• Home
Name Date Size #Lines LOC

..--

pattern_conversion/04-Jul-2025-708450

README.mdD04-Jul-2025237 54

cast_all_constant_to_floating.cppD04-Jul-20252.7 KiB7355

cast_all_constant_to_floating.hD04-Jul-2025217 116

constant_fold.cppD04-Jul-202527 KiB710607

constant_fold.hD04-Jul-2025693 3323

constant_map.cppD04-Jul-202512.4 KiB383341

constant_map.hD04-Jul-20254.5 KiB11881

deduplicate_initializers.cppD04-Jul-20253.3 KiB10283

deduplicate_initializers.hD04-Jul-2025247 159

eliminate_unused_items.cppD04-Jul-2025491 1914

eliminate_unused_items.hD04-Jul-2025343 157

eval_peephole.cppD04-Jul-20255 KiB155123

eval_peephole.hD04-Jul-2025220 148

fixup_onnx_controlflow.cppD04-Jul-202526.8 KiB740551

fixup_onnx_controlflow.hD04-Jul-2025221 116

function_extraction.cppD04-Jul-202537.4 KiB1,185993

function_extraction.hD04-Jul-20252.2 KiB6514

function_substitution.cppD04-Jul-20257.3 KiB196171

function_substitution.hD04-Jul-2025135 105

helper.cppD04-Jul-20259 KiB299261

helper.hD04-Jul-20252.1 KiB7656

list_model_parameters.cppD04-Jul-20256.6 KiB194152

list_model_parameters.hD04-Jul-2025239 127

naming.cppD04-Jul-20255.8 KiB206170

naming.hD04-Jul-2025746 2719

onnx_log.cppD04-Jul-2025517 2820

onnx_log.hD04-Jul-2025584 2416

peephole.cppD04-Jul-202537.7 KiB1,071751

peephole.hD04-Jul-2025213 138

prepare_division_for_onnx.cppD04-Jul-20251.6 KiB4635

prepare_division_for_onnx.hD04-Jul-2025451 185

preprocess_for_onnx.cppD04-Jul-20258.1 KiB232131

preprocess_for_onnx.hD04-Jul-2025155 105

remove_inplace_ops_for_onnx.cppD04-Jul-202530.4 KiB893620

remove_inplace_ops_for_onnx.hD04-Jul-2025201 127

scalar_type_analysis.cppD04-Jul-202516.6 KiB482352

scalar_type_analysis.hD04-Jul-2025280 149

shape_type_inference.cppD04-Jul-202590.8 KiB2,5132,135

shape_type_inference.hD04-Jul-20253.9 KiB9947

unpack_quantized_weights.cppD04-Jul-202529 KiB769661

unpack_quantized_weights.hD04-Jul-2025427 1813

README.md

1The optimization passes in this directory work exclusively on ONNX-style IRs,
2e.g., IRs that have had ToONNX applied to them.  ONNX defines operators
3differently from ATen, so there are different opportunities for peephole
4optimization.
5