.  */
-                phys_addr = (desc & 0xffff0000) | (address & 0xffff);
-                ap = (desc >> (4 + ((address >> 13) & 6))) & 3;
-                break;
-            case 2: /* 4k page.  */
-                phys_addr = (desc & 0xfffff000) | (address & 0xfff);
-                ap = (desc >> (4 + ((address >> 13) & 6))) & 3;
-                break;
-            case 3: /* 1k page.  */
-                if (type == 1) {
-                    if (arm_feature(env, ARM_FEATURE_XSCALE))
-                        phys_addr = (desc & 0xfffff000) | (address & 0xfff);
-                    else {
-                        /* Page translation fault.  */
-                        code = 7;
-                        goto do_fault;
-                    }
-                } else
-                    phys_addr = (desc & 0xfffffc00) | (address & 0x3ff);
-                ap = (desc >> 4) & 3;
-                break;
-            default:
-                /* Never happens, but compiler isn't smart enough to tell.  */
-                abort();
-            }
-            code = 15;
+            /* Section.  */
+            phys_addr = (desc & 0xfff00000) | (address & 0x000fffff);
         }
-        *prot = check_ap(env, ap, domain, access_type, is_user);
-        if (!*prot) {
-            /* Access permission fault.  */
+        ap = ((desc >> 10) & 3) | ((desc >> 13) & 4);
+        xn = desc & (1 << 4);
+        code = 13;
+    } else {
+        /* Lookup l2 entry.  */
+        table = (desc & 0xfffffc00) | ((address >> 10) & 0x3fc);
+        desc = ldl_phys(table);
+        ap = ((desc >> 4) & 3) | ((desc >> 7) & 4);
+        switch (desc & 3) {
+        case 0: /* Page translation fault.  */
+            code = 7;
             goto do_fault;
+        case 1: /* 64k page.  */
+            phys_addr = (desc & 0xffff0000) | (address & 0xffff);
+            xn = desc & (1 << 15);
+            break;
+        case 2: case 3: /* 4k page.  */
+            phys_addr = (desc & 0xfffff000) | (address & 0xfff);
+            xn = desc & 1;
+            break;
+        default:
+            /* Never happens, but compiler isn't smart enough to tell.  */
+            abort();
         }
-        *phys_ptr = phys_addr;
+        code = 15;
+    }
+    if (xn && access_type == 2)
+        goto do_fault;
+
+    *prot = check_ap(env, ap, domain, access_type, is_user);
+    if (!*prot) {
+        /* Access permission fault.  */
+        goto do_fault;
     }
+    *phys_ptr = phys_addr;
     return 0;
 do_fault:
     return code | (domain << 4);
 }
 
+static int get_phys_addr_mpu(CPUState *env, uint32_t address, int access_type,
+			     int is_user, uint32_t *phys_ptr, int *prot)
+{
+    int n;
+    uint32_t mask;
+    uint32_t base;
+
+    *phys_ptr = address;
+    for (n = 7; n >= 0; n--) {
+	base = env->cp15.c6_region[n];
+	if ((base & 1) == 0)
+	    continue;
+	mask = 1 << ((base >> 1) & 0x1f);
+	/* Keep this shift separate from the above to avoid an
+	   (undefined) << 32.  */
+	mask = (mask << 1) - 1;
+	if (((base ^ address) & ~mask) == 0)
+	    break;
+    }
+    if (n < 0)
+	return 2;
+
+    if (access_type == 2) {
+	mask = env->cp15.c5_insn;
+    } else {
+	mask = env->cp15.c5_data;
+    }
+    mask = (mask >> (n * 4)) & 0xf;
+    switch (mask) {
+    case 0:
+	return 1;
+    case 1:
+	if (is_user)
+	  return 1;
+	*prot = PAGE_READ | PAGE_WRITE;
+	break;
+    case 2:
+	*prot = PAGE_READ;
+	if (!is_user)
+	    *prot |= PAGE_WRITE;
+	break;
+    case 3:
+	*prot = PAGE_READ | PAGE_WRITE;
+	break;
+    case 5:
+	if (is_user)
+	    return 1;
+	*prot = PAGE_READ;
+	break;
+    case 6:
+	*prot = PAGE_READ;
+	break;
+    default:
+	/* Bad permission.  */
+	return 1;
+    }
+    return 0;
+}
+
+static inline int get_phys_addr(CPUState *env, uint32_t address,
+                                int access_type, int is_user,
+                                uint32_t *phys_ptr, int *prot)
+{
+    /* Fast Context Switch Extension.  */
+    if (address < 0x02000000)
+        address += env->cp15.c13_fcse;
+
+    if ((env->cp15.c1_sys & 1) == 0) {
+        /* MMU/MPU disabled.  */
+        *phys_ptr = address;
+        *prot = PAGE_READ | PAGE_WRITE;
+        return 0;
+    } else if (arm_feature(env, ARM_FEATURE_MPU)) {
+	return get_phys_addr_mpu(env, address, access_type, is_user, phys_ptr,
+				 prot);
+    } else if (env->cp15.c1_sys & (1 << 23)) {
+        return get_phys_addr_v6(env, address, access_type, is_user, phys_ptr,
+                                prot);
+    } else {
+        return get_phys_addr_v5(env, address, access_type, is_user, phys_ptr,
+                                prot);
+    }
+}
+
 int cpu_arm_handle_mmu_fault (CPUState *env, target_ulong address,
                               int access_type, int mmu_idx, int is_softmmu)
 {
@@ -572,6 +1038,8 @@ int cpu_arm_handle_mmu_fault (CPUState *env, target_ulong address,
         env->exception_index = EXCP_PREFETCH_ABORT;
     } else {
         env->cp15.c5_data = ret;
+        if (access_type == 1 && arm_feature(env, ARM_FEATURE_V6))
+            env->cp15.c5_data |= (1 << 11);
         env->cp15.c6_data = address;
         env->exception_index = EXCP_DATA_ABORT;
     }
@@ -592,6 +1060,24 @@ target_phys_addr_t cpu_get_phys_page_debug(CPUState *env, target_ulong addr)
     return phys_addr;
 }
 
+/* Not really implemented.  Need to figure out a sane way of doing this.
+   Maybe add generic watchpoint support and use that.  */
+
+void helper_mark_exclusive(CPUState *env, uint32_t addr)
+{
+    env->mmon_addr = addr;
+}
+
+int helper_test_exclusive(CPUState *env, uint32_t addr)
+{
+    return (env->mmon_addr != addr);
+}
+
+void helper_clrex(CPUState *env)
+{
+    env->mmon_addr = -1;
+}
+
 void helper_set_cp(CPUState *env, uint32_t insn, uint32_t val)
 {
     int cp_num = (insn >> 8) & 0xf;
@@ -649,13 +1135,20 @@ static uint32_t extended_mpu_ap_bits(uint32_t val)
 
 void helper_set_cp15(CPUState *env, uint32_t insn, uint32_t val)
 {
-    uint32_t op2;
-    uint32_t crm;
+    int op1;
+    int op2;
+    int crm;
 
+    op1 = (insn >> 21) & 7;
     op2 = (insn >> 5) & 7;
     crm = insn & 0xf;
     switch ((insn >> 16) & 0xf) {
-    case 0: /* ID codes.  */
+    case 0:
+        if (((insn >> 21) & 7) == 2) {
+            /* ??? Select cache level.  Ignore.  */
+            return;
+        }
+        /* ID codes.  */
         if (arm_feature(env, ARM_FEATURE_XSCALE))
             break;
         if (arm_feature(env, ARM_FEATURE_OMAPCP))
@@ -672,12 +1165,13 @@ void helper_set_cp15(CPUState *env, uint32_t insn, uint32_t val)
             /* This may enable/disable the MMU, so do a TLB flush.  */
             tlb_flush(env, 1);
             break;
-        case 1:
+        case 1: /* Auxiliary cotrol register.  */
             if (arm_feature(env, ARM_FEATURE_XSCALE)) {
                 env->cp15.c1_xscaleauxcr = val;
                 break;
             }
-            goto bad_reg;
+            /* Not implemented.  */
+            break;
         case 2:
             if (arm_feature(env, ARM_FEATURE_XSCALE))
                 goto bad_reg;
@@ -702,7 +1196,19 @@ void helper_set_cp15(CPUState *env, uint32_t insn, uint32_t val)
                 goto bad_reg;
             }
         } else {
-            env->cp15.c2_base = val;
+	    switch (op2) {
+	    case 0:
+		env->cp15.c2_base0 = val;
+		break;
+	    case 1:
+		env->cp15.c2_base1 = val;
+		break;
+	    case 2:
+		env->cp15.c2_mask = ~(((uint32_t)0xffffffffu) >> val);
+		break;
+	    default:
+		goto bad_reg;
+	    }
         }
         break;
     case 3: /* MMU Domain access control / MPU write buffer control.  */
@@ -751,7 +1257,8 @@ void helper_set_cp15(CPUState *env, uint32_t insn, uint32_t val)
             case 0:
                 env->cp15.c6_data = val;
                 break;
-            case 1:
+            case 1: /* ??? This is WFAR on armv6 */
+            case 2:
                 env->cp15.c6_insn = val;
                 break;
             default:
@@ -763,6 +1270,7 @@ void helper_set_cp15(CPUState *env, uint32_t insn, uint32_t val)
         env->cp15.c15_i_max = 0x000;
         env->cp15.c15_i_min = 0xff0;
         /* No cache, so nothing to do.  */
+        /* ??? MPCore has VA to PA translation functions.  */
         break;
     case 8: /* MMU TLB control.  */
         switch (op2) {
@@ -783,6 +1291,13 @@ void helper_set_cp15(CPUState *env, uint32_t insn, uint32_t val)
             tlb_flush(env, 1);
 #endif
             break;
+        case 2: /* Invalidate on ASID.  */
+            tlb_flush(env, val == 0);
+            break;
+        case 3: /* Invalidate single entry on MVA.  */
+            /* ??? This is like case 1, but ignores ASID.  */
+            tlb_flush(env, 1);
+            break;
         default:
             goto bad_reg;
         }
@@ -792,17 +1307,26 @@ void helper_set_cp15(CPUState *env, uint32_t insn, uint32_t val)
             break;
         switch (crm) {
         case 0: /* Cache lockdown.  */
-            switch (op2) {
-            case 0:
-                env->cp15.c9_data = val;
-                break;
-            case 1:
-                env->cp15.c9_insn = val;
-                break;
-            default:
-                goto bad_reg;
-            }
-            break;
+	    switch (op1) {
+	    case 0: /* L1 cache.  */
+		switch (op2) {
+		case 0:
+		    env->cp15.c9_data = val;
+		    break;
+		case 1:
+		    env->cp15.c9_insn = val;
+		    break;
+		default:
+		    goto bad_reg;
+		}
+		break;
+	    case 1: /* L2 cache.  */
+		/* Ignore writes to L2 lockdown/auxiliary registers.  */
+		break;
+	    default:
+		goto bad_reg;
+	    }
+	    break;
         case 1: /* TCM memory region registers.  */
             /* Not implemented.  */
             goto bad_reg;
@@ -832,6 +1356,15 @@ void helper_set_cp15(CPUState *env, uint32_t insn, uint32_t val)
               tlb_flush(env, 0);
             env->cp15.c13_context = val;
             break;
+        case 2:
+            env->cp15.c13_tls1 = val;
+            break;
+        case 3:
+            env->cp15.c13_tls2 = val;
+            break;
+        case 4:
+            env->cp15.c13_tls3 = val;
+            break;
         default:
             goto bad_reg;
         }
@@ -880,27 +1413,64 @@ void helper_set_cp15(CPUState *env, uint32_t insn, uint32_t val)
     return;
 bad_reg:
     /* ??? For debugging only.  Should raise illegal instruction exception.  */
-    cpu_abort(env, "Unimplemented cp15 register write\n");
+    cpu_abort(env, "Unimplemented cp15 register write (c%d, c%d, {%d, %d})\n",
+              (insn >> 16) & 0xf, crm, op1, op2);
 }
 
 uint32_t helper_get_cp15(CPUState *env, uint32_t insn)
 {
-    uint32_t op2;
-    uint32_t crm;
+    int op1;
+    int op2;
+    int crm;
 
+    op1 = (insn >> 21) & 7;
     op2 = (insn >> 5) & 7;
     crm = insn & 0xf;
     switch ((insn >> 16) & 0xf) {
     case 0: /* ID codes.  */
-        switch (op2) {
-        default: /* Device ID.  */
-            return env->cp15.c0_cpuid;
-        case 1: /* Cache Type.  */
-            return env->cp15.c0_cachetype;
-        case 2: /* TCM status.  */
+        switch (op1) {
+        case 0:
+            switch (crm) {
+            case 0:
+                switch (op2) {
+                case 0: /* Device ID.  */
+                    return env->cp15.c0_cpuid;
+                case 1: /* Cache Type.  */
+		    return env->cp15.c0_cachetype;
+                case 2: /* TCM status.  */
+                    return 0;
+                case 3: /* TLB type register.  */
+                    return 0; /* No lockable TLB entries.  */
+                case 5: /* CPU ID */
+                    return env->cpu_index;
+                default:
+                    goto bad_reg;
+                }
+            case 1:
+                if (!arm_feature(env, ARM_FEATURE_V6))
+                    goto bad_reg;
+                return env->cp15.c0_c1[op2];
+            case 2:
+                if (!arm_feature(env, ARM_FEATURE_V6))
+                    goto bad_reg;
+                return env->cp15.c0_c2[op2];
+            case 3: case 4: case 5: case 6: case 7:
+                return 0;
+            default:
+                goto bad_reg;
+            }
+        case 1:
+            /* These registers aren't documented on arm11 cores.  However
+               Linux looks at them anyway.  */
+            if (!arm_feature(env, ARM_FEATURE_V6))
+                goto bad_reg;
+            if (crm != 0)
+                goto bad_reg;
             if (arm_feature(env, ARM_FEATURE_XSCALE))
                 goto bad_reg;
             return 0;
+        default:
+            goto bad_reg;
         }
     case 1: /* System configuration.  */
         if (arm_feature(env, ARM_FEATURE_OMAPCP))
@@ -909,11 +1479,22 @@ uint32_t helper_get_cp15(CPUState *env, uint32_t insn)
         case 0: /* Control register.  */
             return env->cp15.c1_sys;
         case 1: /* Auxiliary control register.  */
-            if (arm_feature(env, ARM_FEATURE_AUXCR))
-                return 1;
             if (arm_feature(env, ARM_FEATURE_XSCALE))
                 return env->cp15.c1_xscaleauxcr;
-            goto bad_reg;
+            if (!arm_feature(env, ARM_FEATURE_AUXCR))
+                goto bad_reg;
+            switch (ARM_CPUID(env)) {
+            case ARM_CPUID_ARM1026:
+                return 1;
+            case ARM_CPUID_ARM1136:
+                return 7;
+            case ARM_CPUID_ARM11MPCORE:
+                return 1;
+            case ARM_CPUID_CORTEXA8:
+                return 0;
+            default:
+                goto bad_reg;
+            }
         case 2: /* Coprocessor access register.  */
             if (arm_feature(env, ARM_FEATURE_XSCALE))
                 goto bad_reg;
@@ -934,8 +1515,27 @@ uint32_t helper_get_cp15(CPUState *env, uint32_t insn)
                 goto bad_reg;
             }
         } else {
-            return env->cp15.c2_base;
-        }
+	    switch (op2) {
+	    case 0:
+		return env->cp15.c2_base0;
+	    case 1:
+		return env->cp15.c2_base1;
+	    case 2:
+		{
+		    int n;
+		    uint32_t mask;
+		    n = 0;
+		    mask = env->cp15.c2_mask;
+		    while (mask) {
+			n++;
+			mask <<= 1;
+		    }
+		    return n;
+		}
+	    default:
+		goto bad_reg;
+	    }
+	}
     case 3: /* MMU Domain access control / MPU write buffer control.  */
         return env->cp15.c3;
     case 4: /* Reserved.  */
@@ -963,26 +1563,37 @@ uint32_t helper_get_cp15(CPUState *env, uint32_t insn)
         default:
             goto bad_reg;
         }
-    case 6: /* MMU Fault address / MPU base/size.  */
+    case 6: /* MMU Fault address.  */
         if (arm_feature(env, ARM_FEATURE_MPU)) {
-            int n;
-            n = (insn & 0xf);
-            if (n >= 8)
+            if (crm >= 8)
                 goto bad_reg;
-            return env->cp15.c6_region[n];
+            return env->cp15.c6_region[crm];
         } else {
             if (arm_feature(env, ARM_FEATURE_OMAPCP))
                 op2 = 0;
-            switch (op2) {
-            case 0:
-                return env->cp15.c6_data;
-            case 1:
-                /* Arm9 doesn't have an IFAR, but implementing it anyway
-                   shouldn't do any harm.  */
-                return env->cp15.c6_insn;
-            default:
-                goto bad_reg;
-            }
+	    switch (op2) {
+	    case 0:
+		return env->cp15.c6_data;
+	    case 1:
+		if (arm_feature(env, ARM_FEATURE_V6)) {
+		    /* Watchpoint Fault Adrress.  */
+		    return 0; /* Not implemented.  */
+		} else {
+		    /* Instruction Fault Adrress.  */
+		    /* Arm9 doesn't have an IFAR, but implementing it anyway
+		       shouldn't do any harm.  */
+		    return env->cp15.c6_insn;
+		}
+	    case 2:
+		if (arm_feature(env, ARM_FEATURE_V6)) {
+		    /* Instruction Fault Adrress.  */
+		    return env->cp15.c6_insn;
+		} else {
+		    goto bad_reg;
+		}
+	    default:
+		goto bad_reg;
+	    }
         }
     case 7: /* Cache control.  */
         /* ??? This is for test, clean and invaidate operations that set the
@@ -993,13 +1604,23 @@ uint32_t helper_get_cp15(CPUState *env, uint32_t insn)
     case 8: /* MMU TLB control.  */
         goto bad_reg;
     case 9: /* Cache lockdown.  */
-        if (arm_feature(env, ARM_FEATURE_OMAPCP))
+        switch (op1) {
+        case 0: /* L1 cache.  */
+	    if (arm_feature(env, ARM_FEATURE_OMAPCP))
+		return 0;
+            switch (op2) {
+            case 0:
+                return env->cp15.c9_data;
+            case 1:
+                return env->cp15.c9_insn;
+            default:
+                goto bad_reg;
+            }
+        case 1: /* L2 cache */
+            if (crm != 0)
+                goto bad_reg;
+            /* L2 Lockdown and Auxiliary control.  */
             return 0;
-        switch (op2) {
-        case 0:
-            return env->cp15.c9_data;
-        case 1:
-            return env->cp15.c9_insn;
         default:
             goto bad_reg;
         }
@@ -1015,6 +1636,12 @@ uint32_t helper_get_cp15(CPUState *env, uint32_t insn)
             return env->cp15.c13_fcse;
         case 1:
             return env->cp15.c13_context;
+        case 2:
+            return env->cp15.c13_tls1;
+        case 3:
+            return env->cp15.c13_tls2;
+        case 4:
+            return env->cp15.c13_tls3;
         default:
             goto bad_reg;
         }
@@ -1048,10 +1675,125 @@ uint32_t helper_get_cp15(CPUState *env, uint32_t insn)
     }
 bad_reg:
     /* ??? For debugging only.  Should raise illegal instruction exception.  */
-    cpu_abort(env, "Unimplemented cp15 register read\n");
+    cpu_abort(env, "Unimplemented cp15 register read (c%d, c%d, {%d, %d})\n",
+              (insn >> 16) & 0xf, crm, op1, op2);
     return 0;
 }
 
+void helper_set_r13_banked(CPUState *env, int mode, uint32_t val)
+{
+    env->banked_r13[bank_number(mode)] = val;
+}
+
+uint32_t helper_get_r13_banked(CPUState *env, int mode)
+{
+    return env->banked_r13[bank_number(mode)];
+}
+
+uint32_t helper_v7m_mrs(CPUState *env, int reg)
+{
+    switch (reg) {
+    case 0: /* APSR */
+        return xpsr_read(env) & 0xf8000000;
+    case 1: /* IAPSR */
+        return xpsr_read(env) & 0xf80001ff;
+    case 2: /* EAPSR */
+        return xpsr_read(env) & 0xff00fc00;
+    case 3: /* xPSR */
+        return xpsr_read(env) & 0xff00fdff;
+    case 5: /* IPSR */
+        return xpsr_read(env) & 0x000001ff;
+    case 6: /* EPSR */
+        return xpsr_read(env) & 0x0700fc00;
+    case 7: /* IEPSR */
+        return xpsr_read(env) & 0x0700edff;
+    case 8: /* MSP */
+        return env->v7m.current_sp ? env->v7m.other_sp : env->regs[13];
+    case 9: /* PSP */
+        return env->v7m.current_sp ? env->regs[13] : env->v7m.other_sp;
+    case 16: /* PRIMASK */
+        return (env->uncached_cpsr & CPSR_I) != 0;
+    case 17: /* FAULTMASK */
+        return (env->uncached_cpsr & CPSR_F) != 0;
+    case 18: /* BASEPRI */
+    case 19: /* BASEPRI_MAX */
+        return env->v7m.basepri;
+    case 20: /* CONTROL */
+        return env->v7m.control;
+    default:
+        /* ??? For debugging only.  */
+        cpu_abort(env, "Unimplemented system register read (%d)\n", reg);
+        return 0;
+    }
+}
+
+void helper_v7m_msr(CPUState *env, int reg, uint32_t val)
+{
+    switch (reg) {
+    case 0: /* APSR */
+        xpsr_write(env, val, 0xf8000000);
+        break;
+    case 1: /* IAPSR */
+        xpsr_write(env, val, 0xf8000000);
+        break;
+    case 2: /* EAPSR */
+        xpsr_write(env, val, 0xfe00fc00);
+        break;
+    case 3: /* xPSR */
+        xpsr_write(env, val, 0xfe00fc00);
+        break;
+    case 5: /* IPSR */
+        /* IPSR bits are readonly.  */
+        break;
+    case 6: /* EPSR */
+        xpsr_write(env, val, 0x0600fc00);
+        break;
+    case 7: /* IEPSR */
+        xpsr_write(env, val, 0x0600fc00);
+        break;
+    case 8: /* MSP */
+        if (env->v7m.current_sp)
+            env->v7m.other_sp = val;
+        else
+            env->regs[13] = val;
+        break;
+    case 9: /* PSP */
+        if (env->v7m.current_sp)
+            env->regs[13] = val;
+        else
+            env->v7m.other_sp = val;
+        break;
+    case 16: /* PRIMASK */
+        if (val & 1)
+            env->uncached_cpsr |= CPSR_I;
+        else
+            env->uncached_cpsr &= ~CPSR_I;
+        break;
+    case 17: /* FAULTMASK */
+        if (val & 1)
+            env->uncached_cpsr |= CPSR_F;
+        else
+            env->uncached_cpsr &= ~CPSR_F;
+        break;
+    case 18: /* BASEPRI */
+        env->v7m.basepri = val & 0xff;
+        break;
+    case 19: /* BASEPRI_MAX */
+        val &= 0xff;
+        if (val != 0 && (val < env->v7m.basepri || env->v7m.basepri == 0))
+            env->v7m.basepri = val;
+        break;
+    case 20: /* CONTROL */
+        env->v7m.control = val & 3;
+        switch_v7m_sp(env, (val & 2) != 0);
+        break;
+    default:
+        /* ??? For debugging only.  */
+        cpu_abort(env, "Unimplemented system register write (%d)\n", reg);
+        return;
+    }
+}
+
 void cpu_arm_set_cp_io(CPUARMState *env, int cpnum,
                 ARMReadCPFunc *cp_read, ARMWriteCPFunc *cp_write,
                 void *opaque)
diff --git a/target-arm/op.c b/target-arm/op.c
index e8a536c..216944a 100644
--- a/target-arm/op.c
+++ b/target-arm/op.c
@@ -2,7 +2,7 @@
  *  ARM micro operations
  *
  *  Copyright (c) 2003 Fabrice Bellard
- *  Copyright (c) 2005 CodeSourcery, LLC
+ *  Copyright (c) 2005-2007 CodeSourcery, LLC
  *
  * This library is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public
@@ -101,11 +101,6 @@ void OPPROTO op_movl_T0_im(void)
     T0 = PARAM1;
 }
 
-void OPPROTO op_movl_T0_T1(void)
-{
-    T0 = T1;
-}
-
 void OPPROTO op_movl_T1_im(void)
 {
     T1 = PARAM1;
@@ -236,6 +231,11 @@ void OPPROTO op_bicl_T0_T1(void)
     T0 &= ~T1;
 }
 
+void OPPROTO op_notl_T0(void)
+{
+    T0 = ~T0;
+}
+
 void OPPROTO op_notl_T1(void)
 {
     T1 = ~T1;
@@ -351,6 +351,19 @@ void OPPROTO op_test_le(void)
     FORCE_RET();
 }
 
+void OPPROTO op_test_T0(void)
+{
+    if (T0)
+        GOTO_LABEL_PARAM(1);
+    FORCE_RET();
+}
+void OPPROTO op_testn_T0(void)
+{
+    if (!T0)
+        GOTO_LABEL_PARAM(1);
+    FORCE_RET();
+}
+
 void OPPROTO op_goto_tb0(void)
 {
     GOTO_TB(op_goto_tb0, PARAM1, 0);
@@ -368,7 +381,8 @@ void OPPROTO op_exit_tb(void)
 
 void OPPROTO op_movl_T0_cpsr(void)
 {
-    T0 = cpsr_read(env);
+    /* Execution state bits always read as zero.  */
+    T0 = cpsr_read(env) & ~CPSR_EXEC;
     FORCE_RET();
 }
 
@@ -438,6 +452,28 @@ void OPPROTO op_addq_lo_T0_T1(void)
     T0 = res;
 }
 
+/* Dual 16-bit accumulate.  */
+void OPPROTO op_addq_T0_T1_dual(void)
+{
+  uint64_t res;
+  res = ((uint64_t)(env->regs[PARAM2]) << 32) | (env->regs[PARAM1]);
+  res += (int32_t)T0;
+  res += (int32_t)T1;
+  env->regs[PARAM1] = (uint32_t)res;
+  env->regs[PARAM2] = res >> 32;
+}
+
+/* Dual 16-bit subtract accumulate.  */
+void OPPROTO op_subq_T0_T1_dual(void)
+{
+  uint64_t res;
+  res = ((uint64_t)(env->regs[PARAM2]) << 32) | (env->regs[PARAM1]);
+  res += (int32_t)T0;
+  res -= (int32_t)T1;
+  env->regs[PARAM1] = (uint32_t)res;
+  env->regs[PARAM2] = res >> 32;
+}
+
 void OPPROTO op_logicq_cc(void)
 {
     env->NZF = (T1 & 0x80000000) | ((T0 | T1) != 0);
@@ -455,8 +491,21 @@ void OPPROTO op_logicq_cc(void)
 #include "op_mem.h"
 #endif
 
+void OPPROTO op_clrex(void)
+{
+    cpu_lock();
+    helper_clrex(env);
+    cpu_unlock();
+}
+
 /* shifts */
 
+/* Used by NEON.  */
+void OPPROTO op_shll_T0_im(void)
+{
+    T1 = T1 << PARAM1;
+}
+
 /* T1 based */
 
 void OPPROTO op_shll_T1_im(void)
@@ -813,8 +862,39 @@ void OPPROTO op_double_T1_saturate(void)
   FORCE_RET();
 }
 
-/* thumb shift by immediate */
-void OPPROTO op_shll_T0_im_thumb(void)
+/* Unsigned saturating arithmetic for NEON.  */
+void OPPROTO op_addl_T0_T1_usaturate(void)
+{
+  uint32_t res;
+
+  res = T0 + T1;
+  if (res < T0) {
+      env->QF = 1;
+      T0 = 0xffffffff;
+  } else {
+      T0 = res;
+  }
+
+  FORCE_RET();
+}
+
+void OPPROTO op_subl_T0_T1_usaturate(void)
+{
+  uint32_t res;
+
+  res = T0 - T1;
+  if (res > T0) {
+      env->QF = 1;
+      T0 = 0;
+  } else {
+      T0 = res;
+  }
+
+  FORCE_RET();
+}
+
+/* Thumb shift by immediate */
+void OPPROTO op_shll_T0_im_thumb_cc(void)
 {
     int shift;
     shift = PARAM1;
@@ -826,7 +906,13 @@ void OPPROTO op_shll_T0_im_thumb(void)
     FORCE_RET();
 }
 
-void OPPROTO op_shrl_T0_im_thumb(void)
+void OPPROTO op_shll_T0_im_thumb(void)
+{
+    T0 = T0 << PARAM1;
+    FORCE_RET();
+}
+
+void OPPROTO op_shrl_T0_im_thumb_cc(void)
 {
     int shift;
 
@@ -842,7 +928,20 @@ void OPPROTO op_shrl_T0_im_thumb(void)
     FORCE_RET();
 }
 
-void OPPROTO op_sarl_T0_im_thumb(void)
+void OPPROTO op_shrl_T0_im_thumb(void)
+{
+    int shift;
+
+    shift = PARAM1;
+    if (shift == 0) {
+	T0 = 0;
+    } else {
+	T0 = T0 >> shift;
+    }
+    FORCE_RET();
+}
+
+void OPPROTO op_sarl_T0_im_thumb_cc(void)
 {
     int shift;
 
@@ -858,6 +957,19 @@ void OPPROTO op_sarl_T0_im_thumb(void)
     FORCE_RET();
 }
 
+void OPPROTO op_sarl_T0_im_thumb(void)
+{
+    int shift;
+
+    shift = PARAM1;
+    if (shift == 0) {
+	env->CF = T0 & 1;
+    } else {
+	T0 = ((int32_t)T0) >> shift;
+    }
+    FORCE_RET();
+}
+
 /* exceptions */
 
 void OPPROTO op_swi(void)
@@ -891,6 +1003,12 @@ void OPPROTO op_bkpt(void)
     cpu_loop_exit();
 }
 
+void OPPROTO op_exception_exit(void)
+{
+    env->exception_index = EXCP_EXCEPTION_EXIT;
+    cpu_loop_exit();
+}
+
 /* VFP support.  We follow the convention used for VFP instrunctions:
    Single precition routines have a "s" suffix, double precision a
    "d" suffix.  */
@@ -982,6 +1100,28 @@ static inline uint32_t vfp_stoi(float32 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)
 {
@@ -1056,6 +1196,32 @@ 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)
 {
@@ -1142,6 +1308,20 @@ void OPPROTO op_vfp_mdrr(void)
     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);
+}
+
 /* Copy the most significant bit of T0 to all bits of T1.  */
 void OPPROTO op_signbit_T1_T0(void)
 {
@@ -1204,9 +1384,9 @@ void OPPROTO op_movl_user_T0(void)
     FORCE_RET();
 }
 
-void OPPROTO op_movl_T2_T0(void)
+void OPPROTO op_movl_T0_T1(void)
 {
-    T2 = T0;
+    T0 = T1;
 }
 
 void OPPROTO op_movl_T0_T2(void)
@@ -1214,5 +1394,530 @@ void OPPROTO op_movl_T0_T2(void)
     T0 = T2;
 }
 
+void OPPROTO op_movl_T1_T0(void)
+{
+    T1 = T0;
+}
+
+void OPPROTO op_movl_T1_T2(void)
+{
+    T1 = T2;
+}
+
+void OPPROTO op_movl_T2_T0(void)
+{
+    T2 = T0;
+}
+
+/* ARMv6 Media instructions.  */
+
+/* Note that signed overflow is undefined in C.  The following routines are
+   careful to use unsigned types where modulo arithmetic is required.
+   Failure to do so _will_ break on newer gcc.  */
+
+/* Signed saturating arithmetic.  */
+
+/* Perform 16-bit signed satruating addition.  */
+static inline uint16_t add16_sat(uint16_t a, uint16_t b)
+{
+    uint16_t res;
+
+    res = a + b;
+    if (((res ^ a) & 0x8000) && !((a ^ b) & 0x8000)) {
+        if (a & 0x8000)
+            res = 0x8000;
+        else
+            res = 0x7fff;
+    }
+    return res;
+}
+
+/* Perform 8-bit signed satruating addition.  */
+static inline uint8_t add8_sat(uint8_t a, uint8_t b)
+{
+    uint8_t res;
+
+    res = a + b;
+    if (((res ^ a) & 0x80) && !((a ^ b) & 0x80)) {
+        if (a & 0x80)
+            res = 0x80;
+        else
+            res = 0x7f;
+    }
+    return res;
+}
+
+/* Perform 16-bit signed satruating subtraction.  */
+static inline uint16_t sub16_sat(uint16_t a, uint16_t b)
+{
+    uint16_t res;
+
+    res = a - b;
+    if (((res ^ a) & 0x8000) && ((a ^ b) & 0x8000)) {
+        if (a & 0x8000)
+            res = 0x8000;
+        else
+            res = 0x7fff;
+    }
+    return res;
+}
+
+/* Perform 8-bit signed satruating subtraction.  */
+static inline uint8_t sub8_sat(uint8_t a, uint8_t b)
+{
+    uint8_t res;
+
+    res = a - b;
+    if (((res ^ a) & 0x80) && ((a ^ b) & 0x80)) {
+        if (a & 0x80)
+            res = 0x80;
+        else
+            res = 0x7f;
+    }
+    return res;
+}
+
+#define ADD16(a, b, n) RESULT(add16_sat(a, b), n, 16);
+#define SUB16(a, b, n) RESULT(sub16_sat(a, b), n, 16);
+#define ADD8(a, b, n)  RESULT(add8_sat(a, b), n, 8);
+#define SUB8(a, b, n)  RESULT(sub8_sat(a, b), n, 8);
+#define PFX q
+
+#include "op_addsub.h"
+
+/* Unsigned saturating arithmetic.  */
+static inline uint16_t add16_usat(uint16_t a, uint8_t b)
+{
+    uint16_t res;
+    res = a + b;
+    if (res < a)
+        res = 0xffff;
+    return res;
+}
+
+static inline uint16_t sub16_usat(uint16_t a, uint8_t b)
+{
+    if (a < b)
+        return a - b;
+    else
+        return 0;
+}
+
+static inline uint8_t add8_usat(uint8_t a, uint8_t b)
+{
+    uint8_t res;
+    res = a + b;
+    if (res < a)
+        res = 0xff;
+    return res;
+}
+
+static inline uint8_t sub8_usat(uint8_t a, uint8_t b)
+{
+    if (a < b)
+        return a - b;
+    else
+        return 0;
+}
+
+#define ADD16(a, b, n) RESULT(add16_usat(a, b), n, 16);
+#define SUB16(a, b, n) RESULT(sub16_usat(a, b), n, 16);
+#define ADD8(a, b, n)  RESULT(add8_usat(a, b), n, 8);
+#define SUB8(a, b, n)  RESULT(sub8_usat(a, b), n, 8);
+#define PFX uq
+
+#include "op_addsub.h"
+
+/* Signed modulo arithmetic.  */
+#define SARITH16(a, b, n, op) do { \
+    int32_t sum; \
+    sum = (int16_t)((uint16_t)(a) op (uint16_t)(b)); \
+    RESULT(sum, n, 16); \
+    if (sum >= 0) \
+        ge |= 3 << (n * 2); \
+    } while(0)
+
+#define SARITH8(a, b, n, op) do { \
+    int32_t sum; \
+    sum = (int8_t)((uint8_t)(a) op (uint8_t)(b)); \
+    RESULT(sum, n, 8); \
+    if (sum >= 0) \
+        ge |= 1 << n; \
+    } while(0)
+
+
+#define ADD16(a, b, n) SARITH16(a, b, n, +)
+#define SUB16(a, b, n) SARITH16(a, b, n, -)
+#define ADD8(a, b, n)  SARITH8(a, b, n, +)
+#define SUB8(a, b, n)  SARITH8(a, b, n, -)
+#define PFX s
+#define ARITH_GE
+
+#include "op_addsub.h"
+
+/* Unsigned modulo arithmetic.  */
+#define ADD16(a, b, n) do { \
+    uint32_t sum; \
+    sum = (uint32_t)(uint16_t)(a) + (uint32_t)(uint16_t)(b); \
+    RESULT(sum, n, 16); \
+    if ((sum >> 16) == 0) \
+        ge |= 3 << (n * 2); \
+    } while(0)
+
+#define ADD8(a, b, n) do { \
+    uint32_t sum; \
+    sum = (uint32_t)(uint8_t)(a) + (uint32_t)(uint8_t)(b); \
+    RESULT(sum, n, 8); \
+    if ((sum >> 8) == 0) \
+        ge |= 3 << (n * 2); \
+    } while(0)
+
+#define SUB16(a, b, n) do { \
+    uint32_t sum; \
+    sum = (uint32_t)(uint16_t)(a) - (uint32_t)(uint16_t)(b); \
+    RESULT(sum, n, 16); \
+    if ((sum >> 16) == 0) \
+        ge |= 3 << (n * 2); \
+    } while(0)
+
+#define SUB8(a, b, n) do { \
+    uint32_t sum; \
+    sum = (uint32_t)(uint8_t)(a) - (uint32_t)(uint8_t)(b); \
+    RESULT(sum, n, 8); \
+    if ((sum >> 8) == 0) \
+        ge |= 3 << (n * 2); \
+    } while(0)
+
+#define PFX u
+#define ARITH_GE
+
+#include "op_addsub.h"
+
+/* Halved signed arithmetic.  */
+#define ADD16(a, b, n) \
+  RESULT(((int32_t)(int16_t)(a) + (int32_t)(int16_t)(b)) >> 1, n, 16)
+#define SUB16(a, b, n) \
+  RESULT(((int32_t)(int16_t)(a) - (int32_t)(int16_t)(b)) >> 1, n, 16)
+#define ADD8(a, b, n) \
+  RESULT(((int32_t)(int8_t)(a) + (int32_t)(int8_t)(b)) >> 1, n, 8)
+#define SUB8(a, b, n) \
+  RESULT(((int32_t)(int8_t)(a) - (int32_t)(int8_t)(b)) >> 1, n, 8)
+#define PFX sh
+
+#include "op_addsub.h"
+
+/* Halved unsigned arithmetic.  */
+#define ADD16(a, b, n) \
+  RESULT(((uint32_t)(uint16_t)(a) + (uint32_t)(uint16_t)(b)) >> 1, n, 16)
+#define SUB16(a, b, n) \
+  RESULT(((uint32_t)(uint16_t)(a) - (uint32_t)(uint16_t)(b)) >> 1, n, 16)
+#define ADD8(a, b, n) \
+  RESULT(((uint32_t)(uint8_t)(a) + (uint32_t)(uint8_t)(b)) >> 1, n, 8)
+#define SUB8(a, b, n) \
+  RESULT(((uint32_t)(uint8_t)(a) - (uint32_t)(uint8_t)(b)) >> 1, n, 8)
+#define PFX uh
+
+#include "op_addsub.h"
+
+void OPPROTO op_pkhtb_T0_T1(void)
+{
+    T0 = (T0 & 0xffff0000) | (T1 & 0xffff);
+}
+
+void OPPROTO op_pkhbt_T0_T1(void)
+{
+    T0 = (T0 & 0xffff) | (T1 & 0xffff0000);
+}
+void OPPROTO op_rev_T0(void)
+{
+    T0 =  ((T0 & 0xff000000) >> 24)
+        | ((T0 & 0x00ff0000) >> 8)
+        | ((T0 & 0x0000ff00) << 8)
+        | ((T0 & 0x000000ff) << 24);
+}
+
+void OPPROTO op_revh_T0(void)
+{
+    T0 = (T0 >> 16) | (T0 << 16);
+}
+
+void OPPROTO op_rev16_T0(void)
+{
+    T0 =  ((T0 & 0xff000000) >> 8)
+        | ((T0 & 0x00ff0000) << 8)
+        | ((T0 & 0x0000ff00) >> 8)
+        | ((T0 & 0x000000ff) << 8);
+}
+
+void OPPROTO op_revsh_T0(void)
+{
+    T0 = (int16_t)(  ((T0 & 0x0000ff00) >> 8)
+                   | ((T0 & 0x000000ff) << 8));
+}
+
+void OPPROTO op_rbit_T0(void)
+{
+    T0 =  ((T0 & 0xff000000) >> 24)
+        | ((T0 & 0x00ff0000) >> 8)
+        | ((T0 & 0x0000ff00) << 8)
+        | ((T0 & 0x000000ff) << 24);
+    T0 =  ((T0 & 0xf0f0f0f0) >> 4)
+        | ((T0 & 0x0f0f0f0f) << 4);
+    T0 =  ((T0 & 0x88888888) >> 3)
+        | ((T0 & 0x44444444) >> 1)
+        | ((T0 & 0x22222222) << 1)
+        | ((T0 & 0x11111111) << 3);
+}
+
+/* Swap low and high halfwords.  */
+void OPPROTO op_swap_half_T1(void)
+{
+    T1 = (T1 >> 16) | (T1 << 16);
+    FORCE_RET();
+}
+
+/* Dual 16-bit signed multiply.  */
+void OPPROTO op_mul_dual_T0_T1(void)
+{
+    int32_t low;
+    int32_t high;
+    low = (int32_t)(int16_t)T0 * (int32_t)(int16_t)T1;
+    high = (((int32_t)T0) >> 16) * (((int32_t)T1) >> 16);
+    T0 = low;
+    T1 = high;
+}
+
+void OPPROTO op_sel_T0_T1(void)
+{
+    uint32_t mask;
+    uint32_t flags;
+
+    flags = env->GE;
+    mask = 0;
+    if (flags & 1)
+        mask |= 0xff;
+    if (flags & 2)
+        mask |= 0xff00;
+    if (flags & 4)
+        mask |= 0xff0000;
+    if (flags & 8)
+        mask |= 0xff000000;
+    T0 = (T0 & mask) | (T1 & ~mask);
+    FORCE_RET();
+}
+
+void OPPROTO op_roundqd_T0_T1(void)
+{
+    T0 = T1 + ((uint32_t)T0 >> 31);
+}
+
+/* Signed saturation.  */
+static inline uint32_t do_ssat(int32_t val, int shift)
+{
+    int32_t top;
+    uint32_t mask;
+
+    shift = PARAM1;
+    top = val >> shift;
+    mask = (1u << shift) - 1;
+    if (top > 0) {
+        env->QF = 1;
+        return mask;
+    } else if (top < -1) {
+        env->QF = 1;
+        return ~mask;
+    }
+    return val;
+}
+
+/* Unsigned saturation.  */
+static inline uint32_t do_usat(int32_t val, int shift)
+{
+    uint32_t max;
+
+    shift = PARAM1;
+    max = (1u << shift) - 1;
+    if (val < 0) {
+        env->QF = 1;
+        return 0;
+    } else if (val > max) {
+        env->QF = 1;
+        return max;
+    }
+    return val;
+}
+
+/* Signed saturate.  */
+void OPPROTO op_ssat_T1(void)
+{
+    T0 = do_ssat(T0, PARAM1);
+    FORCE_RET();
+}
+
+/* Dual halfword signed saturate.  */
+void OPPROTO op_ssat16_T1(void)
+{
+    uint32_t res;
+
+    res = (uint16_t)do_ssat((int16_t)T0, PARAM1);
+    res |= do_ssat(((int32_t)T0) >> 16, PARAM1) << 16;
+    T0 = res;
+    FORCE_RET();
+}
+
+/* Unsigned saturate.  */
+void OPPROTO op_usat_T1(void)
+{
+    T0 = do_usat(T0, PARAM1);
+    FORCE_RET();
+}
+
+/* Dual halfword unsigned saturate.  */
+void OPPROTO op_usat16_T1(void)
+{
+    uint32_t res;
+
+    res = (uint16_t)do_usat((int16_t)T0, PARAM1);
+    res |= do_usat(((int32_t)T0) >> 16, PARAM1) << 16;
+    T0 = res;
+    FORCE_RET();
+}
+
+/* Dual 16-bit add.  */
+void OPPROTO op_add16_T1_T2(void)
+{
+    uint32_t mask;
+    mask = (T0 & T1) & 0x8000;
+    T0 ^= ~0x8000;
+    T1 ^= ~0x8000;
+    T0 = (T0 + T1) ^ mask;
+}
+
+static inline uint8_t do_usad(uint8_t a, uint8_t b)
+{
+    if (a > b)
+        return a - b;
+    else
+        return b - a;
+}
+
+/* Unsigned sum of absolute byte differences.  */
+void OPPROTO op_usad8_T0_T1(void)
+{
+    uint32_t sum;
+    sum = do_usad(T0, T1);
+    sum += do_usad(T0 >> 8, T1 >> 8);
+    sum += do_usad(T0 >> 16, T1 >>16);
+    sum += do_usad(T0 >> 24, T1 >> 24);
+    T0 = sum;
+}
+
+/* Thumb-2 instructions.  */
+
+/* Insert T1 into T0.  Result goes in T1.  */
+void OPPROTO op_bfi_T1_T0(void)
+{
+    int shift = PARAM1;
+    uint32_t mask = PARAM2;
+    uint32_t bits;
+
+    bits = (T1 << shift) & mask;
+    T1 = (T0 & ~mask) | bits;
+}
+
+/* Unsigned bitfield extract.  */
+void OPPROTO op_ubfx_T1(void)
+{
+    uint32_t shift = PARAM1;
+    uint32_t mask = PARAM2;
+
+    T1 >>= shift;
+    T1 &= mask;
+}
+
+/* Signed bitfield extract.  */
+void OPPROTO op_sbfx_T1(void)
+{
+    uint32_t shift = PARAM1;
+    uint32_t width = PARAM2;
+    int32_t val;
+
+    val = T1 << (32 - (shift + width));
+    T1 = val >> (32 - width);
+}
+
+void OPPROTO op_movtop_T0_im(void)
+{
+    T0 = (T0 & 0xffff) | PARAM1;
+}
+
+/* Used by table branch instructions.  */
+void OPPROTO op_jmp_T0_im(void)
+{
+    env->regs[15] = PARAM1 + (T0 << 1);
+}
+
+void OPPROTO op_set_condexec(void)
+{
+    env->condexec_bits = PARAM1;
+}
+
+void OPPROTO op_sdivl_T0_T1(void)
+{
+  int32_t num;
+  int32_t den;
+  num = T0;
+  den = T1;
+  if (den == 0)
+    T0 = 0;
+  else
+    T0 = num / den;
+  FORCE_RET();
+}
+
+void OPPROTO op_udivl_T0_T1(void)
+{
+  uint32_t num;
+  uint32_t den;
+  num = T0;
+  den = T1;
+  if (den == 0)
+    T0 = 0;
+  else
+    T0 = num / den;
+  FORCE_RET();
+}
+
+void OPPROTO op_movl_T1_r13_banked(void)
+{
+    T1 = helper_get_r13_banked(env, PARAM1);
+}
+
+void OPPROTO op_movl_r13_T1_banked(void)
+{
+    helper_set_r13_banked(env, PARAM1, T1);
+}
+
+void OPPROTO op_v7m_mrs_T0(void)
+{
+    T0 = helper_v7m_mrs(env, PARAM1);
+}
+
+void OPPROTO op_v7m_msr_T0(void)
+{
+    helper_v7m_msr(env, PARAM1, T0);
+}
+
+void OPPROTO op_movl_T0_sp(void)
+{
+    if (PARAM1 == env->v7m.current_sp)
+        T0 = env->regs[13];
+    else
+        T0 = env->v7m.other_sp;
+    FORCE_RET();
+}
+
+#include "op_neon.h"
+
 /* iwMMXt support */
 #include "op_iwmmxt.c"
diff --git a/target-arm/op_addsub.h b/target-arm/op_addsub.h
new file mode 100644
index 0000000..d15360d
--- /dev/null
+++ b/target-arm/op_addsub.h
@@ -0,0 +1,106 @@
+/*
+ * ARMv6 integer SIMD operations.
+ *
+ * Copyright (c) 2007 CodeSourcery.
+ * Written by Paul Brook
+ *
+ * This code is licenced under the GPL.
+ */
+
+#ifdef ARITH_GE
+#define DECLARE_GE uint32_t ge = 0
+#define SET_GE env->GE = ge
+#else
+#define DECLARE_GE do{}while(0)
+#define SET_GE do{}while(0)
+#endif
+
+#define RESULT(val, n, width) \
+    res |= ((uint32_t)(glue(glue(uint,width),_t))(val)) << (n * width)
+
+void OPPROTO glue(glue(op_,PFX),add16_T0_T1)(void)
+{
+    uint32_t res = 0;
+    DECLARE_GE;
+
+    ADD16(T0, T1, 0);
+    ADD16(T0 >> 16, T1 >> 16, 1);
+    SET_GE;
+    T0 = res;
+    FORCE_RET();
+}
+
+void OPPROTO glue(glue(op_,PFX),add8_T0_T1)(void)
+{
+    uint32_t res = 0;
+    DECLARE_GE;
+
+    ADD8(T0, T1, 0);
+    ADD8(T0 >> 8, T1 >> 8, 1);
+    ADD8(T0 >> 16, T1 >> 16, 2);
+    ADD8(T0 >> 24, T1 >> 24, 3);
+    SET_GE;
+    T0 = res;
+    FORCE_RET();
+}
+
+void OPPROTO glue(glue(op_,PFX),sub16_T0_T1)(void)
+{
+    uint32_t res = 0;
+    DECLARE_GE;
+
+    SUB16(T0, T1, 0);
+    SUB16(T0 >> 16, T1 >> 16, 1);
+    SET_GE;
+    T0 = res;
+    FORCE_RET();
+}
+
+void OPPROTO glue(glue(op_,PFX),sub8_T0_T1)(void)
+{
+    uint32_t res = 0;
+    DECLARE_GE;
+
+    SUB8(T0, T1, 0);
+    SUB8(T0 >> 8, T1 >> 8, 1);
+    SUB8(T0 >> 16, T1 >> 16, 2);
+    SUB8(T0 >> 24, T1 >> 24, 3);
+    SET_GE;
+    T0 = res;
+    FORCE_RET();
+}
+
+void OPPROTO glue(glue(op_,PFX),subaddx_T0_T1)(void)
+{
+    uint32_t res = 0;
+    DECLARE_GE;
+
+    ADD16(T0, T1, 0);
+    SUB16(T0 >> 16, T1 >> 16, 1);
+    SET_GE;
+    T0 = res;
+    FORCE_RET();
+}
+
+void OPPROTO glue(glue(op_,PFX),addsubx_T0_T1)(void)
+{
+    uint32_t res = 0;
+    DECLARE_GE;
+
+    SUB16(T0, T1, 0);
+    ADD16(T0 >> 16, T1 >> 16, 1);
+    SET_GE;
+    T0 = res;
+    FORCE_RET();
+}
+
+#undef DECLARE_GE
+#undef SET_GE
+#undef RESULT
+
+#undef ARITH_GE
+#undef PFX
+#undef ADD16
+#undef SUB16
+#undef ADD8
+#undef SUB8
diff --git a/target-arm/op_helper.c b/target-arm/op_helper.c
index 6e14a4d..a9bd95b 100644
--- a/target-arm/op_helper.c
+++ b/target-arm/op_helper.c
@@ -1,7 +1,7 @@
 /*
  *  ARM helper routines
  *
- *  Copyright (c) 2005 CodeSourcery, LLC
+ *  Copyright (c) 2005-2007 CodeSourcery, LLC
  *
  * This library is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public
@@ -175,6 +175,81 @@ void do_vfp_get_fpscr(void)
     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;
+    uint32_t mask;
+    uint32_t tmp;
+    int index;
+    int shift;
+    uint64_t *table;
+    table = (uint64_t *)&env->vfp.regs[rn];
+    val = 0;
+    mask = 0;
+    for (shift = 0; shift < 32; shift += 8) {
+        index = (T1 >> shift) & 0xff;
+        if (index <= maxindex) {
+            tmp = (table[index >> 3] >> (index & 7)) & 0xff;
+            val |= tmp << shift;
+        } else {
+            val |= T0 & (0xff << shift);
+        }
+    }
+    T0 = val;
+}
+
 #if !defined(CONFIG_USER_ONLY)
 
 #define MMUSUFFIX _mmu
@@ -227,5 +302,4 @@ void tlb_fill (target_ulong addr, int is_write, int mmu_idx, void *retaddr)
     }
     env = saved_env;
 }
-
 #endif
diff --git a/target-arm/op_mem.h b/target-arm/op_mem.h
index 6bccb06..b12b63c 100644
--- a/target-arm/op_mem.h
+++ b/target-arm/op_mem.h
@@ -1,5 +1,6 @@
 /* ARM memory operations.  */
 
+void helper_ld(uint32_t);
 /* Load from address T1 into T0.  */
 #define MEM_LD_OP(name) \
 void OPPROTO glue(op_ld##name,MEMSUFFIX)(void) \
@@ -49,6 +50,64 @@ MEM_SWP_OP(l, l)
 
 #undef MEM_SWP_OP
 
+/* Load-locked, store exclusive.  */
+#define EXCLUSIVE_OP(suffix, ldsuffix) \
+void OPPROTO glue(op_ld##suffix##ex,MEMSUFFIX)(void) \
+{ \
+    cpu_lock(); \
+    helper_mark_exclusive(env, T1); \
+    T0 = glue(ld##ldsuffix,MEMSUFFIX)(T1); \
+    cpu_unlock(); \
+    FORCE_RET(); \
+} \
+ \
+void OPPROTO glue(op_st##suffix##ex,MEMSUFFIX)(void) \
+{ \
+    int failed; \
+    cpu_lock(); \
+    failed = helper_test_exclusive(env, T1); \
+    /* ??? Is it safe to hold the cpu lock over a store?  */ \
+    if (!failed) { \
+        glue(st##suffix,MEMSUFFIX)(T1, T0); \
+    } \
+    T0 = failed; \
+    cpu_unlock(); \
+    FORCE_RET(); \
+}
+
+EXCLUSIVE_OP(b, ub)
+EXCLUSIVE_OP(w, uw)
+EXCLUSIVE_OP(l, l)
+
+#undef EXCLUSIVE_OP
+
+/* Load exclusive T0:T1 from address T1.  */
+void OPPROTO glue(op_ldqex,MEMSUFFIX)(void)
+{
+    cpu_lock();
+    helper_mark_exclusive(env, T1);
+    T0 = glue(ldl,MEMSUFFIX)(T1);
+    T1 = glue(ldl,MEMSUFFIX)((T1 + 4));
+    cpu_unlock();
+    FORCE_RET();
+}
+
+/* Store exclusive T0:T2 to address T1.  */
+void OPPROTO glue(op_stqex,MEMSUFFIX)(void)
+{
+    int failed;
+    cpu_lock();
+    failed = helper_test_exclusive(env, T1);
+    /* ??? Is it safe to hold the cpu lock over a store?  */
+    if (!failed) {
+        glue(stl,MEMSUFFIX)(T1, T0);
+        glue(stl,MEMSUFFIX)((T1 + 4), T2);
+    }
+    T0 = failed;
+    cpu_unlock();
+    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) \
diff --git a/target-arm/op_neon.h b/target-arm/op_neon.h
new file mode 100644
index 0000000..232375e
--- /dev/null
+++ b/target-arm/op_neon.h
@@ -0,0 +1,1754 @@
+/*
+ * 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.  */
+
+/* ??? 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)
+
+NEON_OP(getreg_T0)
+{
+    T0 = *(uint32_t *)((char *) env + PARAM1);
+}
+
+NEON_OP(getreg_T1)
+{
+    T1 = *(uint32_t *)((char *) env + PARAM1);
+}
+
+NEON_OP(getreg_T2)
+{
+    T2 = *(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;
+}
+
+NEON_OP(setreg_T2)
+{
+    *(uint32_t *)((char *) env + PARAM1) = T2;
+}
+
+#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();
+}
+
+/* ??? bsl, bif and bit are all the same op, just with the oparands in a
+   differnet order.  It's currently easier to have 3 differnt ops than
+   rearange the operands.  */
+
+/* Bitwise Select.  */
+NEON_OP(bsl)
+{
+    T0 = (T0 & T2) | (T1 & ~T2);
+}
+
+/* Bitwise Insert If True.  */
+NEON_OP(bit)
+{
+    T0 = (T0 & T1) | (T2 & ~T1);
+}
+
+/* Bitwise Insert If False.  */
+NEON_OP(bif)
+{
+    T0 = (T2 & T1) | (T0 & ~T1);
+}
+
+#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 = T2;
+    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 = T2;
+    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 = T2;
+    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 = T2;
+    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 = T2;
+    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 = T2;
+    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
+
+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) \
+{ \
+    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();
+}
+
+/* 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)
+{
+    helper_neon_tbl(PARAM1, PARAM2);
+}
+
+NEON_OP(dup_u8)
+{
+    T0 = (T0 >> PARAM1) & 0xff;
+    T0 |= T0 << 8;
+    T0 |= T0 << 16;
+    FORCE_RET();
+}
+
+/* Helpers for element load/store.  */
+NEON_OP(insert_elt)
+{
+    int shift = PARAM1;
+    uint32_t mask = PARAM2;
+    T2 = (T2 & mask) | (T0 << shift);
+    FORCE_RET();
+}
+
+NEON_OP(extract_elt)
+{
+    int shift = PARAM1;
+    uint32_t mask = PARAM2;
+    T0 = (T2 & mask) >> shift;
+    FORCE_RET();
+}
diff --git a/target-arm/translate.c b/target-arm/translate.c
index bb01f2f..9366bf1 100644
--- a/target-arm/translate.c
+++ b/target-arm/translate.c
@@ -2,7 +2,7 @@
  *  ARM translation
  *
  *  Copyright (c) 2003 Fabrice Bellard
- *  Copyright (c) 2005 CodeSourcery, LLC
+ *  Copyright (c) 2005-2007 CodeSourcery
  *  Copyright (c) 2007 OpenedHand, Ltd.
  *
  * This library is free software; you can redistribute it and/or
@@ -29,9 +29,11 @@
 #include "exec-all.h"
 #include "disas.h"
 
-#define ENABLE_ARCH_5J  0
-#define ENABLE_ARCH_6   1
-#define ENABLE_ARCH_6T2 1
+#define ENABLE_ARCH_5J    0
+#define ENABLE_ARCH_6     arm_feature(env, ARM_FEATURE_V6)
+#define ENABLE_ARCH_6K   arm_feature(env, ARM_FEATURE_V6K)
+#define ENABLE_ARCH_6T2   arm_feature(env, ARM_FEATURE_THUMB2)
+#define ENABLE_ARCH_7     arm_feature(env, ARM_FEATURE_V7)
 
 #define ARCH(x) if (!ENABLE_ARCH_##x) goto illegal_op;
 
@@ -43,6 +45,9 @@ typedef struct DisasContext {
     int condjmp;
     /* The label that will be jumped to when the instruction is skipped.  */
     int condlabel;
+    /* Thumb-2 condtional execution bits.  */
+    int condexec_mask;
+    int condexec_cond;
     struct TranslationBlock *tb;
     int singlestep_enabled;
     int thumb;
@@ -58,7 +63,10 @@ typedef struct DisasContext {
 #define IS_USER(s) (s->user)
 #endif
 
-#define DISAS_JUMP_NEXT 4
+/* These instructions trap after executing, so defer them until after the
+   conditional executions state has been updated.  */
+#define DISAS_WFI 4
+#define DISAS_SWI 5
 
 #ifdef USE_DIRECT_JUMP
 #define TBPARAM(x)
@@ -81,6 +89,51 @@ enum {
 
 #include "gen-op.h"
 
+#define PAS_OP(pfx) {  \
+    gen_op_ ## pfx ## add16_T0_T1, \
+    gen_op_ ## pfx ## addsubx_T0_T1, \
+    gen_op_ ## pfx ## subaddx_T0_T1, \
+    gen_op_ ## pfx ## sub16_T0_T1, \
+    gen_op_ ## pfx ## add8_T0_T1, \
+    NULL, \
+    NULL, \
+    gen_op_ ## pfx ## sub8_T0_T1 }
+
+static GenOpFunc *gen_arm_parallel_addsub[8][8] = {
+    {},
+    PAS_OP(s),
+    PAS_OP(q),
+    PAS_OP(sh),
+    {},
+    PAS_OP(u),
+    PAS_OP(uq),
+    PAS_OP(uh),
+};
+#undef PAS_OP
+
+/* For unknown reasons Arm and Thumb-2 use arbitrarily diffenet encodings.  */
+#define PAS_OP(pfx) {  \
+    gen_op_ ## pfx ## add8_T0_T1, \
+    gen_op_ ## pfx ## add16_T0_T1, \
+    gen_op_ ## pfx ## addsubx_T0_T1, \
+    NULL, \
+    gen_op_ ## pfx ## sub8_T0_T1, \
+    gen_op_ ## pfx ## sub16_T0_T1, \
+    gen_op_ ## pfx ## subaddx_T0_T1, \
+    NULL }
+
+static GenOpFunc *gen_thumb2_parallel_addsub[8][8] = {
+    PAS_OP(s),
+    PAS_OP(q),
+    PAS_OP(sh),
+    {},
+    PAS_OP(u),
+    PAS_OP(uq),
+    PAS_OP(uh),
+    {}
+};
+#undef PAS_OP
+
 static GenOpFunc1 *gen_test_cc[14] = {
     gen_op_test_eq,
     gen_op_test_ne,
@@ -275,6 +328,12 @@ static GenOpFunc1 *gen_op_movl_TN_im[3] = {
     gen_op_movl_T2_im,
 };
 
+static GenOpFunc1 *gen_shift_T0_im_thumb_cc[3] = {
+    gen_op_shll_T0_im_thumb_cc,
+    gen_op_shrl_T0_im_thumb_cc,
+    gen_op_sarl_T0_im_thumb_cc,
+};
+
 static GenOpFunc1 *gen_shift_T0_im_thumb[3] = {
     gen_op_shll_T0_im_thumb,
     gen_op_shrl_T0_im_thumb,
@@ -421,6 +480,15 @@ static inline void gen_vfp_##name(int dp) \
         gen_op_vfp_##name##s();           \
 }
 
+#define VFP_OP1(name)                               \
+static inline void gen_vfp_##name(int dp, int arg)  \
+{                                                   \
+    if (dp)                                         \
+        gen_op_vfp_##name##d(arg);                  \
+    else                                            \
+        gen_op_vfp_##name##s(arg);                  \
+}
+
 VFP_OP(add)
 VFP_OP(sub)
 VFP_OP(mul)
@@ -437,9 +505,25 @@ 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)
 
 #undef VFP_OP
 
+static inline void gen_vfp_fconst(int dp, uint32_t val)
+{
+    if (dp)
+        gen_op_vfp_fconstd(val);
+    else
+        gen_op_vfp_fconsts(val);
+}
+
 static inline void gen_vfp_ld(DisasContext *s, int dp)
 {
     if (dp)
@@ -469,6 +553,20 @@ vfp_reg_offset (int dp, int reg)
           + offsetof(CPU_DoubleU, l.lower);
     }
 }
+
+/* Return the offset of a 32-bit piece of a NEON register.
+   zero is the least significant end of the register.  */
+static inline long
+neon_reg_offset (int reg, int n)
+{
+    int sreg;
+    sreg = reg * 2 + n;
+    return vfp_reg_offset(0, sreg);
+}
+
+#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))
+
 static inline void gen_mov_F0_vreg(int dp, int reg)
 {
     if (dp)
@@ -1582,14 +1680,49 @@ static int disas_cp_insn(CPUState *env, DisasContext *s, uint32_t insn)
     return 0;
 }
 
+static int cp15_user_ok(uint32_t insn)
+{
+    int cpn = (insn >> 16) & 0xf;
+    int cpm = insn & 0xf;
+    int op = ((insn >> 5) & 7) | ((insn >> 18) & 0x38);
+
+    if (cpn == 13 && cpm == 0) {
+        /* TLS register.  */
+        if (op == 2 || (op == 3 && (insn & ARM_CP_RW_BIT)))
+            return 1;
+    }
+    if (cpn == 7) {
+        /* ISB, DSB, DMB.  */
+        if ((cpm == 5 && op == 4)
+                || (cpm == 10 && (op == 4 || op == 5)))
+            return 1;
+    }
+    return 0;
+}
+
 /* Disassemble system coprocessor (cp15) instruction.  Return nonzero if
    instruction is not defined.  */
 static int disas_cp15_insn(CPUState *env, DisasContext *s, uint32_t insn)
 {
     uint32_t rd;
 
-    /* ??? Some cp15 registers are accessible from userspace.  */
-    if (IS_USER(s)) {
+    /* M profile cores use memory mapped registers instead of cp15.  */
+    if (arm_feature(env, ARM_FEATURE_M))
+	return 1;
+
+    if ((insn & (1 << 25)) == 0) {
+        if (insn & (1 << 20)) {
+            /* mrrc */
+            return 1;
+        }
+        /* mcrr.  Used for block cache operations, so implement as no-op.  */
+        return 0;
+    }
+    if ((insn & (1 << 4)) == 0) {
+        /* cdp */
+        return 1;
+    }
+    if (IS_USER(s) && !cp15_user_ok(insn)) {
         return 1;
     }
     if ((insn & 0x0fff0fff) == 0x0e070f90
@@ -1597,8 +1730,7 @@ static int disas_cp15_insn(CPUState *env, DisasContext *s, uint32_t insn)
         /* Wait for interrupt.  */
         gen_op_movl_T0_im((long)s->pc);
         gen_op_movl_reg_TN[0][15]();
-        gen_op_wfi();
-        s->is_jmp = DISAS_JUMP;
+        s->is_jmp = DISAS_WFI;
         return 0;
     }
     rd = (insn >> 12) & 0xf;
@@ -1620,6 +1752,32 @@ static int disas_cp15_insn(CPUState *env, DisasContext *s, uint32_t insn)
     return 0;
 }
 
+#define VFP_REG_SHR(x, n) (((n) > 0) ? (x) >> (n) : (x) << -(n))
+#define VFP_SREG(insn, bigbit, smallbit) \
+  ((VFP_REG_SHR(insn, bigbit - 1) & 0x1e) | (((insn) >> (smallbit)) & 1))
+#define VFP_DREG(reg, insn, bigbit, smallbit) do { \
+    if (arm_feature(env, ARM_FEATURE_VFP3)) { \
+        reg = (((insn) >> (bigbit)) & 0x0f) \
+              | (((insn) >> ((smallbit) - 4)) & 0x10); \
+    } else { \
+        if (insn & (1 << (smallbit))) \
+            return 1; \
+        reg = ((insn) >> (bigbit)) & 0x0f; \
+    }} while (0)
+
+#define VFP_SREG_D(insn) VFP_SREG(insn, 12, 22)
+#define VFP_DREG_D(reg, insn) VFP_DREG(reg, insn, 12, 22)
+#define VFP_SREG_N(insn) VFP_SREG(insn, 16,  7)
+#define VFP_DREG_N(reg, insn) VFP_DREG(reg, insn, 16,  7)
+#define VFP_SREG_M(insn) VFP_SREG(insn,  0,  5)
+#define VFP_DREG_M(reg, insn) VFP_DREG(reg, insn,  0,  5)
+
+static inline int
+vfp_enabled(CPUState * env)
+{
+    return ((env->vfp.xregs[ARM_VFP_FPEXC] & (1 << 30)) != 0);
+}
+
 /* Disassemble a VFP instruction.  Returns nonzero if an error occured
    (ie. an undefined instruction).  */
 static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
@@ -1630,12 +1788,13 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
     if (!arm_feature(env, ARM_FEATURE_VFP))
         return 1;
 
-    if ((env->vfp.xregs[ARM_VFP_FPEXC] & (1 << 30)) == 0) {
-        /* VFP disabled.  Only allow fmxr/fmrx to/from fpexc and fpsid.  */
+    if (!vfp_enabled(env)) {
+        /* VFP disabled.  Only allow fmxr/fmrx to/from some control regs.  */
         if ((insn & 0x0fe00fff) != 0x0ee00a10)
             return 1;
         rn = (insn >> 16) & 0xf;
-        if (rn != 0 && rn != 8)
+        if (rn != ARM_VFP_FPSID && rn != ARM_VFP_FPEXC
+            && rn != ARM_VFP_MVFR1 && rn != ARM_VFP_MVFR0)
             return 1;
     }
     dp = ((insn & 0xf00) == 0xb00);
@@ -1643,44 +1802,129 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
     case 0xe:
         if (insn & (1 << 4)) {
             /* single register transfer */
-            if ((insn & 0x6f) != 0x00)
-                return 1;
             rd = (insn >> 12) & 0xf;
             if (dp) {
-                if (insn & 0x80)
+                int size;
+                int pass;
+
+                VFP_DREG_N(rn, insn);
+                if (insn & 0xf)
                     return 1;
-                rn = (insn >> 16) & 0xf;
-                /* Get the existing value even for arm->vfp moves because
-                   we only set half the register.  */
-                gen_mov_F0_vreg(1, rn);
-                gen_op_vfp_mrrd();
+                if (insn & 0x00c00060
+                    && !arm_feature(env, ARM_FEATURE_NEON))
+                    return 1;
+
+                pass = (insn >> 21) & 1;
+                if (insn & (1 << 22)) {
+                    size = 0;
+                    offset = ((insn >> 5) & 3) * 8;
+                } else if (insn & (1 << 5)) {
+                    size = 1;
+                    offset = (insn & (1 << 6)) ? 16 : 0;
+                } else {
+                    size = 2;
+                    offset = 0;
+                }
                 if (insn & ARM_CP_RW_BIT) {
                     /* vfp->arm */
-                    if (insn & (1 << 21))
-                        gen_movl_reg_T1(s, rd);
-                    else
-                        gen_movl_reg_T0(s, rd);
+                    switch (size) {
+                    case 0:
+                        NEON_GET_REG(T1, rn, pass);
+                        if (offset)
+                            gen_op_shrl_T1_im(offset);
+                        if (insn & (1 << 23))
+                            gen_op_uxtb_T1();
+                        else
+                            gen_op_sxtb_T1();
+                        break;
+                    case 1:
+                        NEON_GET_REG(T1, rn, pass);
+                        if (insn & (1 << 23)) {
+                            if (offset) {
+                                gen_op_shrl_T1_im(16);
+                            } else {
+                                gen_op_uxth_T1();
+                            }
+                        } else {
+                            if (offset) {
+                                gen_op_sarl_T1_im(16);
+                            } else {
+                                gen_op_sxth_T1();
+                            }
+                        }
+                        break;
+                    case 2:
+                        NEON_GET_REG(T1, rn, pass);
+                        break;
+                    }
+                    gen_movl_reg_T1(s, rd);
                 } else {
                     /* arm->vfp */
-                    if (insn & (1 << 21))
-                        gen_movl_T1_reg(s, rd);
-                    else
-                        gen_movl_T0_reg(s, rd);
-                    gen_op_vfp_mdrr();
-                    gen_mov_vreg_F0(dp, rn);
+                    gen_movl_T0_reg(s, rd);
+                    if (insn & (1 << 23)) {
+                        /* VDUP */
+                        if (size == 0) {
+                            gen_op_neon_dup_u8(0);
+                        } else if (size == 1) {
+                            gen_op_neon_dup_low16();
+                        }
+                        NEON_SET_REG(T0, rn, 0);
+                        NEON_SET_REG(T0, rn, 1);
+                    } else {
+                        /* VMOV */
+                        switch (size) {
+                        case 0:
+                            NEON_GET_REG(T2, rn, pass);
+                            gen_op_movl_T1_im(0xff);
+                            gen_op_andl_T0_T1();
+                            gen_op_neon_insert_elt(offset, ~(0xff << offset));
+                            NEON_SET_REG(T2, rn, pass);
+                            break;
+                        case 1:
+                            NEON_GET_REG(T2, rn, pass);
+                            gen_op_movl_T1_im(0xffff);
+                            gen_op_andl_T0_T1();
+                            bank_mask = offset ? 0xffff : 0xffff0000;
+                            gen_op_neon_insert_elt(offset, bank_mask);
+                            NEON_SET_REG(T2, rn, pass);
+                            break;
+                        case 2:
+                            NEON_SET_REG(T0, rn, pass);
+                            break;
+                        }
+                    }
                 }
-            } else {
-                rn = ((insn >> 15) & 0x1e) | ((insn >> 7) & 1);
+            } else { /* !dp */
+                if ((insn & 0x6f) != 0x00)
+                    return 1;
+                rn = VFP_SREG_N(insn);
                 if (insn & ARM_CP_RW_BIT) {
                     /* vfp->arm */
                     if (insn & (1 << 21)) {
                         /* system register */
                         rn >>= 1;
+
                         switch (rn) {
                         case ARM_VFP_FPSID:
+                            /* VFP2 allows access for 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);
+                            break;
                         case ARM_VFP_FPEXC:
+                            if (IS_USER(s))
+                                return 1;
+                            gen_op_vfp_movl_T0_xreg(rn);
+                            break;
                         case ARM_VFP_FPINST:
                         case ARM_VFP_FPINST2:
+                            /* Not present in VFP3.  */
+                            if (IS_USER(s)
+                                || arm_feature(env, ARM_FEATURE_VFP3))
+                                return 1;
                             gen_op_vfp_movl_T0_xreg(rn);
                             break;
                         case ARM_VFP_FPSCR:
@@ -1689,6 +1933,13 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
 			    else
 				gen_op_vfp_movl_T0_fpscr();
                             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);
+                            break;
                         default:
                             return 1;
                         }
@@ -1709,6 +1960,8 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                         /* system register */
                         switch (rn) {
                         case ARM_VFP_FPSID:
+                        case ARM_VFP_MVFR0:
+                        case ARM_VFP_MVFR1:
                             /* Writes are ignored.  */
                             break;
                         case ARM_VFP_FPSCR:
@@ -1716,6 +1969,8 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                             gen_lookup_tb(s);
                             break;
                         case ARM_VFP_FPEXC:
+                            if (IS_USER(s))
+                                return 1;
                             gen_op_vfp_movl_xreg_T0(rn);
                             gen_lookup_tb(s);
                             break;
@@ -1742,38 +1997,31 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                     rn = ((insn >> 15) & 0x1e) | ((insn >> 7) & 1);
                 } else {
                     /* rn is register number */
-                    if (insn & (1 << 7))
-                        return 1;
-                    rn = (insn >> 16) & 0xf;
+                    VFP_DREG_N(rn, insn);
                 }
 
                 if (op == 15 && (rn == 15 || rn > 17)) {
                     /* Integer or single precision destination.  */
-                    rd = ((insn >> 11) & 0x1e) | ((insn >> 22) & 1);
+                    rd = VFP_SREG_D(insn);
                 } else {
-                    if (insn & (1 << 22))
-                        return 1;
-                    rd = (insn >> 12) & 0xf;
+                    VFP_DREG_D(rd, insn);
                 }
 
                 if (op == 15 && (rn == 16 || rn == 17)) {
                     /* Integer source.  */
                     rm = ((insn << 1) & 0x1e) | ((insn >> 5) & 1);
                 } else {
-                    if (insn & (1 << 5))
-                        return 1;
-                    rm = insn & 0xf;
+                    VFP_DREG_M(rm, insn);
                 }
             } else {
-                rn = ((insn >> 15) & 0x1e) | ((insn >> 7) & 1);
+                rn = VFP_SREG_N(insn);
                 if (op == 15 && rn == 15) {
                     /* Double precision destination.  */
-                    if (insn & (1 << 22))
-                        return 1;
-                    rd = (insn >> 12) & 0xf;
-                } else
-                    rd = ((insn >> 11) & 0x1e) | ((insn >> 22) & 1);
-                rm = ((insn << 1) & 0x1e) | ((insn >> 5) & 1);
+                    VFP_DREG_D(rd, insn);
+                } else {
+                    rd = VFP_SREG_D(insn);
+                }
+                rm = VFP_SREG_M(insn);
             }
 
             veclen = env->vfp.vec_len;
@@ -1831,9 +2079,17 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                     gen_mov_F0_vreg(dp, rd);
                     gen_vfp_F1_ld0(dp);
                     break;
+                case 20:
+                case 21:
+                case 22:
+                case 23:
+                    /* Source and destination the same.  */
+                    gen_mov_F0_vreg(dp, rd);
+                    break;
                 default:
                     /* One source operand.  */
                     gen_mov_F0_vreg(dp, rm);
+                    break;
                 }
             } else {
                 /* Two source operands.  */
@@ -1882,6 +2138,27 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                 case 8: /* div: fn / fm */
                     gen_vfp_div(dp);
                     break;
+                case 14: /* fconst */
+                    if (!arm_feature(env, ARM_FEATURE_VFP3))
+                      return 1;
+
+                    n = (insn << 12) & 0x80000000;
+                    i = ((insn >> 12) & 0x70) | (insn & 0xf);
+                    if (dp) {
+                        if (i & 0x40)
+                            i |= 0x3f80;
+                        else
+                            i |= 0x4000;
+                        n |= i << 16;
+                    } else {
+                        if (i & 0x40)
+                            i |= 0x780;
+                        else
+                            i |= 0x800;
+                        n |= i << 19;
+                    }
+                    gen_vfp_fconst(dp, n);
+                    break;
                 case 15: /* extension space */
                     switch (rn) {
                     case 0: /* cpy */
@@ -1921,6 +2198,26 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                     case 17: /* fsito */
                         gen_vfp_sito(dp);
                         break;
+                    case 20: /* fshto */
+                        if (!arm_feature(env, ARM_FEATURE_VFP3))
+                          return 1;
+                        gen_vfp_shto(dp, rm);
+                        break;
+                    case 21: /* fslto */
+                        if (!arm_feature(env, ARM_FEATURE_VFP3))
+                          return 1;
+                        gen_vfp_slto(dp, rm);
+                        break;
+                    case 22: /* fuhto */
+                        if (!arm_feature(env, ARM_FEATURE_VFP3))
+                          return 1;
+                        gen_vfp_uhto(dp, rm);
+                        break;
+                    case 23: /* fulto */
+                        if (!arm_feature(env, ARM_FEATURE_VFP3))
+                          return 1;
+                        gen_vfp_ulto(dp, rm);
+                        break;
                     case 24: /* ftoui */
                         gen_vfp_toui(dp);
                         break;
@@ -1933,6 +2230,26 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                     case 27: /* ftosiz */
                         gen_vfp_tosiz(dp);
                         break;
+                    case 28: /* ftosh */
+                        if (!arm_feature(env, ARM_FEATURE_VFP3))
+                          return 1;
+                        gen_vfp_tosh(dp, rm);
+                        break;
+                    case 29: /* ftosl */
+                        if (!arm_feature(env, ARM_FEATURE_VFP3))
+                          return 1;
+                        gen_vfp_tosl(dp, rm);
+                        break;
+                    case 30: /* ftouh */
+                        if (!arm_feature(env, ARM_FEATURE_VFP3))
+                          return 1;
+                        gen_vfp_touh(dp, rm);
+                        break;
+                    case 31: /* ftoul */
+                        if (!arm_feature(env, ARM_FEATURE_VFP3))
+                          return 1;
+                        gen_vfp_toul(dp, rm);
+                        break;
                     default: /* undefined */
                         printf ("rn:%d\n", rn);
                         return 1;
@@ -1994,16 +2311,15 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
         break;
     case 0xc:
     case 0xd:
-        if (dp && (insn & (1 << 22))) {
+        if (dp && (insn & 0x03e00000) == 0x00400000) {
             /* two-register transfer */
             rn = (insn >> 16) & 0xf;
             rd = (insn >> 12) & 0xf;
             if (dp) {
-                if (insn & (1 << 5))
-                    return 1;
-                rm = insn & 0xf;
-            } else
-                rm = ((insn << 1) & 0x1e) | ((insn >> 5) & 1);
+                VFP_DREG_M(rm, insn);
+            } else {
+                rm = VFP_SREG_M(insn);
+            }
 
             if (insn & ARM_CP_RW_BIT) {
                 /* vfp->arm */
@@ -2040,10 +2356,14 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
             /* Load/store */
             rn = (insn >> 16) & 0xf;
             if (dp)
-                rd = (insn >> 12) & 0xf;
+                VFP_DREG_D(rd, insn);
             else
-                rd = ((insn >> 11) & 0x1e) | ((insn >> 22) & 1);
-            gen_movl_T1_reg(s, rn);
+                rd = VFP_SREG_D(insn);
+            if (s->thumb && rn == 15) {
+                gen_op_movl_T1_im(s->pc & ~2);
+            } else {
+                gen_movl_T1_reg(s, rn);
+            }
             if ((insn & 0x01200000) == 0x01000000) {
                 /* Single load/store */
                 offset = (insn & 0xff) << 2;
@@ -2156,8 +2476,9 @@ static inline void gen_mulxy(int x, int y)
 }
 
 /* Return the mask of PSR bits set by a MSR instruction.  */
-static uint32_t msr_mask(DisasContext *s, int flags, int spsr) {
+static uint32_t msr_mask(CPUState *env, DisasContext *s, int flags, int spsr) {
     uint32_t mask;
+    uint32_t reserved;
 
     mask = 0;
     if (flags & (1 << 0))
@@ -2168,14 +2489,19 @@ static uint32_t msr_mask(DisasContext *s, int flags, int spsr) {
         mask |= 0xff0000;
     if (flags & (1 << 3))
         mask |= 0xff000000;
+
     /* Mask out undefined bits.  */
-    mask &= 0xf90f03ff;
-    /* Mask out state bits.  */
+    mask &= ~CPSR_RESERVED;
+    if (!arm_feature(env, ARM_FEATURE_V6))
+        reserved &= ~(CPSR_E | CPSR_GE);
+    if (!arm_feature(env, ARM_FEATURE_THUMB2))
+        reserved &= ~CPSR_IT;
+    /* Mask out execution state bits.  */
     if (!spsr)
-        mask &= ~0x01000020;
+        reserved &= ~CPSR_EXEC;
     /* Mask out privileged bits.  */
     if (IS_USER(s))
-        mask &= 0xf80f0200;
+        mask &= CPSR_USER;
     return mask;
 }
 
@@ -2194,6 +2520,7 @@ static int gen_set_psr_T0(DisasContext *s, uint32_t mask, int spsr)
     return 0;
 }
 
+/* Generate an old-style exception return.  */
 static void gen_exception_return(DisasContext *s)
 {
     gen_op_movl_reg_TN[0][15]();
@@ -2202,6 +2529,2122 @@ static void gen_exception_return(DisasContext *s)
     s->is_jmp = DISAS_UPDATE;
 }
 
+/* Generate a v6 exception return.  */
+static void gen_rfe(DisasContext *s)
+{
+    gen_op_movl_cpsr_T0(0xffffffff);
+    gen_op_movl_T0_T2();
+    gen_op_movl_reg_TN[0][15]();
+    s->is_jmp = DISAS_UPDATE;
+}
+
+static inline void
+gen_set_condexec (DisasContext *s)
+{
+    if (s->condexec_mask) {
+        gen_op_set_condexec((s->condexec_cond << 4) | (s->condexec_mask >> 1));
+    }
+}
+
+static void gen_nop_hint(DisasContext *s, int val)
+{
+    switch (val) {
+    case 3: /* wfi */
+        gen_op_movl_T0_im((long)s->pc);
+        gen_op_movl_reg_TN[0][15]();
+        s->is_jmp = DISAS_WFI;
+        break;
+    case 2: /* wfe */
+    case 4: /* sev */
+        /* TODO: Implement SEV and WFE.  May help SMP performance.  */
+    default: /* nop */
+        break;
+    }
+}
+
+/* Neon shift by constant.  The actual ops are the same as used for variable
+   shifts.  [OP][U][SIZE]  */
+static GenOpFunc *gen_neon_shift_im[8][2][4] = {
+    { /* 0 */ /* VSHR */
+      {
+        gen_op_neon_shl_u8,
+        gen_op_neon_shl_u16,
+        gen_op_neon_shl_u32,
+        gen_op_neon_shl_u64
+      }, {
+        gen_op_neon_shl_s8,
+        gen_op_neon_shl_s16,
+        gen_op_neon_shl_s32,
+        gen_op_neon_shl_s64
+      }
+    }, { /* 1 */ /* VSRA */
+      {
+        gen_op_neon_shl_u8,
+        gen_op_neon_shl_u16,
+        gen_op_neon_shl_u32,
+        gen_op_neon_shl_u64
+      }, {
+        gen_op_neon_shl_s8,
+        gen_op_neon_shl_s16,
+        gen_op_neon_shl_s32,
+        gen_op_neon_shl_s64
+      }
+    }, { /* 2 */ /* VRSHR */
+      {
+        gen_op_neon_rshl_u8,
+        gen_op_neon_rshl_u16,
+        gen_op_neon_rshl_u32,
+        gen_op_neon_rshl_u64
+      }, {
+        gen_op_neon_rshl_s8,
+        gen_op_neon_rshl_s16,
+        gen_op_neon_rshl_s32,
+        gen_op_neon_rshl_s64
+      }
+    }, { /* 3 */ /* VRSRA */
+      {
+        gen_op_neon_rshl_u8,
+        gen_op_neon_rshl_u16,
+        gen_op_neon_rshl_u32,
+        gen_op_neon_rshl_u64
+      }, {
+        gen_op_neon_rshl_s8,
+        gen_op_neon_rshl_s16,
+        gen_op_neon_rshl_s32,
+        gen_op_neon_rshl_s64
+      }
+    }, { /* 4 */
+      {
+        NULL, NULL, NULL, NULL
+      }, { /* VSRI */
+        gen_op_neon_shl_u8,
+        gen_op_neon_shl_u16,
+        gen_op_neon_shl_u32,
+        gen_op_neon_shl_u64,
+      }
+    }, { /* 5 */
+      { /* VSHL */
+        gen_op_neon_shl_u8,
+        gen_op_neon_shl_u16,
+        gen_op_neon_shl_u32,
+        gen_op_neon_shl_u64,
+      }, { /* VSLI */
+        gen_op_neon_shl_u8,
+        gen_op_neon_shl_u16,
+        gen_op_neon_shl_u32,
+        gen_op_neon_shl_u64,
+      }
+    }, { /* 6 */ /* VQSHL */
+      {
+        gen_op_neon_qshl_u8,
+        gen_op_neon_qshl_u16,
+        gen_op_neon_qshl_u32,
+        gen_op_neon_qshl_u64
+      }, {
+        gen_op_neon_qshl_s8,
+        gen_op_neon_qshl_s16,
+        gen_op_neon_qshl_s32,
+        gen_op_neon_qshl_s64
+      }
+    }, { /* 7 */ /* VQSHLU */
+      {
+        gen_op_neon_qshl_u8,
+        gen_op_neon_qshl_u16,
+        gen_op_neon_qshl_u32,
+        gen_op_neon_qshl_u64
+      }, {
+        gen_op_neon_qshl_u8,
+        gen_op_neon_qshl_u16,
+        gen_op_neon_qshl_u32,
+        gen_op_neon_qshl_u64
+      }
+    }
+};
+
+/* [R][U][size - 1] */
+static GenOpFunc *gen_neon_shift_im_narrow[2][2][3] = {
+    {
+      {
+        gen_op_neon_shl_u16,
+        gen_op_neon_shl_u32,
+        gen_op_neon_shl_u64
+      }, {
+        gen_op_neon_shl_s16,
+        gen_op_neon_shl_s32,
+        gen_op_neon_shl_s64
+      }
+    }, {
+      {
+        gen_op_neon_rshl_u16,
+        gen_op_neon_rshl_u32,
+        gen_op_neon_rshl_u64
+      }, {
+        gen_op_neon_rshl_s16,
+        gen_op_neon_rshl_s32,
+        gen_op_neon_rshl_s64
+      }
+    }
+};
+
+static inline void
+gen_op_neon_narrow_u32 ()
+{
+    /* No-op.  */
+}
+
+static GenOpFunc *gen_neon_narrow[3] = {
+    gen_op_neon_narrow_u8,
+    gen_op_neon_narrow_u16,
+    gen_op_neon_narrow_u32
+};
+
+static GenOpFunc *gen_neon_narrow_satu[3] = {
+    gen_op_neon_narrow_sat_u8,
+    gen_op_neon_narrow_sat_u16,
+    gen_op_neon_narrow_sat_u32
+};
+
+static GenOpFunc *gen_neon_narrow_sats[3] = {
+    gen_op_neon_narrow_sat_s8,
+    gen_op_neon_narrow_sat_s16,
+    gen_op_neon_narrow_sat_s32
+};
+
+static inline int gen_neon_add(int size)
+{
+    switch (size) {
+    case 0: gen_op_neon_add_u8(); break;
+    case 1: gen_op_neon_add_u16(); break;
+    case 2: gen_op_addl_T0_T1(); break;
+    default: return 1;
+    }
+    return 0;
+}
+
+/* 32-bit pairwise ops end up the same as the elementsise versions.  */
+#define gen_op_neon_pmax_s32  gen_op_neon_max_s32
+#define gen_op_neon_pmax_u32  gen_op_neon_max_u32
+#define gen_op_neon_pmin_s32  gen_op_neon_min_s32
+#define gen_op_neon_pmin_u32  gen_op_neon_min_u32
+
+#define GEN_NEON_INTEGER_OP(name) do { \
+    switch ((size << 1) | u) { \
+    case 0: gen_op_neon_##name##_s8(); break; \
+    case 1: gen_op_neon_##name##_u8(); break; \
+    case 2: gen_op_neon_##name##_s16(); break; \
+    case 3: gen_op_neon_##name##_u16(); break; \
+    case 4: gen_op_neon_##name##_s32(); break; \
+    case 5: gen_op_neon_##name##_u32(); break; \
+    default: return 1; \
+    }} while (0)
+
+static inline void
+gen_neon_movl_scratch_T0(int scratch)
+{
+  uint32_t offset;
+
+  offset = offsetof(CPUARMState, vfp.scratch[scratch]);
+  gen_op_neon_setreg_T0(offset);
+}
+
+static inline void
+gen_neon_movl_scratch_T1(int scratch)
+{
+  uint32_t offset;
+
+  offset = offsetof(CPUARMState, vfp.scratch[scratch]);
+  gen_op_neon_setreg_T1(offset);
+}
+
+static inline void
+gen_neon_movl_T0_scratch(int scratch)
+{
+  uint32_t offset;
+
+  offset = offsetof(CPUARMState, vfp.scratch[scratch]);
+  gen_op_neon_getreg_T0(offset);
+}
+
+static inline void
+gen_neon_movl_T1_scratch(int scratch)
+{
+  uint32_t offset;
+
+  offset = offsetof(CPUARMState, vfp.scratch[scratch]);
+  gen_op_neon_getreg_T1(offset);
+}
+
+static inline void gen_op_neon_widen_u32(void)
+{
+    gen_op_movl_T1_im(0);
+}
+
+static inline void gen_neon_get_scalar(int size, int reg)
+{
+    if (size == 1) {
+        NEON_GET_REG(T0, reg >> 1, reg & 1);
+    } else {
+        NEON_GET_REG(T0, reg >> 2, (reg >> 1) & 1);
+        if (reg & 1)
+            gen_op_neon_dup_low16();
+        else
+            gen_op_neon_dup_high16();
+    }
+}
+
+static void gen_neon_unzip(int reg, int q, int tmp, int size)
+{
+    int n;
+
+    for (n = 0; n < q + 1; n += 2) {
+        NEON_GET_REG(T0, reg, n);
+        NEON_GET_REG(T0, reg, n + n);
+        switch (size) {
+        case 0: gen_op_neon_unzip_u8(); break;
+        case 1: gen_op_neon_zip_u16(); break; /* zip and unzip are the same.  */
+        case 2: /* no-op */; break;
+        default: abort();
+        }
+        gen_neon_movl_scratch_T0(tmp + n);
+        gen_neon_movl_scratch_T1(tmp + n + 1);
+    }
+}
+
+static struct {
+    int nregs;
+    int interleave;
+    int spacing;
+} neon_ls_element_type[11] = {
+    {4, 4, 1},
+    {4, 4, 2},
+    {4, 1, 1},
+    {4, 2, 1},
+    {3, 3, 1},
+    {3, 3, 2},
+    {3, 1, 1},
+    {1, 1, 1},
+    {2, 2, 1},
+    {2, 2, 2},
+    {2, 1, 1}
+};
+
+/* Translate a NEON load/store element instruction.  Return nonzero if the
+   instruction is invalid.  */
+static int disas_neon_ls_insn(CPUState * env, DisasContext *s, uint32_t insn)
+{
+    int rd, rn, rm;
+    int op;
+    int nregs;
+    int interleave;
+    int stride;
+    int size;
+    int reg;
+    int pass;
+    int load;
+    int shift;
+    uint32_t mask;
+    int n;
+
+    if (!vfp_enabled(env))
+      return 1;
+    VFP_DREG_D(rd, insn);
+    rn = (insn >> 16) & 0xf;
+    rm = insn & 0xf;
+    load = (insn & (1 << 21)) != 0;
+    if ((insn & (1 << 23)) == 0) {
+        /* Load store all elements.  */
+        op = (insn >> 8) & 0xf;
+        size = (insn >> 6) & 3;
+        if (op > 10 || size == 3)
+            return 1;
+        nregs = neon_ls_element_type[op].nregs;
+        interleave = neon_ls_element_type[op].interleave;
+        gen_movl_T1_reg(s, rn);
+        stride = (1 << size) * interleave;
+        for (reg = 0; reg < nregs; reg++) {
+            if (interleave > 2 || (interleave == 2 && nregs == 2)) {
+                gen_movl_T1_reg(s, rn);
+                gen_op_addl_T1_im((1 << size) * reg);
+            } else if (interleave == 2 && nregs == 4 && reg == 2) {
+                gen_movl_T1_reg(s, rn);
+                gen_op_addl_T1_im(1 << size);
+            }
+            for (pass = 0; pass < 2; pass++) {
+                if (size == 2) {
+                    if (load) {
+                        gen_ldst(ldl, s);
+                        NEON_SET_REG(T0, rd, pass);
+                    } else {
+                        NEON_GET_REG(T0, rd, pass);
+                        gen_ldst(stl, s);
+                    }
+                    gen_op_addl_T1_im(stride);
+                } else if (size == 1) {
+                    if (load) {
+                        gen_ldst(lduw, s);
+                        gen_op_addl_T1_im(stride);
+                        gen_op_movl_T2_T0();
+                        gen_ldst(lduw, s);
+                        gen_op_addl_T1_im(stride);
+                        gen_op_neon_insert_elt(16, 0xffff);
+                        NEON_SET_REG(T2, rd, pass);
+                    } else {
+                        NEON_GET_REG(T2, rd, pass);
+                        gen_op_movl_T0_T2();
+                        gen_ldst(stw, s);
+                        gen_op_addl_T1_im(stride);
+                        gen_op_neon_extract_elt(16, 0xffff0000);
+                        gen_ldst(stw, s);
+                        gen_op_addl_T1_im(stride);
+                    }
+                } else /* size == 0 */ {
+                    if (load) {
+                        mask = 0xff;
+                        for (n = 0; n < 4; n++) {
+                            gen_ldst(ldub, s);
+                            gen_op_addl_T1_im(stride);
+                            if (n == 0) {
+                                gen_op_movl_T2_T0();
+                            } else {
+                                gen_op_neon_insert_elt(n * 8, ~mask);
+                            }
+                            mask <<= 8;
+                        }
+                        NEON_SET_REG(T2, rd, pass);
+                    } else {
+                        NEON_GET_REG(T2, rd, pass);
+                        mask = 0xff;
+                        for (n = 0; n < 4; n++) {
+                            if (n == 0) {
+                                gen_op_movl_T0_T2();
+                            } else {
+                                gen_op_neon_extract_elt(n * 8, mask);
+                            }
+                            gen_ldst(stb, s);
+                            gen_op_addl_T1_im(stride);
+                            mask <<= 8;
+                        }
+                    }
+                }
+            }
+            rd += neon_ls_element_type[op].spacing;
+        }
+        stride = nregs * 8;
+    } else {
+        size = (insn >> 10) & 3;
+        if (size == 3) {
+            /* Load single element to all lanes.  */
+            if (!load)
+                return 1;
+            size = (insn >> 6) & 3;
+            nregs = ((insn >> 8) & 3) + 1;
+            stride = (insn & (1 << 5)) ? 2 : 1;
+            gen_movl_T1_reg(s, rn);
+            for (reg = 0; reg < nregs; reg++) {
+                switch (size) {
+                case 0:
+                    gen_ldst(ldub, s);
+                    gen_op_neon_dup_u8(0);
+                    break;
+                case 1:
+                    gen_ldst(lduw, s);
+                    gen_op_neon_dup_low16();
+                    break;
+                case 2:
+                    gen_ldst(ldl, s);
+                    break;
+                case 3:
+                    return 1;
+                }
+                gen_op_addl_T1_im(1 << size);
+                NEON_SET_REG(T0, rd, 0);
+                NEON_SET_REG(T0, rd, 1);
+                rd += stride;
+            }
+            stride = (1 << size) * nregs;
+        } else {
+            /* Single element.  */
+            pass = (insn >> 7) & 1;
+            switch (size) {
+            case 0:
+                shift = ((insn >> 5) & 3) * 8;
+                mask = 0xff << shift;
+                stride = 1;
+                break;
+            case 1:
+                shift = ((insn >> 6) & 1) * 16;
+                mask = shift ? 0xffff0000 : 0xffff;
+                stride = (insn & (1 << 5)) ? 2 : 1;
+                break;
+            case 2:
+                shift = 0;
+                mask = 0xffffffff;
+                stride = (insn & (1 << 6)) ? 2 : 1;
+                break;
+            default:
+                abort();
+            }
+            nregs = ((insn >> 8) & 3) + 1;
+            gen_movl_T1_reg(s, rn);
+            for (reg = 0; reg < nregs; reg++) {
+                if (load) {
+                    if (size != 2) {
+                        NEON_GET_REG(T2, rd, pass);
+                    }
+                    switch (size) {
+                    case 0:
+                        gen_ldst(ldub, s);
+                        break;
+                    case 1:
+                        gen_ldst(lduw, s);
+                        break;
+                    case 2:
+                        gen_ldst(ldl, s);
+                        NEON_SET_REG(T0, rd, pass);
+                        break;
+                    }
+                    if (size != 2) {
+                        gen_op_neon_insert_elt(shift, ~mask);
+                        NEON_SET_REG(T0, rd, pass);
+                    }
+                } else { /* Store */
+                    if (size == 2) {
+                        NEON_GET_REG(T0, rd, pass);
+                    } else {
+                        NEON_GET_REG(T2, rd, pass);
+                        gen_op_neon_extract_elt(shift, mask);
+                    }
+                    switch (size) {
+                    case 0:
+                        gen_ldst(stb, s);
+                        break;
+                    case 1:
+                        gen_ldst(stw, s);
+                        break;
+                    case 2:
+                        gen_ldst(stl, s);
+                        break;
+                    }
+                }
+                rd += stride;
+                gen_op_addl_T1_im(1 << size);
+            }
+            stride = nregs * (1 << size);
+        }
+    }
+    if (rm != 15) {
+        gen_movl_T1_reg(s, rn);
+        if (rm == 13) {
+            gen_op_addl_T1_im(stride);
+        } else {
+            gen_movl_T2_reg(s, rm);
+            gen_op_addl_T1_T2();
+        }
+        gen_movl_reg_T1(s, rn);
+    }
+    return 0;
+}
+
+/* Translate a NEON data processing instruction.  Return nonzero if the
+   instruction is invalid.
+   In general we process vectors in 32-bit chunks.  This means we can reuse
+   some of the scalar ops, and hopefully the code generated for 32-bit
+   hosts won't be too awful.  The downside is that the few 64-bit operations
+   (mainly shifts) get complicated.  */
+
+static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
+{
+    int op;
+    int q;
+    int rd, rn, rm;
+    int size;
+    int shift;
+    int pass;
+    int count;
+    int pairwise;
+    int u;
+    int n;
+    uint32_t imm;
+
+    if (!vfp_enabled(env))
+      return 1;
+    q = (insn & (1 << 6)) != 0;
+    u = (insn >> 24) & 1;
+    VFP_DREG_D(rd, insn);
+    VFP_DREG_N(rn, insn);
+    VFP_DREG_M(rm, insn);
+    size = (insn >> 20) & 3;
+    if ((insn & (1 << 23)) == 0) {
+        /* Three register same length.  */
+        op = ((insn >> 7) & 0x1e) | ((insn >> 4) & 1);
+        if (size == 3 && (op == 1 || op == 5 || op == 16)) {
+            for (pass = 0; pass < (q ? 2 : 1); pass++) {
+                NEON_GET_REG(T0, rm, pass * 2);
+                NEON_GET_REG(T1, rm, pass * 2 + 1);
+                gen_neon_movl_scratch_T0(0);
+                gen_neon_movl_scratch_T1(1);
+                NEON_GET_REG(T0, rn, pass * 2);
+                NEON_GET_REG(T1, rn, pass * 2 + 1);
+                switch (op) {
+                case 1: /* VQADD */
+                    if (u) {
+                        gen_op_neon_addl_saturate_u64();
+                    } else {
+                        gen_op_neon_addl_saturate_s64();
+                    }
+                    break;
+                case 5: /* VQSUB */
+                    if (u) {
+                        gen_op_neon_subl_saturate_u64();
+                    } else {
+                        gen_op_neon_subl_saturate_s64();
+                    }
+                    break;
+                case 16:
+                    if (u) {
+                        gen_op_neon_subl_u64();
+                    } else {
+                        gen_op_neon_addl_u64();
+                    }
+                    break;
+                default:
+                    abort();
+                }
+                NEON_SET_REG(T0, rd, pass * 2);
+                NEON_SET_REG(T1, rd, pass * 2 + 1);
+            }
+            return 0;
+        }
+        switch (op) {
+        case 8: /* VSHL */
+        case 9: /* VQSHL */
+        case 10: /* VRSHL */
+        case 11: /* VQSHL */
+            /* Shift operations have Rn and Rm reversed.  */
+            {
+                int tmp;
+                tmp = rn;
+                rn = rm;
+                rm = tmp;
+                pairwise = 0;
+            }
+            break;
+        case 20: /* VPMAX */
+        case 21: /* VPMIN */
+        case 23: /* VPADD */
+            pairwise = 1;
+            break;
+        case 26: /* VPADD (float) */
+            pairwise = (u && size < 2);
+            break;
+        case 30: /* VPMIN/VPMAX (float) */
+            pairwise = u;
+            break;
+        default:
+            pairwise = 0;
+            break;
+        }
+        for (pass = 0; pass < (q ? 4 : 2); pass++) {
+
+        if (pairwise) {
+            /* Pairwise.  */
+            if (q)
+                n = (pass & 1) * 2;
+            else
+                n = 0;
+            if (pass < q + 1) {
+                NEON_GET_REG(T0, rn, n);
+                NEON_GET_REG(T1, rn, n + 1);
+            } else {
+                NEON_GET_REG(T0, rm, n);
+                NEON_GET_REG(T1, rm, n + 1);
+            }
+        } else {
+            /* Elementwise.  */
+            NEON_GET_REG(T0, rn, pass);
+            NEON_GET_REG(T1, rm, pass);
+        }
+        switch (op) {
+        case 0: /* VHADD */
+            GEN_NEON_INTEGER_OP(hadd);
+            break;
+        case 1: /* VQADD */
+            switch (size << 1| u) {
+            case 0: gen_op_neon_qadd_s8(); break;
+            case 1: gen_op_neon_qadd_u8(); break;
+            case 2: gen_op_neon_qadd_s16(); break;
+            case 3: gen_op_neon_qadd_u16(); break;
+            case 4: gen_op_addl_T0_T1_saturate(); break;
+            case 5: gen_op_addl_T0_T1_usaturate(); break;
+            default: abort();
+            }
+            break;
+        case 2: /* VRHADD */
+            GEN_NEON_INTEGER_OP(rhadd);
+            break;
+        case 3: /* Logic ops.  */
+            switch ((u << 2) | size) {
+            case 0: /* VAND */
+                gen_op_andl_T0_T1();
+                break;
+            case 1: /* BIC */
+                gen_op_bicl_T0_T1();
+                break;
+            case 2: /* VORR */
+                gen_op_orl_T0_T1();
+                break;
+            case 3: /* VORN */
+                gen_op_notl_T1();
+                gen_op_orl_T0_T1();
+                break;
+            case 4: /* VEOR */
+                gen_op_xorl_T0_T1();
+                break;
+            case 5: /* VBSL */
+                NEON_GET_REG(T2, rd, pass);
+                gen_op_neon_bsl();
+                break;
+            case 6: /* VBIT */
+                NEON_GET_REG(T2, rd, pass);
+                gen_op_neon_bit();
+                break;
+            case 7: /* VBIF */
+                NEON_GET_REG(T2, rd, pass);
+                gen_op_neon_bif();
+                break;
+            }
+            break;
+        case 4: /* VHSUB */
+            GEN_NEON_INTEGER_OP(hsub);
+            break;
+        case 5: /* VQSUB */
+            switch ((size << 1) | u) {
+            case 0: gen_op_neon_qsub_s8(); break;
+            case 1: gen_op_neon_qsub_u8(); break;
+            case 2: gen_op_neon_qsub_s16(); break;
+            case 3: gen_op_neon_qsub_u16(); break;
+            case 4: gen_op_subl_T0_T1_saturate(); break;
+            case 5: gen_op_subl_T0_T1_usaturate(); break;
+            default: abort();
+            }
+            break;
+        case 6: /* VCGT */
+            GEN_NEON_INTEGER_OP(cgt);
+            break;
+        case 7: /* VCGE */
+            GEN_NEON_INTEGER_OP(cge);
+            break;
+        case 8: /* VSHL */
+            switch ((size << 1) | u) {
+            case 0: gen_op_neon_shl_s8(); break;
+            case 1: gen_op_neon_shl_u8(); break;
+            case 2: gen_op_neon_shl_s16(); break;
+            case 3: gen_op_neon_shl_u16(); break;
+            case 4: gen_op_neon_shl_s32(); break;
+            case 5: gen_op_neon_shl_u32(); break;
+#if 0
+            /* ??? Implementing these is tricky because the vector ops work
+               on 32-bit pieces.  */
+            case 6: gen_op_neon_shl_s64(); break;
+            case 7: gen_op_neon_shl_u64(); break;
+#else
+            case 6: case 7: cpu_abort(env, "VSHL.64 not implemented");
+#endif
+            }
+            break;
+        case 9: /* VQSHL */
+            switch ((size << 1) | u) {
+            case 0: gen_op_neon_qshl_s8(); break;
+            case 1: gen_op_neon_qshl_u8(); break;
+            case 2: gen_op_neon_qshl_s16(); break;
+            case 3: gen_op_neon_qshl_u16(); break;
+            case 4: gen_op_neon_qshl_s32(); break;
+            case 5: gen_op_neon_qshl_u32(); break;
+#if 0
+            /* ??? Implementing these is tricky because the vector ops work
+               on 32-bit pieces.  */
+            case 6: gen_op_neon_qshl_s64(); break;
+            case 7: gen_op_neon_qshl_u64(); break;
+#else
+            case 6: case 7: cpu_abort(env, "VQSHL.64 not implemented");
+#endif
+            }
+            break;
+        case 10: /* VRSHL */
+            switch ((size << 1) | u) {
+            case 0: gen_op_neon_rshl_s8(); break;
+            case 1: gen_op_neon_rshl_u8(); break;
+            case 2: gen_op_neon_rshl_s16(); break;
+            case 3: gen_op_neon_rshl_u16(); break;
+            case 4: gen_op_neon_rshl_s32(); break;
+            case 5: gen_op_neon_rshl_u32(); break;
+#if 0
+            /* ??? Implementing these is tricky because the vector ops work
+               on 32-bit pieces.  */
+            case 6: gen_op_neon_rshl_s64(); break;
+            case 7: gen_op_neon_rshl_u64(); break;
+#else
+            case 6: case 7: cpu_abort(env, "VRSHL.64 not implemented");
+#endif
+            }
+            break;
+        case 11: /* VQRSHL */
+            switch ((size << 1) | u) {
+            case 0: gen_op_neon_qrshl_s8(); break;
+            case 1: gen_op_neon_qrshl_u8(); break;
+            case 2: gen_op_neon_qrshl_s16(); break;
+            case 3: gen_op_neon_qrshl_u16(); break;
+            case 4: gen_op_neon_qrshl_s32(); break;
+            case 5: gen_op_neon_qrshl_u32(); break;
+#if 0
+            /* ??? Implementing these is tricky because the vector ops work
+               on 32-bit pieces.  */
+            case 6: gen_op_neon_qrshl_s64(); break;
+            case 7: gen_op_neon_qrshl_u64(); break;
+#else
+            case 6: case 7: cpu_abort(env, "VQRSHL.64 not implemented");
+#endif
+            }
+            break;
+        case 12: /* VMAX */
+            GEN_NEON_INTEGER_OP(max);
+            break;
+        case 13: /* VMIN */
+            GEN_NEON_INTEGER_OP(min);
+            break;
+        case 14: /* VABD */
+            GEN_NEON_INTEGER_OP(abd);
+            break;
+        case 15: /* VABA */
+            GEN_NEON_INTEGER_OP(abd);
+            NEON_GET_REG(T1, rd, pass);
+            gen_neon_add(size);
+            break;
+        case 16:
+            if (!u) { /* VADD */
+                if (gen_neon_add(size))
+                    return 1;
+            } else { /* VSUB */
+                switch (size) {
+                case 0: gen_op_neon_sub_u8(); break;
+                case 1: gen_op_neon_sub_u16(); break;
+                case 2: gen_op_subl_T0_T1(); break;
+                default: return 1;
+                }
+            }
+            break;
+        case 17:
+            if (!u) { /* VTST */
+                switch (size) {
+                case 0: gen_op_neon_tst_u8(); break;
+                case 1: gen_op_neon_tst_u16(); break;
+                case 2: gen_op_neon_tst_u32(); break;
+                default: return 1;
+                }
+            } else { /* VCEQ */
+                switch (size) {
+                case 0: gen_op_neon_ceq_u8(); break;
+                case 1: gen_op_neon_ceq_u16(); break;
+                case 2: gen_op_neon_ceq_u32(); break;
+                default: return 1;
+                }
+            }
+            break;
+        case 18: /* Multiply.  */
+            switch (size) {
+            case 0: gen_op_neon_mul_u8(); break;
+            case 1: gen_op_neon_mul_u16(); break;
+            case 2: gen_op_mul_T0_T1(); break;
+            default: return 1;
+            }
+            NEON_GET_REG(T1, rd, pass);
+            if (u) { /* VMLS */
+                switch (size) {
+                case 0: gen_op_neon_rsb_u8(); break;
+                case 1: gen_op_neon_rsb_u16(); break;
+                case 2: gen_op_rsbl_T0_T1(); break;
+                default: return 1;
+                }
+            } else { /* VMLA */
+                gen_neon_add(size);
+            }
+            break;
+        case 19: /* VMUL */
+            if (u) { /* polynomial */
+                gen_op_neon_mul_p8();
+            } else { /* Integer */
+                switch (size) {
+                case 0: gen_op_neon_mul_u8(); break;
+                case 1: gen_op_neon_mul_u16(); break;
+                case 2: gen_op_mul_T0_T1(); break;
+                default: return 1;
+                }
+            }
+            break;
+        case 20: /* VPMAX */
+            GEN_NEON_INTEGER_OP(pmax);
+            break;
+        case 21: /* VPMIN */
+            GEN_NEON_INTEGER_OP(pmin);
+            break;
+        case 22: /* Hultiply high.  */
+            if (!u) { /* VQDMULH */
+                switch (size) {
+                case 1: gen_op_neon_qdmulh_s16(); break;
+                case 2: gen_op_neon_qdmulh_s32(); break;
+                default: return 1;
+                }
+            } else { /* VQRDHMUL */
+                switch (size) {
+                case 1: gen_op_neon_qrdmulh_s16(); break;
+                case 2: gen_op_neon_qrdmulh_s32(); break;
+                default: return 1;
+                }
+            }
+            break;
+        case 23: /* VPADD */
+            if (u)
+                return 1;
+            switch (size) {
+            case 0: gen_op_neon_padd_u8(); break;
+            case 1: gen_op_neon_padd_u16(); break;
+            case 2: gen_op_addl_T0_T1(); break;
+            default: return 1;
+            }
+            break;
+        case 26: /* Floating point arithnetic.  */
+            switch ((u << 2) | size) {
+            case 0: /* VADD */
+                gen_op_neon_add_f32();
+                break;
+            case 2: /* VSUB */
+                gen_op_neon_sub_f32();
+                break;
+            case 4: /* VPADD */
+                gen_op_neon_add_f32();
+                break;
+            case 6: /* VABD */
+                gen_op_neon_abd_f32();
+                break;
+            default:
+                return 1;
+            }
+            break;
+        case 27: /* Float multiply.  */
+            gen_op_neon_mul_f32();
+            if (!u) {
+                NEON_GET_REG(T1, rd, pass);
+                if (size == 0) {
+                    gen_op_neon_add_f32();
+                } else {
+                    gen_op_neon_rsb_f32();
+                }
+            }
+            break;
+        case 28: /* Float compare.  */
+            if (!u) {
+                gen_op_neon_ceq_f32();
+            } else {
+                if (size == 0)
+                    gen_op_neon_cge_f32();
+                else
+                    gen_op_neon_cgt_f32();
+            }
+            break;
+        case 29: /* Float compare absolute.  */
+            if (!u)
+                return 1;
+            if (size == 0)
+                gen_op_neon_acge_f32();
+            else
+                gen_op_neon_acgt_f32();
+            break;
+        case 30: /* Float min/max.  */
+            if (size == 0)
+                gen_op_neon_max_f32();
+            else
+                gen_op_neon_min_f32();
+            break;
+        case 31:
+            if (size == 0)
+                gen_op_neon_recps_f32();
+            else
+                gen_op_neon_rsqrts_f32();
+            break;
+        default:
+            abort();
+        }
+        /* Save the result.  For elementwise operations we can put it
+           straight into the destination register.  For pairwise operations
+           we have to be careful to avoid clobbering the source operands.  */
+        if (pairwise && rd == rm) {
+            gen_neon_movl_scratch_T0(pass);
+        } else {
+            NEON_SET_REG(T0, rd, pass);
+        }
+
+        } /* for pass */
+        if (pairwise && rd == rm) {
+            for (pass = 0; pass < (q ? 4 : 2); pass++) {
+                gen_neon_movl_T0_scratch(pass);
+                NEON_SET_REG(T0, rd, pass);
+            }
+        }
+    } else if (insn & (1 << 4)) {
+        if ((insn & 0x00380080) != 0) {
+            /* Two registers and shift.  */
+            op = (insn >> 8) & 0xf;
+            if (insn & (1 << 7)) {
+                /* 64-bit shift.   */
+                size = 3;
+            } else {
+                size = 2;
+                while ((insn & (1 << (size + 19))) == 0)
+                    size--;
+            }
+            shift = (insn >> 16) & ((1 << (3 + size)) - 1);
+            /* To avoid excessive dumplication of ops we implement shift
+               by immediate using the variable shift operations.  */
+            if (op < 8) {
+                /* Shift by immediate:
+                   VSHR, VSRA, VRSHR, VRSRA, VSRI, VSHL, VQSHL, VQSHLU.  */
+                /* Right shifts are encoded as N - shift, where N is the
+                   element size in bits.  */
+                if (op <= 4)
+                    shift = shift - (1 << (size + 3));
+                else
+                    shift++;
+                if (size == 3) {
+                    count = q + 1;
+                } else {
+                    count = q ? 4: 2;
+                }
+                switch (size) {
+                case 0:
+                    imm = (uint8_t) shift;
+                    imm |= imm << 8;
+                    imm |= imm << 16;
+                    break;
+                case 1:
+                    imm = (uint16_t) shift;
+                    imm |= imm << 16;
+                    break;
+                case 2:
+                case 3:
+                    imm = shift;
+                    break;
+                default:
+                    abort();
+                }
+
+                for (pass = 0; pass < count; pass++) {
+                    if (size < 3) {
+                        /* Operands in T0 and T1.  */
+                        gen_op_movl_T1_im(imm);
+                        NEON_GET_REG(T0, rm, pass);
+                    } else {
+                        /* Operands in {T0, T1} and env->vfp.scratch.  */
+                        gen_op_movl_T0_im(imm);
+                        gen_neon_movl_scratch_T0(0);
+                        gen_op_movl_T0_im((int32_t)imm >> 31);
+                        gen_neon_movl_scratch_T0(1);
+                        NEON_GET_REG(T0, rm, pass * 2);
+                        NEON_GET_REG(T1, rm, pass * 2 + 1);
+                    }
+
+                    if (gen_neon_shift_im[op][u][size] == NULL)
+                        return 1;
+                    gen_neon_shift_im[op][u][size]();
+
+                    if (op == 1 || op == 3) {
+                        /* Accumulate.  */
+                        if (size == 3) {
+                            gen_neon_movl_scratch_T0(0);
+                            gen_neon_movl_scratch_T1(1);
+                            NEON_GET_REG(T0, rd, pass * 2);
+                            NEON_GET_REG(T1, rd, pass * 2 + 1);
+                            gen_op_neon_addl_u64();
+                        } else {
+                            NEON_GET_REG(T1, rd, pass);
+                            gen_neon_add(size);
+                        }
+                    } else if (op == 4 || (op == 5 && u)) {
+                        /* Insert */
+                        if (size == 3) {
+                            cpu_abort(env, "VS[LR]I.64 not implemented");
+                        }
+                        switch (size) {
+                        case 0:
+                            if (op == 4)
+                                imm = 0xff >> -shift;
+                            else
+                                imm = (uint8_t)(0xff << shift);
+                            imm |= imm << 8;
+                            imm |= imm << 16;
+                            break;
+                        case 1:
+                            if (op == 4)
+                                imm = 0xffff >> -shift;
+                            else
+                                imm = (uint16_t)(0xffff << shift);
+                            imm |= imm << 16;
+                            break;
+                        case 2:
+                            if (op == 4)
+                                imm = 0xffffffffu >> -shift;
+                            else
+                                imm = 0xffffffffu << shift;
+                            break;
+                        default:
+                            abort();
+                        }
+                        NEON_GET_REG(T1, rd, pass);
+                        gen_op_movl_T2_im(imm);
+                        gen_op_neon_bsl();
+                    }
+                    if (size == 3) {
+                        NEON_SET_REG(T0, rd, pass * 2);
+                        NEON_SET_REG(T1, rd, pass * 2 + 1);
+                    } else {
+                        NEON_SET_REG(T0, rd, pass);
+                    }
+                } /* for pass */
+            } else if (op < 10) {
+                /* Shift by immedaiate and narrow:
+                   VSHRN, VRSHRN, VQSHRN, VQRSHRN.  */
+                shift = shift - (1 << (size + 3));
+                size++;
+                if (size == 3) {
+                    count = q + 1;
+                } else {
+                    count = q ? 4: 2;
+                }
+                switch (size) {
+                case 1:
+                    imm = (uint16_t) shift;
+                    imm |= imm << 16;
+                    break;
+                case 2:
+                case 3:
+                    imm = shift;
+                    break;
+                default:
+                    abort();
+                }
+
+                /* Processing MSB first means we need to do less shuffling at
+                   the end.  */
+                for (pass =  count - 1; pass >= 0; pass--) {
+                    /* Avoid clobbering the second operand before it has been
+                       written.  */
+                    n = pass;
+                    if (rd == rm)
+                        n ^= (count - 1);
+                    else
+                        n = pass;
+
+                    if (size < 3) {
+                        /* Operands in T0 and T1.  */
+                        gen_op_movl_T1_im(imm);
+                        NEON_GET_REG(T0, rm, n);
+                    } else {
+                        /* Operands in {T0, T1} and env->vfp.scratch.  */
+                        gen_op_movl_T0_im(imm);
+                        gen_neon_movl_scratch_T0(0);
+                        gen_op_movl_T0_im((int32_t)imm >> 31);
+                        gen_neon_movl_scratch_T0(1);
+                        NEON_GET_REG(T0, rm, n * 2);
+                        NEON_GET_REG(T0, rm, n * 2 + 1);
+                    }
+
+                    gen_neon_shift_im_narrow[q][u][size - 1]();
+
+                    if (size < 3 && (pass & 1) == 0) {
+                        gen_neon_movl_scratch_T0(0);
+                    } else {
+                        uint32_t offset;
+
+                        if (size < 3)
+                            gen_neon_movl_T1_scratch(0);
+
+                        if (op == 8 && !u) {
+                            gen_neon_narrow[size - 1]();
+                        } else {
+                            if (op == 8)
+                                gen_neon_narrow_sats[size - 2]();
+                            else
+                                gen_neon_narrow_satu[size - 1]();
+                        }
+                        if (size == 3)
+                            offset = neon_reg_offset(rd, n);
+                        else
+                            offset = neon_reg_offset(rd, n >> 1);
+                        gen_op_neon_setreg_T0(offset);
+                    }
+                } /* for pass */
+            } else if (op == 10) {
+                /* VSHLL */
+                if (q)
+                    return 1;
+                for (pass = 0; pass < 2; pass++) {
+                    /* Avoid clobbering the input operand.  */
+                    if (rd == rm)
+                        n = 1 - pass;
+                    else
+                        n = pass;
+
+                    NEON_GET_REG(T0, rm, n);
+                    GEN_NEON_INTEGER_OP(widen);
+                    if (shift != 0) {
+                        /* The shift is less than the width of the source
+                           type, so in some cases we can just
+                           shift the whole register.  */
+                        if (size == 1 || (size == 0 && u)) {
+                            gen_op_shll_T0_im(shift);
+                            gen_op_shll_T1_im(shift);
+                        } else {
+                            switch (size) {
+                            case 0: gen_op_neon_shll_u16(shift); break;
+                            case 2: gen_op_neon_shll_u64(shift); break;
+                            default: abort();
+                            }
+                        }
+                    }
+                    NEON_SET_REG(T0, rd, n * 2);
+                    NEON_SET_REG(T1, rd, n * 2 + 1);
+                }
+            } 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));
+                    if (op & 1) {
+                        if (u)
+                            gen_op_vfp_ultos(shift);
+                        else
+                            gen_op_vfp_sltos(shift);
+                    } else {
+                        if (u)
+                            gen_op_vfp_touls(shift);
+                        else
+                            gen_op_vfp_tosls(shift);
+                    }
+                    gen_op_vfp_setreg_F0s(neon_reg_offset(rd, pass));
+                }
+            } else {
+                return 1;
+            }
+        } else { /* (insn & 0x00380080) == 0 */
+            int invert;
+
+            op = (insn >> 8) & 0xf;
+            /* One register and immediate.  */
+            imm = (u << 7) | ((insn >> 12) & 0x70) | (insn & 0xf);
+            invert = (insn & (1 << 5)) != 0;
+            switch (op) {
+            case 0: case 1:
+                /* no-op */
+                break;
+            case 2: case 3:
+                imm <<= 8;
+                break;
+            case 4: case 5:
+                imm <<= 16;
+                break;
+            case 6: case 7:
+                imm <<= 24;
+                break;
+            case 8: case 9:
+                imm |= imm << 16;
+                break;
+            case 10: case 11:
+                imm = (imm << 8) | (imm << 24);
+                break;
+            case 12:
+                imm = (imm < 8) | 0xff;
+                break;
+            case 13:
+                imm = (imm << 16) | 0xffff;
+                break;
+            case 14:
+                imm |= (imm << 8) | (imm << 16) | (imm << 24);
+                if (invert)
+                    imm = ~imm;
+                break;
+            case 15:
+                imm = ((imm & 0x80) << 24) | ((imm & 0x3f) << 19)
+                      | ((imm & 0x40) ? (0x1f << 25) : (1 << 30));
+                break;
+            }
+            if (invert)
+                imm = ~imm;
+
+            if (op != 14 || !invert)
+                gen_op_movl_T1_im(imm);
+
+            for (pass = 0; pass < (q ? 4 : 2); pass++) {
+                if (op & 1 && op < 12) {
+                    NEON_GET_REG(T0, rd, pass);
+                    if (invert) {
+                        /* The immediate value has already been inverted, so
+                           BIC becomes AND.  */
+                        gen_op_andl_T0_T1();
+                    } else {
+                        gen_op_orl_T0_T1();
+                    }
+                    NEON_SET_REG(T0, rd, pass);
+                } else {
+                    if (op == 14 && invert) {
+                        uint32_t tmp;
+                        tmp = 0;
+                        for (n = 0; n < 4; n++) {
+                            if (imm & (1 << (n + (pass & 1) * 4)))
+                                tmp |= 0xff << (n * 8);
+                        }
+                        gen_op_movl_T1_im(tmp);
+                    }
+                    /* VMOV, VMVN.  */
+                    NEON_SET_REG(T1, rd, pass);
+                }
+            }
+        }
+    } else { /* (insn & 0x00800010 == 0x00800010) */
+        if (size != 3) {
+            op = (insn >> 8) & 0xf;
+            if ((insn & (1 << 6)) == 0) {
+                /* Three registers of different lengths.  */
+                int src1_wide;
+                int src2_wide;
+                int prewiden;
+                /* prewiden, src1_wide, src2_wide */
+                static const int neon_3reg_wide[16][3] = {
+                    {1, 0, 0}, /* VADDL */
+                    {1, 1, 0}, /* VADDW */
+                    {1, 0, 0}, /* VSUBL */
+                    {1, 1, 0}, /* VSUBW */
+                    {0, 1, 1}, /* VADDHN */
+                    {0, 0, 0}, /* VABAL */
+                    {0, 1, 1}, /* VSUBHN */
+                    {0, 0, 0}, /* VABDL */
+                    {0, 0, 0}, /* VMLAL */
+                    {0, 0, 0}, /* VQDMLAL */
+                    {0, 0, 0}, /* VMLSL */
+                    {0, 0, 0}, /* VQDMLSL */
+                    {0, 0, 0}, /* Integer VMULL */
+                    {0, 0, 0}, /* VQDMULL */
+                    {0, 0, 0}  /* Polynomial VMULL */
+                };
+
+                prewiden = neon_3reg_wide[op][0];
+                src1_wide = neon_3reg_wide[op][1];
+                src2_wide = neon_3reg_wide[op][2];
+
+                /* Avoid overlapping operands.  Wide source operands are
+                   always aligned so will never overlap with wide
+                   destinations in problematic ways.  */
+                if (rd == rm) {
+                    NEON_GET_REG(T2, rm, 1);
+                } else if (rd == rn) {
+                    NEON_GET_REG(T2, rn, 1);
+                }
+                for (pass = 0; pass < 2; pass++) {
+                    /* Load the second operand into env->vfp.scratch.
+                       Also widen narrow operands.  */
+                    if (pass == 1 && rd == rm) {
+                        if (prewiden) {
+                            gen_op_movl_T0_T2();
+                        } else {
+                            gen_op_movl_T1_T2();
+                        }
+                    } else {
+                        if (src2_wide) {
+                            NEON_GET_REG(T0, rm, pass * 2);
+                            NEON_GET_REG(T1, rm, pass * 2 + 1);
+                        } else {
+                            if (prewiden) {
+                                NEON_GET_REG(T0, rm, pass);
+                            } else {
+                                NEON_GET_REG(T1, rm, pass);
+                            }
+                        }
+                    }
+                    if (prewiden && !src2_wide) {
+                        GEN_NEON_INTEGER_OP(widen);
+                    }
+                    if (prewiden || src2_wide) {
+                        gen_neon_movl_scratch_T0(0);
+                        gen_neon_movl_scratch_T1(1);
+                    }
+
+                    /* Load the first operand.  */
+                    if (pass == 1 && rd == rn) {
+                        gen_op_movl_T0_T2();
+                    } else {
+                        if (src1_wide) {
+                            NEON_GET_REG(T0, rn, pass * 2);
+                            NEON_GET_REG(T1, rn, pass * 2 + 1);
+                        } else {
+                            NEON_GET_REG(T0, rn, pass);
+                        }
+                    }
+                    if (prewiden && !src1_wide) {
+                        GEN_NEON_INTEGER_OP(widen);
+                    }
+                    switch (op) {
+                    case 0: case 1: case 4: /* VADDL, VADDW, VADDHN, VRADDHN */
+                        switch (size) {
+                        case 0: gen_op_neon_addl_u16(); break;
+                        case 1: gen_op_neon_addl_u32(); break;
+                        case 2: gen_op_neon_addl_u64(); break;
+                        default: abort();
+                        }
+                        break;
+                    case 2: case 3: case 6: /* VSUBL, VSUBW, VSUBHL, VRSUBHL */
+                        switch (size) {
+                        case 0: gen_op_neon_subl_u16(); break;
+                        case 1: gen_op_neon_subl_u32(); break;
+                        case 2: gen_op_neon_subl_u64(); break;
+                        default: abort();
+                        }
+                        break;
+                    case 5: case 7: /* VABAL, VABDL */
+                        switch ((size << 1) | u) {
+                        case 0: gen_op_neon_abdl_s16(); break;
+                        case 1: gen_op_neon_abdl_u16(); break;
+                        case 2: gen_op_neon_abdl_s32(); break;
+                        case 3: gen_op_neon_abdl_u32(); break;
+                        case 4: gen_op_neon_abdl_s64(); break;
+                        case 5: gen_op_neon_abdl_u64(); break;
+                        default: abort();
+                        }
+                        break;
+                    case 8: case 9: case 10: case 11: case 12: case 13:
+                        /* VMLAL, VQDMLAL, VMLSL, VQDMLSL, VMULL, VQDMULL */
+                        switch ((size << 1) | u) {
+                        case 0: gen_op_neon_mull_s8(); break;
+                        case 1: gen_op_neon_mull_u8(); break;
+                        case 2: gen_op_neon_mull_s16(); break;
+                        case 3: gen_op_neon_mull_u16(); break;
+                        case 4: gen_op_imull_T0_T1(); break;
+                        case 5: gen_op_mull_T0_T1(); break;
+                        default: abort();
+                        }
+                        break;
+                    case 14: /* Polynomial VMULL */
+                        cpu_abort(env, "Polynomial VMULL not implemented");
+
+                    default: /* 15 is RESERVED.  */
+                        return 1;
+                    }
+                    if (op == 5 || op == 13 || (op >= 8 && op <= 11)) {
+                        /* Accumulate.  */
+                        if (op == 10 || op == 11) {
+                            switch (size) {
+                            case 0: gen_op_neon_negl_u16(); break;
+                            case 1: gen_op_neon_negl_u32(); break;
+                            case 2: gen_op_neon_negl_u64(); break;
+                            default: abort();
+                            }
+                        }
+
+                        gen_neon_movl_scratch_T0(0);
+                        gen_neon_movl_scratch_T1(1);
+
+                        if (op != 13) {
+                            NEON_GET_REG(T0, rd, pass * 2);
+                            NEON_GET_REG(T1, rd, pass * 2 + 1);
+                        }
+
+                        switch (op) {
+                        case 5: case 8: case 10: /* VABAL, VMLAL, VMLSL */
+                            switch (size) {
+                            case 0: gen_op_neon_addl_u16(); break;
+                            case 1: gen_op_neon_addl_u32(); break;
+                            case 2: gen_op_neon_addl_u64(); break;
+                            default: abort();
+                            }
+                            break;
+                        case 9: case 11: /* VQDMLAL, VQDMLSL */
+                            switch (size) {
+                            case 1: gen_op_neon_addl_saturate_s32(); break;
+                            case 2: gen_op_neon_addl_saturate_s64(); break;
+                            default: abort();
+                            }
+                            /* Fall through.  */
+                        case 13: /* VQDMULL */
+                            switch (size) {
+                            case 1: gen_op_neon_addl_saturate_s32(); break;
+                            case 2: gen_op_neon_addl_saturate_s64(); break;
+                            default: abort();
+                            }
+                            break;
+                        default:
+                            abort();
+                        }
+                        NEON_SET_REG(T0, rd, pass * 2);
+                        NEON_SET_REG(T1, rd, pass * 2 + 1);
+                    } else if (op == 4 || op == 6) {
+                        /* Narrowing operation.  */
+                        if (u) {
+                            switch (size) {
+                            case 0: gen_op_neon_narrow_high_u8(); break;
+                            case 1: gen_op_neon_narrow_high_u16(); break;
+                            case 2: gen_op_movl_T0_T1(); break;
+                            default: abort();
+                            }
+                        } else {
+                            switch (size) {
+                            case 0: gen_op_neon_narrow_high_round_u8(); break;
+                            case 1: gen_op_neon_narrow_high_round_u16(); break;
+                            case 2: gen_op_neon_narrow_high_round_u32(); break;
+                            default: abort();
+                            }
+                        }
+                        NEON_SET_REG(T0, rd, pass);
+                    } else {
+                        /* Write back the result.  */
+                        NEON_SET_REG(T0, rd, pass * 2);
+                        NEON_SET_REG(T1, rd, pass * 2 + 1);
+                    }
+                }
+            } else {
+                /* Two registers and a scalar.  */
+                switch (op) {
+                case 0: /* Integer VMLA scalar */
+                case 1: /* Float VMLA scalar */
+                case 4: /* Integer VMLS scalar */
+                case 5: /* Floating point VMLS scalar */
+                case 8: /* Integer VMUL scalar */
+                case 9: /* Floating point VMUL scalar */
+                case 12: /* VQDMULH scalar */
+                case 13: /* VQRDMULH scalar */
+                    gen_neon_get_scalar(size, rm);
+                    gen_op_movl_T2_T0();
+                    for (pass = 0; pass < (u ? 4 : 2); pass++) {
+                        if (pass != 0)
+                            gen_op_movl_T0_T2();
+                        NEON_GET_REG(T1, rn, pass);
+                        if (op == 12) {
+                            if (size == 1) {
+                                gen_op_neon_qdmulh_s16();
+                            } else {
+                                gen_op_neon_qdmulh_s32();
+                            }
+                        } else if (op == 13) {
+                            if (size == 1) {
+                                gen_op_neon_qrdmulh_s16();
+                            } else {
+                                gen_op_neon_qrdmulh_s32();
+                            }
+                        } else if (op & 1) {
+                            gen_op_neon_mul_f32();
+                        } else {
+                            switch (size) {
+                            case 0: gen_op_neon_mul_u8(); break;
+                            case 1: gen_op_neon_mul_u16(); break;
+                            case 2: gen_op_mul_T0_T1(); break;
+                            default: return 1;
+                            }
+                        }
+                        if (op < 8) {
+                            /* Accumulate.  */
+                            NEON_GET_REG(T1, rd, pass);
+                            switch (op) {
+                            case 0:
+                                gen_neon_add(size);
+                                break;
+                            case 1:
+                                gen_op_neon_add_f32();
+                                break;
+                            case 4:
+                                switch (size) {
+                                case 0: gen_op_neon_rsb_u8(); break;
+                                case 1: gen_op_neon_rsb_u16(); break;
+                                case 2: gen_op_rsbl_T0_T1(); break;
+                                default: return 1;
+                                }
+                                break;
+                            case 5:
+                                gen_op_neon_rsb_f32();
+                                break;
+                            default:
+                                abort();
+                            }
+                        }
+                        NEON_SET_REG(T0, rd, pass);
+                    }
+                    break;
+                case 2: /* VMLAL sclar */
+                case 3: /* VQDMLAL scalar */
+                case 6: /* VMLSL scalar */
+                case 7: /* VQDMLSL scalar */
+                case 10: /* VMULL scalar */
+                case 11: /* VQDMULL scalar */
+                    if (rd == rn) {
+                        /* Save overlapping operands before they are
+                           clobbered.  */
+                        NEON_GET_REG(T0, rn, 1);
+                        gen_neon_movl_scratch_T0(2);
+                    }
+                    gen_neon_get_scalar(size, rm);
+                    gen_op_movl_T2_T0();
+                    for (pass = 0; pass < 2; pass++) {
+                        if (pass != 0) {
+                            gen_op_movl_T0_T2();
+                        }
+                        if (pass != 0 && rd == rn) {
+                            gen_neon_movl_T1_scratch(2);
+                        } else {
+                            NEON_GET_REG(T1, rn, pass);
+                        }
+                        switch ((size << 1) | u) {
+                        case 0: gen_op_neon_mull_s8(); break;
+                        case 1: gen_op_neon_mull_u8(); break;
+                        case 2: gen_op_neon_mull_s16(); break;
+                        case 3: gen_op_neon_mull_u16(); break;
+                        case 4: gen_op_imull_T0_T1(); break;
+                        case 5: gen_op_mull_T0_T1(); break;
+                        default: abort();
+                        }
+                        if (op == 6 || op == 7) {
+                            switch (size) {
+                            case 0: gen_op_neon_negl_u16(); break;
+                            case 1: gen_op_neon_negl_u32(); break;
+                            case 2: gen_op_neon_negl_u64(); break;
+                            default: abort();
+                            }
+                        }
+                        gen_neon_movl_scratch_T0(0);
+                        gen_neon_movl_scratch_T1(1);
+                        NEON_GET_REG(T0, rd, pass * 2);
+                        NEON_GET_REG(T1, rd, pass * 2 + 1);
+                        switch (op) {
+                        case 2: case 6:
+                            switch (size) {
+                            case 0: gen_op_neon_addl_u16(); break;
+                            case 1: gen_op_neon_addl_u32(); break;
+                            case 2: gen_op_neon_addl_u64(); break;
+                            default: abort();
+                            }
+                            break;
+                        case 3: case 7:
+                            switch (size) {
+                            case 1:
+                                gen_op_neon_addl_saturate_s32();
+                                gen_op_neon_addl_saturate_s32();
+                                break;
+                            case 2:
+                                gen_op_neon_addl_saturate_s64();
+                                gen_op_neon_addl_saturate_s64();
+                                break;
+                            default: abort();
+                            }
+                            break;
+                        case 10:
+                            /* no-op */
+                            break;
+                        case 11:
+                            switch (size) {
+                            case 1: gen_op_neon_addl_saturate_s32(); break;
+                            case 2: gen_op_neon_addl_saturate_s64(); break;
+                            default: abort();
+                            }
+                            break;
+                        default:
+                            abort();
+                        }
+                        NEON_SET_REG(T0, rd, pass * 2);
+                        NEON_SET_REG(T1, rd, pass * 2 + 1);
+                    }
+                    break;
+                default: /* 14 and 15 are RESERVED */
+                    return 1;
+                }
+            }
+        } else { /* size == 3 */
+            if (!u) {
+                /* Extract.  */
+                int reg;
+                imm = (insn >> 8) & 0xf;
+                reg = rn;
+                count = q ? 4 : 2;
+                n = imm >> 2;
+                NEON_GET_REG(T0, reg, n);
+                for (pass = 0; pass < count; pass++) {
+                    n++;
+                    if (n > count) {
+                        reg = rm;
+                        n -= count;
+                    }
+                    if (imm & 3) {
+                        NEON_GET_REG(T1, reg, n);
+                        gen_op_neon_extract((insn << 3) & 0x1f);
+                    }
+                    /* ??? This is broken if rd and rm overlap */
+                    NEON_SET_REG(T0, rd, pass);
+                    if (imm & 3) {
+                        gen_op_movl_T0_T1();
+                    } else {
+                        NEON_GET_REG(T0, reg, n);
+                    }
+                }
+            } else if ((insn & (1 << 11)) == 0) {
+                /* Two register misc.  */
+                op = ((insn >> 12) & 0x30) | ((insn >> 7) & 0xf);
+                size = (insn >> 18) & 3;
+                switch (op) {
+                case 0: /* VREV64 */
+                    if (size == 3)
+                        return 1;
+                    for (pass = 0; pass < (q ? 2 : 1); pass++) {
+                        NEON_GET_REG(T0, rm, pass * 2);
+                        NEON_GET_REG(T1, rm, pass * 2 + 1);
+                        switch (size) {
+                        case 0: gen_op_rev_T0(); break;
+                        case 1: gen_op_revh_T0(); break;
+                        case 2: /* no-op */ break;
+                        default: abort();
+                        }
+                        NEON_SET_REG(T0, rd, pass * 2 + 1);
+                        if (size == 2) {
+                            NEON_SET_REG(T1, rd, pass * 2);
+                        } else {
+                            gen_op_movl_T0_T1();
+                            switch (size) {
+                            case 0: gen_op_rev_T0(); break;
+                            case 1: gen_op_revh_T0(); break;
+                            default: abort();
+                            }
+                            NEON_SET_REG(T0, rd, pass * 2);
+                        }
+                    }
+                    break;
+                case 4: case 5: /* VPADDL */
+                case 12: case 13: /* VPADAL */
+                    if (size < 2)
+                        goto elementwise;
+                    if (size == 3)
+                        return 1;
+                    for (pass = 0; pass < (q ? 2 : 1); pass++) {
+                        NEON_GET_REG(T0, rm, pass * 2);
+                        NEON_GET_REG(T1, rm, pass * 2 + 1);
+                        if (op & 1)
+                            gen_op_neon_paddl_u32();
+                        else
+                            gen_op_neon_paddl_s32();
+                        if (op >= 12) {
+                            /* Accumulate.  */
+                            gen_neon_movl_scratch_T0(0);
+                            gen_neon_movl_scratch_T1(1);
+
+                            NEON_GET_REG(T0, rd, pass * 2);
+                            NEON_GET_REG(T1, rd, pass * 2 + 1);
+                            gen_op_neon_addl_u64();
+                        }
+                        NEON_SET_REG(T0, rd, pass * 2);
+                        NEON_SET_REG(T1, rd, pass * 2 + 1);
+                    }
+                    break;
+                case 33: /* VTRN */
+                    if (size == 2) {
+                        for (n = 0; n < (q ? 4 : 2); n += 2) {
+                            NEON_GET_REG(T0, rm, n);
+                            NEON_GET_REG(T1, rd, n + 1);
+                            NEON_SET_REG(T1, rm, n);
+                            NEON_SET_REG(T0, rd, n + 1);
+                        }
+                    } else {
+                        goto elementwise;
+                    }
+                    break;
+                case 34: /* VUZP */
+                    /* Reg  Before       After
+                       Rd   A3 A2 A1 A0  B2 B0 A2 A0
+                       Rm   B3 B2 B1 B0  B3 B1 A3 A1
+                     */
+                    if (size == 3)
+                        return 1;
+                    gen_neon_unzip(rd, q, 0, size);
+                    gen_neon_unzip(rm, q, 4, size);
+                    if (q) {
+                        static int unzip_order_q[8] =
+                            {0, 2, 4, 6, 1, 3, 5, 7};
+                        for (n = 0; n < 8; n++) {
+                            int reg = (n < 4) ? rd : rm;
+                            gen_neon_movl_T0_scratch(unzip_order_q[n]);
+                            NEON_SET_REG(T0, reg, n % 4);
+                        }
+                    } else {
+                        static int unzip_order[4] =
+                            {0, 4, 1, 5};
+                        for (n = 0; n < 4; n++) {
+                            int reg = (n < 2) ? rd : rm;
+                            gen_neon_movl_T0_scratch(unzip_order[n]);
+                            NEON_SET_REG(T0, reg, n % 2);
+                        }
+                    }
+                    break;
+                case 35: /* VZIP */
+                    /* Reg  Before       After
+                       Rd   A3 A2 A1 A0  B1 A1 B0 A0
+                       Rm   B3 B2 B1 B0  B3 A3 B2 A2
+                     */
+                    if (size == 3)
+                        return 1;
+                    count = (q ? 4 : 2);
+                    for (n = 0; n < count; n++) {
+                        NEON_GET_REG(T0, rd, n);
+                        NEON_GET_REG(T1, rd, n);
+                        switch (size) {
+                        case 0: gen_op_neon_zip_u8(); break;
+                        case 1: gen_op_neon_zip_u16(); break;
+                        case 2: /* no-op */; break;
+                        default: abort();
+                        }
+                        gen_neon_movl_scratch_T0(n * 2);
+                        gen_neon_movl_scratch_T1(n * 2 + 1);
+                    }
+                    for (n = 0; n < count * 2; n++) {
+                        int reg = (n < count) ? rd : rm;
+                        gen_neon_movl_T0_scratch(n);
+                        NEON_SET_REG(T0, reg, n % count);
+                    }
+                    break;
+                case 36: case 37: /* VMOVN, VQMOVUN, VQMOVN */
+                    for (pass = 0; pass < 2; pass++) {
+                        if (rd == rm + 1) {
+                            n = 1 - pass;
+                        } else {
+                            n = pass;
+                        }
+                        NEON_GET_REG(T0, rm, n * 2);
+                        NEON_GET_REG(T1, rm, n * 2 + 1);
+                        if (op == 36 && q == 0) {
+                            switch (size) {
+                            case 0: gen_op_neon_narrow_u8(); break;
+                            case 1: gen_op_neon_narrow_u16(); break;
+                            case 2: /* no-op */ break;
+                            default: return 1;
+                            }
+                        } else if (q) {
+                            switch (size) {
+                            case 0: gen_op_neon_narrow_sat_u8(); break;
+                            case 1: gen_op_neon_narrow_sat_u16(); break;
+                            case 2: gen_op_neon_narrow_sat_u32(); break;
+                            default: return 1;
+                            }
+                        } else {
+                            switch (size) {
+                            case 0: gen_op_neon_narrow_sat_s8(); break;
+                            case 1: gen_op_neon_narrow_sat_s16(); break;
+                            case 2: gen_op_neon_narrow_sat_s32(); break;
+                            default: return 1;
+                            }
+                        }
+                        NEON_SET_REG(T0, rd, n);
+                    }
+                    break;
+                case 38: /* VSHLL */
+                    if (q)
+                        return 1;
+                    if (rm == rd) {
+                        NEON_GET_REG(T2, rm, 1);
+                    }
+                    for (pass = 0; pass < 2; pass++) {
+                        if (pass == 1 && rm == rd) {
+                            gen_op_movl_T0_T2();
+                        } else {
+                            NEON_GET_REG(T0, rm, pass);
+                        }
+                        switch (size) {
+                        case 0: gen_op_neon_widen_high_u8(); break;
+                        case 1: gen_op_neon_widen_high_u16(); break;
+                        case 2:
+                            gen_op_movl_T1_T0();
+                            gen_op_movl_T0_im(0);
+                            break;
+                        default: return 1;
+                        }
+                        NEON_SET_REG(T0, rd, pass * 2);
+                        NEON_SET_REG(T1, rd, pass * 2 + 1);
+                    }
+                    break;
+                default:
+                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));
+                        } else {
+                            NEON_GET_REG(T0, rm, pass);
+                        }
+                        switch (op) {
+                        case 1: /* VREV32 */
+                            switch (size) {
+                            case 0: gen_op_rev_T0(); break;
+                            case 1: gen_op_revh_T0(); break;
+                            default: return 1;
+                            }
+                            break;
+                        case 2: /* VREV16 */
+                            if (size != 0)
+                                return 1;
+                            gen_op_rev16_T0();
+                            break;
+                        case 4: case 5: /* VPADDL */
+                        case 12: case 13: /* VPADAL */
+                            switch ((size << 1) | (op & 1)) {
+                            case 0: gen_op_neon_paddl_s8(); break;
+                            case 1: gen_op_neon_paddl_u8(); break;
+                            case 2: gen_op_neon_paddl_s16(); break;
+                            case 3: gen_op_neon_paddl_u16(); break;
+                            default: abort();
+                            }
+                            if (op >= 12) {
+                                /* Accumulate */
+                                NEON_GET_REG(T1, rd, pass);
+                                switch (size) {
+                                case 0: gen_op_neon_add_u16(); break;
+                                case 1: gen_op_addl_T0_T1(); break;
+                                default: abort();
+                                }
+                            }
+                            break;
+                        case 8: /* CLS */
+                            switch (size) {
+                            case 0: gen_op_neon_cls_s8(); break;
+                            case 1: gen_op_neon_cls_s16(); break;
+                            case 2: gen_op_neon_cls_s32(); break;
+                            default: return 1;
+                            }
+                            break;
+                        case 9: /* CLZ */
+                            switch (size) {
+                            case 0: gen_op_neon_clz_u8(); break;
+                            case 1: gen_op_neon_clz_u16(); break;
+                            case 2: gen_op_clz_T0(); break;
+                            default: return 1;
+                            }
+                            break;
+                        case 10: /* CNT */
+                            if (size != 0)
+                                return 1;
+                            gen_op_neon_cnt_u8();
+                            break;
+                        case 11: /* VNOT */
+                            if (size != 0)
+                                return 1;
+                            gen_op_notl_T0();
+                            break;
+                        case 14: /* VQABS */
+                            switch (size) {
+                            case 0: gen_op_neon_qabs_s8(); break;
+                            case 1: gen_op_neon_qabs_s16(); break;
+                            case 2: gen_op_neon_qabs_s32(); break;
+                            default: return 1;
+                            }
+                            break;
+                        case 15: /* VQNEG */
+                            switch (size) {
+                            case 0: gen_op_neon_qneg_s8(); break;
+                            case 1: gen_op_neon_qneg_s16(); break;
+                            case 2: gen_op_neon_qneg_s32(); break;
+                            default: return 1;
+                            }
+                            break;
+                        case 16: case 19: /* VCGT #0, VCLE #0 */
+                            gen_op_movl_T1_im(0);
+                            switch(size) {
+                            case 0: gen_op_neon_cgt_s8(); break;
+                            case 1: gen_op_neon_cgt_s16(); break;
+                            case 2: gen_op_neon_cgt_s32(); break;
+                            default: return 1;
+                            }
+                            if (op == 19)
+                                gen_op_notl_T0();
+                            break;
+                        case 17: case 20: /* VCGE #0, VCLT #0 */
+                            gen_op_movl_T1_im(0);
+                            switch(size) {
+                            case 0: gen_op_neon_cge_s8(); break;
+                            case 1: gen_op_neon_cge_s16(); break;
+                            case 2: gen_op_neon_cge_s32(); break;
+                            default: return 1;
+                            }
+                            if (op == 20)
+                                gen_op_notl_T0();
+                            break;
+                        case 18: /* VCEQ #0 */
+                            gen_op_movl_T1_im(0);
+                            switch(size) {
+                            case 0: gen_op_neon_ceq_u8(); break;
+                            case 1: gen_op_neon_ceq_u16(); break;
+                            case 2: gen_op_neon_ceq_u32(); break;
+                            default: return 1;
+                            }
+                            break;
+                        case 22: /* VABS */
+                            switch(size) {
+                            case 0: gen_op_neon_abs_s8(); break;
+                            case 1: gen_op_neon_abs_s16(); break;
+                            case 2: gen_op_neon_abs_s32(); break;
+                            default: return 1;
+                            }
+                            break;
+                        case 23: /* VNEG */
+                            gen_op_movl_T1_im(0);
+                            switch(size) {
+                            case 0: gen_op_neon_rsb_u8(); break;
+                            case 1: gen_op_neon_rsb_u16(); break;
+                            case 2: gen_op_rsbl_T0_T1(); break;
+                            default: return 1;
+                            }
+                            break;
+                        case 24: case 27: /* Float VCGT #0, Float VCLE #0 */
+                            gen_op_movl_T1_im(0);
+                            gen_op_neon_cgt_f32();
+                            if (op == 27)
+                                gen_op_notl_T0();
+                            break;
+                        case 25: case 28: /* Float VCGE #0, Float VCLT #0 */
+                            gen_op_movl_T1_im(0);
+                            gen_op_neon_cge_f32();
+                            if (op == 28)
+                                gen_op_notl_T0();
+                            break;
+                        case 26: /* Float VCEQ #0 */
+                            gen_op_movl_T1_im(0);
+                            gen_op_neon_ceq_f32();
+                            break;
+                        case 30: /* Float VABS */
+                            gen_op_vfp_abss();
+                            break;
+                        case 31: /* Float VNEG */
+                            gen_op_vfp_negs();
+                            break;
+                        case 32: /* VSWP */
+                            NEON_GET_REG(T1, rd, pass);
+                            NEON_SET_REG(T1, rm, pass);
+                            break;
+                        case 33: /* VTRN */
+                            NEON_GET_REG(T1, rd, pass);
+                            switch (size) {
+                            case 0: gen_op_neon_trn_u8(); break;
+                            case 1: gen_op_neon_trn_u16(); break;
+                            case 2: abort();
+                            default: return 1;
+                            }
+                            NEON_SET_REG(T1, rm, pass);
+                            break;
+                        case 56: /* Integer VRECPE */
+                            gen_op_neon_recpe_u32();
+                            break;
+                        case 57: /* Integer VRSQRTE */
+                            gen_op_neon_rsqrte_u32();
+                            break;
+                        case 58: /* Float VRECPE */
+                            gen_op_neon_recpe_f32();
+                            break;
+                        case 59: /* Float VRSQRTE */
+                            gen_op_neon_rsqrte_f32();
+                            break;
+                        case 60: /* VCVT.F32.S32 */
+                            gen_op_vfp_tosizs();
+                            break;
+                        case 61: /* VCVT.F32.U32 */
+                            gen_op_vfp_touizs();
+                            break;
+                        case 62: /* VCVT.S32.F32 */
+                            gen_op_vfp_sitos();
+                            break;
+                        case 63: /* VCVT.U32.F32 */
+                            gen_op_vfp_uitos();
+                            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));
+                        } else {
+                            NEON_SET_REG(T0, rd, pass);
+                        }
+                    }
+                    break;
+                }
+            } else if ((insn & (1 << 10)) == 0) {
+                /* VTBL, VTBX.  */
+                n = (insn >> 5) & 0x18;
+                NEON_GET_REG(T1, rm, 0);
+                if (insn & (1 << 6)) {
+                    NEON_GET_REG(T0, rd, 0);
+                } else {
+                    gen_op_movl_T0_im(0);
+                }
+                gen_op_neon_tbl(rn, n);
+                gen_op_movl_T2_T0();
+                NEON_GET_REG(T1, rm, 1);
+                if (insn & (1 << 6)) {
+                    NEON_GET_REG(T0, rd, 0);
+                } else {
+                    gen_op_movl_T0_im(0);
+                }
+                gen_op_neon_tbl(rn, n);
+                NEON_SET_REG(T2, rd, 0);
+                NEON_SET_REG(T0, rd, 1);
+            } else if ((insn & 0x380) == 0) {
+                /* VDUP */
+                if (insn & (1 << 19)) {
+                    NEON_SET_REG(T0, rm, 1);
+                } else {
+                    NEON_SET_REG(T0, rm, 0);
+                }
+                if (insn & (1 << 16)) {
+                    gen_op_neon_dup_u8(((insn >> 17) & 3) * 8);
+                } else if (insn & (1 << 17)) {
+                    if ((insn >> 18) & 1)
+                        gen_op_neon_dup_high16();
+                    else
+                        gen_op_neon_dup_low16();
+                }
+                for (pass = 0; pass < (q ? 4 : 2); pass++) {
+                    NEON_SET_REG(T0, rd, pass);
+                }
+            } else {
+                return 1;
+            }
+        }
+    }
+    return 0;
+}
+
+static int disas_coproc_insn(CPUState * env, DisasContext *s, uint32_t insn)
+{
+    int cpnum;
+
+    cpnum = (insn >> 8) & 0xf;
+    if (arm_feature(env, ARM_FEATURE_XSCALE)
+	    && ((env->cp15.c15_cpar ^ 0x3fff) & (1 << cpnum)))
+	return 1;
+
+    switch (cpnum) {
+      case 0:
+      case 1:
+	if (arm_feature(env, ARM_FEATURE_IWMMXT)) {
+	    return disas_iwmmxt_insn(env, s, insn);
+	} else if (arm_feature(env, ARM_FEATURE_XSCALE)) {
+	    return disas_dsp_insn(env, s, insn);
+	}
+	return 1;
+    case 10:
+    case 11:
+	return disas_vfp_insn (env, s, insn);
+    case 15:
+	return disas_cp15_insn (env, s, insn);
+    default:
+	/* Unknown coprocessor.  See if the board has hooked it.  */
+	return disas_cp_insn (env, s, insn);
+    }
+}
+
 static void disas_arm_insn(CPUState * env, DisasContext *s)
 {
     unsigned int cond, insn, val, op1, i, shift, rm, rs, rn, rd, sh;
@@ -2209,12 +4652,137 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
     insn = ldl_code(s->pc);
     s->pc += 4;
 
+    /* M variants do not implement ARM mode.  */
+    if (IS_M(env))
+        goto illegal_op;
     cond = insn >> 28;
     if (cond == 0xf){
         /* Unconditional instructions.  */
+        if (((insn >> 25) & 7) == 1) {
+            /* NEON Data processing.  */
+            if (!arm_feature(env, ARM_FEATURE_NEON))
+                goto illegal_op;
+
+            if (disas_neon_data_insn(env, s, insn))
+                goto illegal_op;
+            return;
+        }
+        if ((insn & 0x0f100000) == 0x04000000) {
+            /* NEON load/store.  */
+            if (!arm_feature(env, ARM_FEATURE_NEON))
+                goto illegal_op;
+
+            if (disas_neon_ls_insn(env, s, insn))
+                goto illegal_op;
+            return;
+        }
         if ((insn & 0x0d70f000) == 0x0550f000)
             return; /* PLD */
-        else if ((insn & 0x0e000000) == 0x0a000000) {
+        else if ((insn & 0x0ffffdff) == 0x01010000) {
+            ARCH(6);
+            /* setend */
+            if (insn & (1 << 9)) {
+                /* BE8 mode not implemented.  */
+                goto illegal_op;
+            }
+            return;
+        } else if ((insn & 0x0fffff00) == 0x057ff000) {
+            switch ((insn >> 4) & 0xf) {
+            case 1: /* clrex */
+                ARCH(6K);
+                gen_op_clrex();
+                return;
+            case 4: /* dsb */
+            case 5: /* dmb */
+            case 6: /* isb */
+                ARCH(7);
+                /* We don't emulate caches so these are a no-op.  */
+                return;
+            default:
+                goto illegal_op;
+            }
+        } else if ((insn & 0x0e5fffe0) == 0x084d0500) {
+            /* srs */
+            uint32_t offset;
+            if (IS_USER(s))
+                goto illegal_op;
+            ARCH(6);
+            op1 = (insn & 0x1f);
+            if (op1 == (env->uncached_cpsr & CPSR_M)) {
+                gen_movl_T1_reg(s, 13);
+            } else {
+                gen_op_movl_T1_r13_banked(op1);
+            }
+            i = (insn >> 23) & 3;
+            switch (i) {
+            case 0: offset = -4; break; /* DA */
+            case 1: offset = -8; break; /* DB */
+            case 2: offset = 0; break; /* IA */
+            case 3: offset = 4; break; /* IB */
+            default: abort();
+            }
+            if (offset)
+                gen_op_addl_T1_im(offset);
+            gen_movl_T0_reg(s, 14);
+            gen_ldst(stl, s);
+            gen_op_movl_T0_cpsr();
+            gen_op_addl_T1_im(4);
+            gen_ldst(stl, s);
+            if (insn & (1 << 21)) {
+                /* Base writeback.  */
+                switch (i) {
+                case 0: offset = -8; break;
+                case 1: offset = -4; break;
+                case 2: offset = 4; break;
+                case 3: offset = 0; break;
+                default: abort();
+                }
+                if (offset)
+                    gen_op_addl_T1_im(offset);
+                if (op1 == (env->uncached_cpsr & CPSR_M)) {
+                    gen_movl_reg_T1(s, 13);
+                } else {
+                    gen_op_movl_r13_T1_banked(op1);
+                }
+            }
+        } else if ((insn & 0x0e5fffe0) == 0x081d0a00) {
+            /* rfe */
+            uint32_t offset;
+            if (IS_USER(s))
+                goto illegal_op;
+            ARCH(6);
+            rn = (insn >> 16) & 0xf;
+            gen_movl_T1_reg(s, rn);
+            i = (insn >> 23) & 3;
+            switch (i) {
+            case 0: offset = 0; break; /* DA */
+            case 1: offset = -4; break; /* DB */
+            case 2: offset = 4; break; /* IA */
+            case 3: offset = 8; break; /* IB */
+            default: abort();
+            }
+            if (offset)
+                gen_op_addl_T1_im(offset);
+            /* Load CPSR into T2 and PC into T0.  */
+            gen_ldst(ldl, s);
+            gen_op_movl_T2_T0();
+            gen_op_addl_T1_im(-4);
+            gen_ldst(ldl, s);
+            if (insn & (1 << 21)) {
+                /* Base writeback.  */
+                switch (i) {
+                case 0: offset = -4; break;
+                case 1: offset = 0; break;
+                case 2: offset = 8; break;
+                case 3: offset = 4; break;
+                default: abort();
+                }
+                if (offset)
+                    gen_op_addl_T1_im(offset);
+                gen_movl_reg_T1(s, rn);
+            }
+            gen_rfe(s);
+        } else if ((insn & 0x0e000000) == 0x0a000000) {
             /* branch link and change to thumb (blx <offset>) */
             int32_t offset;
 
@@ -2242,12 +4810,29 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
         } else if ((insn & 0x0f000010) == 0x0e000010) {
             /* Additional coprocessor register transfer.  */
         } else if ((insn & 0x0ff10010) == 0x01000000) {
+            uint32_t mask;
+            uint32_t val;
             /* cps (privileged) */
-        } else if ((insn & 0x0ffffdff) == 0x01010000) {
-            /* setend */
-            if (insn & (1 << 9)) {
-                /* BE8 mode not implemented.  */
-                goto illegal_op;
+            if (IS_USER(s))
+                return;
+            mask = val = 0;
+            if (insn & (1 << 19)) {
+                if (insn & (1 << 8))
+                    mask |= CPSR_A;
+                if (insn & (1 << 7))
+                    mask |= CPSR_I;
+                if (insn & (1 << 6))
+                    mask |= CPSR_F;
+                if (insn & (1 << 18))
+                    val |= mask;
+            }
+            if (insn & (1 << 14)) {
+                mask |= CPSR_M;
+                val |= (insn & 0x1f);
+            }
+            if (mask) {
+                gen_op_movl_T0_im(val);
+                gen_set_psr_T0(s, mask, 0);
             }
             return;
         }
@@ -2259,21 +4844,41 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
         s->condlabel = gen_new_label();
         gen_test_cc[cond ^ 1](s->condlabel);
         s->condjmp = 1;
-        //gen_test_cc[cond ^ 1]((long)s->tb, (long)s->pc);
-        //s->is_jmp = DISAS_JUMP_NEXT;
     }
     if ((insn & 0x0f900000) == 0x03000000) {
-        if ((insn & 0x0fb0f000) != 0x0320f000)
-            goto illegal_op;
-        /* CPSR = immediate */
-        val = insn & 0xff;
-        shift = ((insn >> 8) & 0xf) * 2;
-        if (shift)
-            val = (val >> shift) | (val << (32 - shift));
-        gen_op_movl_T0_im(val);
-        i = ((insn & (1 << 22)) != 0);
-        if (gen_set_psr_T0(s, msr_mask(s, (insn >> 16) & 0xf, i), i))
-            goto illegal_op;
+        if ((insn & (1 << 21)) == 0) {
+            ARCH(6T2);
+            rd = (insn >> 12) & 0xf;
+            val = ((insn >> 4) & 0xf000) | (insn & 0xfff);
+            if ((insn & (1 << 22)) == 0) {
+                /* MOVW */
+                gen_op_movl_T0_im(val);
+            } else {
+                /* MOVT */
+                gen_movl_T0_reg(s, rd);
+                gen_op_movl_T1_im(0xffff);
+                gen_op_andl_T0_T1();
+                gen_op_movl_T1_im(val << 16);
+                gen_op_orl_T0_T1();
+            }
+            gen_movl_reg_T0(s, rd);
+        } else {
+            if (((insn >> 12) & 0xf) != 0xf)
+                goto illegal_op;
+            if (((insn >> 16) & 0xf) == 0) {
+                gen_nop_hint(s, insn & 0xff);
+            } else {
+                /* CPSR = immediate */
+                val = insn & 0xff;
+                shift = ((insn >> 8) & 0xf) * 2;
+                if (shift)
+                    val = (val >> shift) | (val << (32 - shift));
+                gen_op_movl_T0_im(val);
+                i = ((insn & (1 << 22)) != 0);
+                if (gen_set_psr_T0(s, msr_mask(env, s, (insn >> 16) & 0xf, i), i))
+                    goto illegal_op;
+            }
+        }
     } else if ((insn & 0x0f900000) == 0x01000000
                && (insn & 0x00000090) != 0x00000090) {
         /* miscellaneous instructions */
@@ -2286,7 +4891,7 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
                 /* PSR = reg */
                 gen_movl_T0_reg(s, rm);
                 i = ((op1 & 2) != 0);
-                if (gen_set_psr_T0(s, msr_mask(s, (insn >> 16) & 0xf, i), i))
+                if (gen_set_psr_T0(s, msr_mask(env, s, (insn >> 16) & 0xf, i), i))
                     goto illegal_op;
             } else {
                 /* reg = PSR */
@@ -2351,6 +4956,7 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
             gen_movl_reg_T0(s, rd);
             break;
         case 7: /* bkpt */
+            gen_set_condexec(s);
             gen_op_movl_T0_im((long)s->pc - 4);
             gen_op_movl_reg_TN[0][15]();
             gen_op_bkpt();
@@ -2585,19 +5191,28 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
                     rn = (insn >> 12) & 0xf;
                     rs = (insn >> 8) & 0xf;
                     rm = (insn) & 0xf;
-                    if (((insn >> 22) & 3) == 0) {
+                    op1 = (insn >> 20) & 0xf;
+                    switch (op1) {
+                    case 0: case 1: case 2: case 3: case 6:
                         /* 32 bit mul */
                         gen_movl_T0_reg(s, rs);
                         gen_movl_T1_reg(s, rm);
                         gen_op_mul_T0_T1();
-                        if (insn & (1 << 21)) {
+                        if (insn & (1 << 22)) {
+                            /* Subtract (mls) */
+                            ARCH(6T2);
+                            gen_movl_T1_reg(s, rn);
+                            gen_op_rsbl_T0_T1();
+                        } else if (insn & (1 << 21)) {
+                            /* Add */
                             gen_movl_T1_reg(s, rn);
                             gen_op_addl_T0_T1();
                         }
                         if (insn & (1 << 20))
                             gen_op_logic_T0_cc();
                         gen_movl_reg_T0(s, rd);
-                    } else {
+                        break;
+                    default:
                         /* 64 bit mul */
                         gen_movl_T0_reg(s, rs);
                         gen_movl_T1_reg(s, rm);
@@ -2616,13 +5231,22 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
                             gen_op_logicq_cc();
                         gen_movl_reg_T0(s, rn);
                         gen_movl_reg_T1(s, rd);
+                        break;
                     }
                 } else {
                     rn = (insn >> 16) & 0xf;
                     rd = (insn >> 12) & 0xf;
                     if (insn & (1 << 23)) {
                         /* load/store exclusive */
-                        goto illegal_op;
+                        gen_movl_T1_reg(s, rn);
+                        if (insn & (1 << 20)) {
+                            gen_ldst(ldlex, s);
+                        } else {
+                            rm = insn & 0xf;
+                            gen_movl_T0_reg(s, rm);
+                            gen_ldst(stlex, s);
+                        }
+                        gen_movl_reg_T0(s, rd);
                     } else {
                         /* SWP instruction */
                         rm = (insn) & 0xf;
@@ -2708,8 +5332,227 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
             break;
         case 0x4:
         case 0x5:
+            goto do_ldst;
         case 0x6:
         case 0x7:
+            if (insn & (1 << 4)) {
+                ARCH(6);
+                /* Armv6 Media instructions.  */
+                rm = insn & 0xf;
+                rn = (insn >> 16) & 0xf;
+                rd = (insn >> 12) & 0xf;
+                rs = (insn >> 8) & 0xf;
+                switch ((insn >> 23) & 3) {
+                case 0: /* Parallel add/subtract.  */
+                    op1 = (insn >> 20) & 7;
+                    gen_movl_T0_reg(s, rn);
+                    gen_movl_T1_reg(s, rm);
+                    sh = (insn >> 5) & 7;
+                    if ((op1 & 3) == 0 || sh == 5 || sh == 6)
+                        goto illegal_op;
+                    gen_arm_parallel_addsub[op1][sh]();
+                    gen_movl_reg_T0(s, rd);
+                    break;
+                case 1:
+                    if ((insn & 0x00700020) == 0) {
+                        /* Hafword pack.  */
+                        gen_movl_T0_reg(s, rn);
+                        gen_movl_T1_reg(s, rm);
+                        shift = (insn >> 7) & 0x1f;
+                        if (shift)
+                            gen_op_shll_T1_im(shift);
+                        if (insn & (1 << 6))
+                            gen_op_pkhtb_T0_T1();
+                        else
+                            gen_op_pkhbt_T0_T1();
+                        gen_movl_reg_T0(s, rd);
+                    } else if ((insn & 0x00200020) == 0x00200000) {
+                        /* [us]sat */
+                        gen_movl_T1_reg(s, rm);
+                        shift = (insn >> 7) & 0x1f;
+                        if (insn & (1 << 6)) {
+                            if (shift == 0)
+                                shift = 31;
+                            gen_op_sarl_T1_im(shift);
+                        } else {
+                            gen_op_shll_T1_im(shift);
+                        }
+                        sh = (insn >> 16) & 0x1f;
+                        if (sh != 0) {
+                            if (insn & (1 << 22))
+                                gen_op_usat_T1(sh);
+                            else
+                                gen_op_ssat_T1(sh);
+                        }
+                        gen_movl_T1_reg(s, rd);
+                    } else if ((insn & 0x00300fe0) == 0x00200f20) {
+                        /* [us]sat16 */
+                        gen_movl_T1_reg(s, rm);
+                        sh = (insn >> 16) & 0x1f;
+                        if (sh != 0) {
+                            if (insn & (1 << 22))
+                                gen_op_usat16_T1(sh);
+                            else
+                                gen_op_ssat16_T1(sh);
+                        }
+                        gen_movl_T1_reg(s, rd);
+                    } else if ((insn & 0x00700fe0) == 0x00000fa0) {
+                        /* Select bytes.  */
+                        gen_movl_T0_reg(s, rn);
+                        gen_movl_T1_reg(s, rm);
+                        gen_op_sel_T0_T1();
+                        gen_movl_reg_T0(s, rd);
+                    } else if ((insn & 0x000003e0) == 0x00000060) {
+                        gen_movl_T1_reg(s, rm);
+                        shift = (insn >> 10) & 3;
+                        /* ??? In many cases it's not neccessary to do a
+                           rotate, a shift is sufficient.  */
+                        if (shift != 0)
+                            gen_op_rorl_T1_im(shift * 8);
+                        op1 = (insn >> 20) & 7;
+                        switch (op1) {
+                        case 0: gen_op_sxtb16_T1(); break;
+                        case 2: gen_op_sxtb_T1();   break;
+                        case 3: gen_op_sxth_T1();   break;
+                        case 4: gen_op_uxtb16_T1(); break;
+                        case 6: gen_op_uxtb_T1();   break;
+                        case 7: gen_op_uxth_T1();   break;
+                        default: goto illegal_op;
+                        }
+                        if (rn != 15) {
+                            gen_movl_T2_reg(s, rn);
+                            if ((op1 & 3) == 0) {
+                                gen_op_add16_T1_T2();
+                            } else {
+                                gen_op_addl_T1_T2();
+                            }
+                        }
+                        gen_movl_reg_T1(s, rd);
+                    } else if ((insn & 0x003f0f60) == 0x003f0f20) {
+                        /* rev */
+                        gen_movl_T0_reg(s, rm);
+                        if (insn & (1 << 22)) {
+                            if (insn & (1 << 7)) {
+                                gen_op_revsh_T0();
+                            } else {
+                                ARCH(6T2);
+                                gen_op_rbit_T0();
+                            }
+                        } else {
+                            if (insn & (1 << 7))
+                                gen_op_rev16_T0();
+                            else
+                                gen_op_rev_T0();
+                        }
+                        gen_movl_reg_T0(s, rd);
+                    } else {
+                        goto illegal_op;
+                    }
+                    break;
+                case 2: /* Multiplies (Type 3).  */
+                    gen_movl_T0_reg(s, rm);
+                    gen_movl_T1_reg(s, rs);
+                    if (insn & (1 << 20)) {
+                        /* Signed multiply most significant [accumulate].  */
+                        gen_op_imull_T0_T1();
+                        if (insn & (1 << 5))
+                            gen_op_roundqd_T0_T1();
+                        else
+                            gen_op_movl_T0_T1();
+                        if (rn != 15) {
+                            gen_movl_T1_reg(s, rn);
+                            if (insn & (1 << 6)) {
+                                gen_op_addl_T0_T1();
+                            } else {
+                                gen_op_rsbl_T0_T1();
+                            }
+                        }
+                        gen_movl_reg_T0(s, rd);
+                    } else {
+                        if (insn & (1 << 5))
+                            gen_op_swap_half_T1();
+                        gen_op_mul_dual_T0_T1();
+                        if (insn & (1 << 22)) {
+                            if (insn & (1 << 6)) {
+                                /* smlald */
+                                gen_op_addq_T0_T1_dual(rn, rd);
+                            } else {
+                                /* smlsld */
+                                gen_op_subq_T0_T1_dual(rn, rd);
+                            }
+                        } else {
+                            /* This addition cannot overflow.  */
+                            if (insn & (1 << 6)) {
+                                /* sm[ul]sd */
+                                gen_op_subl_T0_T1();
+                            } else {
+                                /* sm[ul]ad */
+                                gen_op_addl_T0_T1();
+                            }
+                            if (rn != 15)
+                              {
+                                gen_movl_T1_reg(s, rn);
+                                gen_op_addl_T0_T1_setq();
+                              }
+                            gen_movl_reg_T0(s, rd);
+                        }
+                    }
+                    break;
+                case 3:
+                    op1 = ((insn >> 17) & 0x38) | ((insn >> 5) & 7);
+                    switch (op1) {
+                    case 0: /* Unsigned sum of absolute differences.  */
+                            goto illegal_op;
+                        gen_movl_T0_reg(s, rm);
+                        gen_movl_T1_reg(s, rs);
+                        gen_op_usad8_T0_T1();
+                        if (rn != 15) {
+                            gen_movl_T1_reg(s, rn);
+                            gen_op_addl_T0_T1();
+                        }
+                        gen_movl_reg_T0(s, rd);
+                        break;
+                    case 0x20: case 0x24: case 0x28: case 0x2c:
+                        /* Bitfield insert/clear.  */
+                        ARCH(6T2);
+                        shift = (insn >> 7) & 0x1f;
+                        i = (insn >> 16) & 0x1f;
+                        i = i + 1 - shift;
+                        if (rm == 15) {
+                            gen_op_movl_T1_im(0);
+                        } else {
+                            gen_movl_T1_reg(s, rm);
+                        }
+                        if (i != 32) {
+                            gen_movl_T0_reg(s, rd);
+                            gen_op_bfi_T1_T0(shift, ((1u << i) - 1) << shift);
+                        }
+                        gen_movl_reg_T1(s, rd);
+                        break;
+                    case 0x12: case 0x16: case 0x1a: case 0x1e: /* sbfx */
+                    case 0x32: case 0x36: case 0x3a: case 0x3e: /* ubfx */
+                        gen_movl_T1_reg(s, rm);
+                        shift = (insn >> 7) & 0x1f;
+                        i = ((insn >> 16) & 0x1f) + 1;
+                        if (shift + i > 32)
+                            goto illegal_op;
+                        if (i < 32) {
+                            if (op1 & 0x20) {
+                                gen_op_ubfx_T1(shift, (1u << i) - 1);
+                            } else {
+                                gen_op_sbfx_T1(shift, i);
+                            }
+                        }
+                        gen_movl_reg_T1(s, rd);
+                        break;
+                    default:
+                        goto illegal_op;
+                    }
+                    break;
+                }
+                break;
+            }
+        do_ldst:
             /* Check for undefined extension instructions
              * per the ARM Bible IE:
              * xxxx 0111 1111 xxxx  xxxx xxxx 1111 xxxx
@@ -2913,49 +5756,18 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
         case 0xd:
         case 0xe:
             /* Coprocessor.  */
-            op1 = (insn >> 8) & 0xf;
-            if (arm_feature(env, ARM_FEATURE_XSCALE) &&
-                    ((env->cp15.c15_cpar ^ 0x3fff) & (1 << op1)))
-                goto illegal_op;
-            switch (op1) {
-            case 0 ... 1:
-                if (arm_feature(env, ARM_FEATURE_IWMMXT)) {
-                    if (disas_iwmmxt_insn(env, s, insn))
-                        goto illegal_op;
-                } else if (arm_feature(env, ARM_FEATURE_XSCALE)) {
-                    if (disas_dsp_insn(env, s, insn))
-                        goto illegal_op;
-                } else
-                    goto illegal_op;
-                break;
-            case 2 ... 9:
-            case 12 ... 14:
-                if (disas_cp_insn (env, s, insn))
-                    goto illegal_op;
-                break;
-            case 10:
-            case 11:
-                if (disas_vfp_insn (env, s, insn))
-                    goto illegal_op;
-                break;
-            case 15:
-                if (disas_cp15_insn (env, s, insn))
-                    goto illegal_op;
-                break;
-            default:
-                /* unknown coprocessor.  */
+            if (disas_coproc_insn(env, s, insn))
                 goto illegal_op;
-            }
             break;
         case 0xf:
             /* swi */
             gen_op_movl_T0_im((long)s->pc);
             gen_op_movl_reg_TN[0][15]();
-            gen_op_swi();
-            s->is_jmp = DISAS_JUMP;
+            s->is_jmp = DISAS_SWI;
             break;
         default:
         illegal_op:
+            gen_set_condexec(s);
             gen_op_movl_T0_im((long)s->pc - 4);
             gen_op_movl_reg_TN[0][15]();
             gen_op_undef_insn();
@@ -2965,12 +5777,1061 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
     }
 }
 
-static void disas_thumb_insn(DisasContext *s)
+/* Return true if this is a Thumb-2 logical op.  */
+static int
+thumb2_logic_op(int op)
+{
+    return (op < 8);
+}
+
+/* Generate code for a Thumb-2 data processing operation.  If CONDS is nonzero
+   then set condition code flags based on the result of the operation.
+   If SHIFTER_OUT is nonzero then set the carry flag for logical operations
+   to the high bit of T1.
+   Returns zero if the opcode is valid.  */
+
+static int
+gen_thumb2_data_op(DisasContext *s, int op, int conds, uint32_t shifter_out)
+{
+    int logic_cc;
+
+    logic_cc = 0;
+    switch (op) {
+    case 0: /* and */
+        gen_op_andl_T0_T1();
+        logic_cc = conds;
+        break;
+    case 1: /* bic */
+        gen_op_bicl_T0_T1();
+        logic_cc = conds;
+        break;
+    case 2: /* orr */
+        gen_op_orl_T0_T1();
+        logic_cc = conds;
+        break;
+    case 3: /* orn */
+        gen_op_notl_T1();
+        gen_op_orl_T0_T1();
+        logic_cc = conds;
+        break;
+    case 4: /* eor */
+        gen_op_xorl_T0_T1();
+        logic_cc = conds;
+        break;
+    case 8: /* add */
+        if (conds)
+            gen_op_addl_T0_T1_cc();
+        else
+            gen_op_addl_T0_T1();
+        break;
+    case 10: /* adc */
+        if (conds)
+            gen_op_adcl_T0_T1_cc();
+        else
+            gen_op_adcl_T0_T1();
+        break;
+    case 11: /* sbc */
+        if (conds)
+            gen_op_sbcl_T0_T1_cc();
+        else
+            gen_op_sbcl_T0_T1();
+        break;
+    case 13: /* sub */
+        if (conds)
+            gen_op_subl_T0_T1_cc();
+        else
+            gen_op_subl_T0_T1();
+        break;
+    case 14: /* rsb */
+        if (conds)
+            gen_op_rsbl_T0_T1_cc();
+        else
+            gen_op_rsbl_T0_T1();
+        break;
+    default: /* 5, 6, 7, 9, 12, 15. */
+        return 1;
+    }
+    if (logic_cc) {
+        gen_op_logic_T0_cc();
+        if (shifter_out)
+            gen_op_mov_CF_T1();
+    }
+    return 0;
+}
+
+/* Translate a 32-bit thumb instruction.  Returns nonzero if the instruction
+   is not legal.  */
+static int disas_thumb2_insn(CPUState *env, DisasContext *s, uint16_t insn_hw1)
+{
+    uint32_t insn, imm, shift, offset, addr;
+    uint32_t rd, rn, rm, rs;
+    int op;
+    int shiftop;
+    int conds;
+    int logic_cc;
+
+    if (!(arm_feature(env, ARM_FEATURE_THUMB2)
+          || arm_feature (env, ARM_FEATURE_M))) {
+        /* Thumb-1 cores may need to tread bl and blx as a pair of
+           16-bit instructions to get correct prefetch abort behavior.  */
+        insn = insn_hw1;
+        if ((insn & (1 << 12)) == 0) {
+            /* Second half of blx.  */
+            offset = ((insn & 0x7ff) << 1);
+            gen_movl_T0_reg(s, 14);
+            gen_op_movl_T1_im(offset);
+            gen_op_addl_T0_T1();
+            gen_op_movl_T1_im(0xfffffffc);
+            gen_op_andl_T0_T1();
+
+            addr = (uint32_t)s->pc;
+            gen_op_movl_T1_im(addr | 1);
+            gen_movl_reg_T1(s, 14);
+            gen_bx(s);
+            return 0;
+        }
+        if (insn & (1 << 11)) {
+            /* Second half of bl.  */
+            offset = ((insn & 0x7ff) << 1) | 1;
+            gen_movl_T0_reg(s, 14);
+            gen_op_movl_T1_im(offset);
+            gen_op_addl_T0_T1();
+
+            addr = (uint32_t)s->pc;
+            gen_op_movl_T1_im(addr | 1);
+            gen_movl_reg_T1(s, 14);
+            gen_bx(s);
+            return 0;
+        }
+        if ((s->pc & ~TARGET_PAGE_MASK) == 0) {
+            /* Instruction spans a page boundary.  Implement it as two
+               16-bit instructions in case the second half causes an
+               prefetch abort.  */
+            offset = ((int32_t)insn << 21) >> 9;
+            addr = s->pc + 2 + offset;
+            gen_op_movl_T0_im(addr);
+            gen_movl_reg_T0(s, 14);
+            return 0;
+        }
+        /* Fall through to 32-bit decode.  */
+    }
+
+    insn = lduw_code(s->pc);
+    s->pc += 2;
+    insn |= (uint32_t)insn_hw1 << 16;
+
+    if ((insn & 0xf800e800) != 0xf000e800) {
+        ARCH(6T2);
+    }
+
+    rn = (insn >> 16) & 0xf;
+    rs = (insn >> 12) & 0xf;
+    rd = (insn >> 8) & 0xf;
+    rm = insn & 0xf;
+    switch ((insn >> 25) & 0xf) {
+    case 0: case 1: case 2: case 3:
+        /* 16-bit instructions.  Should never happen.  */
+        abort();
+    case 4:
+        if (insn & (1 << 22)) {
+            /* Other load/store, table branch.  */
+            if (insn & 0x01200000) {
+                /* Load/store doubleword.  */
+                if (rn == 15) {
+                    gen_op_movl_T1_im(s->pc & ~3);
+                } else {
+                    gen_movl_T1_reg(s, rn);
+                }
+                offset = (insn & 0xff) * 4;
+                if ((insn & (1 << 23)) == 0)
+                    offset = -offset;
+                if (insn & (1 << 24)) {
+                    gen_op_addl_T1_im(offset);
+                    offset = 0;
+                }
+                if (insn & (1 << 20)) {
+                    /* ldrd */
+                    gen_ldst(ldl, s);
+                    gen_movl_reg_T0(s, rs);
+                    gen_op_addl_T1_im(4);
+                    gen_ldst(ldl, s);
+                    gen_movl_reg_T0(s, rd);
+                } else {
+                    /* strd */
+                    gen_movl_T0_reg(s, rs);
+                    gen_ldst(stl, s);
+                    gen_op_addl_T1_im(4);
+                    gen_movl_T0_reg(s, rd);
+                    gen_ldst(stl, s);
+                }
+                if (insn & (1 << 21)) {
+                    /* Base writeback.  */
+                    if (rn == 15)
+                        goto illegal_op;
+                    gen_op_addl_T1_im(offset - 4);
+                    gen_movl_reg_T1(s, rn);
+                }
+            } else if ((insn & (1 << 23)) == 0) {
+                /* Load/store exclusive word.  */
+                gen_movl_T0_reg(s, rd);
+                gen_movl_T1_reg(s, rn);
+                if (insn & (1 << 20)) {
+                    gen_ldst(ldlex, s);
+                } else {
+                    gen_ldst(stlex, s);
+                }
+                gen_movl_reg_T0(s, rd);
+            } else if ((insn & (1 << 6)) == 0) {
+                /* Table Branch.  */
+                if (rn == 15) {
+                    gen_op_movl_T1_im(s->pc);
+                } else {
+                    gen_movl_T1_reg(s, rn);
+                }
+                gen_movl_T2_reg(s, rm);
+                gen_op_addl_T1_T2();
+                if (insn & (1 << 4)) {
+                    /* tbh */
+                    gen_op_addl_T1_T2();
+                    gen_ldst(lduw, s);
+                } else { /* tbb */
+                    gen_ldst(ldub, s);
+                }
+                gen_op_jmp_T0_im(s->pc);
+                s->is_jmp = DISAS_JUMP;
+            } else {
+                /* Load/store exclusive byte/halfword/doubleword.  */
+                op = (insn >> 4) & 0x3;
+                gen_movl_T1_reg(s, rn);
+                if (insn & (1 << 20)) {
+                    switch (op) {
+                    case 0:
+                        gen_ldst(ldbex, s);
+                        break;
+                    case 1:
+                        gen_ldst(ldwex, s);
+                        break;
+                    case 3:
+                        gen_ldst(ldqex, s);
+                        gen_movl_reg_T1(s, rd);
+                        break;
+                    default:
+                        goto illegal_op;
+                    }
+                    gen_movl_reg_T0(s, rs);
+                } else {
+                    gen_movl_T0_reg(s, rs);
+                    switch (op) {
+                    case 0:
+                        gen_ldst(stbex, s);
+                        break;
+                    case 1:
+                        gen_ldst(stwex, s);
+                        break;
+                    case 3:
+                        gen_movl_T2_reg(s, rd);
+                        gen_ldst(stqex, s);
+                        break;
+                    default:
+                        goto illegal_op;
+                    }
+                    gen_movl_reg_T0(s, rm);
+                }
+            }
+        } else {
+            /* Load/store multiple, RFE, SRS.  */
+            if (((insn >> 23) & 1) == ((insn >> 24) & 1)) {
+                /* Not available in user mode.  */
+                if (!IS_USER(s))
+                    goto illegal_op;
+                if (insn & (1 << 20)) {
+                    /* rfe */
+                    gen_movl_T1_reg(s, rn);
+                    if (insn & (1 << 24)) {
+                        gen_op_addl_T1_im(4);
+                    } else {
+                        gen_op_addl_T1_im(-4);
+                    }
+                    /* Load CPSR into T2 and PC into T0.  */
+                    gen_ldst(ldl, s);
+                    gen_op_movl_T2_T0();
+                    gen_op_addl_T1_im(-4);
+                    gen_ldst(ldl, s);
+                    if (insn & (1 << 21)) {
+                        /* Base writeback.  */
+                        if (insn & (1 << 24))
+                            gen_op_addl_T1_im(8);
+                        gen_movl_reg_T1(s, rn);
+                    }
+                    gen_rfe(s);
+                } else {
+                    /* srs */
+                    op = (insn & 0x1f);
+                    if (op == (env->uncached_cpsr & CPSR_M)) {
+                        gen_movl_T1_reg(s, 13);
+                    } else {
+                        gen_op_movl_T1_r13_banked(op);
+                    }
+                    if ((insn & (1 << 24)) == 0) {
+                        gen_op_addl_T1_im(-8);
+                    }
+                    gen_movl_T0_reg(s, 14);
+                    gen_ldst(stl, s);
+                    gen_op_movl_T0_cpsr();
+                    gen_op_addl_T1_im(4);
+                    gen_ldst(stl, s);
+                    if (insn & (1 << 21)) {
+                        if ((insn & (1 << 24)) == 0) {
+                            gen_op_addl_T1_im(-4);
+                        } else {
+                            gen_op_addl_T1_im(4);
+                        }
+                        if (op == (env->uncached_cpsr & CPSR_M)) {
+                            gen_movl_reg_T1(s, 13);
+                        } else {
+                            gen_op_movl_r13_T1_banked(op);
+                        }
+                    }
+                }
+            } else {
+                int i;
+                /* Load/store multiple.  */
+                gen_movl_T1_reg(s, rn);
+                offset = 0;
+                for (i = 0; i < 16; i++) {
+                    if (insn & (1 << i))
+                        offset += 4;
+                }
+                if (insn & (1 << 24)) {
+                    gen_op_addl_T1_im(-offset);
+                }
+
+                for (i = 0; i < 16; i++) {
+                    if ((insn & (1 << i)) == 0)
+                        continue;
+                    if (insn & (1 << 20)) {
+                        /* Load.  */
+                        gen_ldst(ldl, s);
+                        if (i == 15) {
+                            gen_bx(s);
+                        } else {
+                            gen_movl_reg_T0(s, i);
+                        }
+                    } else {
+                        /* Store.  */
+                        gen_movl_T0_reg(s, i);
+                        gen_ldst(stl, s);
+                    }
+                    gen_op_addl_T1_im(4);
+                }
+                if (insn & (1 << 21)) {
+                    /* Base register writeback.  */
+                    if (insn & (1 << 24)) {
+                        gen_op_addl_T1_im(-offset);
+                    }
+                    /* Fault if writeback register is in register list.  */
+                    if (insn & (1 << rn))
+                        goto illegal_op;
+                    gen_movl_reg_T1(s, rn);
+                }
+            }
+        }
+        break;
+    case 5: /* Data processing register constant shift.  */
+        if (rn == 15)
+            gen_op_movl_T0_im(0);
+        else
+            gen_movl_T0_reg(s, rn);
+        gen_movl_T1_reg(s, rm);
+        op = (insn >> 21) & 0xf;
+        shiftop = (insn >> 4) & 3;
+        shift = ((insn >> 6) & 3) | ((insn >> 10) & 0x1c);
+        conds = (insn & (1 << 20)) != 0;
+        logic_cc = (conds && thumb2_logic_op(op));
+        if (shift != 0) {
+            if (logic_cc) {
+                gen_shift_T1_im_cc[shiftop](shift);
+            } else {
+                gen_shift_T1_im[shiftop](shift);
+            }
+        } else if (shiftop != 0) {
+            if (logic_cc) {
+                gen_shift_T1_0_cc[shiftop]();
+            } else {
+                gen_shift_T1_0[shiftop]();
+            }
+        }
+        if (gen_thumb2_data_op(s, op, conds, 0))
+            goto illegal_op;
+        if (rd != 15)
+            gen_movl_reg_T0(s, rd);
+        break;
+    case 13: /* Misc data processing.  */
+        op = ((insn >> 22) & 6) | ((insn >> 7) & 1);
+        if (op < 4 && (insn & 0xf000) != 0xf000)
+            goto illegal_op;
+        switch (op) {
+        case 0: /* Register controlled shift.  */
+            gen_movl_T0_reg(s, rm);
+            gen_movl_T1_reg(s, rn);
+            if ((insn & 0x70) != 0)
+                goto illegal_op;
+            op = (insn >> 21) & 3;
+            if (insn & (1 << 20)) {
+                gen_shift_T1_T0_cc[op]();
+                gen_op_logic_T1_cc();
+            } else {
+                gen_shift_T1_T0[op]();
+            }
+            gen_movl_reg_T1(s, rd);
+            break;
+        case 1: /* Sign/zero extend.  */
+            gen_movl_T1_reg(s, rm);
+            shift = (insn >> 4) & 3;
+            /* ??? In many cases it's not neccessary to do a
+               rotate, a shift is sufficient.  */
+            if (shift != 0)
+                gen_op_rorl_T1_im(shift * 8);
+            op = (insn >> 20) & 7;
+            switch (op) {
+            case 0: gen_op_sxth_T1();   break;
+            case 1: gen_op_uxth_T1();   break;
+            case 2: gen_op_sxtb16_T1(); break;
+            case 3: gen_op_uxtb16_T1(); break;
+            case 4: gen_op_sxtb_T1();   break;
+            case 5: gen_op_uxtb_T1();   break;
+            default: goto illegal_op;
+            }
+            if (rn != 15) {
+                gen_movl_T2_reg(s, rn);
+                if ((op >> 1) == 1) {
+                    gen_op_add16_T1_T2();
+                } else {
+                    gen_op_addl_T1_T2();
+                }
+            }
+            gen_movl_reg_T1(s, rd);
+            break;
+        case 2: /* SIMD add/subtract.  */
+            op = (insn >> 20) & 7;
+            shift = (insn >> 4) & 7;
+            if ((op & 3) == 3 || (shift & 3) == 3)
+                goto illegal_op;
+            gen_movl_T0_reg(s, rn);
+            gen_movl_T1_reg(s, rm);
+            gen_thumb2_parallel_addsub[op][shift]();
+            gen_movl_reg_T0(s, rd);
+            break;
+        case 3: /* Other data processing.  */
+            op = ((insn >> 17) & 0x38) | ((insn >> 4) & 7);
+            if (op < 4) {
+                /* Saturating add/subtract.  */
+                gen_movl_T0_reg(s, rm);
+                gen_movl_T1_reg(s, rn);
+                if (op & 2)
+                    gen_op_double_T1_saturate();
+                if (op & 1)
+                    gen_op_subl_T0_T1_saturate();
+                else
+                    gen_op_addl_T0_T1_saturate();
+            } else {
+                gen_movl_T0_reg(s, rn);
+                switch (op) {
+                case 0x0a: /* rbit */
+                    gen_op_rbit_T0();
+                    break;
+                case 0x08: /* rev */
+                    gen_op_rev_T0();
+                    break;
+                case 0x09: /* rev16 */
+                    gen_op_rev16_T0();
+                    break;
+                case 0x0b: /* revsh */
+                    gen_op_revsh_T0();
+                    break;
+                case 0x10: /* sel */
+                    gen_movl_T1_reg(s, rm);
+                    gen_op_sel_T0_T1();
+                    break;
+                case 0x18: /* clz */
+                    gen_op_clz_T0();
+                    break;
+                default:
+                    goto illegal_op;
+                }
+            }
+            gen_movl_reg_T0(s, rd);
+            break;
+        case 4: case 5: /* 32-bit multiply.  Sum of absolute differences.  */
+            op = (insn >> 4) & 0xf;
+            gen_movl_T0_reg(s, rn);
+            gen_movl_T1_reg(s, rm);
+            switch ((insn >> 20) & 7) {
+            case 0: /* 32 x 32 -> 32 */
+                gen_op_mul_T0_T1();
+                if (rs != 15) {
+                    gen_movl_T1_reg(s, rs);
+                    if (op)
+                        gen_op_rsbl_T0_T1();
+                    else
+                        gen_op_addl_T0_T1();
+                }
+                gen_movl_reg_T0(s, rd);
+                break;
+            case 1: /* 16 x 16 -> 32 */
+                gen_mulxy(op & 2, op & 1);
+                if (rs != 15) {
+                    gen_movl_T1_reg(s, rs);
+                    gen_op_addl_T0_T1_setq();
+                }
+                gen_movl_reg_T0(s, rd);
+                break;
+            case 2: /* Dual multiply add.  */
+            case 4: /* Dual multiply subtract.  */
+                if (op)
+                    gen_op_swap_half_T1();
+                gen_op_mul_dual_T0_T1();
+                /* This addition cannot overflow.  */
+                if (insn & (1 << 22)) {
+                    gen_op_subl_T0_T1();
+                } else {
+                    gen_op_addl_T0_T1();
+                }
+                if (rs != 15)
+                  {
+                    gen_movl_T1_reg(s, rs);
+                    gen_op_addl_T0_T1_setq();
+                  }
+                gen_movl_reg_T0(s, rd);
+                break;
+            case 3: /* 32 * 16 -> 32msb */
+                if (op)
+                    gen_op_sarl_T1_im(16);
+                else
+                    gen_op_sxth_T1();
+                gen_op_imulw_T0_T1();
+                if (rs != 15)
+                  {
+                    gen_movl_T1_reg(s, rs);
+                    gen_op_addl_T0_T1_setq();
+                  }
+                gen_movl_reg_T0(s, rd);
+                break;
+            case 5: case 6: /* 32 * 32 -> 32msb */
+                gen_op_imull_T0_T1();
+                if (insn & (1 << 5))
+                    gen_op_roundqd_T0_T1();
+                else
+                    gen_op_movl_T0_T1();
+                if (rs != 15) {
+                    gen_movl_T1_reg(s, rs);
+                    if (insn & (1 << 21)) {
+                        gen_op_addl_T0_T1();
+                    } else {
+                        gen_op_rsbl_T0_T1();
+                    }
+                }
+                gen_movl_reg_T0(s, rd);
+                break;
+            case 7: /* Unsigned sum of absolute differences.  */
+                gen_op_usad8_T0_T1();
+                if (rs != 15) {
+                    gen_movl_T1_reg(s, rs);
+                    gen_op_addl_T0_T1();
+                }
+                gen_movl_reg_T0(s, rd);
+                break;
+            }
+            break;
+        case 6: case 7: /* 64-bit multiply, Divide.  */
+            op = ((insn >> 4) & 0xf) | ((insn >> 16) & 0x70);
+            gen_movl_T0_reg(s, rn);
+            gen_movl_T1_reg(s, rm);
+            if ((op & 0x50) == 0x10) {
+                /* sdiv, udiv */
+                if (!arm_feature(env, ARM_FEATURE_DIV))
+                    goto illegal_op;
+                if (op & 0x20)
+                    gen_op_udivl_T0_T1();
+                else
+                    gen_op_sdivl_T0_T1();
+                gen_movl_reg_T0(s, rd);
+            } else if ((op & 0xe) == 0xc) {
+                /* Dual multiply accumulate long.  */
+                if (op & 1)
+                    gen_op_swap_half_T1();
+                gen_op_mul_dual_T0_T1();
+                if (op & 0x10) {
+                    gen_op_subl_T0_T1();
+                } else {
+                    gen_op_addl_T0_T1();
+                }
+                gen_op_signbit_T1_T0();
+                gen_op_addq_T0_T1(rs, rd);
+                gen_movl_reg_T0(s, rs);
+                gen_movl_reg_T1(s, rd);
+            } else {
+                if (op & 0x20) {
+                    /* Unsigned 64-bit multiply  */
+                    gen_op_mull_T0_T1();
+                } else {
+                    if (op & 8) {
+                        /* smlalxy */
+                        gen_mulxy(op & 2, op & 1);
+                        gen_op_signbit_T1_T0();
+                    } else {
+                        /* Signed 64-bit multiply  */
+                        gen_op_imull_T0_T1();
+                    }
+                }
+                if (op & 4) {
+                    /* umaal */
+                    gen_op_addq_lo_T0_T1(rs);
+                    gen_op_addq_lo_T0_T1(rd);
+                } else if (op & 0x40) {
+                    /* 64-bit accumulate.  */
+                    gen_op_addq_T0_T1(rs, rd);
+                }
+                gen_movl_reg_T0(s, rs);
+                gen_movl_reg_T1(s, rd);
+            }
+            break;
+        }
+        break;
+    case 6: case 7: case 14: case 15:
+        /* Coprocessor.  */
+        if (((insn >> 24) & 3) == 3) {
+            /* Translate into the equivalent ARM encoding.  */
+            insn = (insn & 0xe2ffffff) | ((insn & (1 << 28)) >> 4);
+            if (disas_neon_data_insn(env, s, insn))
+                goto illegal_op;
+        } else {
+            if (insn & (1 << 28))
+                goto illegal_op;
+            if (disas_coproc_insn (env, s, insn))
+                goto illegal_op;
+        }
+        break;
+    case 8: case 9: case 10: case 11:
+        if (insn & (1 << 15)) {
+            /* Branches, misc control.  */
+            if (insn & 0x5000) {
+                /* Unconditional branch.  */
+                /* signextend(hw1[10:0]) -> offset[:12].  */
+                offset = ((int32_t)insn << 5) >> 9 & ~(int32_t)0xfff;
+                /* hw1[10:0] -> offset[11:1].  */
+                offset |= (insn & 0x7ff) << 1;
+                /* (~hw2[13, 11] ^ offset[24]) -> offset[23,22]
+                   offset[24:22] already have the same value because of the
+                   sign extension above.  */
+                offset ^= ((~insn) & (1 << 13)) << 10;
+                offset ^= ((~insn) & (1 << 11)) << 11;
+
+                addr = s->pc;
+                if (insn & (1 << 14)) {
+                    /* Branch and link.  */
+                    gen_op_movl_T1_im(addr | 1);
+                    gen_movl_reg_T1(s, 14);
+                }
+
+                addr += offset;
+                if (insn & (1 << 12)) {
+                    /* b/bl */
+                    gen_jmp(s, addr);
+                } else {
+                    /* blx */
+                    addr &= ~(uint32_t)2;
+                    gen_op_movl_T0_im(addr);
+                    gen_bx(s);
+                }
+            } else if (((insn >> 23) & 7) == 7) {
+                /* Misc control */
+                if (insn & (1 << 13))
+                    goto illegal_op;
+
+                if (insn & (1 << 26)) {
+                    /* Secure monitor call (v6Z) */
+                    goto illegal_op; /* not implemented.  */
+                } else {
+                    op = (insn >> 20) & 7;
+                    switch (op) {
+                    case 0: /* msr cpsr.  */
+                        if (IS_M(env)) {
+                            gen_op_v7m_msr_T0(insn & 0xff);
+                            gen_movl_reg_T0(s, rn);
+                            gen_lookup_tb(s);
+                            break;
+                        }
+                        /* fall through */
+                    case 1: /* msr spsr.  */
+                        if (IS_M(env))
+                            goto illegal_op;
+                        gen_movl_T0_reg(s, rn);
+                        if (gen_set_psr_T0(s,
+                              msr_mask(env, s, (insn >> 8) & 0xf, op == 1),
+                              op == 1))
+                            goto illegal_op;
+                        break;
+                    case 2: /* cps, nop-hint.  */
+                        if (((insn >> 8) & 7) == 0) {
+                            gen_nop_hint(s, insn & 0xff);
+                        }
+                        /* Implemented as NOP in user mode.  */
+                        if (IS_USER(s))
+                            break;
+                        offset = 0;
+                        imm = 0;
+                        if (insn & (1 << 10)) {
+                            if (insn & (1 << 7))
+                                offset |= CPSR_A;
+                            if (insn & (1 << 6))
+                                offset |= CPSR_I;
+                            if (insn & (1 << 5))
+                                offset |= CPSR_F;
+                            if (insn & (1 << 9))
+                                imm = CPSR_A | CPSR_I | CPSR_F;
+                        }
+                        if (insn & (1 << 8)) {
+                            offset |= 0x1f;
+                            imm |= (insn & 0x1f);
+                        }
+                        if (offset) {
+                            gen_op_movl_T0_im(imm);
+                            gen_set_psr_T0(s, offset, 0);
+                        }
+                        break;
+                    case 3: /* Special control operations.  */
+                        op = (insn >> 4) & 0xf;
+                        switch (op) {
+                        case 2: /* clrex */
+                            gen_op_clrex();
+                            break;
+                        case 4: /* dsb */
+                        case 5: /* dmb */
+                        case 6: /* isb */
+                            /* These execute as NOPs.  */
+                            ARCH(7);
+                            break;
+                        default:
+                            goto illegal_op;
+                        }
+                        break;
+                    case 4: /* bxj */
+                        /* Trivial implementation equivalent to bx.  */
+                        gen_movl_T0_reg(s, rn);
+                        gen_bx(s);
+                        break;
+                    case 5: /* Exception return.  */
+                        /* Unpredictable in user mode.  */
+                        goto illegal_op;
+                    case 6: /* mrs cpsr.  */
+                        if (IS_M(env)) {
+                            gen_op_v7m_mrs_T0(insn & 0xff);
+                        } else {
+                            gen_op_movl_T0_cpsr();
+                        }
+                        gen_movl_reg_T0(s, rd);
+                        break;
+                    case 7: /* mrs spsr.  */
+                        /* Not accessible in user mode.  */
+                        if (IS_USER(s) || IS_M(env))
+                            goto illegal_op;
+                        gen_op_movl_T0_spsr();
+                        gen_movl_reg_T0(s, rd);
+                        break;
+                    }
+                }
+            } else {
+                /* Conditional branch.  */
+                op = (insn >> 22) & 0xf;
+                /* Generate a conditional jump to next instruction.  */
+                s->condlabel = gen_new_label();
+                gen_test_cc[op ^ 1](s->condlabel);
+                s->condjmp = 1;
+
+                /* offset[11:1] = insn[10:0] */
+                offset = (insn & 0x7ff) << 1;
+                /* offset[17:12] = insn[21:16].  */
+                offset |= (insn & 0x003f0000) >> 4;
+                /* offset[31:20] = insn[26].  */
+                offset |= ((int32_t)((insn << 5) & 0x80000000)) >> 11;
+                /* offset[18] = insn[13].  */
+                offset |= (insn & (1 << 13)) << 5;
+                /* offset[19] = insn[11].  */
+                offset |= (insn & (1 << 11)) << 8;
+
+                /* jump to the offset */
+                addr = s->pc + offset;
+                gen_jmp(s, addr);
+            }
+        } else {
+            /* Data processing immediate.  */
+            if (insn & (1 << 25)) {
+                if (insn & (1 << 24)) {
+                    if (insn & (1 << 20))
+                        goto illegal_op;
+                    /* Bitfield/Saturate.  */
+                    op = (insn >> 21) & 7;
+                    imm = insn & 0x1f;
+                    shift = ((insn >> 6) & 3) | ((insn >> 10) & 0x1c);
+                    if (rn == 15)
+                        gen_op_movl_T1_im(0);
+                    else
+                        gen_movl_T1_reg(s, rn);
+                    switch (op) {
+                    case 2: /* Signed bitfield extract.  */
+                        imm++;
+                        if (shift + imm > 32)
+                            goto illegal_op;
+                        if (imm < 32)
+                            gen_op_sbfx_T1(shift, imm);
+                        break;
+                    case 6: /* Unsigned bitfield extract.  */
+                        imm++;
+                        if (shift + imm > 32)
+                            goto illegal_op;
+                        if (imm < 32)
+                            gen_op_ubfx_T1(shift, (1u << imm) - 1);
+                        break;
+                    case 3: /* Bitfield insert/clear.  */
+                        if (imm < shift)
+                            goto illegal_op;
+                        imm = imm + 1 - shift;
+                        if (imm != 32) {
+                            gen_movl_T0_reg(s, rd);
+                            gen_op_bfi_T1_T0(shift, ((1u << imm) - 1) << shift);
+                        }
+                        break;
+                    case 7:
+                        goto illegal_op;
+                    default: /* Saturate.  */
+                        gen_movl_T1_reg(s, rn);
+                        if (shift) {
+                            if (op & 1)
+                                gen_op_sarl_T1_im(shift);
+                            else
+                                gen_op_shll_T1_im(shift);
+                        }
+                        if (op & 4) {
+                            /* Unsigned.  */
+                            gen_op_ssat_T1(imm);
+                            if ((op & 1) && shift == 0)
+                                gen_op_usat16_T1(imm);
+                            else
+                                gen_op_usat_T1(imm);
+                        } else {
+                            /* Signed.  */
+                            gen_op_ssat_T1(imm);
+                            if ((op & 1) && shift == 0)
+                                gen_op_ssat16_T1(imm);
+                            else
+                                gen_op_ssat_T1(imm);
+                        }
+                        break;
+                    }
+                    gen_movl_reg_T1(s, rd);
+                } else {
+                    imm = ((insn & 0x04000000) >> 15)
+                          | ((insn & 0x7000) >> 4) | (insn & 0xff);
+                    if (insn & (1 << 22)) {
+                        /* 16-bit immediate.  */
+                        imm |= (insn >> 4) & 0xf000;
+                        if (insn & (1 << 23)) {
+                            /* movt */
+                            gen_movl_T0_reg(s, rd);
+                            gen_op_movtop_T0_im(imm << 16);
+                        } else {
+                            /* movw */
+                            gen_op_movl_T0_im(imm);
+                        }
+                    } else {
+                        /* Add/sub 12-bit immediate.  */
+                        if (rn == 15) {
+                            addr = s->pc & ~(uint32_t)3;
+                            if (insn & (1 << 23))
+                                addr -= imm;
+                            else
+                                addr += imm;
+                            gen_op_movl_T0_im(addr);
+                        } else {
+                            gen_movl_T0_reg(s, rn);
+                            gen_op_movl_T1_im(imm);
+                            if (insn & (1 << 23))
+                                gen_op_subl_T0_T1();
+                            else
+                                gen_op_addl_T0_T1();
+                        }
+                    }
+                    gen_movl_reg_T0(s, rd);
+                }
+            } else {
+                int shifter_out = 0;
+                /* modified 12-bit immediate.  */
+                shift = ((insn & 0x04000000) >> 23) | ((insn & 0x7000) >> 12);
+                imm = (insn & 0xff);
+                switch (shift) {
+                case 0: /* XY */
+                    /* Nothing to do.  */
+                    break;
+                case 1: /* 00XY00XY */
+                    imm |= imm << 16;
+                    break;
+                case 2: /* XY00XY00 */
+                    imm |= imm << 16;
+                    imm <<= 8;
+                    break;
+                case 3: /* XYXYXYXY */
+                    imm |= imm << 16;
+                    imm |= imm << 8;
+                    break;
+                default: /* Rotated constant.  */
+                    shift = (shift << 1) | (imm >> 7);
+                    imm |= 0x80;
+                    imm = imm << (32 - shift);
+                    shifter_out = 1;
+                    break;
+                }
+                gen_op_movl_T1_im(imm);
+                rn = (insn >> 16) & 0xf;
+                if (rn == 15)
+                    gen_op_movl_T0_im(0);
+                else
+                    gen_movl_T0_reg(s, rn);
+                op = (insn >> 21) & 0xf;
+                if (gen_thumb2_data_op(s, op, (insn & (1 << 20)) != 0,
+                                       shifter_out))
+                    goto illegal_op;
+                rd = (insn >> 8) & 0xf;
+                if (rd != 15) {
+                    gen_movl_reg_T0(s, rd);
+                }
+            }
+        }
+        break;
+    case 12: /* Load/store single data item.  */
+        {
+        int postinc = 0;
+        int writeback = 0;
+        if ((insn & 0x01100000) == 0x01000000) {
+            if (disas_neon_ls_insn(env, s, insn))
+                goto illegal_op;
+            break;
+        }
+        if (rn == 15) {
+            /* PC relative.  */
+            /* s->pc has already been incremented by 4.  */
+            imm = s->pc & 0xfffffffc;
+            if (insn & (1 << 23))
+                imm += insn & 0xfff;
+            else
+                imm -= insn & 0xfff;
+            gen_op_movl_T1_im(imm);
+        } else {
+            gen_movl_T1_reg(s, rn);
+            if (insn & (1 << 23)) {
+                /* Positive offset.  */
+                imm = insn & 0xfff;
+                gen_op_addl_T1_im(imm);
+            } else {
+                op = (insn >> 8) & 7;
+                imm = insn & 0xff;
+                switch (op) {
+                case 0: case 8: /* Shifted Register.  */
+                    shift = (insn >> 4) & 0xf;
+                    if (shift > 3)
+                        goto illegal_op;
+                    gen_movl_T2_reg(s, rm);
+                    if (shift)
+                        gen_op_shll_T2_im(shift);
+                    gen_op_addl_T1_T2();
+                    break;
+                case 4: /* Negative offset.  */
+                    gen_op_addl_T1_im(-imm);
+                    break;
+                case 6: /* User privilege.  */
+                    gen_op_addl_T1_im(imm);
+                    break;
+                case 1: /* Post-decrement.  */
+                    imm = -imm;
+                    /* Fall through.  */
+                case 3: /* Post-increment.  */
+                    gen_op_movl_T2_im(imm);
+                    postinc = 1;
+                    writeback = 1;
+                    break;
+                case 5: /* Pre-decrement.  */
+                    imm = -imm;
+                    /* Fall through.  */
+                case 7: /* Pre-increment.  */
+                    gen_op_addl_T1_im(imm);
+                    writeback = 1;
+                    break;
+                default:
+                    goto illegal_op;
+                }
+            }
+        }
+        op = ((insn >> 21) & 3) | ((insn >> 22) & 4);
+        if (insn & (1 << 20)) {
+            /* Load.  */
+            if (rs == 15 && op != 2) {
+                if (op & 2)
+                    goto illegal_op;
+                /* Memory hint.  Implemented as NOP.  */
+            } else {
+                switch (op) {
+                case 0: gen_ldst(ldub, s); break;
+                case 4: gen_ldst(ldsb, s); break;
+                case 1: gen_ldst(lduw, s); break;
+                case 5: gen_ldst(ldsw, s); break;
+                case 2: gen_ldst(ldl, s); break;
+                default: goto illegal_op;
+                }
+                if (rs == 15) {
+                    gen_bx(s);
+                } else {
+                    gen_movl_reg_T0(s, rs);
+                }
+            }
+        } else {
+            /* Store.  */
+            if (rs == 15)
+                goto illegal_op;
+            gen_movl_T0_reg(s, rs);
+            switch (op) {
+            case 0: gen_ldst(stb, s); break;
+            case 1: gen_ldst(stw, s); break;
+            case 2: gen_ldst(stl, s); break;
+            default: goto illegal_op;
+            }
+        }
+        if (postinc)
+            gen_op_addl_T1_im(imm);
+        if (writeback)
+            gen_movl_reg_T1(s, rn);
+        }
+        break;
+    default:
+        goto illegal_op;
+    }
+    return 0;
+illegal_op:
+    return 1;
+}
+
+static void disas_thumb_insn(CPUState *env, DisasContext *s)
 {
     uint32_t val, insn, op, rm, rn, rd, shift, cond;
     int32_t offset;
     int i;
 
+    if (s->condexec_mask) {
+        cond = s->condexec_cond;
+        s->condlabel = gen_new_label();
+        gen_test_cc[cond ^ 1](s->condlabel);
+        s->condjmp = 1;
+    }
+
     insn = lduw_code(s->pc);
     s->pc += 2;
 
@@ -2990,17 +6851,27 @@ static void disas_thumb_insn(DisasContext *s)
                 rm = (insn >> 6) & 7;
                 gen_movl_T1_reg(s, rm);
             }
-            if (insn & (1 << 9))
-                gen_op_subl_T0_T1_cc();
-            else
-                gen_op_addl_T0_T1_cc();
+            if (insn & (1 << 9)) {
+                if (s->condexec_mask)
+                    gen_op_subl_T0_T1();
+                else
+                    gen_op_subl_T0_T1_cc();
+            } else {
+                if (s->condexec_mask)
+                    gen_op_addl_T0_T1();
+                else
+                    gen_op_addl_T0_T1_cc();
+            }
             gen_movl_reg_T0(s, rd);
         } else {
             /* shift immediate */
             rm = (insn >> 3) & 7;
             shift = (insn >> 6) & 0x1f;
             gen_movl_T0_reg(s, rm);
-            gen_shift_T0_im_thumb[op](shift);
+            if (s->condexec_mask)
+                gen_shift_T0_im_thumb[op](shift);
+            else
+                gen_shift_T0_im_thumb_cc[op](shift);
             gen_movl_reg_T0(s, rd);
         }
         break;
@@ -3016,16 +6887,23 @@ static void disas_thumb_insn(DisasContext *s)
         }
         switch (op) {
         case 0: /* mov */
-            gen_op_logic_T0_cc();
+            if (!s->condexec_mask)
+                gen_op_logic_T0_cc();
             break;
         case 1: /* cmp */
             gen_op_subl_T0_T1_cc();
             break;
         case 2: /* add */
-            gen_op_addl_T0_T1_cc();
+            if (s->condexec_mask)
+                gen_op_addl_T0_T1();
+            else
+                gen_op_addl_T0_T1_cc();
             break;
         case 3: /* sub */
-            gen_op_subl_T0_T1_cc();
+            if (s->condexec_mask)
+                gen_op_subl_T0_T1();
+            else
+                gen_op_subl_T0_T1_cc();
             break;
         }
         if (op != 1)
@@ -3099,33 +6977,57 @@ static void disas_thumb_insn(DisasContext *s)
         switch (op) {
         case 0x0: /* and */
             gen_op_andl_T0_T1();
-            gen_op_logic_T0_cc();
+            if (!s->condexec_mask)
+                gen_op_logic_T0_cc();
             break;
         case 0x1: /* eor */
             gen_op_xorl_T0_T1();
-            gen_op_logic_T0_cc();
+            if (!s->condexec_mask)
+                gen_op_logic_T0_cc();
             break;
         case 0x2: /* lsl */
-            gen_op_shll_T1_T0_cc();
-            gen_op_logic_T1_cc();
+            if (s->condexec_mask) {
+                gen_op_shll_T1_T0();
+            } else {
+                gen_op_shll_T1_T0_cc();
+                gen_op_logic_T1_cc();
+            }
             break;
         case 0x3: /* lsr */
-            gen_op_shrl_T1_T0_cc();
-            gen_op_logic_T1_cc();
+            if (s->condexec_mask) {
+                gen_op_shrl_T1_T0();
+            } else {
+                gen_op_shrl_T1_T0_cc();
+                gen_op_logic_T1_cc();
+            }
             break;
         case 0x4: /* asr */
-            gen_op_sarl_T1_T0_cc();
-            gen_op_logic_T1_cc();
+            if (s->condexec_mask) {
+                gen_op_sarl_T1_T0();
+            } else {
+                gen_op_sarl_T1_T0_cc();
+                gen_op_logic_T1_cc();
+            }
             break;
         case 0x5: /* adc */
-            gen_op_adcl_T0_T1_cc();
+            if (s->condexec_mask)
+                gen_op_adcl_T0_T1();
+            else
+                gen_op_adcl_T0_T1_cc();
             break;
         case 0x6: /* sbc */
-            gen_op_sbcl_T0_T1_cc();
+            if (s->condexec_mask)
+                gen_op_sbcl_T0_T1();
+            else
+                gen_op_sbcl_T0_T1_cc();
             break;
         case 0x7: /* ror */
-            gen_op_rorl_T1_T0_cc();
-            gen_op_logic_T1_cc();
+            if (s->condexec_mask) {
+                gen_op_rorl_T1_T0();
+            } else {
+                gen_op_rorl_T1_T0_cc();
+                gen_op_logic_T1_cc();
+            }
             break;
         case 0x8: /* tst */
             gen_op_andl_T0_T1();
@@ -3133,7 +7035,10 @@ static void disas_thumb_insn(DisasContext *s)
             rd = 16;
             break;
         case 0x9: /* neg */
-            gen_op_subl_T0_T1_cc();
+            if (s->condexec_mask)
+                gen_op_subl_T0_T1();
+            else
+                gen_op_subl_T0_T1_cc();
             break;
         case 0xa: /* cmp */
             gen_op_subl_T0_T1_cc();
@@ -3145,19 +7050,23 @@ static void disas_thumb_insn(DisasContext *s)
             break;
         case 0xc: /* orr */
             gen_op_orl_T0_T1();
-            gen_op_logic_T0_cc();
+            if (!s->condexec_mask)
+                gen_op_logic_T0_cc();
             break;
         case 0xd: /* mul */
             gen_op_mull_T0_T1();
-            gen_op_logic_T0_cc();
+            if (!s->condexec_mask)
+                gen_op_logic_T0_cc();
             break;
         case 0xe: /* bic */
             gen_op_bicl_T0_T1();
-            gen_op_logic_T0_cc();
+            if (!s->condexec_mask)
+                gen_op_logic_T0_cc();
             break;
         case 0xf: /* mvn */
             gen_op_notl_T1();
-            gen_op_logic_T1_cc();
+            if (!s->condexec_mask)
+                gen_op_logic_T1_cc();
             val = 1;
             rm = rd;
             break;
@@ -3323,6 +7232,19 @@ static void disas_thumb_insn(DisasContext *s)
             gen_movl_reg_T1(s, 13);
             break;
 
+        case 2: /* sign/zero extend.  */
+            ARCH(6);
+            rd = insn & 7;
+            rm = (insn >> 3) & 7;
+            gen_movl_T1_reg(s, rm);
+            switch ((insn >> 6) & 3) {
+            case 0: gen_op_sxth_T1(); break;
+            case 1: gen_op_sxtb_T1(); break;
+            case 2: gen_op_uxth_T1(); break;
+            case 3: gen_op_uxtb_T1(); break;
+            }
+            gen_movl_reg_T1(s, rd);
+            break;
         case 4: case 5: case 0xc: case 0xd:
             /* push/pop */
             gen_movl_T1_reg(s, 13);
@@ -3378,13 +7300,82 @@ static void disas_thumb_insn(DisasContext *s)
                 gen_bx(s);
             break;
 
+        case 1: case 3: case 9: case 11: /* czb */
+            rm = insn & 7;
+            gen_movl_T0_reg(s, rm);
+            s->condlabel = gen_new_label();
+            s->condjmp = 1;
+            if (insn & (1 << 11))
+                gen_op_testn_T0(s->condlabel);
+            else
+                gen_op_test_T0(s->condlabel);
+
+            offset = ((insn & 0xf8) >> 2) | (insn & 0x200) >> 3;
+            val = (uint32_t)s->pc + 2;
+            val += offset;
+            gen_jmp(s, val);
+            break;
+
+        case 15: /* IT, nop-hint.  */
+            if ((insn & 0xf) == 0) {
+                gen_nop_hint(s, (insn >> 4) & 0xf);
+                break;
+            }
+            /* If Then.  */
+            s->condexec_cond = (insn >> 4) & 0xe;
+            s->condexec_mask = insn & 0x1f;
+            /* No actual code generated for this insn, just setup state.  */
+            break;
+
         case 0xe: /* bkpt */
+            gen_set_condexec(s);
             gen_op_movl_T0_im((long)s->pc - 2);
             gen_op_movl_reg_TN[0][15]();
             gen_op_bkpt();
             s->is_jmp = DISAS_JUMP;
             break;
 
+        case 0xa: /* rev */
+            ARCH(6);
+            rn = (insn >> 3) & 0x7;
+            rd = insn & 0x7;
+            gen_movl_T0_reg(s, rn);
+            switch ((insn >> 6) & 3) {
+            case 0: gen_op_rev_T0(); break;
+            case 1: gen_op_rev16_T0(); break;
+            case 3: gen_op_revsh_T0(); break;
+            default: goto illegal_op;
+            }
+            gen_movl_reg_T0(s, rd);
+            break;
+
+        case 6: /* cps */
+            ARCH(6);
+            if (IS_USER(s))
+                break;
+            if (IS_M(env)) {
+                val = (insn & (1 << 4)) != 0;
+                gen_op_movl_T0_im(val);
+                /* PRIMASK */
+                if (insn & 1)
+                    gen_op_v7m_msr_T0(16);
+                /* FAULTMASK */
+                if (insn & 2)
+                    gen_op_v7m_msr_T0(17);
+
+                gen_lookup_tb(s);
+            } else {
+                if (insn & (1 << 4))
+                    shift = CPSR_A | CPSR_I | CPSR_F;
+                else
+                    shift = 0;
+
+                val = ((insn & 7) << 6) & shift;
+                gen_op_movl_T0_im(val);
+                gen_set_psr_T0(s, shift, 0);
+            }
+            break;
+
         default:
             goto undef;
         }
@@ -3423,19 +7414,17 @@ static void disas_thumb_insn(DisasContext *s)
 
         if (cond == 0xf) {
             /* swi */
+            gen_set_condexec(s);
             gen_op_movl_T0_im((long)s->pc | 1);
             /* Don't set r15.  */
             gen_op_movl_reg_TN[0][15]();
-            gen_op_swi();
-            s->is_jmp = DISAS_JUMP;
+            s->is_jmp = DISAS_SWI;
             break;
         }
         /* generate a conditional jump to next instruction */
         s->condlabel = gen_new_label();
         gen_test_cc[cond ^ 1](s->condlabel);
         s->condjmp = 1;
-        //gen_test_cc[cond ^ 1]((long)s->tb, (long)s->pc);
-        //s->is_jmp = DISAS_JUMP_NEXT;
         gen_movl_T1_reg(s, 15);
 
         /* jump to the offset */
@@ -3446,22 +7435,12 @@ static void disas_thumb_insn(DisasContext *s)
         break;
 
     case 14:
-        /* unconditional branch */
         if (insn & (1 << 11)) {
-            /* Second half of blx.  */
-            offset = ((insn & 0x7ff) << 1);
-            gen_movl_T0_reg(s, 14);
-            gen_op_movl_T1_im(offset);
-            gen_op_addl_T0_T1();
-            gen_op_movl_T1_im(0xfffffffc);
-            gen_op_andl_T0_T1();
-
-            val = (uint32_t)s->pc;
-            gen_op_movl_T1_im(val | 1);
-            gen_movl_reg_T1(s, 14);
-            gen_bx(s);
+            if (disas_thumb2_insn(env, s, insn))
+              goto undef32;
             break;
         }
+        /* unconditional branch */
         val = (uint32_t)s->pc;
         offset = ((int32_t)insn << 21) >> 21;
         val += (offset << 1) + 2;
@@ -3469,51 +7448,21 @@ static void disas_thumb_insn(DisasContext *s)
         break;
 
     case 15:
-        /* branch and link [and switch to arm] */
-        if ((s->pc & ~TARGET_PAGE_MASK) == 0) {
-            /* Instruction spans a page boundary.  Implement it as two
-               16-bit instructions in case the second half causes an
-               prefetch abort.  */
-            offset = ((int32_t)insn << 21) >> 9;
-            val = s->pc + 2 + offset;
-            gen_op_movl_T0_im(val);
-            gen_movl_reg_T0(s, 14);
-            break;
-        }
-        if (insn & (1 << 11)) {
-            /* Second half of bl.  */
-            offset = ((insn & 0x7ff) << 1) | 1;
-            gen_movl_T0_reg(s, 14);
-            gen_op_movl_T1_im(offset);
-            gen_op_addl_T0_T1();
-
-            val = (uint32_t)s->pc;
-            gen_op_movl_T1_im(val | 1);
-            gen_movl_reg_T1(s, 14);
-            gen_bx(s);
-            break;
-        }
-        offset = ((int32_t)insn << 21) >> 10;
-        insn = lduw_code(s->pc);
-        offset |= insn & 0x7ff;
-
-        val = (uint32_t)s->pc + 2;
-        gen_op_movl_T1_im(val | 1);
-        gen_movl_reg_T1(s, 14);
-
-        val += offset << 1;
-        if (insn & (1 << 12)) {
-            /* bl */
-            gen_jmp(s, val);
-        } else {
-            /* blx */
-            val &= ~(uint32_t)2;
-            gen_op_movl_T0_im(val);
-            gen_bx(s);
-        }
+        if (disas_thumb2_insn(env, s, insn))
+          goto undef32;
+        break;
     }
     return;
+undef32:
+    gen_set_condexec(s);
+    gen_op_movl_T0_im((long)s->pc - 4);
+    gen_op_movl_reg_TN[0][15]();
+    gen_op_undef_insn();
+    s->is_jmp = DISAS_JUMP;
+    return;
+illegal_op:
 undef:
+    gen_set_condexec(s);
     gen_op_movl_T0_im((long)s->pc - 2);
     gen_op_movl_reg_TN[0][15]();
     gen_op_undef_insn();
@@ -3547,21 +7496,44 @@ static inline int gen_intermediate_code_internal(CPUState *env,
     dc->singlestep_enabled = env->singlestep_enabled;
     dc->condjmp = 0;
     dc->thumb = env->thumb;
+    dc->condexec_mask = (env->condexec_bits & 0xf) << 1;
+    dc->condexec_cond = env->condexec_bits >> 4;
     dc->is_mem = 0;
 #if !defined(CONFIG_USER_ONLY)
-    dc->user = (env->uncached_cpsr & 0x1f) == ARM_CPU_MODE_USR;
+    if (IS_M(env)) {
+        dc->user = ((env->v7m.exception == 0) && (env->v7m.control & 1));
+    } else {
+        dc->user = (env->uncached_cpsr & 0x1f) == ARM_CPU_MODE_USR;
+    }
 #endif
     next_page_start = (pc_start & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE;
     nb_gen_labels = 0;
     lj = -1;
+    /* Reset the conditional execution bits immediately. This avoids
+       complications trying to do it at the end of the block.  */
+    if (env->condexec_bits)
+      gen_op_set_condexec(0);
     do {
+#ifndef CONFIG_USER_ONLY
+        if (dc->pc >= 0xfffffff0 && IS_M(env)) {
+            /* We always get here via a jump, so know we are not in a
+               conditional execution block.  */
+            gen_op_exception_exit();
+        }
+#endif
+
         if (env->nb_breakpoints > 0) {
             for(j = 0; j < env->nb_breakpoints; j++) {
                 if (env->breakpoints[j] == dc->pc) {
+                    gen_set_condexec(dc);
                     gen_op_movl_T0_im((long)dc->pc);
                     gen_op_movl_reg_TN[0][15]();
                     gen_op_debug();
                     dc->is_jmp = DISAS_JUMP;
+                    /* Advance PC so that clearing the breakpoint will
+                       invalidate this TB.  */
+                    dc->pc += 2;
+                    goto done_generating;
                     break;
                 }
             }
@@ -3577,10 +7549,19 @@ static inline int gen_intermediate_code_internal(CPUState *env,
             gen_opc_instr_start[lj] = 1;
         }
 
-        if (env->thumb)
-          disas_thumb_insn(dc);
-        else
-          disas_arm_insn(env, dc);
+        if (env->thumb) {
+            disas_thumb_insn(env, dc);
+            if (dc->condexec_mask) {
+                dc->condexec_cond = (dc->condexec_cond & 0xe)
+                                   | ((dc->condexec_mask >> 4) & 1);
+                dc->condexec_mask = (dc->condexec_mask << 1) & 0x1f;
+                if (dc->condexec_mask == 0) {
+                    dc->condexec_cond = 0;
+                }
+            }
+        } else {
+            disas_arm_insn(env, dc);
+        }
 
         if (dc->condjmp && !dc->is_jmp) {
             gen_set_label(dc->condlabel);
@@ -3599,13 +7580,19 @@ static inline int gen_intermediate_code_internal(CPUState *env,
     } while (!dc->is_jmp && gen_opc_ptr < gen_opc_end &&
              !env->singlestep_enabled &&
              dc->pc < next_page_start);
+
     /* At this stage dc->condjmp will only be set when the skipped
-     * instruction was a conditional branch, and the PC has already been
-     * written.  */
+       instruction was a conditional branch or trap, and the PC has
+       already been written.  */
     if (__builtin_expect(env->singlestep_enabled, 0)) {
         /* Make sure the pc is updated, and raise a debug exception.  */
         if (dc->condjmp) {
-            gen_op_debug();
+            gen_set_condexec(dc);
+            if (dc->is_jmp == DISAS_SWI) {
+                gen_op_swi();
+            } else {
+                gen_op_debug();
+            }
             gen_set_label(dc->condlabel);
         }
         if (dc->condjmp || !dc->is_jmp) {
@@ -3613,8 +7600,24 @@ static inline int gen_intermediate_code_internal(CPUState *env,
             gen_op_movl_reg_TN[0][15]();
             dc->condjmp = 0;
         }
-        gen_op_debug();
+        gen_set_condexec(dc);
+        if (dc->is_jmp == DISAS_SWI && !dc->condjmp) {
+            gen_op_swi();
+        } else {
+            /* FIXME: Single stepping a WFI insn will not halt
+               the CPU.  */
+            gen_op_debug();
+        }
     } else {
+        /* While branches must always occur at the end of an IT block,
+           there are a few other things that can cause us to terminate
+           the TB in the middel of an IT block:
+            - Exception generating instructions (bkpt, swi, undefined).
+            - Page boundaries.
+            - Hardware watchpoints.
+           Hardware breakpoints have already been handled and skip this code.
+         */
+        gen_set_condexec(dc);
         switch(dc->is_jmp) {
         case DISAS_NEXT:
             gen_goto_tb(dc, 1, dc->pc);
@@ -3629,13 +7632,21 @@ static inline int gen_intermediate_code_internal(CPUState *env,
         case DISAS_TB_JUMP:
             /* nothing more to generate */
             break;
+        case DISAS_WFI:
+            gen_op_wfi();
+            break;
+        case DISAS_SWI:
+            gen_op_swi();
+            break;
         }
         if (dc->condjmp) {
             gen_set_label(dc->condlabel);
+            gen_set_condexec(dc);
             gen_goto_tb(dc, 1, dc->pc);
             dc->condjmp = 0;
         }
     }
+done_generating:
     *gen_opc_ptr = INDEX_op_end;
 
 #ifdef DEBUG_DISAS
@@ -3676,6 +7687,7 @@ static const char *cpu_mode_names[16] = {
   "usr", "fiq", "irq", "svc", "???", "???", "???", "abt",
   "???", "???", "???", "und", "???", "???", "???", "sys"
 };
+
 void cpu_dump_state(CPUState *env, FILE *f,
                     int (*cpu_fprintf)(FILE *f, const char *fmt, ...),
                     int flags)