Hi Frank, You results are amazing! What kind of patch have you applied? I'm still waiting for my USB-HDD to arrive. I bet your patch will make my nightmare go away. could I try it?
Best regards, Ashuaria Lee 2008/6/3, Frank Xue <[EMAIL PROTECTED]>: > Swami, Lee, > I tested performance using 3 different USB devices (2 USB HDD and 1 > USB Flash Disk). > Here is the performance test result. > > Instructions for testing > ---------------------------------- > writing test: > time dd if=/dev/zero of=/mnt/test.gz bs=1M count=500 && sync > reading test: > time dd if=/mnt/test.gz of=/dev/null bs=1M && sync > > Result: > Device Write Read > ----------------------------------------------------- > USB HDD1 ~9.5MBps ~13.0MBps > USB HDD2 ~7.2MBps ~8.0MBps > USB FD ~3.8MBps ~9.0MB/s > > It seems the performance is pretty good. > I made some minor changes and applied the patch to git 2.6.23, this patch > also works > well on git 2.6.23. > > > Regards > > Frank > > Ashuaria Lee wrote: > > > Hi Subbrathnam, > > > > The only difference I feel is that I used a thumb USB Memory, which is > > SanDisk cruzer micro 1.0GB, and you used USB HDD. > > > > I can't understand Why? my read speed is so high, and why the write > > speed so low. > > > > I'll try the USB hdd when I get one. > > > > Good Luck. > > > > 2008/5/31, Subbrathnam, Swaminathan <[EMAIL PROTECTED]>: > > > > > > > Lee, > > > > > > On the TI DaVinci EVM I am consistently getting around 4-6MBps > throughput writing to the USB HDD's under DMA mode. > > > > > > Many more MUSB fixes need to be done still compared to the 2.6.10 > kernel MUSB driver. I am planning to do the same gradually as seperate > patches. > > > > > > Any way I can reproduce your test setup ? > > > > > > swami > > > > > > ________________________________________ > > > From: Frank Xue [EMAIL PROTECTED] > > > Sent: Wednesday, May 28, 2008 8:24 AM > > > To: Ashuaria Lee > > > Cc: Subbrathnam, Swaminathan; davinci-linux-open-source > > > Subject: Re: USB CPPI DMA patch for DaVinci > > > > > > > > > Lee, Swami, > > > > > > I did a simple and quick test. The kernel was also configured with > > > default setting. > > > It seems this patch fixed USB stability issue. > > > > > > test case: > > > ~ 20 files, about 100MBytes every file. > > > > > > Speed benchmark result on git 2.6.25 > > > --------------------------------------------- > > > Read: 3.6M Bytes/sec > > > Write: 2.8M Bytes/sec > > > > > > Lee, I almost could not believe the writing performance is so poor ( > > > 0.38MB/sec) on your board. > > > :( > > > > > > Regards > > > > > > Frank > > > > > > Ashuaria Lee wrote: > > > > Hi, > > > > > > > > You patch DID work with the stability, and didn't crash using DMA > mode. > > > > And Also I got more speed on reading data from USB-storage. > > > > The reading performance has increased significantly. > > > > But, the write speed has decreased. > > > > > > > > Let me add the screenshot of each results. > > > > > > > > The short summary is > > > > ============================================ > > > > MVL 2.6.10 with patch 45 | GIT 2.6.25 with your DMA patch > > > > Read: 8.78MB/sec | 12MB/sec > > > > Write: 3.7MB/sec | 0.38MB/sec > > > > ============================================ > > > > > > > > The GIT was configured with default setting and only changed to use > DMA for USB. > > > > did I do something wrong? > > > > > > > > Any opinions? > > > > > > > > Best Regards, > > > > Ashuaria Lee > > > > > > > > > > > > > > > > > > > > > > > > > > > -- Regards, Ashuaria Lee _______________________________________________ Davinci-linux-open-source mailing list [email protected] http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
