On Mon, 17 Jul 2006 23:28:40 +0200, Christian MICHON wrote: > On 7/17/06, Fabrice Bellard <[EMAIL PROTECTED]> wrote: >> Sometimes I try :-) But evaluating patches takes a large amount of time >> and in a program like QEMU regressions are difficult to track and very >> easy to introduce... >> >> > not all on this list do contribute patches (in my case, not much > contribution at this level). But I believe we're tons of testers of > incoming patches :) > > you're not alone to test these patches: this is what I mean... > > I remember the asynchronous/dma patch could be also a good candidate for > insertion, if it could be updated...
I actually think the last dma patch I saw was addressing the problem in the wrong way. IIRC, it was using threads? We really ought to do the work to make the block API properly asynchronous. It may seem tedious now, but it would also allow the use of something like linux-aio down the road which has the potential to have a pretty huge performance impact on block performance (with pbrook's recent SCSI emulation). Regards, Anthony Liguori _______________________________________________ Qemu-devel mailing list Qemu-devel@nongnu.org http://lists.nongnu.org/mailman/listinfo/qemu-devel
