Merge remote-tracking branch 'remotes/rth-gitlab/tags/pull-tcg-20210113' into staging

Improvements to tcg constant handling.
Force utf8 for decodetree.

# gpg: Signature made Thu 14 Jan 2021 02:15:42 GMT
# gpg:                using RSA key 7A481E78868B4DB6A85A05C064DF38E8AF7E215F
# gpg:                issuer "richard.henderson@linaro.org"
# gpg: Good signature from "Richard Henderson <richard.henderson@linaro.org>" [full]
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A  05C0 64DF 38E8 AF7E 215F

* remotes/rth-gitlab/tags/pull-tcg-20210113: (24 commits)
  decodetree: Open files with encoding='utf-8'
  tcg/aarch64: Use tcg_constant_vec with tcg vec expanders
  tcg/ppc: Use tcg_constant_vec with tcg vec expanders
  tcg: Remove tcg_gen_dup{8,16,32,64}i_vec
  tcg/i386: Use tcg_constant_vec with tcg vec expanders
  tcg: Add tcg_reg_alloc_dup2
  tcg: Remove movi and dupi opcodes
  tcg/tci: Add special tci_movi_{i32,i64} opcodes
  tcg: Use tcg_constant_{i32,i64,vec} with gvec expanders
  tcg: Use tcg_constant_{i32,i64} with tcg plugins
  tcg: Use tcg_constant_{i32,i64} with tcg int expanders
  tcg: Use tcg_constant_i32 with icount expander
  tcg: Convert tcg_gen_dupi_vec to TCG_CONST
  tcg/optimize: Use tcg_constant_internal with constant folding
  tcg/optimize: Adjust TempOptInfo allocation
  tcg/optimize: Improve find_better_copy
  tcg: Introduce TYPE_CONST temporaries
  tcg: Expand TempOptInfo to 64-bits
  tcg: Rename struct tcg_temp_info to TempOptInfo
  tcg: Expand TCGTemp.val to 64-bits
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

21 files changed, 889 insertions, 668 deletions

diff --git a/accel/tcg/plugin-gen.c b/accel/tcg/plugin-gen.c
index 51580d51a0..e5dc9d0ca9 100644
--- a/accel/tcg/plugin-gen.c
+++ b/accel/tcg/plugin-gen.c
@@ -284,8 +284,8 @@ static TCGOp *copy_extu_i32_i64(TCGOp **begin_op, TCGOp *op)
     if (TCG_TARGET_REG_BITS == 32) {
         /* mov_i32 */
         op = copy_op(begin_op, op, INDEX_op_mov_i32);
-        /* movi_i32 */
-        op = copy_op(begin_op, op, INDEX_op_movi_i32);
+        /* mov_i32 w/ $0 */
+        op = copy_op(begin_op, op, INDEX_op_mov_i32);
     } else {
         /* extu_i32_i64 */
         op = copy_op(begin_op, op, INDEX_op_extu_i32_i64);
@@ -306,39 +306,34 @@ static TCGOp *copy_mov_i64(TCGOp **begin_op, TCGOp *op)
     return op;
 }
 
-static TCGOp *copy_movi_i64(TCGOp **begin_op, TCGOp *op, uint64_t v)
-{
-    if (TCG_TARGET_REG_BITS == 32) {
-        /* 2x movi_i32 */
-        op = copy_op(begin_op, op, INDEX_op_movi_i32);
-        op->args[1] = v;
-
-        op = copy_op(begin_op, op, INDEX_op_movi_i32);
-        op->args[1] = v >> 32;
-    } else {
-        /* movi_i64 */
-        op = copy_op(begin_op, op, INDEX_op_movi_i64);
-        op->args[1] = v;
-    }
-    return op;
-}
-
 static TCGOp *copy_const_ptr(TCGOp **begin_op, TCGOp *op, void *ptr)
 {
     if (UINTPTR_MAX == UINT32_MAX) {
-        /* movi_i32 */
-        op = copy_op(begin_op, op, INDEX_op_movi_i32);
-        op->args[1] = (uintptr_t)ptr;
+        /* mov_i32 */
+        op = copy_op(begin_op, op, INDEX_op_mov_i32);
+        op->args[1] = tcgv_i32_arg(tcg_constant_i32((uintptr_t)ptr));
     } else {
-        /* movi_i64 */
-        op = copy_movi_i64(begin_op, op, (uint64_t)(uintptr_t)ptr);
+        /* mov_i64 */
+        op = copy_op(begin_op, op, INDEX_op_mov_i64);
+        op->args[1] = tcgv_i64_arg(tcg_constant_i64((uintptr_t)ptr));
     }
     return op;
 }
 
 static TCGOp *copy_const_i64(TCGOp **begin_op, TCGOp *op, uint64_t v)
 {
-    return copy_movi_i64(begin_op, op, v);
+    if (TCG_TARGET_REG_BITS == 32) {
+        /* 2x mov_i32 */
+        op = copy_op(begin_op, op, INDEX_op_mov_i32);
+        op->args[1] = tcgv_i32_arg(tcg_constant_i32(v));
+        op = copy_op(begin_op, op, INDEX_op_mov_i32);
+        op->args[1] = tcgv_i32_arg(tcg_constant_i32(v >> 32));
+    } else {
+        /* mov_i64 */
+        op = copy_op(begin_op, op, INDEX_op_mov_i64);
+        op->args[1] = tcgv_i64_arg(tcg_constant_i64(v));
+    }
+    return op;
 }
 
 static TCGOp *copy_extu_tl_i64(TCGOp **begin_op, TCGOp *op)
@@ -486,8 +481,8 @@ static TCGOp *append_mem_cb(const struct qemu_plugin_dyn_cb *cb,
 
     tcg_debug_assert(type == PLUGIN_GEN_CB_MEM);
 
-    /* const_i32 == movi_i32 ("info", so it remains as is) */
-    op = copy_op(&begin_op, op, INDEX_op_movi_i32);
+    /* const_i32 == mov_i32 ("info", so it remains as is) */
+    op = copy_op(&begin_op, op, INDEX_op_mov_i32);
 
     /* const_ptr */
     op = copy_const_ptr(&begin_op, op, cb->userp);
diff --git a/include/exec/gen-icount.h b/include/exec/gen-icount.h
index aa4b44354a..298e01eef4 100644
--- a/include/exec/gen-icount.h
+++ b/include/exec/gen-icount.h
@@ -34,7 +34,7 @@ static inline void gen_io_end(void)
 
 static inline void gen_tb_start(const TranslationBlock *tb)
 {
-    TCGv_i32 count, imm;
+    TCGv_i32 count;
 
     tcg_ctx->exitreq_label = gen_new_label();
     if (tb_cflags(tb) & CF_USE_ICOUNT) {
@@ -48,15 +48,13 @@ static inline void gen_tb_start(const TranslationBlock *tb)
                    offsetof(ArchCPU, env));
 
     if (tb_cflags(tb) & CF_USE_ICOUNT) {
-        imm = tcg_temp_new_i32();
-        /* We emit a movi with a dummy immediate argument. Keep the insn index
-         * of the movi so that we later (when we know the actual insn count)
-         * can update the immediate argument with the actual insn count.  */
-        tcg_gen_movi_i32(imm, 0xdeadbeef);
+        /*
+         * We emit a sub with a dummy immediate argument. Keep the insn index
+         * of the sub so that we later (when we know the actual insn count)
+         * can update the argument with the actual insn count.
+         */
+        tcg_gen_sub_i32(count, count, tcg_constant_i32(0));
         icount_start_insn = tcg_last_op();
-
-        tcg_gen_sub_i32(count, count, imm);
-        tcg_temp_free_i32(imm);
     }
 
     tcg_gen_brcondi_i32(TCG_COND_LT, count, 0, tcg_ctx->exitreq_label);
@@ -74,9 +72,12 @@ static inline void gen_tb_start(const TranslationBlock *tb)
 static inline void gen_tb_end(const TranslationBlock *tb, int num_insns)
 {
     if (tb_cflags(tb) & CF_USE_ICOUNT) {
-        /* Update the num_insn immediate parameter now that we know
-         * the actual insn count.  */
-        tcg_set_insn_param(icount_start_insn, 1, num_insns);
+        /*
+         * Update the num_insn immediate parameter now that we know
+         * the actual insn count.
+         */
+        tcg_set_insn_param(icount_start_insn, 2,
+                           tcgv_i32_arg(tcg_constant_i32(num_insns)));
     }
 
     gen_set_label(tcg_ctx->exitreq_label);
diff --git a/include/tcg/tcg-op.h b/include/tcg/tcg-op.h
index 901b19f32a..2cd1faf9c4 100644
--- a/include/tcg/tcg-op.h
+++ b/include/tcg/tcg-op.h
@@ -271,6 +271,7 @@ void tcg_gen_mb(TCGBar);
 
 /* 32 bit ops */
 
+void tcg_gen_movi_i32(TCGv_i32 ret, int32_t arg);
 void tcg_gen_addi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2);
 void tcg_gen_subfi_i32(TCGv_i32 ret, int32_t arg1, TCGv_i32 arg2);
 void tcg_gen_subi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2);
@@ -349,11 +350,6 @@ static inline void tcg_gen_mov_i32(TCGv_i32 ret, TCGv_i32 arg)
     }
 }
 
-static inline void tcg_gen_movi_i32(TCGv_i32 ret, int32_t arg)
-{
-    tcg_gen_op2i_i32(INDEX_op_movi_i32, ret, arg);
-}
-
 static inline void tcg_gen_ld8u_i32(TCGv_i32 ret, TCGv_ptr arg2,
                                     tcg_target_long offset)
 {
@@ -467,6 +463,7 @@ static inline void tcg_gen_not_i32(TCGv_i32 ret, TCGv_i32 arg)
 
 /* 64 bit ops */
 
+void tcg_gen_movi_i64(TCGv_i64 ret, int64_t arg);
 void tcg_gen_addi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2);
 void tcg_gen_subfi_i64(TCGv_i64 ret, int64_t arg1, TCGv_i64 arg2);
 void tcg_gen_subi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2);
@@ -550,11 +547,6 @@ static inline void tcg_gen_mov_i64(TCGv_i64 ret, TCGv_i64 arg)
     }
 }
 
-static inline void tcg_gen_movi_i64(TCGv_i64 ret, int64_t arg)
-{
-    tcg_gen_op2i_i64(INDEX_op_movi_i64, ret, arg);
-}
-
 static inline void tcg_gen_ld8u_i64(TCGv_i64 ret, TCGv_ptr arg2,
                                     tcg_target_long offset)
 {
@@ -698,7 +690,6 @@ static inline void tcg_gen_sub_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2)
 
 void tcg_gen_discard_i64(TCGv_i64 arg);
 void tcg_gen_mov_i64(TCGv_i64 ret, TCGv_i64 arg);
-void tcg_gen_movi_i64(TCGv_i64 ret, int64_t arg);
 void tcg_gen_ld8u_i64(TCGv_i64 ret, TCGv_ptr arg2, tcg_target_long offset);
 void tcg_gen_ld8s_i64(TCGv_i64 ret, TCGv_ptr arg2, tcg_target_long offset);
 void tcg_gen_ld16u_i64(TCGv_i64 ret, TCGv_ptr arg2, tcg_target_long offset);
@@ -968,10 +959,6 @@ void tcg_gen_mov_vec(TCGv_vec, TCGv_vec);
 void tcg_gen_dup_i32_vec(unsigned vece, TCGv_vec, TCGv_i32);
 void tcg_gen_dup_i64_vec(unsigned vece, TCGv_vec, TCGv_i64);
 void tcg_gen_dup_mem_vec(unsigned vece, TCGv_vec, TCGv_ptr, tcg_target_long);
-void tcg_gen_dup8i_vec(TCGv_vec, uint32_t);
-void tcg_gen_dup16i_vec(TCGv_vec, uint32_t);
-void tcg_gen_dup32i_vec(TCGv_vec, uint32_t);
-void tcg_gen_dup64i_vec(TCGv_vec, uint64_t);
 void tcg_gen_dupi_vec(unsigned vece, TCGv_vec, uint64_t);
 void tcg_gen_add_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b);
 void tcg_gen_sub_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b);
diff --git a/include/tcg/tcg-opc.h b/include/tcg/tcg-opc.h
index 70a76646c4..900984c005 100644
--- a/include/tcg/tcg-opc.h
+++ b/include/tcg/tcg-opc.h
@@ -45,7 +45,6 @@ DEF(br, 0, 0, 1, TCG_OPF_BB_END)
 DEF(mb, 0, 0, 1, 0)
 
 DEF(mov_i32, 1, 1, 0, TCG_OPF_NOT_PRESENT)
-DEF(movi_i32, 1, 0, 1, TCG_OPF_NOT_PRESENT)
 DEF(setcond_i32, 1, 2, 1, 0)
 DEF(movcond_i32, 1, 4, 1, IMPL(TCG_TARGET_HAS_movcond_i32))
 /* load/store */
@@ -111,7 +110,6 @@ DEF(ctz_i32, 1, 2, 0, IMPL(TCG_TARGET_HAS_ctz_i32))
 DEF(ctpop_i32, 1, 1, 0, IMPL(TCG_TARGET_HAS_ctpop_i32))
 
 DEF(mov_i64, 1, 1, 0, TCG_OPF_64BIT | TCG_OPF_NOT_PRESENT)
-DEF(movi_i64, 1, 0, 1, TCG_OPF_64BIT | TCG_OPF_NOT_PRESENT)
 DEF(setcond_i64, 1, 2, 1, IMPL64)
 DEF(movcond_i64, 1, 4, 1, IMPL64 | IMPL(TCG_TARGET_HAS_movcond_i64))
 /* load/store */
@@ -221,7 +219,6 @@ DEF(qemu_st8_i32, 0, TLADDR_ARGS + 1, 1,
 #define IMPLVEC  TCG_OPF_VECTOR | IMPL(TCG_TARGET_MAYBE_vec)
 
 DEF(mov_vec, 1, 1, 0, TCG_OPF_VECTOR | TCG_OPF_NOT_PRESENT)
-DEF(dupi_vec, 1, 0, 1, TCG_OPF_VECTOR | TCG_OPF_NOT_PRESENT)
 
 DEF(dup_vec, 1, 1, 0, IMPLVEC)
 DEF(dup2_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_REG_BITS == 32))
@@ -278,6 +275,14 @@ DEF(last_generic, 0, 0, 0, TCG_OPF_NOT_PRESENT)
 #include "tcg-target.opc.h"
 #endif
 
+#ifdef TCG_TARGET_INTERPRETER
+/* These opcodes are only for use between the tci generator and interpreter. */
+DEF(tci_movi_i32, 1, 0, 1, TCG_OPF_NOT_PRESENT)
+#if TCG_TARGET_REG_BITS == 64
+DEF(tci_movi_i64, 1, 0, 1, TCG_OPF_64BIT | TCG_OPF_NOT_PRESENT)
+#endif
+#endif
+
 #undef TLADDR_ARGS
 #undef DATA64_ARGS
 #undef IMPL
diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index 95fe5604eb..504c5e9bb0 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -483,26 +483,32 @@ typedef enum TCGTempVal {
     TEMP_VAL_CONST,
 } TCGTempVal;
 
+typedef enum TCGTempKind {
+    /* Temp is dead at the end of all basic blocks. */
+    TEMP_NORMAL,
+    /* Temp is saved across basic blocks but dead at the end of TBs. */
+    TEMP_LOCAL,
+    /* Temp is saved across both basic blocks and translation blocks. */
+    TEMP_GLOBAL,
+    /* Temp is in a fixed register. */
+    TEMP_FIXED,
+    /* Temp is a fixed constant. */
+    TEMP_CONST,
+} TCGTempKind;
+
 typedef struct TCGTemp {
     TCGReg reg:8;
     TCGTempVal val_type:8;
     TCGType base_type:8;
     TCGType type:8;
-    unsigned int fixed_reg:1;
+    TCGTempKind kind:3;
     unsigned int indirect_reg:1;
     unsigned int indirect_base:1;
     unsigned int mem_coherent:1;
     unsigned int mem_allocated:1;
-    /* If true, the temp is saved across both basic blocks and
-       translation blocks.  */
-    unsigned int temp_global:1;
-    /* If true, the temp is saved across basic blocks but dead
-       at the end of translation blocks.  If false, the temp is
-       dead at the end of basic blocks.  */
-    unsigned int temp_local:1;
     unsigned int temp_allocated:1;
 
-    tcg_target_long val;
+    int64_t val;
     struct TCGTemp *mem_base;
     intptr_t mem_offset;
     const char *name;
@@ -661,6 +667,7 @@ struct TCGContext {
     QSIMPLEQ_HEAD(, TCGOp) plugin_ops;
 #endif
 
+    GHashTable *const_table[TCG_TYPE_COUNT];
     TCGTempSet free_temps[TCG_TYPE_COUNT * 2];
     TCGTemp temps[TCG_MAX_TEMPS]; /* globals first, temps after */
 
@@ -675,6 +682,11 @@ struct TCGContext {
     target_ulong gen_insn_data[TCG_MAX_INSNS][TARGET_INSN_START_WORDS];
 };
 
+static inline bool temp_readonly(TCGTemp *ts)
+{
+    return ts->kind >= TEMP_FIXED;
+}
+
 extern TCGContext tcg_init_ctx;
 extern __thread TCGContext *tcg_ctx;
 extern const void *tcg_code_gen_epilogue;
@@ -1070,6 +1082,7 @@ TCGOp *tcg_op_insert_after(TCGContext *s, TCGOp *op, TCGOpcode opc);
 
 void tcg_optimize(TCGContext *s);
 
+/* Allocate a new temporary and initialize it with a constant. */
 TCGv_i32 tcg_const_i32(int32_t val);
 TCGv_i64 tcg_const_i64(int64_t val);
 TCGv_i32 tcg_const_local_i32(int32_t val);
@@ -1079,6 +1092,25 @@ TCGv_vec tcg_const_ones_vec(TCGType);
 TCGv_vec tcg_const_zeros_vec_matching(TCGv_vec);
 TCGv_vec tcg_const_ones_vec_matching(TCGv_vec);
 
+/*
+ * Locate or create a read-only temporary that is a constant.
+ * This kind of temporary need not and should not be freed.
+ */
+TCGTemp *tcg_constant_internal(TCGType type, int64_t val);
+
+static inline TCGv_i32 tcg_constant_i32(int32_t val)
+{
+    return temp_tcgv_i32(tcg_constant_internal(TCG_TYPE_I32, val));
+}
+
+static inline TCGv_i64 tcg_constant_i64(int64_t val)
+{
+    return temp_tcgv_i64(tcg_constant_internal(TCG_TYPE_I64, val));
+}
+
+TCGv_vec tcg_constant_vec(TCGType type, unsigned vece, int64_t val);
+TCGv_vec tcg_constant_vec_matching(TCGv_vec match, unsigned vece, int64_t val);
+
 #if UINTPTR_MAX == UINT32_MAX
 # define tcg_const_ptr(x)        ((TCGv_ptr)tcg_const_i32((intptr_t)(x)))
 # define tcg_const_local_ptr(x)  ((TCGv_ptr)tcg_const_local_i32((intptr_t)(x)))
diff --git a/scripts/decodetree.py b/scripts/decodetree.py
index 47aa9caf6d..4637b633e7 100644
--- a/scripts/decodetree.py
+++ b/scripts/decodetree.py
@@ -20,6 +20,7 @@
 # See the syntax and semantics in docs/devel/decodetree.rst.
 #
 
+import io
 import os
 import re
 import sys
@@ -1304,7 +1305,7 @@ def main():
 
     for filename in args:
         input_file = filename
-        f = open(filename, 'r')
+        f = open(filename, 'rt', encoding='utf-8')
         parse_file(f, toppat)
         f.close()
 
@@ -1324,9 +1325,11 @@ def main():
         prop_size(stree)
 
     if output_file:
-        output_fd = open(output_file, 'w')
+        output_fd = open(output_file, 'wt', encoding='utf-8')
     else:
-        output_fd = sys.stdout
+        output_fd = io.TextIOWrapper(sys.stdout.buffer,
+                                     encoding=sys.stdout.encoding,
+                                     errors="ignore")
 
     output_autogen()
     for n in sorted(arguments.keys()):
diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc
index ab199b143f..23954ec7cf 100644
--- a/tcg/aarch64/tcg-target.c.inc
+++ b/tcg/aarch64/tcg-target.c.inc
@@ -857,14 +857,14 @@ static void tcg_out_logicali(TCGContext *s, AArch64Insn insn, TCGType ext,
     tcg_out_insn_3404(s, insn, ext, rd, rn, ext, r, c);
 }
 
-static void tcg_out_dupi_vec(TCGContext *s, TCGType type,
-                             TCGReg rd, tcg_target_long v64)
+static void tcg_out_dupi_vec(TCGContext *s, TCGType type, unsigned vece,
+                             TCGReg rd, int64_t v64)
 {
     bool q = type == TCG_TYPE_V128;
     int cmode, imm8, i;
 
     /* Test all bytes equal first.  */
-    if (v64 == dup_const(MO_8, v64)) {
+    if (vece == MO_8) {
         imm8 = (uint8_t)v64;
         tcg_out_insn(s, 3606, MOVI, q, rd, 0, 0xe, imm8);
         return;
@@ -891,7 +891,7 @@ static void tcg_out_dupi_vec(TCGContext *s, TCGType type,
      * cannot find an expansion there's no point checking a larger
      * width because we already know by replication it cannot match.
      */
-    if (v64 == dup_const(MO_16, v64)) {
+    if (vece == MO_16) {
         uint16_t v16 = v64;
 
         if (is_shimm16(v16, &cmode, &imm8)) {
@@ -910,7 +910,7 @@ static void tcg_out_dupi_vec(TCGContext *s, TCGType type,
         tcg_out_insn(s, 3606, MOVI, q, rd, 0, 0x8, v16 & 0xff);
         tcg_out_insn(s, 3606, ORR, q, rd, 0, 0xa, v16 >> 8);
         return;
-    } else if (v64 == dup_const(MO_32, v64)) {
+    } else if (vece == MO_32) {
         uint32_t v32 = v64;
         uint32_t n32 = ~v32;
 
@@ -1011,13 +1011,6 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg rd,
     case TCG_TYPE_I64:
         tcg_debug_assert(rd < 32);
         break;
-
-    case TCG_TYPE_V64:
-    case TCG_TYPE_V128:
-        tcg_debug_assert(rd >= 32);
-        tcg_out_dupi_vec(s, type, rd, value);
-        return;
-
     default:
         g_assert_not_reached();
     }
@@ -2264,8 +2257,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
 
     case INDEX_op_mov_i32:  /* Always emitted via tcg_out_mov.  */
     case INDEX_op_mov_i64:
-    case INDEX_op_movi_i32: /* Always emitted via tcg_out_movi.  */
-    case INDEX_op_movi_i64:
     case INDEX_op_call:     /* Always emitted via tcg_out_call.  */
     default:
         g_assert_not_reached();
@@ -2442,7 +2433,7 @@ static void tcg_out_vec_op(TCGContext *s, TCGOpcode opc,
                         tcg_out_insn_3617(s, insn, is_q, vece, a0, a1);
                         break;
                     }
-                    tcg_out_dupi_vec(s, type, TCG_VEC_TMP, 0);
+                    tcg_out_dupi_vec(s, type, MO_8, TCG_VEC_TMP, 0);
                     a2 = TCG_VEC_TMP;
                 }
                 insn = cmp_insn[cond];
@@ -2473,7 +2464,6 @@ static void tcg_out_vec_op(TCGContext *s, TCGOpcode opc,
         break;
 
     case INDEX_op_mov_vec:  /* Always emitted via tcg_out_mov.  */
-    case INDEX_op_dupi_vec: /* Always emitted via tcg_out_movi.  */
     case INDEX_op_dup_vec:  /* Always emitted via tcg_out_dup_vec.  */
     default:
         g_assert_not_reached();
@@ -2526,7 +2516,7 @@ void tcg_expand_vec_op(TCGOpcode opc, TCGType type, unsigned vece,
                        TCGArg a0, ...)
 {
     va_list va;
-    TCGv_vec v0, v1, v2, t1, t2;
+    TCGv_vec v0, v1, v2, t1, t2, c1;
     TCGArg a2;
 
     va_start(va, a0);
@@ -2558,8 +2548,8 @@ void tcg_expand_vec_op(TCGOpcode opc, TCGType type, unsigned vece,
 
     case INDEX_op_rotlv_vec:
         t1 = tcg_temp_new_vec(type);
-        tcg_gen_dupi_vec(vece, t1, 8 << vece);
-        tcg_gen_sub_vec(vece, t1, v2, t1);
+        c1 = tcg_constant_vec(type, vece, 8 << vece);
+        tcg_gen_sub_vec(vece, t1, v2, c1);
         /* Right shifts are negative left shifts for AArch64.  */
         vec_gen_3(INDEX_op_shlv_vec, type, vece, tcgv_vec_arg(t1),
                   tcgv_vec_arg(v1), tcgv_vec_arg(t1));
@@ -2572,9 +2562,9 @@ void tcg_expand_vec_op(TCGOpcode opc, TCGType type, unsigned vece,
     case INDEX_op_rotrv_vec:
         t1 = tcg_temp_new_vec(type);
         t2 = tcg_temp_new_vec(type);
+        c1 = tcg_constant_vec(type, vece, 8 << vece);
         tcg_gen_neg_vec(vece, t1, v2);
-        tcg_gen_dupi_vec(vece, t2, 8 << vece);
-        tcg_gen_add_vec(vece, t2, t1, t2);
+        tcg_gen_sub_vec(vece, t2, c1, v2);
         /* Right shifts are negative left shifts for AArch64.  */
         vec_gen_3(INDEX_op_shlv_vec, type, vece, tcgv_vec_arg(t1),
                   tcgv_vec_arg(v1), tcgv_vec_arg(t1));
diff --git a/tcg/arm/tcg-target.c.inc b/tcg/arm/tcg-target.c.inc
index 0fd1126454..c2b26b3c45 100644
--- a/tcg/arm/tcg-target.c.inc
+++ b/tcg/arm/tcg-target.c.inc
@@ -2068,7 +2068,6 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
         break;
 
     case INDEX_op_mov_i32:  /* Always emitted via tcg_out_mov.  */
-    case INDEX_op_movi_i32: /* Always emitted via tcg_out_movi.  */
     case INDEX_op_call:     /* Always emitted via tcg_out_call.  */
     default:
         tcg_abort();
diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc
index 46e856f442..050f3cb0b1 100644
--- a/tcg/i386/tcg-target.c.inc
+++ b/tcg/i386/tcg-target.c.inc
@@ -942,8 +942,8 @@ static bool tcg_out_dupm_vec(TCGContext *s, TCGType type, unsigned vece,
     return true;
 }
 
-static void tcg_out_dupi_vec(TCGContext *s, TCGType type,
-                             TCGReg ret, tcg_target_long arg)
+static void tcg_out_dupi_vec(TCGContext *s, TCGType type, unsigned vece,
+                             TCGReg ret, int64_t arg)
 {
     int vex_l = (type == TCG_TYPE_V256 ? P_VEXL : 0);
 
@@ -956,7 +956,14 @@ static void tcg_out_dupi_vec(TCGContext *s, TCGType type,
         return;
     }
 
-    if (TCG_TARGET_REG_BITS == 64) {
+    if (TCG_TARGET_REG_BITS == 32 && vece < MO_64) {
+        if (have_avx2) {
+            tcg_out_vex_modrm_pool(s, OPC_VPBROADCASTD + vex_l, ret);
+        } else {
+            tcg_out_vex_modrm_pool(s, OPC_VBROADCASTSS, ret);
+        }
+        new_pool_label(s, arg, R_386_32, s->code_ptr - 4, 0);
+    } else {
         if (type == TCG_TYPE_V64) {
             tcg_out_vex_modrm_pool(s, OPC_MOVQ_VqWq, ret);
         } else if (have_avx2) {
@@ -964,40 +971,39 @@ static void tcg_out_dupi_vec(TCGContext *s, TCGType type,
         } else {
             tcg_out_vex_modrm_pool(s, OPC_MOVDDUP, ret);
         }
-        new_pool_label(s, arg, R_386_PC32, s->code_ptr - 4, -4);
-    } else {
-        if (have_avx2) {
-            tcg_out_vex_modrm_pool(s, OPC_VPBROADCASTD + vex_l, ret);
+        if (TCG_TARGET_REG_BITS == 64) {
+            new_pool_label(s, arg, R_386_PC32, s->code_ptr - 4, -4);
         } else {
-            tcg_out_vex_modrm_pool(s, OPC_VBROADCASTSS, ret);
+            new_pool_l2(s, R_386_32, s->code_ptr - 4, 0, arg, arg >> 32);
         }
-        new_pool_label(s, arg, R_386_32, s->code_ptr - 4, 0);
     }
 }
 
-static void tcg_out_movi(TCGContext *s, TCGType type,
-                         TCGReg ret, tcg_target_long arg)
+static void tcg_out_movi_vec(TCGContext *s, TCGType type,
+                             TCGReg ret, tcg_target_long arg)
 {
-    tcg_target_long diff;
-
-    switch (type) {
-    case TCG_TYPE_I32:
-#if TCG_TARGET_REG_BITS == 64
-    case TCG_TYPE_I64:
-#endif
-        if (ret < 16) {
-            break;
-        }
-        /* fallthru */
-    case TCG_TYPE_V64:
-    case TCG_TYPE_V128:
-    case TCG_TYPE_V256:
-        tcg_debug_assert(ret >= 16);
-        tcg_out_dupi_vec(s, type, ret, arg);
+    if (arg == 0) {
+        tcg_out_vex_modrm(s, OPC_PXOR, ret, ret, ret);
         return;
-    default:
-        g_assert_not_reached();
     }
+    if (arg == -1) {
+        tcg_out_vex_modrm(s, OPC_PCMPEQB, ret, ret, ret);
+        return;
+    }
+
+    int rexw = (type == TCG_TYPE_I32 ? 0 : P_REXW);
+    tcg_out_vex_modrm_pool(s, OPC_MOVD_VyEy + rexw, ret);
+    if (TCG_TARGET_REG_BITS == 64) {
+        new_pool_label(s, arg, R_386_PC32, s->code_ptr - 4, -4);
+    } else {
+        new_pool_label(s, arg, R_386_32, s->code_ptr - 4, 0);
+    }
+}
+
+static void tcg_out_movi_int(TCGContext *s, TCGType type,
+                             TCGReg ret, tcg_target_long arg)
+{
+    tcg_target_long diff;
 
     if (arg == 0) {
         tgen_arithr(s, ARITH_XOR, ret, ret);
@@ -1027,6 +1033,25 @@ static void tcg_out_movi(TCGContext *s, TCGType type,
     tcg_out64(s, arg);
 }
 
+static void tcg_out_movi(TCGContext *s, TCGType type,
+                         TCGReg ret, tcg_target_long arg)
+{
+    switch (type) {
+    case TCG_TYPE_I32:
+#if TCG_TARGET_REG_BITS == 64
+    case TCG_TYPE_I64:
+#endif
+        if (ret < 16) {
+            tcg_out_movi_int(s, type, ret, arg);
+        } else {
+            tcg_out_movi_vec(s, type, ret, arg);
+        }
+        break;
+    default:
+        g_assert_not_reached();
+    }
+}
+
 static inline void tcg_out_pushi(TCGContext *s, tcg_target_long val)
 {
     if (val == (int8_t)val) {
@@ -2641,8 +2666,6 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
         break;
     case INDEX_op_mov_i32:  /* Always emitted via tcg_out_mov.  */
     case INDEX_op_mov_i64:
-    case INDEX_op_movi_i32: /* Always emitted via tcg_out_movi.  */
-    case INDEX_op_movi_i64:
     case INDEX_op_call:     /* Always emitted via tcg_out_call.  */
     default:
         tcg_abort();
@@ -2928,7 +2951,6 @@ static void tcg_out_vec_op(TCGContext *s, TCGOpcode opc,
         break;
 
     case INDEX_op_mov_vec:  /* Always emitted via tcg_out_mov.  */
-    case INDEX_op_dupi_vec: /* Always emitted via tcg_out_movi.  */
     case INDEX_op_dup_vec:  /* Always emitted via tcg_out_dup_vec.  */
     default:
         g_assert_not_reached();
@@ -3464,7 +3486,7 @@ static void expand_vec_rotv(TCGType type, unsigned vece, TCGv_vec v0,
 static void expand_vec_mul(TCGType type, unsigned vece,
                            TCGv_vec v0, TCGv_vec v1, TCGv_vec v2)
 {
-    TCGv_vec t1, t2, t3, t4;
+    TCGv_vec t1, t2, t3, t4, zero;
 
     tcg_debug_assert(vece == MO_8);
 
@@ -3482,11 +3504,11 @@ static void expand_vec_mul(TCGType type, unsigned vece,
     case TCG_TYPE_V64:
         t1 = tcg_temp_new_vec(TCG_TYPE_V128);
         t2 = tcg_temp_new_vec(TCG_TYPE_V128);
-        tcg_gen_dup16i_vec(t2, 0);
+        zero = tcg_constant_vec(TCG_TYPE_V128, MO_8, 0);
         vec_gen_3(INDEX_op_x86_punpckl_vec, TCG_TYPE_V128, MO_8,
-                  tcgv_vec_arg(t1), tcgv_vec_arg(v1), tcgv_vec_arg(t2));
+                  tcgv_vec_arg(t1), tcgv_vec_arg(v1), tcgv_vec_arg(zero));
         vec_gen_3(INDEX_op_x86_punpckl_vec, TCG_TYPE_V128, MO_8,
-                  tcgv_vec_arg(t2), tcgv_vec_arg(t2), tcgv_vec_arg(v2));
+                  tcgv_vec_arg(t2), tcgv_vec_arg(zero), tcgv_vec_arg(v2));
         tcg_gen_mul_vec(MO_16, t1, t1, t2);
         tcg_gen_shri_vec(MO_16, t1, t1, 8);
         vec_gen_3(INDEX_op_x86_packus_vec, TCG_TYPE_V128, MO_8,
@@ -3501,15 +3523,15 @@ static void expand_vec_mul(TCGType type, unsigned vece,
         t2 = tcg_temp_new_vec(type);
         t3 = tcg_temp_new_vec(type);
         t4 = tcg_temp_new_vec(type);
-        tcg_gen_dup16i_vec(t4, 0);
+        zero = tcg_constant_vec(TCG_TYPE_V128, MO_8, 0);
         vec_gen_3(INDEX_op_x86_punpckl_vec, type, MO_8,
-                  tcgv_vec_arg(t1), tcgv_vec_arg(v1), tcgv_vec_arg(t4));
+                  tcgv_vec_arg(t1), tcgv_vec_arg(v1), tcgv_vec_arg(zero));
         vec_gen_3(INDEX_op_x86_punpckl_vec, type, MO_8,
-                  tcgv_vec_arg(t2), tcgv_vec_arg(t4), tcgv_vec_arg(v2));
+                  tcgv_vec_arg(t2), tcgv_vec_arg(zero), tcgv_vec_arg(v2));
         vec_gen_3(INDEX_op_x86_punpckh_vec, type, MO_8,
-                  tcgv_vec_arg(t3), tcgv_vec_arg(v1), tcgv_vec_arg(t4));
+                  tcgv_vec_arg(t3), tcgv_vec_arg(v1), tcgv_vec_arg(zero));
         vec_gen_3(INDEX_op_x86_punpckh_vec, type, MO_8,
-                  tcgv_vec_arg(t4), tcgv_vec_arg(t4), tcgv_vec_arg(v2));
+                  tcgv_vec_arg(t4), tcgv_vec_arg(zero), tcgv_vec_arg(v2));
         tcg_gen_mul_vec(MO_16, t1, t1, t2);
         tcg_gen_mul_vec(MO_16, t3, t3, t4);
         tcg_gen_shri_vec(MO_16, t1, t1, 8);
@@ -3537,7 +3559,7 @@ static bool expand_vec_cmp_noinv(TCGType type, unsigned vece, TCGv_vec v0,
         NEED_UMIN = 8,
         NEED_UMAX = 16,
     };
-    TCGv_vec t1, t2;
+    TCGv_vec t1, t2, t3;
     uint8_t fixup;
 
     switch (cond) {
@@ -3608,9 +3630,9 @@ static bool expand_vec_cmp_noinv(TCGType type, unsigned vece, TCGv_vec v0,
     } else if (fixup & NEED_BIAS) {
         t1 = tcg_temp_new_vec(type);
         t2 = tcg_temp_new_vec(type);
-        tcg_gen_dupi_vec(vece, t2, 1ull << ((8 << vece) - 1));
-        tcg_gen_sub_vec(vece, t1, v1, t2);
-        tcg_gen_sub_vec(vece, t2, v2, t2);
+        t3 = tcg_constant_vec(type, vece, 1ull << ((8 << vece) - 1));
+        tcg_gen_sub_vec(vece, t1, v1, t3);
+        tcg_gen_sub_vec(vece, t2, v2, t3);
         v1 = t1;
         v2 = t2;
         cond = tcg_signed_cond(cond);
diff --git a/tcg/mips/tcg-target.c.inc b/tcg/mips/tcg-target.c.inc
index add157f6c3..7293169ab2 100644
--- a/tcg/mips/tcg-target.c.inc
+++ b/tcg/mips/tcg-target.c.inc
@@ -2141,8 +2141,6 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
         break;
     case INDEX_op_mov_i32:  /* Always emitted via tcg_out_mov.  */
     case INDEX_op_mov_i64:
-    case INDEX_op_movi_i32: /* Always emitted via tcg_out_movi.  */
-    case INDEX_op_movi_i64:
     case INDEX_op_call:     /* Always emitted via tcg_out_call.  */
     default:
         tcg_abort();
diff --git a/tcg/optimize.c b/tcg/optimize.c
index 1fb42eb2a9..37c902283e 100644
--- a/tcg/optimize.c
+++ b/tcg/optimize.c
@@ -35,20 +35,20 @@
         glue(glue(case INDEX_op_, x), _i64):    \
         glue(glue(case INDEX_op_, x), _vec)
 
-struct tcg_temp_info {
+typedef struct TempOptInfo {
     bool is_const;
     TCGTemp *prev_copy;
     TCGTemp *next_copy;
-    tcg_target_ulong val;
-    tcg_target_ulong mask;
-};
+    uint64_t val;
+    uint64_t mask;
+} TempOptInfo;
 
-static inline struct tcg_temp_info *ts_info(TCGTemp *ts)
+static inline TempOptInfo *ts_info(TCGTemp *ts)
 {
     return ts->state_ptr;
 }
 
-static inline struct tcg_temp_info *arg_info(TCGArg arg)
+static inline TempOptInfo *arg_info(TCGArg arg)
 {
     return ts_info(arg_temp(arg));
 }
@@ -71,9 +71,9 @@ static inline bool ts_is_copy(TCGTemp *ts)
 /* Reset TEMP's state, possibly removing the temp for the list of copies.  */
 static void reset_ts(TCGTemp *ts)
 {
-    struct tcg_temp_info *ti = ts_info(ts);
-    struct tcg_temp_info *pi = ts_info(ti->prev_copy);
-    struct tcg_temp_info *ni = ts_info(ti->next_copy);
+    TempOptInfo *ti = ts_info(ts);
+    TempOptInfo *pi = ts_info(ti->prev_copy);
+    TempOptInfo *ni = ts_info(ti->next_copy);
 
     ni->prev_copy = ti->prev_copy;
     pi->next_copy = ti->next_copy;
@@ -89,55 +89,67 @@ static void reset_temp(TCGArg arg)
 }
 
 /* Initialize and activate a temporary.  */
-static void init_ts_info(struct tcg_temp_info *infos,
-                         TCGTempSet *temps_used, TCGTemp *ts)
+static void init_ts_info(TCGTempSet *temps_used, TCGTemp *ts)
 {
     size_t idx = temp_idx(ts);
-    if (!test_bit(idx, temps_used->l)) {
-        struct tcg_temp_info *ti = &infos[idx];
+    TempOptInfo *ti;
 
+    if (test_bit(idx, temps_used->l)) {
+        return;
+    }
+    set_bit(idx, temps_used->l);
+
+    ti = ts->state_ptr;
+    if (ti == NULL) {
+        ti = tcg_malloc(sizeof(TempOptInfo));
         ts->state_ptr = ti;
-        ti->next_copy = ts;
-        ti->prev_copy = ts;
+    }
+
+    ti->next_copy = ts;
+    ti->prev_copy = ts;
+    if (ts->kind == TEMP_CONST) {
+        ti->is_const = true;
+        ti->val = ts->val;
+        ti->mask = ts->val;
+        if (TCG_TARGET_REG_BITS > 32 && ts->type == TCG_TYPE_I32) {
+            /* High bits of a 32-bit quantity are garbage.  */
+            ti->mask |= ~0xffffffffull;
+        }
+    } else {
         ti->is_const = false;
         ti->mask = -1;
-        set_bit(idx, temps_used->l);
     }
 }
 
-static void init_arg_info(struct tcg_temp_info *infos,
-                          TCGTempSet *temps_used, TCGArg arg)
+static void init_arg_info(TCGTempSet *temps_used, TCGArg arg)
 {
-    init_ts_info(infos, temps_used, arg_temp(arg));
+    init_ts_info(temps_used, arg_temp(arg));
 }
 
 static TCGTemp *find_better_copy(TCGContext *s, TCGTemp *ts)
 {
-    TCGTemp *i;
+    TCGTemp *i, *g, *l;
 
-    /* If this is already a global, we can't do better. */
-    if (ts->temp_global) {
+    /* If this is already readonly, we can't do better. */
+    if (temp_readonly(ts)) {
         return ts;
     }
 
-    /* Search for a global first. */
+    g = l = NULL;
     for (i = ts_info(ts)->next_copy; i != ts; i = ts_info(i)->next_copy) {
-        if (i->temp_global) {
+        if (temp_readonly(i)) {
             return i;
-        }
-    }
-
-    /* If it is a temp, search for a temp local. */
-    if (!ts->temp_local) {
-        for (i = ts_info(ts)->next_copy; i != ts; i = ts_info(i)->next_copy) {
-            if (ts->temp_local) {
-                return i;
+        } else if (i->kind > ts->kind) {
+            if (i->kind == TEMP_GLOBAL) {
+                g = i;
+            } else if (i->kind == TEMP_LOCAL) {
+                l = i;
             }
         }
     }
 
-    /* Failure to find a better representation, return the same temp. */
-    return ts;
+    /* If we didn't find a better representation, return the same temp. */
+    return g ? g : l ? l : ts;
 }
 
 static bool ts_are_copies(TCGTemp *ts1, TCGTemp *ts2)
@@ -166,45 +178,14 @@ static bool args_are_copies(TCGArg arg1, TCGArg arg2)
     return ts_are_copies(arg_temp(arg1), arg_temp(arg2));
 }
 
-static void tcg_opt_gen_movi(TCGContext *s, TCGOp *op, TCGArg dst, TCGArg val)
-{
-    const TCGOpDef *def;
-    TCGOpcode new_op;
-    tcg_target_ulong mask;
-    struct tcg_temp_info *di = arg_info(dst);
-
-    def = &tcg_op_defs[op->opc];
-    if (def->flags & TCG_OPF_VECTOR) {
-        new_op = INDEX_op_dupi_vec;
-    } else if (def->flags & TCG_OPF_64BIT) {
-        new_op = INDEX_op_movi_i64;
-    } else {
-        new_op = INDEX_op_movi_i32;
-    }
-    op->opc = new_op;
-    /* TCGOP_VECL and TCGOP_VECE remain unchanged.  */
-    op->args[0] = dst;
-    op->args[1] = val;
-
-    reset_temp(dst);
-    di->is_const = true;
-    di->val = val;
-    mask = val;
-    if (TCG_TARGET_REG_BITS > 32 && new_op == INDEX_op_movi_i32) {
-        /* High bits of the destination are now garbage.  */
-        mask |= ~0xffffffffull;
-    }
-    di->mask = mask;
-}
-
 static void tcg_opt_gen_mov(TCGContext *s, TCGOp *op, TCGArg dst, TCGArg src)
 {
     TCGTemp *dst_ts = arg_temp(dst);
     TCGTemp *src_ts = arg_temp(src);
     const TCGOpDef *def;
-    struct tcg_temp_info *di;
-    struct tcg_temp_info *si;
-    tcg_target_ulong mask;
+    TempOptInfo *di;
+    TempOptInfo *si;
+    uint64_t mask;
     TCGOpcode new_op;
 
     if (ts_are_copies(dst_ts, src_ts)) {
@@ -236,7 +217,7 @@ static void tcg_opt_gen_mov(TCGContext *s, TCGOp *op, TCGArg dst, TCGArg src)
     di->mask = mask;
 
     if (src_ts->type == dst_ts->type) {
-        struct tcg_temp_info *ni = ts_info(si->next_copy);
+        TempOptInfo *ni = ts_info(si->next_copy);
 
         di->next_copy = si->next_copy;
         di->prev_copy = src_ts;
@@ -247,7 +228,28 @@ static void tcg_opt_gen_mov(TCGContext *s, TCGOp *op, TCGArg dst, TCGArg src)
     }
 }
 
-static TCGArg do_constant_folding_2(TCGOpcode op, TCGArg x, TCGArg y)
+static void tcg_opt_gen_movi(TCGContext *s, TCGTempSet *temps_used,
+                             TCGOp *op, TCGArg dst, uint64_t val)
+{
+    const TCGOpDef *def = &tcg_op_defs[op->opc];
+    TCGType type;
+    TCGTemp *tv;
+
+    if (def->flags & TCG_OPF_VECTOR) {
+        type = TCGOP_VECL(op) + TCG_TYPE_V64;
+    } else if (def->flags & TCG_OPF_64BIT) {
+        type = TCG_TYPE_I64;
+    } else {
+        type = TCG_TYPE_I32;
+    }
+
+    /* Convert movi to mov with constant temp. */
+    tv = tcg_constant_internal(type, val);
+    init_ts_info(temps_used, tv);
+    tcg_opt_gen_mov(s, op, dst, temp_arg(tv));
+}
+
+static uint64_t do_constant_folding_2(TCGOpcode op, uint64_t x, uint64_t y)
 {
     uint64_t l64, h64;
 
@@ -410,10 +412,10 @@ static TCGArg do_constant_folding_2(TCGOpcode op, TCGArg x, TCGArg y)
     }
 }
 
-static TCGArg do_constant_folding(TCGOpcode op, TCGArg x, TCGArg y)
+static uint64_t do_constant_folding(TCGOpcode op, uint64_t x, uint64_t y)
 {
     const TCGOpDef *def = &tcg_op_defs[op];
-    TCGArg res = do_constant_folding_2(op, x, y);
+    uint64_t res = do_constant_folding_2(op, x, y);
     if (!(def->flags & TCG_OPF_64BIT)) {
         res = (int32_t)res;
     }
@@ -501,8 +503,9 @@ static bool do_constant_folding_cond_eq(TCGCond c)
 static TCGArg do_constant_folding_cond(TCGOpcode op, TCGArg x,
                                        TCGArg y, TCGCond c)
 {
-    tcg_target_ulong xv = arg_info(x)->val;
-    tcg_target_ulong yv = arg_info(y)->val;
+    uint64_t xv = arg_info(x)->val;
+    uint64_t yv = arg_info(y)->val;
+
     if (arg_is_const(x) && arg_is_const(y)) {
         const TCGOpDef *def = &tcg_op_defs[op];
         tcg_debug_assert(!(def->flags & TCG_OPF_VECTOR));
@@ -597,9 +600,8 @@ static bool swap_commutative2(TCGArg *p1, TCGArg *p2)
 /* Propagate constants and copies, fold constant expressions. */
 void tcg_optimize(TCGContext *s)
 {
-    int nb_temps, nb_globals;
+    int nb_temps, nb_globals, i;
     TCGOp *op, *op_next, *prev_mb = NULL;
-    struct tcg_temp_info *infos;
     TCGTempSet temps_used;
 
     /* Array VALS has an element for each temp.
@@ -609,13 +611,15 @@ void tcg_optimize(TCGContext *s)
 
     nb_temps = s->nb_temps;
     nb_globals = s->nb_globals;
-    bitmap_zero(temps_used.l, nb_temps);
-    infos = tcg_malloc(sizeof(struct tcg_temp_info) * nb_temps);
+
+    memset(&temps_used, 0, sizeof(temps_used));
+    for (i = 0; i < nb_temps; ++i) {
+        s->temps[i].state_ptr = NULL;
+    }
 
     QTAILQ_FOREACH_SAFE(op, &s->ops, link, op_next) {
-        tcg_target_ulong mask, partmask, affected;
-        int nb_oargs, nb_iargs, i;
-        TCGArg tmp;
+        uint64_t mask, partmask, affected, tmp;
+        int nb_oargs, nb_iargs;
         TCGOpcode opc = op->opc;
         const TCGOpDef *def = &tcg_op_defs[opc];
 
@@ -627,14 +631,14 @@ void tcg_optimize(TCGContext *s)
             for (i = 0; i < nb_oargs + nb_iargs; i++) {
                 TCGTemp *ts = arg_temp(op->args[i]);
                 if (ts) {
-                    init_ts_info(infos, &temps_used, ts);
+                    init_ts_info(&temps_used, ts);
                 }
             }
         } else {
             nb_oargs = def->nb_oargs;
             nb_iargs = def->nb_iargs;
             for (i = 0; i < nb_oargs + nb_iargs; i++) {
-                init_arg_info(infos, &temps_used, op->args[i]);
+                init_arg_info(&temps_used, op->args[i]);
             }
         }
 
@@ -713,7 +717,7 @@ void tcg_optimize(TCGContext *s)
         CASE_OP_32_64(rotr):
             if (arg_is_const(op->args[1])
                 && arg_info(op->args[1])->val == 0) {
-                tcg_opt_gen_movi(s, op, op->args[0], 0);
+                tcg_opt_gen_movi(s, &temps_used, op, op->args[0], 0);
                 continue;
             }
             break;
@@ -1040,7 +1044,7 @@ void tcg_optimize(TCGContext *s)
 
         if (partmask == 0) {
             tcg_debug_assert(nb_oargs == 1);
-            tcg_opt_gen_movi(s, op, op->args[0], 0);
+            tcg_opt_gen_movi(s, &temps_used, op, op->args[0], 0);
             continue;
         }
         if (affected == 0) {
@@ -1057,7 +1061,7 @@ void tcg_optimize(TCGContext *s)
         CASE_OP_32_64(mulsh):
             if (arg_is_const(op->args[2])
                 && arg_info(op->args[2])->val == 0) {
-                tcg_opt_gen_movi(s, op, op->args[0], 0);
+                tcg_opt_gen_movi(s, &temps_used, op, op->args[0], 0);
                 continue;
             }
             break;
@@ -1084,7 +1088,7 @@ void tcg_optimize(TCGContext *s)
         CASE_OP_32_64_VEC(sub):
         CASE_OP_32_64_VEC(xor):
             if (args_are_copies(op->args[1], op->args[2])) {
-                tcg_opt_gen_movi(s, op, op->args[0], 0);
+                tcg_opt_gen_movi(s, &temps_used, op, op->args[0], 0);
                 continue;
             }
             break;
@@ -1099,16 +1103,12 @@ void tcg_optimize(TCGContext *s)
         CASE_OP_32_64_VEC(mov):
             tcg_opt_gen_mov(s, op, op->args[0], op->args[1]);
             break;
-        CASE_OP_32_64(movi):
-        case INDEX_op_dupi_vec:
-            tcg_opt_gen_movi(s, op, op->args[0], op->args[1]);
-            break;
 
         case INDEX_op_dup_vec:
             if (arg_is_const(op->args[1])) {
                 tmp = arg_info(op->args[1])->val;
                 tmp = dup_const(TCGOP_VECE(op), tmp);
-                tcg_opt_gen_movi(s, op, op->args[0], tmp);
+                tcg_opt_gen_movi(s, &temps_used, op, op->args[0], tmp);
                 break;
             }
             goto do_default;
@@ -1116,11 +1116,10 @@ void tcg_optimize(TCGContext *s)
         case INDEX_op_dup2_vec:
             assert(TCG_TARGET_REG_BITS == 32);
             if (arg_is_const(op->args[1]) && arg_is_const(op->args[2])) {
-                tmp = arg_info(op->args[1])->val;
-                if (tmp == arg_info(op->args[2])->val) {
-                    tcg_opt_gen_movi(s, op, op->args[0], tmp);
-                    break;
-                }
+                tcg_opt_gen_movi(s, &temps_used, op, op->args[0],
+                                 deposit64(arg_info(op->args[1])->val, 32, 32,
+                                           arg_info(op->args[2])->val));
+                break;
             } else if (args_are_copies(op->args[1], op->args[2])) {
                 op->opc = INDEX_op_dup_vec;
                 TCGOP_VECE(op) = MO_32;
@@ -1146,7 +1145,7 @@ void tcg_optimize(TCGContext *s)
         case INDEX_op_extrh_i64_i32:
             if (arg_is_const(op->args[1])) {
                 tmp = do_constant_folding(opc, arg_info(op->args[1])->val, 0);
-                tcg_opt_gen_movi(s, op, op->args[0], tmp);
+                tcg_opt_gen_movi(s, &temps_used, op, op->args[0], tmp);
                 break;
             }
             goto do_default;
@@ -1176,7 +1175,7 @@ void tcg_optimize(TCGContext *s)
             if (arg_is_const(op->args[1]) && arg_is_const(op->args[2])) {
                 tmp = do_constant_folding(opc, arg_info(op->args[1])->val,
                                           arg_info(op->args[2])->val);
-                tcg_opt_gen_movi(s, op, op->args[0], tmp);
+                tcg_opt_gen_movi(s, &temps_used, op, op->args[0], tmp);
                 break;
             }
             goto do_default;
@@ -1187,7 +1186,7 @@ void tcg_optimize(TCGContext *s)
                 TCGArg v = arg_info(op->args[1])->val;
                 if (v != 0) {
                     tmp = do_constant_folding(opc, v, 0);
-                    tcg_opt_gen_movi(s, op, op->args[0], tmp);
+                    tcg_opt_gen_movi(s, &temps_used, op, op->args[0], tmp);
                 } else {
                     tcg_opt_gen_mov(s, op, op->args[0], op->args[2]);
                 }
@@ -1200,7 +1199,7 @@ void tcg_optimize(TCGContext *s)
                 tmp = deposit64(arg_info(op->args[1])->val,
                                 op->args[3], op->args[4],
                                 arg_info(op->args[2])->val);
-                tcg_opt_gen_movi(s, op, op->args[0], tmp);
+                tcg_opt_gen_movi(s, &temps_used, op, op->args[0], tmp);
                 break;
             }
             goto do_default;
@@ -1209,7 +1208,7 @@ void tcg_optimize(TCGContext *s)
             if (arg_is_const(op->args[1])) {
                 tmp = extract64(arg_info(op->args[1])->val,
                                 op->args[2], op->args[3]);
-                tcg_opt_gen_movi(s, op, op->args[0], tmp);
+                tcg_opt_gen_movi(s, &temps_used, op, op->args[0], tmp);
                 break;
             }
             goto do_default;
@@ -1218,23 +1217,24 @@ void tcg_optimize(TCGContext *s)
             if (arg_is_const(op->args[1])) {
                 tmp = sextract64(arg_info(op->args[1])->val,
                                  op->args[2], op->args[3]);
-                tcg_opt_gen_movi(s, op, op->args[0], tmp);
+                tcg_opt_gen_movi(s, &temps_used, op, op->args[0], tmp);
                 break;
             }
             goto do_default;
 
         CASE_OP_32_64(extract2):
             if (arg_is_const(op->args[1]) && arg_is_const(op->args[2])) {
-                TCGArg v1 = arg_info(op->args[1])->val;
-                TCGArg v2 = arg_info(op->args[2])->val;
+                uint64_t v1 = arg_info(op->args[1])->val;
+                uint64_t v2 = arg_info(op->args[2])->val;
+                int shr = op->args[3];
 
                 if (opc == INDEX_op_extract2_i64) {
-                    tmp = (v1 >> op->args[3]) | (v2 << (64 - op->args[3]));
+                    tmp = (v1 >> shr) | (v2 << (64 - shr));
                 } else {
-                    tmp = (int32_t)(((uint32_t)v1 >> op->args[3]) |
-                                    ((uint32_t)v2 << (32 - op->args[3])));
+                    tmp = (int32_t)(((uint32_t)v1 >> shr) |
+                                    ((uint32_t)v2 << (32 - shr)));
                 }
-                tcg_opt_gen_movi(s, op, op->args[0], tmp);
+                tcg_opt_gen_movi(s, &temps_used, op, op->args[0], tmp);
                 break;
             }
             goto do_default;
@@ -1243,7 +1243,7 @@ void tcg_optimize(TCGContext *s)
             tmp = do_constant_folding_cond(opc, op->args[1],
                                            op->args[2], op->args[3]);
             if (tmp != 2) {
-                tcg_opt_gen_movi(s, op, op->args[0], tmp);
+                tcg_opt_gen_movi(s, &temps_used, op, op->args[0], tmp);
                 break;
             }
             goto do_default;
@@ -1253,7 +1253,7 @@ void tcg_optimize(TCGContext *s)
                                            op->args[1], op->args[2]);
             if (tmp != 2) {
                 if (tmp) {
-                    bitmap_zero(temps_used.l, nb_temps);
+                    memset(&temps_used, 0, sizeof(temps_used));
                     op->opc = INDEX_op_br;
                     op->args[0] = op->args[3];
                 } else {
@@ -1271,9 +1271,10 @@ void tcg_optimize(TCGContext *s)
                 break;
             }
             if (arg_is_const(op->args[3]) && arg_is_const(op->args[4])) {
-                tcg_target_ulong tv = arg_info(op->args[3])->val;
-                tcg_target_ulong fv = arg_info(op->args[4])->val;
+                uint64_t tv = arg_info(op->args[3])->val;
+                uint64_t fv = arg_info(op->args[4])->val;
                 TCGCond cond = op->args[5];
+
                 if (fv == 1 && tv == 0) {
                     cond = tcg_invert_cond(cond);
                 } else if (!(tv == 1 && fv == 0)) {
@@ -1298,7 +1299,7 @@ void tcg_optimize(TCGContext *s)
                 uint64_t a = ((uint64_t)ah << 32) | al;
                 uint64_t b = ((uint64_t)bh << 32) | bl;
                 TCGArg rl, rh;
-                TCGOp *op2 = tcg_op_insert_before(s, op, INDEX_op_movi_i32);
+                TCGOp *op2 = tcg_op_insert_before(s, op, INDEX_op_mov_i32);
 
                 if (opc == INDEX_op_add2_i32) {
                     a += b;
@@ -1308,8 +1309,8 @@ void tcg_optimize(TCGContext *s)
 
                 rl = op->args[0];
                 rh = op->args[1];
-                tcg_opt_gen_movi(s, op, rl, (int32_t)a);
-                tcg_opt_gen_movi(s, op2, rh, (int32_t)(a >> 32));
+                tcg_opt_gen_movi(s, &temps_used, op, rl, (int32_t)a);
+                tcg_opt_gen_movi(s, &temps_used, op2, rh, (int32_t)(a >> 32));
                 break;
             }
             goto do_default;
@@ -1320,12 +1321,12 @@ void tcg_optimize(TCGContext *s)
                 uint32_t b = arg_info(op->args[3])->val;
                 uint64_t r = (uint64_t)a * b;
                 TCGArg rl, rh;
-                TCGOp *op2 = tcg_op_insert_before(s, op, INDEX_op_movi_i32);
+                TCGOp *op2 = tcg_op_insert_before(s, op, INDEX_op_mov_i32);
 
                 rl = op->args[0];
                 rh = op->args[1];
-                tcg_opt_gen_movi(s, op, rl, (int32_t)r);
-                tcg_opt_gen_movi(s, op2, rh, (int32_t)(r >> 32));
+                tcg_opt_gen_movi(s, &temps_used, op, rl, (int32_t)r);
+                tcg_opt_gen_movi(s, &temps_used, op2, rh, (int32_t)(r >> 32));
                 break;
             }
             goto do_default;
@@ -1336,7 +1337,7 @@ void tcg_optimize(TCGContext *s)
             if (tmp != 2) {
                 if (tmp) {
             do_brcond_true:
-                    bitmap_zero(temps_used.l, nb_temps);
+                    memset(&temps_used, 0, sizeof(temps_used));
                     op->opc = INDEX_op_br;
                     op->args[0] = op->args[5];
                 } else {
@@ -1352,7 +1353,7 @@ void tcg_optimize(TCGContext *s)
                 /* Simplify LT/GE comparisons vs zero to a single compare
                    vs the high word of the input.  */
             do_brcond_high:
-                bitmap_zero(temps_used.l, nb_temps);
+                memset(&temps_used, 0, sizeof(temps_used));
                 op->opc = INDEX_op_brcond_i32;
                 op->args[0] = op->args[1];
                 op->args[1] = op->args[3];
@@ -1378,7 +1379,7 @@ void tcg_optimize(TCGContext *s)
                     goto do_default;
                 }
             do_brcond_low:
-                bitmap_zero(temps_used.l, nb_temps);
+                memset(&temps_used, 0, sizeof(temps_used));
                 op->opc = INDEX_op_brcond_i32;
                 op->args[1] = op->args[2];
                 op->args[2] = op->args[4];
@@ -1413,7 +1414,7 @@ void tcg_optimize(TCGContext *s)
                                             op->args[5]);
             if (tmp != 2) {
             do_setcond_const:
-                tcg_opt_gen_movi(s, op, op->args[0], tmp);
+                tcg_opt_gen_movi(s, &temps_used, op, op->args[0], tmp);
             } else if ((op->args[5] == TCG_COND_LT
                         || op->args[5] == TCG_COND_GE)
                        && arg_is_const(op->args[3])
@@ -1498,7 +1499,7 @@ void tcg_optimize(TCGContext *s)
                block, otherwise we only trash the output args.  "mask" is
                the non-zero bits mask for the first output arg.  */
             if (def->flags & TCG_OPF_BB_END) {
-                bitmap_zero(temps_used.l, nb_temps);
+                memset(&temps_used, 0, sizeof(temps_used));
             } else {
         do_reset_output:
                 for (i = 0; i < nb_oargs; i++) {
diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc
index 19a4a12f15..cf64892295 100644
--- a/tcg/ppc/tcg-target.c.inc
+++ b/tcg/ppc/tcg-target.c.inc
@@ -912,31 +912,41 @@ static void tcg_out_movi_int(TCGContext *s, TCGType type, TCGReg ret,
     }
 }
 
-static void tcg_out_dupi_vec(TCGContext *s, TCGType type, TCGReg ret,
-                             tcg_target_long val)
+static void tcg_out_dupi_vec(TCGContext *s, TCGType type, unsigned vece,
+                             TCGReg ret, int64_t val)
 {
     uint32_t load_insn;
     int rel, low;
     intptr_t add;
 
-    low = (int8_t)val;
-    if (low >= -16 && low < 16) {
-        if (val == (tcg_target_long)dup_const(MO_8, low)) {
+    switch (vece) {
+    case MO_8:
+        low = (int8_t)val;
+        if (low >= -16 && low < 16) {
             tcg_out32(s, VSPLTISB | VRT(ret) | ((val & 31) << 16));
             return;
         }
-        if (val == (tcg_target_long)dup_const(MO_16, low)) {
+        if (have_isa_3_00) {
+            tcg_out32(s, XXSPLTIB | VRT(ret) | ((val & 0xff) << 11));
+            return;
+        }
+        break;
+
+    case MO_16:
+        low = (int16_t)val;
+        if (low >= -16 && low < 16) {
             tcg_out32(s, VSPLTISH | VRT(ret) | ((val & 31) << 16));
             return;
         }
-        if (val == (tcg_target_long)dup_const(MO_32, low)) {
+        break;
+
+    case MO_32:
+        low = (int32_t)val;
+        if (low >= -16 && low < 16) {
             tcg_out32(s, VSPLTISW | VRT(ret) | ((val & 31) << 16));
             return;
         }
-    }
-    if (have_isa_3_00 && val == (tcg_target_long)dup_const(MO_8, val)) {
-        tcg_out32(s, XXSPLTIB | VRT(ret) | ((val & 0xff) << 11));
-        return;
+        break;
     }
 
     /*
@@ -956,14 +966,15 @@ static void tcg_out_dupi_vec(TCGContext *s, TCGType type, TCGReg ret,
         if (TCG_TARGET_REG_BITS == 64) {
             new_pool_label(s, val, rel, s->code_ptr, add);
         } else {
-            new_pool_l2(s, rel, s->code_ptr, add, val, val);
+            new_pool_l2(s, rel, s->code_ptr, add, val >> 32, val);
         }
     } else {
         load_insn = LVX | VRT(ret) | RB(TCG_REG_TMP1);
         if (TCG_TARGET_REG_BITS == 64) {
             new_pool_l2(s, rel, s->code_ptr, add, val, val);
         } else {
-            new_pool_l4(s, rel, s->code_ptr, add, val, val, val, val);
+            new_pool_l4(s, rel, s->code_ptr, add,
+                        val >> 32, val, val >> 32, val);
         }
     }
 
@@ -987,12 +998,6 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg ret,
         tcg_out_movi_int(s, type, ret, arg, false);
         break;
 
-    case TCG_TYPE_V64:
-    case TCG_TYPE_V128:
-        tcg_debug_assert(ret >= TCG_REG_V0);
-        tcg_out_dupi_vec(s, type, ret, arg);
-        break;
-
     default:
         g_assert_not_reached();
     }
@@ -2972,8 +2977,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args,
 
     case INDEX_op_mov_i32:   /* Always emitted via tcg_out_mov.  */
     case INDEX_op_mov_i64:
-    case INDEX_op_movi_i32:  /* Always emitted via tcg_out_movi.  */
-    case INDEX_op_movi_i64:
     case INDEX_op_call:      /* Always emitted via tcg_out_call.  */
     default:
         tcg_abort();
@@ -3321,7 +3324,6 @@ static void tcg_out_vec_op(TCGContext *s, TCGOpcode opc,
         return;
 
     case INDEX_op_mov_vec:  /* Always emitted via tcg_out_mov.  */
-    case INDEX_op_dupi_vec: /* Always emitted via tcg_out_movi.  */
     case INDEX_op_dup_vec:  /* Always emitted via tcg_out_dup_vec.  */
     default:
         g_assert_not_reached();
@@ -3334,13 +3336,22 @@ static void tcg_out_vec_op(TCGContext *s, TCGOpcode opc,
 static void expand_vec_shi(TCGType type, unsigned vece, TCGv_vec v0,
                            TCGv_vec v1, TCGArg imm, TCGOpcode opci)
 {
-    TCGv_vec t1 = tcg_temp_new_vec(type);
+    TCGv_vec t1;
 
-    /* Splat w/bytes for xxspltib.  */
-    tcg_gen_dupi_vec(MO_8, t1, imm & ((8 << vece) - 1));
+    if (vece == MO_32) {
+        /*
+         * Only 5 bits are significant, and VSPLTISB can represent -16..15.
+         * So using negative numbers gets us the 4th bit easily.
+         */
+        imm = sextract32(imm, 0, 5);
+    } else {
+        imm &= (8 << vece) - 1;
+    }
+
+    /* Splat w/bytes for xxspltib when 2.07 allows MO_64. */
+    t1 = tcg_constant_vec(type, MO_8, imm);
     vec_gen_3(opci, type, vece, tcgv_vec_arg(v0),
               tcgv_vec_arg(v1), tcgv_vec_arg(t1));
-    tcg_temp_free_vec(t1);
 }
 
 static void expand_vec_cmp(TCGType type, unsigned vece, TCGv_vec v0,
@@ -3398,7 +3409,7 @@ static void expand_vec_mul(TCGType type, unsigned vece, TCGv_vec v0,
 {
     TCGv_vec t1 = tcg_temp_new_vec(type);
     TCGv_vec t2 = tcg_temp_new_vec(type);
-    TCGv_vec t3, t4;
+    TCGv_vec c0, c16;
 
     switch (vece) {
     case MO_8:
@@ -3417,21 +3428,22 @@ static void expand_vec_mul(TCGType type, unsigned vece, TCGv_vec v0,
 
     case MO_32:
         tcg_debug_assert(!have_isa_2_07);
-        t3 = tcg_temp_new_vec(type);
-        t4 = tcg_temp_new_vec(type);
-        tcg_gen_dupi_vec(MO_8, t4, -16);
+        /*
+         * Only 5 bits are significant, and VSPLTISB can represent -16..15.
+         * So using -16 is a quick way to represent 16.
+         */
+        c16 = tcg_constant_vec(type, MO_8, -16);
+        c0 = tcg_constant_vec(type, MO_8, 0);
+
         vec_gen_3(INDEX_op_rotlv_vec, type, MO_32, tcgv_vec_arg(t1),
-                  tcgv_vec_arg(v2), tcgv_vec_arg(t4));
+                  tcgv_vec_arg(v2), tcgv_vec_arg(c16));
         vec_gen_3(INDEX_op_ppc_mulou_vec, type, MO_16, tcgv_vec_arg(t2),
                   tcgv_vec_arg(v1), tcgv_vec_arg(v2));
-        tcg_gen_dupi_vec(MO_8, t3, 0);
-        vec_gen_4(INDEX_op_ppc_msum_vec, type, MO_16, tcgv_vec_arg(t3),
-                  tcgv_vec_arg(v1), tcgv_vec_arg(t1), tcgv_vec_arg(t3));
-        vec_gen_3(INDEX_op_shlv_vec, type, MO_32, tcgv_vec_arg(t3),
-                  tcgv_vec_arg(t3), tcgv_vec_arg(t4));
-        tcg_gen_add_vec(MO_32, v0, t2, t3);
-        tcg_temp_free_vec(t3);
-        tcg_temp_free_vec(t4);
+        vec_gen_4(INDEX_op_ppc_msum_vec, type, MO_16, tcgv_vec_arg(t1),
+                  tcgv_vec_arg(v1), tcgv_vec_arg(t1), tcgv_vec_arg(c0));
+        vec_gen_3(INDEX_op_shlv_vec, type, MO_32, tcgv_vec_arg(t1),
+                  tcgv_vec_arg(t1), tcgv_vec_arg(c16));
+        tcg_gen_add_vec(MO_32, v0, t1, t2);
         break;
 
     default:
diff --git a/tcg/riscv/tcg-target.c.inc b/tcg/riscv/tcg-target.c.inc
index c60b91ba58..71c0badc02 100644
--- a/tcg/riscv/tcg-target.c.inc
+++ b/tcg/riscv/tcg-target.c.inc
@@ -1563,8 +1563,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
 
     case INDEX_op_mov_i32:  /* Always emitted via tcg_out_mov.  */
     case INDEX_op_mov_i64:
-    case INDEX_op_movi_i32: /* Always emitted via tcg_out_movi.  */
-    case INDEX_op_movi_i64:
     case INDEX_op_call:     /* Always emitted via tcg_out_call.  */
     default:
         g_assert_not_reached();
diff --git a/tcg/s390/tcg-target.c.inc b/tcg/s390/tcg-target.c.inc
index d7ef079055..8517e55232 100644
--- a/tcg/s390/tcg-target.c.inc
+++ b/tcg/s390/tcg-target.c.inc
@@ -2295,8 +2295,6 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
 
     case INDEX_op_mov_i32:  /* Always emitted via tcg_out_mov.  */
     case INDEX_op_mov_i64:
-    case INDEX_op_movi_i32: /* Always emitted via tcg_out_movi.  */
-    case INDEX_op_movi_i64:
     case INDEX_op_call:     /* Always emitted via tcg_out_call.  */
     default:
         tcg_abort();
diff --git a/tcg/sparc/tcg-target.c.inc b/tcg/sparc/tcg-target.c.inc
index 922ae96481..28b5b6559a 100644
--- a/tcg/sparc/tcg-target.c.inc
+++ b/tcg/sparc/tcg-target.c.inc
@@ -1586,8 +1586,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
 
     case INDEX_op_mov_i32:  /* Always emitted via tcg_out_mov.  */
     case INDEX_op_mov_i64:
-    case INDEX_op_movi_i32: /* Always emitted via tcg_out_movi.  */
-    case INDEX_op_movi_i64:
     case INDEX_op_call:     /* Always emitted via tcg_out_call.  */
     default:
         tcg_abort();
diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c
index 1a41dfa908..498a959839 100644
--- a/tcg/tcg-op-gvec.c
+++ b/tcg/tcg-op-gvec.c
@@ -115,7 +115,7 @@ void tcg_gen_gvec_2_ool(uint32_t dofs, uint32_t aofs,
                         gen_helper_gvec_2 *fn)
 {
     TCGv_ptr a0, a1;
-    TCGv_i32 desc = tcg_const_i32(simd_desc(oprsz, maxsz, data));
+    TCGv_i32 desc = tcg_constant_i32(simd_desc(oprsz, maxsz, data));
 
     a0 = tcg_temp_new_ptr();
     a1 = tcg_temp_new_ptr();
@@ -127,7 +127,6 @@ void tcg_gen_gvec_2_ool(uint32_t dofs, uint32_t aofs,
 
     tcg_temp_free_ptr(a0);
     tcg_temp_free_ptr(a1);
-    tcg_temp_free_i32(desc);
 }
 
 /* Generate a call to a gvec-style helper with two vector operands
@@ -137,7 +136,7 @@ void tcg_gen_gvec_2i_ool(uint32_t dofs, uint32_t aofs, TCGv_i64 c,
                          gen_helper_gvec_2i *fn)
 {
     TCGv_ptr a0, a1;
-    TCGv_i32 desc = tcg_const_i32(simd_desc(oprsz, maxsz, data));
+    TCGv_i32 desc = tcg_constant_i32(simd_desc(oprsz, maxsz, data));
 
     a0 = tcg_temp_new_ptr();
     a1 = tcg_temp_new_ptr();
@@ -149,7 +148,6 @@ void tcg_gen_gvec_2i_ool(uint32_t dofs, uint32_t aofs, TCGv_i64 c,
 
     tcg_temp_free_ptr(a0);
     tcg_temp_free_ptr(a1);
-    tcg_temp_free_i32(desc);
 }
 
 /* Generate a call to a gvec-style helper with three vector operands.  */
@@ -158,7 +156,7 @@ void tcg_gen_gvec_3_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
                         gen_helper_gvec_3 *fn)
 {
     TCGv_ptr a0, a1, a2;
-    TCGv_i32 desc = tcg_const_i32(simd_desc(oprsz, maxsz, data));
+    TCGv_i32 desc = tcg_constant_i32(simd_desc(oprsz, maxsz, data));
 
     a0 = tcg_temp_new_ptr();
     a1 = tcg_temp_new_ptr();
@@ -173,7 +171,6 @@ void tcg_gen_gvec_3_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
     tcg_temp_free_ptr(a0);
     tcg_temp_free_ptr(a1);
     tcg_temp_free_ptr(a2);
-    tcg_temp_free_i32(desc);
 }
 
 /* Generate a call to a gvec-style helper with four vector operands.  */
@@ -182,7 +179,7 @@ void tcg_gen_gvec_4_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
                         int32_t data, gen_helper_gvec_4 *fn)
 {
     TCGv_ptr a0, a1, a2, a3;
-    TCGv_i32 desc = tcg_const_i32(simd_desc(oprsz, maxsz, data));
+    TCGv_i32 desc = tcg_constant_i32(simd_desc(oprsz, maxsz, data));
 
     a0 = tcg_temp_new_ptr();
     a1 = tcg_temp_new_ptr();
@@ -200,7 +197,6 @@ void tcg_gen_gvec_4_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
     tcg_temp_free_ptr(a1);
     tcg_temp_free_ptr(a2);
     tcg_temp_free_ptr(a3);
-    tcg_temp_free_i32(desc);
 }
 
 /* Generate a call to a gvec-style helper with five vector operands.  */
@@ -209,7 +205,7 @@ void tcg_gen_gvec_5_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
                         uint32_t maxsz, int32_t data, gen_helper_gvec_5 *fn)
 {
     TCGv_ptr a0, a1, a2, a3, a4;
-    TCGv_i32 desc = tcg_const_i32(simd_desc(oprsz, maxsz, data));
+    TCGv_i32 desc = tcg_constant_i32(simd_desc(oprsz, maxsz, data));
 
     a0 = tcg_temp_new_ptr();
     a1 = tcg_temp_new_ptr();
@@ -230,7 +226,6 @@ void tcg_gen_gvec_5_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
     tcg_temp_free_ptr(a2);
     tcg_temp_free_ptr(a3);
     tcg_temp_free_ptr(a4);
-    tcg_temp_free_i32(desc);
 }
 
 /* Generate a call to a gvec-style helper with three vector operands
@@ -240,7 +235,7 @@ void tcg_gen_gvec_2_ptr(uint32_t dofs, uint32_t aofs,
                         int32_t data, gen_helper_gvec_2_ptr *fn)
 {
     TCGv_ptr a0, a1;
-    TCGv_i32 desc = tcg_const_i32(simd_desc(oprsz, maxsz, data));
+    TCGv_i32 desc = tcg_constant_i32(simd_desc(oprsz, maxsz, data));
 
     a0 = tcg_temp_new_ptr();
     a1 = tcg_temp_new_ptr();
@@ -252,7 +247,6 @@ void tcg_gen_gvec_2_ptr(uint32_t dofs, uint32_t aofs,
 
     tcg_temp_free_ptr(a0);
     tcg_temp_free_ptr(a1);
-    tcg_temp_free_i32(desc);
 }
 
 /* Generate a call to a gvec-style helper with three vector operands
@@ -262,7 +256,7 @@ void tcg_gen_gvec_3_ptr(uint32_t dofs, uint32_t aofs, uint32_t bofs,
                         int32_t data, gen_helper_gvec_3_ptr *fn)
 {
     TCGv_ptr a0, a1, a2;
-    TCGv_i32 desc = tcg_const_i32(simd_desc(oprsz, maxsz, data));
+    TCGv_i32 desc = tcg_constant_i32(simd_desc(oprsz, maxsz, data));
 
     a0 = tcg_temp_new_ptr();
     a1 = tcg_temp_new_ptr();
@@ -277,7 +271,6 @@ void tcg_gen_gvec_3_ptr(uint32_t dofs, uint32_t aofs, uint32_t bofs,
     tcg_temp_free_ptr(a0);
     tcg_temp_free_ptr(a1);
     tcg_temp_free_ptr(a2);
-    tcg_temp_free_i32(desc);
 }
 
 /* Generate a call to a gvec-style helper with four vector operands
@@ -288,7 +281,7 @@ void tcg_gen_gvec_4_ptr(uint32_t dofs, uint32_t aofs, uint32_t bofs,
                         gen_helper_gvec_4_ptr *fn)
 {
     TCGv_ptr a0, a1, a2, a3;
-    TCGv_i32 desc = tcg_const_i32(simd_desc(oprsz, maxsz, data));
+    TCGv_i32 desc = tcg_constant_i32(simd_desc(oprsz, maxsz, data));
 
     a0 = tcg_temp_new_ptr();
     a1 = tcg_temp_new_ptr();
@@ -306,7 +299,6 @@ void tcg_gen_gvec_4_ptr(uint32_t dofs, uint32_t aofs, uint32_t bofs,
     tcg_temp_free_ptr(a1);
     tcg_temp_free_ptr(a2);
     tcg_temp_free_ptr(a3);
-    tcg_temp_free_i32(desc);
 }
 
 /* Generate a call to a gvec-style helper with five vector operands
@@ -317,7 +309,7 @@ void tcg_gen_gvec_5_ptr(uint32_t dofs, uint32_t aofs, uint32_t bofs,
                         gen_helper_gvec_5_ptr *fn)
 {
     TCGv_ptr a0, a1, a2, a3, a4;
-    TCGv_i32 desc = tcg_const_i32(simd_desc(oprsz, maxsz, data));
+    TCGv_i32 desc = tcg_constant_i32(simd_desc(oprsz, maxsz, data));
 
     a0 = tcg_temp_new_ptr();
     a1 = tcg_temp_new_ptr();
@@ -338,7 +330,6 @@ void tcg_gen_gvec_5_ptr(uint32_t dofs, uint32_t aofs, uint32_t bofs,
     tcg_temp_free_ptr(a2);
     tcg_temp_free_ptr(a3);
     tcg_temp_free_ptr(a4);
-    tcg_temp_free_i32(desc);
 }
 
 /* Return true if we want to implement something of OPRSZ bytes
@@ -605,9 +596,9 @@ static void do_dup(unsigned vece, uint32_t dofs, uint32_t oprsz,
                 || (TCG_TARGET_REG_BITS == 64
                     && (in_c == 0 || in_c == -1
                         || !check_size_impl(oprsz, 4)))) {
-                t_64 = tcg_const_i64(in_c);
+                t_64 = tcg_constant_i64(in_c);
             } else {
-                t_32 = tcg_const_i32(in_c);
+                t_32 = tcg_constant_i32(in_c);
             }
         }
 
@@ -648,11 +639,11 @@ static void do_dup(unsigned vece, uint32_t dofs, uint32_t oprsz,
             t_val = tcg_temp_new_i32();
             tcg_gen_extrl_i64_i32(t_val, in_64);
         } else {
-            t_val = tcg_const_i32(in_c);
+            t_val = tcg_constant_i32(in_c);
         }
         gen_helper_memset(t_ptr, t_ptr, t_val, t_size);
 
-        if (!in_32) {
+        if (in_64) {
             tcg_temp_free_i32(t_val);
         }
         tcg_temp_free_ptr(t_size);
@@ -660,15 +651,14 @@ static void do_dup(unsigned vece, uint32_t dofs, uint32_t oprsz,
         return;
     }
 
-    t_desc = tcg_const_i32(simd_desc(oprsz, maxsz, 0));
+    t_desc = tcg_constant_i32(simd_desc(oprsz, maxsz, 0));
 
     if (vece == MO_64) {
         if (in_64) {
             gen_helper_gvec_dup64(t_ptr, t_desc, in_64);
         } else {
-            t_64 = tcg_const_i64(in_c);
+            t_64 = tcg_constant_i64(in_c);
             gen_helper_gvec_dup64(t_ptr, t_desc, t_64);
-            tcg_temp_free_i64(t_64);
         }
     } else {
         typedef void dup_fn(TCGv_ptr, TCGv_i32, TCGv_i32);
@@ -680,24 +670,23 @@ static void do_dup(unsigned vece, uint32_t dofs, uint32_t oprsz,
 
         if (in_32) {
             fns[vece](t_ptr, t_desc, in_32);
-        } else {
+        } else if (in_64) {
             t_32 = tcg_temp_new_i32();
-            if (in_64) {
-                tcg_gen_extrl_i64_i32(t_32, in_64);
-            } else if (vece == MO_8) {
-                tcg_gen_movi_i32(t_32, in_c & 0xff);
+            tcg_gen_extrl_i64_i32(t_32, in_64);
+            fns[vece](t_ptr, t_desc, t_32);
+            tcg_temp_free_i32(t_32);
+        } else {
+            if (vece == MO_8) {
+                in_c &= 0xff;
             } else if (vece == MO_16) {
-                tcg_gen_movi_i32(t_32, in_c & 0xffff);
-            } else {
-                tcg_gen_movi_i32(t_32, in_c);
+                in_c &= 0xffff;
             }
+            t_32 = tcg_constant_i32(in_c);
             fns[vece](t_ptr, t_desc, t_32);
-            tcg_temp_free_i32(t_32);
         }
     }
 
     tcg_temp_free_ptr(t_ptr);
-    tcg_temp_free_i32(t_desc);
     return;
 
  done:
@@ -1247,10 +1236,9 @@ void tcg_gen_gvec_2i(uint32_t dofs, uint32_t aofs, uint32_t oprsz,
             if (g->fno) {
                 tcg_gen_gvec_2_ool(dofs, aofs, oprsz, maxsz, c, g->fno);
             } else {
-                TCGv_i64 tcg_c = tcg_const_i64(c);
+                TCGv_i64 tcg_c = tcg_constant_i64(c);
                 tcg_gen_gvec_2i_ool(dofs, aofs, tcg_c, oprsz,
                                     maxsz, c, g->fnoi);
-                tcg_temp_free_i64(tcg_c);
             }
             oprsz = maxsz;
         }
@@ -1744,16 +1732,14 @@ static void gen_addv_mask(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b, TCGv_i64 m)
 
 void tcg_gen_vec_add8_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b)
 {
-    TCGv_i64 m = tcg_const_i64(dup_const(MO_8, 0x80));
+    TCGv_i64 m = tcg_constant_i64(dup_const(MO_8, 0x80));
     gen_addv_mask(d, a, b, m);
-    tcg_temp_free_i64(m);
 }
 
 void tcg_gen_vec_add16_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b)
 {
-    TCGv_i64 m = tcg_const_i64(dup_const(MO_16, 0x8000));
+    TCGv_i64 m = tcg_constant_i64(dup_const(MO_16, 0x8000));
     gen_addv_mask(d, a, b, m);
-    tcg_temp_free_i64(m);
 }
 
 void tcg_gen_vec_add32_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b)
@@ -1837,9 +1823,8 @@ void tcg_gen_gvec_adds(unsigned vece, uint32_t dofs, uint32_t aofs,
 void tcg_gen_gvec_addi(unsigned vece, uint32_t dofs, uint32_t aofs,
                        int64_t c, uint32_t oprsz, uint32_t maxsz)
 {
-    TCGv_i64 tmp = tcg_const_i64(c);
+    TCGv_i64 tmp = tcg_constant_i64(c);
     tcg_gen_gvec_adds(vece, dofs, aofs, tmp, oprsz, maxsz);
-    tcg_temp_free_i64(tmp);
 }
 
 static const TCGOpcode vecop_list_sub[] = { INDEX_op_sub_vec, 0 };
@@ -1897,16 +1882,14 @@ static void gen_subv_mask(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b, TCGv_i64 m)
 
 void tcg_gen_vec_sub8_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b)
 {
-    TCGv_i64 m = tcg_const_i64(dup_const(MO_8, 0x80));
+    TCGv_i64 m = tcg_constant_i64(dup_const(MO_8, 0x80));
     gen_subv_mask(d, a, b, m);
-    tcg_temp_free_i64(m);
 }
 
 void tcg_gen_vec_sub16_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b)
 {
-    TCGv_i64 m = tcg_const_i64(dup_const(MO_16, 0x8000));
+    TCGv_i64 m = tcg_constant_i64(dup_const(MO_16, 0x8000));
     gen_subv_mask(d, a, b, m);
-    tcg_temp_free_i64(m);
 }
 
 void tcg_gen_vec_sub32_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b)
@@ -2017,9 +2000,8 @@ void tcg_gen_gvec_muls(unsigned vece, uint32_t dofs, uint32_t aofs,
 void tcg_gen_gvec_muli(unsigned vece, uint32_t dofs, uint32_t aofs,
                        int64_t c, uint32_t oprsz, uint32_t maxsz)
 {
-    TCGv_i64 tmp = tcg_const_i64(c);
+    TCGv_i64 tmp = tcg_constant_i64(c);
     tcg_gen_gvec_muls(vece, dofs, aofs, tmp, oprsz, maxsz);
-    tcg_temp_free_i64(tmp);
 }
 
 void tcg_gen_gvec_ssadd(unsigned vece, uint32_t dofs, uint32_t aofs,
@@ -2076,18 +2058,16 @@ void tcg_gen_gvec_sssub(unsigned vece, uint32_t dofs, uint32_t aofs,
 
 static void tcg_gen_usadd_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b)
 {
-    TCGv_i32 max = tcg_const_i32(-1);
+    TCGv_i32 max = tcg_constant_i32(-1);
     tcg_gen_add_i32(d, a, b);
     tcg_gen_movcond_i32(TCG_COND_LTU, d, d, a, max, d);
-    tcg_temp_free_i32(max);
 }
 
 static void tcg_gen_usadd_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b)
 {
-    TCGv_i64 max = tcg_const_i64(-1);
+    TCGv_i64 max = tcg_constant_i64(-1);
     tcg_gen_add_i64(d, a, b);
     tcg_gen_movcond_i64(TCG_COND_LTU, d, d, a, max, d);
-    tcg_temp_free_i64(max);
 }
 
 void tcg_gen_gvec_usadd(unsigned vece, uint32_t dofs, uint32_t aofs,
@@ -2120,18 +2100,16 @@ void tcg_gen_gvec_usadd(unsigned vece, uint32_t dofs, uint32_t aofs,
 
 static void tcg_gen_ussub_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b)
 {
-    TCGv_i32 min = tcg_const_i32(0);
+    TCGv_i32 min = tcg_constant_i32(0);
     tcg_gen_sub_i32(d, a, b);
     tcg_gen_movcond_i32(TCG_COND_LTU, d, a, b, min, d);
-    tcg_temp_free_i32(min);
 }
 
 static void tcg_gen_ussub_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b)
 {
-    TCGv_i64 min = tcg_const_i64(0);
+    TCGv_i64 min = tcg_constant_i64(0);
     tcg_gen_sub_i64(d, a, b);
     tcg_gen_movcond_i64(TCG_COND_LTU, d, a, b, min, d);
-    tcg_temp_free_i64(min);
 }
 
 void tcg_gen_gvec_ussub(unsigned vece, uint32_t dofs, uint32_t aofs,
@@ -2292,16 +2270,14 @@ static void gen_negv_mask(TCGv_i64 d, TCGv_i64 b, TCGv_i64 m)
 
 void tcg_gen_vec_neg8_i64(TCGv_i64 d, TCGv_i64 b)
 {
-    TCGv_i64 m = tcg_const_i64(dup_const(MO_8, 0x80));
+    TCGv_i64 m = tcg_constant_i64(dup_const(MO_8, 0x80));
     gen_negv_mask(d, b, m);
-    tcg_temp_free_i64(m);
 }
 
 void tcg_gen_vec_neg16_i64(TCGv_i64 d, TCGv_i64 b)
 {
-    TCGv_i64 m = tcg_const_i64(dup_const(MO_16, 0x8000));
+    TCGv_i64 m = tcg_constant_i64(dup_const(MO_16, 0x8000));
     gen_negv_mask(d, b, m);
-    tcg_temp_free_i64(m);
 }
 
 void tcg_gen_vec_neg32_i64(TCGv_i64 d, TCGv_i64 b)
@@ -2570,9 +2546,8 @@ void tcg_gen_gvec_ands(unsigned vece, uint32_t dofs, uint32_t aofs,
 void tcg_gen_gvec_andi(unsigned vece, uint32_t dofs, uint32_t aofs,
                        int64_t c, uint32_t oprsz, uint32_t maxsz)
 {
-    TCGv_i64 tmp = tcg_const_i64(dup_const(vece, c));
+    TCGv_i64 tmp = tcg_constant_i64(dup_const(vece, c));
     tcg_gen_gvec_2s(dofs, aofs, oprsz, maxsz, tmp, &gop_ands);
-    tcg_temp_free_i64(tmp);
 }
 
 static const GVecGen2s gop_xors = {
@@ -2595,9 +2570,8 @@ void tcg_gen_gvec_xors(unsigned vece, uint32_t dofs, uint32_t aofs,
 void tcg_gen_gvec_xori(unsigned vece, uint32_t dofs, uint32_t aofs,
                        int64_t c, uint32_t oprsz, uint32_t maxsz)
 {
-    TCGv_i64 tmp = tcg_const_i64(dup_const(vece, c));
+    TCGv_i64 tmp = tcg_constant_i64(dup_const(vece, c));
     tcg_gen_gvec_2s(dofs, aofs, oprsz, maxsz, tmp, &gop_xors);
-    tcg_temp_free_i64(tmp);
 }
 
 static const GVecGen2s gop_ors = {
@@ -2620,9 +2594,8 @@ void tcg_gen_gvec_ors(unsigned vece, uint32_t dofs, uint32_t aofs,
 void tcg_gen_gvec_ori(unsigned vece, uint32_t dofs, uint32_t aofs,
                       int64_t c, uint32_t oprsz, uint32_t maxsz)
 {
-    TCGv_i64 tmp = tcg_const_i64(dup_const(vece, c));
+    TCGv_i64 tmp = tcg_constant_i64(dup_const(vece, c));
     tcg_gen_gvec_2s(dofs, aofs, oprsz, maxsz, tmp, &gop_ors);
-    tcg_temp_free_i64(tmp);
 }
 
 void tcg_gen_vec_shl8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c)
@@ -3110,9 +3083,9 @@ static void tcg_gen_shlv_mod_vec(unsigned vece, TCGv_vec d,
                                  TCGv_vec a, TCGv_vec b)
 {
     TCGv_vec t = tcg_temp_new_vec_matching(d);
+    TCGv_vec m = tcg_constant_vec_matching(d, vece, (8 << vece) - 1);
 
-    tcg_gen_dupi_vec(vece, t, (8 << vece) - 1);
-    tcg_gen_and_vec(vece, t, t, b);
+    tcg_gen_and_vec(vece, t, b, m);
     tcg_gen_shlv_vec(vece, d, a, t);
     tcg_temp_free_vec(t);
 }
@@ -3173,9 +3146,9 @@ static void tcg_gen_shrv_mod_vec(unsigned vece, TCGv_vec d,
                                  TCGv_vec a, TCGv_vec b)
 {
     TCGv_vec t = tcg_temp_new_vec_matching(d);
+    TCGv_vec m = tcg_constant_vec_matching(d, vece, (8 << vece) - 1);
 
-    tcg_gen_dupi_vec(vece, t, (8 << vece) - 1);
-    tcg_gen_and_vec(vece, t, t, b);
+    tcg_gen_and_vec(vece, t, b, m);
     tcg_gen_shrv_vec(vece, d, a, t);
     tcg_temp_free_vec(t);
 }
@@ -3236,9 +3209,9 @@ static void tcg_gen_sarv_mod_vec(unsigned vece, TCGv_vec d,
                                  TCGv_vec a, TCGv_vec b)
 {
     TCGv_vec t = tcg_temp_new_vec_matching(d);
+    TCGv_vec m = tcg_constant_vec_matching(d, vece, (8 << vece) - 1);
 
-    tcg_gen_dupi_vec(vece, t, (8 << vece) - 1);
-    tcg_gen_and_vec(vece, t, t, b);
+    tcg_gen_and_vec(vece, t, b, m);
     tcg_gen_sarv_vec(vece, d, a, t);
     tcg_temp_free_vec(t);
 }
@@ -3299,9 +3272,9 @@ static void tcg_gen_rotlv_mod_vec(unsigned vece, TCGv_vec d,
                                   TCGv_vec a, TCGv_vec b)
 {
     TCGv_vec t = tcg_temp_new_vec_matching(d);
+    TCGv_vec m = tcg_constant_vec_matching(d, vece, (8 << vece) - 1);
 
-    tcg_gen_dupi_vec(vece, t, (8 << vece) - 1);
-    tcg_gen_and_vec(vece, t, t, b);
+    tcg_gen_and_vec(vece, t, b, m);
     tcg_gen_rotlv_vec(vece, d, a, t);
     tcg_temp_free_vec(t);
 }
@@ -3358,9 +3331,9 @@ static void tcg_gen_rotrv_mod_vec(unsigned vece, TCGv_vec d,
                                   TCGv_vec a, TCGv_vec b)
 {
     TCGv_vec t = tcg_temp_new_vec_matching(d);
+    TCGv_vec m = tcg_constant_vec_matching(d, vece, (8 << vece) - 1);
 
-    tcg_gen_dupi_vec(vece, t, (8 << vece) - 1);
-    tcg_gen_and_vec(vece, t, t, b);
+    tcg_gen_and_vec(vece, t, b, m);
     tcg_gen_rotrv_vec(vece, d, a, t);
     tcg_temp_free_vec(t);
 }
diff --git a/tcg/tcg-op-vec.c b/tcg/tcg-op-vec.c
index cdbf11c573..d19aa7373e 100644
--- a/tcg/tcg-op-vec.c
+++ b/tcg/tcg-op-vec.c
@@ -83,7 +83,6 @@ bool tcg_can_emit_vecop_list(const TCGOpcode *list,
         case INDEX_op_xor_vec:
         case INDEX_op_mov_vec:
         case INDEX_op_dup_vec:
-        case INDEX_op_dupi_vec:
         case INDEX_op_dup2_vec:
         case INDEX_op_ld_vec:
         case INDEX_op_st_vec:
@@ -216,25 +215,17 @@ void tcg_gen_mov_vec(TCGv_vec r, TCGv_vec a)
     }
 }
 
-#define MO_REG  (TCG_TARGET_REG_BITS == 64 ? MO_64 : MO_32)
-
-static void do_dupi_vec(TCGv_vec r, unsigned vece, TCGArg a)
-{
-    TCGTemp *rt = tcgv_vec_temp(r);
-    vec_gen_2(INDEX_op_dupi_vec, rt->base_type, vece, temp_arg(rt), a);
-}
-
 TCGv_vec tcg_const_zeros_vec(TCGType type)
 {
     TCGv_vec ret = tcg_temp_new_vec(type);
-    do_dupi_vec(ret, MO_REG, 0);
+    tcg_gen_dupi_vec(MO_64, ret, 0);
     return ret;
 }
 
 TCGv_vec tcg_const_ones_vec(TCGType type)
 {
     TCGv_vec ret = tcg_temp_new_vec(type);
-    do_dupi_vec(ret, MO_REG, -1);
+    tcg_gen_dupi_vec(MO_64, ret, -1);
     return ret;
 }
 
@@ -250,41 +241,10 @@ TCGv_vec tcg_const_ones_vec_matching(TCGv_vec m)
     return tcg_const_ones_vec(t->base_type);
 }
 
-void tcg_gen_dup64i_vec(TCGv_vec r, uint64_t a)
-{
-    if (TCG_TARGET_REG_BITS == 64) {
-        do_dupi_vec(r, MO_64, a);
-    } else if (a == dup_const(MO_32, a)) {
-        do_dupi_vec(r, MO_32, a);
-    } else {
-        TCGv_i64 c = tcg_const_i64(a);
-        tcg_gen_dup_i64_vec(MO_64, r, c);
-        tcg_temp_free_i64(c);
-    }
-}
-
-void tcg_gen_dup32i_vec(TCGv_vec r, uint32_t a)
-{
-    do_dupi_vec(r, MO_REG, dup_const(MO_32, a));
-}
-
-void tcg_gen_dup16i_vec(TCGv_vec r, uint32_t a)
-{
-    do_dupi_vec(r, MO_REG, dup_const(MO_16, a));
-}
-
-void tcg_gen_dup8i_vec(TCGv_vec r, uint32_t a)
-{
-    do_dupi_vec(r, MO_REG, dup_const(MO_8, a));
-}
-
 void tcg_gen_dupi_vec(unsigned vece, TCGv_vec r, uint64_t a)
 {
-    if (vece == MO_64) {
-        tcg_gen_dup64i_vec(r, a);
-    } else {
-        do_dupi_vec(r, MO_REG, dup_const(vece, a));
-    }
+    TCGTemp *rt = tcgv_vec_temp(r);
+    tcg_gen_mov_vec(r, tcg_constant_vec(rt->base_type, vece, a));
 }
 
 void tcg_gen_dup_i64_vec(unsigned vece, TCGv_vec r, TCGv_i64 a)
@@ -489,8 +449,8 @@ void tcg_gen_abs_vec(unsigned vece, TCGv_vec r, TCGv_vec a)
             if (tcg_can_emit_vec_op(INDEX_op_sari_vec, type, vece) > 0) {
                 tcg_gen_sari_vec(vece, t, a, (8 << vece) - 1);
             } else {
-                do_dupi_vec(t, MO_REG, 0);
-                tcg_gen_cmp_vec(TCG_COND_LT, vece, t, a, t);
+                tcg_gen_cmp_vec(TCG_COND_LT, vece, t, a,
+                                tcg_constant_vec(type, vece, 0));
             }
             tcg_gen_xor_vec(vece, r, a, t);
             tcg_gen_sub_vec(vece, r, r, t);
diff --git a/tcg/tcg-op.c b/tcg/tcg-op.c
index 0374b5d56d..70475773f4 100644
--- a/tcg/tcg-op.c
+++ b/tcg/tcg-op.c
@@ -104,15 +104,18 @@ void tcg_gen_mb(TCGBar mb_type)
 
 /* 32 bit ops */
 
+void tcg_gen_movi_i32(TCGv_i32 ret, int32_t arg)
+{
+    tcg_gen_mov_i32(ret, tcg_constant_i32(arg));
+}
+
 void tcg_gen_addi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
 {
     /* some cases can be optimized here */
     if (arg2 == 0) {
         tcg_gen_mov_i32(ret, arg1);
     } else {
-        TCGv_i32 t0 = tcg_const_i32(arg2);
-        tcg_gen_add_i32(ret, arg1, t0);
-        tcg_temp_free_i32(t0);
+        tcg_gen_add_i32(ret, arg1, tcg_constant_i32(arg2));
     }
 }
 
@@ -122,9 +125,7 @@ void tcg_gen_subfi_i32(TCGv_i32 ret, int32_t arg1, TCGv_i32 arg2)
         /* Don't recurse with tcg_gen_neg_i32.  */
         tcg_gen_op2_i32(INDEX_op_neg_i32, ret, arg2);
     } else {
-        TCGv_i32 t0 = tcg_const_i32(arg1);
-        tcg_gen_sub_i32(ret, t0, arg2);
-        tcg_temp_free_i32(t0);
+        tcg_gen_sub_i32(ret, tcg_constant_i32(arg1), arg2);
     }
 }
 
@@ -134,15 +135,12 @@ void tcg_gen_subi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
     if (arg2 == 0) {
         tcg_gen_mov_i32(ret, arg1);
     } else {
-        TCGv_i32 t0 = tcg_const_i32(arg2);
-        tcg_gen_sub_i32(ret, arg1, t0);
-        tcg_temp_free_i32(t0);
+        tcg_gen_sub_i32(ret, arg1, tcg_constant_i32(arg2));
     }
 }
 
 void tcg_gen_andi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
 {
-    TCGv_i32 t0;
     /* Some cases can be optimized here.  */
     switch (arg2) {
     case 0:
@@ -165,9 +163,8 @@ void tcg_gen_andi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
         }
         break;
     }
-    t0 = tcg_const_i32(arg2);
-    tcg_gen_and_i32(ret, arg1, t0);
-    tcg_temp_free_i32(t0);
+
+    tcg_gen_and_i32(ret, arg1, tcg_constant_i32(arg2));
 }
 
 void tcg_gen_ori_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
@@ -178,9 +175,7 @@ void tcg_gen_ori_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
     } else if (arg2 == 0) {
         tcg_gen_mov_i32(ret, arg1);
     } else {
-        TCGv_i32 t0 = tcg_const_i32(arg2);
-        tcg_gen_or_i32(ret, arg1, t0);
-        tcg_temp_free_i32(t0);
+        tcg_gen_or_i32(ret, arg1, tcg_constant_i32(arg2));
     }
 }
 
@@ -193,9 +188,7 @@ void tcg_gen_xori_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
         /* Don't recurse with tcg_gen_not_i32.  */
         tcg_gen_op2_i32(INDEX_op_not_i32, ret, arg1);
     } else {
-        TCGv_i32 t0 = tcg_const_i32(arg2);
-        tcg_gen_xor_i32(ret, arg1, t0);
-        tcg_temp_free_i32(t0);
+        tcg_gen_xor_i32(ret, arg1, tcg_constant_i32(arg2));
     }
 }
 
@@ -205,9 +198,7 @@ void tcg_gen_shli_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
     if (arg2 == 0) {
         tcg_gen_mov_i32(ret, arg1);
     } else {
-        TCGv_i32 t0 = tcg_const_i32(arg2);
-        tcg_gen_shl_i32(ret, arg1, t0);
-        tcg_temp_free_i32(t0);
+        tcg_gen_shl_i32(ret, arg1, tcg_constant_i32(arg2));
     }
 }
 
@@ -217,9 +208,7 @@ void tcg_gen_shri_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
     if (arg2 == 0) {
         tcg_gen_mov_i32(ret, arg1);
     } else {
-        TCGv_i32 t0 = tcg_const_i32(arg2);
-        tcg_gen_shr_i32(ret, arg1, t0);
-        tcg_temp_free_i32(t0);
+        tcg_gen_shr_i32(ret, arg1, tcg_constant_i32(arg2));
     }
 }
 
@@ -229,9 +218,7 @@ void tcg_gen_sari_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
     if (arg2 == 0) {
         tcg_gen_mov_i32(ret, arg1);
     } else {
-        TCGv_i32 t0 = tcg_const_i32(arg2);
-        tcg_gen_sar_i32(ret, arg1, t0);
-        tcg_temp_free_i32(t0);
+        tcg_gen_sar_i32(ret, arg1, tcg_constant_i32(arg2));
     }
 }
 
@@ -250,9 +237,7 @@ void tcg_gen_brcondi_i32(TCGCond cond, TCGv_i32 arg1, int32_t arg2, TCGLabel *l)
     if (cond == TCG_COND_ALWAYS) {
         tcg_gen_br(l);
     } else if (cond != TCG_COND_NEVER) {
-        TCGv_i32 t0 = tcg_const_i32(arg2);
-        tcg_gen_brcond_i32(cond, arg1, t0, l);
-        tcg_temp_free_i32(t0);
+        tcg_gen_brcond_i32(cond, arg1, tcg_constant_i32(arg2), l);
     }
 }
 
@@ -271,9 +256,7 @@ void tcg_gen_setcond_i32(TCGCond cond, TCGv_i32 ret,
 void tcg_gen_setcondi_i32(TCGCond cond, TCGv_i32 ret,
                           TCGv_i32 arg1, int32_t arg2)
 {
-    TCGv_i32 t0 = tcg_const_i32(arg2);
-    tcg_gen_setcond_i32(cond, ret, arg1, t0);
-    tcg_temp_free_i32(t0);
+    tcg_gen_setcond_i32(cond, ret, arg1, tcg_constant_i32(arg2));
 }
 
 void tcg_gen_muli_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
@@ -283,9 +266,7 @@ void tcg_gen_muli_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
     } else if (is_power_of_2(arg2)) {
         tcg_gen_shli_i32(ret, arg1, ctz32(arg2));
     } else {
-        TCGv_i32 t0 = tcg_const_i32(arg2);
-        tcg_gen_mul_i32(ret, arg1, t0);
-        tcg_temp_free_i32(t0);
+        tcg_gen_mul_i32(ret, arg1, tcg_constant_i32(arg2));
     }
 }
 
@@ -433,9 +414,7 @@ void tcg_gen_clz_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2)
 
 void tcg_gen_clzi_i32(TCGv_i32 ret, TCGv_i32 arg1, uint32_t arg2)
 {
-    TCGv_i32 t = tcg_const_i32(arg2);
-    tcg_gen_clz_i32(ret, arg1, t);
-    tcg_temp_free_i32(t);
+    tcg_gen_clz_i32(ret, arg1, tcg_constant_i32(arg2));
 }
 
 void tcg_gen_ctz_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2)
@@ -468,10 +447,9 @@ void tcg_gen_ctz_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2)
             tcg_gen_clzi_i32(t, t, 32);
             tcg_gen_xori_i32(t, t, 31);
         }
-        z = tcg_const_i32(0);
+        z = tcg_constant_i32(0);
         tcg_gen_movcond_i32(TCG_COND_EQ, ret, arg1, z, arg2, t);
         tcg_temp_free_i32(t);
-        tcg_temp_free_i32(z);
     } else {
         gen_helper_ctz_i32(ret, arg1, arg2);
     }
@@ -487,9 +465,7 @@ void tcg_gen_ctzi_i32(TCGv_i32 ret, TCGv_i32 arg1, uint32_t arg2)
         tcg_gen_ctpop_i32(ret, t);
         tcg_temp_free_i32(t);
     } else {
-        TCGv_i32 t = tcg_const_i32(arg2);
-        tcg_gen_ctz_i32(ret, arg1, t);
-        tcg_temp_free_i32(t);
+        tcg_gen_ctz_i32(ret, arg1, tcg_constant_i32(arg2));
     }
 }
 
@@ -547,9 +523,7 @@ void tcg_gen_rotli_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
     if (arg2 == 0) {
         tcg_gen_mov_i32(ret, arg1);
     } else if (TCG_TARGET_HAS_rot_i32) {
-        TCGv_i32 t0 = tcg_const_i32(arg2);
-        tcg_gen_rotl_i32(ret, arg1, t0);
-        tcg_temp_free_i32(t0);
+        tcg_gen_rotl_i32(ret, arg1, tcg_constant_i32(arg2));
     } else {
         TCGv_i32 t0, t1;
         t0 = tcg_temp_new_i32();
@@ -653,9 +627,8 @@ void tcg_gen_deposit_z_i32(TCGv_i32 ret, TCGv_i32 arg,
         tcg_gen_andi_i32(ret, arg, (1u << len) - 1);
     } else if (TCG_TARGET_HAS_deposit_i32
                && TCG_TARGET_deposit_i32_valid(ofs, len)) {
-        TCGv_i32 zero = tcg_const_i32(0);
+        TCGv_i32 zero = tcg_constant_i32(0);
         tcg_gen_op5ii_i32(INDEX_op_deposit_i32, ret, zero, arg, ofs, len);
-        tcg_temp_free_i32(zero);
     } else {
         /* To help two-operand hosts we prefer to zero-extend first,
            which allows ARG to stay live.  */
@@ -1052,7 +1025,7 @@ void tcg_gen_bswap32_i32(TCGv_i32 ret, TCGv_i32 arg)
     } else {
         TCGv_i32 t0 = tcg_temp_new_i32();
         TCGv_i32 t1 = tcg_temp_new_i32();
-        TCGv_i32 t2 = tcg_const_i32(0x00ff00ff);
+        TCGv_i32 t2 = tcg_constant_i32(0x00ff00ff);
 
                                         /* arg = abcd */
         tcg_gen_shri_i32(t0, arg, 8);   /*  t0 = .abc */
@@ -1067,7 +1040,6 @@ void tcg_gen_bswap32_i32(TCGv_i32 ret, TCGv_i32 arg)
 
         tcg_temp_free_i32(t0);
         tcg_temp_free_i32(t1);
-        tcg_temp_free_i32(t2);
     }
 }
 
@@ -1114,8 +1086,15 @@ void tcg_gen_discard_i64(TCGv_i64 arg)
 
 void tcg_gen_mov_i64(TCGv_i64 ret, TCGv_i64 arg)
 {
-    tcg_gen_mov_i32(TCGV_LOW(ret), TCGV_LOW(arg));
-    tcg_gen_mov_i32(TCGV_HIGH(ret), TCGV_HIGH(arg));
+    TCGTemp *ts = tcgv_i64_temp(arg);
+
+    /* Canonicalize TCGv_i64 TEMP_CONST into TCGv_i32 TEMP_CONST. */
+    if (ts->kind == TEMP_CONST) {
+        tcg_gen_movi_i64(ret, ts->val);
+    } else {
+        tcg_gen_mov_i32(TCGV_LOW(ret), TCGV_LOW(arg));
+        tcg_gen_mov_i32(TCGV_HIGH(ret), TCGV_HIGH(arg));
+    }
 }
 
 void tcg_gen_movi_i64(TCGv_i64 ret, int64_t arg)
@@ -1237,6 +1216,14 @@ void tcg_gen_mul_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2)
     tcg_temp_free_i64(t0);
     tcg_temp_free_i32(t1);
 }
+
+#else
+
+void tcg_gen_movi_i64(TCGv_i64 ret, int64_t arg)
+{
+    tcg_gen_mov_i64(ret, tcg_constant_i64(arg));
+}
+
 #endif /* TCG_TARGET_REG_SIZE == 32 */
 
 void tcg_gen_addi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
@@ -1244,10 +1231,12 @@ void tcg_gen_addi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
     /* some cases can be optimized here */
     if (arg2 == 0) {
         tcg_gen_mov_i64(ret, arg1);
+    } else if (TCG_TARGET_REG_BITS == 64) {
+        tcg_gen_add_i64(ret, arg1, tcg_constant_i64(arg2));
     } else {
-        TCGv_i64 t0 = tcg_const_i64(arg2);
-        tcg_gen_add_i64(ret, arg1, t0);
-        tcg_temp_free_i64(t0);
+        tcg_gen_add2_i32(TCGV_LOW(ret), TCGV_HIGH(ret),
+                         TCGV_LOW(arg1), TCGV_HIGH(arg1),
+                         tcg_constant_i32(arg2), tcg_constant_i32(arg2 >> 32));
     }
 }
 
@@ -1256,10 +1245,12 @@ void tcg_gen_subfi_i64(TCGv_i64 ret, int64_t arg1, TCGv_i64 arg2)
     if (arg1 == 0 && TCG_TARGET_HAS_neg_i64) {
         /* Don't recurse with tcg_gen_neg_i64.  */
         tcg_gen_op2_i64(INDEX_op_neg_i64, ret, arg2);
+    } else if (TCG_TARGET_REG_BITS == 64) {
+        tcg_gen_sub_i64(ret, tcg_constant_i64(arg1), arg2);
     } else {
-        TCGv_i64 t0 = tcg_const_i64(arg1);
-        tcg_gen_sub_i64(ret, t0, arg2);
-        tcg_temp_free_i64(t0);
+        tcg_gen_sub2_i32(TCGV_LOW(ret), TCGV_HIGH(ret),
+                         tcg_constant_i32(arg1), tcg_constant_i32(arg1 >> 32),
+                         TCGV_LOW(arg2), TCGV_HIGH(arg2));
     }
 }
 
@@ -1268,17 +1259,17 @@ void tcg_gen_subi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
     /* some cases can be optimized here */
     if (arg2 == 0) {
         tcg_gen_mov_i64(ret, arg1);
+    } else if (TCG_TARGET_REG_BITS == 64) {
+        tcg_gen_sub_i64(ret, arg1, tcg_constant_i64(arg2));
     } else {
-        TCGv_i64 t0 = tcg_const_i64(arg2);
-        tcg_gen_sub_i64(ret, arg1, t0);
-        tcg_temp_free_i64(t0);
+        tcg_gen_sub2_i32(TCGV_LOW(ret), TCGV_HIGH(ret),
+                         TCGV_LOW(arg1), TCGV_HIGH(arg1),
+                         tcg_constant_i32(arg2), tcg_constant_i32(arg2 >> 32));
     }
 }
 
 void tcg_gen_andi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
 {
-    TCGv_i64 t0;
-
     if (TCG_TARGET_REG_BITS == 32) {
         tcg_gen_andi_i32(TCGV_LOW(ret), TCGV_LOW(arg1), arg2);
         tcg_gen_andi_i32(TCGV_HIGH(ret), TCGV_HIGH(arg1), arg2 >> 32);
@@ -1313,9 +1304,8 @@ void tcg_gen_andi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
         }
         break;
     }
-    t0 = tcg_const_i64(arg2);
-    tcg_gen_and_i64(ret, arg1, t0);
-    tcg_temp_free_i64(t0);
+
+    tcg_gen_and_i64(ret, arg1, tcg_constant_i64(arg2));
 }
 
 void tcg_gen_ori_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
@@ -1331,9 +1321,7 @@ void tcg_gen_ori_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
     } else if (arg2 == 0) {
         tcg_gen_mov_i64(ret, arg1);
     } else {
-        TCGv_i64 t0 = tcg_const_i64(arg2);
-        tcg_gen_or_i64(ret, arg1, t0);
-        tcg_temp_free_i64(t0);
+        tcg_gen_or_i64(ret, arg1, tcg_constant_i64(arg2));
     }
 }
 
@@ -1351,9 +1339,7 @@ void tcg_gen_xori_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
         /* Don't recurse with tcg_gen_not_i64.  */
         tcg_gen_op2_i64(INDEX_op_not_i64, ret, arg1);
     } else {
-        TCGv_i64 t0 = tcg_const_i64(arg2);
-        tcg_gen_xor_i64(ret, arg1, t0);
-        tcg_temp_free_i64(t0);
+        tcg_gen_xor_i64(ret, arg1, tcg_constant_i64(arg2));
     }
 }
 
@@ -1415,9 +1401,7 @@ void tcg_gen_shli_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
     } else if (arg2 == 0) {
         tcg_gen_mov_i64(ret, arg1);
     } else {
-        TCGv_i64 t0 = tcg_const_i64(arg2);
-        tcg_gen_shl_i64(ret, arg1, t0);
-        tcg_temp_free_i64(t0);
+        tcg_gen_shl_i64(ret, arg1, tcg_constant_i64(arg2));
     }
 }
 
@@ -1429,9 +1413,7 @@ void tcg_gen_shri_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
     } else if (arg2 == 0) {
         tcg_gen_mov_i64(ret, arg1);
     } else {
-        TCGv_i64 t0 = tcg_const_i64(arg2);
-        tcg_gen_shr_i64(ret, arg1, t0);
-        tcg_temp_free_i64(t0);
+        tcg_gen_shr_i64(ret, arg1, tcg_constant_i64(arg2));
     }
 }
 
@@ -1443,9 +1425,7 @@ void tcg_gen_sari_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
     } else if (arg2 == 0) {
         tcg_gen_mov_i64(ret, arg1);
     } else {
-        TCGv_i64 t0 = tcg_const_i64(arg2);
-        tcg_gen_sar_i64(ret, arg1, t0);
-        tcg_temp_free_i64(t0);
+        tcg_gen_sar_i64(ret, arg1, tcg_constant_i64(arg2));
     }
 }
 
@@ -1468,12 +1448,17 @@ void tcg_gen_brcond_i64(TCGCond cond, TCGv_i64 arg1, TCGv_i64 arg2, TCGLabel *l)
 
 void tcg_gen_brcondi_i64(TCGCond cond, TCGv_i64 arg1, int64_t arg2, TCGLabel *l)
 {
-    if (cond == TCG_COND_ALWAYS) {
+    if (TCG_TARGET_REG_BITS == 64) {
+        tcg_gen_brcond_i64(cond, arg1, tcg_constant_i64(arg2), l);
+    } else if (cond == TCG_COND_ALWAYS) {
         tcg_gen_br(l);
     } else if (cond != TCG_COND_NEVER) {
-        TCGv_i64 t0 = tcg_const_i64(arg2);
-        tcg_gen_brcond_i64(cond, arg1, t0, l);
-        tcg_temp_free_i64(t0);
+        l->refs++;
+        tcg_gen_op6ii_i32(INDEX_op_brcond2_i32,
+                          TCGV_LOW(arg1), TCGV_HIGH(arg1),
+                          tcg_constant_i32(arg2),
+                          tcg_constant_i32(arg2 >> 32),
+                          cond, label_arg(l));
     }
 }
 
@@ -1499,9 +1484,19 @@ void tcg_gen_setcond_i64(TCGCond cond, TCGv_i64 ret,
 void tcg_gen_setcondi_i64(TCGCond cond, TCGv_i64 ret,
                           TCGv_i64 arg1, int64_t arg2)
 {
-    TCGv_i64 t0 = tcg_const_i64(arg2);
-    tcg_gen_setcond_i64(cond, ret, arg1, t0);
-    tcg_temp_free_i64(t0);
+    if (TCG_TARGET_REG_BITS == 64) {
+        tcg_gen_setcond_i64(cond, ret, arg1, tcg_constant_i64(arg2));
+    } else if (cond == TCG_COND_ALWAYS) {
+        tcg_gen_movi_i64(ret, 1);
+    } else if (cond == TCG_COND_NEVER) {
+        tcg_gen_movi_i64(ret, 0);
+    } else {
+        tcg_gen_op6i_i32(INDEX_op_setcond2_i32, TCGV_LOW(ret),
+                         TCGV_LOW(arg1), TCGV_HIGH(arg1),
+                         tcg_constant_i32(arg2),
+                         tcg_constant_i32(arg2 >> 32), cond);
+        tcg_gen_movi_i32(TCGV_HIGH(ret), 0);
+    }
 }
 
 void tcg_gen_muli_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
@@ -1690,7 +1685,7 @@ void tcg_gen_bswap32_i64(TCGv_i64 ret, TCGv_i64 arg)
     } else {
         TCGv_i64 t0 = tcg_temp_new_i64();
         TCGv_i64 t1 = tcg_temp_new_i64();
-        TCGv_i64 t2 = tcg_const_i64(0x00ff00ff);
+        TCGv_i64 t2 = tcg_constant_i64(0x00ff00ff);
 
                                         /* arg = ....abcd */
         tcg_gen_shri_i64(t0, arg, 8);   /*  t0 = .....abc */
@@ -1706,7 +1701,6 @@ void tcg_gen_bswap32_i64(TCGv_i64 ret, TCGv_i64 arg)
 
         tcg_temp_free_i64(t0);
         tcg_temp_free_i64(t1);
-        tcg_temp_free_i64(t2);
     }
 }
 
@@ -1850,16 +1844,16 @@ void tcg_gen_clzi_i64(TCGv_i64 ret, TCGv_i64 arg1, uint64_t arg2)
     if (TCG_TARGET_REG_BITS == 32
         && TCG_TARGET_HAS_clz_i32
         && arg2 <= 0xffffffffu) {
-        TCGv_i32 t = tcg_const_i32((uint32_t)arg2 - 32);
-        tcg_gen_clz_i32(t, TCGV_LOW(arg1), t);
+        TCGv_i32 t = tcg_temp_new_i32();
+        tcg_gen_clzi_i32(t, TCGV_LOW(arg1), arg2 - 32);
         tcg_gen_addi_i32(t, t, 32);
         tcg_gen_clz_i32(TCGV_LOW(ret), TCGV_HIGH(arg1), t);
         tcg_gen_movi_i32(TCGV_HIGH(ret), 0);
         tcg_temp_free_i32(t);
     } else {
-        TCGv_i64 t = tcg_const_i64(arg2);
-        tcg_gen_clz_i64(ret, arg1, t);
-        tcg_temp_free_i64(t);
+        TCGv_i64 t0 = tcg_const_i64(arg2);
+        tcg_gen_clz_i64(ret, arg1, t0);
+        tcg_temp_free_i64(t0);
     }
 }
 
@@ -1881,7 +1875,7 @@ void tcg_gen_ctz_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2)
             tcg_gen_clzi_i64(t, t, 64);
             tcg_gen_xori_i64(t, t, 63);
         }
-        z = tcg_const_i64(0);
+        z = tcg_constant_i64(0);
         tcg_gen_movcond_i64(TCG_COND_EQ, ret, arg1, z, arg2, t);
         tcg_temp_free_i64(t);
         tcg_temp_free_i64(z);
@@ -1895,8 +1889,8 @@ void tcg_gen_ctzi_i64(TCGv_i64 ret, TCGv_i64 arg1, uint64_t arg2)
     if (TCG_TARGET_REG_BITS == 32
         && TCG_TARGET_HAS_ctz_i32
         && arg2 <= 0xffffffffu) {
-        TCGv_i32 t32 = tcg_const_i32((uint32_t)arg2 - 32);
-        tcg_gen_ctz_i32(t32, TCGV_HIGH(arg1), t32);
+        TCGv_i32 t32 = tcg_temp_new_i32();
+        tcg_gen_ctzi_i32(t32, TCGV_HIGH(arg1), arg2 - 32);
         tcg_gen_addi_i32(t32, t32, 32);
         tcg_gen_ctz_i32(TCGV_LOW(ret), TCGV_LOW(arg1), t32);
         tcg_gen_movi_i32(TCGV_HIGH(ret), 0);
@@ -1911,9 +1905,9 @@ void tcg_gen_ctzi_i64(TCGv_i64 ret, TCGv_i64 arg1, uint64_t arg2)
         tcg_gen_ctpop_i64(ret, t);
         tcg_temp_free_i64(t);
     } else {
-        TCGv_i64 t64 = tcg_const_i64(arg2);
-        tcg_gen_ctz_i64(ret, arg1, t64);
-        tcg_temp_free_i64(t64);
+        TCGv_i64 t0 = tcg_const_i64(arg2);
+        tcg_gen_ctz_i64(ret, arg1, t0);
+        tcg_temp_free_i64(t0);
     }
 }
 
@@ -1969,9 +1963,7 @@ void tcg_gen_rotli_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
     if (arg2 == 0) {
         tcg_gen_mov_i64(ret, arg1);
     } else if (TCG_TARGET_HAS_rot_i64) {
-        TCGv_i64 t0 = tcg_const_i64(arg2);
-        tcg_gen_rotl_i64(ret, arg1, t0);
-        tcg_temp_free_i64(t0);
+        tcg_gen_rotl_i64(ret, arg1, tcg_constant_i64(arg2));
     } else {
         TCGv_i64 t0, t1;
         t0 = tcg_temp_new_i64();
@@ -2089,9 +2081,8 @@ void tcg_gen_deposit_z_i64(TCGv_i64 ret, TCGv_i64 arg,
         tcg_gen_andi_i64(ret, arg, (1ull << len) - 1);
     } else if (TCG_TARGET_HAS_deposit_i64
                && TCG_TARGET_deposit_i64_valid(ofs, len)) {
-        TCGv_i64 zero = tcg_const_i64(0);
+        TCGv_i64 zero = tcg_constant_i64(0);
         tcg_gen_op5ii_i64(INDEX_op_deposit_i64, ret, zero, arg, ofs, len);
-        tcg_temp_free_i64(zero);
     } else {
         if (TCG_TARGET_REG_BITS == 32) {
             if (ofs >= 32) {
@@ -3117,9 +3108,8 @@ void tcg_gen_atomic_cmpxchg_i32(TCGv_i32 retv, TCGv addr, TCGv_i32 cmpv,
 
 #ifdef CONFIG_SOFTMMU
         {
-            TCGv_i32 oi = tcg_const_i32(make_memop_idx(memop & ~MO_SIGN, idx));
-            gen(retv, cpu_env, addr, cmpv, newv, oi);
-            tcg_temp_free_i32(oi);
+            TCGMemOpIdx oi = make_memop_idx(memop & ~MO_SIGN, idx);
+            gen(retv, cpu_env, addr, cmpv, newv, tcg_constant_i32(oi));
         }
 #else
         gen(retv, cpu_env, addr, cmpv, newv);
@@ -3162,9 +3152,8 @@ void tcg_gen_atomic_cmpxchg_i64(TCGv_i64 retv, TCGv addr, TCGv_i64 cmpv,
 
 #ifdef CONFIG_SOFTMMU
         {
-            TCGv_i32 oi = tcg_const_i32(make_memop_idx(memop, idx));
-            gen(retv, cpu_env, addr, cmpv, newv, oi);
-            tcg_temp_free_i32(oi);
+            TCGMemOpIdx oi = make_memop_idx(memop, idx);
+            gen(retv, cpu_env, addr, cmpv, newv, tcg_constant_i32(oi));
         }
 #else
         gen(retv, cpu_env, addr, cmpv, newv);
@@ -3226,9 +3215,8 @@ static void do_atomic_op_i32(TCGv_i32 ret, TCGv addr, TCGv_i32 val,
 
 #ifdef CONFIG_SOFTMMU
     {
-        TCGv_i32 oi = tcg_const_i32(make_memop_idx(memop & ~MO_SIGN, idx));
-        gen(ret, cpu_env, addr, val, oi);
-        tcg_temp_free_i32(oi);
+        TCGMemOpIdx oi = make_memop_idx(memop & ~MO_SIGN, idx);
+        gen(ret, cpu_env, addr, val, tcg_constant_i32(oi));
     }
 #else
     gen(ret, cpu_env, addr, val);
@@ -3272,9 +3260,8 @@ static void do_atomic_op_i64(TCGv_i64 ret, TCGv addr, TCGv_i64 val,
 
 #ifdef CONFIG_SOFTMMU
         {
-            TCGv_i32 oi = tcg_const_i32(make_memop_idx(memop & ~MO_SIGN, idx));
-            gen(ret, cpu_env, addr, val, oi);
-            tcg_temp_free_i32(oi);
+            TCGMemOpIdx oi = make_memop_idx(memop & ~MO_SIGN, idx);
+            gen(ret, cpu_env, addr, val, tcg_constant_i32(oi));
         }
 #else
         gen(ret, cpu_env, addr, val);
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 472bf1755b..8f8badb61c 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -117,8 +117,8 @@ static bool tcg_out_dup_vec(TCGContext *s, TCGType type, unsigned vece,
                             TCGReg dst, TCGReg src);
 static bool tcg_out_dupm_vec(TCGContext *s, TCGType type, unsigned vece,
                              TCGReg dst, TCGReg base, intptr_t offset);
-static void tcg_out_dupi_vec(TCGContext *s, TCGType type,
-                             TCGReg dst, tcg_target_long arg);
+static void tcg_out_dupi_vec(TCGContext *s, TCGType type, unsigned vece,
+                             TCGReg dst, int64_t arg);
 static void tcg_out_vec_op(TCGContext *s, TCGOpcode opc, unsigned vecl,
                            unsigned vece, const TCGArg *args,
                            const int *const_args);
@@ -133,8 +133,8 @@ static inline bool tcg_out_dupm_vec(TCGContext *s, TCGType type, unsigned vece,
 {
     g_assert_not_reached();
 }
-static inline void tcg_out_dupi_vec(TCGContext *s, TCGType type,
-                                    TCGReg dst, tcg_target_long arg)
+static inline void tcg_out_dupi_vec(TCGContext *s, TCGType type, unsigned vece,
+                                    TCGReg dst, int64_t arg)
 {
     g_assert_not_reached();
 }
@@ -1184,6 +1184,13 @@ void tcg_func_start(TCGContext *s)
     /* No temps have been previously allocated for size or locality.  */
     memset(s->free_temps, 0, sizeof(s->free_temps));
 
+    /* No constant temps have been previously allocated. */
+    for (int i = 0; i < TCG_TYPE_COUNT; ++i) {
+        if (s->const_table[i]) {
+            g_hash_table_remove_all(s->const_table[i]);
+        }
+    }
+
     s->nb_ops = 0;
     s->nb_labels = 0;
     s->current_frame_offset = s->frame_start;
@@ -1211,7 +1218,7 @@ static inline TCGTemp *tcg_global_alloc(TCGContext *s)
     tcg_debug_assert(s->nb_globals == s->nb_temps);
     s->nb_globals++;
     ts = tcg_temp_alloc(s);
-    ts->temp_global = 1;
+    ts->kind = TEMP_GLOBAL;
 
     return ts;
 }
@@ -1228,7 +1235,7 @@ static TCGTemp *tcg_global_reg_new_internal(TCGContext *s, TCGType type,
     ts = tcg_global_alloc(s);
     ts->base_type = type;
     ts->type = type;
-    ts->fixed_reg = 1;
+    ts->kind = TEMP_FIXED;
     ts->reg = reg;
     ts->name = name;
     tcg_regset_set_reg(s->reserved_regs, reg);
@@ -1255,13 +1262,19 @@ TCGTemp *tcg_global_mem_new_internal(TCGType type, TCGv_ptr base,
     bigendian = 1;
 #endif
 
-    if (!base_ts->fixed_reg) {
+    switch (base_ts->kind) {
+    case TEMP_FIXED:
+        break;
+    case TEMP_GLOBAL:
         /* We do not support double-indirect registers.  */
         tcg_debug_assert(!base_ts->indirect_reg);
         base_ts->indirect_base = 1;
         s->nb_indirects += (TCG_TARGET_REG_BITS == 32 && type == TCG_TYPE_I64
                             ? 2 : 1);
         indirect_reg = 1;
+        break;
+    default:
+        g_assert_not_reached();
     }
 
     if (TCG_TARGET_REG_BITS == 32 && type == TCG_TYPE_I64) {
@@ -1303,6 +1316,7 @@ TCGTemp *tcg_global_mem_new_internal(TCGType type, TCGv_ptr base,
 TCGTemp *tcg_temp_new_internal(TCGType type, bool temp_local)
 {
     TCGContext *s = tcg_ctx;
+    TCGTempKind kind = temp_local ? TEMP_LOCAL : TEMP_NORMAL;
     TCGTemp *ts;
     int idx, k;
 
@@ -1315,7 +1329,7 @@ TCGTemp *tcg_temp_new_internal(TCGType type, bool temp_local)
         ts = &s->temps[idx];
         ts->temp_allocated = 1;
         tcg_debug_assert(ts->base_type == type);
-        tcg_debug_assert(ts->temp_local == temp_local);
+        tcg_debug_assert(ts->kind == kind);
     } else {
         ts = tcg_temp_alloc(s);
         if (TCG_TARGET_REG_BITS == 32 && type == TCG_TYPE_I64) {
@@ -1324,18 +1338,18 @@ TCGTemp *tcg_temp_new_internal(TCGType type, bool temp_local)
             ts->base_type = type;
             ts->type = TCG_TYPE_I32;
             ts->temp_allocated = 1;
-            ts->temp_local = temp_local;
+            ts->kind = kind;
 
             tcg_debug_assert(ts2 == ts + 1);
             ts2->base_type = TCG_TYPE_I64;
             ts2->type = TCG_TYPE_I32;
             ts2->temp_allocated = 1;
-            ts2->temp_local = temp_local;
+            ts2->kind = kind;
         } else {
             ts->base_type = type;
             ts->type = type;
             ts->temp_allocated = 1;
-            ts->temp_local = temp_local;
+            ts->kind = kind;
         }
     }
 
@@ -1385,6 +1399,11 @@ void tcg_temp_free_internal(TCGTemp *ts)
     TCGContext *s = tcg_ctx;
     int k, idx;
 
+    /* In order to simplify users of tcg_constant_*, silently ignore free. */
+    if (ts->kind == TEMP_CONST) {
+        return;
+    }
+
 #if defined(CONFIG_DEBUG_TCG)
     s->temps_in_use--;
     if (s->temps_in_use < 0) {
@@ -1392,15 +1411,77 @@ void tcg_temp_free_internal(TCGTemp *ts)
     }
 #endif
 
-    tcg_debug_assert(ts->temp_global == 0);
+    tcg_debug_assert(ts->kind < TEMP_GLOBAL);
     tcg_debug_assert(ts->temp_allocated != 0);
     ts->temp_allocated = 0;
 
     idx = temp_idx(ts);
-    k = ts->base_type + (ts->temp_local ? TCG_TYPE_COUNT : 0);
+    k = ts->base_type + (ts->kind == TEMP_NORMAL ? 0 : TCG_TYPE_COUNT);
     set_bit(idx, s->free_temps[k].l);
 }
 
+TCGTemp *tcg_constant_internal(TCGType type, int64_t val)
+{
+    TCGContext *s = tcg_ctx;
+    GHashTable *h = s->const_table[type];
+    TCGTemp *ts;
+
+    if (h == NULL) {
+        h = g_hash_table_new(g_int64_hash, g_int64_equal);
+        s->const_table[type] = h;
+    }
+
+    ts = g_hash_table_lookup(h, &val);
+    if (ts == NULL) {
+        ts = tcg_temp_alloc(s);
+
+        if (TCG_TARGET_REG_BITS == 32 && type == TCG_TYPE_I64) {
+            TCGTemp *ts2 = tcg_temp_alloc(s);
+
+            ts->base_type = TCG_TYPE_I64;
+            ts->type = TCG_TYPE_I32;
+            ts->kind = TEMP_CONST;
+            ts->temp_allocated = 1;
+            /*
+             * Retain the full value of the 64-bit constant in the low
+             * part, so that the hash table works.  Actual uses will
+             * truncate the value to the low part.
+             */
+            ts->val = val;
+
+            tcg_debug_assert(ts2 == ts + 1);
+            ts2->base_type = TCG_TYPE_I64;
+            ts2->type = TCG_TYPE_I32;
+            ts2->kind = TEMP_CONST;
+            ts2->temp_allocated = 1;
+            ts2->val = val >> 32;
+        } else {
+            ts->base_type = type;
+            ts->type = type;
+            ts->kind = TEMP_CONST;
+            ts->temp_allocated = 1;
+            ts->val = val;
+        }
+        g_hash_table_insert(h, &ts->val, ts);
+    }
+
+    return ts;
+}
+
+TCGv_vec tcg_constant_vec(TCGType type, unsigned vece, int64_t val)
+{
+    val = dup_const(vece, val);
+    return temp_tcgv_vec(tcg_constant_internal(type, val));
+}
+
+TCGv_vec tcg_constant_vec_matching(TCGv_vec match, unsigned vece, int64_t val)
+{
+    TCGTemp *t = tcgv_vec_temp(match);
+
+    tcg_debug_assert(t->temp_allocated != 0);
+    return tcg_constant_vec(t->base_type, vece, val);
+}
+
 TCGv_i32 tcg_const_i32(int32_t val)
 {
     TCGv_i32 t0;
@@ -1483,7 +1564,6 @@ bool tcg_op_supported(TCGOpcode op)
         return TCG_TARGET_HAS_goto_ptr;
 
     case INDEX_op_mov_i32:
-    case INDEX_op_movi_i32:
     case INDEX_op_setcond_i32:
     case INDEX_op_brcond_i32:
     case INDEX_op_ld8u_i32:
@@ -1577,7 +1657,6 @@ bool tcg_op_supported(TCGOpcode op)
         return TCG_TARGET_REG_BITS == 32;
 
     case INDEX_op_mov_i64:
-    case INDEX_op_movi_i64:
     case INDEX_op_setcond_i64:
     case INDEX_op_brcond_i64:
     case INDEX_op_ld8u_i64:
@@ -1683,7 +1762,6 @@ bool tcg_op_supported(TCGOpcode op)
 
     case INDEX_op_mov_vec:
     case INDEX_op_dup_vec:
-    case INDEX_op_dupi_vec:
     case INDEX_op_dupm_vec:
     case INDEX_op_ld_vec:
     case INDEX_op_st_vec:
@@ -1930,17 +2008,30 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args)
 static void tcg_reg_alloc_start(TCGContext *s)
 {
     int i, n;
-    TCGTemp *ts;
 
-    for (i = 0, n = s->nb_globals; i < n; i++) {
-        ts = &s->temps[i];
-        ts->val_type = (ts->fixed_reg ? TEMP_VAL_REG : TEMP_VAL_MEM);
-    }
-    for (n = s->nb_temps; i < n; i++) {
-        ts = &s->temps[i];
-        ts->val_type = (ts->temp_local ? TEMP_VAL_MEM : TEMP_VAL_DEAD);
-        ts->mem_allocated = 0;
-        ts->fixed_reg = 0;
+    for (i = 0, n = s->nb_temps; i < n; i++) {
+        TCGTemp *ts = &s->temps[i];
+        TCGTempVal val = TEMP_VAL_MEM;
+
+        switch (ts->kind) {
+        case TEMP_CONST:
+            val = TEMP_VAL_CONST;
+            break;
+        case TEMP_FIXED:
+            val = TEMP_VAL_REG;
+            break;
+        case TEMP_GLOBAL:
+            break;
+        case TEMP_NORMAL:
+            val = TEMP_VAL_DEAD;
+            /* fall through */
+        case TEMP_LOCAL:
+            ts->mem_allocated = 0;
+            break;
+        default:
+            g_assert_not_reached();
+        }
+        ts->val_type = val;
     }
 
     memset(s->reg_to_temp, 0, sizeof(s->reg_to_temp));
@@ -1951,12 +2042,37 @@ static char *tcg_get_arg_str_ptr(TCGContext *s, char *buf, int buf_size,
 {
     int idx = temp_idx(ts);
 
-    if (ts->temp_global) {
+    switch (ts->kind) {
+    case TEMP_FIXED:
+    case TEMP_GLOBAL:
         pstrcpy(buf, buf_size, ts->name);
-    } else if (ts->temp_local) {
+        break;
+    case TEMP_LOCAL:
         snprintf(buf, buf_size, "loc%d", idx - s->nb_globals);
-    } else {
+        break;
+    case TEMP_NORMAL:
         snprintf(buf, buf_size, "tmp%d", idx - s->nb_globals);
+        break;
+    case TEMP_CONST:
+        switch (ts->type) {
+        case TCG_TYPE_I32:
+            snprintf(buf, buf_size, "$0x%x", (int32_t)ts->val);
+            break;
+#if TCG_TARGET_REG_BITS > 32
+        case TCG_TYPE_I64:
+            snprintf(buf, buf_size, "$0x%" PRIx64, ts->val);
+            break;
+#endif
+        case TCG_TYPE_V64:
+        case TCG_TYPE_V128:
+        case TCG_TYPE_V256:
+            snprintf(buf, buf_size, "v%d$0x%" PRIx64,
+                     64 << (ts->type - TCG_TYPE_V64), ts->val);
+            break;
+        default:
+            g_assert_not_reached();
+        }
+        break;
     }
     return buf;
 }
@@ -2547,15 +2663,25 @@ static void la_bb_end(TCGContext *s, int ng, int nt)
 {
     int i;
 
-    for (i = 0; i < ng; ++i) {
-        s->temps[i].state = TS_DEAD | TS_MEM;
-        la_reset_pref(&s->temps[i]);
-    }
-    for (i = ng; i < nt; ++i) {
-        s->temps[i].state = (s->temps[i].temp_local
-                             ? TS_DEAD | TS_MEM
-                             : TS_DEAD);
-        la_reset_pref(&s->temps[i]);
+    for (i = 0; i < nt; ++i) {
+        TCGTemp *ts = &s->temps[i];
+        int state;
+
+        switch (ts->kind) {
+        case TEMP_FIXED:
+        case TEMP_GLOBAL:
+        case TEMP_LOCAL:
+            state = TS_DEAD | TS_MEM;
+            break;
+        case TEMP_NORMAL:
+        case TEMP_CONST:
+            state = TS_DEAD;
+            break;
+        default:
+            g_assert_not_reached();
+        }
+        ts->state = state;
+        la_reset_pref(ts);
     }
 }
 
@@ -2583,14 +2709,24 @@ static void la_bb_sync(TCGContext *s, int ng, int nt)
     la_global_sync(s, ng);
 
     for (int i = ng; i < nt; ++i) {
-        if (s->temps[i].temp_local) {
-            int state = s->temps[i].state;
-            s->temps[i].state = state | TS_MEM;
+        TCGTemp *ts = &s->temps[i];
+        int state;
+
+        switch (ts->kind) {
+        case TEMP_LOCAL:
+            state = ts->state;
+            ts->state = state | TS_MEM;
             if (state != TS_DEAD) {
                 continue;
             }
-        } else {
+            break;
+        case TEMP_NORMAL:
             s->temps[i].state = TS_DEAD;
+            break;
+        case TEMP_CONST:
+            continue;
+        default:
+            g_assert_not_reached();
         }
         la_reset_pref(&s->temps[i]);
     }
@@ -3151,7 +3287,7 @@ static void dump_regs(TCGContext *s)
                    tcg_target_reg_names[ts->mem_base->reg]);
             break;
         case TEMP_VAL_CONST:
-            printf("$0x%" TCG_PRIlx, ts->val);
+            printf("$0x%" PRIx64, ts->val);
             break;
         case TEMP_VAL_DEAD:
             printf("D");
@@ -3191,7 +3327,8 @@ static void check_regs(TCGContext *s)
     }
     for (k = 0; k < s->nb_temps; k++) {
         ts = &s->temps[k];
-        if (ts->val_type == TEMP_VAL_REG && !ts->fixed_reg
+        if (ts->val_type == TEMP_VAL_REG
+            && ts->kind != TEMP_FIXED
             && s->reg_to_temp[ts->reg] != ts) {
             printf("Inconsistency for temp %s:\n",
                    tcg_get_arg_str_ptr(s, buf, sizeof(buf), ts));
@@ -3228,16 +3365,28 @@ static void temp_load(TCGContext *, TCGTemp *, TCGRegSet, TCGRegSet, TCGRegSet);
    mark it free; otherwise mark it dead.  */
 static void temp_free_or_dead(TCGContext *s, TCGTemp *ts, int free_or_dead)
 {
-    if (ts->fixed_reg) {
+    TCGTempVal new_type;
+
+    switch (ts->kind) {
+    case TEMP_FIXED:
         return;
+    case TEMP_GLOBAL:
+    case TEMP_LOCAL:
+        new_type = TEMP_VAL_MEM;
+        break;
+    case TEMP_NORMAL:
+        new_type = free_or_dead < 0 ? TEMP_VAL_MEM : TEMP_VAL_DEAD;
+        break;
+    case TEMP_CONST:
+        new_type = TEMP_VAL_CONST;
+        break;
+    default:
+        g_assert_not_reached();
     }
     if (ts->val_type == TEMP_VAL_REG) {
         s->reg_to_temp[ts->reg] = NULL;
     }
-    ts->val_type = (free_or_dead < 0
-                    || ts->temp_local
-                    || ts->temp_global
-                    ? TEMP_VAL_MEM : TEMP_VAL_DEAD);
+    ts->val_type = new_type;
 }
 
 /* Mark a temporary as dead.  */
@@ -3253,10 +3402,7 @@ static inline void temp_dead(TCGContext *s, TCGTemp *ts)
 static void temp_sync(TCGContext *s, TCGTemp *ts, TCGRegSet allocated_regs,
                       TCGRegSet preferred_regs, int free_or_dead)
 {
-    if (ts->fixed_reg) {
-        return;
-    }
-    if (!ts->mem_coherent) {
+    if (!temp_readonly(ts) && !ts->mem_coherent) {
         if (!ts->mem_allocated) {
             temp_allocate_frame(s, ts);
         }
@@ -3387,7 +3533,27 @@ static void temp_load(TCGContext *s, TCGTemp *ts, TCGRegSet desired_regs,
     case TEMP_VAL_CONST:
         reg = tcg_reg_alloc(s, desired_regs, allocated_regs,
                             preferred_regs, ts->indirect_base);
-        tcg_out_movi(s, ts->type, reg, ts->val);
+        if (ts->type <= TCG_TYPE_I64) {
+            tcg_out_movi(s, ts->type, reg, ts->val);
+        } else {
+            uint64_t val = ts->val;
+            MemOp vece = MO_64;
+
+            /*
+             * Find the minimal vector element that matches the constant.
+             * The targets will, in general, have to do this search anyway,
+             * do this generically.
+             */
+            if (val == dup_const(MO_8, val)) {
+                vece = MO_8;
+            } else if (val == dup_const(MO_16, val)) {
+                vece = MO_16;
+            } else if (val == dup_const(MO_32, val)) {
+                vece = MO_32;
+            }
+
+            tcg_out_dupi_vec(s, ts->type, vece, reg, ts->val);
+        }
         ts->mem_coherent = 0;
         break;
     case TEMP_VAL_MEM:
@@ -3411,7 +3577,7 @@ static void temp_save(TCGContext *s, TCGTemp *ts, TCGRegSet allocated_regs)
 {
     /* The liveness analysis already ensures that globals are back
        in memory. Keep an tcg_debug_assert for safety. */
-    tcg_debug_assert(ts->val_type == TEMP_VAL_MEM || ts->fixed_reg);
+    tcg_debug_assert(ts->val_type == TEMP_VAL_MEM || temp_readonly(ts));
 }
 
 /* save globals to their canonical location and assume they can be
@@ -3436,7 +3602,7 @@ static void sync_globals(TCGContext *s, TCGRegSet allocated_regs)
     for (i = 0, n = s->nb_globals; i < n; i++) {
         TCGTemp *ts = &s->temps[i];
         tcg_debug_assert(ts->val_type != TEMP_VAL_REG
-                         || ts->fixed_reg
+                         || ts->kind == TEMP_FIXED
                          || ts->mem_coherent);
     }
 }
@@ -3449,12 +3615,22 @@ static void tcg_reg_alloc_bb_end(TCGContext *s, TCGRegSet allocated_regs)
 
     for (i = s->nb_globals; i < s->nb_temps; i++) {
         TCGTemp *ts = &s->temps[i];
-        if (ts->temp_local) {
+
+        switch (ts->kind) {
+        case TEMP_LOCAL:
             temp_save(s, ts, allocated_regs);
-        } else {
+            break;
+        case TEMP_NORMAL:
             /* The liveness analysis already ensures that temps are dead.
                Keep an tcg_debug_assert for safety. */
             tcg_debug_assert(ts->val_type == TEMP_VAL_DEAD);
+            break;
+        case TEMP_CONST:
+            /* Similarly, we should have freed any allocated register. */
+            tcg_debug_assert(ts->val_type == TEMP_VAL_CONST);
+            break;
+        default:
+            g_assert_not_reached();
         }
     }
 
@@ -3475,23 +3651,30 @@ static void tcg_reg_alloc_cbranch(TCGContext *s, TCGRegSet allocated_regs)
          * The liveness analysis already ensures that temps are dead.
          * Keep tcg_debug_asserts for safety.
          */
-        if (ts->temp_local) {
+        switch (ts->kind) {
+        case TEMP_LOCAL:
             tcg_debug_assert(ts->val_type != TEMP_VAL_REG || ts->mem_coherent);
-        } else {
+            break;
+        case TEMP_NORMAL:
             tcg_debug_assert(ts->val_type == TEMP_VAL_DEAD);
+            break;
+        case TEMP_CONST:
+            break;
+        default:
+            g_assert_not_reached();
         }
     }
 }
 
 /*
- * Specialized code generation for INDEX_op_movi_*.
+ * Specialized code generation for INDEX_op_mov_* with a constant.
  */
 static void tcg_reg_alloc_do_movi(TCGContext *s, TCGTemp *ots,
                                   tcg_target_ulong val, TCGLifeData arg_life,
                                   TCGRegSet preferred_regs)
 {
     /* ENV should not be modified.  */
-    tcg_debug_assert(!ots->fixed_reg);
+    tcg_debug_assert(!temp_readonly(ots));
 
     /* The movi is not explicitly generated here.  */
     if (ots->val_type == TEMP_VAL_REG) {
@@ -3507,14 +3690,6 @@ static void tcg_reg_alloc_do_movi(TCGContext *s, TCGTemp *ots,
     }
 }
 
-static void tcg_reg_alloc_movi(TCGContext *s, const TCGOp *op)
-{
-    TCGTemp *ots = arg_temp(op->args[0]);
-    tcg_target_ulong val = op->args[1];
-
-    tcg_reg_alloc_do_movi(s, ots, val, op->life, op->output_pref[0]);
-}
-
 /*
  * Specialized code generation for INDEX_op_mov_*.
  */
@@ -3531,7 +3706,7 @@ static void tcg_reg_alloc_mov(TCGContext *s, const TCGOp *op)
     ts = arg_temp(op->args[1]);
 
     /* ENV should not be modified.  */
-    tcg_debug_assert(!ots->fixed_reg);
+    tcg_debug_assert(!temp_readonly(ots));
 
     /* Note that otype != itype for no-op truncation.  */
     otype = ots->type;
@@ -3570,7 +3745,7 @@ static void tcg_reg_alloc_mov(TCGContext *s, const TCGOp *op)
         }
         temp_dead(s, ots);
     } else {
-        if (IS_DEAD_ARG(1) && !ts->fixed_reg) {
+        if (IS_DEAD_ARG(1) && ts->kind != TEMP_FIXED) {
             /* the mov can be suppressed */
             if (ots->val_type == TEMP_VAL_REG) {
                 s->reg_to_temp[ots->reg] = NULL;
@@ -3592,7 +3767,7 @@ static void tcg_reg_alloc_mov(TCGContext *s, const TCGOp *op)
                  * Store the source register into the destination slot
                  * and leave the destination temp as TEMP_VAL_MEM.
                  */
-                assert(!ots->fixed_reg);
+                assert(!temp_readonly(ots));
                 if (!ts->mem_allocated) {
                     temp_allocate_frame(s, ots);
                 }
@@ -3629,7 +3804,7 @@ static void tcg_reg_alloc_dup(TCGContext *s, const TCGOp *op)
     its = arg_temp(op->args[1]);
 
     /* ENV should not be modified.  */
-    tcg_debug_assert(!ots->fixed_reg);
+    tcg_debug_assert(!temp_readonly(ots));
 
     itype = its->type;
     vece = TCGOP_VECE(op);
@@ -3769,45 +3944,42 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
         i_preferred_regs = o_preferred_regs = 0;
         if (arg_ct->ialias) {
             o_preferred_regs = op->output_pref[arg_ct->alias_index];
-            if (ts->fixed_reg) {
-                /* if fixed register, we must allocate a new register
-                   if the alias is not the same register */
-                if (arg != op->args[arg_ct->alias_index]) {
-                    goto allocate_in_reg;
-                }
-            } else {
-                /* if the input is aliased to an output and if it is
-                   not dead after the instruction, we must allocate
-                   a new register and move it */
-                if (!IS_DEAD_ARG(i)) {
-                    goto allocate_in_reg;
-                }
 
-                /* check if the current register has already been allocated
-                   for another input aliased to an output */
-                if (ts->val_type == TEMP_VAL_REG) {
-                    int k2, i2;
-                    reg = ts->reg;
-                    for (k2 = 0 ; k2 < k ; k2++) {
-                        i2 = def->args_ct[nb_oargs + k2].sort_index;
-                        if (def->args_ct[i2].ialias && reg == new_args[i2]) {
-                            goto allocate_in_reg;
-                        }
+            /*
+             * If the input is readonly, then it cannot also be an
+             * output and aliased to itself.  If the input is not
+             * dead after the instruction, we must allocate a new
+             * register and move it.
+             */
+            if (temp_readonly(ts) || !IS_DEAD_ARG(i)) {
+                goto allocate_in_reg;
+            }
+
+            /*
+             * Check if the current register has already been allocated
+             * for another input aliased to an output.
+             */
+            if (ts->val_type == TEMP_VAL_REG) {
+                reg = ts->reg;
+                for (int k2 = 0; k2 < k; k2++) {
+                    int i2 = def->args_ct[nb_oargs + k2].sort_index;
+                    if (def->args_ct[i2].ialias && reg == new_args[i2]) {
+                        goto allocate_in_reg;
                     }
                 }
-                i_preferred_regs = o_preferred_regs;
             }
+            i_preferred_regs = o_preferred_regs;
         }
 
         temp_load(s, ts, arg_ct->regs, i_allocated_regs, i_preferred_regs);
         reg = ts->reg;
 
-        if (tcg_regset_test_reg(arg_ct->regs, reg)) {
-            /* nothing to do : the constraint is satisfied */
-        } else {
-        allocate_in_reg:
-            /* allocate a new register matching the constraint 
-               and move the temporary register into it */
+        if (!tcg_regset_test_reg(arg_ct->regs, reg)) {
+ allocate_in_reg:
+            /*
+             * Allocate a new register matching the constraint
+             * and move the temporary register into it.
+             */
             temp_load(s, ts, tcg_target_available_regs[ts->type],
                       i_allocated_regs, 0);
             reg = tcg_reg_alloc(s, arg_ct->regs, i_allocated_regs,
@@ -3861,7 +4033,7 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
             ts = arg_temp(arg);
 
             /* ENV should not be modified.  */
-            tcg_debug_assert(!ts->fixed_reg);
+            tcg_debug_assert(!temp_readonly(ts));
 
             if (arg_ct->oalias && !const_args[arg_ct->alias_index]) {
                 reg = new_args[arg_ct->alias_index];
@@ -3902,7 +4074,7 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
         ts = arg_temp(op->args[i]);
 
         /* ENV should not be modified.  */
-        tcg_debug_assert(!ts->fixed_reg);
+        tcg_debug_assert(!temp_readonly(ts));
 
         if (NEED_SYNC_ARG(i)) {
             temp_sync(s, ts, o_allocated_regs, 0, IS_DEAD_ARG(i));
@@ -3912,6 +4084,98 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
     }
 }
 
+static bool tcg_reg_alloc_dup2(TCGContext *s, const TCGOp *op)
+{
+    const TCGLifeData arg_life = op->life;
+    TCGTemp *ots, *itsl, *itsh;
+    TCGType vtype = TCGOP_VECL(op) + TCG_TYPE_V64;
+
+    /* This opcode is only valid for 32-bit hosts, for 64-bit elements. */
+    tcg_debug_assert(TCG_TARGET_REG_BITS == 32);
+    tcg_debug_assert(TCGOP_VECE(op) == MO_64);
+
+    ots = arg_temp(op->args[0]);
+    itsl = arg_temp(op->args[1]);
+    itsh = arg_temp(op->args[2]);
+
+    /* ENV should not be modified.  */
+    tcg_debug_assert(!temp_readonly(ots));
+
+    /* Allocate the output register now.  */
+    if (ots->val_type != TEMP_VAL_REG) {
+        TCGRegSet allocated_regs = s->reserved_regs;
+        TCGRegSet dup_out_regs =
+            tcg_op_defs[INDEX_op_dup_vec].args_ct[0].regs;
+
+        /* Make sure to not spill the input registers. */
+        if (!IS_DEAD_ARG(1) && itsl->val_type == TEMP_VAL_REG) {
+            tcg_regset_set_reg(allocated_regs, itsl->reg);
+        }
+        if (!IS_DEAD_ARG(2) && itsh->val_type == TEMP_VAL_REG) {
+            tcg_regset_set_reg(allocated_regs, itsh->reg);
+        }
+
+        ots->reg = tcg_reg_alloc(s, dup_out_regs, allocated_regs,
+                                 op->output_pref[0], ots->indirect_base);
+        ots->val_type = TEMP_VAL_REG;
+        ots->mem_coherent = 0;
+        s->reg_to_temp[ots->reg] = ots;
+    }
+
+    /* Promote dup2 of immediates to dupi_vec. */
+    if (itsl->val_type == TEMP_VAL_CONST && itsh->val_type == TEMP_VAL_CONST) {
+        uint64_t val = deposit64(itsl->val, 32, 32, itsh->val);
+        MemOp vece = MO_64;
+
+        if (val == dup_const(MO_8, val)) {
+            vece = MO_8;
+        } else if (val == dup_const(MO_16, val)) {
+            vece = MO_16;
+        } else if (val == dup_const(MO_32, val)) {
+            vece = MO_32;
+        }
+
+        tcg_out_dupi_vec(s, vtype, vece, ots->reg, val);
+        goto done;
+    }
+
+    /* If the two inputs form one 64-bit value, try dupm_vec. */
+    if (itsl + 1 == itsh && itsl->base_type == TCG_TYPE_I64) {
+        if (!itsl->mem_coherent) {
+            temp_sync(s, itsl, s->reserved_regs, 0, 0);
+        }
+        if (!itsh->mem_coherent) {
+            temp_sync(s, itsh, s->reserved_regs, 0, 0);
+        }
+#ifdef HOST_WORDS_BIGENDIAN
+        TCGTemp *its = itsh;
+#else
+        TCGTemp *its = itsl;
+#endif
+        if (tcg_out_dupm_vec(s, vtype, MO_64, ots->reg,
+                             its->mem_base->reg, its->mem_offset)) {
+            goto done;
+        }
+    }
+
+    /* Fall back to generic expansion. */
+    return false;
+
+ done:
+    if (IS_DEAD_ARG(1)) {
+        temp_dead(s, itsl);
+    }
+    if (IS_DEAD_ARG(2)) {
+        temp_dead(s, itsh);
+    }
+    if (NEED_SYNC_ARG(0)) {
+        temp_sync(s, ots, s->reserved_regs, 0, IS_DEAD_ARG(0));
+    } else if (IS_DEAD_ARG(0)) {
+        temp_dead(s, ots);
+    }
+    return true;
+}
+
 #ifdef TCG_TARGET_STACK_GROWSUP
 #define STACK_DIR(x) (-(x))
 #else
@@ -4034,7 +4298,7 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
         ts = arg_temp(arg);
 
         /* ENV should not be modified.  */
-        tcg_debug_assert(!ts->fixed_reg);
+        tcg_debug_assert(!temp_readonly(ts));
 
         reg = tcg_target_call_oarg_regs[i];
         tcg_debug_assert(s->reg_to_temp[reg] == NULL);
@@ -4298,11 +4562,6 @@ int tcg_gen_code(TCGContext *s, TranslationBlock *tb)
         case INDEX_op_mov_vec:
             tcg_reg_alloc_mov(s, op);
             break;
-        case INDEX_op_movi_i32:
-        case INDEX_op_movi_i64:
-        case INDEX_op_dupi_vec:
-            tcg_reg_alloc_movi(s, op);
-            break;
         case INDEX_op_dup_vec:
             tcg_reg_alloc_dup(s, op);
             break;
@@ -4334,6 +4593,11 @@ int tcg_gen_code(TCGContext *s, TranslationBlock *tb)
         case INDEX_op_call:
             tcg_reg_alloc_call(s, op);
             break;
+        case INDEX_op_dup2_vec:
+            if (tcg_reg_alloc_dup2(s, op)) {
+                break;
+            }
+            /* fall through */
         default:
             /* Sanity check that we've not introduced any unhandled opcodes. */
             tcg_debug_assert(tcg_op_supported(opc));
diff --git a/tcg/tci.c b/tcg/tci.c
index 2bcc4409be..2311aa7d3a 100644
--- a/tcg/tci.c
+++ b/tcg/tci.c
@@ -593,7 +593,7 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
             t1 = tci_read_r32(regs, &tb_ptr);
             tci_write_reg32(regs, t0, t1);
             break;
-        case INDEX_op_movi_i32:
+        case INDEX_op_tci_movi_i32:
             t0 = *tb_ptr++;
             t1 = tci_read_i32(&tb_ptr);
             tci_write_reg32(regs, t0, t1);
@@ -864,7 +864,7 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
             t1 = tci_read_r64(regs, &tb_ptr);
             tci_write_reg64(regs, t0, t1);
             break;
-        case INDEX_op_movi_i64:
+        case INDEX_op_tci_movi_i64:
             t0 = *tb_ptr++;
             t1 = tci_read_i64(&tb_ptr);
             tci_write_reg64(regs, t0, t1);
diff --git a/tcg/tci/tcg-target.c.inc b/tcg/tci/tcg-target.c.inc
index d5a4d9d37c..15981265db 100644
--- a/tcg/tci/tcg-target.c.inc
+++ b/tcg/tci/tcg-target.c.inc
@@ -529,13 +529,13 @@ static void tcg_out_movi(TCGContext *s, TCGType type,
     uint8_t *old_code_ptr = s->code_ptr;
     uint32_t arg32 = arg;
     if (type == TCG_TYPE_I32 || arg == arg32) {
-        tcg_out_op_t(s, INDEX_op_movi_i32);
+        tcg_out_op_t(s, INDEX_op_tci_movi_i32);
         tcg_out_r(s, t0);
         tcg_out32(s, arg32);
     } else {
         tcg_debug_assert(type == TCG_TYPE_I64);
 #if TCG_TARGET_REG_BITS == 64
-        tcg_out_op_t(s, INDEX_op_movi_i64);
+        tcg_out_op_t(s, INDEX_op_tci_movi_i64);
         tcg_out_r(s, t0);
         tcg_out64(s, arg);
 #else
@@ -814,8 +814,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args,
         break;
     case INDEX_op_mov_i32:  /* Always emitted via tcg_out_mov.  */
     case INDEX_op_mov_i64:
-    case INDEX_op_movi_i32: /* Always emitted via tcg_out_movi.  */
-    case INDEX_op_movi_i64:
     case INDEX_op_call:     /* Always emitted via tcg_out_call.  */
     default:
         tcg_abort();