601222 - Blacklist tracejit and methodjit on samsung i9000 devices

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

+++ This bug was initially created as a clone of Bug #600488 +++

Spun off from bug 600596 and bug 600488, which now need to block betaN instead of beta1 (I don't have those powers).

Comment 1

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Attachment: Blacklist JITs on samsung i9000s

On my vibrant, this makes -j testComparisons.js crash go away.  With

    os.environ['JS_IGNORE_JIT_BROKENNESS'] = '1'
    args = [
        JS,
        'basic/testComparisons.js',
        '--show-cmd',
        '--show-output',
        '--avoid-stdio',
        '--jitflags', 'j',
    ]

it comes back.

This is a gross hack, but the precedent is 

                       && JSC::MacroAssemblerX86Common::getSSEState() >=
                          JSC::MacroAssemblerX86Common::HasSSE2

To my knowledge, we don't have a cleaner way of checking the host CPU.

Comment 2

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

http://hg.mozilla.org/mozilla-central/rev/4791cfde3ca0 (relbranch)
http://hg.mozilla.org/mozilla-central/rev/9c2484dac245

Sigh.

Comment 3

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

It struck me while I was walking to dinner that it might have been a good idea to write something to logcat when we disable JIT.  Ah well, maybe can get that in if we respin.

Comment 4

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

It's really too bad that this went in with an env var instead of a pref :(  The jit issue seems fixed with Android 2.2, but it's very difficult to enable the jit now.

Comment 5

[Matt Brubeck (:mbrubeck)]

Maybe we could change the blacklist to also depend on the OS version, or the kernel version.  The leaked Android 2.2 image for Captivate (I assume this is the one that Vlad is testing) uses kernel 2.6.32.9 according to this page (compared to 2.6.29 in the shipping Android 2.1 images):
http://forum.xda-developers.com/showthread.php?t=797282&page=8

Comment 6

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

(In reply to comment #4)
> The
> jit issue seems fixed with Android 2.2, but it's very difficult to enable the
> jit now.

How did you test the JIT?  (It's easy to enable for test with the script here https://wiki.mozilla.org/Mobile/Fennec/Android#JS_trace-tests.)

If it's indeed fixed, yes, we should pass 2.2 through.

Comment 7

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Well, I tested by running sunspider and kraken, and browsing for about 20-30 minutes with no hangs.  I've never been able to browse that long before without issues.  I'll give the tests a try when I get a chance.

Comment 8

[James Slaughter]

The blacklist in the patch doesn't include the string "GT-I9000" which is what my (UK/European, non-carrier-model) Galaxy S has for 'Hardware' according to /proc/cpuinfo. Should it?

Comment 9

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

If there's a significant number of these running 2.1 (eclair), then yes.

Comment 10

[James Slaughter]

Samsung have not yet officially released Froyo firmware for the device; besides those using leaked or community builds, every Galaxy S runs 2.1-update1. Also, note that updates are via Samsung's "Kies" software rather than OTA, so uptake of 2.2 (rumoured to arrive later this month) will be slow. The device is on the majority of UK carriers and is reasonably popular.

Comment 11

[Oldarney Man Slaughter]

Samsung has officially realeased froyo for Galaxy S phones, most carriers in the US have not decided to release it yet. Although fully legit leaked versions of froyo for galaxy s phones are available.