• State Accepted Answer
  • +1 person also asked this people also asked this
  • Locked Locked
  • Replies 2 replies
  • Subscribers 136 subscribers
  • Views 4849 views
  • Users 0 members are here
Options
Related
How was your experience today?
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

ARM MALI400 GPU VP and PP are not working parallel. Why ?

Ravinder

Dear ARM Team,

 

We are working on Mali400-MP2 HW and  MALI DDK version r5p0-01rel0.

We are seeing almost all our applications are running serially one after other for CPU, GPU-VP, GPU-FP.

I attached one of our application performance chart captured using the ARM DS-5 streamline tool.

 

Questions:

  1. Why our applications are not able to process parallel on GPU-VP and GPU-VP, what are we missing in our Graphics solution ?
  2. How can we write an application which makes more efficient use of the CPU and GPU (GPU-VP, GPU-FP).
  3. Do you have any guidelines to achieve better parallelism and performance ?
  4. Can we get an example application which was written to get more advantage of Mali400 GPU ?

Please help me on this.

Thanks,

Ravinder Are

Parents
  • +1
    Why our applications are not able to process parallel on GPU-VP and GPU-VP, what are we missing in our Graphics solution ?

    Guessing slightly without seeing what your API sequence is looking like, but at a guess it looks like you are breaking the pipeline by enforcing some point of serialization in the GL behaviour. Do you have a glFinish() or a glReadPixels() anywhere in your per-frame rendering sequence? If you do please remove them. If you can explain why you have them we can hopefully provide some alternative.

    Cheers,
    Pete

Reply
  • +1
    Why our applications are not able to process parallel on GPU-VP and GPU-VP, what are we missing in our Graphics solution ?

    Guessing slightly without seeing what your API sequence is looking like, but at a guess it looks like you are breaking the pipeline by enforcing some point of serialization in the GL behaviour. Do you have a glFinish() or a glReadPixels() anywhere in your per-frame rendering sequence? If you do please remove them. If you can explain why you have them we can hopefully provide some alternative.

    Cheers,
    Pete

Children
No data