• State
  • Locked Locked
  • Replies 2 replies
  • Subscribers 101 subscribers
  • Views 2168 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

OMAP-L138 MMC/SD Linux driver performance in PSP 03.20.00.14 and PSP-03.21.00.03

Andrea Galbusera
Prodigy 35 points
Other Parts Discussed in Thread: OMAP-L138

Hello,

I'm evaluating the opportunity to move a project I work on, from Linux PSP 03.20.00.14 to the most recent PSP 03.21.00.03.

One point I'm concerned with is the MMC/SD driver performance, since we use SD for both system boot and data acquisition.

By comparing "Device Driver Features and Performance Guide" for both PSP releases (links [1] and [2] respectively) it appears that throughput performance have dropped on the OMAP-L138 by more than 60%. This may perhaps be due to different testing conditions, that do not show up from the table anyway. Since I noticed from the same benchmark pages that other drivers also seem to suffer of performance drops, I'm asking if anyone had the opportunity to compare the two PSP performances, especially with respect to MMC/SD functionality.

If such a degradation is confirmed, is there any known reason for that? Anyway if this is the case, I'd probably drop the idea to migrate to the most recent kernel. 

Thank you in advance,

Regards,

Andrea

Links:

[1] http://processors.wiki.ti.com/index.php/DaVinci_PSP_03.20.00.14_Device_Driver_Features_and_Performance_Guide#Performance_and_Benchmarks_3

[2] http://processors.wiki.ti.com/index.php/DaVinci_PSP_03.21.00.04_Device_Driver_Features_and_Performance_Guide#Performance_and_Benchmarks_3

  • 0
    TI__Expert 5170 points

    Hi Andrea,

    Overall the performance of some of the drivers has come down. This is due to the fact that we have moved from Linux Kernel v2.6.33-rc4 (in 03.20.00.14) to v2.6.37 (in 03.21.00.04). Also earlier we used to calculate the performance by writing zeroes to the device but now we are writing a file created from /dev/random device for read/write. You can use 'dd' command available in most of the Linux file systems to verify the extent to which the performance has varied between the two releases.

    Regards, Sudhakar

  • 0 in reply to Sudhakar Rajashekhara
    Expert 1935 points

    so which is it?  are you saying the newer kernel is causing it to be slower, or that you have changed the benchmark?  any chance of getting an apple-to-apples officially from TI?

    btw, there's an excellent program for benchmarking block media, i recommend it.  gives you a bunch of very useful tests, no need to write your own.  It's called IOZone (www.iozone.org) and builds easily for arm/linux. At least, it did for iMX and NXP ARMs, haven't tried it under the TI toolchain.