Re: [gentoo-user] Re: Firefox-10.0.1 fails to compile on x86

[Michael Mol]

On Thu, Feb 23, 2012 at 2:55 PM, Nikos Chantziaras <realnc@×××××.de> wrote:
> On 23/02/12 21:42, Michael Mol wrote:
>>
>> On Thu, Feb 23, 2012 at 2:36 PM, Nikos Chantziaras<realnc@×××××.de>
>>  wrote:
>>>
>>> On 23/02/12 12:44, Mick wrote:
>>>>
>>>> The irony is that older boxen which would benefit most from building
>>>> from
>>>> source are constrained in resources to achieve this and have to resort
>>>> to
>>>> installing bin packages.
>>>
>>>
>>> I doubt that the bin package will be slower than the one compiled from
>>> source.  I predict the reverse, in fact.  The bin package will perform
>>> better.
>>
>>
>> That seems a strange prediction. What drives that hunch?
>
>
> The PGO optimized build that Mozilla is shipping.  You can also build with
> PGO from source, but that means building FF *twice* in a row (by enabling
> the "pgo" USE flag).  I doubt that with the old laptop anyone is building FF
> twice with PGO, and that means that the -bin package should be faster.
>
> Furthermore, FF is build using its own CFLAGS.  They are the same in the
> source build as well as in the -bin package.  The only difference is
> probably the -march option.  And that doesn't make much difference to begin
> with (after -march=i686, gains are very minimal).

I knew and forgot about PGO, but I didn't realize there was a USE flag
for it. Neat. I'll be enabling that.

I disagree with the idea that keeping things down around -march=i686
provides only minimal gains. SSE and SSE2 instructions carry a big
benefit for numerical operations, especially those which can be
parallelized, but not enough to justify batching into a GPU. AVX will
be adding operations which allow more useful and flexible use of
registers. Simply bumping up to x86-64 from simple x86 doubles your
GPRs, which gives the compiler all kinds of room to work with.

If the combination of those things doesn't significantly benefit a
program written in C or C++, then I suspect there's something
dreadfully wrong with the architecture of that codebase.


-- 
:wq