frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-11-16 23:44:53 +08:00
Code Issues Packages Projects Releases Wiki Activity

dynamo tracing perf: cache cleaned_instructions: 33.7 -> 30.0 (#143070)

Browse Source 
See #143056 for overall docs.

This PR: Cache the interesting/expensive bits of `cleaned_instructions()`

Pull Request resolved: https://github.com/pytorch/pytorch/pull/143070
Approved by: https://github.com/jansel
This commit is contained in:
Aaron Orenstein
2024-12-23 08:32:43 -08:00
committed by PyTorch MergeBot
parent 51a7ecde80
commit 3df12d38cf
2 changed files with 63 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
benchmarks/dynamo/pr_time_benchmarks/expected_results.csv
View File

@ -18,7 +18,7 @@ add_loop_inductor_gpu,compile_time_instruction_count,27530000000,0.015
basic_modules_ListOfLinears_eager,compile_time_instruction_count,980406675,0.015
basic_modules_ListOfLinears_eager,compile_time_instruction_count,930000000,0.015
@ -38,7 +38,7 @@ update_hint_regression,compile_time_instruction_count,1686000000,0.02
sum_floordiv_regression,compile_time_instruction_count,1049000000,0.015
sum_floordiv_regression,compile_time_instruction_count,1040000000,0.015
@ -50,7 +50,7 @@ aotdispatcher_inference_nosubclass_cpu,compile_time_instruction_count,2018000000
aotdispatcher_inference_subclass_cpu,compile_time_instruction_count,5848000000,0.015
aotdispatcher_inference_subclass_cpu,compile_time_instruction_count,5800000000,0.015
@ -62,4 +62,4 @@ aotdispatcher_training_nosubclass_cpu,compile_time_instruction_count,3863000000,
aotdispatcher_training_subclass_cpu,compile_time_instruction_count,10390000000,0.015
aotdispatcher_training_subclass_cpu,compile_time_instruction_count,10340000000,0.015

1 add_loop_eager compile_time_instruction_count 3066000000 0.015
18
19
20
21
22
23
24
38
39
40
41
42
43
44
50
51
52
53
54
55
56
62
63
64
65