Open Bug 1939434 Opened 1 year ago Updated 7 months ago

Nightly is 5.5x (Was:12x) slower on a js demo at https://manueldois.github.io/Random-tests/Array%20hashing/dist/index.html with BigInt

Categories

(Core :: JavaScript Engine, task, P3)

Product:
Core
Component:
JavaScript Engine
Type:
task

Tracking

()

People

(Reporter: mayankleoboy1, Unassigned)

References

(Blocks 1 open bug,
URL
)

Details

Go to https://manueldois.github.io/Random-tests/Array%20hashing/dist/index.html
Click on "run test with array with 200000 tuples"

Chrome: https://share.firefox.dev/4jaEANF (1.3s)
Nightly: https://share.firefox.dev/3W27fKB (15s)

Summary: Nightly us 12x slower on a js demo at https://manueldois.github.io/Random-tests/Array%20hashing/dist/index.html → Nightly is 12x slower on a js demo at https://manueldois.github.io/Random-tests/Array%20hashing/dist/index.html with BigInt

This is basically spending all of its time under JS::BigInt::mul...

Depends on: 1940058

The patch from bug 1940058 should help a bit. Further improvements need more advanced integer multiplication algorithms (Karatsuba, Schönhage–Strassen, etc.) than the simple O(n^2) algorithm we currently use. FWIW the largest BigInt in this benchmark has 9105 BigInt digits (on 64-bit).

Severity: -- → S3
Priority: -- → P3

Latest profile:https://share.firefox.dev/3PuxCFp (7.5s)
So we get 50% improvement from bug 1940058 (same as what anba tested earlier)

Nightly: https://share.firefox.dev/45ZgDCP (7.3s)
Chrome: https://share.firefox.dev/4p38bew (1.3s)

Summary: Nightly is 12x slower on a js demo at https://manueldois.github.io/Random-tests/Array%20hashing/dist/index.html with BigInt → Nightly is 5.5x (Was:12x) slower on a js demo at https://manueldois.github.io/Random-tests/Array%20hashing/dist/index.html with BigInt
You need to log in before you can comment on or make changes to this bug.