fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-06 02:58:05 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

aco/ra: discard tmp_file after get_regs_for_copies fails

Browse source 
get_regs_for_copies() might change tmp_file, so we need to create a new
one.

Recursively handle_call() with an modified register_file instead of
repeating get_regs_for_copies().

This also updates register_file for the caller.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38811>
This commit is contained in:
Rhys Perry 2025-12-02 11:25:49 +00:00 • committed by Marge Bot
parent 159154a741
commit 6e6e40ab5c
1 changed files with 8 additions and 19 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

27
src/amd/compiler/aco_register_allocation.cpp
View file

@ -3614,19 +3614,15 @@ emit_parallel_copy(ra_ctx& ctx, std::vector<parallelcopy>& copies,
}
std::vector<Instruction*>
split_blocking_vectors(ra_ctx& ctx, std::vector<unsigned>& vars, RegisterFile& file)
split_blocking_vectors(ra_ctx& ctx, std::vector<unsigned>& vars, RegisterFile& register_file)
{
std::vector<unsigned> blocked_defs;
std::vector<Instruction*> splits;
for (auto var_it = vars.begin(); var_it != vars.end();) {
unsigned id = *var_it;
for (unsigned id : vars) {
RegClass rc = ctx.program->temp_rc[id];
PhysReg start = ctx.assignments[id].reg;
if (rc.size() == 1) {
++var_it;
if (rc.size() == 1)
continue;
}
aco_ptr<Instruction> split = aco_ptr<Instruction>(
create_instruction(aco_opcode::p_split_vector, Format::PSEUDO, 1, rc.size()));
@ -3639,17 +3635,11 @@ split_blocking_vectors(ra_ctx& ctx, std::vector<unsigned>& vars, RegisterFile& f
ctx.assignments.emplace_back(reg, def_rc);
split->definitions[def_idx] = Definition(tmp, reg);
if (get_reg_specified(ctx, file, def_rc, split, reg, 0))
file.fill(split->definitions[def_idx]);
else
blocked_defs.push_back(tmp.id());
register_file.fill(split->definitions[def_idx]);
}
splits.push_back(split.release());
var_it = vars.erase(var_it);
}
vars.insert(vars.end(), blocked_defs.begin(), blocked_defs.end());
return splits;
}
@ -3725,14 +3715,13 @@ handle_call(ra_ctx& ctx, aco_ptr<Instruction>& instr, BITSET_DECLARE(call_clobbe
bool success = get_regs_for_copies(ctx, tmp_file, copies, vars, instr, PhysRegInterval{});
if (success) {
parallelcopy.insert(parallelcopy.end(), copies.begin(), copies.end());
update_renames(ctx, register_file, parallelcopy, instr);
} else {
vector_splits = split_blocking_vectors(ctx, vars, tmp_file);
success = get_regs_for_copies(ctx, tmp_file, parallelcopy, vars, instr, PhysRegInterval{});
assert(vector_splits.empty());
/* With all blocking vars being scalar, assigning registers should always succeed. */
assert(success);
vector_splits = split_blocking_vectors(ctx, vars, register_file);
handle_call(ctx, instr, call_clobbered_regs, vector_splits, parallelcopy, register_file);
}
update_renames(ctx, register_file, parallelcopy, instr);
}
} /* end namespace */