[maurycy]

@pablogsal AMA. This is a sneak peak, only get_stack_trace for now. I'm playing with async. I'm cooking it for the past two weeks. It's not bad so far:

Metric	Baseline (main 025cc26)	This PR	Change
Work rate	339,576.5 calls/sec	3,774,163.0 calls/sec	11.11x faster
Sample rate	325,989.3 samples/sec	2,842,399.3 samples/sec	8.72x faster
Avg call time	2.94 µs	0.26 µs	-91.2%
Total attempts	979,732	8,532,147	8.71x more
Successful samples	977,970	8,527,201	-
Failed samples	1,762	4,946	-
Success rate	99.82%	99.94%	+0.12 pp
Work efficiency	96.2%	75.4%	-

I compared binary outputs. I will open a PR in python/cpython once I'm happy but more than open for any comments.

2026-06-04T00:00:55.450053000+0200 maurycy@gimel /Users/maurycy/src/github.com/maurycy/cpython (vmremap 7299a47) % sudo ./python.exe Tools/inspection/benchmark_external_inspection.py --duration 3

============================================================
get_stack_trace() Benchmark Results
============================================================

Basic Statistics:
  Wall time:           3.000 seconds
  Total attempts:      8,532,147
  Successful samples:  8,527,201
  Failed samples:      4,946
  Success rate:        99.94%

Performance Metrics:
  Average call time:   0.26 µs
  Work rate:           3774163.0 calls/sec
  Sample rate:         2842399.3 samples/sec
  Total work time:     2.261 seconds
  Work efficiency:     75.4%

v.

2026-06-04T00:03:25.240297000+0200 maurycy@gimel /Users/maurycy/work/cpython-main (main 025cc26) % sudo ./python.exe Tools/inspection/benchmark_external_inspection.py --duration 3

============================================================
get_stack_trace() Benchmark Results
============================================================

Basic Statistics:
  Wall time:           3.000 seconds
  Total attempts:      979,732
  Successful samples:  977,970
  Failed samples:      1,762
  Success rate:        99.82%

Performance Metrics:
  Average call time:   2.94 µs
  Work rate:           339576.5 calls/sec
  Sample rate:         325989.3 samples/sec
  Total work time:     2.885 seconds
  Work efficiency:     96.2%

ref https://github.com/apple-oss-distributions/xnu

[pablogsal]

Dude these numbers are amazing!

[maurycy]

python#161011