On Thu, Mar 22, 2012 at 2:50 AM, Kevin Kofler <kevin.kofler@xxxxxxxxx> wrote: > Chris Tyler wrote: >> On Thu, 2012-03-22 at 02:38 +0100, Kevin Kofler wrote: >>> And finally, for our build speed issue, the practical consideration will >>> be whether the parallelism will actually speed our builds up. Right now >>> our builds are either serial or have portions parallelized with "make >>> -j", which assumes a single multi-CPU computer (but the "multi-core" ARM >>> setups actually present themselves as a multi-computer cluster, which is >>> not supported by "make -j", not as a multi-CPU computer), so the >>> parallelism does little to the latency of an individual build (though of >>> course it does help the overall throughput). >> >> Actually, there's both: ARM scales to multiple cores per CPU (dual-core >> and quad-core are common, and very high core counts are on the horizon), >> and vendors are preparing many-CPU boxes (e.g., HP Redstone, with 288 >> quad core (+1 management core) systems in 4U). > > But there are x86 CPUs with more than 4 cores, and multi-CPU SMP systems > which still present themselves as one (multi-CPU/core) computer. IIRC, our > x86 Koji builders have 16 cores per machine (might be even more by now, not > sure). But as you said yourself in an earlier thread a lot of compilation isn't massively parallel so massive amount of cores for building isn't necessarily as much a win as pure GHz. On that front the current A15 gen which is arriving now easily does the 2.5 - 3 ghz that the intel platforms do (yes, I know they go to 3.6 but they're not regularly used primarily due to heat) and most of the recent wins on Intel architecture has been for media related things through various SSE versions and other offload functions for things like crypto all of which aren't massively used in standard compilation and all of which have similar functionality on ARM platforms. Peter -- devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/devel
