AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

ir3: calculate sstall/systall across blocks

Browse Source 
Resetting the ss/sy delays at the start of blocks would underestimate
the actual delays at runtime. Make the estimate more accurate by keeping
track of outstanding delays at the end of blocks and setting the initial
delays of blocks to the maximum of their predecessor blocks.

Signed-off-by: Job Noorman <jnoorman@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34108>
This commit is contained in:
Job Noorman
2025-05-09 11:26:05 +02:00
committed by Marge Bot
parent 13f9135e85
commit 0977863a8e
1 changed files with 30 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/freedreno/ir3/ir3.c
+30 -11
View File
@@ -281,8 +281,25 @@ ir3_collect_info(struct ir3_shader_variant *v)
/* Full and half aliases do not overlap so treat them as !mergedregs. */
regmask_init(&aliases, false);
struct block_data {
unsigned sfu_delay;
unsigned mem_delay;
};
void *mem_ctx = ralloc_context(NULL);
foreach_block (block, &shader->block_list) {
int sfu_delay = 0, mem_delay = 0;
block->data = rzalloc(mem_ctx, struct block_data);
}
foreach_block (block, &shader->block_list) {
struct block_data *bd = block->data;
for (unsigned i = 0; i < block->predecessors_count; i++) {
struct block_data *pbd = block->predecessors[i]->data;
bd->sfu_delay = MAX2(bd->sfu_delay, pbd->sfu_delay);
bd->mem_delay = MAX2(bd->mem_delay, pbd->mem_delay);
}
foreach_instr (instr, &block->instr_list) {
@@ -367,28 +384,28 @@ ir3_collect_info(struct ir3_shader_variant *v)
if (instr->flags & IR3_INSTR_SS) {
info->ss++;
info->sstall += sfu_delay;
sfu_delay = 0;
info->sstall += bd->sfu_delay;
bd->sfu_delay = 0;
}
if (instr->flags & IR3_INSTR_SY) {
info->sy++;
info->systall += mem_delay;
mem_delay = 0;
info->systall += bd->mem_delay;
bd->mem_delay = 0;
}
if (is_ss_producer(instr)) {
sfu_delay = soft_ss_delay(instr);
bd->sfu_delay = soft_ss_delay(instr);
} else {
int n = MIN2(sfu_delay, 1 + instr->repeat + instr->nop);
sfu_delay -= n;
int n = MIN2(bd->sfu_delay, 1 + instr->repeat + instr->nop);
bd->sfu_delay -= n;
}
if (is_sy_producer(instr)) {
mem_delay = soft_sy_delay(instr, shader);
bd->mem_delay = soft_sy_delay(instr, shader);
} else {
int n = MIN2(mem_delay, 1 + instr->repeat + instr->nop);
mem_delay -= n;
int n = MIN2(bd->mem_delay, 1 + instr->repeat + instr->nop);
bd->mem_delay -= n;
}
} else {
unsigned instrs_count = 1 + instr->repeat + instr->nop;
@@ -471,6 +488,8 @@ ir3_collect_info(struct ir3_shader_variant *v)
compiler, regs_count, info->double_threadsize);
info->max_waves = MIN2(reg_independent_max_waves, reg_dependent_max_waves);
assert(info->max_waves <= v->compiler->max_waves);
ralloc_free(mem_ctx);
}
static struct ir3_register *