[PULL 10/25] target/i386: change X86_ENTRYwr to use T0, use it for moves

[Paolo Bonzini]

Just like X86_ENTRYr, X86_ENTRYwr is easily changed to use only T0.
In this case, the motivation is to use it for the MOV instruction
family.  The case when you need to preserve the input value is the
odd one, as it is used basically only for BLS* instructions.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 target/i386/tcg/decode-new.c.inc | 48 ++++++++++++++++----------------
 target/i386/tcg/emit.c.inc       |  2 +-
 2 files changed, 25 insertions(+), 25 deletions(-)

diff --git a/target/i386/tcg/decode-new.c.inc b/target/i386/tcg/decode-new.c.inc
index f9d3e2577b2..d41002e2f5c 100644
--- a/target/i386/tcg/decode-new.c.inc
+++ b/target/i386/tcg/decode-new.c.inc
@@ -180,7 +180,7 @@
 #define X86_OP_ENTRYrr(op, op0, s0, op1, s1, ...)                 \
     X86_OP_ENTRY3(op, None, None, op0, s0, op1, s1, ## __VA_ARGS__)
 #define X86_OP_ENTRYwr(op, op0, s0, op1, s1, ...)                 \
-    X86_OP_ENTRY3(op, op0, s0, None, None, op1, s1, ## __VA_ARGS__)
+    X86_OP_ENTRY3(op, op0, s0, op1, s1, None, None, ## __VA_ARGS__)
 #define X86_OP_ENTRY2(op, op0, s0, op1, s1, ...)                  \
     X86_OP_ENTRY3(op, op0, s0, 2op, s0, op1, s1, ## __VA_ARGS__)
 #define X86_OP_ENTRYw(op, op0, s0, ...)                           \
@@ -612,15 +612,15 @@ static const X86OpEntry opcodes_0F38_00toEF[240] = {
 /* five rows for no prefix, 66, F3, F2, 66+F2  */
 static const X86OpEntry opcodes_0F38_F0toFF[16][5] = {
     [0] = {
-        X86_OP_ENTRY3(MOVBE, G,y, M,y, None,None, cpuid(MOVBE)),
-        X86_OP_ENTRY3(MOVBE, G,w, M,w, None,None, cpuid(MOVBE)),
+        X86_OP_ENTRYwr(MOVBE, G,y, M,y, cpuid(MOVBE)),
+        X86_OP_ENTRYwr(MOVBE, G,w, M,w, cpuid(MOVBE)),
         {},
         X86_OP_ENTRY2(CRC32, G,d, E,b, cpuid(SSE42)),
         X86_OP_ENTRY2(CRC32, G,d, E,b, cpuid(SSE42)),
     },
     [1] = {
-        X86_OP_ENTRY3(MOVBE, M,y, G,y, None,None, cpuid(MOVBE)),
-        X86_OP_ENTRY3(MOVBE, M,w, G,w, None,None, cpuid(MOVBE)),
+        X86_OP_ENTRYwr(MOVBE, M,y, G,y, cpuid(MOVBE)),
+        X86_OP_ENTRYwr(MOVBE, M,w, G,w, cpuid(MOVBE)),
         {},
         X86_OP_ENTRY2(CRC32, G,d, E,y, cpuid(SSE42)),
         X86_OP_ENTRY2(CRC32, G,d, E,w, cpuid(SSE42)),
@@ -1586,18 +1586,18 @@ static const X86OpEntry opcodes_root[256] = {
     [0x7E] = X86_OP_ENTRYr(Jcc, J,b),
     [0x7F] = X86_OP_ENTRYr(Jcc, J,b),
 
-    [0x88] = X86_OP_ENTRY3(MOV, E,b, G,b, None, None),
-    [0x89] = X86_OP_ENTRY3(MOV, E,v, G,v, None, None),
-    [0x8A] = X86_OP_ENTRY3(MOV, G,b, E,b, None, None),
-    [0x8B] = X86_OP_ENTRY3(MOV, G,v, E,v, None, None),
-    /* Missing in Table A-2: memory destination is always 16-bit.  */
-    [0x8C] = X86_OP_ENTRY3(MOV, E,v, S,w, None, None, op0_Mw),
-    [0x8D] = X86_OP_ENTRY3(LEA, G,v, M,v, None, None, noseg),
-    [0x8E] = X86_OP_ENTRY3(MOV, S,w, E,w, None, None),
+    [0x88] = X86_OP_ENTRYwr(MOV, E,b, G,b),
+    [0x89] = X86_OP_ENTRYwr(MOV, E,v, G,v),
+    [0x8A] = X86_OP_ENTRYwr(MOV, G,b, E,b),
+    [0x8B] = X86_OP_ENTRYwr(MOV, G,v, E,v),
+     /* Missing in Table A-2: memory destination is always 16-bit.  */
+    [0x8C] = X86_OP_ENTRYwr(MOV, E,v, S,w, op0_Mw),
+    [0x8D] = X86_OP_ENTRYwr(LEA, G,v, M,v, noseg),
+    [0x8E] = X86_OP_ENTRYwr(MOV, S,w, E,w),
     [0x8F] = X86_OP_GROUPw(group1A, E,v),
 
     [0x98] = X86_OP_ENTRY1(CBW,    0,v), /* rAX */
-    [0x99] = X86_OP_ENTRY3(CWD,    2,v, 0,v, None, None), /* rDX, rAX */
+    [0x99] = X86_OP_ENTRYwr(CWD,   2,v, 0,v), /* rDX, rAX */
     [0x9A] = X86_OP_ENTRYrr(CALLF, I_unsigned,p, I_unsigned,w, chk(i64)),
     [0x9B] = X86_OP_ENTRY0(WAIT),
     [0x9C] = X86_OP_ENTRY0(PUSHF,  chk(vm86_iopl) svm(PUSHF)),
@@ -1607,22 +1607,22 @@ static const X86OpEntry opcodes_root[256] = {
 
     [0xA8] = X86_OP_ENTRYrr(AND, 0,b, I,b),   /* AL, Ib */
     [0xA9] = X86_OP_ENTRYrr(AND, 0,v, I,z),   /* rAX, Iz */
-    [0xAA] = X86_OP_ENTRY3(STOS, Y,b, 0,b, None, None),
-    [0xAB] = X86_OP_ENTRY3(STOS, Y,v, 0,v, None, None),
+    [0xAA] = X86_OP_ENTRYwr(STOS, Y,b, 0,b),
+    [0xAB] = X86_OP_ENTRYwr(STOS, Y,v, 0,v),
     /* Manual writeback because REP LODS (!) has to write EAX/RAX after every 
LODS.  */
     [0xAC] = X86_OP_ENTRYr(LODS, X,b),
     [0xAD] = X86_OP_ENTRYr(LODS, X,v),
     [0xAE] = X86_OP_ENTRYrr(SCAS, 0,b, Y,b),
     [0xAF] = X86_OP_ENTRYrr(SCAS, 0,v, Y,v),
 
-    [0xB8] = X86_OP_ENTRY3(MOV, LoBits,v, I,v, None, None),
-    [0xB9] = X86_OP_ENTRY3(MOV, LoBits,v, I,v, None, None),
-    [0xBA] = X86_OP_ENTRY3(MOV, LoBits,v, I,v, None, None),
-    [0xBB] = X86_OP_ENTRY3(MOV, LoBits,v, I,v, None, None),
-    [0xBC] = X86_OP_ENTRY3(MOV, LoBits,v, I,v, None, None),
-    [0xBD] = X86_OP_ENTRY3(MOV, LoBits,v, I,v, None, None),
-    [0xBE] = X86_OP_ENTRY3(MOV, LoBits,v, I,v, None, None),
-    [0xBF] = X86_OP_ENTRY3(MOV, LoBits,v, I,v, None, None),
+    [0xB8] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v),
+    [0xB9] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v),
+    [0xBA] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v),
+    [0xBB] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v),
+    [0xBC] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v),
+    [0xBD] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v),
+    [0xBE] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v),
+    [0xBF] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v),
 
     [0xC8] = X86_OP_ENTRYrr(ENTER, I,w, I,b),
     [0xC9] = X86_OP_ENTRY1(LEAVE, A,d64),
diff --git a/target/i386/tcg/emit.c.inc b/target/i386/tcg/emit.c.inc
index 797e6e81406..78d89db57cd 100644
--- a/target/i386/tcg/emit.c.inc
+++ b/target/i386/tcg/emit.c.inc
@@ -1796,7 +1796,7 @@ static void gen_IN(DisasContext *s, X86DecodedInsn 
*decode)
     MemOp ot = decode->op[0].ot;
     TCGv_i32 port = tcg_temp_new_i32();
 
-    tcg_gen_trunc_tl_i32(port, s->T1);
+    tcg_gen_trunc_tl_i32(port, s->T0);
     tcg_gen_ext16u_i32(port, port);
     if (!gen_check_io(s, ot, port, SVM_IOIO_TYPE_MASK)) {
         return;
-- 
2.45.1