Am 08.11.2015 um 00:56 schrieb Kevin Kofler:
Reindl Harald wrote:if there would be no difference kernel upstream won't invest that much time for runtime-cpu-detection (look at the bootlog on different hardware)Exactly BECAUSE performance-critical software normally does runtime detection (except where upstream is really really lazy), hardcoding a requirement for newer instruction sets does not buy us anything.
define performance criticalour cms systems got 5 times faster within 6 months by micro-optimizig each alone would not had any effect
The absence of a portable fallback also makes the software unusable on our secondary architectures, and in some cases (such as Darktable), even on the PRIMARY architecture ARM. (Now, whether ARM should be a primary architecture to begin with is a different matter, but it currently is.) So I really don't understand those upstreams that provide ONLY platform- specific vector code with no portable version
i understand them well - not endless time and write effecient software without test on dozens of environments older than 10 years - as C programmer if i would write new code these days i would even go so far and rely on AVX
yes, that can't be done for a distribution, but i know that all machines sorrounding me are SandyBridge bases, even the network routers for small offices and so why would i care about anything else?
signature.asc
Description: OpenPGP digital signature
-- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
