Hi Bernard, Yes the 4 GB/s is a mistake ... by 1000x ;) I meant to write 4 MB/s.
I should look at porting the git kernel to DM355. Using 2.6.10 drivers feels old-fashioned at times. If it's been ported to DM6446, and the MV kernels have the DM6446/DM355 register differences coded, than I guess it shouldn't be huge amount of work. -- Pete On Wed, Apr 23, 2008 at 10:24 PM, Bernard Blackham <[EMAIL PROTECTED]> wrote: > > I am using a kernel retrieved from TI's site on Apr 20/08. I built the > > kernel with the only change to the config being the inclusion of the > > resizer driver. > > > > On Windows Vista, my 150x 2GB SD card gets a write speed of > 4 GB/s > > over a 128 MB file. > > It's hard to believe those numbers are correct. You're probably hitting the > disk cache at 4GB/s with your 128MB file (and then not even that fast - > there are probably gross rounding errors because the actual time was so > small). It doesn't actually hit the SD card at this speed. The SD card speed > can't be more than 25MB/s (4-bit interface * 50MHz clock), and it's even > less than that due to SD being a high latency protocol. In reality, it's > closer to 20MB/s if everything was running perfectly optimally. > > > > On the embedded system my write speed is 135 kB/s - an unbelievable 3%!! > > Well, yes that is pretty poor. At least on the DM6446, the git kernel has a > different driver with which I was getting speeds of around 2-3 MB/s for > write, and a 5-6 MB/s on reads (at the 2.4.24 branch). > > > > Has anyone looked at the SD driver and/or SD interface? I was not > > planning on touching this driver, never mind having to rewrite big > > chunks of it :( > > I don't know if anybody is using the git kernel on the DM355. But I'm > willing to bet that it would be less effort to port git to DM355 than the SD > driver to MV's 2.6.10. > > Cheers, > Bernard. > > > _______________________________________________ Davinci-linux-open-source mailing list [email protected] http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
