This is the mail archive of the binutils@sourceware.org mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [patch] Use unaligned access on x86_64


ping.

On 27 April 2015 at 15:09, Rafael EspÃndola <rafael.espindola@gmail.com> wrote:
> I tested this on my linux workstation with a debug build of clang. The
> results are still very interesting.
>
> The perf logs are attached.
>
> Cheers,
> Rafael
>
>
> On 24 April 2015 at 10:54, Rafael EspÃndola <rafael.espindola@gmail.com> wrote:
>> Hi,
>>
>> x86_64 has exquisite support for unaligned loads. It is a shame not to use it.
>>
>> The attached patch avoids aligning archive members on x86_64. The
>> results when linking clang are very interesting:
>>
>> * massif reports that the malloc memory usage goes from 331,295,192
>> bytes to just 133,415,136 bytes.
>>
>> * the linking time (30 runs average) goes from
>>
>> 1.310065610 seconds time elapsed ( +-  0.19% )
>>
>> to
>>
>> 1.162564763 seconds time elapsed ( +-  0.14% )
>>
>>
>> Cheers,
>> Rafael
>>
>>
>> 2015-04-24  Rafael Ãvila de EspÃndola <rafael.espindola@gmail.com>
>>
>> * fileread.cc (find_or_make_view): Don't align on x86_64.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]