Home
last modified time | relevance | path

Searched refs:parallelcopies (Results 1 – 8 of 8) sorted by relevance

/third_party/mesa3d/src/amd/compiler/
Daco_lower_to_cssa.cpp67 std::vector<std::vector<copy>> parallelcopies; /* copies per block */ member
77 ctx.parallelcopies.resize(ctx.program->blocks.size()); in collect_parallelcopies()
119 ctx.parallelcopies[preds[i]].emplace_back(copy{Definition(tmp), op}); in collect_parallelcopies()
206 for (const copy& cp : ctx.parallelcopies[block_idx]) { in intersects()
444 if (ctx.parallelcopies[i].empty()) in emit_parallelcopies()
452 for (const copy& cp : ctx.parallelcopies[i]) { in emit_parallelcopies()
Daco_register_allocation.cpp767 std::vector<std::pair<Operand, Definition>>& parallelcopies, in update_renames() argument
771 for (std::pair<Operand, Definition>& copy : parallelcopies) { in update_renames()
779 auto it = parallelcopies.begin(); in update_renames()
780 while (it != parallelcopies.end()) { in update_renames()
794 it = parallelcopies.erase(it); in update_renames()
803 for (std::pair<Operand, Definition>& other : parallelcopies) { in update_renames()
809 it = parallelcopies.erase(it); in update_renames()
842 for (std::pair<Operand, Definition>& pc : parallelcopies) { in update_renames()
1041 std::vector<std::pair<Operand, Definition>>& parallelcopies, in get_reg_for_create_vector_copy() argument
1078 parallelcopies.emplace_back(instr->operands[i], pc_def); in get_reg_for_create_vector_copy()
[all …]
DREADME.md93 The next step is a pass out of SSA by inserting parallelcopies at the end of blocks to match the ph…
/third_party/mesa3d/docs/relnotes/
D20.0.0.rst767 - aco: handle phi affinities transitively through parallelcopies
768 - aco: ignore parallelcopies to the same register on jump threading
D21.1.0.rst1362 - aco/validate: ensure that Operand and Definition size matches for parallelcopies
1366 - aco/optimizer: convert extract_vector with index 0 into parallelcopies if possible
1409 - aco/cssa: don't create parallelcopies for constants and exec
D22.1.0.rst863 - aco/ra: fix stride check on subdword parallelcopies for create_vector
D20.2.0.rst1234 - aco: coalesce parallelcopies during register allocation
D21.2.0.rst1466 - ir3: Improve printing of array parallelcopies/phis