[Qemu-devel] [PATCH 08/10] tcg: Make call address a constant parameter

[Richard Henderson]

Avoid allocating a tcg temporary to hold the constant address,
and instead place it directly into the op_call arguments.

At the same time, convert to the newly introduced tcg_out_call
backend function, rather than invoking tcg_out_op for the call.

Signed-off-by: Richard Henderson <address@hidden>
---
 tcg/optimize.c |  75 +++++++++++++++------------------
 tcg/tcg-op.h   |  16 ++-----
 tcg/tcg-opc.h  |   2 +-
 tcg/tcg.c      | 129 +++++++++++++++------------------------------------------
 tcg/tcg.h      |   2 +-
 5 files changed, 72 insertions(+), 152 deletions(-)

diff --git a/tcg/optimize.c b/tcg/optimize.c
index 0302f4f..3a504a1 100644
--- a/tcg/optimize.c
+++ b/tcg/optimize.c
@@ -513,12 +513,8 @@ static bool swap_commutative2(TCGArg *p1, TCGArg *p2)
 static TCGArg *tcg_constant_folding(TCGContext *s, uint16_t *tcg_opc_ptr,
                                     TCGArg *args, TCGOpDef *tcg_op_defs)
 {
-    int i, nb_ops, op_index, nb_temps, nb_globals, nb_call_args;
-    tcg_target_ulong mask, affected;
-    TCGOpcode op;
-    const TCGOpDef *def;
+    int nb_ops, op_index, nb_temps, nb_globals;
     TCGArg *gen_args;
-    TCGArg tmp;
 
     /* Array VALS has an element for each temp.
        If this temp holds a constant then its value is kept in VALS' element.
@@ -532,22 +528,27 @@ static TCGArg *tcg_constant_folding(TCGContext *s, 
uint16_t *tcg_opc_ptr,
     nb_ops = tcg_opc_ptr - s->gen_opc_buf;
     gen_args = args;
     for (op_index = 0; op_index < nb_ops; op_index++) {
-        op = s->gen_opc_buf[op_index];
-        def = &tcg_op_defs[op];
-        /* Do copy propagation */
+        TCGOpcode op = s->gen_opc_buf[op_index];
+        const TCGOpDef *def = &tcg_op_defs[op];
+        tcg_target_ulong mask, affected;
+        int nb_oargs, nb_iargs, nb_args, i;
+        TCGArg tmp;
+
         if (op == INDEX_op_call) {
-            int nb_oargs = args[0] >> 16;
-            int nb_iargs = args[0] & 0xffff;
-            for (i = nb_oargs + 1; i < nb_oargs + nb_iargs + 1; i++) {
-                if (temps[args[i]].state == TCG_TEMP_COPY) {
-                    args[i] = find_better_copy(s, args[i]);
-                }
-            }
+            *gen_args++ = tmp = *args++;
+            nb_oargs = tmp >> 16;
+            nb_iargs = tmp & 0xffff;
+            nb_args = nb_oargs + nb_iargs + def->nb_cargs;
         } else {
-            for (i = def->nb_oargs; i < def->nb_oargs + def->nb_iargs; i++) {
-                if (temps[args[i]].state == TCG_TEMP_COPY) {
-                    args[i] = find_better_copy(s, args[i]);
-                }
+            nb_oargs = def->nb_oargs;
+            nb_iargs = def->nb_iargs;
+            nb_args = def->nb_args;
+        }
+
+        /* Do copy propagation */
+        for (i = nb_oargs; i < nb_oargs + nb_iargs; i++) {
+            if (temps[args[i]].state == TCG_TEMP_COPY) {
+                args[i] = find_better_copy(s, args[i]);
             }
         }
 
@@ -882,7 +883,7 @@ static TCGArg *tcg_constant_folding(TCGContext *s, uint16_t 
*tcg_opc_ptr,
 
         CASE_OP_32_64(qemu_ld):
             {
-                TCGMemOp mop = args[def->nb_oargs + def->nb_iargs];
+                TCGMemOp mop = args[nb_oargs + nb_iargs];
                 if (!(mop & MO_SIGN)) {
                     mask = (2ULL << ((8 << (mop & MO_SIZE)) - 1)) - 1;
                 }
@@ -900,15 +901,15 @@ static TCGArg *tcg_constant_folding(TCGContext *s, 
uint16_t *tcg_opc_ptr,
         }
 
         if (mask == 0) {
-            assert(def->nb_oargs == 1);
+            assert(nb_oargs == 1);
             s->gen_opc_buf[op_index] = op_to_movi(op);
             tcg_opt_gen_movi(gen_args, args[0], 0);
-            args += def->nb_oargs + def->nb_iargs + def->nb_cargs;
+            args += nb_args;
             gen_args += 2;
             continue;
         }
         if (affected == 0) {
-            assert(def->nb_oargs == 1);
+            assert(nb_oargs == 1);
             if (temps_are_copies(args[0], args[1])) {
                 s->gen_opc_buf[op_index] = INDEX_op_nop;
             } else if (temps[args[1]].state != TCG_TEMP_CONST) {
@@ -920,7 +921,7 @@ static TCGArg *tcg_constant_folding(TCGContext *s, uint16_t 
*tcg_opc_ptr,
                 tcg_opt_gen_movi(gen_args, args[0], temps[args[1]].val);
                 gen_args += 2;
             }
-            args += def->nb_iargs + 1;
+            args += nb_args;
             continue;
         }
 
@@ -1246,24 +1247,13 @@ static TCGArg *tcg_constant_folding(TCGContext *s, 
uint16_t *tcg_opc_ptr,
             break;
 
         case INDEX_op_call:
-            nb_call_args = (args[0] >> 16) + (args[0] & 0xffff);
-            if (!(args[nb_call_args + 1] & (TCG_CALL_NO_READ_GLOBALS |
-                                            TCG_CALL_NO_WRITE_GLOBALS))) {
+            if (!(args[nb_oargs + nb_iargs + 1]
+                  & (TCG_CALL_NO_READ_GLOBALS | TCG_CALL_NO_WRITE_GLOBALS))) {
                 for (i = 0; i < nb_globals; i++) {
                     reset_temp(i);
                 }
             }
-            for (i = 0; i < (args[0] >> 16); i++) {
-                reset_temp(args[i + 1]);
-            }
-            i = nb_call_args + 3;
-            while (i) {
-                *gen_args = *args;
-                args++;
-                gen_args++;
-                i--;
-            }
-            break;
+            goto do_reset_output;
 
         default:
         do_default:
@@ -1275,7 +1265,8 @@ static TCGArg *tcg_constant_folding(TCGContext *s, 
uint16_t *tcg_opc_ptr,
             if (def->flags & TCG_OPF_BB_END) {
                 reset_all_temps(nb_temps);
             } else {
-                for (i = 0; i < def->nb_oargs; i++) {
+        do_reset_output:
+                for (i = 0; i < nb_oargs; i++) {
                     reset_temp(args[i]);
                     /* Save the corresponding known-zero bits mask for the
                        first output argument (only one supported so far). */
@@ -1284,11 +1275,11 @@ static TCGArg *tcg_constant_folding(TCGContext *s, 
uint16_t *tcg_opc_ptr,
                     }
                 }
             }
-            for (i = 0; i < def->nb_args; i++) {
+            for (i = 0; i < nb_args; i++) {
                 gen_args[i] = args[i];
             }
-            args += def->nb_args;
-            gen_args += def->nb_args;
+            args += nb_args;
+            gen_args += nb_args;
             break;
         }
     }
diff --git a/tcg/tcg-op.h b/tcg/tcg-op.h
index 6dbbb38..bdd0139 100644
--- a/tcg/tcg-op.h
+++ b/tcg/tcg-op.h
@@ -390,11 +390,7 @@ static inline int tcg_gen_sizemask(int n, int is_64bit, 
int is_signed)
 static inline void tcg_gen_helperN(void *func, int flags, int sizemask,
                                    TCGArg ret, int nargs, TCGArg *args)
 {
-    TCGv_ptr fn;
-    fn = tcg_const_ptr(func);
-    tcg_gen_callN(&tcg_ctx, fn, flags, sizemask, ret,
-                  nargs, args);
-    tcg_temp_free_ptr(fn);
+    tcg_gen_callN(&tcg_ctx, func, flags, sizemask, ret, nargs, args);
 }
 
 /* Note: Both tcg_gen_helper32() and tcg_gen_helper64() are currently
@@ -405,29 +401,23 @@ static inline void tcg_gen_helperN(void *func, int flags, 
int sizemask,
 static inline void tcg_gen_helper32(void *func, int sizemask, TCGv_i32 ret,
                                     TCGv_i32 a, TCGv_i32 b)
 {
-    TCGv_ptr fn;
     TCGArg args[2];
-    fn = tcg_const_ptr(func);
     args[0] = GET_TCGV_I32(a);
     args[1] = GET_TCGV_I32(b);
-    tcg_gen_callN(&tcg_ctx, fn,
+    tcg_gen_callN(&tcg_ctx, func,
                   TCG_CALL_NO_READ_GLOBALS | TCG_CALL_NO_SIDE_EFFECTS,
                   sizemask, GET_TCGV_I32(ret), 2, args);
-    tcg_temp_free_ptr(fn);
 }
 
 static inline void tcg_gen_helper64(void *func, int sizemask, TCGv_i64 ret,
                                     TCGv_i64 a, TCGv_i64 b)
 {
-    TCGv_ptr fn;
     TCGArg args[2];
-    fn = tcg_const_ptr(func);
     args[0] = GET_TCGV_I64(a);
     args[1] = GET_TCGV_I64(b);
-    tcg_gen_callN(&tcg_ctx, fn,
+    tcg_gen_callN(&tcg_ctx, func,
                   TCG_CALL_NO_READ_GLOBALS | TCG_CALL_NO_SIDE_EFFECTS,
                   sizemask, GET_TCGV_I64(ret), 2, args);
-    tcg_temp_free_ptr(fn);
 }
 
 /* 32 bit ops */
diff --git a/tcg/tcg-opc.h b/tcg/tcg-opc.h
index adb6ce1..1716848 100644
--- a/tcg/tcg-opc.h
+++ b/tcg/tcg-opc.h
@@ -40,7 +40,7 @@ DEF(discard, 1, 0, 0, TCG_OPF_NOT_PRESENT)
 DEF(set_label, 0, 0, 1, TCG_OPF_BB_END | TCG_OPF_NOT_PRESENT)
 
 /* variable number of parameters */
-DEF(call, 0, 1, 2, TCG_OPF_CALL_CLOBBER)
+DEF(call, 0, 0, 3, TCG_OPF_CALL_CLOBBER)
 
 DEF(br, 0, 0, 1, TCG_OPF_BB_END)
 
diff --git a/tcg/tcg.c b/tcg/tcg.c
index a835632..08f255e 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -101,6 +101,7 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, const 
TCGArg *args,
                        const int *const_args);
 static void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg, TCGReg arg1,
                        intptr_t arg2);
+static void tcg_out_call(TCGContext *s, tcg_insn_unit *target);
 static int tcg_target_const_match(tcg_target_long val, TCGType type,
                                   const TCGArgConstraint *arg_ct);
 static void tcg_out_tb_init(TCGContext *s);
@@ -705,7 +706,7 @@ int tcg_check_temp_count(void)
 /* Note: we convert the 64 bit args to 32 bit and do some alignment
    and endian swap. Maybe it would be better to do the alignment
    and endian swap in tcg_reg_alloc_call(). */
-void tcg_gen_callN(TCGContext *s, TCGv_ptr func, unsigned int flags,
+void tcg_gen_callN(TCGContext *s, void *func, unsigned int flags,
                    int sizemask, TCGArg ret, int nargs, TCGArg *args)
 {
     int i;
@@ -832,11 +833,10 @@ void tcg_gen_callN(TCGContext *s, TCGv_ptr func, unsigned 
int flags,
         *s->gen_opparam_ptr++ = args[i];
         real_args++;
     }
-    *s->gen_opparam_ptr++ = GET_TCGV_PTR(func);
-
+    *s->gen_opparam_ptr++ = (uintptr_t)func;
     *s->gen_opparam_ptr++ = flags;
 
-    *nparam = (nb_rets << 16) | (real_args + 1);
+    *nparam = (nb_rets << 16) | real_args;
 
     /* total parameters, needed to go backward in the instruction stream */
     *s->gen_opparam_ptr++ = 1 + nb_rets + real_args + 3;
@@ -1243,49 +1243,21 @@ void tcg_dump_ops(TCGContext *s)
             nb_iargs = arg & 0xffff;
             nb_cargs = def->nb_cargs;
 
-            qemu_log(" %s ", def->name);
-
-            /* function name */
-            qemu_log("%s",
-                     tcg_get_arg_str_idx(s, buf, sizeof(buf),
-                                         args[nb_oargs + nb_iargs - 1]));
-            /* flags */
-            qemu_log(",$0x%" TCG_PRIlx, args[nb_oargs + nb_iargs]);
-            /* nb out args */
-            qemu_log(",$%d", nb_oargs);
-            for(i = 0; i < nb_oargs; i++) {
-                qemu_log(",");
-                qemu_log("%s", tcg_get_arg_str_idx(s, buf, sizeof(buf),
+            /* function name, flags, out args */
+            qemu_log(" %s %s,$0x%" TCG_PRIlx ",$%d", def->name,
+                     tcg_find_helper(s, args[nb_oargs + nb_iargs]),
+                     args[nb_oargs + nb_iargs + 1], nb_oargs);
+            for (i = 0; i < nb_oargs; i++) {
+                qemu_log(",%s", tcg_get_arg_str_idx(s, buf, sizeof(buf),
                                                    args[i]));
             }
-            for(i = 0; i < (nb_iargs - 1); i++) {
-                qemu_log(",");
-                if (args[nb_oargs + i] == TCG_CALL_DUMMY_ARG) {
-                    qemu_log("<dummy>");
-                } else {
-                    qemu_log("%s", tcg_get_arg_str_idx(s, buf, sizeof(buf),
-                                                       args[nb_oargs + i]));
-                }
-            }
-        } else if (c == INDEX_op_movi_i32 || c == INDEX_op_movi_i64) {
-            tcg_target_ulong val;
-            const char *name;
-
-            nb_oargs = def->nb_oargs;
-            nb_iargs = def->nb_iargs;
-            nb_cargs = def->nb_cargs;
-            qemu_log(" %s %s,$", def->name,
-                     tcg_get_arg_str_idx(s, buf, sizeof(buf), args[0]));
-            val = args[1];
-            name = tcg_find_helper(s, val);
-            if (name) {
-                qemu_log("%s", name);
-            } else {
-                if (c == INDEX_op_movi_i32) {
-                    qemu_log("0x%x", (uint32_t)val);
-                } else {
-                    qemu_log("0x%" PRIx64 , (uint64_t)val);
+            for (i = 0; i < nb_iargs; i++) {
+                TCGArg arg = args[nb_oargs + i];
+                const char *t = "<dummy>";
+                if (arg != TCG_CALL_DUMMY_ARG) {
+                    t = tcg_get_arg_str_idx(s, buf, sizeof(buf), arg);
                 }
+                qemu_log(",%s", t);
             }
         } else {
             qemu_log(" %s ", def->name);
@@ -1548,9 +1520,9 @@ static inline void tcg_la_bb_end(TCGContext *s, uint8_t 
*dead_temps,
    temporaries are removed. */
 static void tcg_liveness_analysis(TCGContext *s)
 {
-    int i, op_index, nb_args, nb_iargs, nb_oargs, arg, nb_ops;
+    int i, op_index, nb_args, nb_iargs, nb_oargs, nb_ops;
     TCGOpcode op, op_new, op_new2;
-    TCGArg *args;
+    TCGArg *args, arg;
     const TCGOpDef *def;
     uint8_t *dead_temps, *mem_temps;
     uint16_t dead_args;
@@ -1580,15 +1552,15 @@ static void tcg_liveness_analysis(TCGContext *s)
 
                 nb_args = args[-1];
                 args -= nb_args;
-                nb_iargs = args[0] & 0xffff;
-                nb_oargs = args[0] >> 16;
-                args++;
-                call_flags = args[nb_oargs + nb_iargs];
+                arg = *args++;
+                nb_iargs = arg & 0xffff;
+                nb_oargs = arg >> 16;
+                call_flags = args[nb_oargs + nb_iargs + 1];
 
                 /* pure functions can be removed if their result is not
                    used */
                 if (call_flags & TCG_CALL_NO_SIDE_EFFECTS) {
-                    for(i = 0; i < nb_oargs; i++) {
+                    for (i = 0; i < nb_oargs; i++) {
                         arg = args[i];
                         if (!dead_temps[arg] || mem_temps[arg]) {
                             goto do_not_remove_call;
@@ -1602,7 +1574,7 @@ static void tcg_liveness_analysis(TCGContext *s)
                     /* output args are dead */
                     dead_args = 0;
                     sync_args = 0;
-                    for(i = 0; i < nb_oargs; i++) {
+                    for (i = 0; i < nb_oargs; i++) {
                         arg = args[i];
                         if (dead_temps[arg]) {
                             dead_args |= (1 << i);
@@ -1625,7 +1597,7 @@ static void tcg_liveness_analysis(TCGContext *s)
                     }
 
                     /* input args are live */
-                    for(i = nb_oargs; i < nb_iargs + nb_oargs; i++) {
+                    for (i = nb_oargs; i < nb_iargs + nb_oargs; i++) {
                         arg = args[i];
                         if (arg != TCG_CALL_DUMMY_ARG) {
                             if (dead_temps[arg]) {
@@ -2372,26 +2344,27 @@ static int tcg_reg_alloc_call(TCGContext *s, const 
TCGOpDef *def,
                               uint16_t dead_args, uint8_t sync_args)
 {
     int nb_iargs, nb_oargs, flags, nb_regs, i, reg, nb_params;
-    TCGArg arg, func_arg;
+    TCGArg arg;
     TCGTemp *ts;
     intptr_t stack_offset;
     size_t call_stack_size;
-    uintptr_t func_addr;
-    int const_func_arg, allocate_args;
+    tcg_insn_unit *func_addr;
+    int allocate_args;
     TCGRegSet allocated_regs;
-    const TCGArgConstraint *arg_ct;
 
     arg = *args++;
 
     nb_oargs = arg >> 16;
     nb_iargs = arg & 0xffff;
-    nb_params = nb_iargs - 1;
+    nb_params = nb_iargs;
 
-    flags = args[nb_oargs + nb_iargs];
+    func_addr = (tcg_insn_unit *)(intptr_t)args[nb_oargs + nb_iargs];
+    flags = args[nb_oargs + nb_iargs + 1];
 
     nb_regs = ARRAY_SIZE(tcg_target_call_iarg_regs);
-    if (nb_regs > nb_params)
+    if (nb_regs > nb_params) {
         nb_regs = nb_params;
+    }
 
     /* assign stack slots first */
     call_stack_size = (nb_params - nb_regs) * sizeof(tcg_target_long);
@@ -2459,40 +2432,6 @@ static int tcg_reg_alloc_call(TCGContext *s, const 
TCGOpDef *def,
         }
     }
     
-    /* assign function address */
-    func_arg = args[nb_oargs + nb_iargs - 1];
-    arg_ct = &def->args_ct[0];
-    ts = &s->temps[func_arg];
-    func_addr = ts->val;
-    const_func_arg = 0;
-    if (ts->val_type == TEMP_VAL_MEM) {
-        reg = tcg_reg_alloc(s, arg_ct->u.regs, allocated_regs);
-        tcg_out_ld(s, ts->type, reg, ts->mem_reg, ts->mem_offset);
-        func_arg = reg;
-        tcg_regset_set_reg(allocated_regs, reg);
-    } else if (ts->val_type == TEMP_VAL_REG) {
-        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, ts->type, reg, ts->reg);
-        }
-        func_arg = reg;
-        tcg_regset_set_reg(allocated_regs, reg);
-    } else if (ts->val_type == TEMP_VAL_CONST) {
-        if (tcg_target_const_match(func_addr, ts->type, arg_ct)) {
-            const_func_arg = 1;
-            func_arg = func_addr;
-        } else {
-            reg = tcg_reg_alloc(s, arg_ct->u.regs, allocated_regs);
-            tcg_out_movi(s, ts->type, reg, func_addr);
-            func_arg = reg;
-            tcg_regset_set_reg(allocated_regs, reg);
-        }
-    } else {
-        tcg_abort();
-    }
-        
-    
     /* mark dead temporaries and free the associated registers */
     for(i = nb_oargs; i < nb_iargs + nb_oargs; i++) {
         if (IS_DEAD_ARG(i)) {
@@ -2517,7 +2456,7 @@ static int tcg_reg_alloc_call(TCGContext *s, const 
TCGOpDef *def,
         save_globals(s, allocated_regs);
     }
 
-    tcg_out_op(s, opc, &func_arg, &const_func_arg);
+    tcg_out_call(s, func_addr);
 
     /* assign output registers and emit moves if needed */
     for(i = 0; i < nb_oargs; i++) {
diff --git a/tcg/tcg.h b/tcg/tcg.h
index 4f4b1fa..fbc9310 100644
--- a/tcg/tcg.h
+++ b/tcg/tcg.h
@@ -725,7 +725,7 @@ void tcg_add_target_add_op_defs(const TCGTargetOpDef 
*tdefs);
 #define tcg_temp_free_ptr(T) tcg_temp_free_i64(TCGV_PTR_TO_NAT(T))
 #endif
 
-void tcg_gen_callN(TCGContext *s, TCGv_ptr func, unsigned int flags,
+void tcg_gen_callN(TCGContext *s, void *func, unsigned int flags,
                    int sizemask, TCGArg ret, int nargs, TCGArg *args);
 
 void tcg_gen_shifti_i64(TCGv_i64 ret, TCGv_i64 arg1,
-- 
1.9.0