On 2023-04-20 16:19:41 +0200, to...@tuxteam.de wrote: > On Thu, Apr 20, 2023 at 04:13:48PM +0200, Vincent Lefevre wrote: > > [...] > > > "marginally faster" is incorrect, but this probably depends on > > how fast the disk is. For the MPFR svn-to-git conversion with > > reposurgeon 2 years ago, I had to use /dev/shm (tmpfs) because > > it was awfully slow on /tmp (but the SSD disk was rather slow: > > the new one I got several months later was about 50 times as > > fast, IIRC). > > Hm. Interesting. I'd expected the buffer cache to paper over the > difference (provided you have enough RAM for that, but since we > are comparing that to putting things in tmpfs, by definition, you > have :)
I think that the issue is that even though there is a buffer cache, data are written to disk, even when this is not needed for really temporary data. These write operations make the system slow when it tries to read other data from the disk. -- Vincent Lefèvre <vinc...@vinc17.net> - Web: <https://www.vinc17.net/> 100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/> Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)
