Initial import of abandoned HQEMU version 2.5.2HEAD master

author: Timothy Pearson <tpearson@raptorengineering.com> 2019-05-11 15:12:49 -0500
committer: Timothy Pearson <tpearson@raptorengineering.com> 2019-05-11 15:12:49 -0500
commit: 9e80202352dd49bdd9e67b8b906d86f058431505 (patch)
tree: 5673c17aad6e3833da8c4ff21b5a11f666ec9fbe /src/target-tilegx
download: hqemu-9e80202352dd49bdd9e67b8b906d86f058431505.zip
hqemu-9e80202352dd49bdd9e67b8b906d86f058431505.tar.gz
9 files changed, 4791 insertions, 0 deletions
diff --git a/src/target-tilegx/Makefile.objs b/src/target-tilegx/Makefile.objs
new file mode 100644
index 0000000..0db778f
--- /dev/null
+++ b/src/target-tilegx/Makefile.objs
@@ -0,0 +1 @@
+obj-y += cpu.o translate.o helper.o simd_helper.o
diff --git a/src/target-tilegx/cpu.c b/src/target-tilegx/cpu.c
new file mode 100644
index 0000000..c249704
--- /dev/null
+++ b/src/target-tilegx/cpu.c
@@ -0,0 +1,185 @@
+/*
+ * QEMU TILE-Gx CPU
+ *
+ *  Copyright (c) 2015 Chen Gang
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see
+ * <http://www.gnu.org/licenses/lgpl-2.1.html>
+ */
+
+#include "cpu.h"
+#include "qemu-common.h"
+#include "hw/qdev-properties.h"
+#include "migration/vmstate.h"
+#include "linux-user/syscall_defs.h"
+
+static void tilegx_cpu_dump_state(CPUState *cs, FILE *f,
+                                  fprintf_function cpu_fprintf, int flags)
+{
+    static const char * const reg_names[TILEGX_R_COUNT] = {
+         "r0",  "r1",  "r2",  "r3",  "r4",  "r5",  "r6",  "r7",
+         "r8",  "r9", "r10", "r11", "r12", "r13", "r14", "r15",
+        "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23",
+        "r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31",
+        "r32", "r33", "r34", "r35", "r36", "r37", "r38", "r39",
+        "r40", "r41", "r42", "r43", "r44", "r45", "r46", "r47",
+        "r48", "r49", "r50", "r51",  "bp",  "tp",  "sp",  "lr"
+    };
+
+    TileGXCPU *cpu = TILEGX_CPU(cs);
+    CPUTLGState *env = &cpu->env;
+    int i;
+
+    for (i = 0; i < TILEGX_R_COUNT; i++) {
+        cpu_fprintf(f, "%-4s" TARGET_FMT_lx "%s",
+                    reg_names[i], env->regs[i],
+                    (i % 4) == 3 ? "\n" : " ");
+    }
+    cpu_fprintf(f, "PC  " TARGET_FMT_lx " CEX " TARGET_FMT_lx "\n\n",
+                env->pc, env->spregs[TILEGX_SPR_CMPEXCH]);
+}
+
+TileGXCPU *cpu_tilegx_init(const char *cpu_model)
+{
+    TileGXCPU *cpu;
+
+    cpu = TILEGX_CPU(object_new(TYPE_TILEGX_CPU));
+
+    object_property_set_bool(OBJECT(cpu), true, "realized", NULL);
+
+    return cpu;
+}
+
+static void tilegx_cpu_set_pc(CPUState *cs, vaddr value)
+{
+    TileGXCPU *cpu = TILEGX_CPU(cs);
+
+    cpu->env.pc = value;
+}
+
+static bool tilegx_cpu_has_work(CPUState *cs)
+{
+    return true;
+}
+
+static void tilegx_cpu_reset(CPUState *s)
+{
+    TileGXCPU *cpu = TILEGX_CPU(s);
+    TileGXCPUClass *tcc = TILEGX_CPU_GET_CLASS(cpu);
+    CPUTLGState *env = &cpu->env;
+
+    tcc->parent_reset(s);
+
+    memset(env, 0, sizeof(CPUTLGState));
+    tlb_flush(s, 1);
+}
+
+static void tilegx_cpu_realizefn(DeviceState *dev, Error **errp)
+{
+    CPUState *cs = CPU(dev);
+    TileGXCPUClass *tcc = TILEGX_CPU_GET_CLASS(dev);
+
+    cpu_reset(cs);
+    qemu_init_vcpu(cs);
+
+    tcc->parent_realize(dev, errp);
+}
+
+static void tilegx_cpu_initfn(Object *obj)
+{
+    CPUState *cs = CPU(obj);
+    TileGXCPU *cpu = TILEGX_CPU(obj);
+    CPUTLGState *env = &cpu->env;
+    static bool tcg_initialized;
+
+    cs->env_ptr = env;
+    cpu_exec_init(cs, &error_abort);
+
+    if (tcg_enabled() && !tcg_initialized) {
+        tcg_initialized = true;
+        tilegx_tcg_init();
+    }
+}
+
+static void tilegx_cpu_do_interrupt(CPUState *cs)
+{
+    cs->exception_index = -1;
+}
+
+static int tilegx_cpu_handle_mmu_fault(CPUState *cs, vaddr address, int rw,
+                                       int mmu_idx)
+{
+    TileGXCPU *cpu = TILEGX_CPU(cs);
+
+    /* The sigcode field will be filled in by do_signal in main.c.  */
+    cs->exception_index = TILEGX_EXCP_SIGNAL;
+    cpu->env.excaddr = address;
+    cpu->env.signo = TARGET_SIGSEGV;
+    cpu->env.sigcode = 0;
+
+    return 1;
+}
+
+static bool tilegx_cpu_exec_interrupt(CPUState *cs, int interrupt_request)
+{
+    if (interrupt_request & CPU_INTERRUPT_HARD) {
+        tilegx_cpu_do_interrupt(cs);
+        return true;
+    }
+    return false;
+}
+
+static void tilegx_cpu_class_init(ObjectClass *oc, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(oc);
+    CPUClass *cc = CPU_CLASS(oc);
+    TileGXCPUClass *tcc = TILEGX_CPU_CLASS(oc);
+
+    tcc->parent_realize = dc->realize;
+    dc->realize = tilegx_cpu_realizefn;
+
+    tcc->parent_reset = cc->reset;
+    cc->reset = tilegx_cpu_reset;
+
+    cc->has_work = tilegx_cpu_has_work;
+    cc->do_interrupt = tilegx_cpu_do_interrupt;
+    cc->cpu_exec_interrupt = tilegx_cpu_exec_interrupt;
+    cc->dump_state = tilegx_cpu_dump_state;
+    cc->set_pc = tilegx_cpu_set_pc;
+    cc->handle_mmu_fault = tilegx_cpu_handle_mmu_fault;
+    cc->gdb_num_core_regs = 0;
+
+    /*
+     * Reason: tilegx_cpu_initfn() calls cpu_exec_init(), which saves
+     * the object in cpus -> dangling pointer after final
+     * object_unref().
+     */
+    dc->cannot_destroy_with_object_finalize_yet = true;
+}
+
+static const TypeInfo tilegx_cpu_type_info = {
+    .name = TYPE_TILEGX_CPU,
+    .parent = TYPE_CPU,
+    .instance_size = sizeof(TileGXCPU),
+    .instance_init = tilegx_cpu_initfn,
+    .class_size = sizeof(TileGXCPUClass),
+    .class_init = tilegx_cpu_class_init,
+};
+
+static void tilegx_cpu_register_types(void)
+{
+    type_register_static(&tilegx_cpu_type_info);
+}
+
+type_init(tilegx_cpu_register_types)
diff --git a/src/target-tilegx/cpu.h b/src/target-tilegx/cpu.h
new file mode 100644
index 0000000..03df107
--- /dev/null
+++ b/src/target-tilegx/cpu.h
@@ -0,0 +1,182 @@
+/*
+ *  TILE-Gx virtual CPU header
+ *
+ *  Copyright (c) 2015 Chen Gang
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see <http://www.gnu.org/licenses/>.
+ */
+#ifndef CPU_TILEGX_H
+#define CPU_TILEGX_H
+
+#include "config.h"
+#include "qemu-common.h"
+
+#define TARGET_LONG_BITS 64
+
+#define CPUArchState struct CPUTLGState
+
+#include "exec/cpu-defs.h"
+
+
+/* TILE-Gx common register alias */
+#define TILEGX_R_RE    0   /*  0 register, for function/syscall return value */
+#define TILEGX_R_ERR   1   /*  1 register, for syscall errno flag */
+#define TILEGX_R_NR    10  /* 10 register, for syscall number */
+#define TILEGX_R_BP    52  /* 52 register, optional frame pointer */
+#define TILEGX_R_TP    53  /* TP register, thread local storage data */
+#define TILEGX_R_SP    54  /* SP register, stack pointer */
+#define TILEGX_R_LR    55  /* LR register, may save pc, but it is not pc */
+#define TILEGX_R_COUNT 56  /* Only 56 registers are really useful */
+#define TILEGX_R_SN    56  /* SN register, obsoleted, it likes zero register */
+#define TILEGX_R_IDN0  57  /* IDN0 register, cause IDN_ACCESS exception */
+#define TILEGX_R_IDN1  58  /* IDN1 register, cause IDN_ACCESS exception */
+#define TILEGX_R_UDN0  59  /* UDN0 register, cause UDN_ACCESS exception */
+#define TILEGX_R_UDN1  60  /* UDN1 register, cause UDN_ACCESS exception */
+#define TILEGX_R_UDN2  61  /* UDN2 register, cause UDN_ACCESS exception */
+#define TILEGX_R_UDN3  62  /* UDN3 register, cause UDN_ACCESS exception */
+#define TILEGX_R_ZERO  63  /* Zero register, always zero */
+#define TILEGX_R_NOREG 255 /* Invalid register value */
+
+/* TILE-Gx special registers used by outside */
+enum {
+    TILEGX_SPR_CMPEXCH = 0,
+    TILEGX_SPR_CRITICAL_SEC = 1,
+    TILEGX_SPR_SIM_CONTROL = 2,
+    TILEGX_SPR_EX_CONTEXT_0_0 = 3,
+    TILEGX_SPR_EX_CONTEXT_0_1 = 4,
+    TILEGX_SPR_COUNT
+};
+
+/* Exception numbers */
+typedef enum {
+    TILEGX_EXCP_NONE = 0,
+    TILEGX_EXCP_SYSCALL = 1,
+    TILEGX_EXCP_SIGNAL = 2,
+    TILEGX_EXCP_OPCODE_UNKNOWN = 0x101,
+    TILEGX_EXCP_OPCODE_UNIMPLEMENTED = 0x102,
+    TILEGX_EXCP_OPCODE_CMPEXCH = 0x103,
+    TILEGX_EXCP_OPCODE_CMPEXCH4 = 0x104,
+    TILEGX_EXCP_OPCODE_EXCH = 0x105,
+    TILEGX_EXCP_OPCODE_EXCH4 = 0x106,
+    TILEGX_EXCP_OPCODE_FETCHADD = 0x107,
+    TILEGX_EXCP_OPCODE_FETCHADD4 = 0x108,
+    TILEGX_EXCP_OPCODE_FETCHADDGEZ = 0x109,
+    TILEGX_EXCP_OPCODE_FETCHADDGEZ4 = 0x10a,
+    TILEGX_EXCP_OPCODE_FETCHAND = 0x10b,
+    TILEGX_EXCP_OPCODE_FETCHAND4 = 0x10c,
+    TILEGX_EXCP_OPCODE_FETCHOR = 0x10d,
+    TILEGX_EXCP_OPCODE_FETCHOR4 = 0x10e,
+    TILEGX_EXCP_REG_IDN_ACCESS = 0x181,
+    TILEGX_EXCP_REG_UDN_ACCESS = 0x182,
+    TILEGX_EXCP_UNALIGNMENT = 0x201,
+    TILEGX_EXCP_DBUG_BREAK = 0x301
+} TileExcp;
+
+typedef struct CPUTLGState {
+    uint64_t regs[TILEGX_R_COUNT];     /* Common used registers by outside */
+    uint64_t spregs[TILEGX_SPR_COUNT]; /* Special used registers by outside */
+    uint64_t pc;                       /* Current pc */
+
+#if defined(CONFIG_USER_ONLY)
+    uint64_t excaddr;                  /* exception address */
+    uint64_t atomic_srca;              /* Arguments to atomic "exceptions" */
+    uint64_t atomic_srcb;
+    uint32_t atomic_dstr;
+    uint32_t signo;                    /* Signal number */
+    uint32_t sigcode;                  /* Signal code */
+#endif
+
+    CPU_COMMON
+} CPUTLGState;
+
+#include "qom/cpu.h"
+
+#define TYPE_TILEGX_CPU "tilegx-cpu"
+
+#define TILEGX_CPU_CLASS(klass) \
+    OBJECT_CLASS_CHECK(TileGXCPUClass, (klass), TYPE_TILEGX_CPU)
+#define TILEGX_CPU(obj) \
+    OBJECT_CHECK(TileGXCPU, (obj), TYPE_TILEGX_CPU)
+#define TILEGX_CPU_GET_CLASS(obj) \
+    OBJECT_GET_CLASS(TileGXCPUClass, (obj), TYPE_TILEGX_CPU)
+
+/**
+ * TileGXCPUClass:
+ * @parent_realize: The parent class' realize handler.
+ * @parent_reset: The parent class' reset handler.
+ *
+ * A Tile-Gx CPU model.
+ */
+typedef struct TileGXCPUClass {
+    /*< private >*/
+    CPUClass parent_class;
+    /*< public >*/
+
+    DeviceRealize parent_realize;
+    void (*parent_reset)(CPUState *cpu);
+} TileGXCPUClass;
+
+/**
+ * TileGXCPU:
+ * @env: #CPUTLGState
+ *
+ * A Tile-GX CPU.
+ */
+typedef struct TileGXCPU {
+    /*< private >*/
+    CPUState parent_obj;
+    /*< public >*/
+
+    CPUTLGState env;
+} TileGXCPU;
+
+static inline TileGXCPU *tilegx_env_get_cpu(CPUTLGState *env)
+{
+    return container_of(env, TileGXCPU, env);
+}
+
+#define ENV_GET_CPU(e) CPU(tilegx_env_get_cpu(e))
+
+#define ENV_OFFSET offsetof(TileGXCPU, env)
+
+/* TILE-Gx memory attributes */
+#define TARGET_PAGE_BITS 16  /* TILE-Gx uses 64KB page size */
+#define TARGET_PHYS_ADDR_SPACE_BITS 42
+#define TARGET_VIRT_ADDR_SPACE_BITS 64
+#define MMU_USER_IDX    0  /* Current memory operation is in user mode */
+
+#include "exec/cpu-all.h"
+
+void tilegx_tcg_init(void);
+int cpu_tilegx_exec(CPUState *s);
+int cpu_tilegx_signal_handler(int host_signum, void *pinfo, void *puc);
+
+TileGXCPU *cpu_tilegx_init(const char *cpu_model);
+
+#define cpu_init(cpu_model) CPU(cpu_tilegx_init(cpu_model))
+
+#define cpu_exec cpu_tilegx_exec
+#define cpu_signal_handler cpu_tilegx_signal_handler
+
+static inline void cpu_get_tb_cpu_state(CPUTLGState *env, target_ulong *pc,
+                                        target_ulong *cs_base, int *flags)
+{
+    *pc = env->pc;
+    *cs_base = 0;
+    *flags = 0;
+}
+
+#include "exec/exec-all.h"
+
+#endif
diff --git a/src/target-tilegx/helper.c b/src/target-tilegx/helper.c
new file mode 100644
index 0000000..dda821f
--- /dev/null
+++ b/src/target-tilegx/helper.c
@@ -0,0 +1,161 @@
+/*
+ * QEMU TILE-Gx helpers
+ *
+ *  Copyright (c) 2015 Chen Gang
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see
+ * <http://www.gnu.org/licenses/lgpl-2.1.html>
+ */
+
+#include "cpu.h"
+#include "qemu-common.h"
+#include "exec/helper-proto.h"
+#include <zlib.h> /* For crc32 */
+#include "syscall_defs.h"
+
+void helper_exception(CPUTLGState *env, uint32_t excp)
+{
+    CPUState *cs = CPU(tilegx_env_get_cpu(env));
+
+    cs->exception_index = excp;
+    cpu_loop_exit(cs);
+}
+
+void helper_ext01_ics(CPUTLGState *env)
+{
+    uint64_t val = env->spregs[TILEGX_SPR_EX_CONTEXT_0_1];
+
+    switch (val) {
+    case 0:
+    case 1:
+        env->spregs[TILEGX_SPR_CRITICAL_SEC] = val;
+        break;
+    default:
+#if defined(CONFIG_USER_ONLY)
+        env->signo = TARGET_SIGILL;
+        env->sigcode = TARGET_ILL_ILLOPC;
+        helper_exception(env, TILEGX_EXCP_SIGNAL);
+#else
+        helper_exception(env, TILEGX_EXCP_OPCODE_UNIMPLEMENTED);
+#endif
+        break;
+    }
+}
+
+uint64_t helper_cntlz(uint64_t arg)
+{
+    return clz64(arg);
+}
+
+uint64_t helper_cnttz(uint64_t arg)
+{
+    return ctz64(arg);
+}
+
+uint64_t helper_pcnt(uint64_t arg)
+{
+    return ctpop64(arg);
+}
+
+uint64_t helper_revbits(uint64_t arg)
+{
+    return revbit64(arg);
+}
+
+/*
+ * Functional Description
+ *     uint64_t a = rf[SrcA];
+ *     uint64_t b = rf[SrcB];
+ *     uint64_t d = rf[Dest];
+ *     uint64_t output = 0;
+ *     unsigned int counter;
+ *     for (counter = 0; counter < (WORD_SIZE / BYTE_SIZE); counter++)
+ *     {
+ *         int sel = getByte (b, counter) & 0xf;
+ *         uint8_t byte = (sel < 8) ? getByte (d, sel) : getByte (a, (sel - 8));
+ *         output = setByte (output, counter, byte);
+ *     }
+ *     rf[Dest] = output;
+ */
+uint64_t helper_shufflebytes(uint64_t dest, uint64_t srca, uint64_t srcb)
+{
+    uint64_t vdst = 0;
+    int count;
+
+    for (count = 0; count < 64; count += 8) {
+        uint64_t sel = srcb >> count;
+        uint64_t src = (sel & 8) ? srca : dest;
+        vdst |= extract64(src, (sel & 7) * 8, 8) << count;
+    }
+
+    return vdst;
+}
+
+uint64_t helper_crc32_8(uint64_t accum, uint64_t input)
+{
+    uint8_t buf = input;
+
+    /* zlib crc32 converts the accumulator and output to one's complement.  */
+    return crc32(accum ^ 0xffffffff, &buf, 1) ^ 0xffffffff;
+}
+
+uint64_t helper_crc32_32(uint64_t accum, uint64_t input)
+{
+    uint8_t buf[4];
+
+    stl_le_p(buf, input);
+
+    /* zlib crc32 converts the accumulator and output to one's complement.  */
+    return crc32(accum ^ 0xffffffff, buf, 4) ^ 0xffffffff;
+}
+
+uint64_t helper_cmula(uint64_t srcd, uint64_t srca, uint64_t srcb)
+{
+    uint32_t reala = (int16_t)srca;
+    uint32_t imaga = (int16_t)(srca >> 16);
+    uint32_t realb = (int16_t)srcb;
+    uint32_t imagb = (int16_t)(srcb >> 16);
+    uint32_t reald = srcd;
+    uint32_t imagd = srcd >> 32;
+    uint32_t realr = reala * realb - imaga * imagb + reald;
+    uint32_t imagr = reala * imagb + imaga * realb + imagd;
+
+    return deposit64(realr, 32, 32, imagr);
+}
+
+uint64_t helper_cmulaf(uint64_t srcd, uint64_t srca, uint64_t srcb)
+{
+    uint32_t reala = (int16_t)srca;
+    uint32_t imaga = (int16_t)(srca >> 16);
+    uint32_t realb = (int16_t)srcb;
+    uint32_t imagb = (int16_t)(srcb >> 16);
+    uint32_t reald = (int16_t)srcd;
+    uint32_t imagd = (int16_t)(srcd >> 16);
+    int32_t realr = reala * realb - imaga * imagb;
+    int32_t imagr = reala * imagb + imaga * realb;
+
+    return deposit32((realr >> 15) + reald, 16, 16, (imagr >> 15) + imagd);
+}
+
+uint64_t helper_cmul2(uint64_t srca, uint64_t srcb, int shift, int round)
+{
+    uint32_t reala = (int16_t)srca;
+    uint32_t imaga = (int16_t)(srca >> 16);
+    uint32_t realb = (int16_t)srcb;
+    uint32_t imagb = (int16_t)(srcb >> 16);
+    int32_t realr = reala * realb - imaga * imagb + round;
+    int32_t imagr = reala * imagb + imaga * realb + round;
+
+    return deposit32(realr >> shift, 16, 16, imagr >> shift);
+}
diff --git a/src/target-tilegx/helper.h b/src/target-tilegx/helper.h
new file mode 100644
index 0000000..9281d0f
--- /dev/null
+++ b/src/target-tilegx/helper.h
@@ -0,0 +1,26 @@
+DEF_HELPER_2(exception, noreturn, env, i32)
+DEF_HELPER_1(ext01_ics, void, env)
+DEF_HELPER_FLAGS_1(cntlz, TCG_CALL_NO_RWG_SE, i64, i64)
+DEF_HELPER_FLAGS_1(cnttz, TCG_CALL_NO_RWG_SE, i64, i64)
+DEF_HELPER_FLAGS_1(pcnt, TCG_CALL_NO_RWG_SE, i64, i64)
+DEF_HELPER_FLAGS_1(revbits, TCG_CALL_NO_RWG_SE, i64, i64)
+DEF_HELPER_FLAGS_3(shufflebytes, TCG_CALL_NO_RWG_SE, i64, i64, i64, i64)
+DEF_HELPER_FLAGS_2(crc32_8, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_2(crc32_32, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_3(cmula, TCG_CALL_NO_RWG_SE, i64, i64, i64, i64)
+DEF_HELPER_FLAGS_3(cmulaf, TCG_CALL_NO_RWG_SE, i64, i64, i64, i64)
+DEF_HELPER_FLAGS_4(cmul2, TCG_CALL_NO_RWG_SE, i64, i64, i64, int, int)
+
+DEF_HELPER_FLAGS_2(v1int_h, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_2(v1int_l, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_2(v2int_h, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_2(v2int_l, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+
+DEF_HELPER_FLAGS_2(v1multu, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_2(v2mults, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_2(v1shl, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_2(v1shru, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_2(v1shrs, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_2(v2shl, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_2(v2shru, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_2(v2shrs, TCG_CALL_NO_RWG_SE, i64, i64, i64)
diff --git a/src/target-tilegx/opcode_tilegx.h b/src/target-tilegx/opcode_tilegx.h
new file mode 100644
index 0000000..989436d
--- /dev/null
+++ b/src/target-tilegx/opcode_tilegx.h
@@ -0,0 +1,1406 @@
+/* TILE-Gx opcode information.
+ *
+ * Copyright 2011 Tilera Corporation. All Rights Reserved.
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful, but
+ *   WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ *   NON INFRINGEMENT.  See the GNU General Public License for
+ *   more details.
+ *
+ *
+ *
+ *
+ *
+ */
+
+#ifndef __ARCH_OPCODE_H__
+#define __ARCH_OPCODE_H__
+
+#ifndef __ASSEMBLER__
+
+typedef uint64_t tilegx_bundle_bits;
+
+/* These are the bits that determine if a bundle is in the X encoding. */
+#define TILEGX_BUNDLE_MODE_MASK ((tilegx_bundle_bits)3 << 62)
+
+enum
+{
+  /* Maximum number of instructions in a bundle (2 for X, 3 for Y). */
+  TILEGX_MAX_INSTRUCTIONS_PER_BUNDLE = 3,
+
+  /* How many different pipeline encodings are there? X0, X1, Y0, Y1, Y2. */
+  TILEGX_NUM_PIPELINE_ENCODINGS = 5,
+
+  /* Log base 2 of TILEGX_BUNDLE_SIZE_IN_BYTES. */
+  TILEGX_LOG2_BUNDLE_SIZE_IN_BYTES = 3,
+
+  /* Instructions take this many bytes. */
+  TILEGX_BUNDLE_SIZE_IN_BYTES = 1 << TILEGX_LOG2_BUNDLE_SIZE_IN_BYTES,
+
+  /* Log base 2 of TILEGX_BUNDLE_ALIGNMENT_IN_BYTES. */
+  TILEGX_LOG2_BUNDLE_ALIGNMENT_IN_BYTES = 3,
+
+  /* Bundles should be aligned modulo this number of bytes. */
+  TILEGX_BUNDLE_ALIGNMENT_IN_BYTES =
+    (1 << TILEGX_LOG2_BUNDLE_ALIGNMENT_IN_BYTES),
+
+  /* Number of registers (some are magic, such as network I/O). */
+  TILEGX_NUM_REGISTERS = 64,
+};
+
+/* Make a few "tile_" variables to simplify common code between
+   architectures.  */
+
+typedef tilegx_bundle_bits tile_bundle_bits;
+#define TILE_BUNDLE_SIZE_IN_BYTES TILEGX_BUNDLE_SIZE_IN_BYTES
+#define TILE_BUNDLE_ALIGNMENT_IN_BYTES TILEGX_BUNDLE_ALIGNMENT_IN_BYTES
+#define TILE_LOG2_BUNDLE_ALIGNMENT_IN_BYTES \
+  TILEGX_LOG2_BUNDLE_ALIGNMENT_IN_BYTES
+#define TILE_BPT_BUNDLE TILEGX_BPT_BUNDLE
+
+/* 64-bit pattern for a { bpt ; nop } bundle. */
+#define TILEGX_BPT_BUNDLE 0x286a44ae51485000ULL
+
+static inline unsigned int
+get_BFEnd_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 12)) & 0x3f);
+}
+
+static inline unsigned int
+get_BFOpcodeExtension_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 24)) & 0xf);
+}
+
+static inline unsigned int
+get_BFStart_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 18)) & 0x3f);
+}
+
+static inline unsigned int
+get_BrOff_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 31)) & 0x0000003f) |
+         (((unsigned int)(n >> 37)) & 0x0001ffc0);
+}
+
+static inline unsigned int
+get_BrType_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 54)) & 0x1f);
+}
+
+static inline unsigned int
+get_Dest_Imm8_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 31)) & 0x0000003f) |
+         (((unsigned int)(n >> 43)) & 0x000000c0);
+}
+
+static inline unsigned int
+get_Dest_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 0)) & 0x3f);
+}
+
+static inline unsigned int
+get_Dest_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 31)) & 0x3f);
+}
+
+static inline unsigned int
+get_Dest_Y0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 0)) & 0x3f);
+}
+
+static inline unsigned int
+get_Dest_Y1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 31)) & 0x3f);
+}
+
+static inline unsigned int
+get_Imm16_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 12)) & 0xffff);
+}
+
+static inline unsigned int
+get_Imm16_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 43)) & 0xffff);
+}
+
+static inline unsigned int
+get_Imm8OpcodeExtension_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 20)) & 0xff);
+}
+
+static inline unsigned int
+get_Imm8OpcodeExtension_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 51)) & 0xff);
+}
+
+static inline unsigned int
+get_Imm8_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 12)) & 0xff);
+}
+
+static inline unsigned int
+get_Imm8_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 43)) & 0xff);
+}
+
+static inline unsigned int
+get_Imm8_Y0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 12)) & 0xff);
+}
+
+static inline unsigned int
+get_Imm8_Y1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 43)) & 0xff);
+}
+
+static inline unsigned int
+get_JumpOff_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 31)) & 0x7ffffff);
+}
+
+static inline unsigned int
+get_JumpOpcodeExtension_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 58)) & 0x1);
+}
+
+static inline unsigned int
+get_MF_Imm14_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 37)) & 0x3fff);
+}
+
+static inline unsigned int
+get_MT_Imm14_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 31)) & 0x0000003f) |
+         (((unsigned int)(n >> 37)) & 0x00003fc0);
+}
+
+static inline unsigned int
+get_Mode(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 62)) & 0x3);
+}
+
+static inline unsigned int
+get_Opcode_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 28)) & 0x7);
+}
+
+static inline unsigned int
+get_Opcode_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 59)) & 0x7);
+}
+
+static inline unsigned int
+get_Opcode_Y0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 27)) & 0xf);
+}
+
+static inline unsigned int
+get_Opcode_Y1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 58)) & 0xf);
+}
+
+static inline unsigned int
+get_Opcode_Y2(tilegx_bundle_bits n)
+{
+  return (((n >> 26)) & 0x00000001) |
+         (((unsigned int)(n >> 56)) & 0x00000002);
+}
+
+static inline unsigned int
+get_RRROpcodeExtension_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 18)) & 0x3ff);
+}
+
+static inline unsigned int
+get_RRROpcodeExtension_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 49)) & 0x3ff);
+}
+
+static inline unsigned int
+get_RRROpcodeExtension_Y0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 18)) & 0x3);
+}
+
+static inline unsigned int
+get_RRROpcodeExtension_Y1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 49)) & 0x3);
+}
+
+static inline unsigned int
+get_ShAmt_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 12)) & 0x3f);
+}
+
+static inline unsigned int
+get_ShAmt_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 43)) & 0x3f);
+}
+
+static inline unsigned int
+get_ShAmt_Y0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 12)) & 0x3f);
+}
+
+static inline unsigned int
+get_ShAmt_Y1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 43)) & 0x3f);
+}
+
+static inline unsigned int
+get_ShiftOpcodeExtension_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 18)) & 0x3ff);
+}
+
+static inline unsigned int
+get_ShiftOpcodeExtension_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 49)) & 0x3ff);
+}
+
+static inline unsigned int
+get_ShiftOpcodeExtension_Y0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 18)) & 0x3);
+}
+
+static inline unsigned int
+get_ShiftOpcodeExtension_Y1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 49)) & 0x3);
+}
+
+static inline unsigned int
+get_SrcA_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 6)) & 0x3f);
+}
+
+static inline unsigned int
+get_SrcA_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 37)) & 0x3f);
+}
+
+static inline unsigned int
+get_SrcA_Y0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 6)) & 0x3f);
+}
+
+static inline unsigned int
+get_SrcA_Y1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 37)) & 0x3f);
+}
+
+static inline unsigned int
+get_SrcA_Y2(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 20)) & 0x3f);
+}
+
+static inline unsigned int
+get_SrcBDest_Y2(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 51)) & 0x3f);
+}
+
+static inline unsigned int
+get_SrcB_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 12)) & 0x3f);
+}
+
+static inline unsigned int
+get_SrcB_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 43)) & 0x3f);
+}
+
+static inline unsigned int
+get_SrcB_Y0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 12)) & 0x3f);
+}
+
+static inline unsigned int
+get_SrcB_Y1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 43)) & 0x3f);
+}
+
+static inline unsigned int
+get_UnaryOpcodeExtension_X0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 12)) & 0x3f);
+}
+
+static inline unsigned int
+get_UnaryOpcodeExtension_X1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 43)) & 0x3f);
+}
+
+static inline unsigned int
+get_UnaryOpcodeExtension_Y0(tilegx_bundle_bits num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((n >> 12)) & 0x3f);
+}
+
+static inline unsigned int
+get_UnaryOpcodeExtension_Y1(tilegx_bundle_bits n)
+{
+  return (((unsigned int)(n >> 43)) & 0x3f);
+}
+
+
+static inline int
+sign_extend(int n, int num_bits)
+{
+  int shift = (int)(sizeof(int) * 8 - num_bits);
+  return (n << shift) >> shift;
+}
+
+
+
+static inline tilegx_bundle_bits
+create_BFEnd_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3f) << 12);
+}
+
+static inline tilegx_bundle_bits
+create_BFOpcodeExtension_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0xf) << 24);
+}
+
+static inline tilegx_bundle_bits
+create_BFStart_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3f) << 18);
+}
+
+static inline tilegx_bundle_bits
+create_BrOff_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x0000003f)) << 31) |
+         (((tilegx_bundle_bits)(n & 0x0001ffc0)) << 37);
+}
+
+static inline tilegx_bundle_bits
+create_BrType_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x1f)) << 54);
+}
+
+static inline tilegx_bundle_bits
+create_Dest_Imm8_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x0000003f)) << 31) |
+         (((tilegx_bundle_bits)(n & 0x000000c0)) << 43);
+}
+
+static inline tilegx_bundle_bits
+create_Dest_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3f) << 0);
+}
+
+static inline tilegx_bundle_bits
+create_Dest_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3f)) << 31);
+}
+
+static inline tilegx_bundle_bits
+create_Dest_Y0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3f) << 0);
+}
+
+static inline tilegx_bundle_bits
+create_Dest_Y1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3f)) << 31);
+}
+
+static inline tilegx_bundle_bits
+create_Imm16_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0xffff) << 12);
+}
+
+static inline tilegx_bundle_bits
+create_Imm16_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0xffff)) << 43);
+}
+
+static inline tilegx_bundle_bits
+create_Imm8OpcodeExtension_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0xff) << 20);
+}
+
+static inline tilegx_bundle_bits
+create_Imm8OpcodeExtension_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0xff)) << 51);
+}
+
+static inline tilegx_bundle_bits
+create_Imm8_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0xff) << 12);
+}
+
+static inline tilegx_bundle_bits
+create_Imm8_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0xff)) << 43);
+}
+
+static inline tilegx_bundle_bits
+create_Imm8_Y0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0xff) << 12);
+}
+
+static inline tilegx_bundle_bits
+create_Imm8_Y1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0xff)) << 43);
+}
+
+static inline tilegx_bundle_bits
+create_JumpOff_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x7ffffff)) << 31);
+}
+
+static inline tilegx_bundle_bits
+create_JumpOpcodeExtension_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x1)) << 58);
+}
+
+static inline tilegx_bundle_bits
+create_MF_Imm14_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3fff)) << 37);
+}
+
+static inline tilegx_bundle_bits
+create_MT_Imm14_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x0000003f)) << 31) |
+         (((tilegx_bundle_bits)(n & 0x00003fc0)) << 37);
+}
+
+static inline tilegx_bundle_bits
+create_Mode(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3)) << 62);
+}
+
+static inline tilegx_bundle_bits
+create_Opcode_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x7) << 28);
+}
+
+static inline tilegx_bundle_bits
+create_Opcode_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x7)) << 59);
+}
+
+static inline tilegx_bundle_bits
+create_Opcode_Y0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0xf) << 27);
+}
+
+static inline tilegx_bundle_bits
+create_Opcode_Y1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0xf)) << 58);
+}
+
+static inline tilegx_bundle_bits
+create_Opcode_Y2(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x00000001) << 26) |
+         (((tilegx_bundle_bits)(n & 0x00000002)) << 56);
+}
+
+static inline tilegx_bundle_bits
+create_RRROpcodeExtension_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3ff) << 18);
+}
+
+static inline tilegx_bundle_bits
+create_RRROpcodeExtension_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3ff)) << 49);
+}
+
+static inline tilegx_bundle_bits
+create_RRROpcodeExtension_Y0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3) << 18);
+}
+
+static inline tilegx_bundle_bits
+create_RRROpcodeExtension_Y1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3)) << 49);
+}
+
+static inline tilegx_bundle_bits
+create_ShAmt_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3f) << 12);
+}
+
+static inline tilegx_bundle_bits
+create_ShAmt_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3f)) << 43);
+}
+
+static inline tilegx_bundle_bits
+create_ShAmt_Y0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3f) << 12);
+}
+
+static inline tilegx_bundle_bits
+create_ShAmt_Y1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3f)) << 43);
+}
+
+static inline tilegx_bundle_bits
+create_ShiftOpcodeExtension_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3ff) << 18);
+}
+
+static inline tilegx_bundle_bits
+create_ShiftOpcodeExtension_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3ff)) << 49);
+}
+
+static inline tilegx_bundle_bits
+create_ShiftOpcodeExtension_Y0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3) << 18);
+}
+
+static inline tilegx_bundle_bits
+create_ShiftOpcodeExtension_Y1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3)) << 49);
+}
+
+static inline tilegx_bundle_bits
+create_SrcA_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3f) << 6);
+}
+
+static inline tilegx_bundle_bits
+create_SrcA_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3f)) << 37);
+}
+
+static inline tilegx_bundle_bits
+create_SrcA_Y0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3f) << 6);
+}
+
+static inline tilegx_bundle_bits
+create_SrcA_Y1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3f)) << 37);
+}
+
+static inline tilegx_bundle_bits
+create_SrcA_Y2(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3f) << 20);
+}
+
+static inline tilegx_bundle_bits
+create_SrcBDest_Y2(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3f)) << 51);
+}
+
+static inline tilegx_bundle_bits
+create_SrcB_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3f) << 12);
+}
+
+static inline tilegx_bundle_bits
+create_SrcB_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3f)) << 43);
+}
+
+static inline tilegx_bundle_bits
+create_SrcB_Y0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3f) << 12);
+}
+
+static inline tilegx_bundle_bits
+create_SrcB_Y1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3f)) << 43);
+}
+
+static inline tilegx_bundle_bits
+create_UnaryOpcodeExtension_X0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3f) << 12);
+}
+
+static inline tilegx_bundle_bits
+create_UnaryOpcodeExtension_X1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3f)) << 43);
+}
+
+static inline tilegx_bundle_bits
+create_UnaryOpcodeExtension_Y0(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return ((n & 0x3f) << 12);
+}
+
+static inline tilegx_bundle_bits
+create_UnaryOpcodeExtension_Y1(int num)
+{
+  const unsigned int n = (unsigned int)num;
+  return (((tilegx_bundle_bits)(n & 0x3f)) << 43);
+}
+
+
+enum
+{
+  ADDI_IMM8_OPCODE_X0 = 1,
+  ADDI_IMM8_OPCODE_X1 = 1,
+  ADDI_OPCODE_Y0 = 0,
+  ADDI_OPCODE_Y1 = 1,
+  ADDLI_OPCODE_X0 = 1,
+  ADDLI_OPCODE_X1 = 0,
+  ADDXI_IMM8_OPCODE_X0 = 2,
+  ADDXI_IMM8_OPCODE_X1 = 2,
+  ADDXI_OPCODE_Y0 = 1,
+  ADDXI_OPCODE_Y1 = 2,
+  ADDXLI_OPCODE_X0 = 2,
+  ADDXLI_OPCODE_X1 = 1,
+  ADDXSC_RRR_0_OPCODE_X0 = 1,
+  ADDXSC_RRR_0_OPCODE_X1 = 1,
+  ADDX_RRR_0_OPCODE_X0 = 2,
+  ADDX_RRR_0_OPCODE_X1 = 2,
+  ADDX_RRR_0_OPCODE_Y0 = 0,
+  ADDX_RRR_0_OPCODE_Y1 = 0,
+  ADD_RRR_0_OPCODE_X0 = 3,
+  ADD_RRR_0_OPCODE_X1 = 3,
+  ADD_RRR_0_OPCODE_Y0 = 1,
+  ADD_RRR_0_OPCODE_Y1 = 1,
+  ANDI_IMM8_OPCODE_X0 = 3,
+  ANDI_IMM8_OPCODE_X1 = 3,
+  ANDI_OPCODE_Y0 = 2,
+  ANDI_OPCODE_Y1 = 3,
+  AND_RRR_0_OPCODE_X0 = 4,
+  AND_RRR_0_OPCODE_X1 = 4,
+  AND_RRR_5_OPCODE_Y0 = 0,
+  AND_RRR_5_OPCODE_Y1 = 0,
+  BEQZT_BRANCH_OPCODE_X1 = 16,
+  BEQZ_BRANCH_OPCODE_X1 = 17,
+  BFEXTS_BF_OPCODE_X0 = 4,
+  BFEXTU_BF_OPCODE_X0 = 5,
+  BFINS_BF_OPCODE_X0 = 6,
+  BF_OPCODE_X0 = 3,
+  BGEZT_BRANCH_OPCODE_X1 = 18,
+  BGEZ_BRANCH_OPCODE_X1 = 19,
+  BGTZT_BRANCH_OPCODE_X1 = 20,
+  BGTZ_BRANCH_OPCODE_X1 = 21,
+  BLBCT_BRANCH_OPCODE_X1 = 22,
+  BLBC_BRANCH_OPCODE_X1 = 23,
+  BLBST_BRANCH_OPCODE_X1 = 24,
+  BLBS_BRANCH_OPCODE_X1 = 25,
+  BLEZT_BRANCH_OPCODE_X1 = 26,
+  BLEZ_BRANCH_OPCODE_X1 = 27,
+  BLTZT_BRANCH_OPCODE_X1 = 28,
+  BLTZ_BRANCH_OPCODE_X1 = 29,
+  BNEZT_BRANCH_OPCODE_X1 = 30,
+  BNEZ_BRANCH_OPCODE_X1 = 31,
+  BRANCH_OPCODE_X1 = 2,
+  CMOVEQZ_RRR_0_OPCODE_X0 = 5,
+  CMOVEQZ_RRR_4_OPCODE_Y0 = 0,
+  CMOVNEZ_RRR_0_OPCODE_X0 = 6,
+  CMOVNEZ_RRR_4_OPCODE_Y0 = 1,
+  CMPEQI_IMM8_OPCODE_X0 = 4,
+  CMPEQI_IMM8_OPCODE_X1 = 4,
+  CMPEQI_OPCODE_Y0 = 3,
+  CMPEQI_OPCODE_Y1 = 4,
+  CMPEQ_RRR_0_OPCODE_X0 = 7,
+  CMPEQ_RRR_0_OPCODE_X1 = 5,
+  CMPEQ_RRR_3_OPCODE_Y0 = 0,
+  CMPEQ_RRR_3_OPCODE_Y1 = 2,
+  CMPEXCH4_RRR_0_OPCODE_X1 = 6,
+  CMPEXCH_RRR_0_OPCODE_X1 = 7,
+  CMPLES_RRR_0_OPCODE_X0 = 8,
+  CMPLES_RRR_0_OPCODE_X1 = 8,
+  CMPLES_RRR_2_OPCODE_Y0 = 0,
+  CMPLES_RRR_2_OPCODE_Y1 = 0,
+  CMPLEU_RRR_0_OPCODE_X0 = 9,
+  CMPLEU_RRR_0_OPCODE_X1 = 9,
+  CMPLEU_RRR_2_OPCODE_Y0 = 1,
+  CMPLEU_RRR_2_OPCODE_Y1 = 1,
+  CMPLTSI_IMM8_OPCODE_X0 = 5,
+  CMPLTSI_IMM8_OPCODE_X1 = 5,
+  CMPLTSI_OPCODE_Y0 = 4,
+  CMPLTSI_OPCODE_Y1 = 5,
+  CMPLTS_RRR_0_OPCODE_X0 = 10,
+  CMPLTS_RRR_0_OPCODE_X1 = 10,
+  CMPLTS_RRR_2_OPCODE_Y0 = 2,
+  CMPLTS_RRR_2_OPCODE_Y1 = 2,
+  CMPLTUI_IMM8_OPCODE_X0 = 6,
+  CMPLTUI_IMM8_OPCODE_X1 = 6,
+  CMPLTU_RRR_0_OPCODE_X0 = 11,
+  CMPLTU_RRR_0_OPCODE_X1 = 11,
+  CMPLTU_RRR_2_OPCODE_Y0 = 3,
+  CMPLTU_RRR_2_OPCODE_Y1 = 3,
+  CMPNE_RRR_0_OPCODE_X0 = 12,
+  CMPNE_RRR_0_OPCODE_X1 = 12,
+  CMPNE_RRR_3_OPCODE_Y0 = 1,
+  CMPNE_RRR_3_OPCODE_Y1 = 3,
+  CMULAF_RRR_0_OPCODE_X0 = 13,
+  CMULA_RRR_0_OPCODE_X0 = 14,
+  CMULFR_RRR_0_OPCODE_X0 = 15,
+  CMULF_RRR_0_OPCODE_X0 = 16,
+  CMULHR_RRR_0_OPCODE_X0 = 17,
+  CMULH_RRR_0_OPCODE_X0 = 18,
+  CMUL_RRR_0_OPCODE_X0 = 19,
+  CNTLZ_UNARY_OPCODE_X0 = 1,
+  CNTLZ_UNARY_OPCODE_Y0 = 1,
+  CNTTZ_UNARY_OPCODE_X0 = 2,
+  CNTTZ_UNARY_OPCODE_Y0 = 2,
+  CRC32_32_RRR_0_OPCODE_X0 = 20,
+  CRC32_8_RRR_0_OPCODE_X0 = 21,
+  DBLALIGN2_RRR_0_OPCODE_X0 = 22,
+  DBLALIGN2_RRR_0_OPCODE_X1 = 13,
+  DBLALIGN4_RRR_0_OPCODE_X0 = 23,
+  DBLALIGN4_RRR_0_OPCODE_X1 = 14,
+  DBLALIGN6_RRR_0_OPCODE_X0 = 24,
+  DBLALIGN6_RRR_0_OPCODE_X1 = 15,
+  DBLALIGN_RRR_0_OPCODE_X0 = 25,
+  DRAIN_UNARY_OPCODE_X1 = 1,
+  DTLBPR_UNARY_OPCODE_X1 = 2,
+  EXCH4_RRR_0_OPCODE_X1 = 16,
+  EXCH_RRR_0_OPCODE_X1 = 17,
+  FDOUBLE_ADDSUB_RRR_0_OPCODE_X0 = 26,
+  FDOUBLE_ADD_FLAGS_RRR_0_OPCODE_X0 = 27,
+  FDOUBLE_MUL_FLAGS_RRR_0_OPCODE_X0 = 28,
+  FDOUBLE_PACK1_RRR_0_OPCODE_X0 = 29,
+  FDOUBLE_PACK2_RRR_0_OPCODE_X0 = 30,
+  FDOUBLE_SUB_FLAGS_RRR_0_OPCODE_X0 = 31,
+  FDOUBLE_UNPACK_MAX_RRR_0_OPCODE_X0 = 32,
+  FDOUBLE_UNPACK_MIN_RRR_0_OPCODE_X0 = 33,
+  FETCHADD4_RRR_0_OPCODE_X1 = 18,
+  FETCHADDGEZ4_RRR_0_OPCODE_X1 = 19,
+  FETCHADDGEZ_RRR_0_OPCODE_X1 = 20,
+  FETCHADD_RRR_0_OPCODE_X1 = 21,
+  FETCHAND4_RRR_0_OPCODE_X1 = 22,
+  FETCHAND_RRR_0_OPCODE_X1 = 23,
+  FETCHOR4_RRR_0_OPCODE_X1 = 24,
+  FETCHOR_RRR_0_OPCODE_X1 = 25,
+  FINV_UNARY_OPCODE_X1 = 3,
+  FLUSHWB_UNARY_OPCODE_X1 = 4,
+  FLUSH_UNARY_OPCODE_X1 = 5,
+  FNOP_UNARY_OPCODE_X0 = 3,
+  FNOP_UNARY_OPCODE_X1 = 6,
+  FNOP_UNARY_OPCODE_Y0 = 3,
+  FNOP_UNARY_OPCODE_Y1 = 8,
+  FSINGLE_ADD1_RRR_0_OPCODE_X0 = 34,
+  FSINGLE_ADDSUB2_RRR_0_OPCODE_X0 = 35,
+  FSINGLE_MUL1_RRR_0_OPCODE_X0 = 36,
+  FSINGLE_MUL2_RRR_0_OPCODE_X0 = 37,
+  FSINGLE_PACK1_UNARY_OPCODE_X0 = 4,
+  FSINGLE_PACK1_UNARY_OPCODE_Y0 = 4,
+  FSINGLE_PACK2_RRR_0_OPCODE_X0 = 38,
+  FSINGLE_SUB1_RRR_0_OPCODE_X0 = 39,
+  ICOH_UNARY_OPCODE_X1 = 7,
+  ILL_UNARY_OPCODE_X1 = 8,
+  ILL_UNARY_OPCODE_Y1 = 9,
+  IMM8_OPCODE_X0 = 4,
+  IMM8_OPCODE_X1 = 3,
+  INV_UNARY_OPCODE_X1 = 9,
+  IRET_UNARY_OPCODE_X1 = 10,
+  JALRP_UNARY_OPCODE_X1 = 11,
+  JALRP_UNARY_OPCODE_Y1 = 10,
+  JALR_UNARY_OPCODE_X1 = 12,
+  JALR_UNARY_OPCODE_Y1 = 11,
+  JAL_JUMP_OPCODE_X1 = 0,
+  JRP_UNARY_OPCODE_X1 = 13,
+  JRP_UNARY_OPCODE_Y1 = 12,
+  JR_UNARY_OPCODE_X1 = 14,
+  JR_UNARY_OPCODE_Y1 = 13,
+  JUMP_OPCODE_X1 = 4,
+  J_JUMP_OPCODE_X1 = 1,
+  LD1S_ADD_IMM8_OPCODE_X1 = 7,
+  LD1S_OPCODE_Y2 = 0,
+  LD1S_UNARY_OPCODE_X1 = 15,
+  LD1U_ADD_IMM8_OPCODE_X1 = 8,
+  LD1U_OPCODE_Y2 = 1,
+  LD1U_UNARY_OPCODE_X1 = 16,
+  LD2S_ADD_IMM8_OPCODE_X1 = 9,
+  LD2S_OPCODE_Y2 = 2,
+  LD2S_UNARY_OPCODE_X1 = 17,
+  LD2U_ADD_IMM8_OPCODE_X1 = 10,
+  LD2U_OPCODE_Y2 = 3,
+  LD2U_UNARY_OPCODE_X1 = 18,
+  LD4S_ADD_IMM8_OPCODE_X1 = 11,
+  LD4S_OPCODE_Y2 = 1,
+  LD4S_UNARY_OPCODE_X1 = 19,
+  LD4U_ADD_IMM8_OPCODE_X1 = 12,
+  LD4U_OPCODE_Y2 = 2,
+  LD4U_UNARY_OPCODE_X1 = 20,
+  LDNA_UNARY_OPCODE_X1 = 21,
+  LDNT1S_ADD_IMM8_OPCODE_X1 = 13,
+  LDNT1S_UNARY_OPCODE_X1 = 22,
+  LDNT1U_ADD_IMM8_OPCODE_X1 = 14,
+  LDNT1U_UNARY_OPCODE_X1 = 23,
+  LDNT2S_ADD_IMM8_OPCODE_X1 = 15,
+  LDNT2S_UNARY_OPCODE_X1 = 24,
+  LDNT2U_ADD_IMM8_OPCODE_X1 = 16,
+  LDNT2U_UNARY_OPCODE_X1 = 25,
+  LDNT4S_ADD_IMM8_OPCODE_X1 = 17,
+  LDNT4S_UNARY_OPCODE_X1 = 26,
+  LDNT4U_ADD_IMM8_OPCODE_X1 = 18,
+  LDNT4U_UNARY_OPCODE_X1 = 27,
+  LDNT_ADD_IMM8_OPCODE_X1 = 19,
+  LDNT_UNARY_OPCODE_X1 = 28,
+  LD_ADD_IMM8_OPCODE_X1 = 20,
+  LD_OPCODE_Y2 = 3,
+  LD_UNARY_OPCODE_X1 = 29,
+  LNK_UNARY_OPCODE_X1 = 30,
+  LNK_UNARY_OPCODE_Y1 = 14,
+  LDNA_ADD_IMM8_OPCODE_X1 = 21,
+  MFSPR_IMM8_OPCODE_X1 = 22,
+  MF_UNARY_OPCODE_X1 = 31,
+  MM_BF_OPCODE_X0 = 7,
+  MNZ_RRR_0_OPCODE_X0 = 40,
+  MNZ_RRR_0_OPCODE_X1 = 26,
+  MNZ_RRR_4_OPCODE_Y0 = 2,
+  MNZ_RRR_4_OPCODE_Y1 = 2,
+  MODE_OPCODE_YA2 = 1,
+  MODE_OPCODE_YB2 = 2,
+  MODE_OPCODE_YC2 = 3,
+  MTSPR_IMM8_OPCODE_X1 = 23,
+  MULAX_RRR_0_OPCODE_X0 = 41,
+  MULAX_RRR_3_OPCODE_Y0 = 2,
+  MULA_HS_HS_RRR_0_OPCODE_X0 = 42,
+  MULA_HS_HS_RRR_9_OPCODE_Y0 = 0,
+  MULA_HS_HU_RRR_0_OPCODE_X0 = 43,
+  MULA_HS_LS_RRR_0_OPCODE_X0 = 44,
+  MULA_HS_LU_RRR_0_OPCODE_X0 = 45,
+  MULA_HU_HU_RRR_0_OPCODE_X0 = 46,
+  MULA_HU_HU_RRR_9_OPCODE_Y0 = 1,
+  MULA_HU_LS_RRR_0_OPCODE_X0 = 47,
+  MULA_HU_LU_RRR_0_OPCODE_X0 = 48,
+  MULA_LS_LS_RRR_0_OPCODE_X0 = 49,
+  MULA_LS_LS_RRR_9_OPCODE_Y0 = 2,
+  MULA_LS_LU_RRR_0_OPCODE_X0 = 50,
+  MULA_LU_LU_RRR_0_OPCODE_X0 = 51,
+  MULA_LU_LU_RRR_9_OPCODE_Y0 = 3,
+  MULX_RRR_0_OPCODE_X0 = 52,
+  MULX_RRR_3_OPCODE_Y0 = 3,
+  MUL_HS_HS_RRR_0_OPCODE_X0 = 53,
+  MUL_HS_HS_RRR_8_OPCODE_Y0 = 0,
+  MUL_HS_HU_RRR_0_OPCODE_X0 = 54,
+  MUL_HS_LS_RRR_0_OPCODE_X0 = 55,
+  MUL_HS_LU_RRR_0_OPCODE_X0 = 56,
+  MUL_HU_HU_RRR_0_OPCODE_X0 = 57,
+  MUL_HU_HU_RRR_8_OPCODE_Y0 = 1,
+  MUL_HU_LS_RRR_0_OPCODE_X0 = 58,
+  MUL_HU_LU_RRR_0_OPCODE_X0 = 59,
+  MUL_LS_LS_RRR_0_OPCODE_X0 = 60,
+  MUL_LS_LS_RRR_8_OPCODE_Y0 = 2,
+  MUL_LS_LU_RRR_0_OPCODE_X0 = 61,
+  MUL_LU_LU_RRR_0_OPCODE_X0 = 62,
+  MUL_LU_LU_RRR_8_OPCODE_Y0 = 3,
+  MZ_RRR_0_OPCODE_X0 = 63,
+  MZ_RRR_0_OPCODE_X1 = 27,
+  MZ_RRR_4_OPCODE_Y0 = 3,
+  MZ_RRR_4_OPCODE_Y1 = 3,
+  NAP_UNARY_OPCODE_X1 = 32,
+  NOP_UNARY_OPCODE_X0 = 5,
+  NOP_UNARY_OPCODE_X1 = 33,
+  NOP_UNARY_OPCODE_Y0 = 5,
+  NOP_UNARY_OPCODE_Y1 = 15,
+  NOR_RRR_0_OPCODE_X0 = 64,
+  NOR_RRR_0_OPCODE_X1 = 28,
+  NOR_RRR_5_OPCODE_Y0 = 1,
+  NOR_RRR_5_OPCODE_Y1 = 1,
+  ORI_IMM8_OPCODE_X0 = 7,
+  ORI_IMM8_OPCODE_X1 = 24,
+  OR_RRR_0_OPCODE_X0 = 65,
+  OR_RRR_0_OPCODE_X1 = 29,
+  OR_RRR_5_OPCODE_Y0 = 2,
+  OR_RRR_5_OPCODE_Y1 = 2,
+  PCNT_UNARY_OPCODE_X0 = 6,
+  PCNT_UNARY_OPCODE_Y0 = 6,
+  REVBITS_UNARY_OPCODE_X0 = 7,
+  REVBITS_UNARY_OPCODE_Y0 = 7,
+  REVBYTES_UNARY_OPCODE_X0 = 8,
+  REVBYTES_UNARY_OPCODE_Y0 = 8,
+  ROTLI_SHIFT_OPCODE_X0 = 1,
+  ROTLI_SHIFT_OPCODE_X1 = 1,
+  ROTLI_SHIFT_OPCODE_Y0 = 0,
+  ROTLI_SHIFT_OPCODE_Y1 = 0,
+  ROTL_RRR_0_OPCODE_X0 = 66,
+  ROTL_RRR_0_OPCODE_X1 = 30,
+  ROTL_RRR_6_OPCODE_Y0 = 0,
+  ROTL_RRR_6_OPCODE_Y1 = 0,
+  RRR_0_OPCODE_X0 = 5,
+  RRR_0_OPCODE_X1 = 5,
+  RRR_0_OPCODE_Y0 = 5,
+  RRR_0_OPCODE_Y1 = 6,
+  RRR_1_OPCODE_Y0 = 6,
+  RRR_1_OPCODE_Y1 = 7,
+  RRR_2_OPCODE_Y0 = 7,
+  RRR_2_OPCODE_Y1 = 8,
+  RRR_3_OPCODE_Y0 = 8,
+  RRR_3_OPCODE_Y1 = 9,
+  RRR_4_OPCODE_Y0 = 9,
+  RRR_4_OPCODE_Y1 = 10,
+  RRR_5_OPCODE_Y0 = 10,
+  RRR_5_OPCODE_Y1 = 11,
+  RRR_6_OPCODE_Y0 = 11,
+  RRR_6_OPCODE_Y1 = 12,
+  RRR_7_OPCODE_Y0 = 12,
+  RRR_7_OPCODE_Y1 = 13,
+  RRR_8_OPCODE_Y0 = 13,
+  RRR_9_OPCODE_Y0 = 14,
+  SHIFT_OPCODE_X0 = 6,
+  SHIFT_OPCODE_X1 = 6,
+  SHIFT_OPCODE_Y0 = 15,
+  SHIFT_OPCODE_Y1 = 14,
+  SHL16INSLI_OPCODE_X0 = 7,
+  SHL16INSLI_OPCODE_X1 = 7,
+  SHL1ADDX_RRR_0_OPCODE_X0 = 67,
+  SHL1ADDX_RRR_0_OPCODE_X1 = 31,
+  SHL1ADDX_RRR_7_OPCODE_Y0 = 1,
+  SHL1ADDX_RRR_7_OPCODE_Y1 = 1,
+  SHL1ADD_RRR_0_OPCODE_X0 = 68,
+  SHL1ADD_RRR_0_OPCODE_X1 = 32,
+  SHL1ADD_RRR_1_OPCODE_Y0 = 0,
+  SHL1ADD_RRR_1_OPCODE_Y1 = 0,
+  SHL2ADDX_RRR_0_OPCODE_X0 = 69,
+  SHL2ADDX_RRR_0_OPCODE_X1 = 33,
+  SHL2ADDX_RRR_7_OPCODE_Y0 = 2,
+  SHL2ADDX_RRR_7_OPCODE_Y1 = 2,
+  SHL2ADD_RRR_0_OPCODE_X0 = 70,
+  SHL2ADD_RRR_0_OPCODE_X1 = 34,
+  SHL2ADD_RRR_1_OPCODE_Y0 = 1,
+  SHL2ADD_RRR_1_OPCODE_Y1 = 1,
+  SHL3ADDX_RRR_0_OPCODE_X0 = 71,
+  SHL3ADDX_RRR_0_OPCODE_X1 = 35,
+  SHL3ADDX_RRR_7_OPCODE_Y0 = 3,
+  SHL3ADDX_RRR_7_OPCODE_Y1 = 3,
+  SHL3ADD_RRR_0_OPCODE_X0 = 72,
+  SHL3ADD_RRR_0_OPCODE_X1 = 36,
+  SHL3ADD_RRR_1_OPCODE_Y0 = 2,
+  SHL3ADD_RRR_1_OPCODE_Y1 = 2,
+  SHLI_SHIFT_OPCODE_X0 = 2,
+  SHLI_SHIFT_OPCODE_X1 = 2,
+  SHLI_SHIFT_OPCODE_Y0 = 1,
+  SHLI_SHIFT_OPCODE_Y1 = 1,
+  SHLXI_SHIFT_OPCODE_X0 = 3,
+  SHLXI_SHIFT_OPCODE_X1 = 3,
+  SHLX_RRR_0_OPCODE_X0 = 73,
+  SHLX_RRR_0_OPCODE_X1 = 37,
+  SHL_RRR_0_OPCODE_X0 = 74,
+  SHL_RRR_0_OPCODE_X1 = 38,
+  SHL_RRR_6_OPCODE_Y0 = 1,
+  SHL_RRR_6_OPCODE_Y1 = 1,
+  SHRSI_SHIFT_OPCODE_X0 = 4,
+  SHRSI_SHIFT_OPCODE_X1 = 4,
+  SHRSI_SHIFT_OPCODE_Y0 = 2,
+  SHRSI_SHIFT_OPCODE_Y1 = 2,
+  SHRS_RRR_0_OPCODE_X0 = 75,
+  SHRS_RRR_0_OPCODE_X1 = 39,
+  SHRS_RRR_6_OPCODE_Y0 = 2,
+  SHRS_RRR_6_OPCODE_Y1 = 2,
+  SHRUI_SHIFT_OPCODE_X0 = 5,
+  SHRUI_SHIFT_OPCODE_X1 = 5,
+  SHRUI_SHIFT_OPCODE_Y0 = 3,
+  SHRUI_SHIFT_OPCODE_Y1 = 3,
+  SHRUXI_SHIFT_OPCODE_X0 = 6,
+  SHRUXI_SHIFT_OPCODE_X1 = 6,
+  SHRUX_RRR_0_OPCODE_X0 = 76,
+  SHRUX_RRR_0_OPCODE_X1 = 40,
+  SHRU_RRR_0_OPCODE_X0 = 77,
+  SHRU_RRR_0_OPCODE_X1 = 41,
+  SHRU_RRR_6_OPCODE_Y0 = 3,
+  SHRU_RRR_6_OPCODE_Y1 = 3,
+  SHUFFLEBYTES_RRR_0_OPCODE_X0 = 78,
+  ST1_ADD_IMM8_OPCODE_X1 = 25,
+  ST1_OPCODE_Y2 = 0,
+  ST1_RRR_0_OPCODE_X1 = 42,
+  ST2_ADD_IMM8_OPCODE_X1 = 26,
+  ST2_OPCODE_Y2 = 1,
+  ST2_RRR_0_OPCODE_X1 = 43,
+  ST4_ADD_IMM8_OPCODE_X1 = 27,
+  ST4_OPCODE_Y2 = 2,
+  ST4_RRR_0_OPCODE_X1 = 44,
+  STNT1_ADD_IMM8_OPCODE_X1 = 28,
+  STNT1_RRR_0_OPCODE_X1 = 45,
+  STNT2_ADD_IMM8_OPCODE_X1 = 29,
+  STNT2_RRR_0_OPCODE_X1 = 46,
+  STNT4_ADD_IMM8_OPCODE_X1 = 30,
+  STNT4_RRR_0_OPCODE_X1 = 47,
+  STNT_ADD_IMM8_OPCODE_X1 = 31,
+  STNT_RRR_0_OPCODE_X1 = 48,
+  ST_ADD_IMM8_OPCODE_X1 = 32,
+  ST_OPCODE_Y2 = 3,
+  ST_RRR_0_OPCODE_X1 = 49,
+  SUBXSC_RRR_0_OPCODE_X0 = 79,
+  SUBXSC_RRR_0_OPCODE_X1 = 50,
+  SUBX_RRR_0_OPCODE_X0 = 80,
+  SUBX_RRR_0_OPCODE_X1 = 51,
+  SUBX_RRR_0_OPCODE_Y0 = 2,
+  SUBX_RRR_0_OPCODE_Y1 = 2,
+  SUB_RRR_0_OPCODE_X0 = 81,
+  SUB_RRR_0_OPCODE_X1 = 52,
+  SUB_RRR_0_OPCODE_Y0 = 3,
+  SUB_RRR_0_OPCODE_Y1 = 3,
+  SWINT0_UNARY_OPCODE_X1 = 34,
+  SWINT1_UNARY_OPCODE_X1 = 35,
+  SWINT2_UNARY_OPCODE_X1 = 36,
+  SWINT3_UNARY_OPCODE_X1 = 37,
+  TBLIDXB0_UNARY_OPCODE_X0 = 9,
+  TBLIDXB0_UNARY_OPCODE_Y0 = 9,
+  TBLIDXB1_UNARY_OPCODE_X0 = 10,
+  TBLIDXB1_UNARY_OPCODE_Y0 = 10,
+  TBLIDXB2_UNARY_OPCODE_X0 = 11,
+  TBLIDXB2_UNARY_OPCODE_Y0 = 11,
+  TBLIDXB3_UNARY_OPCODE_X0 = 12,
+  TBLIDXB3_UNARY_OPCODE_Y0 = 12,
+  UNARY_RRR_0_OPCODE_X0 = 82,
+  UNARY_RRR_0_OPCODE_X1 = 53,
+  UNARY_RRR_1_OPCODE_Y0 = 3,
+  UNARY_RRR_1_OPCODE_Y1 = 3,
+  V1ADDI_IMM8_OPCODE_X0 = 8,
+  V1ADDI_IMM8_OPCODE_X1 = 33,
+  V1ADDUC_RRR_0_OPCODE_X0 = 83,
+  V1ADDUC_RRR_0_OPCODE_X1 = 54,
+  V1ADD_RRR_0_OPCODE_X0 = 84,
+  V1ADD_RRR_0_OPCODE_X1 = 55,
+  V1ADIFFU_RRR_0_OPCODE_X0 = 85,
+  V1AVGU_RRR_0_OPCODE_X0 = 86,
+  V1CMPEQI_IMM8_OPCODE_X0 = 9,
+  V1CMPEQI_IMM8_OPCODE_X1 = 34,
+  V1CMPEQ_RRR_0_OPCODE_X0 = 87,
+  V1CMPEQ_RRR_0_OPCODE_X1 = 56,
+  V1CMPLES_RRR_0_OPCODE_X0 = 88,
+  V1CMPLES_RRR_0_OPCODE_X1 = 57,
+  V1CMPLEU_RRR_0_OPCODE_X0 = 89,
+  V1CMPLEU_RRR_0_OPCODE_X1 = 58,
+  V1CMPLTSI_IMM8_OPCODE_X0 = 10,
+  V1CMPLTSI_IMM8_OPCODE_X1 = 35,
+  V1CMPLTS_RRR_0_OPCODE_X0 = 90,
+  V1CMPLTS_RRR_0_OPCODE_X1 = 59,
+  V1CMPLTUI_IMM8_OPCODE_X0 = 11,
+  V1CMPLTUI_IMM8_OPCODE_X1 = 36,
+  V1CMPLTU_RRR_0_OPCODE_X0 = 91,
+  V1CMPLTU_RRR_0_OPCODE_X1 = 60,
+  V1CMPNE_RRR_0_OPCODE_X0 = 92,
+  V1CMPNE_RRR_0_OPCODE_X1 = 61,
+  V1DDOTPUA_RRR_0_OPCODE_X0 = 161,
+  V1DDOTPUSA_RRR_0_OPCODE_X0 = 93,
+  V1DDOTPUS_RRR_0_OPCODE_X0 = 94,
+  V1DDOTPU_RRR_0_OPCODE_X0 = 162,
+  V1DOTPA_RRR_0_OPCODE_X0 = 95,
+  V1DOTPUA_RRR_0_OPCODE_X0 = 163,
+  V1DOTPUSA_RRR_0_OPCODE_X0 = 96,
+  V1DOTPUS_RRR_0_OPCODE_X0 = 97,
+  V1DOTPU_RRR_0_OPCODE_X0 = 164,
+  V1DOTP_RRR_0_OPCODE_X0 = 98,
+  V1INT_H_RRR_0_OPCODE_X0 = 99,
+  V1INT_H_RRR_0_OPCODE_X1 = 62,
+  V1INT_L_RRR_0_OPCODE_X0 = 100,
+  V1INT_L_RRR_0_OPCODE_X1 = 63,
+  V1MAXUI_IMM8_OPCODE_X0 = 12,
+  V1MAXUI_IMM8_OPCODE_X1 = 37,
+  V1MAXU_RRR_0_OPCODE_X0 = 101,
+  V1MAXU_RRR_0_OPCODE_X1 = 64,
+  V1MINUI_IMM8_OPCODE_X0 = 13,
+  V1MINUI_IMM8_OPCODE_X1 = 38,
+  V1MINU_RRR_0_OPCODE_X0 = 102,
+  V1MINU_RRR_0_OPCODE_X1 = 65,
+  V1MNZ_RRR_0_OPCODE_X0 = 103,
+  V1MNZ_RRR_0_OPCODE_X1 = 66,
+  V1MULTU_RRR_0_OPCODE_X0 = 104,
+  V1MULUS_RRR_0_OPCODE_X0 = 105,
+  V1MULU_RRR_0_OPCODE_X0 = 106,
+  V1MZ_RRR_0_OPCODE_X0 = 107,
+  V1MZ_RRR_0_OPCODE_X1 = 67,
+  V1SADAU_RRR_0_OPCODE_X0 = 108,
+  V1SADU_RRR_0_OPCODE_X0 = 109,
+  V1SHLI_SHIFT_OPCODE_X0 = 7,
+  V1SHLI_SHIFT_OPCODE_X1 = 7,
+  V1SHL_RRR_0_OPCODE_X0 = 110,
+  V1SHL_RRR_0_OPCODE_X1 = 68,
+  V1SHRSI_SHIFT_OPCODE_X0 = 8,
+  V1SHRSI_SHIFT_OPCODE_X1 = 8,
+  V1SHRS_RRR_0_OPCODE_X0 = 111,
+  V1SHRS_RRR_0_OPCODE_X1 = 69,
+  V1SHRUI_SHIFT_OPCODE_X0 = 9,
+  V1SHRUI_SHIFT_OPCODE_X1 = 9,
+  V1SHRU_RRR_0_OPCODE_X0 = 112,
+  V1SHRU_RRR_0_OPCODE_X1 = 70,
+  V1SUBUC_RRR_0_OPCODE_X0 = 113,
+  V1SUBUC_RRR_0_OPCODE_X1 = 71,
+  V1SUB_RRR_0_OPCODE_X0 = 114,
+  V1SUB_RRR_0_OPCODE_X1 = 72,
+  V2ADDI_IMM8_OPCODE_X0 = 14,
+  V2ADDI_IMM8_OPCODE_X1 = 39,
+  V2ADDSC_RRR_0_OPCODE_X0 = 115,
+  V2ADDSC_RRR_0_OPCODE_X1 = 73,
+  V2ADD_RRR_0_OPCODE_X0 = 116,
+  V2ADD_RRR_0_OPCODE_X1 = 74,
+  V2ADIFFS_RRR_0_OPCODE_X0 = 117,
+  V2AVGS_RRR_0_OPCODE_X0 = 118,
+  V2CMPEQI_IMM8_OPCODE_X0 = 15,
+  V2CMPEQI_IMM8_OPCODE_X1 = 40,
+  V2CMPEQ_RRR_0_OPCODE_X0 = 119,
+  V2CMPEQ_RRR_0_OPCODE_X1 = 75,
+  V2CMPLES_RRR_0_OPCODE_X0 = 120,
+  V2CMPLES_RRR_0_OPCODE_X1 = 76,
+  V2CMPLEU_RRR_0_OPCODE_X0 = 121,
+  V2CMPLEU_RRR_0_OPCODE_X1 = 77,
+  V2CMPLTSI_IMM8_OPCODE_X0 = 16,
+  V2CMPLTSI_IMM8_OPCODE_X1 = 41,
+  V2CMPLTS_RRR_0_OPCODE_X0 = 122,
+  V2CMPLTS_RRR_0_OPCODE_X1 = 78,
+  V2CMPLTUI_IMM8_OPCODE_X0 = 17,
+  V2CMPLTUI_IMM8_OPCODE_X1 = 42,
+  V2CMPLTU_RRR_0_OPCODE_X0 = 123,
+  V2CMPLTU_RRR_0_OPCODE_X1 = 79,
+  V2CMPNE_RRR_0_OPCODE_X0 = 124,
+  V2CMPNE_RRR_0_OPCODE_X1 = 80,
+  V2DOTPA_RRR_0_OPCODE_X0 = 125,
+  V2DOTP_RRR_0_OPCODE_X0 = 126,
+  V2INT_H_RRR_0_OPCODE_X0 = 127,
+  V2INT_H_RRR_0_OPCODE_X1 = 81,
+  V2INT_L_RRR_0_OPCODE_X0 = 128,
+  V2INT_L_RRR_0_OPCODE_X1 = 82,
+  V2MAXSI_IMM8_OPCODE_X0 = 18,
+  V2MAXSI_IMM8_OPCODE_X1 = 43,
+  V2MAXS_RRR_0_OPCODE_X0 = 129,
+  V2MAXS_RRR_0_OPCODE_X1 = 83,
+  V2MINSI_IMM8_OPCODE_X0 = 19,
+  V2MINSI_IMM8_OPCODE_X1 = 44,
+  V2MINS_RRR_0_OPCODE_X0 = 130,
+  V2MINS_RRR_0_OPCODE_X1 = 84,
+  V2MNZ_RRR_0_OPCODE_X0 = 131,
+  V2MNZ_RRR_0_OPCODE_X1 = 85,
+  V2MULFSC_RRR_0_OPCODE_X0 = 132,
+  V2MULS_RRR_0_OPCODE_X0 = 133,
+  V2MULTS_RRR_0_OPCODE_X0 = 134,
+  V2MZ_RRR_0_OPCODE_X0 = 135,
+  V2MZ_RRR_0_OPCODE_X1 = 86,
+  V2PACKH_RRR_0_OPCODE_X0 = 136,
+  V2PACKH_RRR_0_OPCODE_X1 = 87,
+  V2PACKL_RRR_0_OPCODE_X0 = 137,
+  V2PACKL_RRR_0_OPCODE_X1 = 88,
+  V2PACKUC_RRR_0_OPCODE_X0 = 138,
+  V2PACKUC_RRR_0_OPCODE_X1 = 89,
+  V2SADAS_RRR_0_OPCODE_X0 = 139,
+  V2SADAU_RRR_0_OPCODE_X0 = 140,
+  V2SADS_RRR_0_OPCODE_X0 = 141,
+  V2SADU_RRR_0_OPCODE_X0 = 142,
+  V2SHLI_SHIFT_OPCODE_X0 = 10,
+  V2SHLI_SHIFT_OPCODE_X1 = 10,
+  V2SHLSC_RRR_0_OPCODE_X0 = 143,
+  V2SHLSC_RRR_0_OPCODE_X1 = 90,
+  V2SHL_RRR_0_OPCODE_X0 = 144,
+  V2SHL_RRR_0_OPCODE_X1 = 91,
+  V2SHRSI_SHIFT_OPCODE_X0 = 11,
+  V2SHRSI_SHIFT_OPCODE_X1 = 11,
+  V2SHRS_RRR_0_OPCODE_X0 = 145,
+  V2SHRS_RRR_0_OPCODE_X1 = 92,
+  V2SHRUI_SHIFT_OPCODE_X0 = 12,
+  V2SHRUI_SHIFT_OPCODE_X1 = 12,
+  V2SHRU_RRR_0_OPCODE_X0 = 146,
+  V2SHRU_RRR_0_OPCODE_X1 = 93,
+  V2SUBSC_RRR_0_OPCODE_X0 = 147,
+  V2SUBSC_RRR_0_OPCODE_X1 = 94,
+  V2SUB_RRR_0_OPCODE_X0 = 148,
+  V2SUB_RRR_0_OPCODE_X1 = 95,
+  V4ADDSC_RRR_0_OPCODE_X0 = 149,
+  V4ADDSC_RRR_0_OPCODE_X1 = 96,
+  V4ADD_RRR_0_OPCODE_X0 = 150,
+  V4ADD_RRR_0_OPCODE_X1 = 97,
+  V4INT_H_RRR_0_OPCODE_X0 = 151,
+  V4INT_H_RRR_0_OPCODE_X1 = 98,
+  V4INT_L_RRR_0_OPCODE_X0 = 152,
+  V4INT_L_RRR_0_OPCODE_X1 = 99,
+  V4PACKSC_RRR_0_OPCODE_X0 = 153,
+  V4PACKSC_RRR_0_OPCODE_X1 = 100,
+  V4SHLSC_RRR_0_OPCODE_X0 = 154,
+  V4SHLSC_RRR_0_OPCODE_X1 = 101,
+  V4SHL_RRR_0_OPCODE_X0 = 155,
+  V4SHL_RRR_0_OPCODE_X1 = 102,
+  V4SHRS_RRR_0_OPCODE_X0 = 156,
+  V4SHRS_RRR_0_OPCODE_X1 = 103,
+  V4SHRU_RRR_0_OPCODE_X0 = 157,
+  V4SHRU_RRR_0_OPCODE_X1 = 104,
+  V4SUBSC_RRR_0_OPCODE_X0 = 158,
+  V4SUBSC_RRR_0_OPCODE_X1 = 105,
+  V4SUB_RRR_0_OPCODE_X0 = 159,
+  V4SUB_RRR_0_OPCODE_X1 = 106,
+  WH64_UNARY_OPCODE_X1 = 38,
+  XORI_IMM8_OPCODE_X0 = 20,
+  XORI_IMM8_OPCODE_X1 = 45,
+  XOR_RRR_0_OPCODE_X0 = 160,
+  XOR_RRR_0_OPCODE_X1 = 107,
+  XOR_RRR_5_OPCODE_Y0 = 3,
+  XOR_RRR_5_OPCODE_Y1 = 3
+};
+
+
+#endif /* __ASSEMBLER__ */
+
+#endif /* __ARCH_OPCODE_H__ */
diff --git a/src/target-tilegx/simd_helper.c b/src/target-tilegx/simd_helper.c
new file mode 100644
index 0000000..6d7bb5c
--- /dev/null
+++ b/src/target-tilegx/simd_helper.c
@@ -0,0 +1,165 @@
+/*
+ * QEMU TILE-Gx helpers
+ *
+ *  Copyright (c) 2015 Chen Gang
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see
+ * <http://www.gnu.org/licenses/lgpl-2.1.html>
+ */
+
+#include "cpu.h"
+#include "qemu-common.h"
+#include "exec/helper-proto.h"
+
+
+/* Broadcast a value to all elements of a vector.  */
+#define V1(X)      (((X) & 0xff) * 0x0101010101010101ull)
+#define V2(X)      (((X) & 0xffff) * 0x0001000100010001ull)
+
+
+uint64_t helper_v1multu(uint64_t a, uint64_t b)
+{
+    uint64_t r = 0;
+    int i;
+
+    for (i = 0; i < 64; i += 8) {
+        unsigned ae = extract64(a, i, 8);
+        unsigned be = extract64(b, i, 8);
+        r = deposit64(r, i, 8, ae * be);
+    }
+    return r;
+}
+
+uint64_t helper_v2mults(uint64_t a, uint64_t b)
+{
+    uint64_t r = 0;
+    int i;
+
+    /* While the instruction talks about signed inputs, with a
+       truncated result the sign of the inputs doesn't matter.  */
+    for (i = 0; i < 64; i += 16) {
+        unsigned ae = extract64(a, i, 16);
+        unsigned be = extract64(b, i, 16);
+        r = deposit64(r, i, 16, ae * be);
+    }
+    return r;
+}
+
+uint64_t helper_v1shl(uint64_t a, uint64_t b)
+{
+    uint64_t m;
+
+    b &= 7;
+    m = V1(0xff >> b);
+    return (a & m) << b;
+}
+
+uint64_t helper_v2shl(uint64_t a, uint64_t b)
+{
+    uint64_t m;
+
+    b &= 15;
+    m = V2(0xffff >> b);
+    return (a & m) << b;
+}
+
+uint64_t helper_v1shru(uint64_t a, uint64_t b)
+{
+    uint64_t m;
+
+    b &= 7;
+    m = V1(0xff << b);
+    return (a & m) >> b;
+}
+
+uint64_t helper_v2shru(uint64_t a, uint64_t b)
+{
+    uint64_t m;
+
+    b &= 15;
+    m = V2(0xffff << b);
+    return (a & m) >> b;
+}
+
+uint64_t helper_v1shrs(uint64_t a, uint64_t b)
+{
+    uint64_t r = 0;
+    int i;
+
+    b &= 7;
+    for (i = 0; i < 64; i += 8) {
+        r = deposit64(r, i, 8, sextract64(a, i + b, 8 - b));
+    }
+    return r;
+}
+
+uint64_t helper_v2shrs(uint64_t a, uint64_t b)
+{
+    uint64_t r = 0;
+    int i;
+
+    b &= 15;
+    for (i = 0; i < 64; i += 16) {
+        r = deposit64(r, i, 16, sextract64(a, i + b, 16 - b));
+    }
+    return r;
+}
+
+uint64_t helper_v1int_h(uint64_t a, uint64_t b)
+{
+    uint64_t r = 0;
+    int i;
+
+    for (i = 0; i < 32; i += 8) {
+        r = deposit64(r, 2 * i + 8, 8, extract64(a, i + 32, 8));
+        r = deposit64(r, 2 * i, 8, extract64(b, i + 32, 8));
+    }
+    return r;
+}
+
+uint64_t helper_v1int_l(uint64_t a, uint64_t b)
+{
+    uint64_t r = 0;
+    int i;
+
+    for (i = 0; i < 32; i += 8) {
+        r = deposit64(r, 2 * i + 8, 8, extract64(a, i, 8));
+        r = deposit64(r, 2 * i, 8, extract64(b, i, 8));
+    }
+    return r;
+}
+
+uint64_t helper_v2int_h(uint64_t a, uint64_t b)
+{
+    uint64_t r = 0;
+    int i;
+
+    for (i = 0; i < 32; i += 16) {
+        r = deposit64(r, 2 * i + 16, 16, extract64(a, i + 32, 16));
+        r = deposit64(r, 2 * i, 16, extract64(b, i + 32, 16));
+    }
+    return r;
+}
+
+uint64_t helper_v2int_l(uint64_t a, uint64_t b)
+{
+    uint64_t r = 0;
+    int i;
+
+    for (i = 0; i < 32; i += 16) {
+        r = deposit64(r, 2 * i + 16, 16, extract64(a, i, 16));
+        r = deposit64(r, 2 * i, 16, extract64(b, i, 16));
+    }
+    return r;
+}
diff --git a/src/target-tilegx/spr_def_64.h b/src/target-tilegx/spr_def_64.h
new file mode 100644
index 0000000..67a6c17
--- /dev/null
+++ b/src/target-tilegx/spr_def_64.h
@@ -0,0 +1,216 @@
+/*
+ * Copyright 2011 Tilera Corporation. All Rights Reserved.
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful, but
+ *   WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ *   NON INFRINGEMENT.  See the GNU General Public License for
+ *   more details.
+ */
+
+#ifndef __DOXYGEN__
+
+#ifndef __ARCH_SPR_DEF_64_H__
+#define __ARCH_SPR_DEF_64_H__
+
+#define SPR_AUX_PERF_COUNT_0 0x2105
+#define SPR_AUX_PERF_COUNT_1 0x2106
+#define SPR_AUX_PERF_COUNT_CTL 0x2107
+#define SPR_AUX_PERF_COUNT_STS 0x2108
+#define SPR_CMPEXCH_VALUE 0x2780
+#define SPR_CYCLE 0x2781
+#define SPR_DONE 0x2705
+#define SPR_DSTREAM_PF 0x2706
+#define SPR_EVENT_BEGIN 0x2782
+#define SPR_EVENT_END 0x2783
+#define SPR_EX_CONTEXT_0_0 0x2580
+#define SPR_EX_CONTEXT_0_1 0x2581
+#define SPR_EX_CONTEXT_0_1__PL_SHIFT 0
+#define SPR_EX_CONTEXT_0_1__PL_RMASK 0x3
+#define SPR_EX_CONTEXT_0_1__PL_MASK  0x3
+#define SPR_EX_CONTEXT_0_1__ICS_SHIFT 2
+#define SPR_EX_CONTEXT_0_1__ICS_RMASK 0x1
+#define SPR_EX_CONTEXT_0_1__ICS_MASK  0x4
+#define SPR_EX_CONTEXT_1_0 0x2480
+#define SPR_EX_CONTEXT_1_1 0x2481
+#define SPR_EX_CONTEXT_1_1__PL_SHIFT 0
+#define SPR_EX_CONTEXT_1_1__PL_RMASK 0x3
+#define SPR_EX_CONTEXT_1_1__PL_MASK  0x3
+#define SPR_EX_CONTEXT_1_1__ICS_SHIFT 2
+#define SPR_EX_CONTEXT_1_1__ICS_RMASK 0x1
+#define SPR_EX_CONTEXT_1_1__ICS_MASK  0x4
+#define SPR_EX_CONTEXT_2_0 0x2380
+#define SPR_EX_CONTEXT_2_1 0x2381
+#define SPR_EX_CONTEXT_2_1__PL_SHIFT 0
+#define SPR_EX_CONTEXT_2_1__PL_RMASK 0x3
+#define SPR_EX_CONTEXT_2_1__PL_MASK  0x3
+#define SPR_EX_CONTEXT_2_1__ICS_SHIFT 2
+#define SPR_EX_CONTEXT_2_1__ICS_RMASK 0x1
+#define SPR_EX_CONTEXT_2_1__ICS_MASK  0x4
+#define SPR_FAIL 0x2707
+#define SPR_IDN_AVAIL_EN 0x1a05
+#define SPR_IDN_DATA_AVAIL 0x0a80
+#define SPR_IDN_DEADLOCK_TIMEOUT 0x1806
+#define SPR_IDN_DEMUX_COUNT_0 0x0a05
+#define SPR_IDN_DEMUX_COUNT_1 0x0a06
+#define SPR_IDN_DIRECTION_PROTECT 0x1405
+#define SPR_IDN_PENDING 0x0a08
+#define SPR_ILL_TRANS_REASON__I_STREAM_VA_RMASK 0x1
+#define SPR_INTCTRL_0_STATUS 0x2505
+#define SPR_INTCTRL_1_STATUS 0x2405
+#define SPR_INTCTRL_2_STATUS 0x2305
+#define SPR_INTERRUPT_CRITICAL_SECTION 0x2708
+#define SPR_INTERRUPT_MASK_0 0x2506
+#define SPR_INTERRUPT_MASK_1 0x2406
+#define SPR_INTERRUPT_MASK_2 0x2306
+#define SPR_INTERRUPT_MASK_RESET_0 0x2507
+#define SPR_INTERRUPT_MASK_RESET_1 0x2407
+#define SPR_INTERRUPT_MASK_RESET_2 0x2307
+#define SPR_INTERRUPT_MASK_SET_0 0x2508
+#define SPR_INTERRUPT_MASK_SET_1 0x2408
+#define SPR_INTERRUPT_MASK_SET_2 0x2308
+#define SPR_INTERRUPT_VECTOR_BASE_0 0x2509
+#define SPR_INTERRUPT_VECTOR_BASE_1 0x2409
+#define SPR_INTERRUPT_VECTOR_BASE_2 0x2309
+#define SPR_INTERRUPT_VECTOR_BASE_3 0x2209
+#define SPR_IPI_EVENT_0 0x1f05
+#define SPR_IPI_EVENT_1 0x1e05
+#define SPR_IPI_EVENT_2 0x1d05
+#define SPR_IPI_EVENT_RESET_0 0x1f06
+#define SPR_IPI_EVENT_RESET_1 0x1e06
+#define SPR_IPI_EVENT_RESET_2 0x1d06
+#define SPR_IPI_EVENT_SET_0 0x1f07
+#define SPR_IPI_EVENT_SET_1 0x1e07
+#define SPR_IPI_EVENT_SET_2 0x1d07
+#define SPR_IPI_MASK_0 0x1f08
+#define SPR_IPI_MASK_1 0x1e08
+#define SPR_IPI_MASK_2 0x1d08
+#define SPR_IPI_MASK_RESET_0 0x1f09
+#define SPR_IPI_MASK_RESET_1 0x1e09
+#define SPR_IPI_MASK_RESET_2 0x1d09
+#define SPR_IPI_MASK_SET_0 0x1f0a
+#define SPR_IPI_MASK_SET_1 0x1e0a
+#define SPR_IPI_MASK_SET_2 0x1d0a
+#define SPR_MPL_AUX_PERF_COUNT_SET_0 0x2100
+#define SPR_MPL_AUX_PERF_COUNT_SET_1 0x2101
+#define SPR_MPL_AUX_PERF_COUNT_SET_2 0x2102
+#define SPR_MPL_AUX_TILE_TIMER_SET_0 0x1700
+#define SPR_MPL_AUX_TILE_TIMER_SET_1 0x1701
+#define SPR_MPL_AUX_TILE_TIMER_SET_2 0x1702
+#define SPR_MPL_IDN_ACCESS_SET_0 0x0a00
+#define SPR_MPL_IDN_ACCESS_SET_1 0x0a01
+#define SPR_MPL_IDN_ACCESS_SET_2 0x0a02
+#define SPR_MPL_IDN_AVAIL_SET_0 0x1a00
+#define SPR_MPL_IDN_AVAIL_SET_1 0x1a01
+#define SPR_MPL_IDN_AVAIL_SET_2 0x1a02
+#define SPR_MPL_IDN_COMPLETE_SET_0 0x0500
+#define SPR_MPL_IDN_COMPLETE_SET_1 0x0501
+#define SPR_MPL_IDN_COMPLETE_SET_2 0x0502
+#define SPR_MPL_IDN_FIREWALL_SET_0 0x1400
+#define SPR_MPL_IDN_FIREWALL_SET_1 0x1401
+#define SPR_MPL_IDN_FIREWALL_SET_2 0x1402
+#define SPR_MPL_IDN_TIMER_SET_0 0x1800
+#define SPR_MPL_IDN_TIMER_SET_1 0x1801
+#define SPR_MPL_IDN_TIMER_SET_2 0x1802
+#define SPR_MPL_INTCTRL_0_SET_0 0x2500
+#define SPR_MPL_INTCTRL_0_SET_1 0x2501
+#define SPR_MPL_INTCTRL_0_SET_2 0x2502
+#define SPR_MPL_INTCTRL_1_SET_0 0x2400
+#define SPR_MPL_INTCTRL_1_SET_1 0x2401
+#define SPR_MPL_INTCTRL_1_SET_2 0x2402
+#define SPR_MPL_INTCTRL_2_SET_0 0x2300
+#define SPR_MPL_INTCTRL_2_SET_1 0x2301
+#define SPR_MPL_INTCTRL_2_SET_2 0x2302
+#define SPR_MPL_IPI_0 0x1f04
+#define SPR_MPL_IPI_0_SET_0 0x1f00
+#define SPR_MPL_IPI_0_SET_1 0x1f01
+#define SPR_MPL_IPI_0_SET_2 0x1f02
+#define SPR_MPL_IPI_1 0x1e04
+#define SPR_MPL_IPI_1_SET_0 0x1e00
+#define SPR_MPL_IPI_1_SET_1 0x1e01
+#define SPR_MPL_IPI_1_SET_2 0x1e02
+#define SPR_MPL_IPI_2 0x1d04
+#define SPR_MPL_IPI_2_SET_0 0x1d00
+#define SPR_MPL_IPI_2_SET_1 0x1d01
+#define SPR_MPL_IPI_2_SET_2 0x1d02
+#define SPR_MPL_PERF_COUNT_SET_0 0x2000
+#define SPR_MPL_PERF_COUNT_SET_1 0x2001
+#define SPR_MPL_PERF_COUNT_SET_2 0x2002
+#define SPR_MPL_UDN_ACCESS_SET_0 0x0b00
+#define SPR_MPL_UDN_ACCESS_SET_1 0x0b01
+#define SPR_MPL_UDN_ACCESS_SET_2 0x0b02
+#define SPR_MPL_UDN_AVAIL_SET_0 0x1b00
+#define SPR_MPL_UDN_AVAIL_SET_1 0x1b01
+#define SPR_MPL_UDN_AVAIL_SET_2 0x1b02
+#define SPR_MPL_UDN_COMPLETE_SET_0 0x0600
+#define SPR_MPL_UDN_COMPLETE_SET_1 0x0601
+#define SPR_MPL_UDN_COMPLETE_SET_2 0x0602
+#define SPR_MPL_UDN_FIREWALL_SET_0 0x1500
+#define SPR_MPL_UDN_FIREWALL_SET_1 0x1501
+#define SPR_MPL_UDN_FIREWALL_SET_2 0x1502
+#define SPR_MPL_UDN_TIMER_SET_0 0x1900
+#define SPR_MPL_UDN_TIMER_SET_1 0x1901
+#define SPR_MPL_UDN_TIMER_SET_2 0x1902
+#define SPR_MPL_WORLD_ACCESS_SET_0 0x2700
+#define SPR_MPL_WORLD_ACCESS_SET_1 0x2701
+#define SPR_MPL_WORLD_ACCESS_SET_2 0x2702
+#define SPR_PASS 0x2709
+#define SPR_PERF_COUNT_0 0x2005
+#define SPR_PERF_COUNT_1 0x2006
+#define SPR_PERF_COUNT_CTL 0x2007
+#define SPR_PERF_COUNT_DN_CTL 0x2008
+#define SPR_PERF_COUNT_STS 0x2009
+#define SPR_PROC_STATUS 0x2784
+#define SPR_SIM_CONTROL 0x2785
+#define SPR_SINGLE_STEP_CONTROL_0 0x0405
+#define SPR_SINGLE_STEP_CONTROL_0__CANCELED_MASK  0x1
+#define SPR_SINGLE_STEP_CONTROL_0__INHIBIT_MASK  0x2
+#define SPR_SINGLE_STEP_CONTROL_1 0x0305
+#define SPR_SINGLE_STEP_CONTROL_1__CANCELED_MASK  0x1
+#define SPR_SINGLE_STEP_CONTROL_1__INHIBIT_MASK  0x2
+#define SPR_SINGLE_STEP_CONTROL_2 0x0205
+#define SPR_SINGLE_STEP_CONTROL_2__CANCELED_MASK  0x1
+#define SPR_SINGLE_STEP_CONTROL_2__INHIBIT_MASK  0x2
+#define SPR_SINGLE_STEP_EN_0_0 0x250a
+#define SPR_SINGLE_STEP_EN_0_1 0x240a
+#define SPR_SINGLE_STEP_EN_0_2 0x230a
+#define SPR_SINGLE_STEP_EN_1_0 0x250b
+#define SPR_SINGLE_STEP_EN_1_1 0x240b
+#define SPR_SINGLE_STEP_EN_1_2 0x230b
+#define SPR_SINGLE_STEP_EN_2_0 0x250c
+#define SPR_SINGLE_STEP_EN_2_1 0x240c
+#define SPR_SINGLE_STEP_EN_2_2 0x230c
+#define SPR_SYSTEM_SAVE_0_0 0x2582
+#define SPR_SYSTEM_SAVE_0_1 0x2583
+#define SPR_SYSTEM_SAVE_0_2 0x2584
+#define SPR_SYSTEM_SAVE_0_3 0x2585
+#define SPR_SYSTEM_SAVE_1_0 0x2482
+#define SPR_SYSTEM_SAVE_1_1 0x2483
+#define SPR_SYSTEM_SAVE_1_2 0x2484
+#define SPR_SYSTEM_SAVE_1_3 0x2485
+#define SPR_SYSTEM_SAVE_2_0 0x2382
+#define SPR_SYSTEM_SAVE_2_1 0x2383
+#define SPR_SYSTEM_SAVE_2_2 0x2384
+#define SPR_SYSTEM_SAVE_2_3 0x2385
+#define SPR_TILE_COORD 0x270b
+#define SPR_TILE_RTF_HWM 0x270c
+#define SPR_TILE_TIMER_CONTROL 0x1605
+#define SPR_UDN_AVAIL_EN 0x1b05
+#define SPR_UDN_DATA_AVAIL 0x0b80
+#define SPR_UDN_DEADLOCK_TIMEOUT 0x1906
+#define SPR_UDN_DEMUX_COUNT_0 0x0b05
+#define SPR_UDN_DEMUX_COUNT_1 0x0b06
+#define SPR_UDN_DEMUX_COUNT_2 0x0b07
+#define SPR_UDN_DEMUX_COUNT_3 0x0b08
+#define SPR_UDN_DIRECTION_PROTECT 0x1505
+#define SPR_UDN_PENDING 0x0b0a
+#define SPR_WATCH_MASK 0x200a
+#define SPR_WATCH_VAL 0x200b
+
+#endif /* !defined(__ARCH_SPR_DEF_64_H__) */
+
+#endif /* !defined(__DOXYGEN__) */
diff --git a/src/target-tilegx/translate.c b/src/target-tilegx/translate.c
new file mode 100644
index 0000000..354f25a
--- /dev/null
+++ b/src/target-tilegx/translate.c
@@ -0,0 +1,2449 @@
+/*
+ * QEMU TILE-Gx CPU
+ *
+ *  Copyright (c) 2015 Chen Gang
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see
+ * <http://www.gnu.org/licenses/lgpl-2.1.html>
+ */
+
+#include "cpu.h"
+#include "qemu/log.h"
+#include "disas/disas.h"
+#include "tcg-op.h"
+#include "exec/cpu_ldst.h"
+#include "linux-user/syscall_defs.h"
+
+#include "opcode_tilegx.h"
+#include "spr_def_64.h"
+
+#define FMT64X                          "%016" PRIx64
+
+static TCGv_ptr cpu_env;
+static TCGv cpu_pc;
+static TCGv cpu_regs[TILEGX_R_COUNT];
+
+static const char * const reg_names[64] = {
+     "r0",  "r1",  "r2",  "r3",  "r4",  "r5",  "r6",  "r7",
+     "r8",  "r9", "r10", "r11", "r12", "r13", "r14", "r15",
+    "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23",
+    "r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31",
+    "r32", "r33", "r34", "r35", "r36", "r37", "r38", "r39",
+    "r40", "r41", "r42", "r43", "r44", "r45", "r46", "r47",
+    "r48", "r49", "r50", "r51",  "bp",  "tp",  "sp",  "lr",
+    "sn", "idn0", "idn1", "udn0", "udn1", "udn2", "udn2", "zero"
+};
+
+/* Modified registers are cached in temporaries until the end of the bundle. */
+typedef struct {
+    unsigned reg;
+    TCGv val;
+} DisasContextTemp;
+
+#define MAX_WRITEBACK 4
+
+/* This is the state at translation time.  */
+typedef struct {
+    uint64_t pc;		/* Current pc */
+
+    TCGv zero;                  /* For zero register */
+
+    DisasContextTemp wb[MAX_WRITEBACK];
+    int num_wb;
+    int mmuidx;
+    bool exit_tb;
+    TileExcp atomic_excp;
+
+    struct {
+        TCGCond cond;    /* branch condition */
+        TCGv dest;       /* branch destination */
+        TCGv val1;       /* value to be compared against zero, for cond */
+    } jmp;               /* Jump object, only once in each TB block */
+} DisasContext;
+
+#include "exec/gen-icount.h"
+
+/* Differentiate the various pipe encodings.  */
+#define TY_X0  0
+#define TY_X1  1
+#define TY_Y0  2
+#define TY_Y1  3
+
+/* Remerge the base opcode and extension fields for switching.
+   The X opcode fields are 3 bits; Y0/Y1 opcode fields are 4 bits;
+   Y2 opcode field is 2 bits.  */
+#define OE(OP, EXT, XY) (TY_##XY + OP * 4 + EXT * 64)
+
+/* Similar, but for Y2 only.  */
+#define OEY2(OP, MODE) (OP + MODE * 4)
+
+/* Similar, but make sure opcode names match up.  */
+#define OE_RR_X0(E)    OE(RRR_0_OPCODE_X0, E##_UNARY_OPCODE_X0, X0)
+#define OE_RR_X1(E)    OE(RRR_0_OPCODE_X1, E##_UNARY_OPCODE_X1, X1)
+#define OE_RR_Y0(E)    OE(RRR_1_OPCODE_Y0, E##_UNARY_OPCODE_Y0, Y0)
+#define OE_RR_Y1(E)    OE(RRR_1_OPCODE_Y1, E##_UNARY_OPCODE_Y1, Y1)
+#define OE_RRR(E,N,XY) OE(RRR_##N##_OPCODE_##XY, E##_RRR_##N##_OPCODE_##XY, XY)
+#define OE_IM(E,XY)    OE(IMM8_OPCODE_##XY, E##_IMM8_OPCODE_##XY, XY)
+#define OE_SH(E,XY)    OE(SHIFT_OPCODE_##XY, E##_SHIFT_OPCODE_##XY, XY)
+
+#define V1_IMM(X)      (((X) & 0xff) * 0x0101010101010101ull)
+#define V2_IMM(X)      (((X) & 0xffff) * 0x0001000100010001ull)
+
+
+static void gen_exception(DisasContext *dc, TileExcp num)
+{
+    TCGv_i32 tmp;
+
+    tcg_gen_movi_tl(cpu_pc, dc->pc + TILEGX_BUNDLE_SIZE_IN_BYTES);
+
+    tmp = tcg_const_i32(num);
+    gen_helper_exception(cpu_env, tmp);
+    tcg_temp_free_i32(tmp);
+    dc->exit_tb = true;
+}
+
+static bool check_gr(DisasContext *dc, uint8_t reg)
+{
+    if (likely(reg < TILEGX_R_COUNT)) {
+        return true;
+    }
+
+    switch (reg) {
+    case TILEGX_R_SN:
+    case TILEGX_R_ZERO:
+        break;
+    case TILEGX_R_IDN0:
+    case TILEGX_R_IDN1:
+        gen_exception(dc, TILEGX_EXCP_REG_IDN_ACCESS);
+        break;
+    case TILEGX_R_UDN0:
+    case TILEGX_R_UDN1:
+    case TILEGX_R_UDN2:
+    case TILEGX_R_UDN3:
+        gen_exception(dc, TILEGX_EXCP_REG_UDN_ACCESS);
+        break;
+    default:
+        g_assert_not_reached();
+    }
+    return false;
+}
+
+static TCGv load_zero(DisasContext *dc)
+{
+    if (TCGV_IS_UNUSED_I64(dc->zero)) {
+        dc->zero = tcg_const_i64(0);
+    }
+    return dc->zero;
+}
+
+static TCGv load_gr(DisasContext *dc, unsigned reg)
+{
+    if (check_gr(dc, reg)) {
+        return cpu_regs[reg];
+    }
+    return load_zero(dc);
+}
+
+static TCGv dest_gr(DisasContext *dc, unsigned reg)
+{
+    int n;
+
+    /* Skip the result, mark the exception if necessary, and continue */
+    check_gr(dc, reg);
+
+    n = dc->num_wb++;
+    dc->wb[n].reg = reg;
+    return dc->wb[n].val = tcg_temp_new_i64();
+}
+
+static void gen_saturate_op(TCGv tdest, TCGv tsrca, TCGv tsrcb,
+                            void (*operate)(TCGv, TCGv, TCGv))
+{
+    TCGv t0 = tcg_temp_new();
+
+    tcg_gen_ext32s_tl(tdest, tsrca);
+    tcg_gen_ext32s_tl(t0, tsrcb);
+    operate(tdest, tdest, t0);
+
+    tcg_gen_movi_tl(t0, 0x7fffffff);
+    tcg_gen_movcond_tl(TCG_COND_GT, tdest, tdest, t0, t0, tdest);
+    tcg_gen_movi_tl(t0, -0x80000000LL);
+    tcg_gen_movcond_tl(TCG_COND_LT, tdest, tdest, t0, t0, tdest);
+
+    tcg_temp_free(t0);
+}
+
+static void gen_atomic_excp(DisasContext *dc, unsigned dest, TCGv tdest,
+                            TCGv tsrca, TCGv tsrcb, TileExcp excp)
+{
+#ifdef CONFIG_USER_ONLY
+    TCGv_i32 t;
+
+    tcg_gen_st_tl(tsrca, cpu_env, offsetof(CPUTLGState, atomic_srca));
+    tcg_gen_st_tl(tsrcb, cpu_env, offsetof(CPUTLGState, atomic_srcb));
+    t = tcg_const_i32(dest);
+    tcg_gen_st_i32(t, cpu_env, offsetof(CPUTLGState, atomic_dstr));
+    tcg_temp_free_i32(t);
+
+    /* We're going to write the real result in the exception.  But in
+       the meantime we've already created a writeback register, and
+       we don't want that to remain uninitialized.  */
+    tcg_gen_movi_tl(tdest, 0);
+
+    /* Note that we need to delay issuing the exception that implements
+       the atomic operation until after writing back the results of the
+       instruction occupying the X0 pipe.  */
+    dc->atomic_excp = excp;
+#else
+    gen_exception(dc, TILEGX_EXCP_OPCODE_UNIMPLEMENTED);
+#endif
+}
+
+/* Shift the 128-bit value TSRCA:TSRCD right by the number of bytes
+   specified by the bottom 3 bits of TSRCB, and set TDEST to the
+   low 64 bits of the resulting value.  */
+static void gen_dblalign(TCGv tdest, TCGv tsrcd, TCGv tsrca, TCGv tsrcb)
+{
+    TCGv t0 = tcg_temp_new();
+
+    tcg_gen_andi_tl(t0, tsrcb, 7);
+    tcg_gen_shli_tl(t0, t0, 3);
+    tcg_gen_shr_tl(tdest, tsrcd, t0);
+
+    /* We want to do "t0 = tsrca << (64 - t0)".  Two's complement
+       arithmetic on a 6-bit field tells us that 64 - t0 is equal
+       to (t0 ^ 63) + 1.  So we can do the shift in two parts,
+       neither of which will be an invalid shift by 64.  */
+    tcg_gen_xori_tl(t0, t0, 63);
+    tcg_gen_shl_tl(t0, tsrca, t0);
+    tcg_gen_shli_tl(t0, t0, 1);
+    tcg_gen_or_tl(tdest, tdest, t0);
+
+    tcg_temp_free(t0);
+}
+
+/* Similarly, except that the 128-bit value is TSRCA:TSRCB, and the
+   right shift is an immediate.  */
+static void gen_dblaligni(TCGv tdest, TCGv tsrca, TCGv tsrcb, int shr)
+{
+    TCGv t0 = tcg_temp_new();
+
+    tcg_gen_shri_tl(t0, tsrcb, shr);
+    tcg_gen_shli_tl(tdest, tsrca, 64 - shr);
+    tcg_gen_or_tl(tdest, tdest, t0);
+
+    tcg_temp_free(t0);
+}
+
+typedef enum {
+    LU, LS, HU, HS
+} MulHalf;
+
+static void gen_ext_half(TCGv d, TCGv s, MulHalf h)
+{
+    switch (h) {
+    case LU:
+        tcg_gen_ext32u_tl(d, s);
+        break;
+    case LS:
+        tcg_gen_ext32s_tl(d, s);
+        break;
+    case HU:
+        tcg_gen_shri_tl(d, s, 32);
+        break;
+    case HS:
+        tcg_gen_sari_tl(d, s, 32);
+        break;
+    }
+}
+
+static void gen_mul_half(TCGv tdest, TCGv tsrca, TCGv tsrcb,
+                         MulHalf ha, MulHalf hb)
+{
+    TCGv t = tcg_temp_new();
+    gen_ext_half(t, tsrca, ha);
+    gen_ext_half(tdest, tsrcb, hb);
+    tcg_gen_mul_tl(tdest, tdest, t);
+    tcg_temp_free(t);
+}
+
+static void gen_cmul2(TCGv tdest, TCGv tsrca, TCGv tsrcb, int sh, int rd)
+{
+    TCGv_i32 tsh = tcg_const_i32(sh);
+    TCGv_i32 trd = tcg_const_i32(rd);
+    gen_helper_cmul2(tdest, tsrca, tsrcb, tsh, trd);
+    tcg_temp_free_i32(tsh);
+    tcg_temp_free_i32(trd);
+}
+
+static TileExcp gen_st_opcode(DisasContext *dc, unsigned dest, unsigned srca,
+                              unsigned srcb, TCGMemOp memop, const char *name)
+{
+    if (dest) {
+        return TILEGX_EXCP_OPCODE_UNKNOWN;
+    }
+
+    tcg_gen_qemu_st_tl(load_gr(dc, srcb), load_gr(dc, srca),
+		       dc->mmuidx, memop);
+
+    qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s", name,
+                  reg_names[srca], reg_names[srcb]);
+    return TILEGX_EXCP_NONE;
+}
+
+static TileExcp gen_st_add_opcode(DisasContext *dc, unsigned srca, unsigned srcb,
+                                  int imm, TCGMemOp memop, const char *name)
+{
+    TCGv tsrca = load_gr(dc, srca);
+    TCGv tsrcb = load_gr(dc, srcb);
+
+    tcg_gen_qemu_st_tl(tsrcb, tsrca, dc->mmuidx, memop);
+    tcg_gen_addi_tl(dest_gr(dc, srca), tsrca, imm);
+
+    qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s, %d", name,
+                  reg_names[srca], reg_names[srcb], imm);
+    return TILEGX_EXCP_NONE;
+}
+
+/* Equality comparison with zero can be done quickly and efficiently.  */
+static void gen_v1cmpeq0(TCGv v)
+{
+    TCGv m = tcg_const_tl(V1_IMM(0x7f));
+    TCGv c = tcg_temp_new();
+
+    /* ~(((v & m) + m) | m | v).  Sets the msb for each byte == 0.  */
+    tcg_gen_and_tl(c, v, m);
+    tcg_gen_add_tl(c, c, m);
+    tcg_gen_or_tl(c, c, m);
+    tcg_gen_nor_tl(c, c, v);
+    tcg_temp_free(m);
+
+    /* Shift the msb down to form the lsb boolean result.  */
+    tcg_gen_shri_tl(v, c, 7);
+    tcg_temp_free(c);
+}
+
+static void gen_v1cmpne0(TCGv v)
+{
+    TCGv m = tcg_const_tl(V1_IMM(0x7f));
+    TCGv c = tcg_temp_new();
+
+    /* (((v & m) + m) | v) & ~m.  Sets the msb for each byte != 0.  */
+    tcg_gen_and_tl(c, v, m);
+    tcg_gen_add_tl(c, c, m);
+    tcg_gen_or_tl(c, c, v);
+    tcg_gen_andc_tl(c, c, m);
+    tcg_temp_free(m);
+
+    /* Shift the msb down to form the lsb boolean result.  */
+    tcg_gen_shri_tl(v, c, 7);
+    tcg_temp_free(c);
+}
+
+/* Vector addition can be performed via arithmetic plus masking.  It is
+   efficient this way only for 4 or more elements.  */
+static void gen_v12add(TCGv tdest, TCGv tsrca, TCGv tsrcb, uint64_t sign)
+{
+    TCGv tmask = tcg_const_tl(~sign);
+    TCGv t0 = tcg_temp_new();
+    TCGv t1 = tcg_temp_new();
+
+    /* ((a & ~sign) + (b & ~sign)) ^ ((a ^ b) & sign).  */
+    tcg_gen_and_tl(t0, tsrca, tmask);
+    tcg_gen_and_tl(t1, tsrcb, tmask);
+    tcg_gen_add_tl(tdest, t0, t1);
+    tcg_gen_xor_tl(t0, tsrca, tsrcb);
+    tcg_gen_andc_tl(t0, t0, tmask);
+    tcg_gen_xor_tl(tdest, tdest, t0);
+
+    tcg_temp_free(t1);
+    tcg_temp_free(t0);
+    tcg_temp_free(tmask);
+}
+
+/* Similarly for vector subtraction.  */
+static void gen_v12sub(TCGv tdest, TCGv tsrca, TCGv tsrcb, uint64_t sign)
+{
+    TCGv tsign = tcg_const_tl(sign);
+    TCGv t0 = tcg_temp_new();
+    TCGv t1 = tcg_temp_new();
+
+    /* ((a | sign) - (b & ~sign)) ^ ((a ^ ~b) & sign).  */
+    tcg_gen_or_tl(t0, tsrca, tsign);
+    tcg_gen_andc_tl(t1, tsrcb, tsign);
+    tcg_gen_sub_tl(tdest, t0, t1);
+    tcg_gen_eqv_tl(t0, tsrca, tsrcb);
+    tcg_gen_and_tl(t0, t0, tsign);
+    tcg_gen_xor_tl(tdest, tdest, t0);
+
+    tcg_temp_free(t1);
+    tcg_temp_free(t0);
+    tcg_temp_free(tsign);
+}
+
+static void gen_v4sh(TCGv d64, TCGv a64, TCGv b64,
+                     void (*generate)(TCGv_i32, TCGv_i32, TCGv_i32))
+{
+    TCGv_i32 al = tcg_temp_new_i32();
+    TCGv_i32 ah = tcg_temp_new_i32();
+    TCGv_i32 bl = tcg_temp_new_i32();
+
+    tcg_gen_extr_i64_i32(al, ah, a64);
+    tcg_gen_extrl_i64_i32(bl, b64);
+    tcg_gen_andi_i32(bl, bl, 31);
+    generate(al, al, bl);
+    generate(ah, ah, bl);
+    tcg_gen_concat_i32_i64(d64, al, ah);
+
+    tcg_temp_free_i32(al);
+    tcg_temp_free_i32(ah);
+    tcg_temp_free_i32(bl);
+}
+
+static void gen_v4op(TCGv d64, TCGv a64, TCGv b64,
+                     void (*generate)(TCGv_i32, TCGv_i32, TCGv_i32))
+{
+    TCGv_i32 al = tcg_temp_new_i32();
+    TCGv_i32 ah = tcg_temp_new_i32();
+    TCGv_i32 bl = tcg_temp_new_i32();
+    TCGv_i32 bh = tcg_temp_new_i32();
+
+    tcg_gen_extr_i64_i32(al, ah, a64);
+    tcg_gen_extr_i64_i32(bl, bh, b64);
+    generate(al, al, bl);
+    generate(ah, ah, bh);
+    tcg_gen_concat_i32_i64(d64, al, ah);
+
+    tcg_temp_free_i32(al);
+    tcg_temp_free_i32(ah);
+    tcg_temp_free_i32(bl);
+    tcg_temp_free_i32(bh);
+}
+
+static TileExcp gen_signal(DisasContext *dc, int signo, int sigcode,
+                           const char *mnemonic)
+{
+    TCGv_i32 t0 = tcg_const_i32(signo);
+    TCGv_i32 t1 = tcg_const_i32(sigcode);
+
+    tcg_gen_st_i32(t0, cpu_env, offsetof(CPUTLGState, signo));
+    tcg_gen_st_i32(t1, cpu_env, offsetof(CPUTLGState, sigcode));
+
+    tcg_temp_free_i32(t1);
+    tcg_temp_free_i32(t0);
+
+    qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s", mnemonic);
+    return TILEGX_EXCP_SIGNAL;
+}
+
+static bool parse_from_addli(uint64_t bundle, int *signo, int *sigcode)
+{
+    int imm;
+
+    if ((get_Opcode_X0(bundle) != ADDLI_OPCODE_X0)
+        || (get_Dest_X0(bundle) != TILEGX_R_ZERO)
+        || (get_SrcA_X0(bundle) != TILEGX_R_ZERO)) {
+        return false;
+    }
+
+    imm = get_Imm16_X0(bundle);
+    *signo = imm & 0x3f;
+    *sigcode = (imm >> 6) & 0xf;
+
+    /* ??? The linux kernel validates both signo and the sigcode vs the
+       known max for each signal.  Don't bother here.  */
+    return true;
+}
+
+static TileExcp gen_specill(DisasContext *dc, unsigned dest, unsigned srca,
+                            uint64_t bundle)
+{
+    const char *mnemonic;
+    int signo;
+    int sigcode;
+
+    if (dest == 0x1c && srca == 0x25) {
+        signo = TARGET_SIGTRAP;
+        sigcode = TARGET_TRAP_BRKPT;
+        mnemonic = "bpt";
+    } else if (dest == 0x1d && srca == 0x25
+               && parse_from_addli(bundle, &signo, &sigcode)) {
+        mnemonic = "raise";
+    } else {
+        signo = TARGET_SIGILL;
+        sigcode = TARGET_ILL_ILLOPC;
+        mnemonic = "ill";
+    }
+
+    return gen_signal(dc, signo, sigcode, mnemonic);
+}
+
+static TileExcp gen_rr_opcode(DisasContext *dc, unsigned opext,
+                              unsigned dest, unsigned srca, uint64_t bundle)
+{
+    TCGv tdest, tsrca;
+    const char *mnemonic;
+    TCGMemOp memop;
+    TileExcp ret = TILEGX_EXCP_NONE;
+    bool prefetch_nofault = false;
+
+    /* Eliminate instructions with no output before doing anything else.  */
+    switch (opext) {
+    case OE_RR_Y0(NOP):
+    case OE_RR_Y1(NOP):
+    case OE_RR_X0(NOP):
+    case OE_RR_X1(NOP):
+        mnemonic = "nop";
+        goto done0;
+    case OE_RR_Y0(FNOP):
+    case OE_RR_Y1(FNOP):
+    case OE_RR_X0(FNOP):
+    case OE_RR_X1(FNOP):
+        mnemonic = "fnop";
+        goto done0;
+    case OE_RR_X1(DRAIN):
+        mnemonic = "drain";
+        goto done0;
+    case OE_RR_X1(FLUSHWB):
+        mnemonic = "flushwb";
+        goto done0;
+    case OE_RR_X1(ILL):
+        return gen_specill(dc, dest, srca, bundle);
+    case OE_RR_Y1(ILL):
+        return gen_signal(dc, TARGET_SIGILL, TARGET_ILL_ILLOPC, "ill");
+    case OE_RR_X1(MF):
+        mnemonic = "mf";
+        goto done0;
+    case OE_RR_X1(NAP):
+        /* ??? This should yield, especially in system mode.  */
+        mnemonic = "nap";
+        goto done0;
+    case OE_RR_X1(IRET):
+        gen_helper_ext01_ics(cpu_env);
+        dc->jmp.cond = TCG_COND_ALWAYS;
+        dc->jmp.dest = tcg_temp_new();
+        tcg_gen_ld_tl(dc->jmp.dest, cpu_env,
+                      offsetof(CPUTLGState, spregs[TILEGX_SPR_EX_CONTEXT_0_0]));
+        tcg_gen_andi_tl(dc->jmp.dest, dc->jmp.dest, ~7);
+        mnemonic = "iret";
+        goto done0;
+    case OE_RR_X1(SWINT0):
+    case OE_RR_X1(SWINT2):
+    case OE_RR_X1(SWINT3):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RR_X1(SWINT1):
+        ret = TILEGX_EXCP_SYSCALL;
+        mnemonic = "swint1";
+    done0:
+        if (srca || dest) {
+            return TILEGX_EXCP_OPCODE_UNKNOWN;
+        }
+        qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s", mnemonic);
+        return ret;
+
+    case OE_RR_X1(DTLBPR):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RR_X1(FINV):
+        mnemonic = "finv";
+        goto done1;
+    case OE_RR_X1(FLUSH):
+        mnemonic = "flush";
+        goto done1;
+    case OE_RR_X1(ICOH):
+        mnemonic = "icoh";
+        goto done1;
+    case OE_RR_X1(INV):
+        mnemonic = "inv";
+        goto done1;
+    case OE_RR_X1(WH64):
+        mnemonic = "wh64";
+        goto done1;
+    case OE_RR_X1(JRP):
+    case OE_RR_Y1(JRP):
+        mnemonic = "jrp";
+        goto do_jr;
+    case OE_RR_X1(JR):
+    case OE_RR_Y1(JR):
+        mnemonic = "jr";
+        goto do_jr;
+    case OE_RR_X1(JALRP):
+    case OE_RR_Y1(JALRP):
+        mnemonic = "jalrp";
+        goto do_jalr;
+    case OE_RR_X1(JALR):
+    case OE_RR_Y1(JALR):
+        mnemonic = "jalr";
+    do_jalr:
+        tcg_gen_movi_tl(dest_gr(dc, TILEGX_R_LR),
+                        dc->pc + TILEGX_BUNDLE_SIZE_IN_BYTES);
+    do_jr:
+        dc->jmp.cond = TCG_COND_ALWAYS;
+        dc->jmp.dest = tcg_temp_new();
+        tcg_gen_andi_tl(dc->jmp.dest, load_gr(dc, srca), ~7);
+    done1:
+        if (dest) {
+            return TILEGX_EXCP_OPCODE_UNKNOWN;
+        }
+        qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s", mnemonic, reg_names[srca]);
+        return ret;
+    }
+
+    tdest = dest_gr(dc, dest);
+    tsrca = load_gr(dc, srca);
+
+    switch (opext) {
+    case OE_RR_X0(CNTLZ):
+    case OE_RR_Y0(CNTLZ):
+        gen_helper_cntlz(tdest, tsrca);
+        mnemonic = "cntlz";
+        break;
+    case OE_RR_X0(CNTTZ):
+    case OE_RR_Y0(CNTTZ):
+        gen_helper_cnttz(tdest, tsrca);
+        mnemonic = "cnttz";
+        break;
+    case OE_RR_X0(FSINGLE_PACK1):
+    case OE_RR_Y0(FSINGLE_PACK1):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RR_X1(LD1S):
+        memop = MO_SB;
+        mnemonic = "ld1s"; /* prefetch_l1_fault */
+        goto do_load;
+    case OE_RR_X1(LD1U):
+        memop = MO_UB;
+        mnemonic = "ld1u"; /* prefetch, prefetch_l1 */
+        prefetch_nofault = (dest == TILEGX_R_ZERO);
+        goto do_load;
+    case OE_RR_X1(LD2S):
+        memop = MO_TESW;
+        mnemonic = "ld2s"; /* prefetch_l2_fault */
+        goto do_load;
+    case OE_RR_X1(LD2U):
+        memop = MO_TEUW;
+        mnemonic = "ld2u"; /* prefetch_l2 */
+        prefetch_nofault = (dest == TILEGX_R_ZERO);
+        goto do_load;
+    case OE_RR_X1(LD4S):
+        memop = MO_TESL;
+        mnemonic = "ld4s"; /* prefetch_l3_fault */
+        goto do_load;
+    case OE_RR_X1(LD4U):
+        memop = MO_TEUL;
+        mnemonic = "ld4u"; /* prefetch_l3 */
+        prefetch_nofault = (dest == TILEGX_R_ZERO);
+        goto do_load;
+    case OE_RR_X1(LDNT1S):
+        memop = MO_SB;
+        mnemonic = "ldnt1s";
+        goto do_load;
+    case OE_RR_X1(LDNT1U):
+        memop = MO_UB;
+        mnemonic = "ldnt1u";
+        goto do_load;
+    case OE_RR_X1(LDNT2S):
+        memop = MO_TESW;
+        mnemonic = "ldnt2s";
+        goto do_load;
+    case OE_RR_X1(LDNT2U):
+        memop = MO_TEUW;
+        mnemonic = "ldnt2u";
+        goto do_load;
+    case OE_RR_X1(LDNT4S):
+        memop = MO_TESL;
+        mnemonic = "ldnt4s";
+        goto do_load;
+    case OE_RR_X1(LDNT4U):
+        memop = MO_TEUL;
+        mnemonic = "ldnt4u";
+        goto do_load;
+    case OE_RR_X1(LDNT):
+        memop = MO_TEQ;
+        mnemonic = "ldnt";
+        goto do_load;
+    case OE_RR_X1(LD):
+        memop = MO_TEQ;
+        mnemonic = "ld";
+    do_load:
+        if (!prefetch_nofault) {
+            tcg_gen_qemu_ld_tl(tdest, tsrca, dc->mmuidx, memop);
+        }
+        break;
+    case OE_RR_X1(LDNA):
+        tcg_gen_andi_tl(tdest, tsrca, ~7);
+        tcg_gen_qemu_ld_tl(tdest, tdest, dc->mmuidx, MO_TEQ);
+        mnemonic = "ldna";
+        break;
+    case OE_RR_X1(LNK):
+    case OE_RR_Y1(LNK):
+        if (srca) {
+            return TILEGX_EXCP_OPCODE_UNKNOWN;
+        }
+        tcg_gen_movi_tl(tdest, dc->pc + TILEGX_BUNDLE_SIZE_IN_BYTES);
+        mnemonic = "lnk";
+        break;
+    case OE_RR_X0(PCNT):
+    case OE_RR_Y0(PCNT):
+        gen_helper_pcnt(tdest, tsrca);
+        mnemonic = "pcnt";
+        break;
+    case OE_RR_X0(REVBITS):
+    case OE_RR_Y0(REVBITS):
+        gen_helper_revbits(tdest, tsrca);
+        mnemonic = "revbits";
+        break;
+    case OE_RR_X0(REVBYTES):
+    case OE_RR_Y0(REVBYTES):
+        tcg_gen_bswap64_tl(tdest, tsrca);
+        mnemonic = "revbytes";
+        break;
+    case OE_RR_X0(TBLIDXB0):
+    case OE_RR_Y0(TBLIDXB0):
+        tcg_gen_deposit_tl(tdest, load_gr(dc, dest), tsrca, 2, 8);
+        mnemonic = "tblidxb0";
+        break;
+    case OE_RR_X0(TBLIDXB1):
+    case OE_RR_Y0(TBLIDXB1):
+        tcg_gen_shri_tl(tdest, tsrca, 8);
+        tcg_gen_deposit_tl(tdest, load_gr(dc, dest), tdest, 2, 8);
+        mnemonic = "tblidxb1";
+        break;
+    case OE_RR_X0(TBLIDXB2):
+    case OE_RR_Y0(TBLIDXB2):
+        tcg_gen_shri_tl(tdest, tsrca, 16);
+        tcg_gen_deposit_tl(tdest, load_gr(dc, dest), tdest, 2, 8);
+        mnemonic = "tblidxb2";
+        break;
+    case OE_RR_X0(TBLIDXB3):
+    case OE_RR_Y0(TBLIDXB3):
+        tcg_gen_shri_tl(tdest, tsrca, 24);
+        tcg_gen_deposit_tl(tdest, load_gr(dc, dest), tdest, 2, 8);
+        mnemonic = "tblidxb3";
+        break;
+    default:
+        return TILEGX_EXCP_OPCODE_UNKNOWN;
+    }
+
+    qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s", mnemonic,
+                  reg_names[dest], reg_names[srca]);
+    return ret;
+}
+
+static TileExcp gen_rrr_opcode(DisasContext *dc, unsigned opext,
+                               unsigned dest, unsigned srca, unsigned srcb)
+{
+    TCGv tdest = dest_gr(dc, dest);
+    TCGv tsrca = load_gr(dc, srca);
+    TCGv tsrcb = load_gr(dc, srcb);
+    TCGv t0;
+    const char *mnemonic;
+
+    switch (opext) {
+    case OE_RRR(ADDXSC, 0, X0):
+    case OE_RRR(ADDXSC, 0, X1):
+        gen_saturate_op(tdest, tsrca, tsrcb, tcg_gen_add_tl);
+        mnemonic = "addxsc";
+        break;
+    case OE_RRR(ADDX, 0, X0):
+    case OE_RRR(ADDX, 0, X1):
+    case OE_RRR(ADDX, 0, Y0):
+    case OE_RRR(ADDX, 0, Y1):
+        tcg_gen_add_tl(tdest, tsrca, tsrcb);
+        tcg_gen_ext32s_tl(tdest, tdest);
+        mnemonic = "addx";
+        break;
+    case OE_RRR(ADD, 0, X0):
+    case OE_RRR(ADD, 0, X1):
+    case OE_RRR(ADD, 0, Y0):
+    case OE_RRR(ADD, 0, Y1):
+        tcg_gen_add_tl(tdest, tsrca, tsrcb);
+        mnemonic = "add";
+        break;
+    case OE_RRR(AND, 0, X0):
+    case OE_RRR(AND, 0, X1):
+    case OE_RRR(AND, 5, Y0):
+    case OE_RRR(AND, 5, Y1):
+        tcg_gen_and_tl(tdest, tsrca, tsrcb);
+        mnemonic = "and";
+        break;
+    case OE_RRR(CMOVEQZ, 0, X0):
+    case OE_RRR(CMOVEQZ, 4, Y0):
+        tcg_gen_movcond_tl(TCG_COND_EQ, tdest, tsrca, load_zero(dc),
+                           tsrcb, load_gr(dc, dest));
+        mnemonic = "cmoveqz";
+        break;
+    case OE_RRR(CMOVNEZ, 0, X0):
+    case OE_RRR(CMOVNEZ, 4, Y0):
+        tcg_gen_movcond_tl(TCG_COND_NE, tdest, tsrca, load_zero(dc),
+                           tsrcb, load_gr(dc, dest));
+        mnemonic = "cmovnez";
+        break;
+    case OE_RRR(CMPEQ, 0, X0):
+    case OE_RRR(CMPEQ, 0, X1):
+    case OE_RRR(CMPEQ, 3, Y0):
+    case OE_RRR(CMPEQ, 3, Y1):
+        tcg_gen_setcond_tl(TCG_COND_EQ, tdest, tsrca, tsrcb);
+        mnemonic = "cmpeq";
+        break;
+    case OE_RRR(CMPEXCH4, 0, X1):
+        gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
+                        TILEGX_EXCP_OPCODE_CMPEXCH4);
+        mnemonic = "cmpexch4";
+        break;
+    case OE_RRR(CMPEXCH, 0, X1):
+        gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
+                        TILEGX_EXCP_OPCODE_CMPEXCH);
+        mnemonic = "cmpexch";
+        break;
+    case OE_RRR(CMPLES, 0, X0):
+    case OE_RRR(CMPLES, 0, X1):
+    case OE_RRR(CMPLES, 2, Y0):
+    case OE_RRR(CMPLES, 2, Y1):
+        tcg_gen_setcond_tl(TCG_COND_LE, tdest, tsrca, tsrcb);
+        mnemonic = "cmples";
+        break;
+    case OE_RRR(CMPLEU, 0, X0):
+    case OE_RRR(CMPLEU, 0, X1):
+    case OE_RRR(CMPLEU, 2, Y0):
+    case OE_RRR(CMPLEU, 2, Y1):
+        tcg_gen_setcond_tl(TCG_COND_LEU, tdest, tsrca, tsrcb);
+        mnemonic = "cmpleu";
+        break;
+    case OE_RRR(CMPLTS, 0, X0):
+    case OE_RRR(CMPLTS, 0, X1):
+    case OE_RRR(CMPLTS, 2, Y0):
+    case OE_RRR(CMPLTS, 2, Y1):
+        tcg_gen_setcond_tl(TCG_COND_LT, tdest, tsrca, tsrcb);
+        mnemonic = "cmplts";
+        break;
+    case OE_RRR(CMPLTU, 0, X0):
+    case OE_RRR(CMPLTU, 0, X1):
+    case OE_RRR(CMPLTU, 2, Y0):
+    case OE_RRR(CMPLTU, 2, Y1):
+        tcg_gen_setcond_tl(TCG_COND_LTU, tdest, tsrca, tsrcb);
+        mnemonic = "cmpltu";
+        break;
+    case OE_RRR(CMPNE, 0, X0):
+    case OE_RRR(CMPNE, 0, X1):
+    case OE_RRR(CMPNE, 3, Y0):
+    case OE_RRR(CMPNE, 3, Y1):
+        tcg_gen_setcond_tl(TCG_COND_NE, tdest, tsrca, tsrcb);
+        mnemonic = "cmpne";
+        break;
+    case OE_RRR(CMULAF, 0, X0):
+        gen_helper_cmulaf(tdest, load_gr(dc, dest), tsrca, tsrcb);
+        mnemonic = "cmulaf";
+        break;
+    case OE_RRR(CMULA, 0, X0):
+        gen_helper_cmula(tdest, load_gr(dc, dest), tsrca, tsrcb);
+        mnemonic = "cmula";
+        break;
+    case OE_RRR(CMULFR, 0, X0):
+        gen_cmul2(tdest, tsrca, tsrcb, 15, 1 << 14);
+        mnemonic = "cmulfr";
+        break;
+    case OE_RRR(CMULF, 0, X0):
+        gen_cmul2(tdest, tsrca, tsrcb, 15, 0);
+        mnemonic = "cmulf";
+        break;
+    case OE_RRR(CMULHR, 0, X0):
+        gen_cmul2(tdest, tsrca, tsrcb, 16, 1 << 15);
+        mnemonic = "cmulhr";
+        break;
+    case OE_RRR(CMULH, 0, X0):
+        gen_cmul2(tdest, tsrca, tsrcb, 16, 0);
+        mnemonic = "cmulh";
+        break;
+    case OE_RRR(CMUL, 0, X0):
+        gen_helper_cmula(tdest, load_zero(dc), tsrca, tsrcb);
+        mnemonic = "cmul";
+        break;
+    case OE_RRR(CRC32_32, 0, X0):
+        gen_helper_crc32_32(tdest, tsrca, tsrcb);
+        mnemonic = "crc32_32";
+        break;
+    case OE_RRR(CRC32_8, 0, X0):
+        gen_helper_crc32_8(tdest, tsrca, tsrcb);
+        mnemonic = "crc32_8";
+        break;
+    case OE_RRR(DBLALIGN2, 0, X0):
+    case OE_RRR(DBLALIGN2, 0, X1):
+        gen_dblaligni(tdest, tsrca, tsrcb, 16);
+        mnemonic = "dblalign2";
+        break;
+    case OE_RRR(DBLALIGN4, 0, X0):
+    case OE_RRR(DBLALIGN4, 0, X1):
+        gen_dblaligni(tdest, tsrca, tsrcb, 32);
+        mnemonic = "dblalign4";
+        break;
+    case OE_RRR(DBLALIGN6, 0, X0):
+    case OE_RRR(DBLALIGN6, 0, X1):
+        gen_dblaligni(tdest, tsrca, tsrcb, 48);
+        mnemonic = "dblalign6";
+        break;
+    case OE_RRR(DBLALIGN, 0, X0):
+        gen_dblalign(tdest, load_gr(dc, dest), tsrca, tsrcb);
+        mnemonic = "dblalign";
+        break;
+    case OE_RRR(EXCH4, 0, X1):
+        gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
+                        TILEGX_EXCP_OPCODE_EXCH4);
+        mnemonic = "exch4";
+        break;
+    case OE_RRR(EXCH, 0, X1):
+        gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
+                        TILEGX_EXCP_OPCODE_EXCH);
+        mnemonic = "exch";
+        break;
+    case OE_RRR(FDOUBLE_ADDSUB, 0, X0):
+    case OE_RRR(FDOUBLE_ADD_FLAGS, 0, X0):
+    case OE_RRR(FDOUBLE_MUL_FLAGS, 0, X0):
+    case OE_RRR(FDOUBLE_PACK1, 0, X0):
+    case OE_RRR(FDOUBLE_PACK2, 0, X0):
+    case OE_RRR(FDOUBLE_SUB_FLAGS, 0, X0):
+    case OE_RRR(FDOUBLE_UNPACK_MAX, 0, X0):
+    case OE_RRR(FDOUBLE_UNPACK_MIN, 0, X0):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(FETCHADD4, 0, X1):
+        gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
+                        TILEGX_EXCP_OPCODE_FETCHADD4);
+        mnemonic = "fetchadd4";
+        break;
+    case OE_RRR(FETCHADDGEZ4, 0, X1):
+        gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
+                        TILEGX_EXCP_OPCODE_FETCHADDGEZ4);
+        mnemonic = "fetchaddgez4";
+        break;
+    case OE_RRR(FETCHADDGEZ, 0, X1):
+        gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
+                        TILEGX_EXCP_OPCODE_FETCHADDGEZ);
+        mnemonic = "fetchaddgez";
+        break;
+    case OE_RRR(FETCHADD, 0, X1):
+        gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
+                        TILEGX_EXCP_OPCODE_FETCHADD);
+        mnemonic = "fetchadd";
+        break;
+    case OE_RRR(FETCHAND4, 0, X1):
+        gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
+                        TILEGX_EXCP_OPCODE_FETCHAND4);
+        mnemonic = "fetchand4";
+        break;
+    case OE_RRR(FETCHAND, 0, X1):
+        gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
+                        TILEGX_EXCP_OPCODE_FETCHAND);
+        mnemonic = "fetchand";
+        break;
+    case OE_RRR(FETCHOR4, 0, X1):
+        gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
+                        TILEGX_EXCP_OPCODE_FETCHOR4);
+        mnemonic = "fetchor4";
+        break;
+    case OE_RRR(FETCHOR, 0, X1):
+        gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
+                        TILEGX_EXCP_OPCODE_FETCHOR);
+        mnemonic = "fetchor";
+        break;
+    case OE_RRR(FSINGLE_ADD1, 0, X0):
+    case OE_RRR(FSINGLE_ADDSUB2, 0, X0):
+    case OE_RRR(FSINGLE_MUL1, 0, X0):
+    case OE_RRR(FSINGLE_MUL2, 0, X0):
+    case OE_RRR(FSINGLE_PACK2, 0, X0):
+    case OE_RRR(FSINGLE_SUB1, 0, X0):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(MNZ, 0, X0):
+    case OE_RRR(MNZ, 0, X1):
+    case OE_RRR(MNZ, 4, Y0):
+    case OE_RRR(MNZ, 4, Y1):
+        t0 = load_zero(dc);
+        tcg_gen_movcond_tl(TCG_COND_NE, tdest, tsrca, t0, tsrcb, t0);
+        mnemonic = "mnz";
+        break;
+    case OE_RRR(MULAX, 0, X0):
+    case OE_RRR(MULAX, 3, Y0):
+        tcg_gen_mul_tl(tdest, tsrca, tsrcb);
+        tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
+        tcg_gen_ext32s_tl(tdest, tdest);
+        mnemonic = "mulax";
+        break;
+    case OE_RRR(MULA_HS_HS, 0, X0):
+    case OE_RRR(MULA_HS_HS, 9, Y0):
+        gen_mul_half(tdest, tsrca, tsrcb, HS, HS);
+        tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
+        mnemonic = "mula_hs_hs";
+        break;
+    case OE_RRR(MULA_HS_HU, 0, X0):
+        gen_mul_half(tdest, tsrca, tsrcb, HS, HU);
+        tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
+        mnemonic = "mula_hs_hu";
+        break;
+    case OE_RRR(MULA_HS_LS, 0, X0):
+        gen_mul_half(tdest, tsrca, tsrcb, HS, LS);
+        tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
+        mnemonic = "mula_hs_ls";
+        break;
+    case OE_RRR(MULA_HS_LU, 0, X0):
+        gen_mul_half(tdest, tsrca, tsrcb, HS, LU);
+        tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
+        mnemonic = "mula_hs_lu";
+        break;
+    case OE_RRR(MULA_HU_HU, 0, X0):
+    case OE_RRR(MULA_HU_HU, 9, Y0):
+        gen_mul_half(tdest, tsrca, tsrcb, HU, HU);
+        tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
+        mnemonic = "mula_hu_hu";
+        break;
+    case OE_RRR(MULA_HU_LS, 0, X0):
+        gen_mul_half(tdest, tsrca, tsrcb, HU, LS);
+        tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
+        mnemonic = "mula_hu_ls";
+        break;
+    case OE_RRR(MULA_HU_LU, 0, X0):
+        gen_mul_half(tdest, tsrca, tsrcb, HU, LU);
+        tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
+        mnemonic = "mula_hu_lu";
+        break;
+    case OE_RRR(MULA_LS_LS, 0, X0):
+    case OE_RRR(MULA_LS_LS, 9, Y0):
+        gen_mul_half(tdest, tsrca, tsrcb, LS, LS);
+        tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
+        mnemonic = "mula_ls_ls";
+        break;
+    case OE_RRR(MULA_LS_LU, 0, X0):
+        gen_mul_half(tdest, tsrca, tsrcb, LS, LU);
+        tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
+        mnemonic = "mula_ls_lu";
+        break;
+    case OE_RRR(MULA_LU_LU, 0, X0):
+    case OE_RRR(MULA_LU_LU, 9, Y0):
+        gen_mul_half(tdest, tsrca, tsrcb, LU, LU);
+        tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
+        mnemonic = "mula_lu_lu";
+        break;
+    case OE_RRR(MULX, 0, X0):
+    case OE_RRR(MULX, 3, Y0):
+        tcg_gen_mul_tl(tdest, tsrca, tsrcb);
+        tcg_gen_ext32s_tl(tdest, tdest);
+        mnemonic = "mulx";
+        break;
+    case OE_RRR(MUL_HS_HS, 0, X0):
+    case OE_RRR(MUL_HS_HS, 8, Y0):
+        gen_mul_half(tdest, tsrca, tsrcb, HS, HS);
+        mnemonic = "mul_hs_hs";
+        break;
+    case OE_RRR(MUL_HS_HU, 0, X0):
+        gen_mul_half(tdest, tsrca, tsrcb, HS, HU);
+        mnemonic = "mul_hs_hu";
+        break;
+    case OE_RRR(MUL_HS_LS, 0, X0):
+        gen_mul_half(tdest, tsrca, tsrcb, HS, LS);
+        mnemonic = "mul_hs_ls";
+        break;
+    case OE_RRR(MUL_HS_LU, 0, X0):
+        gen_mul_half(tdest, tsrca, tsrcb, HS, LU);
+        mnemonic = "mul_hs_lu";
+        break;
+    case OE_RRR(MUL_HU_HU, 0, X0):
+    case OE_RRR(MUL_HU_HU, 8, Y0):
+        gen_mul_half(tdest, tsrca, tsrcb, HU, HU);
+        mnemonic = "mul_hu_hu";
+        break;
+    case OE_RRR(MUL_HU_LS, 0, X0):
+        gen_mul_half(tdest, tsrca, tsrcb, HU, LS);
+        mnemonic = "mul_hu_ls";
+        break;
+    case OE_RRR(MUL_HU_LU, 0, X0):
+        gen_mul_half(tdest, tsrca, tsrcb, HU, LU);
+        mnemonic = "mul_hu_lu";
+        break;
+    case OE_RRR(MUL_LS_LS, 0, X0):
+    case OE_RRR(MUL_LS_LS, 8, Y0):
+        gen_mul_half(tdest, tsrca, tsrcb, LS, LS);
+        mnemonic = "mul_ls_ls";
+        break;
+    case OE_RRR(MUL_LS_LU, 0, X0):
+        gen_mul_half(tdest, tsrca, tsrcb, LS, LU);
+        mnemonic = "mul_ls_lu";
+        break;
+    case OE_RRR(MUL_LU_LU, 0, X0):
+    case OE_RRR(MUL_LU_LU, 8, Y0):
+        gen_mul_half(tdest, tsrca, tsrcb, LU, LU);
+        mnemonic = "mul_lu_lu";
+        break;
+    case OE_RRR(MZ, 0, X0):
+    case OE_RRR(MZ, 0, X1):
+    case OE_RRR(MZ, 4, Y0):
+    case OE_RRR(MZ, 4, Y1):
+        t0 = load_zero(dc);
+        tcg_gen_movcond_tl(TCG_COND_EQ, tdest, tsrca, t0, tsrcb, t0);
+        mnemonic = "mz";
+        break;
+    case OE_RRR(NOR, 0, X0):
+    case OE_RRR(NOR, 0, X1):
+    case OE_RRR(NOR, 5, Y0):
+    case OE_RRR(NOR, 5, Y1):
+        tcg_gen_nor_tl(tdest, tsrca, tsrcb);
+        mnemonic = "nor";
+        break;
+    case OE_RRR(OR, 0, X0):
+    case OE_RRR(OR, 0, X1):
+    case OE_RRR(OR, 5, Y0):
+    case OE_RRR(OR, 5, Y1):
+        tcg_gen_or_tl(tdest, tsrca, tsrcb);
+        mnemonic = "or";
+        break;
+    case OE_RRR(ROTL, 0, X0):
+    case OE_RRR(ROTL, 0, X1):
+    case OE_RRR(ROTL, 6, Y0):
+    case OE_RRR(ROTL, 6, Y1):
+        tcg_gen_andi_tl(tdest, tsrcb, 63);
+        tcg_gen_rotl_tl(tdest, tsrca, tdest);
+        mnemonic = "rotl";
+        break;
+    case OE_RRR(SHL1ADDX, 0, X0):
+    case OE_RRR(SHL1ADDX, 0, X1):
+    case OE_RRR(SHL1ADDX, 7, Y0):
+    case OE_RRR(SHL1ADDX, 7, Y1):
+        tcg_gen_shli_tl(tdest, tsrca, 1);
+        tcg_gen_add_tl(tdest, tdest, tsrcb);
+        tcg_gen_ext32s_tl(tdest, tdest);
+        mnemonic = "shl1addx";
+        break;
+    case OE_RRR(SHL1ADD, 0, X0):
+    case OE_RRR(SHL1ADD, 0, X1):
+    case OE_RRR(SHL1ADD, 1, Y0):
+    case OE_RRR(SHL1ADD, 1, Y1):
+        tcg_gen_shli_tl(tdest, tsrca, 1);
+        tcg_gen_add_tl(tdest, tdest, tsrcb);
+        mnemonic = "shl1add";
+        break;
+    case OE_RRR(SHL2ADDX, 0, X0):
+    case OE_RRR(SHL2ADDX, 0, X1):
+    case OE_RRR(SHL2ADDX, 7, Y0):
+    case OE_RRR(SHL2ADDX, 7, Y1):
+        tcg_gen_shli_tl(tdest, tsrca, 2);
+        tcg_gen_add_tl(tdest, tdest, tsrcb);
+        tcg_gen_ext32s_tl(tdest, tdest);
+        mnemonic = "shl2addx";
+        break;
+    case OE_RRR(SHL2ADD, 0, X0):
+    case OE_RRR(SHL2ADD, 0, X1):
+    case OE_RRR(SHL2ADD, 1, Y0):
+    case OE_RRR(SHL2ADD, 1, Y1):
+        tcg_gen_shli_tl(tdest, tsrca, 2);
+        tcg_gen_add_tl(tdest, tdest, tsrcb);
+        mnemonic = "shl2add";
+        break;
+    case OE_RRR(SHL3ADDX, 0, X0):
+    case OE_RRR(SHL3ADDX, 0, X1):
+    case OE_RRR(SHL3ADDX, 7, Y0):
+    case OE_RRR(SHL3ADDX, 7, Y1):
+        tcg_gen_shli_tl(tdest, tsrca, 3);
+        tcg_gen_add_tl(tdest, tdest, tsrcb);
+        tcg_gen_ext32s_tl(tdest, tdest);
+        mnemonic = "shl3addx";
+        break;
+    case OE_RRR(SHL3ADD, 0, X0):
+    case OE_RRR(SHL3ADD, 0, X1):
+    case OE_RRR(SHL3ADD, 1, Y0):
+    case OE_RRR(SHL3ADD, 1, Y1):
+        tcg_gen_shli_tl(tdest, tsrca, 3);
+        tcg_gen_add_tl(tdest, tdest, tsrcb);
+        mnemonic = "shl3add";
+        break;
+    case OE_RRR(SHLX, 0, X0):
+    case OE_RRR(SHLX, 0, X1):
+        tcg_gen_andi_tl(tdest, tsrcb, 31);
+        tcg_gen_shl_tl(tdest, tsrca, tdest);
+        tcg_gen_ext32s_tl(tdest, tdest);
+        mnemonic = "shlx";
+        break;
+    case OE_RRR(SHL, 0, X0):
+    case OE_RRR(SHL, 0, X1):
+    case OE_RRR(SHL, 6, Y0):
+    case OE_RRR(SHL, 6, Y1):
+        tcg_gen_andi_tl(tdest, tsrcb, 63);
+        tcg_gen_shl_tl(tdest, tsrca, tdest);
+        mnemonic = "shl";
+        break;
+    case OE_RRR(SHRS, 0, X0):
+    case OE_RRR(SHRS, 0, X1):
+    case OE_RRR(SHRS, 6, Y0):
+    case OE_RRR(SHRS, 6, Y1):
+        tcg_gen_andi_tl(tdest, tsrcb, 63);
+        tcg_gen_sar_tl(tdest, tsrca, tdest);
+        mnemonic = "shrs";
+        break;
+    case OE_RRR(SHRUX, 0, X0):
+    case OE_RRR(SHRUX, 0, X1):
+        t0 = tcg_temp_new();
+        tcg_gen_andi_tl(t0, tsrcb, 31);
+        tcg_gen_ext32u_tl(tdest, tsrca);
+        tcg_gen_shr_tl(tdest, tdest, t0);
+        tcg_gen_ext32s_tl(tdest, tdest);
+        tcg_temp_free(t0);
+        mnemonic = "shrux";
+        break;
+    case OE_RRR(SHRU, 0, X0):
+    case OE_RRR(SHRU, 0, X1):
+    case OE_RRR(SHRU, 6, Y0):
+    case OE_RRR(SHRU, 6, Y1):
+        tcg_gen_andi_tl(tdest, tsrcb, 63);
+        tcg_gen_shr_tl(tdest, tsrca, tdest);
+        mnemonic = "shru";
+        break;
+    case OE_RRR(SHUFFLEBYTES, 0, X0):
+        gen_helper_shufflebytes(tdest, load_gr(dc, dest), tsrca, tsrca);
+        mnemonic = "shufflebytes";
+        break;
+    case OE_RRR(SUBXSC, 0, X0):
+    case OE_RRR(SUBXSC, 0, X1):
+        gen_saturate_op(tdest, tsrca, tsrcb, tcg_gen_sub_tl);
+        mnemonic = "subxsc";
+        break;
+    case OE_RRR(SUBX, 0, X0):
+    case OE_RRR(SUBX, 0, X1):
+    case OE_RRR(SUBX, 0, Y0):
+    case OE_RRR(SUBX, 0, Y1):
+        tcg_gen_sub_tl(tdest, tsrca, tsrcb);
+        tcg_gen_ext32s_tl(tdest, tdest);
+        mnemonic = "subx";
+        break;
+    case OE_RRR(SUB, 0, X0):
+    case OE_RRR(SUB, 0, X1):
+    case OE_RRR(SUB, 0, Y0):
+    case OE_RRR(SUB, 0, Y1):
+        tcg_gen_sub_tl(tdest, tsrca, tsrcb);
+        mnemonic = "sub";
+        break;
+    case OE_RRR(V1ADDUC, 0, X0):
+    case OE_RRR(V1ADDUC, 0, X1):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V1ADD, 0, X0):
+    case OE_RRR(V1ADD, 0, X1):
+        gen_v12add(tdest, tsrca, tsrcb, V1_IMM(0x80));
+        mnemonic = "v1add";
+        break;
+    case OE_RRR(V1ADIFFU, 0, X0):
+    case OE_RRR(V1AVGU, 0, X0):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V1CMPEQ, 0, X0):
+    case OE_RRR(V1CMPEQ, 0, X1):
+        tcg_gen_xor_tl(tdest, tsrca, tsrcb);
+        gen_v1cmpeq0(tdest);
+        mnemonic = "v1cmpeq";
+        break;
+    case OE_RRR(V1CMPLES, 0, X0):
+    case OE_RRR(V1CMPLES, 0, X1):
+    case OE_RRR(V1CMPLEU, 0, X0):
+    case OE_RRR(V1CMPLEU, 0, X1):
+    case OE_RRR(V1CMPLTS, 0, X0):
+    case OE_RRR(V1CMPLTS, 0, X1):
+    case OE_RRR(V1CMPLTU, 0, X0):
+    case OE_RRR(V1CMPLTU, 0, X1):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V1CMPNE, 0, X0):
+    case OE_RRR(V1CMPNE, 0, X1):
+        tcg_gen_xor_tl(tdest, tsrca, tsrcb);
+        gen_v1cmpne0(tdest);
+        mnemonic = "v1cmpne";
+        break;
+    case OE_RRR(V1DDOTPUA, 0, X0):
+    case OE_RRR(V1DDOTPUSA, 0, X0):
+    case OE_RRR(V1DDOTPUS, 0, X0):
+    case OE_RRR(V1DDOTPU, 0, X0):
+    case OE_RRR(V1DOTPA, 0, X0):
+    case OE_RRR(V1DOTPUA, 0, X0):
+    case OE_RRR(V1DOTPUSA, 0, X0):
+    case OE_RRR(V1DOTPUS, 0, X0):
+    case OE_RRR(V1DOTPU, 0, X0):
+    case OE_RRR(V1DOTP, 0, X0):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V1INT_H, 0, X0):
+    case OE_RRR(V1INT_H, 0, X1):
+        gen_helper_v1int_h(tdest, tsrca, tsrcb);
+        mnemonic = "v1int_h";
+        break;
+    case OE_RRR(V1INT_L, 0, X0):
+    case OE_RRR(V1INT_L, 0, X1):
+        gen_helper_v1int_l(tdest, tsrca, tsrcb);
+        mnemonic = "v1int_l";
+        break;
+    case OE_RRR(V1MAXU, 0, X0):
+    case OE_RRR(V1MAXU, 0, X1):
+    case OE_RRR(V1MINU, 0, X0):
+    case OE_RRR(V1MINU, 0, X1):
+    case OE_RRR(V1MNZ, 0, X0):
+    case OE_RRR(V1MNZ, 0, X1):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V1MULTU, 0, X0):
+        gen_helper_v1multu(tdest, tsrca, tsrcb);
+        mnemonic = "v1multu";
+        break;
+    case OE_RRR(V1MULUS, 0, X0):
+    case OE_RRR(V1MULU, 0, X0):
+    case OE_RRR(V1MZ, 0, X0):
+    case OE_RRR(V1MZ, 0, X1):
+    case OE_RRR(V1SADAU, 0, X0):
+    case OE_RRR(V1SADU, 0, X0):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V1SHL, 0, X0):
+    case OE_RRR(V1SHL, 0, X1):
+        gen_helper_v1shl(tdest, tsrca, tsrcb);
+        mnemonic = "v1shl";
+        break;
+    case OE_RRR(V1SHRS, 0, X0):
+    case OE_RRR(V1SHRS, 0, X1):
+        gen_helper_v1shrs(tdest, tsrca, tsrcb);
+        mnemonic = "v1shrs";
+        break;
+    case OE_RRR(V1SHRU, 0, X0):
+    case OE_RRR(V1SHRU, 0, X1):
+        gen_helper_v1shru(tdest, tsrca, tsrcb);
+        mnemonic = "v1shru";
+        break;
+    case OE_RRR(V1SUBUC, 0, X0):
+    case OE_RRR(V1SUBUC, 0, X1):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V1SUB, 0, X0):
+    case OE_RRR(V1SUB, 0, X1):
+        gen_v12sub(tdest, tsrca, tsrcb, V1_IMM(0x80));
+        mnemonic = "v1sub";
+        break;
+    case OE_RRR(V2ADDSC, 0, X0):
+    case OE_RRR(V2ADDSC, 0, X1):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V2ADD, 0, X0):
+    case OE_RRR(V2ADD, 0, X1):
+        gen_v12add(tdest, tsrca, tsrcb, V2_IMM(0x8000));
+        mnemonic = "v2add";
+        break;
+    case OE_RRR(V2ADIFFS, 0, X0):
+    case OE_RRR(V2AVGS, 0, X0):
+    case OE_RRR(V2CMPEQ, 0, X0):
+    case OE_RRR(V2CMPEQ, 0, X1):
+    case OE_RRR(V2CMPLES, 0, X0):
+    case OE_RRR(V2CMPLES, 0, X1):
+    case OE_RRR(V2CMPLEU, 0, X0):
+    case OE_RRR(V2CMPLEU, 0, X1):
+    case OE_RRR(V2CMPLTS, 0, X0):
+    case OE_RRR(V2CMPLTS, 0, X1):
+    case OE_RRR(V2CMPLTU, 0, X0):
+    case OE_RRR(V2CMPLTU, 0, X1):
+    case OE_RRR(V2CMPNE, 0, X0):
+    case OE_RRR(V2CMPNE, 0, X1):
+    case OE_RRR(V2DOTPA, 0, X0):
+    case OE_RRR(V2DOTP, 0, X0):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V2INT_H, 0, X0):
+    case OE_RRR(V2INT_H, 0, X1):
+        gen_helper_v2int_h(tdest, tsrca, tsrcb);
+        mnemonic = "v2int_h";
+        break;
+    case OE_RRR(V2INT_L, 0, X0):
+    case OE_RRR(V2INT_L, 0, X1):
+        gen_helper_v2int_l(tdest, tsrca, tsrcb);
+        mnemonic = "v2int_l";
+        break;
+    case OE_RRR(V2MAXS, 0, X0):
+    case OE_RRR(V2MAXS, 0, X1):
+    case OE_RRR(V2MINS, 0, X0):
+    case OE_RRR(V2MINS, 0, X1):
+    case OE_RRR(V2MNZ, 0, X0):
+    case OE_RRR(V2MNZ, 0, X1):
+    case OE_RRR(V2MULFSC, 0, X0):
+    case OE_RRR(V2MULS, 0, X0):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V2MULTS, 0, X0):
+        gen_helper_v2mults(tdest, tsrca, tsrcb);
+        mnemonic = "v2mults";
+        break;
+    case OE_RRR(V2MZ, 0, X0):
+    case OE_RRR(V2MZ, 0, X1):
+    case OE_RRR(V2PACKH, 0, X0):
+    case OE_RRR(V2PACKH, 0, X1):
+    case OE_RRR(V2PACKL, 0, X0):
+    case OE_RRR(V2PACKL, 0, X1):
+    case OE_RRR(V2PACKUC, 0, X0):
+    case OE_RRR(V2PACKUC, 0, X1):
+    case OE_RRR(V2SADAS, 0, X0):
+    case OE_RRR(V2SADAU, 0, X0):
+    case OE_RRR(V2SADS, 0, X0):
+    case OE_RRR(V2SADU, 0, X0):
+    case OE_RRR(V2SHLSC, 0, X0):
+    case OE_RRR(V2SHLSC, 0, X1):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V2SHL, 0, X0):
+    case OE_RRR(V2SHL, 0, X1):
+        gen_helper_v2shl(tdest, tsrca, tsrcb);
+        mnemonic = "v2shl";
+        break;
+    case OE_RRR(V2SHRS, 0, X0):
+    case OE_RRR(V2SHRS, 0, X1):
+        gen_helper_v2shrs(tdest, tsrca, tsrcb);
+        mnemonic = "v2shrs";
+        break;
+    case OE_RRR(V2SHRU, 0, X0):
+    case OE_RRR(V2SHRU, 0, X1):
+        gen_helper_v2shru(tdest, tsrca, tsrcb);
+        mnemonic = "v2shru";
+        break;
+    case OE_RRR(V2SUBSC, 0, X0):
+    case OE_RRR(V2SUBSC, 0, X1):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V2SUB, 0, X0):
+    case OE_RRR(V2SUB, 0, X1):
+        gen_v12sub(tdest, tsrca, tsrcb, V2_IMM(0x8000));
+        mnemonic = "v2sub";
+        break;
+    case OE_RRR(V4ADDSC, 0, X0):
+    case OE_RRR(V4ADDSC, 0, X1):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V4ADD, 0, X0):
+    case OE_RRR(V4ADD, 0, X1):
+        gen_v4op(tdest, tsrca, tsrcb, tcg_gen_add_i32);
+        mnemonic = "v4add";
+        break;
+    case OE_RRR(V4INT_H, 0, X0):
+    case OE_RRR(V4INT_H, 0, X1):
+        tcg_gen_shri_tl(tdest, tsrcb, 32);
+        tcg_gen_deposit_tl(tdest, tsrca, tdest, 0, 32);
+        mnemonic = "v4int_h";
+        break;
+    case OE_RRR(V4INT_L, 0, X0):
+    case OE_RRR(V4INT_L, 0, X1):
+        tcg_gen_deposit_tl(tdest, tsrcb, tsrca, 32, 32);
+        mnemonic = "v4int_l";
+        break;
+    case OE_RRR(V4PACKSC, 0, X0):
+    case OE_RRR(V4PACKSC, 0, X1):
+    case OE_RRR(V4SHLSC, 0, X0):
+    case OE_RRR(V4SHLSC, 0, X1):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V4SHL, 0, X0):
+    case OE_RRR(V4SHL, 0, X1):
+        gen_v4sh(tdest, tsrca, tsrcb, tcg_gen_shl_i32);
+        mnemonic = "v4shl";
+        break;
+    case OE_RRR(V4SHRS, 0, X0):
+    case OE_RRR(V4SHRS, 0, X1):
+        gen_v4sh(tdest, tsrca, tsrcb, tcg_gen_sar_i32);
+        mnemonic = "v4shrs";
+        break;
+    case OE_RRR(V4SHRU, 0, X0):
+    case OE_RRR(V4SHRU, 0, X1):
+        gen_v4sh(tdest, tsrca, tsrcb, tcg_gen_shr_i32);
+        mnemonic = "v4shru";
+        break;
+    case OE_RRR(V4SUBSC, 0, X0):
+    case OE_RRR(V4SUBSC, 0, X1):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_RRR(V4SUB, 0, X0):
+    case OE_RRR(V4SUB, 0, X1):
+        gen_v4op(tdest, tsrca, tsrcb, tcg_gen_sub_i32);
+        mnemonic = "v2sub";
+        break;
+    case OE_RRR(XOR, 0, X0):
+    case OE_RRR(XOR, 0, X1):
+    case OE_RRR(XOR, 5, Y0):
+    case OE_RRR(XOR, 5, Y1):
+        tcg_gen_xor_tl(tdest, tsrca, tsrcb);
+        mnemonic = "xor";
+        break;
+    default:
+        return TILEGX_EXCP_OPCODE_UNKNOWN;
+    }
+
+    qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s, %s", mnemonic,
+                  reg_names[dest], reg_names[srca], reg_names[srcb]);
+    return TILEGX_EXCP_NONE;
+}
+
+static TileExcp gen_rri_opcode(DisasContext *dc, unsigned opext,
+                               unsigned dest, unsigned srca, int imm)
+{
+    TCGv tdest = dest_gr(dc, dest);
+    TCGv tsrca = load_gr(dc, srca);
+    bool prefetch_nofault = false;
+    const char *mnemonic;
+    TCGMemOp memop;
+    int i2, i3;
+    TCGv t0;
+
+    switch (opext) {
+    case OE(ADDI_OPCODE_Y0, 0, Y0):
+    case OE(ADDI_OPCODE_Y1, 0, Y1):
+    case OE_IM(ADDI, X0):
+    case OE_IM(ADDI, X1):
+        tcg_gen_addi_tl(tdest, tsrca, imm);
+        mnemonic = "addi";
+        break;
+    case OE(ADDXI_OPCODE_Y0, 0, Y0):
+    case OE(ADDXI_OPCODE_Y1, 0, Y1):
+    case OE_IM(ADDXI, X0):
+    case OE_IM(ADDXI, X1):
+        tcg_gen_addi_tl(tdest, tsrca, imm);
+        tcg_gen_ext32s_tl(tdest, tdest);
+        mnemonic = "addxi";
+        break;
+    case OE(ANDI_OPCODE_Y0, 0, Y0):
+    case OE(ANDI_OPCODE_Y1, 0, Y1):
+    case OE_IM(ANDI, X0):
+    case OE_IM(ANDI, X1):
+        tcg_gen_andi_tl(tdest, tsrca, imm);
+        mnemonic = "andi";
+        break;
+    case OE(CMPEQI_OPCODE_Y0, 0, Y0):
+    case OE(CMPEQI_OPCODE_Y1, 0, Y1):
+    case OE_IM(CMPEQI, X0):
+    case OE_IM(CMPEQI, X1):
+        tcg_gen_setcondi_tl(TCG_COND_EQ, tdest, tsrca, imm);
+        mnemonic = "cmpeqi";
+        break;
+    case OE(CMPLTSI_OPCODE_Y0, 0, Y0):
+    case OE(CMPLTSI_OPCODE_Y1, 0, Y1):
+    case OE_IM(CMPLTSI, X0):
+    case OE_IM(CMPLTSI, X1):
+        tcg_gen_setcondi_tl(TCG_COND_LT, tdest, tsrca, imm);
+        mnemonic = "cmpltsi";
+        break;
+    case OE_IM(CMPLTUI, X0):
+    case OE_IM(CMPLTUI, X1):
+        tcg_gen_setcondi_tl(TCG_COND_LTU, tdest, tsrca, imm);
+        mnemonic = "cmpltui";
+        break;
+    case OE_IM(LD1S_ADD, X1):
+        memop = MO_SB;
+        mnemonic = "ld1s_add"; /* prefetch_add_l1_fault */
+        goto do_load_add;
+    case OE_IM(LD1U_ADD, X1):
+        memop = MO_UB;
+        mnemonic = "ld1u_add"; /* prefetch_add_l1 */
+        prefetch_nofault = (dest == TILEGX_R_ZERO);
+        goto do_load_add;
+    case OE_IM(LD2S_ADD, X1):
+        memop = MO_TESW;
+        mnemonic = "ld2s_add"; /* prefetch_add_l2_fault */
+        goto do_load_add;
+    case OE_IM(LD2U_ADD, X1):
+        memop = MO_TEUW;
+        mnemonic = "ld2u_add"; /* prefetch_add_l2 */
+        prefetch_nofault = (dest == TILEGX_R_ZERO);
+        goto do_load_add;
+    case OE_IM(LD4S_ADD, X1):
+        memop = MO_TESL;
+        mnemonic = "ld4s_add"; /* prefetch_add_l3_fault */
+        goto do_load_add;
+    case OE_IM(LD4U_ADD, X1):
+        memop = MO_TEUL;
+        mnemonic = "ld4u_add"; /* prefetch_add_l3 */
+        prefetch_nofault = (dest == TILEGX_R_ZERO);
+        goto do_load_add;
+    case OE_IM(LDNT1S_ADD, X1):
+        memop = MO_SB;
+        mnemonic = "ldnt1s_add";
+        goto do_load_add;
+    case OE_IM(LDNT1U_ADD, X1):
+        memop = MO_UB;
+        mnemonic = "ldnt1u_add";
+        goto do_load_add;
+    case OE_IM(LDNT2S_ADD, X1):
+        memop = MO_TESW;
+        mnemonic = "ldnt2s_add";
+        goto do_load_add;
+    case OE_IM(LDNT2U_ADD, X1):
+        memop = MO_TEUW;
+        mnemonic = "ldnt2u_add";
+        goto do_load_add;
+    case OE_IM(LDNT4S_ADD, X1):
+        memop = MO_TESL;
+        mnemonic = "ldnt4s_add";
+        goto do_load_add;
+    case OE_IM(LDNT4U_ADD, X1):
+        memop = MO_TEUL;
+        mnemonic = "ldnt4u_add";
+        goto do_load_add;
+    case OE_IM(LDNT_ADD, X1):
+        memop = MO_TEQ;
+        mnemonic = "ldnt_add";
+        goto do_load_add;
+    case OE_IM(LD_ADD, X1):
+        memop = MO_TEQ;
+        mnemonic = "ld_add";
+    do_load_add:
+        if (!prefetch_nofault) {
+            tcg_gen_qemu_ld_tl(tdest, tsrca, dc->mmuidx, memop);
+        }
+        tcg_gen_addi_tl(dest_gr(dc, srca), tsrca, imm);
+        break;
+    case OE_IM(LDNA_ADD, X1):
+        tcg_gen_andi_tl(tdest, tsrca, ~7);
+        tcg_gen_qemu_ld_tl(tdest, tdest, dc->mmuidx, MO_TEQ);
+        tcg_gen_addi_tl(dest_gr(dc, srca), tsrca, imm);
+        mnemonic = "ldna_add";
+        break;
+    case OE_IM(ORI, X0):
+    case OE_IM(ORI, X1):
+        tcg_gen_ori_tl(tdest, tsrca, imm);
+        mnemonic = "ori";
+        break;
+    case OE_IM(V1ADDI, X0):
+    case OE_IM(V1ADDI, X1):
+        t0 = tcg_const_tl(V1_IMM(imm));
+        gen_v12add(tdest, tsrca, t0, V1_IMM(0x80));
+        tcg_temp_free(t0);
+        mnemonic = "v1addi";
+        break;
+    case OE_IM(V1CMPEQI, X0):
+    case OE_IM(V1CMPEQI, X1):
+        tcg_gen_xori_tl(tdest, tsrca, V1_IMM(imm));
+        gen_v1cmpeq0(tdest);
+        mnemonic = "v1cmpeqi";
+        break;
+    case OE_IM(V1CMPLTSI, X0):
+    case OE_IM(V1CMPLTSI, X1):
+    case OE_IM(V1CMPLTUI, X0):
+    case OE_IM(V1CMPLTUI, X1):
+    case OE_IM(V1MAXUI, X0):
+    case OE_IM(V1MAXUI, X1):
+    case OE_IM(V1MINUI, X0):
+    case OE_IM(V1MINUI, X1):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_IM(V2ADDI, X0):
+    case OE_IM(V2ADDI, X1):
+        t0 = tcg_const_tl(V2_IMM(imm));
+        gen_v12add(tdest, tsrca, t0, V2_IMM(0x8000));
+        tcg_temp_free(t0);
+        mnemonic = "v2addi";
+        break;
+    case OE_IM(V2CMPEQI, X0):
+    case OE_IM(V2CMPEQI, X1):
+    case OE_IM(V2CMPLTSI, X0):
+    case OE_IM(V2CMPLTSI, X1):
+    case OE_IM(V2CMPLTUI, X0):
+    case OE_IM(V2CMPLTUI, X1):
+    case OE_IM(V2MAXSI, X0):
+    case OE_IM(V2MAXSI, X1):
+    case OE_IM(V2MINSI, X0):
+    case OE_IM(V2MINSI, X1):
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    case OE_IM(XORI, X0):
+    case OE_IM(XORI, X1):
+        tcg_gen_xori_tl(tdest, tsrca, imm);
+        mnemonic = "xori";
+        break;
+
+    case OE_SH(ROTLI, X0):
+    case OE_SH(ROTLI, X1):
+    case OE_SH(ROTLI, Y0):
+    case OE_SH(ROTLI, Y1):
+        tcg_gen_rotli_tl(tdest, tsrca, imm);
+        mnemonic = "rotli";
+        break;
+    case OE_SH(SHLI, X0):
+    case OE_SH(SHLI, X1):
+    case OE_SH(SHLI, Y0):
+    case OE_SH(SHLI, Y1):
+        tcg_gen_shli_tl(tdest, tsrca, imm);
+        mnemonic = "shli";
+        break;
+    case OE_SH(SHLXI, X0):
+    case OE_SH(SHLXI, X1):
+        tcg_gen_shli_tl(tdest, tsrca, imm & 31);
+        tcg_gen_ext32s_tl(tdest, tdest);
+        mnemonic = "shlxi";
+        break;
+    case OE_SH(SHRSI, X0):
+    case OE_SH(SHRSI, X1):
+    case OE_SH(SHRSI, Y0):
+    case OE_SH(SHRSI, Y1):
+        tcg_gen_sari_tl(tdest, tsrca, imm);
+        mnemonic = "shrsi";
+        break;
+    case OE_SH(SHRUI, X0):
+    case OE_SH(SHRUI, X1):
+    case OE_SH(SHRUI, Y0):
+    case OE_SH(SHRUI, Y1):
+        tcg_gen_shri_tl(tdest, tsrca, imm);
+        mnemonic = "shrui";
+        break;
+    case OE_SH(SHRUXI, X0):
+    case OE_SH(SHRUXI, X1):
+        if ((imm & 31) == 0) {
+            tcg_gen_ext32s_tl(tdest, tsrca);
+        } else {
+            tcg_gen_ext32u_tl(tdest, tsrca);
+            tcg_gen_shri_tl(tdest, tdest, imm & 31);
+        }
+        mnemonic = "shlxi";
+        break;
+    case OE_SH(V1SHLI, X0):
+    case OE_SH(V1SHLI, X1):
+        i2 = imm & 7;
+        i3 = 0xff >> i2;
+        tcg_gen_andi_tl(tdest, tsrca, V1_IMM(i3));
+        tcg_gen_shli_tl(tdest, tdest, i2);
+        mnemonic = "v1shli";
+        break;
+    case OE_SH(V1SHRSI, X0):
+    case OE_SH(V1SHRSI, X1):
+        t0 = tcg_const_tl(imm & 7);
+        gen_helper_v1shrs(tdest, tsrca, t0);
+        tcg_temp_free(t0);
+        mnemonic = "v1shrsi";
+        break;
+    case OE_SH(V1SHRUI, X0):
+    case OE_SH(V1SHRUI, X1):
+        i2 = imm & 7;
+        i3 = (0xff << i2) & 0xff;
+        tcg_gen_andi_tl(tdest, tsrca, V1_IMM(i3));
+        tcg_gen_shri_tl(tdest, tdest, i2);
+        mnemonic = "v1shrui";
+        break;
+    case OE_SH(V2SHLI, X0):
+    case OE_SH(V2SHLI, X1):
+        i2 = imm & 15;
+        i3 = 0xffff >> i2;
+        tcg_gen_andi_tl(tdest, tsrca, V2_IMM(i3));
+        tcg_gen_shli_tl(tdest, tdest, i2);
+        mnemonic = "v2shli";
+        break;
+    case OE_SH(V2SHRSI, X0):
+    case OE_SH(V2SHRSI, X1):
+        t0 = tcg_const_tl(imm & 15);
+        gen_helper_v2shrs(tdest, tsrca, t0);
+        tcg_temp_free(t0);
+        mnemonic = "v2shrsi";
+        break;
+    case OE_SH(V2SHRUI, X0):
+    case OE_SH(V2SHRUI, X1):
+        i2 = imm & 15;
+        i3 = (0xffff << i2) & 0xffff;
+        tcg_gen_andi_tl(tdest, tsrca, V2_IMM(i3));
+        tcg_gen_shri_tl(tdest, tdest, i2);
+        mnemonic = "v2shrui";
+        break;
+
+    case OE(ADDLI_OPCODE_X0, 0, X0):
+    case OE(ADDLI_OPCODE_X1, 0, X1):
+        tcg_gen_addi_tl(tdest, tsrca, imm);
+        mnemonic = "addli";
+        break;
+    case OE(ADDXLI_OPCODE_X0, 0, X0):
+    case OE(ADDXLI_OPCODE_X1, 0, X1):
+        tcg_gen_addi_tl(tdest, tsrca, imm);
+        tcg_gen_ext32s_tl(tdest, tdest);
+        mnemonic = "addxli";
+        break;
+    case OE(SHL16INSLI_OPCODE_X0, 0, X0):
+    case OE(SHL16INSLI_OPCODE_X1, 0, X1):
+        tcg_gen_shli_tl(tdest, tsrca, 16);
+        tcg_gen_ori_tl(tdest, tdest, imm & 0xffff);
+        mnemonic = "shl16insli";
+        break;
+
+    default:
+        return TILEGX_EXCP_OPCODE_UNKNOWN;
+    }
+
+    qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s, %d", mnemonic,
+                  reg_names[dest], reg_names[srca], imm);
+    return TILEGX_EXCP_NONE;
+}
+
+static TileExcp gen_bf_opcode_x0(DisasContext *dc, unsigned ext,
+                                 unsigned dest, unsigned srca,
+                                 unsigned bfs, unsigned bfe)
+{
+    TCGv tdest = dest_gr(dc, dest);
+    TCGv tsrca = load_gr(dc, srca);
+    TCGv tsrcd;
+    int len;
+    const char *mnemonic;
+
+    /* The bitfield is either between E and S inclusive,
+       or up from S and down from E inclusive.  */
+    if (bfs <= bfe) {
+        len = bfe - bfs + 1;
+    } else {
+        len = (64 - bfs) + (bfe + 1);
+    }
+
+    switch (ext) {
+    case BFEXTU_BF_OPCODE_X0:
+        if (bfs == 0 && bfe == 7) {
+            tcg_gen_ext8u_tl(tdest, tsrca);
+        } else if (bfs == 0 && bfe == 15) {
+            tcg_gen_ext16u_tl(tdest, tsrca);
+        } else if (bfs == 0 && bfe == 31) {
+            tcg_gen_ext32u_tl(tdest, tsrca);
+        } else {
+            int rol = 63 - bfe;
+            if (bfs <= bfe) {
+                tcg_gen_shli_tl(tdest, tsrca, rol);
+            } else {
+                tcg_gen_rotli_tl(tdest, tsrca, rol);
+            }
+            tcg_gen_shri_tl(tdest, tdest, (bfs + rol) & 63);
+        }
+        mnemonic = "bfextu";
+        break;
+
+    case BFEXTS_BF_OPCODE_X0:
+        if (bfs == 0 && bfe == 7) {
+            tcg_gen_ext8s_tl(tdest, tsrca);
+        } else if (bfs == 0 && bfe == 15) {
+            tcg_gen_ext16s_tl(tdest, tsrca);
+        } else if (bfs == 0 && bfe == 31) {
+            tcg_gen_ext32s_tl(tdest, tsrca);
+        } else {
+            int rol = 63 - bfe;
+            if (bfs <= bfe) {
+                tcg_gen_shli_tl(tdest, tsrca, rol);
+            } else {
+                tcg_gen_rotli_tl(tdest, tsrca, rol);
+            }
+            tcg_gen_sari_tl(tdest, tdest, (bfs + rol) & 63);
+        }
+        mnemonic = "bfexts";
+        break;
+
+    case BFINS_BF_OPCODE_X0:
+        tsrcd = load_gr(dc, dest);
+        if (bfs <= bfe) {
+            tcg_gen_deposit_tl(tdest, tsrcd, tsrca, bfs, len);
+        } else {
+            tcg_gen_rotri_tl(tdest, tsrcd, bfs);
+            tcg_gen_deposit_tl(tdest, tdest, tsrca, 0, len);
+            tcg_gen_rotli_tl(tdest, tdest, bfs);
+        }
+        mnemonic = "bfins";
+        break;
+
+    case MM_BF_OPCODE_X0:
+        tsrcd = load_gr(dc, dest);
+        if (bfs == 0) {
+            tcg_gen_deposit_tl(tdest, tsrca, tsrcd, 0, len);
+        } else {
+            uint64_t mask = len == 64 ? -1 : rol64((1ULL << len) - 1, bfs);
+            TCGv tmp = tcg_const_tl(mask);
+
+            tcg_gen_and_tl(tdest, tsrcd, tmp);
+            tcg_gen_andc_tl(tmp, tsrca, tmp);
+            tcg_gen_or_tl(tdest, tdest, tmp);
+            tcg_temp_free(tmp);
+        }
+        mnemonic = "mm";
+        break;
+
+    default:
+        return TILEGX_EXCP_OPCODE_UNKNOWN;
+    }
+
+    qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s, %u, %u", mnemonic,
+                  reg_names[dest], reg_names[srca], bfs, bfe);
+    return TILEGX_EXCP_NONE;
+}
+
+static TileExcp gen_branch_opcode_x1(DisasContext *dc, unsigned ext,
+                                     unsigned srca, int off)
+{
+    target_ulong tgt = dc->pc + off * TILEGX_BUNDLE_SIZE_IN_BYTES;
+    const char *mnemonic;
+
+    dc->jmp.dest = tcg_const_tl(tgt);
+    dc->jmp.val1 = tcg_temp_new();
+    tcg_gen_mov_tl(dc->jmp.val1, load_gr(dc, srca));
+
+    /* Note that the "predict taken" opcodes have bit 0 clear.
+       Therefore, fold the two cases together by setting bit 0.  */
+    switch (ext | 1) {
+    case BEQZ_BRANCH_OPCODE_X1:
+        dc->jmp.cond = TCG_COND_EQ;
+        mnemonic = "beqz";
+        break;
+    case BNEZ_BRANCH_OPCODE_X1:
+        dc->jmp.cond = TCG_COND_NE;
+        mnemonic = "bnez";
+        break;
+    case BGEZ_BRANCH_OPCODE_X1:
+        dc->jmp.cond = TCG_COND_GE;
+        mnemonic = "bgez";
+        break;
+    case BGTZ_BRANCH_OPCODE_X1:
+        dc->jmp.cond = TCG_COND_GT;
+        mnemonic = "bgtz";
+        break;
+    case BLEZ_BRANCH_OPCODE_X1:
+        dc->jmp.cond = TCG_COND_LE;
+        mnemonic = "blez";
+        break;
+    case BLTZ_BRANCH_OPCODE_X1:
+        dc->jmp.cond = TCG_COND_LT;
+        mnemonic = "bltz";
+        break;
+    case BLBC_BRANCH_OPCODE_X1:
+        dc->jmp.cond = TCG_COND_EQ;
+        tcg_gen_andi_tl(dc->jmp.val1, dc->jmp.val1, 1);
+        mnemonic = "blbc";
+        break;
+    case BLBS_BRANCH_OPCODE_X1:
+        dc->jmp.cond = TCG_COND_NE;
+        tcg_gen_andi_tl(dc->jmp.val1, dc->jmp.val1, 1);
+        mnemonic = "blbs";
+        break;
+    default:
+        return TILEGX_EXCP_OPCODE_UNKNOWN;
+    }
+
+    if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
+        qemu_log("%s%s %s, " TARGET_FMT_lx " <%s>",
+                 mnemonic, ext & 1 ? "" : "t",
+                 reg_names[srca], tgt, lookup_symbol(tgt));
+    }
+    return TILEGX_EXCP_NONE;
+}
+
+static TileExcp gen_jump_opcode_x1(DisasContext *dc, unsigned ext, int off)
+{
+    target_ulong tgt = dc->pc + off * TILEGX_BUNDLE_SIZE_IN_BYTES;
+    const char *mnemonic = "j";
+
+    /* The extension field is 1 bit, therefore we only have JAL and J.  */
+    if (ext == JAL_JUMP_OPCODE_X1) {
+        tcg_gen_movi_tl(dest_gr(dc, TILEGX_R_LR),
+                        dc->pc + TILEGX_BUNDLE_SIZE_IN_BYTES);
+        mnemonic = "jal";
+    }
+    dc->jmp.cond = TCG_COND_ALWAYS;
+    dc->jmp.dest = tcg_const_tl(tgt);
+
+    if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
+        qemu_log("%s " TARGET_FMT_lx " <%s>",
+                 mnemonic, tgt, lookup_symbol(tgt));
+    }
+    return TILEGX_EXCP_NONE;
+}
+
+typedef struct {
+    const char *name;
+    intptr_t offset;
+    void (*get)(TCGv, TCGv_ptr);
+    void (*put)(TCGv_ptr, TCGv);
+} TileSPR;
+
+static const TileSPR *find_spr(unsigned spr)
+{
+    /* Allow the compiler to construct the binary search tree.  */
+#define D(N, O, G, P) \
+    case SPR_##N: { static const TileSPR x = { #N, O, G, P }; return &x; }
+
+    switch (spr) {
+    D(CMPEXCH_VALUE,
+      offsetof(CPUTLGState, spregs[TILEGX_SPR_CMPEXCH]), 0, 0)
+    D(INTERRUPT_CRITICAL_SECTION,
+      offsetof(CPUTLGState, spregs[TILEGX_SPR_CRITICAL_SEC]), 0, 0)
+    D(SIM_CONTROL,
+      offsetof(CPUTLGState, spregs[TILEGX_SPR_SIM_CONTROL]), 0, 0)
+    D(EX_CONTEXT_0_0,
+      offsetof(CPUTLGState, spregs[TILEGX_SPR_EX_CONTEXT_0_0]), 0, 0)
+    D(EX_CONTEXT_0_1,
+      offsetof(CPUTLGState, spregs[TILEGX_SPR_EX_CONTEXT_0_1]), 0, 0)
+    }
+
+#undef D
+
+    qemu_log_mask(LOG_UNIMP, "UNIMP SPR %u\n", spr);
+    return NULL;
+}
+
+static TileExcp gen_mtspr_x1(DisasContext *dc, unsigned spr, unsigned srca)
+{
+    const TileSPR *def = find_spr(spr);
+    TCGv tsrca;
+
+    if (def == NULL) {
+        qemu_log_mask(CPU_LOG_TB_IN_ASM, "mtspr spr[%u], %s", spr, reg_names[srca]);
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    }
+
+    tsrca = load_gr(dc, srca);
+    if (def->put) {
+        def->put(cpu_env, tsrca);
+    } else {
+        tcg_gen_st_tl(tsrca, cpu_env, def->offset);
+    }
+    qemu_log_mask(CPU_LOG_TB_IN_ASM, "mtspr %s, %s", def->name, reg_names[srca]);
+    return TILEGX_EXCP_NONE;
+}
+
+static TileExcp gen_mfspr_x1(DisasContext *dc, unsigned dest, unsigned spr)
+{
+    const TileSPR *def = find_spr(spr);
+    TCGv tdest;
+
+    if (def == NULL) {
+        qemu_log_mask(CPU_LOG_TB_IN_ASM, "mtspr %s, spr[%u]", reg_names[dest], spr);
+        return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
+    }
+
+    tdest = dest_gr(dc, dest);
+    if (def->get) {
+        def->get(tdest, cpu_env);
+    } else {
+        tcg_gen_ld_tl(tdest, cpu_env, def->offset);
+    }
+    qemu_log_mask(CPU_LOG_TB_IN_ASM, "mfspr %s, %s", reg_names[dest], def->name);
+    return TILEGX_EXCP_NONE;
+}
+
+static TileExcp decode_y0(DisasContext *dc, tilegx_bundle_bits bundle)
+{
+    unsigned opc = get_Opcode_Y0(bundle);
+    unsigned ext = get_RRROpcodeExtension_Y0(bundle);
+    unsigned dest = get_Dest_Y0(bundle);
+    unsigned srca = get_SrcA_Y0(bundle);
+    unsigned srcb;
+    int imm;
+
+    switch (opc) {
+    case RRR_1_OPCODE_Y0:
+        if (ext == UNARY_RRR_1_OPCODE_Y0) {
+            ext = get_UnaryOpcodeExtension_Y0(bundle);
+            return gen_rr_opcode(dc, OE(opc, ext, Y0), dest, srca, bundle);
+        }
+        /* fallthru */
+    case RRR_0_OPCODE_Y0:
+    case RRR_2_OPCODE_Y0:
+    case RRR_3_OPCODE_Y0:
+    case RRR_4_OPCODE_Y0:
+    case RRR_5_OPCODE_Y0:
+    case RRR_6_OPCODE_Y0:
+    case RRR_7_OPCODE_Y0:
+    case RRR_8_OPCODE_Y0:
+    case RRR_9_OPCODE_Y0:
+        srcb = get_SrcB_Y0(bundle);
+        return gen_rrr_opcode(dc, OE(opc, ext, Y0), dest, srca, srcb);
+
+    case SHIFT_OPCODE_Y0:
+        ext = get_ShiftOpcodeExtension_Y0(bundle);
+        imm = get_ShAmt_Y0(bundle);
+        return gen_rri_opcode(dc, OE(opc, ext, Y0), dest, srca, imm);
+
+    case ADDI_OPCODE_Y0:
+    case ADDXI_OPCODE_Y0:
+    case ANDI_OPCODE_Y0:
+    case CMPEQI_OPCODE_Y0:
+    case CMPLTSI_OPCODE_Y0:
+        imm = (int8_t)get_Imm8_Y0(bundle);
+        return gen_rri_opcode(dc, OE(opc, 0, Y0), dest, srca, imm);
+
+    default:
+        return TILEGX_EXCP_OPCODE_UNKNOWN;
+    }
+}
+
+static TileExcp decode_y1(DisasContext *dc, tilegx_bundle_bits bundle)
+{
+    unsigned opc = get_Opcode_Y1(bundle);
+    unsigned ext = get_RRROpcodeExtension_Y1(bundle);
+    unsigned dest = get_Dest_Y1(bundle);
+    unsigned srca = get_SrcA_Y1(bundle);
+    unsigned srcb;
+    int imm;
+
+    switch (get_Opcode_Y1(bundle)) {
+    case RRR_1_OPCODE_Y1:
+        if (ext == UNARY_RRR_1_OPCODE_Y0) {
+            ext = get_UnaryOpcodeExtension_Y1(bundle);
+            return gen_rr_opcode(dc, OE(opc, ext, Y1), dest, srca, bundle);
+        }
+        /* fallthru */
+    case RRR_0_OPCODE_Y1:
+    case RRR_2_OPCODE_Y1:
+    case RRR_3_OPCODE_Y1:
+    case RRR_4_OPCODE_Y1:
+    case RRR_5_OPCODE_Y1:
+    case RRR_6_OPCODE_Y1:
+    case RRR_7_OPCODE_Y1:
+        srcb = get_SrcB_Y1(bundle);
+        return gen_rrr_opcode(dc, OE(opc, ext, Y1), dest, srca, srcb);
+
+    case SHIFT_OPCODE_Y1:
+        ext = get_ShiftOpcodeExtension_Y1(bundle);
+        imm = get_ShAmt_Y1(bundle);
+        return gen_rri_opcode(dc, OE(opc, ext, Y1), dest, srca, imm);
+
+    case ADDI_OPCODE_Y1:
+    case ADDXI_OPCODE_Y1:
+    case ANDI_OPCODE_Y1:
+    case CMPEQI_OPCODE_Y1:
+    case CMPLTSI_OPCODE_Y1:
+        imm = (int8_t)get_Imm8_Y1(bundle);
+        return gen_rri_opcode(dc, OE(opc, 0, Y1), dest, srca, imm);
+
+    default:
+        return TILEGX_EXCP_OPCODE_UNKNOWN;
+    }
+}
+
+static TileExcp decode_y2(DisasContext *dc, tilegx_bundle_bits bundle)
+{
+    unsigned mode = get_Mode(bundle);
+    unsigned opc = get_Opcode_Y2(bundle);
+    unsigned srca = get_SrcA_Y2(bundle);
+    unsigned srcbdest = get_SrcBDest_Y2(bundle);
+    const char *mnemonic;
+    TCGMemOp memop;
+    bool prefetch_nofault = false;
+
+    switch (OEY2(opc, mode)) {
+    case OEY2(LD1S_OPCODE_Y2, MODE_OPCODE_YA2):
+        memop = MO_SB;
+        mnemonic = "ld1s"; /* prefetch_l1_fault */
+        goto do_load;
+    case OEY2(LD1U_OPCODE_Y2, MODE_OPCODE_YA2):
+        memop = MO_UB;
+        mnemonic = "ld1u"; /* prefetch, prefetch_l1 */
+        prefetch_nofault = (srcbdest == TILEGX_R_ZERO);
+        goto do_load;
+    case OEY2(LD2S_OPCODE_Y2, MODE_OPCODE_YA2):
+        memop = MO_TESW;
+        mnemonic = "ld2s"; /* prefetch_l2_fault */
+        goto do_load;
+    case OEY2(LD2U_OPCODE_Y2, MODE_OPCODE_YA2):
+        memop = MO_TEUW;
+        mnemonic = "ld2u"; /* prefetch_l2 */
+        prefetch_nofault = (srcbdest == TILEGX_R_ZERO);
+        goto do_load;
+    case OEY2(LD4S_OPCODE_Y2, MODE_OPCODE_YB2):
+        memop = MO_TESL;
+        mnemonic = "ld4s"; /* prefetch_l3_fault */
+        goto do_load;
+    case OEY2(LD4U_OPCODE_Y2, MODE_OPCODE_YB2):
+        memop = MO_TEUL;
+        mnemonic = "ld4u"; /* prefetch_l3 */
+        prefetch_nofault = (srcbdest == TILEGX_R_ZERO);
+        goto do_load;
+    case OEY2(LD_OPCODE_Y2, MODE_OPCODE_YB2):
+        memop = MO_TEQ;
+        mnemonic = "ld";
+    do_load:
+        if (!prefetch_nofault) {
+            tcg_gen_qemu_ld_tl(dest_gr(dc, srcbdest), load_gr(dc, srca),
+                               dc->mmuidx, memop);
+        }
+        qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s", mnemonic,
+                      reg_names[srcbdest], reg_names[srca]);
+        return TILEGX_EXCP_NONE;
+
+    case OEY2(ST1_OPCODE_Y2, MODE_OPCODE_YC2):
+        return gen_st_opcode(dc, 0, srca, srcbdest, MO_UB, "st1");
+    case OEY2(ST2_OPCODE_Y2, MODE_OPCODE_YC2):
+        return gen_st_opcode(dc, 0, srca, srcbdest, MO_TEUW, "st2");
+    case OEY2(ST4_OPCODE_Y2, MODE_OPCODE_YC2):
+        return gen_st_opcode(dc, 0, srca, srcbdest, MO_TEUL, "st4");
+    case OEY2(ST_OPCODE_Y2, MODE_OPCODE_YC2):
+        return gen_st_opcode(dc, 0, srca, srcbdest, MO_TEQ, "st");
+
+    default:
+        return TILEGX_EXCP_OPCODE_UNKNOWN;
+    }
+}
+
+static TileExcp decode_x0(DisasContext *dc, tilegx_bundle_bits bundle)
+{
+    unsigned opc = get_Opcode_X0(bundle);
+    unsigned dest = get_Dest_X0(bundle);
+    unsigned srca = get_SrcA_X0(bundle);
+    unsigned ext, srcb, bfs, bfe;
+    int imm;
+
+    switch (opc) {
+    case RRR_0_OPCODE_X0:
+        ext = get_RRROpcodeExtension_X0(bundle);
+        if (ext == UNARY_RRR_0_OPCODE_X0) {
+            ext = get_UnaryOpcodeExtension_X0(bundle);
+            return gen_rr_opcode(dc, OE(opc, ext, X0), dest, srca, bundle);
+        }
+        srcb = get_SrcB_X0(bundle);
+        return gen_rrr_opcode(dc, OE(opc, ext, X0), dest, srca, srcb);
+
+    case SHIFT_OPCODE_X0:
+        ext = get_ShiftOpcodeExtension_X0(bundle);
+        imm = get_ShAmt_X0(bundle);
+        return gen_rri_opcode(dc, OE(opc, ext, X0), dest, srca, imm);
+
+    case IMM8_OPCODE_X0:
+        ext = get_Imm8OpcodeExtension_X0(bundle);
+        imm = (int8_t)get_Imm8_X0(bundle);
+        return gen_rri_opcode(dc, OE(opc, ext, X0), dest, srca, imm);
+
+    case BF_OPCODE_X0:
+        ext = get_BFOpcodeExtension_X0(bundle);
+        bfs = get_BFStart_X0(bundle);
+        bfe = get_BFEnd_X0(bundle);
+        return gen_bf_opcode_x0(dc, ext, dest, srca, bfs, bfe);
+
+    case ADDLI_OPCODE_X0:
+    case SHL16INSLI_OPCODE_X0:
+    case ADDXLI_OPCODE_X0:
+        imm = (int16_t)get_Imm16_X0(bundle);
+        return gen_rri_opcode(dc, OE(opc, 0, X0), dest, srca, imm);
+
+    default:
+        return TILEGX_EXCP_OPCODE_UNKNOWN;
+    }
+}
+
+static TileExcp decode_x1(DisasContext *dc, tilegx_bundle_bits bundle)
+{
+    unsigned opc = get_Opcode_X1(bundle);
+    unsigned dest = get_Dest_X1(bundle);
+    unsigned srca = get_SrcA_X1(bundle);
+    unsigned ext, srcb;
+    int imm;
+
+    switch (opc) {
+    case RRR_0_OPCODE_X1:
+        ext = get_RRROpcodeExtension_X1(bundle);
+        srcb = get_SrcB_X1(bundle);
+        switch (ext) {
+        case UNARY_RRR_0_OPCODE_X1:
+            ext = get_UnaryOpcodeExtension_X1(bundle);
+            return gen_rr_opcode(dc, OE(opc, ext, X1), dest, srca, bundle);
+        case ST1_RRR_0_OPCODE_X1:
+            return gen_st_opcode(dc, dest, srca, srcb, MO_UB, "st1");
+        case ST2_RRR_0_OPCODE_X1:
+            return gen_st_opcode(dc, dest, srca, srcb, MO_TEUW, "st2");
+        case ST4_RRR_0_OPCODE_X1:
+            return gen_st_opcode(dc, dest, srca, srcb, MO_TEUL, "st4");
+        case STNT1_RRR_0_OPCODE_X1:
+            return gen_st_opcode(dc, dest, srca, srcb, MO_UB, "stnt1");
+        case STNT2_RRR_0_OPCODE_X1:
+            return gen_st_opcode(dc, dest, srca, srcb, MO_TEUW, "stnt2");
+        case STNT4_RRR_0_OPCODE_X1:
+            return gen_st_opcode(dc, dest, srca, srcb, MO_TEUL, "stnt4");
+        case STNT_RRR_0_OPCODE_X1:
+            return gen_st_opcode(dc, dest, srca, srcb, MO_TEQ, "stnt");
+        case ST_RRR_0_OPCODE_X1:
+            return gen_st_opcode(dc, dest, srca, srcb, MO_TEQ, "st");
+        }
+        return gen_rrr_opcode(dc, OE(opc, ext, X1), dest, srca, srcb);
+
+    case SHIFT_OPCODE_X1:
+        ext = get_ShiftOpcodeExtension_X1(bundle);
+        imm = get_ShAmt_X1(bundle);
+        return gen_rri_opcode(dc, OE(opc, ext, X1), dest, srca, imm);
+
+    case IMM8_OPCODE_X1:
+        ext = get_Imm8OpcodeExtension_X1(bundle);
+        imm = (int8_t)get_Dest_Imm8_X1(bundle);
+        srcb = get_SrcB_X1(bundle);
+        switch (ext) {
+        case ST1_ADD_IMM8_OPCODE_X1:
+            return gen_st_add_opcode(dc, srca, srcb, imm, MO_UB, "st1_add");
+        case ST2_ADD_IMM8_OPCODE_X1:
+            return gen_st_add_opcode(dc, srca, srcb, imm, MO_TEUW, "st2_add");
+        case ST4_ADD_IMM8_OPCODE_X1:
+            return gen_st_add_opcode(dc, srca, srcb, imm, MO_TEUL, "st4_add");
+        case STNT1_ADD_IMM8_OPCODE_X1:
+            return gen_st_add_opcode(dc, srca, srcb, imm, MO_UB, "stnt1_add");
+        case STNT2_ADD_IMM8_OPCODE_X1:
+            return gen_st_add_opcode(dc, srca, srcb, imm, MO_TEUW, "stnt2_add");
+        case STNT4_ADD_IMM8_OPCODE_X1:
+            return gen_st_add_opcode(dc, srca, srcb, imm, MO_TEUL, "stnt4_add");
+        case STNT_ADD_IMM8_OPCODE_X1:
+            return gen_st_add_opcode(dc, srca, srcb, imm, MO_TEQ, "stnt_add");
+        case ST_ADD_IMM8_OPCODE_X1:
+            return gen_st_add_opcode(dc, srca, srcb, imm, MO_TEQ, "st_add");
+        case MFSPR_IMM8_OPCODE_X1:
+            return gen_mfspr_x1(dc, dest, get_MF_Imm14_X1(bundle));
+        case MTSPR_IMM8_OPCODE_X1:
+            return gen_mtspr_x1(dc, get_MT_Imm14_X1(bundle), srca);
+        }
+        imm = (int8_t)get_Imm8_X1(bundle);
+        return gen_rri_opcode(dc, OE(opc, ext, X1), dest, srca, imm);
+
+    case BRANCH_OPCODE_X1:
+        ext = get_BrType_X1(bundle);
+        imm = sextract32(get_BrOff_X1(bundle), 0, 17);
+        return gen_branch_opcode_x1(dc, ext, srca, imm);
+
+    case JUMP_OPCODE_X1:
+        ext = get_JumpOpcodeExtension_X1(bundle);
+        imm = sextract32(get_JumpOff_X1(bundle), 0, 27);
+        return gen_jump_opcode_x1(dc, ext, imm);
+
+    case ADDLI_OPCODE_X1:
+    case SHL16INSLI_OPCODE_X1:
+    case ADDXLI_OPCODE_X1:
+        imm = (int16_t)get_Imm16_X1(bundle);
+        return gen_rri_opcode(dc, OE(opc, 0, X1), dest, srca, imm);
+
+    default:
+        return TILEGX_EXCP_OPCODE_UNKNOWN;
+    }
+}
+
+static void notice_excp(DisasContext *dc, uint64_t bundle,
+                        const char *type, TileExcp excp)
+{
+    if (likely(excp == TILEGX_EXCP_NONE)) {
+        return;
+    }
+    gen_exception(dc, excp);
+    switch (excp) {
+    case TILEGX_EXCP_OPCODE_UNIMPLEMENTED:
+        qemu_log_mask(LOG_UNIMP, "UNIMP %s, [" FMT64X "]\n", type, bundle);
+        break;
+    case TILEGX_EXCP_OPCODE_UNKNOWN:
+        qemu_log_mask(LOG_UNIMP, "UNKNOWN %s, [" FMT64X "]\n", type, bundle);
+        break;
+    default:
+        break;
+    }
+}
+
+static void translate_one_bundle(DisasContext *dc, uint64_t bundle)
+{
+    int i;
+
+    for (i = 0; i < ARRAY_SIZE(dc->wb); i++) {
+        DisasContextTemp *wb = &dc->wb[i];
+        wb->reg = TILEGX_R_NOREG;
+        TCGV_UNUSED_I64(wb->val);
+    }
+    dc->num_wb = 0;
+
+    qemu_log_mask(CPU_LOG_TB_IN_ASM, "  %" PRIx64 ":  { ", dc->pc);
+    if (get_Mode(bundle)) {
+        notice_excp(dc, bundle, "y0", decode_y0(dc, bundle));
+        qemu_log_mask(CPU_LOG_TB_IN_ASM, " ; ");
+        notice_excp(dc, bundle, "y1", decode_y1(dc, bundle));
+        qemu_log_mask(CPU_LOG_TB_IN_ASM, " ; ");
+        notice_excp(dc, bundle, "y2", decode_y2(dc, bundle));
+    } else {
+        notice_excp(dc, bundle, "x0", decode_x0(dc, bundle));
+        qemu_log_mask(CPU_LOG_TB_IN_ASM, " ; ");
+        notice_excp(dc, bundle, "x1", decode_x1(dc, bundle));
+    }
+    qemu_log_mask(CPU_LOG_TB_IN_ASM, " }\n");
+
+    for (i = dc->num_wb - 1; i >= 0; --i) {
+        DisasContextTemp *wb = &dc->wb[i];
+        if (wb->reg < TILEGX_R_COUNT) {
+            tcg_gen_mov_i64(cpu_regs[wb->reg], wb->val);
+        }
+        tcg_temp_free_i64(wb->val);
+    }
+
+    if (dc->jmp.cond != TCG_COND_NEVER) {
+        if (dc->jmp.cond == TCG_COND_ALWAYS) {
+            tcg_gen_mov_i64(cpu_pc, dc->jmp.dest);
+        } else {
+            TCGv next = tcg_const_i64(dc->pc + TILEGX_BUNDLE_SIZE_IN_BYTES);
+            tcg_gen_movcond_i64(dc->jmp.cond, cpu_pc,
+                                dc->jmp.val1, load_zero(dc),
+                                dc->jmp.dest, next);
+            tcg_temp_free_i64(dc->jmp.val1);
+            tcg_temp_free_i64(next);
+        }
+        tcg_temp_free_i64(dc->jmp.dest);
+        tcg_gen_exit_tb(0);
+        dc->exit_tb = true;
+    } else if (dc->atomic_excp != TILEGX_EXCP_NONE) {
+        gen_exception(dc, dc->atomic_excp);
+    }
+}
+
+void gen_intermediate_code(CPUTLGState *env, struct TranslationBlock *tb)
+{
+    TileGXCPU *cpu = tilegx_env_get_cpu(env);
+    DisasContext ctx;
+    DisasContext *dc = &ctx;
+    CPUState *cs = CPU(cpu);
+    uint64_t pc_start = tb->pc;
+    uint64_t next_page_start = (pc_start & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE;
+    int num_insns = 0;
+    int max_insns = tb->cflags & CF_COUNT_MASK;
+
+    dc->pc = pc_start;
+    dc->mmuidx = 0;
+    dc->exit_tb = false;
+    dc->atomic_excp = TILEGX_EXCP_NONE;
+    dc->jmp.cond = TCG_COND_NEVER;
+    TCGV_UNUSED_I64(dc->jmp.dest);
+    TCGV_UNUSED_I64(dc->jmp.val1);
+    TCGV_UNUSED_I64(dc->zero);
+
+    if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
+        qemu_log("IN: %s\n", lookup_symbol(pc_start));
+    }
+    if (!max_insns) {
+        max_insns = CF_COUNT_MASK;
+    }
+    if (cs->singlestep_enabled || singlestep) {
+        max_insns = 1;
+    }
+    if (max_insns > TCG_MAX_INSNS) {
+        max_insns = TCG_MAX_INSNS;
+    }
+    gen_tb_start(tb);
+
+    while (1) {
+        tcg_gen_insn_start(dc->pc);
+        num_insns++;
+
+        translate_one_bundle(dc, cpu_ldq_data(env, dc->pc));
+
+        if (dc->exit_tb) {
+            /* PC updated and EXIT_TB/GOTO_TB/exception emitted.  */
+            break;
+        }
+        dc->pc += TILEGX_BUNDLE_SIZE_IN_BYTES;
+        if (num_insns >= max_insns
+            || dc->pc >= next_page_start
+            || tcg_op_buf_full()) {
+            /* Ending the TB due to TB size or page boundary.  Set PC.  */
+            tcg_gen_movi_tl(cpu_pc, dc->pc);
+            tcg_gen_exit_tb(0);
+            break;
+        }
+    }
+
+    gen_tb_end(tb, num_insns);
+    tb->size = dc->pc - pc_start;
+    tb->icount = num_insns;
+
+    qemu_log_mask(CPU_LOG_TB_IN_ASM, "\n");
+}
+
+void restore_state_to_opc(CPUTLGState *env, TranslationBlock *tb,
+                          target_ulong *data)
+{
+    env->pc = data[0];
+}
+
+void tilegx_tcg_init(void)
+{
+    int i;
+
+    cpu_env = tcg_global_reg_new_ptr(TCG_AREG0, "env");
+    cpu_pc = tcg_global_mem_new_i64(TCG_AREG0, offsetof(CPUTLGState, pc), "pc");
+    for (i = 0; i < TILEGX_R_COUNT; i++) {
+        cpu_regs[i] = tcg_global_mem_new_i64(TCG_AREG0,
+                                             offsetof(CPUTLGState, regs[i]),
+                                             reg_names[i]);
+    }
+}
author	Timothy Pearson <tpearson@raptorengineering.com>	2019-05-11 15:12:49 -0500
committer	Timothy Pearson <tpearson@raptorengineering.com>	2019-05-11 15:12:49 -0500
commit	9e80202352dd49bdd9e67b8b906d86f058431505 (patch)
tree	5673c17aad6e3833da8c4ff21b5a11f666ec9fbe /src/target-tilegx
download	hqemu-9e80202352dd49bdd9e67b8b906d86f058431505.zip hqemu-9e80202352dd49bdd9e67b8b906d86f058431505.tar.gz