Richard Henderson <richard.hender...@linaro.org> writes:
> Do not allocate a large block for indexing. Instead, allocate
> for each temporary as they are seen.
>
> In general, this will use less memory, if we consider that most
> TBs do not touch every target register. This also allows us to
> allocate TempOptInfo for new temps created during optimization.
>
> Signed-off-by: Richard Henderson <richard.hender...@linaro.org>
Reviewed-by: Alex Bennée <alex.ben...@linaro.org>
> ---
> tcg/optimize.c | 60 ++++++++++++++++++++++++++++----------------------
> 1 file changed, 34 insertions(+), 26 deletions(-)
>
> diff --git a/tcg/optimize.c b/tcg/optimize.c
> index b86bf3d707..d36d7e1d7f 100644
> --- a/tcg/optimize.c
> +++ b/tcg/optimize.c
> @@ -89,35 +89,41 @@ static void reset_temp(TCGArg arg)
> }
>
> /* Initialize and activate a temporary. */
> -static void init_ts_info(TempOptInfo *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)) {
> - TempOptInfo *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;
> - if (ts->kind == TEMP_CONST) {
> - ti->is_const = true;
> - ti->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;
> + }
> +
> + 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;
> }
> - set_bit(idx, temps_used->l);
> + } else {
> + ti->is_const = false;
> + ti->mask = -1;
> }
> }
>
> -static void init_arg_info(TempOptInfo *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)
> @@ -603,9 +609,8 @@ 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;
> - TempOptInfo *infos;
> TCGTempSet temps_used;
>
> /* Array VALS has an element for each temp.
> @@ -615,12 +620,15 @@ 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(TempOptInfo) * nb_temps);
> + 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;
> + int nb_oargs, nb_iargs;
> TCGArg tmp;
> TCGOpcode opc = op->opc;
> const TCGOpDef *def = &tcg_op_defs[opc];
> @@ -633,14 +641,14 @@ 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]);
> }
> }
--
Alex Bennée
