need to measure the effect of ICF (identical code folding) for Chrome on ChromeOS |
|||
Issue descriptionWe need to measure the effect of ICF on Chrome for ChromeOS. Since we are using PIE, the effect may not be much. Sri has some ideas on how to improve ICF in combination with PIE. Lets see how much are we getting now. Also, why hasn't anyone complained about the possible stack traces issue?
,
Apr 18 2016
BTW, I think we have ICF and PIE turned on for Chrome Linux.
,
Apr 18 2016
has anyone complained about stack traces issues on Chrome for Linux?
,
Apr 19 2016
Probably here and there, but I didn't find any examples when querying crbug.com for: OS=Linux ICF backtrace. Also tried s/backtrace/stacktrace/
,
Apr 19 2016
Some data about Chrome size:
with icf without icf
on arm: 75277531 80324343
on x86_64 105057474 112265666
,
Apr 19 2016
I assume this is with --icf=safe right? Would you mind measuring with --icf=all too if it is not too much trouble? Thanks Sri
,
Apr 19 2016
This is --icf=all. Do you need the data for --icf=safe?
,
Apr 19 2016
Oh ok. No, that is fine. --icf=all is the best you can get. Thanks for clarifying that. Sri
,
Apr 19 2016
do you want the data without PIE?
,
Apr 19 2016
No, thanks. The reason I asked originally is because I suspected you are using --icf=safe which does not give good code size reductions with PIE. If you are using --icf=all you are getting the best from ICF. Thanks Sri
,
Apr 21 2016
Close this?
,
May 23 2016
Bulk verified
,
May 23 2016
bulk verified |
|||
►
Sign in to add a comment |
|||
Comment 1 by llozano@chromium.org
, Apr 18 2016Status: Assigned (was: Untriaged)