tcg: Implement insert_op_before

Rather reserving space in the op stream for optimization,
let the optimizer add ops as necessary.

Reviewed-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
Signed-off-by: Richard Henderson <rth@twiddle.net>

3 files changed, 35 insertions, 44 deletions

diff --git a/tcg/optimize.c b/tcg/optimize.c
index 973fbb4..067917c 100644
--- a/tcg/optimize.c
+++ b/tcg/optimize.c
@@ -67,6 +67,37 @@ static void reset_temp(TCGArg temp)
     temps[temp].mask = -1;
 }
 
+static TCGOp *insert_op_before(TCGContext *s, TCGOp *old_op,
+                                TCGOpcode opc, int nargs)
+{
+    int oi = s->gen_next_op_idx;
+    int pi = s->gen_next_parm_idx;
+    int prev = old_op->prev;
+    int next = old_op - s->gen_op_buf;
+    TCGOp *new_op;
+
+    tcg_debug_assert(oi < OPC_BUF_SIZE);
+    tcg_debug_assert(pi + nargs <= OPPARAM_BUF_SIZE);
+    s->gen_next_op_idx = oi + 1;
+    s->gen_next_parm_idx = pi + nargs;
+
+    new_op = &s->gen_op_buf[oi];
+    *new_op = (TCGOp){
+        .opc = opc,
+        .args = pi,
+        .prev = prev,
+        .next = next
+    };
+    if (prev >= 0) {
+        s->gen_op_buf[prev].next = oi;
+    } else {
+        s->gen_first_op_idx = oi;
+    }
+    old_op->prev = oi;
+
+    return new_op;
+}
+
 /* Reset all temporaries, given that there are NB_TEMPS of them.  */
 static void reset_all_temps(int nb_temps)
 {
@@ -1108,8 +1139,8 @@ static void tcg_constant_folding(TCGContext *s)
                 uint64_t a = ((uint64_t)ah << 32) | al;
                 uint64_t b = ((uint64_t)bh << 32) | bl;
                 TCGArg rl, rh;
-                TCGOp *op2;
-                TCGArg *args2;
+                TCGOp *op2 = insert_op_before(s, op, INDEX_op_movi_i32, 2);
+                TCGArg *args2 = &s->gen_opparam_buf[op2->args];
 
                 if (opc == INDEX_op_add2_i32) {
                     a += b;
@@ -1117,15 +1148,6 @@ static void tcg_constant_folding(TCGContext *s)
                     a -= b;
                 }
 
-                /* We emit the extra nop when we emit the add2/sub2.  */
-                op2 = &s->gen_op_buf[oi_next];
-                assert(op2->opc == INDEX_op_nop);
-
-                /* But we still have to allocate args for the op.  */
-                op2->args = s->gen_next_parm_idx;
-                s->gen_next_parm_idx += 2;
-                args2 = &s->gen_opparam_buf[op2->args];
-
                 rl = args[0];
                 rh = args[1];
                 tcg_opt_gen_movi(s, op, args, opc, rl, (uint32_t)a);
@@ -1144,17 +1166,8 @@ static void tcg_constant_folding(TCGContext *s)
                 uint32_t b = temps[args[3]].val;
                 uint64_t r = (uint64_t)a * b;
                 TCGArg rl, rh;
-                TCGOp *op2;
-                TCGArg *args2;
-
-                /* We emit the extra nop when we emit the mulu2.  */
-                op2 = &s->gen_op_buf[oi_next];
-                assert(op2->opc == INDEX_op_nop);
-
-                /* But we still have to allocate args for the op.  */
-                op2->args = s->gen_next_parm_idx;
-                s->gen_next_parm_idx += 2;
-                args2 = &s->gen_opparam_buf[op2->args];
+                TCGOp *op2 = insert_op_before(s, op, INDEX_op_movi_i32, 2);
+                TCGArg *args2 = &s->gen_opparam_buf[op2->args];
 
                 rl = args[0];
                 rh = args[1];
diff --git a/tcg/tcg-op.c b/tcg/tcg-op.c
index cbaa15c..afa351d 100644
--- a/tcg/tcg-op.c
+++ b/tcg/tcg-op.c
@@ -57,11 +57,6 @@ static void tcg_emit_op(TCGContext *ctx, TCGOpcode opc, int args)
     };
 }
 
-void tcg_gen_op0(TCGContext *ctx, TCGOpcode opc)
-{
-    tcg_emit_op(ctx, opc, -1);
-}
-
 void tcg_gen_op1(TCGContext *ctx, TCGOpcode opc, TCGArg a1)
 {
     int pi = ctx->gen_next_parm_idx;
@@ -571,8 +566,6 @@ void tcg_gen_add2_i32(TCGv_i32 rl, TCGv_i32 rh, TCGv_i32 al,
 {
     if (TCG_TARGET_HAS_add2_i32) {
         tcg_gen_op6_i32(INDEX_op_add2_i32, rl, rh, al, ah, bl, bh);
-        /* Allow the optimizer room to replace add2 with two moves.  */
-        tcg_gen_op0(&tcg_ctx, INDEX_op_nop);
     } else {
         TCGv_i64 t0 = tcg_temp_new_i64();
         TCGv_i64 t1 = tcg_temp_new_i64();
@@ -590,8 +583,6 @@ void tcg_gen_sub2_i32(TCGv_i32 rl, TCGv_i32 rh, TCGv_i32 al,
 {
     if (TCG_TARGET_HAS_sub2_i32) {
         tcg_gen_op6_i32(INDEX_op_sub2_i32, rl, rh, al, ah, bl, bh);
-        /* Allow the optimizer room to replace sub2 with two moves.  */
-        tcg_gen_op0(&tcg_ctx, INDEX_op_nop);
     } else {
         TCGv_i64 t0 = tcg_temp_new_i64();
         TCGv_i64 t1 = tcg_temp_new_i64();
@@ -608,8 +599,6 @@ void tcg_gen_mulu2_i32(TCGv_i32 rl, TCGv_i32 rh, TCGv_i32 arg1, TCGv_i32 arg2)
 {
     if (TCG_TARGET_HAS_mulu2_i32) {
         tcg_gen_op4_i32(INDEX_op_mulu2_i32, rl, rh, arg1, arg2);
-        /* Allow the optimizer room to replace mulu2 with two moves.  */
-        tcg_gen_op0(&tcg_ctx, INDEX_op_nop);
     } else if (TCG_TARGET_HAS_muluh_i32) {
         TCGv_i32 t = tcg_temp_new_i32();
         tcg_gen_op3_i32(INDEX_op_mul_i32, t, arg1, arg2);
@@ -632,8 +621,6 @@ void tcg_gen_muls2_i32(TCGv_i32 rl, TCGv_i32 rh, TCGv_i32 arg1, TCGv_i32 arg2)
 {
     if (TCG_TARGET_HAS_muls2_i32) {
         tcg_gen_op4_i32(INDEX_op_muls2_i32, rl, rh, arg1, arg2);
-        /* Allow the optimizer room to replace muls2 with two moves.  */
-        tcg_gen_op0(&tcg_ctx, INDEX_op_nop);
     } else if (TCG_TARGET_HAS_mulsh_i32) {
         TCGv_i32 t = tcg_temp_new_i32();
         tcg_gen_op3_i32(INDEX_op_mul_i32, t, arg1, arg2);
@@ -1648,8 +1635,6 @@ void tcg_gen_add2_i64(TCGv_i64 rl, TCGv_i64 rh, TCGv_i64 al,
 {
     if (TCG_TARGET_HAS_add2_i64) {
         tcg_gen_op6_i64(INDEX_op_add2_i64, rl, rh, al, ah, bl, bh);
-        /* Allow the optimizer room to replace add2 with two moves.  */
-        tcg_gen_op0(&tcg_ctx, INDEX_op_nop);
     } else {
         TCGv_i64 t0 = tcg_temp_new_i64();
         TCGv_i64 t1 = tcg_temp_new_i64();
@@ -1668,8 +1653,6 @@ void tcg_gen_sub2_i64(TCGv_i64 rl, TCGv_i64 rh, TCGv_i64 al,
 {
     if (TCG_TARGET_HAS_sub2_i64) {
         tcg_gen_op6_i64(INDEX_op_sub2_i64, rl, rh, al, ah, bl, bh);
-        /* Allow the optimizer room to replace sub2 with two moves.  */
-        tcg_gen_op0(&tcg_ctx, INDEX_op_nop);
     } else {
         TCGv_i64 t0 = tcg_temp_new_i64();
         TCGv_i64 t1 = tcg_temp_new_i64();
@@ -1687,8 +1670,6 @@ void tcg_gen_mulu2_i64(TCGv_i64 rl, TCGv_i64 rh, TCGv_i64 arg1, TCGv_i64 arg2)
 {
     if (TCG_TARGET_HAS_mulu2_i64) {
         tcg_gen_op4_i64(INDEX_op_mulu2_i64, rl, rh, arg1, arg2);
-        /* Allow the optimizer room to replace mulu2 with two moves.  */
-        tcg_gen_op0(&tcg_ctx, INDEX_op_nop);
     } else if (TCG_TARGET_HAS_muluh_i64) {
         TCGv_i64 t = tcg_temp_new_i64();
         tcg_gen_op3_i64(INDEX_op_mul_i64, t, arg1, arg2);
@@ -1708,8 +1689,6 @@ void tcg_gen_muls2_i64(TCGv_i64 rl, TCGv_i64 rh, TCGv_i64 arg1, TCGv_i64 arg2)
 {
     if (TCG_TARGET_HAS_muls2_i64) {
         tcg_gen_op4_i64(INDEX_op_muls2_i64, rl, rh, arg1, arg2);
-        /* Allow the optimizer room to replace muls2 with two moves.  */
-        tcg_gen_op0(&tcg_ctx, INDEX_op_nop);
     } else if (TCG_TARGET_HAS_mulsh_i64) {
         TCGv_i64 t = tcg_temp_new_i64();
         tcg_gen_op3_i64(INDEX_op_mul_i64, t, arg1, arg2);
diff --git a/tcg/tcg-op.h b/tcg/tcg-op.h
index eacfd8a..96adf9a 100644
--- a/tcg/tcg-op.h
+++ b/tcg/tcg-op.h
@@ -28,7 +28,6 @@
 
 /* Basic output routines.  Not for general consumption.  */
 
-void tcg_gen_op0(TCGContext *, TCGOpcode);
 void tcg_gen_op1(TCGContext *, TCGOpcode, TCGArg);
 void tcg_gen_op2(TCGContext *, TCGOpcode, TCGArg, TCGArg);
 void tcg_gen_op3(TCGContext *, TCGOpcode, TCGArg, TCGArg, TCGArg);