[Webkit-unassigned] [Bug 123891] [arm] prepareExecutableCopy should not do direct branches

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=123891





--- Comment #4 from Julien Brianceau <jbriance at cisco.com>  2013-11-06 07:15:05 PST ---
No I didn't run these benchmarks because SunSpider and V8 are crashing with current implementation.

I took subset of SunSpider and V8 to get tests that are not crashing with the current implementation:
    sunspider-1.0/controlflow-recursive.js
    sunspider-1.0/date-format-tofte.js
    sunspider-1.0/date-format-xparb.js
    sunspider-1.0/math-cordic.js
    sunspider-1.0/math-partial-sums.js
    sunspider-1.0/math-spectral-norm.js
    sunspider-1.0/regexp-dna.js
    sunspider-1.0/string-base64.js
    sunspider-1.0/string-fasta.js
    sunspider-1.0/string-tagcloud.js
    sunspider-1.0/string-unpack-code.js
    sunspider-1.0/string-validate-input.js
    v8-v6/v8-deltablue.js
    v8-v6/v8-earley-boyer.js
    v8-v6/v8-raytrace.js
    v8-v6/v8-regexp.js
    v8-v6/v8-richards.js
    v8-v6/v8-splay.js



I run all of them 5 times, and here are the results:
* r158741 without the patch:
    real    0m25.940s
    user    0m24.270s
    sys     0m1.540s

* r158741 with the patch:
    real    0m27.035s
    user    0m25.330s
    sys     0m1.670s


So you're right, there is a slight performance decrease. I'll take a look to check if it's possible to detect and mark those jumps as well.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.