fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-25 00:00:11 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

aco/lower_branches: do eliminate_useless_exec_writes_in_block() during branch lowering.

Browse source 
Totals from 728 (0.92% of 79395) affected shaders: (Navi31)

Instrs: 452926 -> 452161 (-0.17%)
CodeSize: 2255536 -> 2252504 (-0.13%)
Latency: 1683404 -> 1683470 (+0.00%); split: -0.01%, +0.01%
InvThroughput: 210887 -> 210888 (+0.00%); split: -0.00%, +0.00%
SALU: 77865 -> 77106 (-0.97%)
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32477>
This commit is contained in:
Daniel Schürmann 2024-11-21 10:40:39 +01:00 • committed by Marge Bot
parent eecdb45d61
commit 2b5a893e29
1 changed files with 64 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

65
src/amd/compiler/aco_lower_branches.cpp
View file

@ -12,8 +12,11 @@ namespace {
struct branch_ctx {
Program* program;
std::vector<bool> blocks_incoming_exec_used;
branch_ctx(Program* program_) : program(program_) {}
branch_ctx(Program* program_)
: program(program_), blocks_incoming_exec_used(program_->blocks.size(), true)
{}
};
void
@ -35,6 +38,65 @@ remove_linear_successor(branch_ctx& ctx, Block& block, uint32_t succ_index)
}
}
void
eliminate_useless_exec_writes_in_block(branch_ctx& ctx, Block& block)
{
/* Check if any successor needs the outgoing exec mask from the current block. */
bool exec_write_used;
if (block.kind & block_kind_end_with_regs) {
/* Last block of a program with succeed shader part should respect final exec write. */
exec_write_used = true;
} else if (block.linear_succs.empty() && !block.instructions.empty() &&
block.instructions.back()->opcode == aco_opcode::s_setpc_b64) {
/* This block ends in a long jump and exec might be needed for the next shader part. */
exec_write_used = true;
} else {
/* blocks_incoming_exec_used is initialized to true, so this is correct even for loops. */
exec_write_used =
std::any_of(block.linear_succs.begin(), block.linear_succs.end(),
[&ctx](int succ_idx) { return ctx.blocks_incoming_exec_used[succ_idx]; });
}
/* Go through all instructions and eliminate useless exec writes. */
for (int i = block.instructions.size() - 1; i >= 0; --i) {
aco_ptr<Instruction>& instr = block.instructions[i];
/* See if the current instruction needs or writes exec. */
bool needs_exec = needs_exec_mask(instr.get());
bool writes_exec =
instr->writes_exec() && instr->definitions[0].regClass() == ctx.program->lane_mask;
/* See if we found an unused exec write. */
if (writes_exec && !exec_write_used) {
/* Don't eliminate an instruction that writes registers other than exec and scc.
* It is possible that this is eg. an s_and_saveexec and the saved value is
* used by a later branch.
*/
bool writes_other = std::any_of(instr->definitions.begin(), instr->definitions.end(),
[](const Definition& def) -> bool
{ return def.physReg() != exec && def.physReg() != scc; });
if (!writes_other) {
instr.reset();
continue;
}
}
/* For a newly encountered exec write, clear the used flag. */
if (writes_exec)
exec_write_used = false;
/* If the current instruction needs exec, mark it as used. */
exec_write_used |= needs_exec;
}
/* Remember if the current block needs an incoming exec mask from its predecessors. */
ctx.blocks_incoming_exec_used[block.index] = exec_write_used;
/* Cleanup: remove deleted instructions from the vector. */
auto new_end = std::remove(block.instructions.begin(), block.instructions.end(), nullptr);
block.instructions.resize(new_end - block.instructions.begin());
}
/**
* Check if the branch instruction can be removed:
* This is beneficial when executing the next block with an empty exec mask
@ -196,6 +258,7 @@ lower_branches(Program* program)
for (int i = program->blocks.size() - 1; i >= 0; i--) {
Block& block = program->blocks[i];
lower_branch_instruction(ctx, block);
eliminate_useless_exec_writes_in_block(ctx, block);
}
}