OpenHarmony/third_party_mesa3d
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

nir/gcm: allow derivative dependent intrinisics to be moved earlier

Browse Source 
We can't move them later as we could move them into non-uniform
control flow, but moving them earlier should be fine.

This helps avoid a bunch of spilling in unigine shaders due to
moving the tex instructions sources earlier (outside if branches)
but not the instruction itself.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4636>
This commit is contained in:
Timothy Arceri
2019-03-22 15:32:22 +11:00
committed by Marge Bot
parent 50a6dd0d65
commit bf4a6c99d2
1 changed files with 14 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
src/compiler/nir/nir_opt_gcm.c
View File

@@ -55,9 +55,10 @@ struct gcm_instr_info {
/* Flags used in the instr->pass_flags field for various instruction states */ /* Flags used in the instr->pass_flags field for various instruction states */
enum { enum {
GCM_INSTR_PINNED = (1 << 0), GCM_INSTR_PINNED = (1 << 0),
GCM_INSTR_SCHEDULED_EARLY = (1 << 1), GCM_INSTR_SCHEDULE_EARLIER_ONLY = (1 << 1),
GCM_INSTR_SCHEDULED_LATE = (1 << 2), GCM_INSTR_SCHEDULED_EARLY = (1 << 2),
GCM_INSTR_PLACED = (1 << 3), GCM_INSTR_SCHEDULED_LATE = (1 << 3),
GCM_INSTR_PLACED = (1 << 4),
}; };
struct gcm_state { struct gcm_state {
@@ -133,8 +134,8 @@ gcm_pin_instructions(nir_function_impl *impl, struct gcm_state *state)
case nir_op_fddy_fine: case nir_op_fddy_fine:
case nir_op_fddx_coarse: case nir_op_fddx_coarse:
case nir_op_fddy_coarse: case nir_op_fddy_coarse:
/* These can only go in uniform control flow; pin them for now */ /* These can only go in uniform control flow */
instr->pass_flags = GCM_INSTR_PINNED; instr->pass_flags = GCM_INSTR_SCHEDULE_EARLIER_ONLY;
break; break;
default: default:
@@ -145,7 +146,7 @@ gcm_pin_instructions(nir_function_impl *impl, struct gcm_state *state)
case nir_instr_type_tex: case nir_instr_type_tex:
if (nir_tex_instr_has_implicit_derivative(nir_instr_as_tex(instr))) if (nir_tex_instr_has_implicit_derivative(nir_instr_as_tex(instr)))
instr->pass_flags = GCM_INSTR_PINNED; instr->pass_flags = GCM_INSTR_SCHEDULE_EARLIER_ONLY;
break; break;
case nir_instr_type_deref: case nir_instr_type_deref:
@@ -353,6 +354,12 @@ gcm_schedule_late_def(nir_ssa_def *def, void *void_state)
return true; return true;
} }
if (def->parent_instr->pass_flags & GCM_INSTR_SCHEDULE_EARLIER_ONLY &&
lca != def->parent_instr->block &&
nir_block_dominates(def->parent_instr->block, lca)) {
lca = def->parent_instr->block;
}
/* We now have the LCA of all of the uses. If our invariants hold, /* We now have the LCA of all of the uses. If our invariants hold,
* this is dominated by the block that we chose when scheduling early. * this is dominated by the block that we chose when scheduling early.
* We now walk up the dominance tree and pick the lowest block that is * We now walk up the dominance tree and pick the lowest block that is