Patchwork [1/2] tcg: Add TYPE parameter to tcg_out_mov.

login
register
mail settings
Submitter Richard Henderson
Date June 3, 2010, 12:26 a.m.
Message ID <1275524816-4187-2-git-send-email-rth@twiddle.net>
Download mbox | patch
Permalink /patch/54441/
State New
Headers show

Comments

Richard Henderson - June 3, 2010, 12:26 a.m.
Mirror tcg_out_movi in having a TYPE parameter.  This allows x86_64
to perform the move at the proper width, which may elide a REX prefix.

Introduce a TCG_TYPE_REG enumerator to represent the "native width"
of the host register, and to distinguish the usage from "pointer data"
as represented by the existing TCG_TYPE_PTR.

Update all targets to match.

Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/arm/tcg-target.c    |    2 +-
 tcg/hppa/tcg-target.c   |   38 ++++++++++++++++++------------------
 tcg/i386/tcg-target.c   |   36 +++++++++++++++++-----------------
 tcg/ia64/tcg-target.c   |    3 +-
 tcg/mips/tcg-target.c   |   28 +++++++++++++-------------
 tcg/ppc/tcg-target.c    |   48 +++++++++++++++++++++++-----------------------
 tcg/ppc64/tcg-target.c  |   10 ++++----
 tcg/s390/tcg-target.c   |    2 +-
 tcg/sparc/tcg-target.c  |   10 ++++----
 tcg/tcg.c               |   12 +++++-----
 tcg/tcg.h               |   11 ++++++++-
 tcg/x86_64/tcg-target.c |   45 +++++++++++++++++--------------------------
 12 files changed, 122 insertions(+), 123 deletions(-)
Blue Swirl - June 4, 2010, 7:19 p.m.
On Thu, Jun 3, 2010 at 12:26 AM, Richard Henderson <rth@twiddle.net> wrote:
> Mirror tcg_out_movi in having a TYPE parameter.  This allows x86_64
> to perform the move at the proper width, which may elide a REX prefix.
>
> Introduce a TCG_TYPE_REG enumerator to represent the "native width"
> of the host register, and to distinguish the usage from "pointer data"
> as represented by the existing TCG_TYPE_PTR.
>
> Update all targets to match.
>
> Signed-off-by: Richard Henderson <rth@twiddle.net>

Sparc part looks fine to me.

> +
> +    /* An alias for the size of the native pointer.  We don't currently
> +       support any hosts with 64-bit registers and 32-bit pointers.  */

Sparc32plus would match, though the ABI does not guarantee that higher
32 bits are preserved.
Richard Henderson - June 4, 2010, 7:34 p.m.
On 06/04/2010 12:19 PM, Blue Swirl wrote:
> Sparc32plus would match, though the ABI does not guarantee that higher
> 32 bits are preserved.

Actually, it does for %g and %o registers.  See section 2.1.1.

I've thought before about how to adjust the __sparcv8plus__ TCG
to use full 64-bit registers and operations when required, but
have not yet gotten around to it.

All of sparc, mips, ppc, and s390 have a mode with 32-bit pointers
running on a machine with 64-bit registers and operations available.

Unfortunately the hppa linux kernel doesn't save the full 64-bit 
register on interrupt, so we're strictly limited to pa1.1 there,
even when running on a pa2 machine.

Something for the future...


r~

Patch

diff --git a/tcg/arm/tcg-target.c b/tcg/arm/tcg-target.c
index 8d23f47..b3169a9 100644
--- a/tcg/arm/tcg-target.c
+++ b/tcg/arm/tcg-target.c
@@ -1798,7 +1798,7 @@  static void tcg_out_addi(TCGContext *s, int reg, tcg_target_long val)
     }
 }
 
-static inline void tcg_out_mov(TCGContext *s, int ret, int arg)
+static inline void tcg_out_mov(TCGContext *s, TCGType type, int ret, int arg)
 {
     tcg_out_dat_reg(s, COND_AL, ARITH_MOV, ret, 0, arg, SHIFT_IMM_LSL(0));
 }
diff --git a/tcg/hppa/tcg-target.c b/tcg/hppa/tcg-target.c
index 558c21f..a5f2162 100644
--- a/tcg/hppa/tcg-target.c
+++ b/tcg/hppa/tcg-target.c
@@ -338,7 +338,7 @@  static int tcg_target_const_match(tcg_target_long val,
 /* supplied by libgcc */
 extern void *__canonicalize_funcptr_for_compare(void *);
 
-static void tcg_out_mov(TCGContext *s, int ret, int arg)
+static void tcg_out_mov(TCGContext *s, TCGType type, int ret, int arg)
 {
     /* PA1.1 defines COPY as OR r,0,t; PA2.0 defines COPY as LDO 0(r),t
        but hppa-dis.c is unaware of this definition */
@@ -498,7 +498,7 @@  static void tcg_out_ori(TCGContext *s, int ret, int arg, tcg_target_ulong m)
     }
     assert(bs1 == 32 || (1ul << bs1) > m);
 
-    tcg_out_mov(s, ret, arg);
+    tcg_out_mov(s, TCG_TYPE_I32, ret, arg);
     tcg_out32(s, INSN_DEPI | INSN_R2(ret) | INSN_IM5(-1)
               | INSN_SHDEP_CP(31 - bs0) | INSN_DEP_LEN(bs1 - bs0));
 }
@@ -528,7 +528,7 @@  static void tcg_out_andi(TCGContext *s, int ret, int arg, tcg_target_ulong m)
     if (ls1 == 32) {
         tcg_out_extr(s, ret, arg, 0, ls0, 0);
     } else {
-        tcg_out_mov(s, ret, arg);
+        tcg_out_mov(s, TCG_TYPE_I32, ret, arg);
         tcg_out32(s, INSN_DEPI | INSN_R2(ret) | INSN_IM5(0)
                   | INSN_SHDEP_CP(31 - ls0) | INSN_DEP_LEN(ls1 - ls0));
     }
@@ -608,7 +608,7 @@  static void tcg_out_rotr(TCGContext *s, int ret, int arg, int creg)
 static void tcg_out_bswap16(TCGContext *s, int ret, int arg, int sign)
 {
     if (ret != arg) {
-        tcg_out_mov(s, ret, arg);             /* arg =  xxAB */
+        tcg_out_mov(s, TCG_TYPE_I32, ret, arg); /* arg =  xxAB */
     }
     tcg_out_dep(s, ret, ret, 16, 8);          /* ret =  xBAB */
     tcg_out_extr(s, ret, ret, 8, 16, sign);   /* ret =  ..BA */
@@ -638,7 +638,7 @@  static void tcg_out_call(TCGContext *s, void *func)
         tcg_out32(s, INSN_LDIL | INSN_R2(TCG_REG_R20) | reassemble_21(hi));
         tcg_out32(s, INSN_BLE_SR4 | INSN_R2(TCG_REG_R20)
                   | reassemble_17(lo >> 2));
-        tcg_out_mov(s, TCG_REG_RP, TCG_REG_R31);
+        tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_RP, TCG_REG_R31);
     }
 }
 
@@ -685,7 +685,7 @@  static void tcg_out_add2(TCGContext *s, int destl, int desth,
     }
     tcg_out_arith(s, desth, ah, bh, INSN_ADDC);
 
-    tcg_out_mov(s, destl, tmp);
+    tcg_out_mov(s, TCG_TYPE_I32, destl, tmp);
 }
 
 static void tcg_out_sub2(TCGContext *s, int destl, int desth, int al, int ah,
@@ -706,7 +706,7 @@  static void tcg_out_sub2(TCGContext *s, int destl, int desth, int al, int ah,
     }
     tcg_out_arith(s, desth, ah, bh, INSN_SUBB);
 
-    tcg_out_mov(s, destl, tmp);
+    tcg_out_mov(s, TCG_TYPE_I32, destl, tmp);
 }
 
 static void tcg_out_branch(TCGContext *s, int label_index, int nul)
@@ -869,7 +869,7 @@  static void tcg_out_setcond2(TCGContext *s, int cond, TCGArg ret,
         break;
     }
 
-    tcg_out_mov(s, ret, scratch);
+    tcg_out_mov(s, TCG_TYPE_I32, ret, scratch);
 }
 
 #if defined(CONFIG_SOFTMMU)
@@ -1048,9 +1048,9 @@  static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int opc)
     tcg_out_label(s, lab1, (tcg_target_long)s->code_ptr);
 
     argreg = TCG_REG_R26;
-    tcg_out_mov(s, argreg--, addrlo_reg);
+    tcg_out_mov(s, TCG_TYPE_I32, argreg--, addrlo_reg);
     if (TARGET_LONG_BITS == 64) {
-        tcg_out_mov(s, argreg--, addrhi_reg);
+        tcg_out_mov(s, TCG_TYPE_I32, argreg--, addrhi_reg);
     }
     tcg_out_movi(s, TCG_TYPE_I32, argreg, mem_index);
 
@@ -1071,11 +1071,11 @@  static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int opc)
         break;
     case 2:
     case 2 | 4:
-        tcg_out_mov(s, datalo_reg, TCG_REG_RET0);
+        tcg_out_mov(s, TCG_TYPE_I32, datalo_reg, TCG_REG_RET0);
         break;
     case 3:
-        tcg_out_mov(s, datahi_reg, TCG_REG_RET0);
-        tcg_out_mov(s, datalo_reg, TCG_REG_RET1);
+        tcg_out_mov(s, TCG_TYPE_I32, datahi_reg, TCG_REG_RET0);
+        tcg_out_mov(s, TCG_TYPE_I32, datalo_reg, TCG_REG_RET1);
         break;
     default:
         tcg_abort();
@@ -1167,9 +1167,9 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int opc)
     tcg_out_label(s, lab1, (tcg_target_long)s->code_ptr);
 
     argreg = TCG_REG_R26;
-    tcg_out_mov(s, argreg--, addrlo_reg);
+    tcg_out_mov(s, TCG_TYPE_I32, argreg--, addrlo_reg);
     if (TARGET_LONG_BITS == 64) {
-        tcg_out_mov(s, argreg--, addrhi_reg);
+        tcg_out_mov(s, TCG_TYPE_I32, argreg--, addrhi_reg);
     }
 
     switch(opc) {
@@ -1182,7 +1182,7 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int opc)
         tcg_out_movi(s, TCG_TYPE_I32, argreg, mem_index);
         break;
     case 2:
-        tcg_out_mov(s, argreg--, datalo_reg);
+        tcg_out_mov(s, TCG_TYPE_I32, argreg--, datalo_reg);
         tcg_out_movi(s, TCG_TYPE_I32, argreg, mem_index);
         break;
     case 3:
@@ -1196,8 +1196,8 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int opc)
             argreg = TCG_REG_R20;
             tcg_out_movi(s, TCG_TYPE_I32, argreg, mem_index);
         }
-        tcg_out_mov(s, TCG_REG_R23, datahi_reg);
-        tcg_out_mov(s, TCG_REG_R24, datalo_reg);
+        tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_R23, datahi_reg);
+        tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_R24, datalo_reg);
         tcg_out_st(s, TCG_TYPE_I32, argreg, TCG_REG_SP,
                    TCG_TARGET_CALL_STACK_OFFSET - 4);
         break;
@@ -1637,7 +1637,7 @@  void tcg_target_qemu_prologue(TCGContext *s)
 
     /* Jump to TB, and adjust R18 to be the return address.  */
     tcg_out32(s, INSN_BLE_SR4 | INSN_R2(TCG_REG_R26));
-    tcg_out_mov(s, TCG_REG_R18, TCG_REG_R31);
+    tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_R18, TCG_REG_R31);
 
     /* Restore callee saved registers.  */
     tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_RP, TCG_REG_SP, -frame_size - 20);
diff --git a/tcg/i386/tcg-target.c b/tcg/i386/tcg-target.c
index 3600c35..8745ad1 100644
--- a/tcg/i386/tcg-target.c
+++ b/tcg/i386/tcg-target.c
@@ -348,7 +348,7 @@  static inline void tgen_arithr(TCGContext *s, int subop, int dest, int src)
     tcg_out_modrm(s, OPC_ARITH_GvEv + (subop << 3), dest, src);
 }
 
-static inline void tcg_out_mov(TCGContext *s, int ret, int arg)
+static inline void tcg_out_mov(TCGContext *s, TCGType type, int ret, int arg)
 {
     if (arg != ret) {
         tcg_out_modrm(s, OPC_MOVL_GvEv, ret, arg);
@@ -733,8 +733,8 @@  static inline void tcg_out_tlb_load(TCGContext *s, int addrlo_idx,
     const int r0 = TCG_REG_EAX;
     const int r1 = TCG_REG_EDX;
 
-    tcg_out_mov(s, r1, addrlo);
-    tcg_out_mov(s, r0, addrlo);
+    tcg_out_mov(s, TCG_TYPE_I32, r1, addrlo);
+    tcg_out_mov(s, TCG_TYPE_I32, r0, addrlo);
 
     tcg_out_shifti(s, SHIFT_SHR, r1, TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS);
 
@@ -748,7 +748,7 @@  static inline void tcg_out_tlb_load(TCGContext *s, int addrlo_idx,
     /* cmp 0(r1), r0 */
     tcg_out_modrm_offset(s, OPC_CMP_GvEv, r0, r1, 0);
 
-    tcg_out_mov(s, r0, addrlo);
+    tcg_out_mov(s, TCG_TYPE_I32, r0, addrlo);
 
     /* jne label1 */
     tcg_out8(s, OPC_JCC_short + JCC_JNE);
@@ -881,7 +881,7 @@  static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args,
     /* EAX is already loaded.  */
     arg_idx = 1;
     if (TARGET_LONG_BITS == 64) {
-        tcg_out_mov(s, tcg_target_call_iarg_regs[arg_idx++],
+        tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[arg_idx++],
                     args[addrlo_idx + 1]);
     }
     tcg_out_movi(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[arg_idx],
@@ -903,16 +903,16 @@  static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args,
         break;
     case 2:
     default:
-        tcg_out_mov(s, data_reg, TCG_REG_EAX);
+        tcg_out_mov(s, TCG_TYPE_I32, data_reg, TCG_REG_EAX);
         break;
     case 3:
         if (data_reg == TCG_REG_EDX) {
             /* xchg %edx, %eax */
             tcg_out_opc(s, OPC_XCHG_ax_r32 + TCG_REG_EDX);
-            tcg_out_mov(s, data_reg2, TCG_REG_EAX);
+            tcg_out_mov(s, TCG_TYPE_I32, data_reg2, TCG_REG_EAX);
         } else {
-            tcg_out_mov(s, data_reg, TCG_REG_EAX);
-            tcg_out_mov(s, data_reg2, TCG_REG_EDX);
+            tcg_out_mov(s, TCG_TYPE_I32, data_reg, TCG_REG_EAX);
+            tcg_out_mov(s, TCG_TYPE_I32, data_reg2, TCG_REG_EDX);
         }
         break;
     }
@@ -945,7 +945,7 @@  static void tcg_out_qemu_st_direct(TCGContext *s, int datalo, int datahi,
         break;
     case 1:
         if (bswap) {
-            tcg_out_mov(s, scratch, datalo);
+            tcg_out_mov(s, TCG_TYPE_I32, scratch, datalo);
             tcg_out_rolw_8(s, scratch);
             datalo = scratch;
         }
@@ -955,7 +955,7 @@  static void tcg_out_qemu_st_direct(TCGContext *s, int datalo, int datahi,
         break;
     case 2:
         if (bswap) {
-            tcg_out_mov(s, scratch, datalo);
+            tcg_out_mov(s, TCG_TYPE_I32, scratch, datalo);
             tcg_out_bswap32(s, scratch);
             datalo = scratch;
         }
@@ -963,10 +963,10 @@  static void tcg_out_qemu_st_direct(TCGContext *s, int datalo, int datahi,
         break;
     case 3:
         if (bswap) {
-            tcg_out_mov(s, scratch, datahi);
+            tcg_out_mov(s, TCG_TYPE_I32, scratch, datahi);
             tcg_out_bswap32(s, scratch);
             tcg_out_st(s, TCG_TYPE_I32, scratch, base, ofs);
-            tcg_out_mov(s, scratch, datalo);
+            tcg_out_mov(s, TCG_TYPE_I32, scratch, datalo);
             tcg_out_bswap32(s, scratch);
             tcg_out_st(s, TCG_TYPE_I32, scratch, base, ofs + 4);
         } else {
@@ -1022,9 +1022,9 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args,
 
     /* XXX: move that code at the end of the TB */
     if (TARGET_LONG_BITS == 32) {
-        tcg_out_mov(s, TCG_REG_EDX, data_reg);
+        tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_EDX, data_reg);
         if (opc == 3) {
-            tcg_out_mov(s, TCG_REG_ECX, data_reg2);
+            tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_ECX, data_reg2);
             tcg_out_pushi(s, mem_index);
             stack_adjust = 4;
         } else {
@@ -1033,13 +1033,13 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args,
         }
     } else {
         if (opc == 3) {
-            tcg_out_mov(s, TCG_REG_EDX, args[addrlo_idx + 1]);
+            tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_EDX, args[addrlo_idx + 1]);
             tcg_out_pushi(s, mem_index);
             tcg_out_push(s, data_reg2);
             tcg_out_push(s, data_reg);
             stack_adjust = 12;
         } else {
-            tcg_out_mov(s, TCG_REG_EDX, args[addrlo_idx + 1]);
+            tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_EDX, args[addrlo_idx + 1]);
             switch(opc) {
             case 0:
                 tcg_out_ext8u(s, TCG_REG_ECX, data_reg);
@@ -1048,7 +1048,7 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args,
                 tcg_out_ext16u(s, TCG_REG_ECX, data_reg);
                 break;
             case 2:
-                tcg_out_mov(s, TCG_REG_ECX, data_reg);
+                tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_ECX, data_reg);
                 break;
             }
             tcg_out_pushi(s, mem_index);
diff --git a/tcg/ia64/tcg-target.c b/tcg/ia64/tcg-target.c
index 905f48b..0d275e9 100644
--- a/tcg/ia64/tcg-target.c
+++ b/tcg/ia64/tcg-target.c
@@ -827,7 +827,8 @@  static inline void tcg_out_bundle(TCGContext *s, int template,
     s->code_ptr += 16;
 }
 
-static inline void tcg_out_mov(TCGContext *s, TCGArg ret, TCGArg arg)
+static inline void tcg_out_mov(TCGContext *s, TCGType type,
+                               TCGArg ret, TCGArg arg)
 {
     tcg_out_bundle(s, mmI,
                    tcg_opc_m48(TCG_REG_P0, OPC_NOP_M48, 0),
diff --git a/tcg/mips/tcg-target.c b/tcg/mips/tcg-target.c
index f38eb28..8d9c12f 100644
--- a/tcg/mips/tcg-target.c
+++ b/tcg/mips/tcg-target.c
@@ -377,7 +377,7 @@  static inline void tcg_out_nop(TCGContext *s)
     tcg_out32(s, 0);
 }
 
-static inline void tcg_out_mov(TCGContext *s, int ret, int arg)
+static inline void tcg_out_mov(TCGContext *s, TCGType type, int ret, int arg)
 {
     tcg_out_opc_reg(s, OPC_ADDU, ret, arg, TCG_REG_ZERO);
 }
@@ -849,9 +849,9 @@  static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args,
 
     /* slow path */
     sp_args = TCG_REG_A0;
-    tcg_out_mov(s, sp_args++, addr_reg1);
+    tcg_out_mov(s, TCG_TYPE_I32, sp_args++, addr_reg1);
 # if TARGET_LONG_BITS == 64
-    tcg_out_mov(s, sp_args++, addr_reg2);
+    tcg_out_mov(s, TCG_TYPE_I32, sp_args++, addr_reg2);
 # endif
     tcg_out_movi(s, TCG_TYPE_I32, sp_args++, mem_index);
     tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_T9, (tcg_target_long)qemu_ld_helpers[s_bits]);
@@ -872,11 +872,11 @@  static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args,
         tcg_out_ext16s(s, data_reg1, TCG_REG_V0);
         break;
     case 2:
-        tcg_out_mov(s, data_reg1, TCG_REG_V0);
+        tcg_out_mov(s, TCG_TYPE_I32, data_reg1, TCG_REG_V0);
         break;
     case 3:
-        tcg_out_mov(s, data_reg2, TCG_REG_V1);
-        tcg_out_mov(s, data_reg1, TCG_REG_V0);
+        tcg_out_mov(s, TCG_TYPE_I32, data_reg2, TCG_REG_V1);
+        tcg_out_mov(s, TCG_TYPE_I32, data_reg1, TCG_REG_V0);
         break;
     default:
         tcg_abort();
@@ -1035,9 +1035,9 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args,
 
     /* slow path */
     sp_args = TCG_REG_A0;
-    tcg_out_mov(s, sp_args++, addr_reg1);
+    tcg_out_mov(s, TCG_TYPE_I32, sp_args++, addr_reg1);
 # if TARGET_LONG_BITS == 64
-    tcg_out_mov(s, sp_args++, addr_reg2);
+    tcg_out_mov(s, TCG_TYPE_I32, sp_args++, addr_reg2);
 # endif
     switch(opc) {
     case 0:
@@ -1047,12 +1047,12 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args,
         tcg_out_opc_imm(s, OPC_ANDI, sp_args++, data_reg1, 0xffff);
         break;
     case 2:
-        tcg_out_mov(s, sp_args++, data_reg1);
+        tcg_out_mov(s, TCG_TYPE_I32, sp_args++, data_reg1);
         break;
     case 3:
         sp_args = (sp_args + 1) & ~1;
-        tcg_out_mov(s, sp_args++, data_reg1);
-        tcg_out_mov(s, sp_args++, data_reg2);
+        tcg_out_mov(s, TCG_TYPE_I32, sp_args++, data_reg1);
+        tcg_out_mov(s, TCG_TYPE_I32, sp_args++, data_reg2);
         break;
     default:
         tcg_abort();
@@ -1165,7 +1165,7 @@  static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
         break;
 
     case INDEX_op_mov_i32:
-        tcg_out_mov(s, args[0], args[1]);
+        tcg_out_mov(s, TCG_TYPE_I32, args[0], args[1]);
         break;
     case INDEX_op_movi_i32:
         tcg_out_movi(s, TCG_TYPE_I32, args[0], args[1]);
@@ -1216,7 +1216,7 @@  static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
              tcg_out_opc_reg(s, OPC_ADDU, args[1], args[3], args[5]);
         }
         tcg_out_opc_reg(s, OPC_ADDU, args[1], args[1], TCG_REG_T0);
-        tcg_out_mov(s, args[0], TCG_REG_AT);
+        tcg_out_mov(s, TCG_TYPE_I32, args[0], TCG_REG_AT);
         break;
     case INDEX_op_sub_i32:
         if (const_args[2]) {
@@ -1238,7 +1238,7 @@  static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
              tcg_out_opc_reg(s, OPC_SUBU, args[1], args[3], args[5]);
         }
         tcg_out_opc_reg(s, OPC_SUBU, args[1], args[1], TCG_REG_T0);
-        tcg_out_mov(s, args[0], TCG_REG_AT);
+        tcg_out_mov(s, TCG_TYPE_I32, args[0], TCG_REG_AT);
         break;
     case INDEX_op_mul_i32:
         tcg_out_opc_reg(s, OPC_MULT, 0, args[1], args[2]);
diff --git a/tcg/ppc/tcg-target.c b/tcg/ppc/tcg-target.c
index 2b85928..ce078e4 100644
--- a/tcg/ppc/tcg-target.c
+++ b/tcg/ppc/tcg-target.c
@@ -437,7 +437,7 @@  static const uint32_t tcg_to_bc[10] = {
     [TCG_COND_GTU] = BC | BI (7, CR_GT) | BO_COND_TRUE,
 };
 
-static void tcg_out_mov(TCGContext *s, int ret, int arg)
+static void tcg_out_mov(TCGContext *s, TCGType type, int ret, int arg)
 {
     tcg_out32 (s, OR | SAB (arg, ret, arg));
 }
@@ -591,11 +591,11 @@  static void tcg_out_qemu_ld (TCGContext *s, const TCGArg *args, int opc)
 
     /* slow path */
 #if TARGET_LONG_BITS == 32
-    tcg_out_mov (s, 3, addr_reg);
+    tcg_out_mov (s, TCG_TYPE_I32, 3, addr_reg);
     tcg_out_movi (s, TCG_TYPE_I32, 4, mem_index);
 #else
-    tcg_out_mov (s, 3, addr_reg2);
-    tcg_out_mov (s, 4, addr_reg);
+    tcg_out_mov (s, TCG_TYPE_I32, 3, addr_reg2);
+    tcg_out_mov (s, TCG_TYPE_I32, 4, addr_reg);
     tcg_out_movi (s, TCG_TYPE_I32, 5, mem_index);
 #endif
 
@@ -611,23 +611,23 @@  static void tcg_out_qemu_ld (TCGContext *s, const TCGArg *args, int opc)
     case 1:
     case 2:
         if (data_reg != 3)
-            tcg_out_mov (s, data_reg, 3);
+            tcg_out_mov (s, TCG_TYPE_I32, data_reg, 3);
         break;
     case 3:
         if (data_reg == 3) {
             if (data_reg2 == 4) {
-                tcg_out_mov (s, 0, 4);
-                tcg_out_mov (s, 4, 3);
-                tcg_out_mov (s, 3, 0);
+                tcg_out_mov (s, TCG_TYPE_I32, 0, 4);
+                tcg_out_mov (s, TCG_TYPE_I32, 4, 3);
+                tcg_out_mov (s, TCG_TYPE_I32, 3, 0);
             }
             else {
-                tcg_out_mov (s, data_reg2, 3);
-                tcg_out_mov (s, 3, 4);
+                tcg_out_mov (s, TCG_TYPE_I32, data_reg2, 3);
+                tcg_out_mov (s, TCG_TYPE_I32, 3, 4);
             }
         }
         else {
-            if (data_reg != 4) tcg_out_mov (s, data_reg, 4);
-            if (data_reg2 != 3) tcg_out_mov (s, data_reg2, 3);
+            if (data_reg != 4) tcg_out_mov (s, TCG_TYPE_I32, data_reg, 4);
+            if (data_reg2 != 3) tcg_out_mov (s, TCG_TYPE_I32, data_reg2, 3);
         }
         break;
     }
@@ -705,7 +705,7 @@  static void tcg_out_qemu_ld (TCGContext *s, const TCGArg *args, int opc)
             if (r0 == data_reg2) {
                 tcg_out32 (s, LWZ | RT (0) | RA (r0));
                 tcg_out32 (s, LWZ | RT (data_reg) | RA (r0) | 4);
-                tcg_out_mov (s, data_reg2, 0);
+                tcg_out_mov (s, TCG_TYPE_I32, data_reg2, 0);
             }
             else {
                 tcg_out32 (s, LWZ | RT (data_reg2) | RA (r0));
@@ -787,11 +787,11 @@  static void tcg_out_qemu_st (TCGContext *s, const TCGArg *args, int opc)
 
     /* slow path */
 #if TARGET_LONG_BITS == 32
-    tcg_out_mov (s, 3, addr_reg);
+    tcg_out_mov (s, TCG_TYPE_I32, 3, addr_reg);
     ir = 4;
 #else
-    tcg_out_mov (s, 3, addr_reg2);
-    tcg_out_mov (s, 4, addr_reg);
+    tcg_out_mov (s, TCG_TYPE_I32, 3, addr_reg2);
+    tcg_out_mov (s, TCG_TYPE_I32, 4, addr_reg);
 #ifdef TCG_TARGET_CALL_ALIGN_ARGS
     ir = 5;
 #else
@@ -817,14 +817,14 @@  static void tcg_out_qemu_st (TCGContext *s, const TCGArg *args, int opc)
                        | ME (31)));
         break;
     case 2:
-        tcg_out_mov (s, ir, data_reg);
+        tcg_out_mov (s, TCG_TYPE_I32, ir, data_reg);
         break;
     case 3:
 #ifdef TCG_TARGET_CALL_ALIGN_ARGS
         ir = 5;
 #endif
-        tcg_out_mov (s, ir++, data_reg2);
-        tcg_out_mov (s, ir, data_reg);
+        tcg_out_mov (s, TCG_TYPE_I32, ir++, data_reg2);
+        tcg_out_mov (s, TCG_TYPE_I32, ir, data_reg);
         break;
     }
     ir++;
@@ -1526,7 +1526,7 @@  static void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args,
         if (args[0] == args[2] || args[0] == args[3]) {
             tcg_out32 (s, MULLW | TAB (0, args[2], args[3]));
             tcg_out32 (s, MULHWU | TAB (args[1], args[2], args[3]));
-            tcg_out_mov (s, args[0], 0);
+            tcg_out_mov (s, TCG_TYPE_I32, args[0], 0);
         }
         else {
             tcg_out32 (s, MULLW | TAB (args[0], args[2], args[3]));
@@ -1584,7 +1584,7 @@  static void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args,
     case INDEX_op_rotr_i32:
         if (const_args[2]) {
             if (!args[2]) {
-                tcg_out_mov (s, args[0], args[1]);
+                tcg_out_mov (s, TCG_TYPE_I32, args[0], args[1]);
             }
             else {
                 tcg_out32 (s, RLWINM
@@ -1612,7 +1612,7 @@  static void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args,
         if (args[0] == args[3] || args[0] == args[5]) {
             tcg_out32 (s, ADDC | TAB (0, args[2], args[4]));
             tcg_out32 (s, ADDE | TAB (args[1], args[3], args[5]));
-            tcg_out_mov (s, args[0], 0);
+            tcg_out_mov (s, TCG_TYPE_I32, args[0], 0);
         }
         else {
             tcg_out32 (s, ADDC | TAB (args[0], args[2], args[4]));
@@ -1623,7 +1623,7 @@  static void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args,
         if (args[0] == args[3] || args[0] == args[5]) {
             tcg_out32 (s, SUBFC | TAB (0, args[4], args[2]));
             tcg_out32 (s, SUBFE | TAB (args[1], args[5], args[3]));
-            tcg_out_mov (s, args[0], 0);
+            tcg_out_mov (s, TCG_TYPE_I32, args[0], 0);
         }
         else {
             tcg_out32 (s, SUBFC | TAB (args[0], args[4], args[2]));
@@ -1782,7 +1782,7 @@  static void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args,
                 );
 
             if (!a0) {
-                tcg_out_mov (s, args[0], a0);
+                tcg_out_mov (s, TCG_TYPE_I32, args[0], a0);
             }
         }
         break;
diff --git a/tcg/ppc64/tcg-target.c b/tcg/ppc64/tcg-target.c
index 0b6c61f..2d436a5 100644
--- a/tcg/ppc64/tcg-target.c
+++ b/tcg/ppc64/tcg-target.c
@@ -435,7 +435,7 @@  static const uint32_t tcg_to_bc[10] = {
     [TCG_COND_GTU] = BC | BI (7, CR_GT) | BO_COND_TRUE,
 };
 
-static void tcg_out_mov (TCGContext *s, int ret, int arg)
+static void tcg_out_mov (TCGContext *s, TCGType type, int ret, int arg)
 {
     tcg_out32 (s, OR | SAB (arg, ret, arg));
 }
@@ -644,7 +644,7 @@  static void tcg_out_qemu_ld (TCGContext *s, const TCGArg *args, int opc)
 #endif
 
     /* slow path */
-    tcg_out_mov (s, 3, addr_reg);
+    tcg_out_mov (s, TCG_TYPE_I64, 3, addr_reg);
     tcg_out_movi (s, TCG_TYPE_I64, 4, mem_index);
 
     tcg_out_call (s, (tcg_target_long) qemu_ld_helpers[s_bits], 1);
@@ -664,7 +664,7 @@  static void tcg_out_qemu_ld (TCGContext *s, const TCGArg *args, int opc)
     case 2:
     case 3:
         if (data_reg != 3)
-            tcg_out_mov (s, data_reg, 3);
+            tcg_out_mov (s, TCG_TYPE_I64, data_reg, 3);
         break;
     }
     label2_ptr = s->code_ptr;
@@ -746,7 +746,7 @@  static void tcg_out_qemu_ld (TCGContext *s, const TCGArg *args, int opc)
         else tcg_out32 (s, LDX | TAB (data_reg, rbase, r0));
 #else
         if (bswap) {
-            tcg_out_movi32 (s, 0, 4);
+            tcg_out_movi32 (s, TCG_TYPE_I64, 0, 4);
             tcg_out32 (s, LWBRX | RT (data_reg) | RB (r0));
             tcg_out32 (s, LWBRX | RT (      r1) | RA (r0));
             tcg_out_rld (s, RLDIMI, data_reg, r1, 32, 0);
@@ -790,7 +790,7 @@  static void tcg_out_qemu_st (TCGContext *s, const TCGArg *args, int opc)
 #endif
 
     /* slow path */
-    tcg_out_mov (s, 3, addr_reg);
+    tcg_out_mov (s, TCG_TYPE_I64, 3, addr_reg);
     tcg_out_rld (s, RLDICL, 4, data_reg, 0, 64 - (1 << (3 + opc)));
     tcg_out_movi (s, TCG_TYPE_I64, 5, mem_index);
 
diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c
index 265194a..06b6db3 100644
--- a/tcg/s390/tcg-target.c
+++ b/tcg/s390/tcg-target.c
@@ -94,7 +94,7 @@  void tcg_target_qemu_prologue(TCGContext *s)
     /* gets called with KVM */
 }
 
-static inline void tcg_out_mov(TCGContext *s, int ret, int arg)
+static inline void tcg_out_mov(TCGContext *s, TCGType type, int ret, int arg)
 {
     tcg_abort();
 }
diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c
index e460d44..9f970cd 100644
--- a/tcg/sparc/tcg-target.c
+++ b/tcg/sparc/tcg-target.c
@@ -304,7 +304,7 @@  static void tcg_out_arithc(TCGContext *s, int rd, int rs1,
               | (val2const ? INSN_IMM13(val2) : INSN_RS2(val2)));
 }
 
-static inline void tcg_out_mov(TCGContext *s, int ret, int arg)
+static inline void tcg_out_mov(TCGContext *s, TCGType type, int ret, int arg)
 {
     tcg_out_arith(s, ret, arg, TCG_REG_G0, ARITH_OR);
 }
@@ -795,7 +795,7 @@  static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args,
     tcg_out32(s, 0);
 
     /* mov (delay slot) */
-    tcg_out_mov(s, arg0, addr_reg);
+    tcg_out_mov(s, TCG_TYPE_PTR, arg0, addr_reg);
 
     /* mov */
     tcg_out_movi(s, TCG_TYPE_I32, arg1, mem_index);
@@ -845,7 +845,7 @@  static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args,
     case 3:
     default:
         /* mov */
-        tcg_out_mov(s, data_reg, arg0);
+        tcg_out_mov(s, TCG_TYPE_REG, data_reg, arg0);
         break;
     }
 
@@ -1007,10 +1007,10 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args,
     tcg_out32(s, 0);
 
     /* mov (delay slot) */
-    tcg_out_mov(s, arg0, addr_reg);
+    tcg_out_mov(s, TCG_TYPE_PTR, arg0, addr_reg);
 
     /* mov */
-    tcg_out_mov(s, arg1, data_reg);
+    tcg_out_mov(s, TCG_TYPE_REG, arg1, data_reg);
 
     /* mov */
     tcg_out_movi(s, TCG_TYPE_I32, arg2, mem_index);
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 880e7ce..32f9e17 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -1547,7 +1547,7 @@  static void tcg_reg_alloc_mov(TCGContext *s, const TCGOpDef *def,
                 reg = tcg_reg_alloc(s, arg_ct->u.regs, s->reserved_regs);
             }
             if (ts->reg != reg) {
-                tcg_out_mov(s, reg, ts->reg);
+                tcg_out_mov(s, ots->type, reg, ts->reg);
             }
         }
     } else if (ts->val_type == TEMP_VAL_MEM) {
@@ -1652,7 +1652,7 @@  static void tcg_reg_alloc_op(TCGContext *s,
             /* allocate a new register matching the constraint 
                and move the temporary register into it */
             reg = tcg_reg_alloc(s, arg_ct->u.regs, allocated_regs);
-            tcg_out_mov(s, reg, ts->reg);
+            tcg_out_mov(s, ts->type, reg, ts->reg);
         }
         new_args[i] = reg;
         const_args[i] = 0;
@@ -1734,7 +1734,7 @@  static void tcg_reg_alloc_op(TCGContext *s,
         ts = &s->temps[args[i]];
         reg = new_args[i];
         if (ts->fixed_reg && ts->reg != reg) {
-            tcg_out_mov(s, ts->reg, reg);
+            tcg_out_mov(s, ts->type, ts->reg, reg);
         }
     }
 }
@@ -1820,7 +1820,7 @@  static int tcg_reg_alloc_call(TCGContext *s, const TCGOpDef *def,
             tcg_reg_free(s, reg);
             if (ts->val_type == TEMP_VAL_REG) {
                 if (ts->reg != reg) {
-                    tcg_out_mov(s, reg, ts->reg);
+                    tcg_out_mov(s, ts->type, reg, ts->reg);
                 }
             } else if (ts->val_type == TEMP_VAL_MEM) {
                 tcg_out_ld(s, ts->type, reg, ts->mem_reg, ts->mem_offset);
@@ -1849,7 +1849,7 @@  static int tcg_reg_alloc_call(TCGContext *s, const TCGOpDef *def,
         reg = ts->reg;
         if (!tcg_regset_test_reg(arg_ct->u.regs, reg)) {
             reg = tcg_reg_alloc(s, arg_ct->u.regs, allocated_regs);
-            tcg_out_mov(s, reg, ts->reg);
+            tcg_out_mov(s, ts->type, reg, ts->reg);
         }
         func_arg = reg;
         tcg_regset_set_reg(allocated_regs, reg);
@@ -1908,7 +1908,7 @@  static int tcg_reg_alloc_call(TCGContext *s, const TCGOpDef *def,
         assert(s->reg_to_temp[reg] == -1);
         if (ts->fixed_reg) {
             if (ts->reg != reg) {
-                tcg_out_mov(s, ts->reg, reg);
+                tcg_out_mov(s, ts->type, ts->reg, reg);
             }
         } else {
             if (ts->val_type == TEMP_VAL_REG)
diff --git a/tcg/tcg.h b/tcg/tcg.h
index 5853823..d6cb9e7 100644
--- a/tcg/tcg.h
+++ b/tcg/tcg.h
@@ -101,11 +101,18 @@  typedef enum TCGType {
     TCG_TYPE_I64,
     TCG_TYPE_COUNT, /* number of different types */
 
+    /* An alias for the size of the host register.  */
 #if TCG_TARGET_REG_BITS == 32
-    TCG_TYPE_PTR = TCG_TYPE_I32,
+    TCG_TYPE_REG = TCG_TYPE_I32,
 #else
-    TCG_TYPE_PTR = TCG_TYPE_I64,
+    TCG_TYPE_REG = TCG_TYPE_I64,
 #endif
+
+    /* An alias for the size of the native pointer.  We don't currently
+       support any hosts with 64-bit registers and 32-bit pointers.  */
+    TCG_TYPE_PTR = TCG_TYPE_REG,
+
+    /* An alias for the size of the target "long", aka register.  */
 #if TARGET_LONG_BITS == 64
     TCG_TYPE_TL = TCG_TYPE_I64,
 #else
diff --git a/tcg/x86_64/tcg-target.c b/tcg/x86_64/tcg-target.c
index 3892f75..9844a57 100644
--- a/tcg/x86_64/tcg-target.c
+++ b/tcg/x86_64/tcg-target.c
@@ -354,9 +354,10 @@  static void tcg_out_modrm_offset2(TCGContext *s, int opc, int r, int rm,
 }
 #endif
 
-static inline void tcg_out_mov(TCGContext *s, int ret, int arg)
+static inline void tcg_out_mov(TCGContext *s, TCGType type, int ret, int arg)
 {
-    tcg_out_modrm(s, 0x8b | P_REXW, ret, arg);
+    int rexw = (type == TCG_TYPE_I64 ? P_REXW : 0);
+    tcg_out_modrm(s, 0x8b | rexw, ret, arg);
 }
 
 static inline void tcg_out_movi(TCGContext *s, TCGType type, 
@@ -577,11 +578,8 @@  static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args,
     rexw = P_REXW;
 #endif
 #if defined(CONFIG_SOFTMMU)
-    /* mov */
-    tcg_out_modrm(s, 0x8b | rexw, r1, addr_reg);
-
-    /* mov */
-    tcg_out_modrm(s, 0x8b | rexw, r0, addr_reg);
+    tcg_out_mov(s, TCG_TYPE_TL, r1, addr_reg);
+    tcg_out_mov(s, TCG_TYPE_TL, r0, addr_reg);
  
     tcg_out_modrm(s, 0xc1 | rexw, 5, r1); /* shr $x, r1 */
     tcg_out8(s, TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS); 
@@ -599,8 +597,7 @@  static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args,
     /* cmp 0(r1), r0 */
     tcg_out_modrm_offset(s, 0x3b | rexw, r0, r1, 0);
     
-    /* mov */
-    tcg_out_modrm(s, 0x8b | rexw, r0, addr_reg);
+    tcg_out_mov(s, TCG_TYPE_TL, r0, addr_reg);
     
     /* je label1 */
     tcg_out8(s, 0x70 + JCC_JE);
@@ -634,11 +631,10 @@  static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args,
         break;
     case 2:
     default:
-        /* movl */
-        tcg_out_modrm(s, 0x8b, data_reg, TCG_REG_RAX);
+        tcg_out_mov(s, TCG_TYPE_I32, data_reg, TCG_REG_RAX);
         break;
     case 3:
-        tcg_out_mov(s, data_reg, TCG_REG_RAX);
+        tcg_out_mov(s, TCG_TYPE_I64, data_reg, TCG_REG_RAX);
         break;
     }
 
@@ -772,11 +768,8 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args,
     rexw = P_REXW;
 #endif
 #if defined(CONFIG_SOFTMMU)
-    /* mov */
-    tcg_out_modrm(s, 0x8b | rexw, r1, addr_reg);
-
-    /* mov */
-    tcg_out_modrm(s, 0x8b | rexw, r0, addr_reg);
+    tcg_out_mov(s, TCG_TYPE_TL, r1, addr_reg);
+    tcg_out_mov(s, TCG_TYPE_TL, r0, addr_reg);
  
     tcg_out_modrm(s, 0xc1 | rexw, 5, r1); /* shr $x, r1 */
     tcg_out8(s, TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS); 
@@ -793,10 +786,9 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args,
 
     /* cmp 0(r1), r0 */
     tcg_out_modrm_offset(s, 0x3b | rexw, r0, r1, 0);
-    
-    /* mov */
-    tcg_out_modrm(s, 0x8b | rexw, r0, addr_reg);
-    
+
+    tcg_out_mov(s, TCG_TYPE_TL, r0, addr_reg);
+
     /* je label1 */
     tcg_out8(s, 0x70 + JCC_JE);
     label1_ptr = s->code_ptr;
@@ -813,12 +805,11 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args,
         tcg_out_modrm(s, 0xb7 | P_EXT, TCG_REG_RSI, data_reg);
         break;
     case 2:
-        /* movl */
-        tcg_out_modrm(s, 0x8b, TCG_REG_RSI, data_reg);
+        tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_RSI, data_reg);
         break;
     default:
     case 3:
-        tcg_out_mov(s, TCG_REG_RSI, data_reg);
+        tcg_out_mov(s, TCG_TYPE_I64, TCG_REG_RSI, data_reg);
         break;
     }
     tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_RDX, mem_index);
@@ -863,7 +854,7 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args,
         break;
     case 1:
         if (bswap) {
-            tcg_out_modrm(s, 0x8b, r1, data_reg); /* movl */
+            tcg_out_mov(s, TCG_TYPE_I32, r1, data_reg);
             tcg_out8(s, 0x66); /* rolw $8, %ecx */
             tcg_out_modrm(s, 0xc1, 0, r1);
             tcg_out8(s, 8);
@@ -875,7 +866,7 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args,
         break;
     case 2:
         if (bswap) {
-            tcg_out_modrm(s, 0x8b, r1, data_reg); /* movl */
+            tcg_out_mov(s, TCG_TYPE_I32, r1, data_reg);
             /* bswap data_reg */
             tcg_out_opc(s, (0xc8 + r1) | P_EXT, 0, r1, 0);
             data_reg = r1;
@@ -885,7 +876,7 @@  static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args,
         break;
     case 3:
         if (bswap) {
-            tcg_out_mov(s, r1, data_reg);
+            tcg_out_mov(s, TCG_TYPE_I64, r1, data_reg);
             /* bswap data_reg */
             tcg_out_opc(s, (0xc8 + r1) | P_EXT | P_REXW, 0, r1, 0);
             data_reg = r1;