This is the mail archive of the
mailing list for the GCC project.
Re: Type inheritance graph analysis & speculative devirtualization, part 7/7 (speculative devirtualizatoin)
- From: Jan Hubicka <hubicka at ucw dot cz>
- To: Taras Glek <tglek at mozilla dot com>
- Cc: Jan Hubicka <hubicka at ucw dot cz>, gcc-patches at gcc dot gnu dot org, jason at redhat dot com, Nathan Froyd <froydnj at mozilla dot com>, glandium at mozilla dot com, marxin dot liska at gmail dot com
- Date: Tue, 3 Sep 2013 18:24:27 +0200
- Subject: Re: Type inheritance graph analysis & speculative devirtualization, part 7/7 (speculative devirtualizatoin)
- Authentication-results: sourceware.org; auth=none
- References: <20130901135714 dot GA23527 at kam dot mff dot cuni dot cz> <52260602 dot 5090500 at mozilla dot com>
> What is the footprint impact of speculative devirtualization?
It is less than 2% of text section and once we solve problems with ipa-prop tracking,
I hope it will be less.
I hope we now understand better how to devirtualize and I think we can improve non-speculative
> Firefox is sensitive to footprint increases, especially on Android startup.
That is what Martin's patch is expected to solve. With his ordering only
fraction of the text and rodata section needs to be preloaded. We are still
chasing some issues there, but it seems that "only" about 10MB is needed
(out of 60MB binary). Not accounting dynamic linking of data.rel.ro that
adds around another 5MB. It depends on FDO though.