[Webkit-unassigned] [Bug 144039] [EFL] Enable FTL JIT by default on AArch64

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=144039

--- Comment #7 from Yusuke Suzuki <utatane.tea at gmail.com> ---
(In reply to comment #6)
> Update: There are a few crashes/failures with enabled FTL JIT on AArch64
> Linux: (tested on r213461)

Maybe some of them can be fixed since they are not AArch64 specific thing I think.
But others are a bit difficult...

> 
> stress/array-concat-spread-object.js.ftl-no-cjit-validate-sampling-profiler:
> Exception: failed normally with an object
> stress/array-concat-spread-object.js.ftl-no-cjit-validate-sampling-profiler:
> ERROR: Unexpected exit code: 3

This should be related to https://bugs.webkit.org/show_bug.cgi?id=153704.
So it should be debugged even in x86_64 environment. I'll check this this weekend.

> stress/class-syntax-super-in-eval.js.ftl-eager-no-cjit: Segmentation fault
> stress/class-syntax-super-in-eval.js.ftl-eager-no-cjit: ERROR: Unexpected
> exit code: 139

I think it's difficult to debug without AArch64 machine.
Can anyone investigate this issue?

> microbenchmarks/generator-with-several-types.js.ftl-eager-no-cjit: Timed out
> after 207.000000 sec seconds!
> microbenchmarks/generator-with-several-types.js.ftl-eager-no-cjit:
> Segmentation fault
> microbenchmarks/generator-with-several-types.js.ftl-eager-no-cjit: ERROR:
> Unexpected exit code: 139

I guess this is pure timeout. Can anyone investigate it?
If so, just moving it to slowMicrobenchmarks directory solves the problem.


> microbenchmarks/varargs-call.js.ftl-no-cjit-validate-sampling-profiler:
> Segmentation fault
> microbenchmarks/varargs-call.js.ftl-no-cjit-validate-sampling-profiler:
> ERROR: Unexpected exit code: 139

And I think this is also related to https://bugs.webkit.org/show_bug.cgi?id=153704.
I believe varargs forwarding in FTL have some bugs related to SP update & signal stack.

> 
> 
> If there is any volunteer to check and fix these crashes, we could enable 
> it in the near future. Unfortunately I don't and won't have time to do
> it myself.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20170307/f5374040/attachment.html>