diff options
| author | Rob Clark <[email protected]> | 2020-03-04 10:51:10 -0800 |
|---|---|---|
| committer | Marge Bot <[email protected]> | 2020-03-10 16:01:39 +0000 |
| commit | 752b9985bed171a39bb439421d0e2cd8d0ab82aa (patch) | |
| tree | b7743907be8ad1cbae4994de6ebfb2057568ac7b /src/freedreno | |
| parent | 64ae2ef8bbc63750346345e331750f0e0c643103 (diff) | |
freedreno/ir3: add simplified stall estimation
Doesn't take into account stalls that result from a register written in
a different block, etc. But this should be more useful than just using
number of (ss)'s by trying to estimate how costly a given sync is.
Signed-off-by: Rob Clark <[email protected]>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4071>
Diffstat (limited to 'src/freedreno')
| -rw-r--r-- | src/freedreno/ir3/ir3.c | 12 | ||||
| -rw-r--r-- | src/freedreno/ir3/ir3.h | 3 |
2 files changed, 14 insertions, 1 deletions
diff --git a/src/freedreno/ir3/ir3.c b/src/freedreno/ir3/ir3.c index 0dbe9bba460..bf842701868 100644 --- a/src/freedreno/ir3/ir3.c +++ b/src/freedreno/ir3/ir3.c @@ -911,6 +911,8 @@ void * ir3_assemble(struct ir3 *shader, struct ir3_info *info, ptr = dwords = calloc(4, info->sizedwords); foreach_block (block, &shader->block_list) { + unsigned sfu_delay = 0; + foreach_instr (instr, &block->instr_list) { int ret = emit[opc_cat(instr->opc)](instr, dwords, info); if (ret) @@ -925,11 +927,19 @@ void * ir3_assemble(struct ir3 *shader, struct ir3_info *info, info->nops_count += 1 + instr->repeat; dwords += 2; - if (instr->flags & IR3_INSTR_SS) + if (instr->flags & IR3_INSTR_SS) { info->ss++; + info->sstall += sfu_delay; + } if (instr->flags & IR3_INSTR_SY) info->sy++; + + if (is_sfu(instr)) { + sfu_delay = 10; + } else if (sfu_delay > 0) { + sfu_delay--; + } } } diff --git a/src/freedreno/ir3/ir3.h b/src/freedreno/ir3/ir3.h index b66d8e2d6fd..1a3edbc4530 100644 --- a/src/freedreno/ir3/ir3.h +++ b/src/freedreno/ir3/ir3.h @@ -59,6 +59,9 @@ struct ir3_info { /* number of sync bits: */ uint16_t ss, sy; + /* estimate of number of cycles stalled on (ss) */ + uint16_t sstall; + uint16_t last_baryf; /* instruction # of last varying fetch */ }; |