[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Changing the default 32-bit x86 arch for Fedora 12



On Tue, Jun 16, 2009 at 9:07 PM, Josh Boyer wrote:
On Tue, Jun 16, 2009 at 08:16:36PM -0400, Orcan Ogetbil wrote:
>
>- Let's keep F-12 the same: ppc, ppc64, i586, x86_64
>- Since ppc and ppc64 are going to be dropped from F-13, fill in the
>blank spot with i686+SSE2, i.e. F-13: i586, i686+SSE2, x86_64
>
>Everyone happy?

No.  But that isn't the point.

Someone else ask what the real benefit to moving to i686+SSE2 is.
I haven't seen overwhelming evidence that a huge benefit exists.
I think somone is working on gathering more data, but unless it shows
massive gains (1-2% is not massive), then I don't see why we'd change
anything at all.

I'm sure I'm missing something here.

I want to learn more about this SSE/SSE2 business.  Some libraries and applications, especially audio or video related ones, contain assembly code to enable what is called "SSE(2) optimizations". Sometimes this code isn't effective on hardware that doesn't support it. Sometimes there is no runtime detection, and the SSE(2) optimized program fails on unsupported hardware.

The thing is, usually these kind of programs run so slow on older hardware, particularly when dealing with low latency recording, that it doesn't make sense to disable SSE(2) optimizations even if there is no runtime detection. So certain old hardware gets discarded in this case.

Now where does the "i686+SSE2" come into play? Does this SSE2 have any effect on those programs that do not contain SSE(2) related assembly code? Is this 1-2% improvement that you are mentioning only about these kind of programs (that do not contain assembly code)?

Please enlighten me, or point me to some reading.

Thanks,
Orcan

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]