From mboxrd@z Thu Jan 1 00:00:00 1970 From: Stephen Warren Date: Tue, 20 Oct 2015 14:40:12 -0600 Subject: [U-Boot] [PATCH V3 05/10] fat: ff: read max contiguous file data In-Reply-To: <20151020191021.GS23893@bill-the-cat> References: <1443765973-5897-1-git-send-email-swarren@wwwdotorg.org> <1443765973-5897-5-git-send-email-swarren@wwwdotorg.org> <56268E9C.6080803@wwwdotorg.org> <20151020191021.GS23893@bill-the-cat> Message-ID: <5626A6AC.6040105@wwwdotorg.org> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de On 10/20/2015 01:10 PM, Tom Rini wrote: > On Tue, Oct 20, 2015 at 12:57:32PM -0600, Stephen Warren wrote: >> On 10/02/2015 12:06 AM, Stephen Warren wrote: >>> Enhance f_read() to find the maximum contiguous set of clusters to read, >>> and read it all at once (which is fast) rather one by one (which is >>> slow). >> >> Hmm. I had hoped that the author of ff.c would accept this patch >> upstream, so we could pick up a later upstream version that included >> this patch. However, it seems the author of ff.c has a policy of not >> accepting outside contributions: >> >> http://elm-chan.org/fsw/ff/bd/?show=2472 >> (That's a link to the author's reply to my patch, on the forum >> system associated with his/her SW) > > The bit about the license is at > http://elm-chan.org/fsw/ff/en/appnote.html#license > >> I wonder how much of a liability incorporating ff.c into U-Boot will >> be, if we can't ever get any fixes merged upstream. Perhaps we just >> fork it, although I had hoped we'd be able to keep picking up new >> versions. > > Arg, that really does take away one of the potential nice features. I > guess, sadly, at this point I'd rather stick with the version we have > unless you want to deal with re-syncing their releases but still > effectively doing a fork (so that we can also make use of caches which I > think you said before you thought might be part of the performance > problem. FWIW, the caching (of FAT bitmap, not CPU memory) turned out not to be an issue at all; it was just my wild conjecture before I'd investigated the read performance issues.