Front page | perl.perl5.porters |
Postings from April 2013
On Thu, Apr 04, 2013 at 06:37:32AM -0400, bulk 88 wrote: Quoting in full intentionally: > > Date: Thu, 4 Apr 2013 09:46:35 +0100 > > From: nick@ccl4.org > > To: bulk88@hotmail.com > > CC: corion@cpan.org; perl5-porters@perl.org > > Subject: Re: [PATCH] On eliminating external tools from the release process > > > > > > > > On other OSes, prefetching or hinting to the IO scheduler that we are > > > doing sequential reading will serialize and buffer instead of interleave the > > > reads. On others OSes, zero copy DMA to DMA copying is done with > > > kernel help. A for loop through stdio does none of that. > > > > Sure, but in this case, the scripts will immediately block awaiting the > > output, on a local machine which is only being used interactively for this > > task, so disk IO speed is what's going to matter, not how to reduce CPU > > use. > > > Disk IO doesn't matter if it was copied to ram disk cache in 1 ms. Then the > new file is available for reading and writing,, and control is returned back > to me and I can go on, but in the background the disk is trashing for next 3 minutes, > but it doesn't affect anything visually. > > > Ultimately, for the release process, we care about elapsed operator time, > So performance does matter during the build process? You've snipped the part where I said that other parts of the process will consume more time. That is a material part of my point. Optimise the slow parts. This won't be one of them. Nicholas ClarkThread Previous | Thread Next