vectorcall - Branch - ifduyue/python-xxhash

perf: METH_FASTCALL, tp_vectorcall, and related optimizations

#147Merged

Comparing

vectorcall

(

a974b6b

) with

master

(

1585d2b

)

Untouched: 10

New: 24

Skipped: 86

Benchmarks

120 total

test_xxh32_ctor_empty

tests/test_benchmark.py

N/A

N/A84.2 µs

test_xxh32_ctor

tests/test_benchmark.py

N/A

N/A85.9 µs

test_xxh32_ctor_seed

tests/test_benchmark.py

N/A

N/A86.8 µs

test_xxh3_64_intdigest_5b_seed_kw

tests/test_benchmark.py

N/A

N/A85.6 µs

test_xxh32_hexdigest_5b

tests/test_benchmark.py

N/A

N/A85.1 µs

test_xxh64_intdigest_5b

tests/test_benchmark.py

N/A

N/A84.8 µs

test_xxh64_stream_intdigest_2mb

tests/test_benchmark.py

N/A

N/A1.8 ms

test_xxh64_ctor

tests/test_benchmark.py

N/A

N/A86.7 µs

test_xxh64_intdigest_str

tests/test_benchmark.py

N/A

N/A86 µs

test_xxh32_intdigest_5b_seed_kw

tests/test_benchmark.py

N/A

N/A85.5 µs

test_xxh32_intdigest_5b_seed

tests/test_benchmark.py

N/A

N/A87.1 µs

test_xxh3_128_ctor

tests/test_benchmark.py

N/A

N/A87.8 µs

test_xxh32_intdigest_str

tests/test_benchmark.py

N/A

N/A85.4 µs

test_xxh3_128_intdigest_5b

tests/test_benchmark.py

N/A

N/A86.9 µs

test_xxh3_128_intdigest_str

tests/test_benchmark.py

N/A

N/A87 µs

test_xxh3_64_ctor

tests/test_benchmark.py

N/A

N/A87.5 µs

test_xxh3_64_stream_intdigest_2mb

tests/test_benchmark.py

N/A

N/A1.6 ms

test_xxh3_64_intdigest_5b

tests/test_benchmark.py

N/A

N/A84.7 µs

test_xxh3_64_intdigest_str

tests/test_benchmark.py

N/A

N/A85.3 µs

test_xxh64_hexdigest_5b

tests/test_benchmark.py

N/A

N/A86.5 µs

test_xxh32_stream_intdigest_2mb

tests/test_benchmark.py

N/A

N/A2.2 ms

test_xxh32_intdigest_5b

tests/test_benchmark.py

N/A

N/A85 µs

test_xxh64_intdigest_5b_seed_kw

tests/test_benchmark.py

N/A

N/A85.5 µs

test_xxh3_128_stream_intdigest_2mb

tests/test_benchmark.py

N/A

N/A1.6 ms

test_xxh3_128_intdigest_1kb

tests/test_benchmark.py

89.4 µs89 µs

Commits

Click on a commit to change the comparison range

Base

master

1585d2b

+0.75%

perf: add tp_vectorcall to all four hash types

1585d2b

19 hours ago

by ifduyue

-1.26%

fix: remove unsupported min_rounds from codspeed benchmark markers

f49eaa4

19 hours ago

by ifduyue

+0.05%

fix: don't overwrite tp_alloc error with PyErr_NoMemory in vectorcall

da76b27

19 hours ago

by ifduyue

-0.02%

fix: release buffer on tp_alloc failure in vectorcall functions

9647f32

18 hours ago

by ifduyue

+0.02%

refactor: reuse PYXXH*_do_update helpers in vectorcall functions

cba0f28

18 hours ago

by ifduyue

+0.06%

fix: use buf.obj instead of buf.buf for buffer-acquired check

14675ad

18 hours ago

by ifduyue

+0.28%

bench: add 2MB throughput benchmarks

a974b6b

17 hours ago

by ifduyue

Home Terms Privacy Docs