GeezerVison is showing github.com — crypto: add buffering to randomInt by tniessen · Pull Request #35110 · nodejs/node
Conversation
This adds a buffer to randomInt to prevent calling randomBytes on each invocation. It also simplifies the actual calculation to always be synchronous, and doesn't duplicate the logic for asynchronous behavior. However, this change does include additional code to manage the cache.
Performance is improved significantly, by large factors, for synchronous calls and sequential asynchronous calls. For parallel calls, performance is about the same when requesting a few hundred random integers at a time.
Performance is significantly worse when requesting a large number of random integers asynchronously and in parallel. The reason is that this implementation delays all requests until the buffer has been refilled with random data, then serves as many requests as it can, then delays the remaining requests until the buffer has been refilled again, and so on. With a buffer size of 3KiB, 500 random integers can be provided without having to call randomBytes. When the benchmark requests 100,000 random integers asynchronously and in parallel, instead of processing 100,000 calls to randomBytes in parallel, the buffered implementation provides 200 batches of 500 numbers each sequentially.
- We could add additional code to handle such cases and avoid this performance penalty, but I decided to leave the additional complexity out for now. I assume that most use cases will require a sequence of random numbers, and not necessarily a large amount of random numbers at once.
- We could also change the buffer size, but that would require more memory, and could lead to slightly longer delays when refilling the buffer.
- We could also disable buffering for async in general, and only use it for sync calls. That would make any single call to
randomInt with a callback slow, but would make generating huge amounts of random integers at the same time faster.
Benchmark results based on 30 runs (3988798fa4bdbd3ade4cb23289ec0ae05621d438 compared to the master branch at b5a47ca):
confidence improvement accuracy (*) (**) (***)
crypto/randomInt.js n=1000 max=100 min=-100 mode='async-parallel' 7.61 % ±8.19% ±10.90% ±14.19%
crypto/randomInt.js n=1000 max=100 min=-100 mode='async-sequential' *** 113.40 % ±9.73% ±12.98% ±16.98%
crypto/randomInt.js n=1000 max=100 min=-100 mode='sync' *** 151.27 % ±4.81% ±6.41% ±8.36%
crypto/randomInt.js n=1000 max=100 min=-10000 mode='async-parallel' 1.72 % ±8.22% ±10.97% ±14.35%
crypto/randomInt.js n=1000 max=100 min=-10000 mode='async-sequential' *** 110.33 % ±9.82% ±13.12% ±17.17%
crypto/randomInt.js n=1000 max=100 min=-10000 mode='sync' *** 138.80 % ±7.85% ±10.54% ±13.92%
crypto/randomInt.js n=1000 max=100 min=-140737488355327 mode='async-parallel' *** 39.70 % ±9.72% ±12.99% ±17.03%
crypto/randomInt.js n=1000 max=100 min=-140737488355327 mode='async-sequential' *** 277.28 % ±12.17% ±16.35% ±21.61%
crypto/randomInt.js n=1000 max=100 min=-140737488355327 mode='sync' *** 264.60 % ±9.85% ±13.13% ±17.16%
crypto/randomInt.js n=1000 max=10000 min=-100 mode='async-parallel' * 6.06 % ±5.92% ±7.88% ±10.26%
crypto/randomInt.js n=1000 max=10000 min=-100 mode='async-sequential' *** 97.52 % ±10.61% ±14.20% ±18.65%
crypto/randomInt.js n=1000 max=10000 min=-100 mode='sync' *** 138.26 % ±8.36% ±11.26% ±14.94%
crypto/randomInt.js n=1000 max=10000 min=-10000 mode='async-parallel' -4.46 % ±8.33% ±11.16% ±14.69%
crypto/randomInt.js n=1000 max=10000 min=-10000 mode='async-sequential' *** 103.98 % ±10.82% ±14.46% ±18.96%
crypto/randomInt.js n=1000 max=10000 min=-10000 mode='sync' *** 141.13 % ±3.42% ±4.56% ±5.95%
crypto/randomInt.js n=1000 max=10000 min=-140737488355327 mode='async-parallel' *** 40.64 % ±4.59% ±6.11% ±7.95%
crypto/randomInt.js n=1000 max=10000 min=-140737488355327 mode='async-sequential' *** 263.09 % ±23.93% ±32.22% ±42.72%
crypto/randomInt.js n=1000 max=10000 min=-140737488355327 mode='sync' *** 246.81 % ±4.33% ±5.83% ±7.72%
crypto/randomInt.js n=1000 max=140737488355328 min=-100 mode='async-parallel' *** 37.40 % ±7.18% ±9.59% ±12.55%
crypto/randomInt.js n=1000 max=140737488355328 min=-100 mode='async-sequential' *** 275.54 % ±11.41% ±15.31% ±20.20%
crypto/randomInt.js n=1000 max=140737488355328 min=-100 mode='sync' *** 247.66 % ±3.52% ±4.71% ±6.17%
crypto/randomInt.js n=1000 max=140737488355328 min=-10000 mode='async-parallel' *** 36.03 % ±7.56% ±10.11% ±13.27%
crypto/randomInt.js n=1000 max=140737488355328 min=-10000 mode='async-sequential' *** 254.88 % ±18.48% ±24.85% ±32.86%
crypto/randomInt.js n=1000 max=140737488355328 min=-10000 mode='sync' *** 251.64 % ±3.86% ±5.16% ±6.76%
crypto/randomInt.js n=1000 max=140737488355328 min=-140737488355327 mode='async-parallel' -0.22 % ±7.35% ±9.80% ±12.78%
crypto/randomInt.js n=1000 max=140737488355328 min=-140737488355327 mode='async-sequential' *** 104.97 % ±12.95% ±17.36% ±22.87%
crypto/randomInt.js n=1000 max=140737488355328 min=-140737488355327 mode='sync' *** 138.63 % ±4.24% ±5.71% ±7.56%
crypto/randomInt.js n=100000 max=100 min=-100 mode='async-parallel' *** -41.12 % ±1.21% ±1.62% ±2.13%
crypto/randomInt.js n=100000 max=100 min=-100 mode='async-sequential' *** 691.82 % ±7.30% ±9.81% ±12.97%
crypto/randomInt.js n=100000 max=100 min=-100 mode='sync' *** 1310.10 % ±36.17% ±48.74% ±64.70%
crypto/randomInt.js n=100000 max=100 min=-10000 mode='async-parallel' *** -40.20 % ±1.28% ±1.72% ±2.26%
crypto/randomInt.js n=100000 max=100 min=-10000 mode='async-sequential' *** 691.30 % ±8.86% ±11.91% ±15.77%
crypto/randomInt.js n=100000 max=100 min=-10000 mode='sync' *** 1290.11 % ±55.81% ±75.21% ±99.85%
crypto/randomInt.js n=100000 max=100 min=-140737488355327 mode='async-parallel' *** -38.02 % ±0.83% ±1.11% ±1.47%
crypto/randomInt.js n=100000 max=100 min=-140737488355327 mode='async-sequential' *** 1434.15 % ±22.58% ±30.43% ±40.39%
crypto/randomInt.js n=100000 max=100 min=-140737488355327 mode='sync' *** 2233.63 % ±61.72% ±83.18% ±110.43%
crypto/randomInt.js n=100000 max=10000 min=-100 mode='async-parallel' *** -40.04 % ±1.48% ±1.99% ±2.64%
crypto/randomInt.js n=100000 max=10000 min=-100 mode='async-sequential' *** 694.65 % ±7.53% ±10.12% ±13.38%
crypto/randomInt.js n=100000 max=10000 min=-100 mode='sync' *** 1338.29 % ±31.28% ±42.15% ±55.95%
crypto/randomInt.js n=100000 max=10000 min=-10000 mode='async-parallel' *** -39.62 % ±1.15% ±1.54% ±2.04%
crypto/randomInt.js n=100000 max=10000 min=-10000 mode='async-sequential' *** 685.36 % ±8.82% ±11.87% ±15.71%
crypto/randomInt.js n=100000 max=10000 min=-10000 mode='sync' *** 1286.39 % ±43.29% ±58.33% ±77.43%
crypto/randomInt.js n=100000 max=10000 min=-140737488355327 mode='async-parallel' *** -38.56 % ±1.02% ±1.37% ±1.80%
crypto/randomInt.js n=100000 max=10000 min=-140737488355327 mode='async-sequential' *** 1450.61 % ±18.13% ±24.43% ±32.43%
crypto/randomInt.js n=100000 max=10000 min=-140737488355327 mode='sync' *** 2137.98 % ±87.00% ±117.25% ±155.67%
crypto/randomInt.js n=100000 max=140737488355328 min=-100 mode='async-parallel' *** -39.14 % ±0.82% ±1.09% ±1.43%
crypto/randomInt.js n=100000 max=140737488355328 min=-100 mode='async-sequential' *** 1453.44 % ±16.70% ±22.50% ±29.87%
crypto/randomInt.js n=100000 max=140737488355328 min=-100 mode='sync' *** 2161.60 % ±72.54% ±97.76% ±129.79%
crypto/randomInt.js n=100000 max=140737488355328 min=-10000 mode='async-parallel' *** -38.39 % ±1.10% ±1.47% ±1.93%
crypto/randomInt.js n=100000 max=140737488355328 min=-10000 mode='async-sequential' *** 1449.92 % ±13.92% ±18.74% ±24.86%
crypto/randomInt.js n=100000 max=140737488355328 min=-10000 mode='sync' *** 2169.74 % ±67.67% ±91.20% ±121.08%
crypto/randomInt.js n=100000 max=140737488355328 min=-140737488355327 mode='async-parallel' *** -51.28 % ±1.46% ±1.96% ±2.57%
crypto/randomInt.js n=100000 max=140737488355328 min=-140737488355327 mode='async-sequential' *** 695.29 % ±8.29% ±11.15% ±14.75%
crypto/randomInt.js n=100000 max=140737488355328 min=-140737488355327 mode='sync' *** 1271.56 % ±36.90% ±49.73% ±66.01%
Be aware that when doing many comparisons the risk of a false-positive
result increases. In this case there are 54 comparisons, you can thus
expect the following amount of false-positive results:
2.70 false positives, when considering a 5% risk acceptance (*, **, ***),
0.54 false positives, when considering a 1% risk acceptance (**, ***),
0.05 false positives, when considering a 0.1% risk acceptance (***)
Checklist
-
make -j4 test (UNIX), or vcbuild test (Windows) passes
- tests and/or benchmarks are included
- commit message follows commit guidelines
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
bricss reacted with thumbs up emoji
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
|
I'm decidedly 'meh' on this. No one's been complaining about its speed so far.
|
bricss, BridgeAR, and puzpuzpuz reacted with confused emoji
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
|
@bnoordhuis How about adding this for the sync case only? That's a relatively small change with a big payoff, and I think the performance of the sync case is more critical anyway. The downside is that this approach results in the existing code duplication between sync and async behavior, which essentially implements the entire thing twice.
|
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
|
@nodejs/crypto can you review this please? Those benchmark results look very good!
|
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
using an array for this is usually not performing as expected for high numbers. Using a linked list solves this problem.
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we have fixed-queue which also might be appropriate?
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It might make sense to introduce a random buffer internal utility that can be used in multiple places (here, the randomuuid fn, etc). That can be done later tho
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
puzpuzpuz reacted with thumbs up emoji
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
|
@mcollina I doubled the cache size, but a more efficient way to improve performance might be to reduce the number of bytes that is used for each random integer. In most cases, we don't actually need 6 bytes, probably ony 2 to 4 bytes. (It depends on the use case, of course.)
|
mcollina reacted with thumbs up emoji
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
|
@tniessen could you resolve conflicts on this one?
|
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
tniessen
commented
Jan 18, 2021
•
edited
Loading
Uh oh!
There was an error while loading. Please reload this page.
|
Benchmark results still look good:
Table
| n |
max |
min |
mode |
confidence |
improvement |
accuracy (*) |
accuracy (**) |
accuracy (***) |
| 100000 |
10000 |
-10000 |
'async-parallel' |
*** |
41.05 % |
±4.60% |
±6.17% |
±8.12% |
| 100000 |
10000 |
-10000 |
'async-sequential' |
*** |
1931.62 % |
±106.63% |
±143.69% |
±190.75% |
| 100000 |
10000 |
-10000 |
'sync' |
*** |
1976.14 % |
±114.57% |
±154.41% |
±204.98% |
| 100000 |
10000 |
-100 |
'async-parallel' |
*** |
43.43 % |
±5.20% |
±6.95% |
±9.10% |
| 100000 |
10000 |
-100 |
'async-sequential' |
*** |
1888.73 % |
±99.87% |
±134.60% |
±178.69% |
| 100000 |
10000 |
-100 |
'sync' |
*** |
1895.70 % |
±99.80% |
±134.49% |
±178.54% |
| 100000 |
10000 |
-140737488355327 |
'async-parallel' |
*** |
44.39 % |
±4.65% |
±6.23% |
±8.22% |
| 100000 |
10000 |
-140737488355327 |
'async-sequential' |
*** |
3455.63 % |
±230.82% |
±311.08% |
±412.99% |
| 100000 |
10000 |
-140737488355327 |
'sync' |
*** |
3056.68 % |
±158.66% |
±213.82% |
±283.87% |
| 100000 |
100 |
-10000 |
'async-parallel' |
*** |
39.66 % |
±5.26% |
±7.05% |
±9.27% |
| 100000 |
100 |
-10000 |
'async-sequential' |
*** |
1936.95 % |
±97.77% |
±131.76% |
±174.89% |
| 100000 |
100 |
-10000 |
'sync' |
*** |
1885.94 % |
±75.29% |
±101.47% |
±134.70% |
| 100000 |
100 |
-100 |
'async-parallel' |
*** |
39.74 % |
±5.70% |
±7.64% |
±10.06% |
| 100000 |
100 |
-100 |
'async-sequential' |
*** |
1878.43 % |
±88.73% |
±119.57% |
±158.72% |
| 100000 |
100 |
-100 |
'sync' |
*** |
1892.60 % |
±96.45% |
±129.98% |
±172.55% |
| 100000 |
100 |
-140737488355327 |
'async-parallel' |
*** |
40.54 % |
±4.49% |
±6.03% |
±7.96% |
| 100000 |
100 |
-140737488355327 |
'async-sequential' |
*** |
3335.10 % |
±234.14% |
±315.55% |
±418.91% |
| 100000 |
100 |
-140737488355327 |
'sync' |
*** |
3140.61 % |
±170.90% |
±230.32% |
±305.77% |
| 100000 |
140737488355328 |
-10000 |
'async-parallel' |
*** |
47.40 % |
±4.92% |
±6.60% |
±8.71% |
| 100000 |
140737488355328 |
-10000 |
'async-sequential' |
*** |
3300.80 % |
±182.10% |
±245.41% |
±325.80% |
| 100000 |
140737488355328 |
-10000 |
'sync' |
*** |
2980.74 % |
±144.38% |
±194.58% |
±258.31% |
| 100000 |
140737488355328 |
-100 |
'async-parallel' |
*** |
48.80 % |
±5.10% |
±6.86% |
±9.06% |
| 100000 |
140737488355328 |
-100 |
'async-sequential' |
*** |
3385.05 % |
±175.80% |
±236.92% |
±314.53% |
| 100000 |
140737488355328 |
-100 |
'sync' |
*** |
3041.96 % |
±176.29% |
±237.58% |
±315.41% |
| 100000 |
140737488355328 |
-140737488355327 |
'async-parallel' |
*** |
21.46 % |
±4.38% |
±5.86% |
±7.71% |
| 100000 |
140737488355328 |
-140737488355327 |
'async-sequential' |
*** |
2011.00 % |
±146.63% |
±197.62% |
±262.35% |
| 100000 |
140737488355328 |
-140737488355327 |
'sync' |
*** |
1863.61 % |
±92.30% |
±124.39% |
±165.14% |
| 1000 |
10000 |
-10000 |
'async-parallel' |
*** |
52.00 % |
±11.51% |
±15.44% |
±20.35% |
| 1000 |
10000 |
-10000 |
'async-sequential' |
*** |
159.33 % |
±10.13% |
±13.58% |
±17.88% |
| 1000 |
10000 |
-10000 |
'sync' |
*** |
545.42 % |
±38.66% |
±52.05% |
±69.01% |
| 1000 |
10000 |
-100 |
'async-parallel' |
*** |
38.86 % |
±6.89% |
±9.18% |
±11.99% |
| 1000 |
10000 |
-100 |
'async-sequential' |
*** |
156.91 % |
±13.86% |
±18.61% |
±24.56% |
| 1000 |
10000 |
-100 |
'sync' |
*** |
605.82 % |
±54.76% |
±73.78% |
±97.89% |
| 1000 |
10000 |
-140737488355327 |
'async-parallel' |
*** |
78.15 % |
±5.72% |
±7.61% |
±9.90% |
| 1000 |
10000 |
-140737488355327 |
'async-sequential' |
*** |
331.03 % |
±20.68% |
±27.83% |
±36.86% |
| 1000 |
10000 |
-140737488355327 |
'sync' |
*** |
516.57 % |
±42.54% |
±57.30% |
±76.01% |
| 1000 |
100 |
-10000 |
'async-parallel' |
*** |
36.48 % |
±8.30% |
±11.06% |
±14.44% |
| 1000 |
100 |
-10000 |
'async-sequential' |
*** |
152.74 % |
±10.06% |
±13.46% |
±17.65% |
| 1000 |
100 |
-10000 |
'sync' |
*** |
566.73 % |
±35.87% |
±48.30% |
±64.04% |
| 1000 |
100 |
-100 |
'async-parallel' |
*** |
31.18 % |
±6.98% |
±9.30% |
±12.11% |
| 1000 |
100 |
-100 |
'async-sequential' |
*** |
161.54 % |
±11.41% |
±15.28% |
±20.10% |
| 1000 |
100 |
-100 |
'sync' |
*** |
543.65 % |
±36.10% |
±48.61% |
±64.46% |
| 1000 |
100 |
-140737488355327 |
'async-parallel' |
*** |
97.30 % |
±10.66% |
±14.31% |
±18.89% |
| 1000 |
100 |
-140737488355327 |
'async-sequential' |
*** |
353.64 % |
±29.03% |
±39.09% |
±51.84% |
| 1000 |
100 |
-140737488355327 |
'sync' |
*** |
494.78 % |
±20.94% |
±28.15% |
±37.24% |
| 1000 |
140737488355328 |
-10000 |
'async-parallel' |
*** |
82.55 % |
±10.74% |
±14.38% |
±18.90% |
| 1000 |
140737488355328 |
-10000 |
'async-sequential' |
*** |
325.50 % |
±17.69% |
±23.80% |
±31.51% |
| 1000 |
140737488355328 |
-10000 |
'sync' |
*** |
525.51 % |
±45.32% |
±61.05% |
±80.97% |
| 1000 |
140737488355328 |
-100 |
'async-parallel' |
*** |
80.98 % |
±11.91% |
±15.94% |
±20.93% |
| 1000 |
140737488355328 |
-100 |
'async-sequential' |
*** |
325.43 % |
±15.49% |
±20.82% |
±27.53% |
| 1000 |
140737488355328 |
-100 |
'sync' |
*** |
483.61 % |
±36.11% |
±48.61% |
±64.44% |
| 1000 |
140737488355328 |
-140737488355327 |
'async-parallel' |
*** |
47.12 % |
±7.40% |
±9.88% |
±12.92% |
| 1000 |
140737488355328 |
-140737488355327 |
'async-sequential' |
*** |
164.94 % |
±12.99% |
±17.42% |
±22.93% |
| 1000 |
140737488355328 |
-140737488355327 |
'sync' |
*** |
601.63 % |
±45.89% |
±61.83% |
±82.04% |
Unformatted output
confidence improvement accuracy (*) (**) (***)
crypto/randomInt.js n=100000 max=10000 min=-10000 mode='async-parallel' *** 41.05 % ±4.60% ±6.17% ±8.12%
crypto/randomInt.js n=100000 max=10000 min=-10000 mode='async-sequential' *** 1931.62 % ±106.63% ±143.69% ±190.75%
crypto/randomInt.js n=100000 max=10000 min=-10000 mode='sync' *** 1976.14 % ±114.57% ±154.41% ±204.98%
crypto/randomInt.js n=100000 max=10000 min=-100 mode='async-parallel' *** 43.43 % ±5.20% ±6.95% ±9.10%
crypto/randomInt.js n=100000 max=10000 min=-100 mode='async-sequential' *** 1888.73 % ±99.87% ±134.60% ±178.69%
crypto/randomInt.js n=100000 max=10000 min=-100 mode='sync' *** 1895.70 % ±99.80% ±134.49% ±178.54%
crypto/randomInt.js n=100000 max=10000 min=-140737488355327 mode='async-parallel' *** 44.39 % ±4.65% ±6.23% ±8.22%
crypto/randomInt.js n=100000 max=10000 min=-140737488355327 mode='async-sequential' *** 3455.63 % ±230.82% ±311.08% ±412.99%
crypto/randomInt.js n=100000 max=10000 min=-140737488355327 mode='sync' *** 3056.68 % ±158.66% ±213.82% ±283.87%
crypto/randomInt.js n=100000 max=100 min=-10000 mode='async-parallel' *** 39.66 % ±5.26% ±7.05% ±9.27%
crypto/randomInt.js n=100000 max=100 min=-10000 mode='async-sequential' *** 1936.95 % ±97.77% ±131.76% ±174.89%
crypto/randomInt.js n=100000 max=100 min=-10000 mode='sync' *** 1885.94 % ±75.29% ±101.47% ±134.70%
crypto/randomInt.js n=100000 max=100 min=-100 mode='async-parallel' *** 39.74 % ±5.70% ±7.64% ±10.06%
crypto/randomInt.js n=100000 max=100 min=-100 mode='async-sequential' *** 1878.43 % ±88.73% ±119.57% ±158.72%
crypto/randomInt.js n=100000 max=100 min=-100 mode='sync' *** 1892.60 % ±96.45% ±129.98% ±172.55%
crypto/randomInt.js n=100000 max=100 min=-140737488355327 mode='async-parallel' *** 40.54 % ±4.49% ±6.03% ±7.96%
crypto/randomInt.js n=100000 max=100 min=-140737488355327 mode='async-sequential' *** 3335.10 % ±234.14% ±315.55% ±418.91%
crypto/randomInt.js n=100000 max=100 min=-140737488355327 mode='sync' *** 3140.61 % ±170.90% ±230.32% ±305.77%
crypto/randomInt.js n=100000 max=140737488355328 min=-10000 mode='async-parallel' *** 47.40 % ±4.92% ±6.60% ±8.71%
crypto/randomInt.js n=100000 max=140737488355328 min=-10000 mode='async-sequential' *** 3300.80 % ±182.10% ±245.41% ±325.80%
crypto/randomInt.js n=100000 max=140737488355328 min=-10000 mode='sync' *** 2980.74 % ±144.38% ±194.58% ±258.31%
crypto/randomInt.js n=100000 max=140737488355328 min=-100 mode='async-parallel' *** 48.80 % ±5.10% ±6.86% ±9.06%
crypto/randomInt.js n=100000 max=140737488355328 min=-100 mode='async-sequential' *** 3385.05 % ±175.80% ±236.92% ±314.53%
crypto/randomInt.js n=100000 max=140737488355328 min=-100 mode='sync' *** 3041.96 % ±176.29% ±237.58% ±315.41%
crypto/randomInt.js n=100000 max=140737488355328 min=-140737488355327 mode='async-parallel' *** 21.46 % ±4.38% ±5.86% ±7.71%
crypto/randomInt.js n=100000 max=140737488355328 min=-140737488355327 mode='async-sequential' *** 2011.00 % ±146.63% ±197.62% ±262.35%
crypto/randomInt.js n=100000 max=140737488355328 min=-140737488355327 mode='sync' *** 1863.61 % ±92.30% ±124.39% ±165.14%
crypto/randomInt.js n=1000 max=10000 min=-10000 mode='async-parallel' *** 52.00 % ±11.51% ±15.44% ±20.35%
crypto/randomInt.js n=1000 max=10000 min=-10000 mode='async-sequential' *** 159.33 % ±10.13% ±13.58% ±17.88%
crypto/randomInt.js n=1000 max=10000 min=-10000 mode='sync' *** 545.42 % ±38.66% ±52.05% ±69.01%
crypto/randomInt.js n=1000 max=10000 min=-100 mode='async-parallel' *** 38.86 % ±6.89% ±9.18% ±11.99%
crypto/randomInt.js n=1000 max=10000 min=-100 mode='async-sequential' *** 156.91 % ±13.86% ±18.61% ±24.56%
crypto/randomInt.js n=1000 max=10000 min=-100 mode='sync' *** 605.82 % ±54.76% ±73.78% ±97.89%
crypto/randomInt.js n=1000 max=10000 min=-140737488355327 mode='async-parallel' *** 78.15 % ±5.72% ±7.61% ±9.90%
crypto/randomInt.js n=1000 max=10000 min=-140737488355327 mode='async-sequential' *** 331.03 % ±20.68% ±27.83% ±36.86%
crypto/randomInt.js n=1000 max=10000 min=-140737488355327 mode='sync' *** 516.57 % ±42.54% ±57.30% ±76.01%
crypto/randomInt.js n=1000 max=100 min=-10000 mode='async-parallel' *** 36.48 % ±8.30% ±11.06% ±14.44%
crypto/randomInt.js n=1000 max=100 min=-10000 mode='async-sequential' *** 152.74 % ±10.06% ±13.46% ±17.65%
crypto/randomInt.js n=1000 max=100 min=-10000 mode='sync' *** 566.73 % ±35.87% ±48.30% ±64.04%
crypto/randomInt.js n=1000 max=100 min=-100 mode='async-parallel' *** 31.18 % ±6.98% ±9.30% ±12.11%
crypto/randomInt.js n=1000 max=100 min=-100 mode='async-sequential' *** 161.54 % ±11.41% ±15.28% ±20.10%
crypto/randomInt.js n=1000 max=100 min=-100 mode='sync' *** 543.65 % ±36.10% ±48.61% ±64.46%
crypto/randomInt.js n=1000 max=100 min=-140737488355327 mode='async-parallel' *** 97.30 % ±10.66% ±14.31% ±18.89%
crypto/randomInt.js n=1000 max=100 min=-140737488355327 mode='async-sequential' *** 353.64 % ±29.03% ±39.09% ±51.84%
crypto/randomInt.js n=1000 max=100 min=-140737488355327 mode='sync' *** 494.78 % ±20.94% ±28.15% ±37.24%
crypto/randomInt.js n=1000 max=140737488355328 min=-10000 mode='async-parallel' *** 82.55 % ±10.74% ±14.38% ±18.90%
crypto/randomInt.js n=1000 max=140737488355328 min=-10000 mode='async-sequential' *** 325.50 % ±17.69% ±23.80% ±31.51%
crypto/randomInt.js n=1000 max=140737488355328 min=-10000 mode='sync' *** 525.51 % ±45.32% ±61.05% ±80.97%
crypto/randomInt.js n=1000 max=140737488355328 min=-100 mode='async-parallel' *** 80.98 % ±11.91% ±15.94% ±20.93%
crypto/randomInt.js n=1000 max=140737488355328 min=-100 mode='async-sequential' *** 325.43 % ±15.49% ±20.82% ±27.53%
crypto/randomInt.js n=1000 max=140737488355328 min=-100 mode='sync' *** 483.61 % ±36.11% ±48.61% ±64.44%
crypto/randomInt.js n=1000 max=140737488355328 min=-140737488355327 mode='async-parallel' *** 47.12 % ±7.40% ±9.88% ±12.92%
crypto/randomInt.js n=1000 max=140737488355328 min=-140737488355327 mode='async-sequential' *** 164.94 % ±12.99% ±17.42% ±22.93%
crypto/randomInt.js n=1000 max=140737488355328 min=-140737488355327 mode='sync' *** 601.63 % ±45.89% ±61.83% ±82.04%
Be aware that when doing many comparisons the risk of a false-positive
result increases. In this case there are 54 comparisons, you can thus
expect the following amount of false-positive results:
2.70 false positives, when considering a 5% risk acceptance (*, **, ***),
0.54 false positives, when considering a 1% risk acceptance (**, ***),
0.05 false positives, when considering a 0.1% risk acceptance (***)
|
puzpuzpuz reacted with thumbs up emoji
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
An idea for potential enhancement: consider adding options arg with disableEntropyCache setting to allow disabling the random cache. This way will make the API consistent with randomUUID (see #36729).
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Consider using secureBuffer here to integrate with secure heap (see randomUUID's random cache and #36779).
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for the suggestion. We could do that, but UUIDs and random ints always result in insecure memory allocations anyway that mostly destroy any security guarantees. JavaScript is terrible for secure memory management.
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
|
I believe this was technically ready to land a while ago, but I just addressed @aduh95's comments regarding primordials.
|
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
This comment has been minimized.
This comment has been minimized.
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
|
Benchmark results still look good:
Unformatted output
15:15:01 confidence improvement accuracy (*) (**) (***)
15:15:01 crypto/randomInt.jsn=100000 max=10000 min=-10000 mode='async-parallel' *** 24.11 % ±3.79% ±5.04% ±6.57%
15:15:01 crypto/randomInt.jsn=100000 max=10000 min=-10000 mode='async-sequential' *** 1854.25 % ±100.81% ±135.85% ±180.34%
15:15:01 crypto/randomInt.jsn=100000 max=10000 min=-10000 mode='sync' *** 1692.53 % ±86.05% ±115.97% ±153.96%
15:15:01 crypto/randomInt.jsn=100000 max=10000 min=-100 mode='async-parallel' *** 26.16 % ±4.40% ±5.86% ±7.66%
15:15:01 crypto/randomInt.jsn=100000 max=10000 min=-100 mode='async-sequential' *** 1908.77 % ±126.68% ±170.72% ±226.62%
15:15:01 crypto/randomInt.jsn=100000 max=10000 min=-100 mode='sync' *** 1681.47 % ±83.93% ±113.11% ±150.16%
15:15:01 crypto/randomInt.jsn=100000 max=10000 min=-140737488355327 mode='async-parallel' *** 20.07 % ±5.60% ±7.53% ±9.95%
15:15:01 crypto/randomInt.jsn=100000 max=10000 min=-140737488355327 mode='async-sequential' *** 3343.76 % ±182.99% ±246.62% ±327.41%
15:15:01 crypto/randomInt.jsn=100000 max=10000 min=-140737488355327 mode='sync' *** 2995.21 % ±184.03% ±248.02% ±329.26%
15:15:01 crypto/randomInt.jsn=100000 max=100 min=-10000 mode='async-parallel' *** 28.82 % ±4.26% ±5.68% ±7.42%
15:15:01 crypto/randomInt.jsn=100000 max=100 min=-10000 mode='async-sequential' *** 1854.69 % ±91.38% ±123.14% ±163.47%
15:15:01 crypto/randomInt.jsn=100000 max=100 min=-10000 mode='sync' *** 1725.99 % ±72.64% ±97.89% ±129.94%
15:15:01 crypto/randomInt.jsn=100000 max=100 min=-100 mode='async-parallel' *** 20.69 % ±3.74% ±4.98% ±6.50%
15:15:01 crypto/randomInt.jsn=100000 max=100 min=-100 mode='async-sequential' *** 1849.93 % ±89.96% ±121.23% ±160.92%
15:15:01 crypto/randomInt.jsn=100000 max=100 min=-100 mode='sync' *** 1624.21 % ±42.71% ±57.56% ±76.40%
15:15:01 crypto/randomInt.jsn=100000 max=100 min=-140737488355327 mode='async-parallel' *** 21.91 % ±5.37% ±7.21% ±9.54%
15:15:01 crypto/randomInt.jsn=100000 max=100 min=-140737488355327 mode='async-sequential' *** 3501.94 % ±202.95% ±273.51% ±363.11%
15:15:01 crypto/randomInt.jsn=100000 max=100 min=-140737488355327 mode='sync' *** 3044.37 % ±186.10% ±250.80% ±332.97%
15:15:01 crypto/randomInt.jsn=100000 max=140737488355328 min=-10000 mode='async-parallel' *** 22.47 % ±6.02% ±8.10% ±10.71%
15:15:01 crypto/randomInt.jsn=100000 max=140737488355328 min=-10000 mode='async-sequential' *** 3494.36 % ±215.78% ±290.80% ±386.06%
15:15:01 crypto/randomInt.jsn=100000 max=140737488355328 min=-10000 mode='sync' *** 3017.06 % ±166.74% ±224.71% ±298.33%
15:15:01 crypto/randomInt.jsn=100000 max=140737488355328 min=-100 mode='async-parallel' *** 29.59 % ±4.95% ±6.63% ±8.74%
15:15:01 crypto/randomInt.jsn=100000 max=140737488355328 min=-100 mode='async-sequential' *** 3281.68 % ±179.42% ±241.80% ±320.99%
15:15:01 crypto/randomInt.jsn=100000 max=140737488355328 min=-100 mode='sync' *** 2985.85 % ±242.80% ±327.22% ±434.41%
15:15:01 crypto/randomInt.jsn=100000 max=140737488355328 min=-140737488355327 mode='async-parallel' *** -6.84 % ±3.26% ±4.34% ±5.65%
15:15:01 crypto/randomInt.jsn=100000 max=140737488355328 min=-140737488355327 mode='async-sequential' *** 1861.71 % ±83.21% ±112.12% ±148.83%
15:15:01 crypto/randomInt.jsn=100000 max=140737488355328 min=-140737488355327 mode='sync' *** 1652.79 % ±101.12% ±136.28% ±180.92%
15:15:01 crypto/randomInt.jsn=1000 max=10000 min=-10000 mode='async-parallel' *** 55.56 % ±6.94% ±9.29% ±12.21%
15:15:01 crypto/randomInt.jsn=1000 max=10000 min=-10000 mode='async-sequential' *** 184.54 % ±14.83% ±19.90% ±26.24%
15:15:01 crypto/randomInt.jsn=1000 max=10000 min=-10000 mode='sync' *** 491.12 % ±24.91% ±33.46% ±44.18%
15:15:01 crypto/randomInt.jsn=1000 max=10000 min=-100 mode='async-parallel' *** 39.02 % ±6.75% ±8.98% ±11.70%
15:15:01 crypto/randomInt.jsn=1000 max=10000 min=-100 mode='async-sequential' *** 198.34 % ±14.60% ±19.62% ±25.95%
15:15:01 crypto/randomInt.jsn=1000 max=10000 min=-100 mode='sync' *** 500.96 % ±26.60% ±35.77% ±47.35%
15:15:01 crypto/randomInt.jsn=1000 max=10000 min=-140737488355327 mode='async-parallel' *** 85.51 % ±10.84% ±14.54% ±19.18%
15:15:01 crypto/randomInt.jsn=1000 max=10000 min=-140737488355327 mode='async-sequential' *** 363.90 % ±16.79% ±22.56% ±29.80%
15:15:01 crypto/randomInt.jsn=1000 max=10000 min=-140737488355327 mode='sync' *** 704.50 % ±57.53% ±77.49% ±102.80%
15:15:01 crypto/randomInt.jsn=1000 max=100 min=-10000 mode='async-parallel' *** 51.90 % ±12.02% ±16.13% ±21.27%
15:15:01 crypto/randomInt.jsn=1000 max=100 min=-10000 mode='async-sequential' *** 200.13 % ±17.37% ±23.32% ±30.80%
15:15:01 crypto/randomInt.jsn=1000 max=100 min=-10000 mode='sync' *** 499.05 % ±30.34% ±40.84% ±54.13%
15:15:01 crypto/randomInt.jsn=1000 max=100 min=-100 mode='async-parallel' *** 36.25 % ±9.41% ±12.52% ±16.30%
15:15:01 crypto/randomInt.jsn=1000 max=100 min=-100 mode='async-sequential' *** 191.84 % ±16.23% ±21.81% ±28.84%
15:15:01 crypto/randomInt.jsn=1000 max=100 min=-100 mode='sync' *** 465.86 % ±19.74% ±26.53% ±35.09%
15:15:01 crypto/randomInt.jsn=1000 max=100 min=-140737488355327 mode='async-parallel' *** 93.88 % ±12.16% ±16.32% ±21.52%
15:15:01 crypto/randomInt.jsn=1000 max=100 min=-140737488355327 mode='async-sequential' *** 388.58 % ±27.68% ±37.28% ±49.46%
15:15:01 crypto/randomInt.jsn=1000 max=100 min=-140737488355327 mode='sync' *** 772.60 % ±59.93% ±80.74% ±107.14%
15:15:01 crypto/randomInt.jsn=1000 max=140737488355328 min=-10000 mode='async-parallel' *** 76.59 % ±8.42% ±11.20% ±14.58%
15:15:01 crypto/randomInt.jsn=1000 max=140737488355328 min=-10000 mode='async-sequential' *** 384.21 % ±29.08% ±39.17% ±51.95%
15:15:01 crypto/randomInt.jsn=1000 max=140737488355328 min=-10000 mode='sync' *** 662.43 % ±42.43% ±57.11% ±75.66%
15:15:01 crypto/randomInt.jsn=1000 max=140737488355328 min=-100 mode='async-parallel' *** 72.87 % ±8.36% ±11.13% ±14.50%
15:15:01 crypto/randomInt.jsn=1000 max=140737488355328 min=-100 mode='async-sequential' *** 384.58 % ±30.60% ±41.19% ±54.58%
15:15:01 crypto/randomInt.jsn=1000 max=140737488355328 min=-100 mode='sync' *** 658.07 % ±39.90% ±53.70% ±71.14%
15:15:01 crypto/randomInt.jsn=1000 max=140737488355328 min=-140737488355327 mode='async-parallel' *** 40.61 % ±6.91% ±9.19% ±11.96%
15:15:01 crypto/randomInt.jsn=1000 max=140737488355328 min=-140737488355327 mode='async-sequential' *** 203.08 % ±20.27% ±27.28% ±36.14%
15:15:01 crypto/randomInt.jsn=1000 max=140737488355328 min=-140737488355327 mode='sync' *** 521.38 % ±40.28% ±54.24% ±71.94%
15:15:01
15:15:01 Be aware that when doing many comparisons the risk of a false-positive
15:15:01 result increases. In this case, there are 54 comparisons, you can thus
15:15:01 expect the following amount of false-positive results:
15:15:01 2.70 false positives, when considering a 5% risk acceptance (*, **, ***),
15:15:01 0.54 false positives, when considering a 1% risk acceptance (**, ***),
15:15:01 0.05 false positives, when considering a 0.1% risk acceptance (***)
|
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
|
Landed in 5dae7d6, thanks for reviewing.
|
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
gengjiawen
added a commit
to gengjiawen/node
that referenced
this pull request
Oct 8, 2026
When an asynchronous randomInt() call finds the cache empty, it queues
itself and starts randomFill() on the shared cache buffer. A synchronous
call made before that job completes refills the same buffer with
randomFillSync() and returns values from offset 0. The completion
callback then reset the offset to 0, so the queued call and later calls
returned the bytes that synchronous calls had already used. The
threadpool job and randomFillSync() could also write the buffer at the
same time.
Fill a separate buffer asynchronously and copy it into the cache once
the job is done.
Fixes: nodejs#66595
Refs: nodejs#35110
Signed-off-by: Jiawen Geng <[email protected]>
Assisted-by: claude:opus-5.5
gengjiawen
added a commit
to gengjiawen/node
that referenced
this pull request
Oct 13, 2026
When an asynchronous randomInt() call finds the cache empty, it queues
itself and starts randomFill() on the shared cache buffer. A synchronous
call made before that job completes refills the same buffer with
randomFillSync() and returns values from offset 0. The completion
callback then reset the offset to 0, so the queued call and later calls
returned the bytes that synchronous calls had already used. The
threadpool job and randomFillSync() could also write the buffer at the
same time.
Fill a separate buffer asynchronously and copy it into the cache once
the job is done.
Fixes: nodejs#66595
Refs: nodejs#35110
Signed-off-by: Jiawen Geng <[email protected]>
Assisted-by: claude:opus-5.5
nodejs-github-bot
pushed a commit
that referenced
this pull request
Oct 15, 2026
When an asynchronous randomInt() call finds the cache empty, it queues
itself and starts randomFill() on the shared cache buffer. A synchronous
call made before that job completes refills the same buffer with
randomFillSync() and returns values from offset 0. The completion
callback then reset the offset to 0, so the queued call and later calls
returned the bytes that synchronous calls had already used. The
threadpool job and randomFillSync() could also write the buffer at the
same time.
Fill a separate buffer asynchronously and copy it into the cache once
the job is done.
Fixes: #66595
Refs: #35110
Signed-off-by: Jiawen Geng <[email protected]>
Assisted-by: claude:opus-5.5
PR-URL: #66597
Reviewed-By: Filip Skokan <[email protected]>
aduh95
pushed a commit
that referenced
this pull request
Nov 12, 2026
When an asynchronous randomInt() call finds the cache empty, it queues
itself and starts randomFill() on the shared cache buffer. A synchronous
call made before that job completes refills the same buffer with
randomFillSync() and returns values from offset 0. The completion
callback then reset the offset to 0, so the queued call and later calls
returned the bytes that synchronous calls had already used. The
threadpool job and randomFillSync() could also write the buffer at the
same time.
Fill a separate buffer asynchronously and copy it into the cache once
the job is done.
Fixes: #66595
Refs: #35110
Signed-off-by: Jiawen Geng <[email protected]>
Assisted-by: claude:opus-5.5
PR-URL: #66597
Reviewed-By: Filip Skokan <[email protected]>
Uh oh!
There was an error while loading. Please reload this page.
Assignees
No one assigned
Labels
benchmark
Issues and PRs related to Node.js benchmarks and benchmarking infrastructure.
crypto
Issues and PRs related to the crypto subsystem.
performance
Issues and PRs related to the performance of Node.js.
Milestone
No milestone
Development
Successfully merging this pull request may close these issues.
Uh oh!
There was an error while loading. Please reload this page.
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.
|
mcollina
left a comment
