Improve performance of vnodeId generation #2978

developit · 2021-02-06T18:07:33Z

This PR removes the need for an options._vnodeId shared counter by creating separate "keyspaces" within vnodeId (vnode._original) using signed integers. Despite not sharing the ID counter, h() allocates positive monotonic ID's and jsx() negative IDs, which prevents collisions. It also doubles the number of available ID values. The performance improvement here is a result of avoiding a property get + set on every VNode, not sure if that will come through in the benchmarks since we already do a loop and set many properties on props.

Another option that might be worth testing would be to keep IDs unsigned by having jsx() generate IDs starting at 9e9. This would create a collision scenario if h() were to create 9 billion VNodes.

…ces for h() and jsx() via a signed int

github-actions · 2021-02-06T18:07:49Z

📊 Tachometer Benchmark Results

Summary

duration

02_replace1k: unsure 🔍 -5% - +14% (-7.02ms - +19.12ms)
^{preact-local vs preact-master}
03_update10th1k_x16: unsure 🔍 -5% - +2% (-1.74ms - +0.65ms)
^{preact-local vs preact-master}
07_create10k: unsure 🔍 -1% - +1% (-11.90ms - +9.02ms)
^{preact-local vs preact-master}
filter_list: unsure 🔍 -4% - +3% (-1.22ms - +0.91ms)
^{preact-local vs preact-master}
hydrate1k: faster ✔ 0% - 4% (0.18ms - 5.15ms)
^{preact-local vs preact-master}
many_updates: unsure 🔍 -13% - +6% (-4.18ms - +2.14ms)
^{preact-local vs preact-master}
text_update: unsure 🔍 -4% - +4% (-0.17ms - +0.15ms)
^{preact-local vs preact-master}

usedJSHeapSize

02_replace1k: unsure 🔍 -0% - +0% (-0.02ms - +0.01ms)
^{preact-local vs preact-master}
03_update10th1k_x16: unsure 🔍 -0% - +0% (-0.01ms - +0.01ms)
^{preact-local vs preact-master}
07_create10k: unsure 🔍 -0% - +0% (-0.00ms - +0.00ms)
^{preact-local vs preact-master}
filter_list: unsure 🔍 -1% - +2% (-0.02ms - +0.03ms)
^{preact-local vs preact-master}
hydrate1k: unsure 🔍 -0% - +0% (-0.02ms - +0.01ms)
^{preact-local vs preact-master}
many_updates: unsure 🔍 -0% - +0% (-0.01ms - +0.02ms)
^{preact-local vs preact-master}
text_update: unsure 🔍 -0% - -0% (-0.00ms - -0.00ms)
^{preact-local vs preact-master}

Results

02_replace1k

Browser: chrome-headless 90.0.4430.93
Sample size: 80
Built by: Benchmarks #263
Commit: 477ab13

duration

Version	Avg time	vs preact-master	vs preact-local
preact-master	131.82ms - 147.00ms	-	unsure 🔍 -13% - +5% -19.12ms - +7.02ms
preact-local	134.83ms - 156.10ms	unsure 🔍 -5% - +14% -7.02ms - +19.12ms	-

usedJSHeapSize

Version	Avg time	vs preact-master	vs preact-local
preact-master	3.59ms - 3.61ms	-	unsure 🔍 -0% - +0% -0.01ms - +0.02ms
preact-local	3.59ms - 3.61ms	unsure 🔍 -0% - +0% -0.02ms - +0.01ms	-

run-warmup-0

Version	Avg time	vs preact-master	vs preact-local
preact-master	55.94ms - 64.15ms	-	unsure 🔍 -7% - +12% -4.24ms - +6.84ms
preact-local	55.02ms - 62.46ms	unsure 🔍 -11% - +7% -6.84ms - +4.24ms	-

run-warmup-1

Version	Avg time	vs preact-master	vs preact-local
preact-master	86.81ms - 98.33ms	-	unsure 🔍 -10% - +9% -9.48ms - +8.22ms
preact-local	86.48ms - 99.92ms	unsure 🔍 -9% - +10% -8.22ms - +9.48ms	-

run-warmup-2

Version	Avg time	vs preact-master	vs preact-local
preact-master	93.83ms - 109.50ms	-	unsure 🔍 -11% - +11% -10.88ms - +11.20ms
preact-local	93.73ms - 109.29ms	unsure 🔍 -11% - +11% -11.20ms - +10.88ms	-

run-warmup-3

Version	Avg time	vs preact-master	vs preact-local
preact-master	53.24ms - 61.64ms	-	unsure 🔍 -14% - +8% -8.24ms - +5.10ms
preact-local	53.83ms - 64.20ms	unsure 🔍 -9% - +14% -5.10ms - +8.24ms	-

run-warmup-4

Version	Avg time	vs preact-master	vs preact-local
preact-master	93.39ms - 111.90ms	-	unsure 🔍 -4% - +21% -3.54ms - +19.46ms
preact-local	87.86ms - 101.50ms	unsure 🔍 -18% - +3% -19.46ms - +3.54ms	-

run-final

Version	Avg time	vs preact-master	vs preact-local
preact-master	51.10ms - 56.64ms	-	unsure 🔍 -13% - +4% -7.34ms - +2.40ms
preact-local	52.33ms - 60.35ms	unsure 🔍 -5% - +14% -2.40ms - +7.34ms	-

03_update10th1k_x16

Browser: chrome-headless 90.0.4430.93
Sample size: 100
Built by: Benchmarks #263
Commit: 477ab13

duration

Version	Avg time	vs preact-master	vs preact-local
preact-master	34.95ms - 36.66ms	-	unsure 🔍 -2% - +5% -0.65ms - +1.74ms
preact-local	34.42ms - 36.10ms	unsure 🔍 -5% - +2% -1.74ms - +0.65ms	-

usedJSHeapSize

Version	Avg time	vs preact-master	vs preact-local
preact-master	3.65ms - 3.66ms	-	unsure 🔍 -0% - +0% -0.01ms - +0.01ms
preact-local	3.66ms - 3.66ms	unsure 🔍 -0% - +0% -0.01ms - +0.01ms	-

07_create10k

Browser: chrome-headless 90.0.4430.93
Sample size: 50
Built by: Benchmarks #263
Commit: 477ab13

duration

Version	Avg time	vs preact-master	vs preact-local
preact-master	1314.35ms - 1328.09ms	-	unsure 🔍 -1% - +1% -9.02ms - +11.90ms
preact-local	1311.90ms - 1327.67ms	unsure 🔍 -1% - +1% -11.90ms - +9.02ms	-

usedJSHeapSize

Version	Avg time	vs preact-master	vs preact-local
preact-master	25.98ms - 25.98ms	-	unsure 🔍 -0% - +0% -0.00ms - +0.00ms
preact-local	25.98ms - 25.98ms	unsure 🔍 -0% - +0% -0.00ms - +0.00ms	-

filter_list

Browser: chrome-headless 90.0.4430.93
Sample size: 70
Built by: Benchmarks #263
Commit: 477ab13

duration

Version	Avg time	vs preact-master	vs preact-local
preact-master	27.12ms - 28.75ms	-	unsure 🔍 -3% - +4% -0.91ms - +1.22ms
preact-local	27.09ms - 28.47ms	unsure 🔍 -4% - +3% -1.22ms - +0.91ms	-

usedJSHeapSize

Version	Avg time	vs preact-master	vs preact-local
preact-master	1.64ms - 1.68ms	-	unsure 🔍 -2% - +1% -0.03ms - +0.02ms
preact-local	1.64ms - 1.68ms	unsure 🔍 -1% - +2% -0.02ms - +0.03ms	-

hydrate1k

Browser: chrome-headless 90.0.4430.93
Sample size: 60
Built by: Benchmarks #263
Commit: 477ab13

duration

Version	Avg time	vs preact-master	vs preact-local
preact-master	115.46ms - 118.75ms	-	slower ❌ 0% - 5% 0.18ms - 5.15ms
preact-local	112.58ms - 116.31ms	faster ✔ 0% - 4% 0.18ms - 5.15ms	-

usedJSHeapSize

Version	Avg time	vs preact-master	vs preact-local
preact-master	6.33ms - 6.36ms	-	unsure 🔍 -0% - +0% -0.01ms - +0.02ms
preact-local	6.33ms - 6.35ms	unsure 🔍 -0% - +0% -0.02ms - +0.01ms	-

many_updates

Browser: chrome-headless 90.0.4430.93
Sample size: 80
Built by: Benchmarks #263
Commit: 477ab13

duration

Version	Avg time	vs preact-master	vs preact-local
preact-master	29.41ms - 34.55ms	-	unsure 🔍 -7% - +14% -2.14ms - +4.18ms
preact-local	29.12ms - 32.80ms	unsure 🔍 -13% - +6% -4.18ms - +2.14ms	-

usedJSHeapSize

Version	Avg time	vs preact-master	vs preact-local
preact-master	4.84ms - 4.87ms	-	unsure 🔍 -0% - +0% -0.02ms - +0.01ms
preact-local	4.85ms - 4.87ms	unsure 🔍 -0% - +0% -0.01ms - +0.02ms	-

text_update

Browser: chrome-headless 90.0.4430.93
Sample size: 230
Built by: Benchmarks #263
Commit: 477ab13

duration

Version	Avg time	vs preact-master	vs preact-local
preact-master	3.66ms - 3.92ms	-	unsure 🔍 -4% - +4% -0.15ms - +0.17ms
preact-local	3.69ms - 3.87ms	unsure 🔍 -4% - +4% -0.17ms - +0.15ms	-

usedJSHeapSize

Version	Avg time	vs preact-master	vs preact-local
preact-master	0.83ms - 0.83ms	-	unsure 🔍 +0% - +0% +0.00ms - +0.00ms
preact-local	0.83ms - 0.83ms	unsure 🔍 -0% - -0% -0.00ms - -0.00ms	-

_{tachometer-reporter-action v2 for Benchmarks}

coveralls · 2021-02-06T18:09:37Z

Coverage increased (+0.0007%) to 99.626% when pulling 477ab13 on perf-vnodeid into 84b1331 on master.

github-actions · 2021-02-06T18:10:56Z

Size Change: -25 B (0%)

Total Size: 42 kB

Filename	Size	Change
`dist/preact.js`	3.98 kB	+12 B (0%)
`dist/preact.min.js`	4 kB	-5 B (0%)
`dist/preact.module.js`	4 kB	-1 B
`dist/preact.umd.js`	4.04 kB	-3 B (0%)
`jsx-runtime/dist/jsxRuntime.js`	317 B	-10 B (3%)
`jsx-runtime/dist/jsxRuntime.module.js`	327 B	-8 B (2%)
`jsx-runtime/dist/jsxRuntime.umd.js`	395 B	-10 B (2%)

ℹ️ View Unchanged

Filename	Size	Change
`compat/dist/compat.js`	3.39 kB	0 B
`compat/dist/compat.module.js`	3.38 kB	0 B
`compat/dist/compat.umd.js`	3.44 kB	0 B
`debug/dist/debug.js`	2.99 kB	0 B
`debug/dist/debug.module.js`	2.98 kB	0 B
`debug/dist/debug.umd.js`	3.07 kB	0 B
`devtools/dist/devtools.js`	232 B	0 B
`devtools/dist/devtools.module.js`	241 B	0 B
`devtools/dist/devtools.umd.js`	308 B	0 B
`hooks/dist/hooks.js`	1.13 kB	0 B
`hooks/dist/hooks.module.js`	1.15 kB	0 B
`hooks/dist/hooks.umd.js`	1.2 kB	0 B
`test-utils/dist/testUtils.js`	437 B	0 B
`test-utils/dist/testUtils.module.js`	439 B	0 B
`test-utils/dist/testUtils.umd.js`	515 B	0 B

_{compressed-size-action}

marvinhagemeister

Nice one!

Improve performance of vnodeId generation by creating separate keyspa…

971bc48

…ces for h() and jsx() via a signed int

developit added a commit that referenced this pull request Feb 6, 2021

alternative version of #2978 using unsigned ints

d3f21ae

developit mentioned this pull request Feb 6, 2021

Perf vnodeid alt #2979

Closed

Merge branch 'master' into perf-vnodeid

477ab13

marvinhagemeister approved these changes May 10, 2021

View reviewed changes

JoviDeCroock approved these changes May 10, 2021

View reviewed changes

marvinhagemeister merged commit fec28e0 into master May 10, 2021

marvinhagemeister deleted the perf-vnodeid branch May 10, 2021 15:32

developit mentioned this pull request May 11, 2021

[restructure] Use signed ints for VNode ID generation #3145

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Improve performance of vnodeId generation #2978

Improve performance of vnodeId generation #2978

developit commented Feb 6, 2021 •

edited

Loading

github-actions bot commented Feb 6, 2021 •

edited

Loading

📊 Tachometer Benchmark Results

Summary

duration

usedJSHeapSize

Results

duration

usedJSHeapSize

run-warmup-0

run-warmup-1

run-warmup-2

run-warmup-3

run-warmup-4

run-final

duration

usedJSHeapSize

duration

usedJSHeapSize

duration

usedJSHeapSize

duration

usedJSHeapSize

duration

usedJSHeapSize

duration

usedJSHeapSize

coveralls commented Feb 6, 2021 •

edited

Loading

github-actions bot commented Feb 6, 2021 •

edited

Loading

marvinhagemeister left a comment

Improve performance of vnodeId generation #2978

Improve performance of vnodeId generation #2978

Conversation

developit commented Feb 6, 2021 • edited Loading

github-actions bot commented Feb 6, 2021 • edited Loading

📊 Tachometer Benchmark Results

Summary

duration

usedJSHeapSize

Results

duration

usedJSHeapSize

run-warmup-0

run-warmup-1

run-warmup-2

run-warmup-3

run-warmup-4

run-final

duration

usedJSHeapSize

duration

usedJSHeapSize

duration

usedJSHeapSize

duration

usedJSHeapSize

duration

usedJSHeapSize

duration

usedJSHeapSize

coveralls commented Feb 6, 2021 • edited Loading

github-actions bot commented Feb 6, 2021 • edited Loading

marvinhagemeister left a comment

Choose a reason for hiding this comment

developit commented Feb 6, 2021 •

edited

Loading

github-actions bot commented Feb 6, 2021 •

edited

Loading

coveralls commented Feb 6, 2021 •

edited

Loading

github-actions bot commented Feb 6, 2021 •

edited

Loading