Thanks for the clarification. Regards, Jeff -----Original Message----- From: Tomohiro Kusumi [mailto:kusumi.tomohiro@xxxxxxxxx] Sent: Wednesday, September 6, 2017 10:56 AM To: Jeff Furlong <jeff.furlong@xxxxxxx> Cc: fio@xxxxxxxxxxxxxxx Subject: Re: fio patch question on filesetup: align layout buffer 2017-08-31 1:27 GMT+03:00 Jeff Furlong <jeff.furlong@xxxxxxx>: > Hi All, > I see the patch at http://git.kernel.dk/cgit/fio/commit/?id=8c43ba6259409244de64f0b1d388a029e38cf3e5 and just wondered if it is still compatible with: > > .. option:: iomem_align=int, mem_align=int > > This indicates the memory alignment of the I/O memory buffers. Note that > the given alignment is applied to the first I/O unit buffer, if using > :option:`iodepth` the alignment of the following buffers are given by the > :option:`bs` used. In other words, if using a :option:`bs` that is a > multiple of the page sized in the system, all buffers will be aligned to > this value. If using a :option:`bs` that is not page aligned, the alignment > of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and > :option:`bs` used. > > For example, if we set mem_align to something larger than page_size or bs, what does fio_memalign use? And if we set mem_align to something between page_size and bs, what does fio_memalign use? Thanks. mem_align wasn't used in the layout thing, and never needed. fio_memalign() is an alternative for posix_memalign() rather than something to do with that option. fio_memalign() takes an alignment by argument, and the argument for this was page_size. page_size works for direct i/o, as it meets sector size alignment (at least for 512, 2K, 4K on all supported archs), which is what Linux and many others require. Btw, I'm not completely sure why fio_memalign() allocates doubled `size'. Some of the aligned malloc implementations I've seen did take of size*2 or alignment*2, but those didn't explicitly take the footer (or header) size, whereas in fio it explicitly has sizeof(*f) after the data size and max possible space for alignment. Since f->offset is always smaller than the alignment size, it seems to me size*1 is enough, but I might be missing something. > > Regards, > Jeff > > -- > To unsubscribe from this list: send the line "unsubscribe fio" in the > body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at > http://vger.kernel.org/majordomo-info.html ��.n��������+%������w��{.n�������^n�r������&��z�ޗ�zf���h���~����������_��+v���)ߣ�
