[LLVMdev] LTO question

Eric Christopher echristo at gmail.com
Sun Dec 28 13:08:41 PST 2014


On Sun Dec 28 2014 at 12:17:22 PM Philip Reames <listmail at philipreames.com>
wrote:

> On 12/26/2014 03:46 PM, Xinliang David Li wrote:
>
> Vikram, LLVM can handle Firefox size programs. Honza wrote two good
> articles about LTO.
>
>
> http://hubicka.blogspot.com/2014/04/linktime-optimization-in-gcc-1-brief.html
>
> http://hubicka.blogspot.com/2014/04/linktime-optimization-in-gcc-2-firefox.html
>
> Comparison with LLVM is described in the second article. It took about
> 40min to finish building Firefox with llvm using lto and -g. The
> following is a quote:
>
> "This graph shows issues with debug info memory use. LLVM goes up to
> 35GB. LLVM developers are also working on debug info merging
> improvements (equivalent to what GCC's type merging is) and the
> situation has improved in last two releases until the current shape.
> Older LLVM checkouts happily run out of 60GB memory & 60GB swap on my
> machine.".
>
> One thing I want to highlight in the above is that this is a build with
> debug information.  My understanding from listening in on conversations at
> the developers meeting is that LTO w/o debug information scales
> substantially better.  Most of the discussion of memory usage is specific
> to retaining full debug info while doing LTO.  Either compiling without
> debug info at all, or compiling with only line table information results in
> dramatically reduced memory usage.
>
> Note: I have no direct experience here; I'm just repeating what I've
> heard.
>
>
This is correct, it's also an area of active development. :)

Thanks!

-eric
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20141228/a487f22b/attachment.html>


More information about the llvm-dev mailing list