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

brw: Delete pull constant lowering

Browse Source 
Now that we never shrink ranges in the backend, we never lower push
constants to pull constants late in the backend either.  get_pull_loc
will never return true, and so all of brw_lower_constant_loads becomes
a noop.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32841>
This commit is contained in:
Kenneth Graunke
2025-01-01 04:56:53 -08:00
committed by Marge Bot
parent 1ab4fe2dd6
commit 93e186e1a4
4 changed files with 0 additions and 99 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
src/intel/compiler/brw_fs.cpp
View File

@@ -1149,31 +1149,6 @@ fs_visitor::assign_constant_locations()
push_constant_loc[u] = u;
}
bool
fs_visitor::get_pull_locs(const brw_reg &src,
unsigned *out_surf_index,
unsigned *out_pull_index)
{
assert(src.file == UNIFORM);
if (src.nr < UBO_START)
return false;
const struct brw_ubo_range *range =
&prog_data->ubo_ranges[src.nr - UBO_START];
/* If this access is in our (reduced) range, use the push data. */
if (src.offset / 32 < range->length)
return false;
*out_surf_index = range->block;
*out_pull_index = (32 * range->start + src.offset) / 4;
prog_data->has_ubo_pull = true;
return true;
}
/**
* Get the mask of SIMD channels enabled during dispatch and not yet disabled
* by discard. Due to the layout of the sample mask in the fragment shader

2
src/intel/compiler/brw_fs.h
View File

@@ -313,8 +313,6 @@ public:
unsigned payload_node_count,
int *payload_last_use_ip) const;
void assign_constant_locations();
bool get_pull_locs(const brw_reg &src, unsigned *out_surf_index,
unsigned *out_pull_index);
void invalidate_analysis(brw::analysis_dependency_class c);
void vfail(const char *msg, va_list args);

71
src/intel/compiler/brw_lower.cpp
View File

@@ -49,77 +49,6 @@ brw_lower_scalar_fp64_MAD(fs_visitor &s)
return progress;
}
/**
* Replace UNIFORM register file access with either UNIFORM_PULL_CONSTANT_LOAD
* or VARYING_PULL_CONSTANT_LOAD instructions which load values into VGRFs.
*/
bool
brw_lower_constant_loads(fs_visitor &s)
{
unsigned index, pull_index;
bool progress = false;
foreach_block_and_inst_safe (block, fs_inst, inst, s.cfg) {
/* Set up the annotation tracking for new generated instructions. */
const fs_builder ibld(&s, block, inst);
for (int i = 0; i < inst->sources; i++) {
if (inst->src[i].file != UNIFORM)
continue;
/* We'll handle this case later */
if (inst->opcode == SHADER_OPCODE_MOV_INDIRECT && i == 0)
continue;
if (!s.get_pull_locs(inst->src[i], &index, &pull_index))
continue;
assert(inst->src[i].stride == 0);
const unsigned block_sz = 64; /* Fetch one cacheline at a time. */
const fs_builder ubld = ibld.exec_all().group(block_sz / 4, 0);
const brw_reg dst = ubld.vgrf(BRW_TYPE_UD);
const unsigned base = pull_index * 4;
brw_reg srcs[PULL_UNIFORM_CONSTANT_SRCS];
srcs[PULL_UNIFORM_CONSTANT_SRC_SURFACE] = brw_imm_ud(index);
srcs[PULL_UNIFORM_CONSTANT_SRC_OFFSET] = brw_imm_ud(base & ~(block_sz - 1));
srcs[PULL_UNIFORM_CONSTANT_SRC_SIZE] = brw_imm_ud(block_sz);
ubld.emit(FS_OPCODE_UNIFORM_PULL_CONSTANT_LOAD, dst,
srcs, PULL_UNIFORM_CONSTANT_SRCS);
/* Rewrite the instruction to use the temporary VGRF. */
inst->src[i].file = VGRF;
inst->src[i].nr = dst.nr;
inst->src[i].offset = (base & (block_sz - 1)) +
inst->src[i].offset % 4;
progress = true;
}
if (inst->opcode == SHADER_OPCODE_MOV_INDIRECT &&
inst->src[0].file == UNIFORM) {
if (!s.get_pull_locs(inst->src[0], &index, &pull_index))
continue;
ibld.VARYING_PULL_CONSTANT_LOAD(inst->dst,
brw_imm_ud(index),
brw_reg() /* surface_handle */,
inst->src[1],
pull_index * 4, 4, 1);
inst->remove(block);
progress = true;
}
}
s.invalidate_analysis(DEPENDENCY_INSTRUCTIONS);
return progress;
}
bool
brw_lower_load_payload(fs_visitor &s)
{

1
src/intel/compiler/brw_opt.cpp
View File

@@ -44,7 +44,6 @@ brw_optimize(fs_visitor &s)
})
s.assign_constant_locations();
OPT(brw_lower_constant_loads);
if (s.compiler->lower_dpas)
OPT(brw_lower_dpas);