Javascript Performance Thread

[streetwolf]

To be honest I haven't run any of these benchmarks for a long time until I saw that IM hit the hourlies and I just wanted to see what I'd get. I know they are lower than the last time I ran them but a lot has happened since then and now. And really, do these benchmarks really mean all that much?

[mikeclackler]

a link to some test results for sunspider, kraken, v8, and octane between firefox before/after ion, current safari and current chrome canary

http://pastebin.mozilla.org/1816714

[Omega X]

dvander blog post: https://blog.mozilla.org/javascript/201 ... irefox-18/

[HeinzDo]

With last hourly, i got a error by RayTrace on V8 Benchmark Suite - version 7.

[herom]

JM+TI+ION=JM+TI now?
that's a real regression!

[KWierso]

JM+TI+ION=JM+TI now?
that's a real regression!

Erm, being on par or better than what it was without Ion while not even coming close to being done optimizing things is a regression?

[herom]

JM+TI+ION=JM+TI now?
that's a real regression!

Erm, being on par or better than what it was without Ion while not even coming close to being done optimizing things is a regression?

I was talking about the recent regression as it was on arewefastyet.com
but they got it fixed quickly, so... never mind.

[Tametomo]

I was talking about the recent regression as it was on arewefastyet.com
but they got it fixed quickly, so... never mind.

Well, for the 32-bit machine. The 64-bit machine thinks that JM+TI is JM+TI+Ion at the moment.

Probably will get fixed soon enough anyways.

[phuzi0n]

It looks like it's just a bug with AWFY running the wrong engine now that it has landed on mc.

[iamhere]

Machines 8 & 9 still broken :/. Machine 8 doubly so (it has stopped recording Firefox engines and I'm guessing it will have the same issue as Machine 9 when fixed.

Machine 8
http://arewefastyet.com/?view=regress&machine=8
http://arewefastyet.com/?runs=650&view= ... &machine=8

Machine 9
http://arewefastyet.com/?view=regress&machine=9
http://arewefastyet.com/?runs=650&view= ... &machine=9

[jandem]

Machines 8 & 9 still broken :/. Machine 8 doubly so (it has stopped recording Firefox engines and I'm guessing it will have the same issue as Machine 9 when fixed.

Yeah, but 11 & 12 were bought to replace 8 & 9 so I don't expect them to be fixed anytime soon...

[Tametomo]

Looks like parts 2 and 3 of bug 786126 probably have caused some serious regressions to misc-bugs-612930-solve-sudoku and misc-bugs-608733-trace-compiler, as well as minor regressions to ss-raytrace and kraken-crypto-pbkdf2 which are showing up on the 64-bit AWFY machine.

Not that I think that the javascript developers check this thread too much, and they might notice it on their own, but still good to give a heads up just in case, since sometimes these regressions that occurred aren't immediately obvious at times (especially when there's more than one at once, and one ends up getting forgotten).

EDIT: It also appears that there was a further regression as well which might have been caused when fixing some security bugs by Nicolas B. Pierron (somewhere in bugs 794286, 787848, 787309, and 792220). They've regressed v8 rather significantly on DeltaBlue, EarleyBoyer and RayTrace, as well as kraken-crypto-ccm and kraken-crypto-pbkdf2 again.

EDIT 2: The regressions in the last edit have been fixed from backing all of those bugs out, but the original regressions still stand, it looks like.

EDIT 3: All regressions fixed now.

[jiangiang]

v8 take the crown from ionMonkey in kraken test

http://arewefastyet.com/?a=b&view=regress

[JayhawksRock]

v8 take the crown from ionMonkey in kraken test

http://arewefastyet.com/?a=b&view=regress

OK.. so what does that mean to me while I am reading emails or browsing this forum?

[KWierso]

v8 take the crown from ionMonkey in kraken test

http://arewefastyet.com/?a=b&view=regress

20 ms faster? Oh the humanity!