WIP http_parser: don't reuse HTTPParser - #24330
Closed
mmarchini wants to merge 2 commits into
Closed
Conversation
Copy link
Copy Markdown
Contributor
Sorry, something went wrong.
Copy link
Copy Markdown
Collaborator
|
@mmarchini build started: https://ci.nodejs.org/blue/organizations/jenkins/node-test-pull-request-lite-pipeline/detail/node-test-pull-request-lite-pipeline/1589/pipeline |
Sorry, something went wrong.
As discussed in nodejs/diagnostics#248 and nodejs#21313, reusing HTTPParser resource is a blocker for landing `require('async_hooks').currentAsyncResource()`, since reusing an async resource interanlly would make it infeasible to use them reliably in WeakMaps. Two suggestions came up: have a wrapper around the HTTPParser which we would use as the async resource, or stop reusing HTTPParser in our HTTP server/client code. This commit implements the latter, since we have a better GC now and reusing HTTPParser might make sense anymore. This also will avoid one extra JS->C++ call in some cases, which should improve performance for these cases.
mmarchini
force-pushed
the
dont-reuse-httpparser
branch
from
c642522 to
73e2db6
Compare
November 13, 2018 03:08
Copy link
Copy Markdown
Contributor
Author
|
CI: https://ci.nodejs.org/job/node-test-pull-request/18566/ |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
Author
|
Regressions on http/client-request-body seems to be the most significant: confidence improvement accuracy (*) (**) (***) http/client-request-body.js method='end' len=1024 type='asc' dur=5 * -8.12 % ±6.52% ±8.67% ±11.29% http/client-request-body.js method='end' len=1024 type='buf' dur=5 * -6.33 % ±5.59% ±7.44% ±9.69% http/client-request-body.js method='end' len=1024 type='utf' dur=5 * -7.74 % ±6.60% ±8.79% ±11.44% http/client-request-body.js method='end' len=256 type='asc' dur=5 * -8.53 % ±6.42% ±8.55% ±11.13% http/client-request-body.js method='end' len=256 type='buf' dur=5 -4.67 % ±5.86% ±7.80% ±10.15% http/client-request-body.js method='end' len=256 type='utf' dur=5 * -6.33 % ±5.36% ±7.13% ±9.28% http/client-request-body.js method='end' len=32 type='asc' dur=5 *** -12.60 % ±5.93% ±7.89% ±10.27% http/client-request-body.js method='end' len=32 type='buf' dur=5 0.08 % ±6.70% ±8.91% ±11.60% http/client-request-body.js method='end' len=32 type='utf' dur=5 * -7.62 % ±6.36% ±8.47% ±11.05% http/client-request-body.js method='write' len=1024 type='asc' dur=5 -3.80 % ±6.07% ±8.08% ±10.53% http/client-request-body.js method='write' len=1024 type='buf' dur=5 *** -14.17 % ±5.60% ±7.46% ±9.70% http/client-request-body.js method='write' len=1024 type='utf' dur=5 ** -8.91 % ±6.61% ±8.80% ±11.46% http/client-request-body.js method='write' len=256 type='asc' dur=5 ** -10.05 % ±7.20% ±9.58% ±12.48% http/client-request-body.js method='write' len=256 type='buf' dur=5 * -7.62 % ±6.95% ±9.25% ±12.05% http/client-request-body.js method='write' len=256 type='utf' dur=5 -6.19 % ±6.24% ±8.30% ±10.81% http/client-request-body.js method='write' len=32 type='asc' dur=5 -5.66 % ±6.47% ±8.60% ±11.20% http/client-request-body.js method='write' len=32 type='buf' dur=5 -1.43 % ±5.45% ±7.27% ±9.50% http/client-request-body.js method='write' len=32 type='utf' dur=5 *** -14.11 % ±5.77% ±7.68% ±9.99% Be aware that when doing many comparisons the risk of a false-positive result increases. In this case there are 18 comparisons, you can thus expect the following amount of false-positive results: 0.90 false positives, when considering a 5% risk acceptance (*, **, ***), 0.18 false positives, when considering a 1% risk acceptance (**, ***), 0.02 false positives, when considering a 0.1% risk acceptance (***)
I'll see if there's anything else we can optimize to compensate for these regressions. Also, I'm not sure if we can break the HTTPParser API. It's not documented, and this PR removes the .reinitialize() method. @mcollina @Flarna @AndreasMadsen PTAL a let me know what you think. Full HTTP Benchmark Resultsconfidence improvement accuracy (*) (**) (***) http/bench-parser.js n=100000 len=16 -2.87 % ±4.72% ±6.28% ±8.17% http/bench-parser.js n=100000 len=32 * -3.76 % ±3.61% ±4.82% ±6.29% http/bench-parser.js n=100000 len=4 * -2.98 % ±2.37% ±3.15% ±4.10% http/bench-parser.js n=100000 len=8 -3.20 % ±4.06% ±5.40% ±7.04% http/check_invalid_header_char.js n=1000000 input='' -1.95 % ±4.82% ±6.41% ±8.35% http/check_invalid_header_char.js n=1000000 input='\177' -1.12 % ±3.00% ±3.99% ±5.20% http/check_invalid_header_char.js n=1000000 input='1' 0.37 % ±4.01% ±5.34% ±6.97% http/check_invalid_header_char.js n=1000000 input='20091' -2.47 % ±4.19% ±5.61% ±7.37% http/check_invalid_header_char.js n=1000000 input='close' -0.95 % ±4.13% ±5.51% ±7.19% http/check_invalid_header_char.js n=1000000 input='en-US' -1.37 % ±4.28% ±5.70% ±7.43% http/check_invalid_header_char.js n=1000000 input='foo\\nbar' -1.36 % ±4.04% ±5.38% ±7.01% http/check_invalid_header_char.js n=1000000 input='group_acmeair' 2.34 % ±3.57% ±4.75% ±6.20% http/check_invalid_header_char.js n=1000000 input='gzip' -0.10 % ±4.40% ±5.85% ±7.63% http/check_invalid_header_char.js n=1000000 input='keep-alive' 1.84 % ±4.15% ±5.52% ±7.18% http/check_invalid_header_char.js n=1000000 input='LONG_AND_INVALID' 0.47 % ±1.99% ±2.65% ±3.44% http/check_invalid_header_char.js n=1000000 input='private' 2.75 % ±3.79% ±5.07% ±6.64% http/check_invalid_header_char.js n=1000000 input='SAMEORIGIN' -3.61 % ±4.07% ±5.43% ±7.09% http/check_invalid_header_char.js n=1000000 input='Sat, 07 May 2016 16:54:48 GMT' 0.46 % ±2.59% ±3.45% ±4.49% http/check_invalid_header_char.js n=1000000 input='text/html; charset=utf-8' 0.86 % ±3.87% ±5.15% ±6.71% http/check_invalid_header_char.js n=1000000 input='text/plain' -2.86 % ±3.89% ±5.18% ±6.75% http/check_invalid_header_char.js n=1000000 input='\\t\\t\\t\\t\\t\\t\\t\\t\\t\\tFoo bar baz' -2.64 % ±3.54% ±4.71% ±6.13% http/check_invalid_header_char.js n=1000000 input='中文呢' 2.97 % ±4.57% ±6.10% ±7.98% http/check_is_http_token.js n=1000000 key=':' -2.30 % ±4.08% ±5.46% ±7.17% http/check_is_http_token.js n=1000000 key='((((())))' * 5.75 % ±5.45% ±7.29% ±9.57% http/check_is_http_token.js n=1000000 key='@@' 3.72 % ±4.72% ±6.28% ±8.18% http/check_is_http_token.js n=1000000 key='Accept-Ranges' 0.29 % ±3.70% ±4.93% ±6.42% http/check_is_http_token.js n=1000000 key=':alternate-protocol' -0.01 % ±5.30% ±7.05% ±9.18% http/check_is_http_token.js n=1000000 key='alternate-protocol:' -0.11 % ±2.14% ±2.85% ±3.72% http/check_is_http_token.js n=1000000 key='alternate-protocol' -3.76 % ±4.29% ±5.71% ±7.45% http/check_is_http_token.js n=1000000 key='alt-svc' 2.00 % ±3.56% ±4.75% ±6.22% http/check_is_http_token.js n=1000000 key='Cache-Control' 2.31 % ±3.31% ±4.42% ±5.77% http/check_is_http_token.js n=1000000 key='Connection' 2.00 % ±4.50% ±6.00% ±7.81% http/check_is_http_token.js n=1000000 key='Content-Encoding' -0.28 % ±3.47% ±4.61% ±6.01% http/check_is_http_token.js n=1000000 key='content-length' -1.58 % ±3.88% ±5.18% ±6.76% http/check_is_http_token.js n=1000000 key='Content-Location' 0.13 % ±2.85% ±3.80% ±4.96% http/check_is_http_token.js n=1000000 key='content-type' -2.50 % ±3.16% ±4.21% ±5.49% http/check_is_http_token.js n=1000000 key='Content-Type' 1.39 % ±3.68% ±4.90% ±6.38% http/check_is_http_token.js n=1000000 key='date' 0.34 % ±4.66% ±6.21% ±8.09% http/check_is_http_token.js n=1000000 key='ETag' -0.92 % ±2.93% ±3.90% ±5.08% http/check_is_http_token.js n=1000000 key='Expires' 1.48 % ±2.74% ±3.65% ±4.75% http/check_is_http_token.js n=1000000 key='Keep-Alive' * -4.26 % ±4.07% ±5.44% ±7.12% http/check_is_http_token.js n=1000000 key='Last-Modified' -0.46 % ±4.21% ±5.60% ±7.30% http/check_is_http_token.js n=1000000 key='location' -3.81 % ±4.37% ±5.82% ±7.60% http/check_is_http_token.js n=1000000 key='server' -1.71 % ±3.16% ±4.20% ±5.47% http/check_is_http_token.js n=1000000 key='Server' 1.06 % ±4.90% ±6.52% ±8.49% http/check_is_http_token.js n=1000000 key='status' -0.61 % ±4.00% ±5.33% ±6.95% http/check_is_http_token.js n=1000000 key='TCN' 2.94 % ±3.68% ±4.92% ±6.45% http/check_is_http_token.js n=1000000 key='Transfer-Encoding' 0.45 % ±3.12% ±4.15% ±5.42% http/check_is_http_token.js n=1000000 key='Vary' -2.76 % ±3.58% ±4.79% ±6.28% http/check_is_http_token.js n=1000000 key='version' -4.42 % ±4.75% ±6.35% ±8.32% http/check_is_http_token.js n=1000000 key='x-frame-options' -2.81 % ±3.87% ±5.15% ±6.70% http/check_is_http_token.js n=1000000 key='x-xss-protection' -1.62 % ±3.63% ±4.83% ±6.30% http/check_is_http_token.js n=1000000 key='中文呢' -0.27 % ±5.57% ±7.41% ±9.64% http/chunked.js c=100 len=1 n=16 benchmarker='wrk' -0.15 % ±0.49% ±0.66% ±0.85% http/chunked.js c=100 len=1 n=1 benchmarker='wrk' -0.04 % ±0.30% ±0.40% ±0.52% http/chunked.js c=100 len=1 n=4 benchmarker='wrk' -0.10 % ±0.31% ±0.41% ±0.53% http/chunked.js c=100 len=1 n=8 benchmarker='wrk' -0.26 % ±0.35% ±0.47% ±0.61% http/chunked.js c=100 len=256 n=16 benchmarker='wrk' 0.03 % ±0.67% ±0.90% ±1.17% http/chunked.js c=100 len=256 n=1 benchmarker='wrk' -0.17 % ±0.30% ±0.41% ±0.53% http/chunked.js c=100 len=256 n=4 benchmarker='wrk' 0.10 % ±0.33% ±0.44% ±0.57% http/chunked.js c=100 len=256 n=8 benchmarker='wrk' -0.02 % ±0.28% ±0.37% ±0.49% http/chunked.js c=100 len=64 n=16 benchmarker='wrk' -0.09 % ±0.49% ±0.65% ±0.85% http/chunked.js c=100 len=64 n=1 benchmarker='wrk' 0.05 % ±0.29% ±0.39% ±0.51% http/chunked.js c=100 len=64 n=4 benchmarker='wrk' -0.04 % ±0.26% ±0.34% ±0.44% http/chunked.js c=100 len=64 n=8 benchmarker='wrk' -0.09 % ±0.27% ±0.35% ±0.46% http/client-request-body.js method='end' len=1024 type='asc' dur=5 * -8.12 % ±6.52% ±8.67% ±11.29% http/client-request-body.js method='end' len=1024 type='buf' dur=5 * -6.33 % ±5.59% ±7.44% ±9.69% http/client-request-body.js method='end' len=1024 type='utf' dur=5 * -7.74 % ±6.60% ±8.79% ±11.44% http/client-request-body.js method='end' len=256 type='asc' dur=5 * -8.53 % ±6.42% ±8.55% ±11.13% http/client-request-body.js method='end' len=256 type='buf' dur=5 -4.67 % ±5.86% ±7.80% ±10.15% http/client-request-body.js method='end' len=256 type='utf' dur=5 * -6.33 % ±5.36% ±7.13% ±9.28% http/client-request-body.js method='end' len=32 type='asc' dur=5 *** -12.60 % ±5.93% ±7.89% ±10.27% http/client-request-body.js method='end' len=32 type='buf' dur=5 0.08 % ±6.70% ±8.91% ±11.60% http/client-request-body.js method='end' len=32 type='utf' dur=5 * -7.62 % ±6.36% ±8.47% ±11.05% http/client-request-body.js method='write' len=1024 type='asc' dur=5 -3.80 % ±6.07% ±8.08% ±10.53% http/client-request-body.js method='write' len=1024 type='buf' dur=5 *** -14.17 % ±5.60% ±7.46% ±9.70% http/client-request-body.js method='write' len=1024 type='utf' dur=5 ** -8.91 % ±6.61% ±8.80% ±11.46% http/client-request-body.js method='write' len=256 type='asc' dur=5 ** -10.05 % ±7.20% ±9.58% ±12.48% http/client-request-body.js method='write' len=256 type='buf' dur=5 * -7.62 % ±6.95% ±9.25% ±12.05% http/client-request-body.js method='write' len=256 type='utf' dur=5 -6.19 % ±6.24% ±8.30% ±10.81% http/client-request-body.js method='write' len=32 type='asc' dur=5 -5.66 % ±6.47% ±8.60% ±11.20% http/client-request-body.js method='write' len=32 type='buf' dur=5 -1.43 % ±5.45% ±7.27% ±9.50% http/client-request-body.js method='write' len=32 type='utf' dur=5 *** -14.11 % ±5.77% ±7.68% ±9.99% http/cluster.js c=500 len=102400 type='buffer' benchmarker='wrk' 1.55 % ±3.87% ±5.16% ±6.71% http/cluster.js c=500 len=102400 type='bytes' benchmarker='wrk' -0.05 % ±4.17% ±5.55% ±7.22% http/cluster.js c=500 len=1024 type='buffer' benchmarker='wrk' -1.69 % ±3.57% ±4.75% ±6.19% http/cluster.js c=500 len=1024 type='bytes' benchmarker='wrk' -2.44 % ±3.45% ±4.59% ±5.97% http/cluster.js c=500 len=4 type='buffer' benchmarker='wrk' 0.37 % ±3.51% ±4.67% ±6.08% http/cluster.js c=500 len=4 type='bytes' benchmarker='wrk' 0.46 % ±4.00% ±5.32% ±6.92% http/cluster.js c=50 len=102400 type='buffer' benchmarker='wrk' * -3.43 % ±3.40% ±4.53% ±5.90% http/cluster.js c=50 len=102400 type='bytes' benchmarker='wrk' -2.09 % ±3.42% ±4.56% ±5.94% http/cluster.js c=50 len=1024 type='buffer' benchmarker='wrk' 0.26 % ±2.49% ±3.32% ±4.32% http/cluster.js c=50 len=1024 type='bytes' benchmarker='wrk' 1.52 % ±2.58% ±3.43% ±4.46% http/cluster.js c=50 len=4 type='buffer' benchmarker='wrk' -1.20 % ±2.62% ±3.48% ±4.54% http/cluster.js c=50 len=4 type='bytes' benchmarker='wrk' -0.57 % ±2.60% ±3.46% ±4.51% http/create-clientrequest.js n=1000000 len=1 -2.40 % ±3.74% ±4.97% ±6.48% http/create-clientrequest.js n=1000000 len=128 -2.20 % ±4.04% ±5.38% ±7.00% http/create-clientrequest.js n=1000000 len=16 -1.70 % ±3.81% ±5.07% ±6.60% http/create-clientrequest.js n=1000000 len=32 ** -5.55 % ±3.99% ±5.32% ±6.96% http/create-clientrequest.js n=1000000 len=64 -2.73 % ±3.83% ±5.10% ±6.63% http/create-clientrequest.js n=1000000 len=8 -1.84 % ±3.59% ±4.78% ±6.23% http/end-vs-write-end.js method='end' c=100 len=1048576 type='asc' benchmarker='wrk' 2.03 % ±5.86% ±7.80% ±10.16% http/end-vs-write-end.js method='end' c=100 len=1048576 type='buf' benchmarker='wrk' -0.29 % ±6.21% ±8.26% ±10.75% http/end-vs-write-end.js method='end' c=100 len=1048576 type='utf' benchmarker='wrk' -1.81 % ±5.65% ±7.52% ±9.79% http/end-vs-write-end.js method='end' c=100 len=131072 type='asc' benchmarker='wrk' 0.01 % ±3.75% ±4.99% ±6.50% http/end-vs-write-end.js method='end' c=100 len=131072 type='buf' benchmarker='wrk' 0.50 % ±4.05% ±5.39% ±7.02% http/end-vs-write-end.js method='end' c=100 len=131072 type='utf' benchmarker='wrk' 0.35 % ±3.92% ±5.22% ±6.80% http/end-vs-write-end.js method='end' c=100 len=262144 type='asc' benchmarker='wrk' 3.11 % ±3.91% ±5.23% ±6.84% http/end-vs-write-end.js method='end' c=100 len=262144 type='buf' benchmarker='wrk' 1.14 % ±4.90% ±6.52% ±8.49% http/end-vs-write-end.js method='end' c=100 len=262144 type='utf' benchmarker='wrk' -1.23 % ±4.11% ±5.48% ±7.13% http/end-vs-write-end.js method='end' c=100 len=65536 type='asc' benchmarker='wrk' * 2.51 % ±2.48% ±3.30% ±4.30% http/end-vs-write-end.js method='end' c=100 len=65536 type='buf' benchmarker='wrk' 4.03 % ±5.86% ±7.80% ±10.16% http/end-vs-write-end.js method='end' c=100 len=65536 type='utf' benchmarker='wrk' -0.43 % ±2.38% ±3.17% ±4.12% http/end-vs-write-end.js method='write' c=100 len=1048576 type='asc' benchmarker='wrk' -3.62 % ±4.49% ±5.97% ±7.77% http/end-vs-write-end.js method='write' c=100 len=1048576 type='buf' benchmarker='wrk' 0.63 % ±7.58% ±10.09% ±13.14% http/end-vs-write-end.js method='write' c=100 len=1048576 type='utf' benchmarker='wrk' -2.76 % ±4.68% ±6.23% ±8.11% http/end-vs-write-end.js method='write' c=100 len=131072 type='asc' benchmarker='wrk' -1.66 % ±2.00% ±2.66% ±3.46% http/end-vs-write-end.js method='write' c=100 len=131072 type='buf' benchmarker='wrk' -1.95 % ±3.71% ±4.94% ±6.44% http/end-vs-write-end.js method='write' c=100 len=131072 type='utf' benchmarker='wrk' * -3.29 % ±3.15% ±4.20% ±5.48% http/end-vs-write-end.js method='write' c=100 len=262144 type='asc' benchmarker='wrk' -0.31 % ±3.29% ±4.38% ±5.70% http/end-vs-write-end.js method='write' c=100 len=262144 type='buf' benchmarker='wrk' -2.42 % ±4.17% ±5.55% ±7.22% http/end-vs-write-end.js method='write' c=100 len=262144 type='utf' benchmarker='wrk' -1.59 % ±2.62% ±3.49% ±4.55% http/end-vs-write-end.js method='write' c=100 len=65536 type='asc' benchmarker='wrk' -1.98 % ±2.92% ±3.88% ±5.05% http/end-vs-write-end.js method='write' c=100 len=65536 type='buf' benchmarker='wrk' -1.14 % ±5.20% ±6.92% ±9.01% http/end-vs-write-end.js method='write' c=100 len=65536 type='utf' benchmarker='wrk' -1.34 % ±3.09% ±4.11% ±5.36% http/headers.js n=1000 duplicates=100 benchmarker='wrk' -0.21 % ±2.05% ±2.72% ±3.55% http/headers.js n=1000 duplicates=1 benchmarker='wrk' -0.13 % ±2.38% ±3.17% ±4.13% http/headers.js n=10 duplicates=100 benchmarker='wrk' -0.66 % ±4.58% ±6.10% ±7.94% http/headers.js n=10 duplicates=1 benchmarker='wrk' -0.48 % ±3.99% ±5.31% ±6.94% http/http_server_for_chunky_client.js type='send' n=2000 len=1 -1.08 % ±9.30% ±12.38% ±16.11% http/http_server_for_chunky_client.js type='send' n=2000 len=128 -2.23 % ±9.81% ±13.05% ±16.98% http/http_server_for_chunky_client.js type='send' n=2000 len=16 3.12 % ±10.26% ±13.65% ±17.78% http/http_server_for_chunky_client.js type='send' n=2000 len=32 1.30 % ±11.09% ±14.75% ±19.20% http/http_server_for_chunky_client.js type='send' n=2000 len=4 4.60 % ±11.80% ±15.71% ±20.46% http/http_server_for_chunky_client.js type='send' n=2000 len=64 7.43 % ±9.88% ±13.16% ±17.14% http/http_server_for_chunky_client.js type='send' n=2000 len=8 -3.12 % ±10.83% ±14.41% ±18.77% http/http_server_for_chunky_client.js type='send' n=500 len=1 0.77 % ±4.93% ±6.57% ±8.55% http/http_server_for_chunky_client.js type='send' n=500 len=128 * -6.70 % ±5.66% ±7.56% ±9.88% http/http_server_for_chunky_client.js type='send' n=500 len=16 1.04 % ±6.39% ±8.50% ±11.07% http/http_server_for_chunky_client.js type='send' n=500 len=32 -0.63 % ±6.21% ±8.27% ±10.78% http/http_server_for_chunky_client.js type='send' n=500 len=4 -0.61 % ±6.00% ±7.99% ±10.41% http/http_server_for_chunky_client.js type='send' n=500 len=64 -3.43 % ±5.05% ±6.72% ±8.76% http/http_server_for_chunky_client.js type='send' n=500 len=8 2.63 % ±6.24% ±8.30% ±10.80% http/http_server_for_chunky_client.js type='send' n=50 len=1 -0.35 % ±5.31% ±7.07% ±9.21% http/http_server_for_chunky_client.js type='send' n=50 len=128 0.96 % ±5.03% ±6.72% ±8.78% http/http_server_for_chunky_client.js type='send' n=50 len=16 1.77 % ±6.02% ±8.01% ±10.43% http/http_server_for_chunky_client.js type='send' n=50 len=32 -2.29 % ±5.69% ±7.58% ±9.87% http/http_server_for_chunky_client.js type='send' n=50 len=4 -0.49 % ±5.50% ±7.32% ±9.53% http/http_server_for_chunky_client.js type='send' n=50 len=64 4.48 % ±5.31% ±7.08% ±9.22% http/http_server_for_chunky_client.js type='send' n=50 len=8 -1.81 % ±5.92% ±7.88% ±10.27% http/http_server_for_chunky_client.js type='send' n=5 len=1 -2.76 % ±5.12% ±6.82% ±8.88% http/http_server_for_chunky_client.js type='send' n=5 len=128 -1.40 % ±8.42% ±11.22% ±14.64% http/http_server_for_chunky_client.js type='send' n=5 len=16 -3.46 % ±9.00% ±11.98% ±15.62% http/http_server_for_chunky_client.js type='send' n=5 len=32 6.27 % ±8.85% ±11.79% ±15.36% http/http_server_for_chunky_client.js type='send' n=5 len=4 -2.48 % ±7.84% ±10.43% ±13.58% http/http_server_for_chunky_client.js type='send' n=5 len=64 0.57 % ±7.18% ±9.55% ±12.44% http/http_server_for_chunky_client.js type='send' n=5 len=8 1.25 % ±8.12% ±10.82% ±14.09% http/set_header.js n=1000000 value='Connection' -0.18 % ±2.35% ±3.13% ±4.08% http/set_header.js n=1000000 value='Content-Length' ** -3.13 % ±1.97% ±2.62% ±3.42% http/set_header.js n=1000000 value='Content-Type' ** 4.40 % ±3.12% ±4.16% ±5.43% http/set_header.js n=1000000 value='Set-Cookie' -0.47 % ±2.23% ±2.97% ±3.87% http/set_header.js n=1000000 value='Transfer-Encoding' -0.99 % ±2.50% ±3.34% ±4.38% http/set_header.js n=1000000 value='Vary' 0.63 % ±3.02% ±4.02% ±5.25% http/set_header.js n=1000000 value='X-Powered-By' 1.11 % ±2.15% ±2.87% ±3.74% http/set-header.js res='normal' benchmarker='wrk' ** -5.08 % ±3.81% ±5.07% ±6.59% http/set-header.js res='setHeader' benchmarker='wrk' -1.82 % ±4.05% ±5.39% ±7.01% http/set-header.js res='setHeaderWH' benchmarker='wrk' -2.22 % ±4.15% ±5.53% ±7.21% http/simple.js chunkedEnc=0 c=500 chunks=1 len=102400 type='buffer' benchmarker='wrk' -0.57 % ±5.01% ±6.67% ±8.68% http/simple.js chunkedEnc=0 c=500 chunks=1 len=102400 type='bytes' benchmarker='wrk' -0.63 % ±2.70% ±3.60% ±4.68% http/simple.js chunkedEnc=0 c=500 chunks=1 len=1024 type='buffer' benchmarker='wrk' -2.54 % ±4.77% ±6.35% ±8.27% http/simple.js chunkedEnc=0 c=500 chunks=1 len=1024 type='bytes' benchmarker='wrk' 0.11 % ±4.59% ±6.10% ±7.95% http/simple.js chunkedEnc=0 c=500 chunks=1 len=4 type='buffer' benchmarker='wrk' 3.52 % ±4.76% ±6.33% ±8.24% http/simple.js chunkedEnc=0 c=500 chunks=1 len=4 type='bytes' benchmarker='wrk' 0.23 % ±3.93% ±5.23% ±6.81% http/simple.js chunkedEnc=0 c=500 chunks=4 len=102400 type='buffer' benchmarker='wrk' -0.64 % ±5.56% ±7.40% ±9.64% http/simple.js chunkedEnc=0 c=500 chunks=4 len=102400 type='bytes' benchmarker='wrk' -2.00 % ±5.13% ±6.82% ±8.89% http/simple.js chunkedEnc=0 c=500 chunks=4 len=1024 type='buffer' benchmarker='wrk' -0.84 % ±3.74% ±4.97% ±6.47% http/simple.js chunkedEnc=0 c=500 chunks=4 len=1024 type='bytes' benchmarker='wrk' -1.44 % ±4.31% ±5.74% ±7.48% http/simple.js chunkedEnc=0 c=500 chunks=4 len=4 type='buffer' benchmarker='wrk' 0.64 % ±5.27% ±7.01% ±9.12% http/simple.js chunkedEnc=0 c=500 chunks=4 len=4 type='bytes' benchmarker='wrk' 0.77 % ±4.63% ±6.16% ±8.01% http/simple.js chunkedEnc=0 c=50 chunks=1 len=102400 type='buffer' benchmarker='wrk' -1.89 % ±4.29% ±5.71% ±7.44% http/simple.js chunkedEnc=0 c=50 chunks=1 len=102400 type='bytes' benchmarker='wrk' 0.91 % ±2.29% ±3.05% ±4.00% http/simple.js chunkedEnc=0 c=50 chunks=1 len=1024 type='buffer' benchmarker='wrk' -1.53 % ±4.33% ±5.76% ±7.49% http/simple.js chunkedEnc=0 c=50 chunks=1 len=1024 type='bytes' benchmarker='wrk' -2.75 % ±3.84% ±5.11% ±6.66% http/simple.js chunkedEnc=0 c=50 chunks=1 len=4 type='buffer' benchmarker='wrk' 0.24 % ±4.02% ±5.35% ±6.97% http/simple.js chunkedEnc=0 c=50 chunks=1 len=4 type='bytes' benchmarker='wrk' -0.84 % ±3.46% ±4.60% ±5.99% http/simple.js chunkedEnc=0 c=50 chunks=4 len=102400 type='buffer' benchmarker='wrk' 2.48 % ±5.43% ±7.23% ±9.41% http/simple.js chunkedEnc=0 c=50 chunks=4 len=102400 type='bytes' benchmarker='wrk' 0.13 % ±3.03% ±4.03% ±5.25% http/simple.js chunkedEnc=0 c=50 chunks=4 len=1024 type='buffer' benchmarker='wrk' -1.72 % ±3.07% ±4.09% ±5.32% http/simple.js chunkedEnc=0 c=50 chunks=4 len=1024 type='bytes' benchmarker='wrk' -0.47 % ±3.95% ±5.26% ±6.85% http/simple.js chunkedEnc=0 c=50 chunks=4 len=4 type='buffer' benchmarker='wrk' 1.16 % ±4.20% ±5.59% ±7.28% http/simple.js chunkedEnc=0 c=50 chunks=4 len=4 type='bytes' benchmarker='wrk' -1.61 % ±3.09% ±4.11% ±5.35% http/simple.js chunkedEnc=1 c=500 chunks=1 len=102400 type='buffer' benchmarker='wrk' -4.29 % ±4.77% ±6.35% ±8.27% http/simple.js chunkedEnc=1 c=500 chunks=1 len=102400 type='bytes' benchmarker='wrk' -0.74 % ±3.88% ±5.17% ±6.73% http/simple.js chunkedEnc=1 c=500 chunks=1 len=1024 type='buffer' benchmarker='wrk' -0.06 % ±5.12% ±6.81% ±8.87% http/simple.js chunkedEnc=1 c=500 chunks=1 len=1024 type='bytes' benchmarker='wrk' 2.09 % ±4.51% ±6.00% ±7.81% http/simple.js chunkedEnc=1 c=500 chunks=1 len=4 type='buffer' benchmarker='wrk' -1.42 % ±4.60% ±6.12% ±7.98% http/simple.js chunkedEnc=1 c=500 chunks=1 len=4 type='bytes' benchmarker='wrk' 3.07 % ±3.90% ±5.18% ±6.75% http/simple.js chunkedEnc=1 c=500 chunks=4 len=102400 type='buffer' benchmarker='wrk' 0.36 % ±4.04% ±5.38% ±7.00% http/simple.js chunkedEnc=1 c=500 chunks=4 len=102400 type='bytes' benchmarker='wrk' -1.99 % ±3.71% ±4.94% ±6.43% http/simple.js chunkedEnc=1 c=500 chunks=4 len=1024 type='buffer' benchmarker='wrk' -1.53 % ±4.45% ±5.92% ±7.70% http/simple.js chunkedEnc=1 c=500 chunks=4 len=1024 type='bytes' benchmarker='wrk' -1.68 % ±4.93% ±6.56% ±8.54% http/simple.js chunkedEnc=1 c=500 chunks=4 len=4 type='buffer' benchmarker='wrk' -0.22 % ±5.78% ±7.69% ±10.01% http/simple.js chunkedEnc=1 c=500 chunks=4 len=4 type='bytes' benchmarker='wrk' 1.39 % ±4.41% ±5.87% ±7.64% http/simple.js chunkedEnc=1 c=50 chunks=1 len=102400 type='buffer' benchmarker='wrk' 0.53 % ±4.79% ±6.37% ±8.30% http/simple.js chunkedEnc=1 c=50 chunks=1 len=102400 type='bytes' benchmarker='wrk' -0.86 % ±2.03% ±2.70% ±3.51% http/simple.js chunkedEnc=1 c=50 chunks=1 len=1024 type='buffer' benchmarker='wrk' 2.60 % ±3.74% ±4.98% ±6.49% http/simple.js chunkedEnc=1 c=50 chunks=1 len=1024 type='bytes' benchmarker='wrk' -0.39 % ±3.99% ±5.32% ±6.93% http/simple.js chunkedEnc=1 c=50 chunks=1 len=4 type='buffer' benchmarker='wrk' -1.30 % ±3.77% ±5.02% ±6.54% http/simple.js chunkedEnc=1 c=50 chunks=1 len=4 type='bytes' benchmarker='wrk' 1.49 % ±3.60% ±4.79% ±6.23% http/simple.js chunkedEnc=1 c=50 chunks=4 len=102400 type='buffer' benchmarker='wrk' 1.13 % ±4.63% ±6.17% ±8.04% http/simple.js chunkedEnc=1 c=50 chunks=4 len=102400 type='bytes' benchmarker='wrk' -0.38 % ±2.61% ±3.47% ±4.53% http/simple.js chunkedEnc=1 c=50 chunks=4 len=1024 type='buffer' benchmarker='wrk' 1.27 % ±3.24% ±4.31% ±5.62% http/simple.js chunkedEnc=1 c=50 chunks=4 len=1024 type='bytes' benchmarker='wrk' -0.92 % ±2.75% ±3.66% ±4.76% http/simple.js chunkedEnc=1 c=50 chunks=4 len=4 type='buffer' benchmarker='wrk' -3.05 % ±3.11% ±4.13% ±5.38% http/simple.js chunkedEnc=1 c=50 chunks=4 len=4 type='bytes' benchmarker='wrk' 1.37 % ±3.83% ±5.10% ±6.64% http/upgrade.js n=1000 ** -5.39 % ±3.59% ±4.78% ±6.23% http/upgrade.js n=5 -0.94 % ±2.73% ±3.63% ±4.72% Be aware that when doing many comparisons the risk of a false-positive result increases. In this case there are 217 comparisons, you can thus expect the following amount of false-positive results: 10.85 false positives, when considering a 5% risk acceptance (*, **, ***), 2.17 false positives, when considering a 1% risk acceptance (**, ***), 0.22 false positives, when considering a 0.1% risk acceptance (***) Notifying upstream projects of job completion Finished: SUCCESS (edit by @AndreasMadsen: fixed formatting in <details>.) |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
Author
Sorry, something went wrong.
Copy link
Copy Markdown
Member
|
Yeah, http/client-request-body.js doesn't look good. There are some others that could be problematic too. Increased Benchmark CI to 100 runs. There are quite a few comparisons, so I would like a good low confidence threshold (alpha) if possible. https://ci.nodejs.org/view/Node.js%20benchmark/job/benchmark-node-micro-benchmarks/274/ It will take quite a while, so feel free to stop it if there is something else that is more important. |
Sorry, something went wrong.
Copy link
Copy Markdown
Member
|
Picking only the most significant results:
confidence improvement accuracy (*) (**) (***)
http/client-request-body.js method='end' len=1024 type='asc' dur=5 *** -7.39 % ±3.44% ±4.54% ±5.83%
http/client-request-body.js method='end' len=1024 type='utf' dur=5 *** -5.50 % ±3.18% ±4.19% ±5.38%
http/client-request-body.js method='end' len=256 type='asc' dur=5 *** -5.59 % ±3.11% ±4.10% ±5.27%
http/client-request-body.js method='end' len=32 type='buf' dur=5 *** -6.26 % ±3.18% ±4.19% ±5.38%
http/client-request-body.js method='write' len=1024 type='asc' dur=5 *** -6.94 % ±3.52% ±4.65% ±5.97%
http/client-request-body.js method='write' len=1024 type='utf' dur=5 *** -6.67 % ±3.54% ±4.67% ±5.99%
http/client-request-body.js method='write' len=256 type='asc' dur=5 *** -5.65 % ±3.17% ±4.18% ±5.37%
http/client-request-body.js method='write' len=256 type='buf' dur=5 *** -7.99 % ±2.79% ±3.67% ±4.72%
http/client-request-body.js method='write' len=256 type='utf' dur=5 *** -6.70 % ±3.34% ±4.41% ±5.66%
http/client-request-body.js method='write' len=32 type='asc' dur=5 *** -6.13 % ±3.56% ±4.69% ±6.03%
http/client-request-body.js method='write' len=32 type='buf' dur=5 *** -6.48 % ±3.47% ±4.57% ±5.87%
http/client-request-body.js method='write' len=32 type='utf' dur=5 *** -6.78 % ±3.31% ±4.36% ±5.60%
http/set_header.js n=1000000 value='Connection' *** 2.81 % ±1.31% ±1.72% ±2.21%
http/set_header.js n=1000000 value='Content-Length' *** 2.77 % ±0.87% ±1.15% ±1.48%
http/set_header.js n=1000000 value='Content-Type' *** 2.70 % ±1.30% ±1.71% ±2.20%
http/set_header.js n=1000000 value='Set-Cookie' *** 2.90 % ±1.14% ±1.51% ±1.94%
http/set_header.js n=1000000 value='Transfer-Encoding' *** 2.34 % ±1.23% ±1.62% ±2.08%
http/set_header.js n=1000000 value='X-Powered-By' *** 3.64 % ±1.30% ±1.71% ±2.20%
http/upgrade.js n=5 *** -4.00 % ±1.54% ±2.03% ±2.61%
Tbh, I am not a fan of doing this – I was actually thinking about introducing similar mechanisms for other parts of Node, because there is noticable overhead associated with creating C++ objects. If we want less of a performance impact, we probably need to address that in V8 in some way before going this route… |
Sorry, something went wrong.
Copy link
Copy Markdown
Member
|
@mmarchini As you pinged me to comment on this. I'm sorry but I feel not qualified to rate the results of these benchmarks. Clearly more overhead is bad but I can't tell the impact of these micro benchmarks to real world apps. I'm sure there are others in node community with more experience in this then me. @addaleax Just for my personal interest. Do you have some hint or link to a more in depth info what exactly is causing the overhead with C++ objects? e.g. is it the need that GC has to call a finalizer or the crossing of JS/C++ border? I googled around but haven't found details. |
Sorry, something went wrong.
Copy link
Copy Markdown
Member
|
@Flarna It’s the crossing of the JS/C++ border – there’s a lot of overhead in both directions… I didn’t see GC pop up as a significant slowdown in benchmarks, actually. And I wouldn’t expect finalizers to come with a lot of overhead, as that doesn’t require V8 to change engine state in some way… |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
Author
The goal here is to make currentAsyncResource() return a unique resource for each async operation. Another approach @Flarna suggested is wrapping HTTPParser and using this wrapper as the resource. But this solution would have the same performance implications I think. I'm open to suggestions :)
I think it's a good performance optimization as long as it doesn't hurt Node.js debugability and observability.
That would help I think. @hashseed @bmeurer any thoughts on this? |
Sorry, something went wrong.
Copy link
Copy Markdown
Member
I mean, the IncomingMessage object is essentially that wrapper, right…? That would work for me… |
Sorry, something went wrong.
Copy link
Copy Markdown
Member
|
Sounds good; using IncomingMessage has one more advantage - it's already public so one less undocumented/internal object exposed. I'm not yet that familiar with the node source code but it seems to me that HTTPParser is "automatically" an AsyncResource as it is a subclass of AsyncWrap. Not sure if this is moved easily from C++ to Javascript. |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
Author
|
I like this idea, I'm trying it right now :) |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
Author
|
IncomingMessage works for HTTPParser.REQUEST :D Should we use ClientRequest for HTTPParser.RESPONSE? (ref: https://github.com/nodejs/node/blob/master/lib/_http_client.js#L631-L635) |
Sorry, something went wrong.
Copy link
Copy Markdown
Member
|
+1.
Il giorno mar 20 nov 2018 alle 03:06 Matheus Marchini <
notifications@github.com> ha scritto:
… IncomingMessage works for HTTPParser.REQUEST :D
Should we use ClientRequest for HTTPParser.RESPONSE? (ref:
https://github.com/nodejs/node/blob/master/lib/_http_client.js#L631-L635)
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#24330 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AADL44CPpfqGHEnSfEynqjwrrO897J_dks5uw2O8gaJpZM4YaziK>
.
|
Sorry, something went wrong.
Copy link
Copy Markdown
Member
|
Using ClientRequest sound reasonable in current implementation even it's not that intuitive that a parser used for the response is bound to the object used for the request. I think we should not use the same resource type (e.g. HTTPPARSER) for both cases if the object passed is different. |
Sorry, something went wrong.
Copy link
Copy Markdown
Member
@addaleax Hmm, but why is it slower then. The initial approach doesn't increase the number of native calls just changes from reinitialize to createParser which is mostly doing the same. |
Sorry, something went wrong.
Copy link
Copy Markdown
Member
Regarding that and previusely work that have been done using JavaScript. Would it not be possible to compile llhttp or http_parser to WebAssembly and remove that overhead? |
Sorry, something went wrong.
addaleax
pushed a commit
to Drieger/node
that referenced
this pull request
Apr 21, 2019
Change resource being used, previously HTTParser was being reused. We are now using IncomingMessage and ClientRequest objects. The goal here is to make the async resource unique for each async operatio Refs: nodejs#24330 Refs: nodejs/diagnostics#248 Refs: nodejs#21313 Co-authored-by: Matheus Marchini <mat@mmarchini.me>
mmarchini
pushed a commit
that referenced
this pull request
Apr 22, 2019
Change resource being used, previously HTTParser was being reused. We are now using IncomingMessage and ClientRequest objects. The goal here is to make the async resource unique for each async operatio Refs: #24330 Refs: nodejs/diagnostics#248 Refs: #21313 Co-authored-by: Matheus Marchini <mat@mmarchini.me> PR-URL: #25094 Reviewed-By: Matteo Collina <matteo.collina@gmail.com> Reviewed-By: Benedikt Meurer <benedikt.meurer@gmail.com> Reviewed-By: Anna Henningsen <anna@addaleax.net>
3 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
As discussed in nodejs/diagnostics#248 and
#21313, reusing HTTPParser resource
is a blocker for landing
require('async_hooks').currentAsyncResource(), since reusing an async
resource interanlly would make it infeasible to use them reliably in
WeakMaps. Two suggestions came up: have a wrapper around the HTTPParser
which we would use as the async resource, or stop reusing HTTPParser in
our HTTP server/client code.
This commit implements the latter, since we have a better GC now and
reusing HTTPParser might make sense anymore. This also will avoid one
extra JS->C++ call in some cases, which should improve performance for
these cases.
Checklist