Hi Wesley, On Mon, Jan 10, 2011 at 23:33:39, Wesley J. Landaker wrote: > Hi, > > First of all, thanks for the response! > > On Mon, January 10, 2011 06:21, Nori, Sekhar wrote: > > On Fri, Jan 07, 2011 at 02:00:05, Wesley J. Landaker wrote: > > The embedded software forum on e2e.ti.com is another place > > to ask questions on TI software. > > > > http://e2e.ti.com/support/embedded/f/354.aspx > > Thanks for the pointers. I've posted this same query in the OMAP forum, and > might try this form as well if I can't get an answer.
There should be no need to duplicate. Folks on the OMAP forum would move the post to the embedded software forum if they feel the question is better answered there. > > > I just tested SATA with an OMAP-L138 board. The base board is > > 1016660 Rev A board with a 1016841 Rev B SoM. With a Foxconn > > E124936 SATA cable connected to WD5001AALS 500 GB hard drive > > > > I used TI PSP release 03.20.00.14 for this test. > > > > Performance data on SATA is available here: > > > > http://processors.wiki.ti.com/index.php/DaVinci_PSP_03.20.00.14_Device_Driver_Features_and_Performance_Guide#SATA > > I've seen that page, and I get similar results. The main reason the folks > I'm working with asked me to help them look into things is that they are > under the impression that the OMAP-L138 should be able to easily do 50 MB/s. > > For instance, in this thread, a TI employee ("clam") forwarded a claim that > raw SATA tests on the OMAP-L138 could do 120 MB/s. Yes, I think that's correct. SATA driver for TI's DSP/BIOS for the pin-compatible DSP-only chip C6748 is able to do 120 MBps on read and 84 MBps on write. Of course, the two operating systems are very different; DSP/BIOS being more lightweight. Also, I guess buffer copies would be avoided in DSP/BIOS. Please see the DSP/BIOS driver's datasheet here: http://software-dl.ti.com/dsps/dsps_public_sw/psp/BIOSPSP/01_30_01/content/C6748_BIOSPSP_Datasheet.pdf > > http://e2e.ti.com/support/dsp/sitara_arm174_microprocessors/f/47/p/57324/231489.aspx > > Since the wiki (and our data) only shows about 25 MB/s, only a fraction of > that > speed, the my thought was that I would track down the bottleneck > (hypothesizing > that it may be in the Linux kernel driver, for example). > > Unfortunately, I have scoured the internet and have only found claims of > people > getting the similar 25 MB/s number ... I've yet to see anyone actually claim > that they have gotten sustained SATA write performance better than that, which > makes me wonder if it's not a Linux kernel problem, but just a limitation of > the device. I think it is a Linux kernel limitation. > > Hence my questions. =) > > >> * Are there any tricks or gotchas in using SATA with the OMAPL-138 that > >> I should watch out for? > > > > I believe some older LogicPD EVMs had some issues with SATA. These > > have been addressed in newer versions. The board I used is a new > > board. You can get details of the fixes made from Logic or you can > > also contact TI's OMAP-L138 hardware support here: > > http://e2e.ti.com/support/dsp/omap_applications_processors/f/42.aspx > > Thanks for mentioning this. I'll have to look into this to see if this is > contributing to our stability/functionality problem, which I think is possibly > unrelated to our throughput issue. > > > The platform AHCI patches have not been tested on OMAP-L138 > > AFAIK. I have started poking around to see if I can get > > platform AHCI working on OMAP-L138 quickly, but I cant give > > you an ETA at the moment. > > My main interest at the moment would only be if there platform AHCI was known > to give better performance, i.e. if it fixed a previously known bottleneck. > From my own poking around this doesn't seem to be the case, but I'd be happy > to be wrong. Okay I got basic SATA (without port-multiplier support) working on 2.6.37 with platform AHCI driver today. The patches aren't in a shape to share on the list yet. I can forward those to you separately for testing if you want to see if platform AHCI improves performance. Like you I am skeptical if this will really help. Using zero copy mechanisms like sendfile() might help more - but that largely depends on application design. > > >> Second, I can only get -- in the > >> very > >> best case, after tweaking kernel IO schedulers, trying different > >> filesystems > >> vs. raw device access, etc -- is around 25 MB/s sustained write > >> performance. > >> This seems like a common performance number I've seen reported by various > >> people, but it's incredibly slow compared to what SATA should be capable > >> of. We > >> need more like 50 MB/s or more, which is still quite conservative from a > >> SATA > >> perspective. > > > > Hmm, no idea about this one. This seems in line with what is reported in > > datasheets. I assume you are using 456MHz version of the device. > > Yes, that's right. BTW, you mention "datasheets". I wasn't able to find any > peformance number claims other than the FAT and ext2 performance numbers on > the > TI wiki. Is/are there a different datasheet(s) that also has SATA performance > numbers? Or is the wiki what you are refering to? I was referring to the same "Features and performance guide" wiki page. > > > I would suggest getting a recent LogicPD EVM if you are trying SATA. > > Also, I think the SATA cable has a role to play in getting a stable > > connection. > > I've tried lots of SATA cables, but I'll look into if we have old LogicPD > EVMs. > I got handed about 5 boxes of EVMs and a load of extra SOMs when I said I'd > help get to the bottom of this. I suppose it's possible that they are all > "old" > EVMs. > > But if even with a new EVM, or even a different board, is there any reason you > know of to suspect that I'd get more than the max sustained write performance > we saw previously and that is cited on the wiki, at around ~25 MB/s? I don't think changing the EVM will change performance. Thanks, Sekhar _______________________________________________ Davinci-linux-open-source mailing list [email protected] http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
