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

glsl: drop opt_dead_code_local

Browse Source 
This does nothing useful anymore as we now convert to nir at compile
time which will handle all this for us.

shader-db radeonsi:

TOTALS FROM AFFECTED SHADERS (63/168079)
  SGPRS: 3248.00 -> 3208.00 (-1.23 %)
  VGPRS: 2224.00 -> 2228.00 (0.18 %)
  Spilled SGPRs: 0.00 -> 0.00 (0.00 %)
  Spilled VGPRs: 0.00 -> 0.00 (0.00 %)
  Private memory VGPRs: 0.00 -> 0.00 (0.00 %)
  Scratch size: 0.00 -> 0.00 (0.00 %) dwords per thread
  Code Size: 138484.00 -> 138068.00 (-0.30 %) bytes
  Max Waves: 877.00 -> 877.00 (0.00 %)
  Outputs: 0.00 -> 0.00 (0.00 %)
  Patch Outputs: 0.00 -> 0.00 (0.00 %)

shader-db Iris (BDW):

total instructions in shared programs: 17805897 -> 17805917 (<.01%)
instructions in affected programs: 1240 -> 1260 (1.61%)
helped: 0
HURT: 8
HURT stats (abs)   min: 1 max: 4 x̄: 2.50 x̃: 2
HURT stats (rel)   min: 0.39% max: 7.14% x̄: 4.26% x̃: 4.06%
95% mean confidence interval for instructions value: 1.61 3.39
95% mean confidence interval for instructions %-change: 2.01% 6.51%
Instructions are HURT.

total cycles in shared programs: 856868505 -> 856876266 (<.01%)
cycles in affected programs: 2879959 -> 2887720 (0.27%)
helped: 79
HURT: 100
helped stats (abs) min: 1 max: 742 x̄: 61.96 x̃: 12
helped stats (rel) min: <.01% max: 41.84% x̄: 1.17% x̃: 0.20%
HURT stats (abs)   min: 1 max: 1231 x̄: 126.56 x̃: 14
HURT stats (rel)   min: <.01% max: 33.98% x̄: 3.32% x̃: 0.30%
95% mean confidence interval for cycles value: 7.37 79.35
95% mean confidence interval for cycles %-change: 0.29% 2.38%
Cycles are HURT.

LOST:   1
GAINED: 4

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33008>
This commit is contained in:
Timothy Arceri
2025-01-14 16:38:56 +11:00
committed by Marge Bot
parent 5bd9693578
commit 6bca0cc3d9
3 changed files with 0 additions and 355 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
src/compiler/glsl/glsl_parser_extras.cpp
View File

@@ -2507,7 +2507,6 @@ do_common_optimization(exec_list *ir, bool linked,
OPT(opt_flip_matrices, ir);
OPT(do_dead_code_unlinked, ir);
OPT(do_dead_code_local, ir);
OPT(do_tree_grafting, ir);
OPT(do_minmax_prune, ir);
OPT(do_rebalance_tree, ir);

1
src/compiler/glsl/meson.build
View File

@@ -178,7 +178,6 @@ files_libglsl = files(
'opt_algebraic.cpp',
'opt_dead_builtin_variables.cpp',
'opt_dead_code.cpp',
'opt_dead_code_local.cpp',
'opt_flatten_nested_if_blocks.cpp',
'opt_flip_matrices.cpp',
'opt_function_inlining.cpp',

353
src/compiler/glsl/opt_dead_code_local.cpp
View File

@@ -1,353 +0,0 @@
/*
* Copyright © 2010 Intel Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS IN THE SOFTWARE.
*/
/**
* \file opt_dead_code_local.cpp
*
* Eliminates local dead assignments from the code.
*
* This operates on basic blocks, tracking assignments and finding if
* they're used before the variable is completely reassigned.
*
* Compare this to ir_dead_code.cpp, which operates globally looking
* for assignments to variables that are never read.
*/
#include "ir.h"
#include "ir_basic_block.h"
#include "ir_optimization.h"
#include "compiler/glsl_types.h"
static bool debug = false;
namespace {
class assignment_entry : public exec_node
{
public:
/* override operator new from exec_node */
DECLARE_LINEAR_ZALLOC_CXX_OPERATORS(assignment_entry)
assignment_entry(ir_variable *lhs, ir_assignment *ir)
{
assert(lhs);
assert(ir);
this->lhs = lhs;
this->ir = ir;
this->unused = ir->write_mask;
}
ir_variable *lhs;
ir_assignment *ir;
/* bitmask of xyzw channels written that haven't been used so far. */
int unused;
};
class kill_for_derefs_visitor : public ir_hierarchical_visitor {
public:
using ir_hierarchical_visitor::visit;
kill_for_derefs_visitor(exec_list *assignments)
{
this->assignments = assignments;
}
void use_channels(ir_variable *const var, int used)
{
foreach_in_list_safe(assignment_entry, entry, this->assignments) {
if (entry->lhs == var) {
if (glsl_type_is_scalar(var->type) || glsl_type_is_vector(var->type)) {
if (debug)
printf("used %s (0x%01x - 0x%01x)\n", entry->lhs->name,
entry->unused, used & 0xf);
entry->unused &= ~used;
if (!entry->unused)
entry->remove();
} else {
if (debug)
printf("used %s\n", entry->lhs->name);
entry->remove();
}
}
}
}
virtual ir_visitor_status visit(ir_dereference_variable *ir)
{
use_channels(ir->var, ~0);
return visit_continue;
}
virtual ir_visitor_status visit(ir_swizzle *ir)
{
ir_dereference_variable *deref = ir->val->as_dereference_variable();
if (!deref)
return visit_continue;
int used = 0;
used |= 1 << ir->mask.x;
if (ir->mask.num_components > 1)
used |= 1 << ir->mask.y;
if (ir->mask.num_components > 2)
used |= 1 << ir->mask.z;
if (ir->mask.num_components > 3)
used |= 1 << ir->mask.w;
use_channels(deref->var, used);
return visit_continue_with_parent;
}
virtual ir_visitor_status visit_leave(ir_emit_vertex *)
{
/* For the purpose of dead code elimination, emitting a vertex counts as
* "reading" all of the currently assigned output variables.
*/
foreach_in_list_safe(assignment_entry, entry, this->assignments) {
if (entry->lhs->data.mode == ir_var_shader_out) {
if (debug)
printf("kill %s\n", entry->lhs->name);
entry->remove();
}
}
return visit_continue;
}
private:
exec_list *assignments;
};
class array_index_visit : public ir_hierarchical_visitor {
public:
array_index_visit(ir_hierarchical_visitor *v)
{
this->visitor = v;
}
virtual ir_visitor_status visit_enter(class ir_dereference_array *ir)
{
ir->array_index->accept(visitor);
return visit_continue;
}
static void run(ir_instruction *ir, ir_hierarchical_visitor *v)
{
array_index_visit top_visit(v);
ir->accept(& top_visit);
}
ir_hierarchical_visitor *visitor;
};
} /* unnamed namespace */
/**
* Adds an entry to the available copy list if it's a plain assignment
* of a variable to a variable.
*/
static bool
process_assignment(linear_ctx *lin_ctx, ir_assignment *ir, exec_list *assignments)
{
ir_variable *var = NULL;
bool progress = false;
kill_for_derefs_visitor v(assignments);
/* If this is an assignment of the form "foo = foo;", remove the whole
* instruction and be done with it.
*/
const ir_variable *const lhs_var = ir->whole_variable_written();
if (lhs_var != NULL && lhs_var == ir->rhs->whole_variable_referenced()) {
ir->remove();
return true;
}
/* Kill assignment entries for things used to produce this assignment. */
ir->rhs->accept(&v);
/* Kill assignment enties used as array indices.
*/
array_index_visit::run(ir->lhs, &v);
var = ir->lhs->variable_referenced();
assert(var);
/* Now, check if we did a whole-variable assignment. */
ir_dereference_variable *deref_var = ir->lhs->as_dereference_variable();
/* If it's a vector type, we can do per-channel elimination of
* use of the RHS.
*/
if (deref_var && (glsl_type_is_scalar(deref_var->var->type) ||
glsl_type_is_vector(deref_var->var->type))) {
if (debug)
printf("looking for %s.0x%01x to remove\n", var->name,
ir->write_mask);
foreach_in_list_safe(assignment_entry, entry, assignments) {
if (entry->lhs != var)
continue;
/* Skip if the assignment we're trying to eliminate isn't a plain
* variable deref. */
if (entry->ir->lhs->ir_type != ir_type_dereference_variable)
continue;
int remove = entry->unused & ir->write_mask;
if (debug) {
printf("%s 0x%01x - 0x%01x = 0x%01x\n",
var->name,
entry->ir->write_mask,
remove, entry->ir->write_mask & ~remove);
}
if (remove) {
progress = true;
if (debug) {
printf("rewriting:\n ");
entry->ir->print();
printf("\n");
}
entry->ir->write_mask &= ~remove;
entry->unused &= ~remove;
if (entry->ir->write_mask == 0) {
/* Delete the dead assignment. */
entry->ir->remove();
entry->remove();
} else {
void *mem_ctx = ralloc_parent(entry->ir);
/* Reswizzle the RHS arguments according to the new
* write_mask.
*/
unsigned components[4];
unsigned channels = 0;
unsigned next = 0;
for (int i = 0; i < 4; i++) {
if ((entry->ir->write_mask | remove) & (1 << i)) {
if (!(remove & (1 << i)))
components[channels++] = next;
next++;
}
}
entry->ir->rhs = new(mem_ctx) ir_swizzle(entry->ir->rhs,
components,
channels);
if (debug) {
printf("to:\n ");
entry->ir->print();
printf("\n");
}
}
}
}
} else if (ir->whole_variable_written() != NULL) {
/* We did a whole-variable assignment. So, any instruction in
* the assignment list with the same LHS is dead.
*/
if (debug)
printf("looking for %s to remove\n", var->name);
foreach_in_list_safe(assignment_entry, entry, assignments) {
if (entry->lhs == var) {
if (debug)
printf("removing %s\n", var->name);
entry->ir->remove();
entry->remove();
progress = true;
}
}
}
/* Add this instruction to the assignment list available to be removed. */
assignment_entry *entry = new(lin_ctx) assignment_entry(var, ir);
assignments->push_tail(entry);
if (debug) {
printf("add %s\n", var->name);
printf("current entries\n");
foreach_in_list(assignment_entry, entry, assignments) {
printf(" %s (0x%01x)\n", entry->lhs->name, entry->unused);
}
}
return progress;
}
static void
dead_code_local_basic_block(ir_instruction *first,
ir_instruction *last,
void *data)
{
ir_instruction *ir;
exec_node *node, *node_next;
/* List of avaialble_copy */
exec_list assignments;
bool *out_progress = (bool *)data;
bool progress = false;
void *ctx = ralloc_context(NULL);
linear_ctx *lin_ctx = linear_context(ctx);
/* Safe looping, since process_assignment */
for (node = first, node_next = first->next;;
node = node_next, node_next = node->next) {
ir = (ir_instruction *) node;
ir_assignment *ir_assign = ir->as_assignment();
if (debug) {
ir->print();
printf("\n");
}
if (ir_assign) {
progress = process_assignment(lin_ctx, ir_assign, &assignments) ||
progress;
} else {
kill_for_derefs_visitor kill(&assignments);
ir->accept(&kill);
}
if (ir == last)
break;
}
*out_progress = progress;
ralloc_free(ctx);
}
/**
* Does a copy propagation pass on the code present in the instruction stream.
*/
bool
do_dead_code_local(exec_list *instructions)
{
bool progress = false;
call_for_basic_blocks(instructions, dead_code_local_basic_block, &progress);
return progress;
}