patchback/backports/3.12/f7cac7e63f18691e4261af353e84f9073b16624a/pr-10601 - Branch - aio-libs/aiohttp

[PR #10601/f7cac7e6 backport][3.12] Reduce WebSocket buffer slicing overhead

#10640Merged

Comparing

patchback/backports/3.12/f7cac7e63f18691e4261af353e84f9073b16624a/pr-10601

(

105229c

) with

3.12

(

5e20fe1

)

+17%

Improvements: 1

Regressions: 0

Untouched: 46

New: 0

Dropped: 0

Ignored: 2

Benchmarks

Improved

test_read_one_hundred_websocket_text_messages[pyloop]

tests/test_benchmarks_http_websocket.py::test_read_one_hundred_websocket_text_messages[pyloop]

+17%

307 µs

263 µs

Passed

test_load_cookies_into_temp_cookiejar

tests/test_benchmarks_cookiejar.py::test_load_cookies_into_temp_cookiejar

+1%

202 µs

200.3 µs

test_read_large_binary_websocket_messages[pyloop]

tests/test_benchmarks_http_websocket.py::test_read_large_binary_websocket_messages[pyloop]

+1%

9.1 ms

9 ms

test_client_request_update_cookies[pyloop]

tests/test_benchmarks_client_request.py::test_client_request_update_cookies[pyloop]

+1%

170.1 µs

168.9 µs

test_web_response_with_bytes_body

tests/test_benchmarks_web_response.py::test_web_response_with_bytes_body

+1%

540.9 µs

537.8 µs

test_resolve_static_root_route[pyloop]

tests/test_benchmarks_web_urldispatcher.py::test_resolve_static_root_route[pyloop]

987.1 µs

983.2 µs

test_resolve_gitapi_root[pyloop]

tests/test_benchmarks_web_urldispatcher.py::test_resolve_gitapi_root[pyloop]

2 ms

test_resolve_dynamic_resource_url_with_many_dynamic_routes[pyloop]

tests/test_benchmarks_web_urldispatcher.py::test_resolve_dynamic_resource_url_with_many_dynamic_routes[pyloop]

3.5 ms

test_resolve_single_fixed_url_with_many_routes[pyloop]

tests/test_benchmarks_web_urldispatcher.py::test_resolve_single_fixed_url_with_many_routes[pyloop]

891.8 µs

889.4 µs

test_resolve_multiple_level_fixed_url_with_many_routes[pyloop]

tests/test_benchmarks_web_urldispatcher.py::test_resolve_multiple_level_fixed_url_with_many_routes[pyloop]

8.5 ms

test_create_client_request_with_headers[pyloop]

tests/test_benchmarks_client_request.py::test_create_client_request_with_headers[pyloop]

70.6 µs

70.5 µs

test_web_response_with_text_body

tests/test_benchmarks_web_response.py::test_web_response_with_text_body

654.3 µs

653.9 µs

test_resolve_root_route_with_many_fixed_routes[pyloop]

tests/test_benchmarks_web_urldispatcher.py::test_resolve_root_route_with_many_fixed_routes[pyloop]

881.2 µs

880.8 µs

test_resolve_multiple_fixed_url_with_many_routes[pyloop]

tests/test_benchmarks_web_urldispatcher.py::test_resolve_multiple_fixed_url_with_many_routes[pyloop]

2.2 ms

test_get_request_with_251308_compressed_chunked_payload[pyloop]

tests/test_benchmarks_client.py::test_get_request_with_251308_compressed_chunked_payload[pyloop]

426 ms

425.9 ms

test_send_one_hundred_websocket_compressed_messages[pyloop]

tests/test_benchmarks_http_websocket.py::test_send_one_hundred_websocket_compressed_messages[pyloop]

3 ms

test_simple_web_stream_response

tests/test_benchmarks_web_response.py::test_simple_web_stream_response

240 µs

240.1 µs

test_simple_web_response

tests/test_benchmarks_web_response.py::test_simple_web_response

443.2 µs

443.5 µs

test_resolve_prefix_resources_many_prefix_many_plain[pyloop]

tests/test_benchmarks_web_urldispatcher.py::test_resolve_prefix_resources_many_prefix_many_plain[pyloop]

4.9 ms

test_send_client_request_one_hundred[pyloop]

tests/test_benchmarks_client_request.py::test_send_client_request_one_hundred[pyloop]

2.4 ms

test_send_one_hundred_large_websocket_text_messages[pyloop]

tests/test_benchmarks_http_websocket.py::test_send_one_hundred_large_websocket_text_messages[pyloop]

533.1 µs

533.8 µs

test_resolve_dynamic_resource_url_with_many_static_routes[pyloop]

tests/test_benchmarks_web_urldispatcher.py::test_resolve_dynamic_resource_url_with_many_static_routes[pyloop]

3.6 ms

test_simple_web_file_response_not_modified[pyloop]

tests/test_benchmarks_web_fileresponse.py::test_simple_web_file_response_not_modified[pyloop]

55.5 ms

test_send_one_hundred_websocket_text_messages[pyloop]

tests/test_benchmarks_http_websocket.py::test_send_one_hundred_websocket_text_messages[pyloop]

558.9 µs

560 µs

test_simple_web_file_sendfile_fallback_response[pyloop]

tests/test_benchmarks_web_fileresponse.py::test_simple_web_file_sendfile_fallback_response[pyloop]

84.6 ms

84.8 ms

test_one_thousand_large_round_trip_websocket_text_messages[pyloop]

tests/test_benchmarks_client_ws.py::test_one_thousand_large_round_trip_websocket_text_messages[pyloop]

23.2 ms

23.3 ms

test_resolve_gitapi_subapps[pyloop]

tests/test_benchmarks_web_urldispatcher.py::test_resolve_gitapi_subapps[pyloop]

295.6 ms

296.3 ms

test_resolve_root_route[pyloop]

tests/test_benchmarks_web_urldispatcher.py::test_resolve_root_route[pyloop]

884.7 µs

886.9 µs

test_resolve_dynamic_resource_url_with_many_dynamic_routes_with_common_prefix[pyloop]

tests/test_benchmarks_web_urldispatcher.py::test_resolve_dynamic_resource_url_with_many_dynamic_routes_with_common_prefix[pyloop]

246 ms

246.8 ms

test_serialize_headers

tests/test_benchmarks_http_writer.py::test_serialize_headers

1 ms

test_one_hundred_get_requests_with_512kib_content_length_payload[pyloop]

tests/test_benchmarks_client.py::test_one_hundred_get_requests_with_512kib_content_length_payload[pyloop]

164.7 ms

165.4 ms

test_send_one_hundred_websocket_text_messages_with_mask[pyloop]

tests/test_benchmarks_http_websocket.py::test_send_one_hundred_websocket_text_messages_with_mask[pyloop]

795.3 µs

798.7 µs

test_one_hundred_simple_get_requests_multiple_methods_route[pyloop]

tests/test_benchmarks_client.py::test_one_hundred_simple_get_requests_multiple_methods_route[pyloop]

-1%

32 ms

32.2 ms

test_one_hundred_simple_get_requests[pyloop]

tests/test_benchmarks_client.py::test_one_hundred_simple_get_requests[pyloop]

-1%

32.2 ms

32.5 ms

test_ten_web_middlewares[pyloop]

tests/test_benchmarks_web_middleware.py::test_ten_web_middlewares[pyloop]

-1%

35 ms

35.3 ms

test_web_response_with_headers

tests/test_benchmarks_web_response.py::test_web_response_with_headers

-1%

880.9 µs

889.3 µs

test_create_client_request_with_cookies[pyloop]

tests/test_benchmarks_client_request.py::test_create_client_request_with_cookies[pyloop]

-1%

125.1 µs

126.3 µs

test_one_hundred_get_requests_with_30000_content_length_payload[pyloop]

tests/test_benchmarks_client.py::test_one_hundred_get_requests_with_30000_content_length_payload[pyloop]

-1%

39.4 ms

39.8 ms

test_one_hundred_simple_post_requests[pyloop]

tests/test_benchmarks_client.py::test_one_hundred_simple_post_requests[pyloop]

-1%

36.7 ms

37.1 ms

test_one_hundred_get_requests_with_1024_chunked_payload[pyloop]

tests/test_benchmarks_client.py::test_one_hundred_get_requests_with_1024_chunked_payload[pyloop]

-1%

35.5 ms

35.9 ms

test_one_hundred_get_requests_with_1024_content_length_payload[pyloop]

tests/test_benchmarks_client.py::test_one_hundred_get_requests_with_1024_content_length_payload[pyloop]

-1%

35.2 ms

35.5 ms

test_one_hundred_get_requests_with_30000_chunked_payload[pyloop]

tests/test_benchmarks_client.py::test_one_hundred_get_requests_with_30000_chunked_payload[pyloop]

-1%

40.6 ms

41 ms

test_resolve_gitapi[pyloop]

tests/test_benchmarks_web_urldispatcher.py::test_resolve_gitapi[pyloop]

-1%

293.5 ms

297.3 ms

test_simple_web_file_response[pyloop]

tests/test_benchmarks_web_fileresponse.py::test_simple_web_file_response[pyloop]

-1%

78.1 ms

79.1 ms

test_one_hundred_json_post_requests[pyloop]

tests/test_benchmarks_client.py::test_one_hundred_json_post_requests[pyloop]

-1%

38.5 ms

39.1 ms

test_one_thousand_round_trip_websocket_binary_messages[pyloop]

tests/test_benchmarks_client_ws.py::test_one_thousand_round_trip_websocket_binary_messages[pyloop]

-4%

17 ms

17.7 ms

test_one_thousand_round_trip_websocket_text_messages[pyloop]

tests/test_benchmarks_client_ws.py::test_one_thousand_round_trip_websocket_text_messages[pyloop]

-4%

17.5 ms

18.2 ms

Ignored

test_one_hundred_get_requests_iter_chunks_on_512kib_chunked_payload[pyloop]Ignored

tests/test_benchmarks_client.py::test_one_hundred_get_requests_iter_chunks_on_512kib_chunked_payload[pyloop]

-19%

96.9 ms

120.1 ms

test_one_hundred_get_requests_with_512kib_chunked_payload[pyloop]Ignored

tests/test_benchmarks_client.py::test_one_hundred_get_requests_with_512kib_chunked_payload[pyloop]

+13%

188.6 ms

167 ms

Commits

Click on a commit to change the comparison range

Base

3.12

5e20fe1

+17%

Reduce WebSocket buffer slicing overhead (#10601)  ## What do these changes do? Use a `const unsigned char *` for the buffer (Cython will automatically extract is using `__Pyx_PyBytes_AsUString`) as its a lot faster than copying around `PyBytes` objects. We do need to be careful that all slices are bounded and we bound check everything to make sure we do not do an out of bounds read since Cython does not bounds check C strings. I checked that all accesses to `buf_cstr` are proceeded by a bounds check but it would be good to get another set of eyes on that to verify in the `self._state == READ_PAYLOAD` block that we will never try to read out of bounds. <img width="376" alt="Screenshot 2025-03-19 at 10 21 54 AM" src="https://github.com/user-attachments/assets/a340ffa2-f09b-4aff-a4f7-c487dae186c8" /> ## Are there changes in behavior for the user? performance improvement ## Is it a substantial burden for the maintainers to support this? no There is a small risk that someone could remove a bounds check in the future and create a memory safety issue, however in this case its likely we would already be trying to read data that wasn't there if we are missing the bounds checking so the pure python version would throw if we are testing properly. --------- Co-authored-by: Sam Bull <git@sambull.org> (cherry picked from commit f7cac7e63f18691e4261af353e84f9073b16624a)

105229c

15 days ago

by bdraco

Home Terms Privacy Docs