We just got around to trying with both F_RDAHEAD and F_NOCACHE. They seem to make no difference at all on the one system we tested. http://pastebin.com/ZKjVA94t
I also just noted that your successful test was on "External 3GB WD Reds on USB 3 in raid 0" - what filesystem was on the drives? Do I assume you have no spinner that isn't part of a RAID? On Thu, Mar 5, 2015 at 8:48 PM, Hamilton Feltman <[email protected]> wrote: > > > We don't disable caching on other platforms. And unless you modified > run-readtest.sh, we aren't disabling it on OS X either in the test (that > requires an extra -D flag to be passed to readtest). > > > Caching was disabled, but I had tried a lot of different tests, one of the > was to remove lookahead reads and inadvertently it’s still in there. So > this explains the more ‘deterministic’ results. But I think you’re going to > need this also, since what you’re doing is writing your own file cache, and > you don’t want the filesystem to think it knows better. > > if (fcntl(fd, F_RDAHEAD, 0) == -1) { > fprintf (stderr, "Cannot set F_RDAHEAD on file #%d\n", n); > } > > Then a message shows up saying "disk system cannot keep up". Then you > start digging ... and digging .. and digging ... and you find that for 8 > seconds the disk was wasn't managing more than 5MB/sec. And eventually you > end up with readtest.c :) > > > Ok, that clarifies the issue drastically. One thing to check is possibly > something else hitting the disk, like spotlight. Anything else touching the > disk at that seekrate, and the bandwidth is destroyed. You can disable > spotlight for that drive (System preferences -> Spotlight -> Privacy). If > it’s not spotlight, a utility called fseventer will show system wide file > access, but it doesn’t look like mavericks is supported yet. But there has > got to be a command line util that can watch fsevent. > > Something to try anyway. > > >
_______________________________________________ Do not post admin requests to the list. They will be ignored. Coreaudio-api mailing list ([email protected]) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/coreaudio-api/archive%40mail-archive.com This email sent to [email protected]
