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.

Just a few observation with the A8 2.xx WEC7 BSP

 

1.      1.  Touch screen driver sets up the debounce setting for the GPIO but never calls the GPIOIOCTL.

2.       2. BSP catalog still has a bunch of codec BspItemId's in it even though these items have been moved out to the 3rd Party catalog.

3.       3. Vfp2fpcrt.dll still included even though I am not sure it is meaningful any more.  Not sure.

4.       4. Object files in binary version of DVSDK gets nuked on clean rebuild.

5.       5. HSUSB host still has suspend/resume issues. Still throws exceptions..mostly does not work. The whole USB power implementation leaves much to be desired.

6.       6. OTG port when configured as USB host only does not enumerate clients when attached at power up. Issuing suspend/resume does re-enumerate clients.  Also get AbortQueue and re-enumerations periodically.  

7.       7. SMSC Ethernet driver was moved to common directory but still has platform specific dependencies of LAN_IRQ_GPIO.  This causes issues when building for platforms that do not need or uses this driver.

 

 

DV

  • David

    Thanks for your inputs. While we have taken care of some of them post 2.0 release, for others I will open defects in our database. Can you please provide some more details on issues:

    1) #2 - not sure which BSPItemId you are refering to.

    2) #5 - Do you see this on TI EVM as well? We did some improvements in 2.0 BSP and didnt see such exceptions in around 900+ iterations of suspend resume.

    3) #6 - what changes do you do to OTG driver for host only mode.

     

    Thanks
       Madhvi

     

  • For #2 just open the BSP catalog in source mode ... you will see them.

    For #5 I am using both Gumstix (3530) and Beagleboard (3730) platforms ... pretty much same issues. I was using both the CE6 and the WEC7 version of the EHCIPDD system.c ... neather with very much luck but the CE6 version was working somewhat better under CE6. Looks like the only difference was the power switching was moved to the single threaded poweroff function for some strange reasion.

    For #6 just just connected highspeed hub (like the old'en days when HSUSB did not work at all) to OTG port. No code changes at all.Both Gumstix and BeagleBoard use 65950's PHY.

     

     

  • A few more observations:

     

     

    With CE6 A8 1.xx all the OpenGL PowerVR demos run very well. Several do not run under WEC7 (same exact platform), for example ChameleonMan does not run but puts up a "ERROR:GL_IMG_vertex_program not supported by this platform". Several others throw this error under WEC7 but work on CE6. I also noticed the ones that do run (OGLVASE), run quite a bit slower under WEC7 than under CE6 ...again same exact platform and speed (I believe).

    Also, I was able to get some of the DSP demo files to play. davincieffect.aac works and one of the TI demo videos works but others do not.  The omap4ad1_h264vidmpg1xxxx.avi actually play ok but the omap4ad1_mpeg4xxxx.avi plays upside down.  Several other .3GP videos don't work at all or play at a very slow rate. These all worked under CE6. I suspect the codec filters or the new video player to be the issue.   

     

  • DavidVescovi said:

    For #2 just open the BSP catalog in source mode ... you will see them.

    yeah, saw it. I will clean up.

    DavidVescovi said:

    For #5 I am using both Gumstix (3530) and Beagleboard (3730) platforms ... pretty much same issues. I was using both the CE6 and the WEC7 version of the EHCIPDD system.c ... neather with very much luck but the CE6 version was working somewhat better under CE6. Looks like the only difference was the power switching was moved to the single threaded poweroff function for some strange reasion.

    Can you resend this info? I am confused which one is better CE6: or WEC7? Also, are you using the EHCIPDD driver as is from 2.x release or have you done any mods to the driver for your platforms?

     

  • DavidVescovi said:

    Also, I was able to get some of the DSP demo files to play. davincieffect.aac works and one of the TI demo videos works but others do not.  The omap4ad1_h264vidmpg1xxxx.avi actually play ok but the omap4ad1_mpeg4xxxx.avi plays upside down.  Several other .3GP videos don't work at all or play at a very slow rate. These all worked under CE6. I suspect the codec filters or the new video player to be the issue.   

    which player are you using - ceplayit or silverlight video player? When it plays upside done, can you check if its using GDI or DDraw mode? A simple way to check is in TIMM_getDisplayBufViddec2() function to see if the video format requested is YUV (ddraw) or RGB (GDI). I know in RGB conversion we flip the video frame upside down since CE6 GDI needed the output in that form. I dont know if this is still the case for WEC7 - you may need some experimentation here.

    Have you ported the 3GP filter for WEC7? I am not sure what the porting effort is here.

     

     

  • David,

    Thanks once again for your feedback. Regarding your PowerVR observation: what we see with PowerVR between WinCE6 and WinCE7 is different from yours. Please find some demo performance numbers here.  http://processors.wiki.ti.com/index.php/WinCE_Comparative_Benchmarks#Demo_Performance_Comparison

    On an average we see better performance with WEC7 for the same demos (but the demos were recompiled and link  with the updated libs that are part of new SDK). Some demo have wide variation in fps based on the scene (ogles2skybox2) and hence are not suitable for comparison. But oglesvase is a good demo for comparison.

    Few points.

    1. PowerVR DDK 1.5 was used with WinCE6 and DDK 1.7 with WEC7.  

    2. Can you try the demos that are released as a part of the PowerVR-SDK in each of the release. If I remember the demos link with updated libs that are part of the new SDK and hence if using a demo (binary) from an older release may not give the best performance. 

    3. If you were to recompile the application that you see are not working with WEC7 (instead of running the executable as is), do they run correctly? 

    Jatin

  • Thanks Jatin,

    I probably was using the older demos under WEV7.

    I will investigate further and get back to you.

    Thanks for the help.

     

    DV

  • Was indeed using the older PowerVR demos.

    With new demos all OpenGL demos work on WEC7 w/35xx and 37xx platforms. 

    Thank you for pointing this out.

     

    DV