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

glMapBufferRange overhead second inquery

I had a topic question on the old forum which shown an issue with the Mali-T6xx driver being terribly slow with both glBufferSubData and glMapBufferRange with the unsync bit set.

I was wondering if there has been any information on if the situation has improved on your end. I've followed the guide here for installing a Unix system on my Chromebook with the latest drivers you provide to the public and the issue is still there.

Any way to get an update on if its fixed and just pending an update somewhere?

Parents
  • Hi sonicadvance1,

    Is there any chance you could post a reproducing test case so we can clearly see what issue you are running in to?

    The devil is in the detail for most cases of this type of performance bottlenecks, as they usually stem from OpenGL ES's synchronous view of the world interacting badly with the reality of a pipelined rendering architecture. For example, if you try to modify data which is still referenced by a pending rendering operation in the command stream we either need to clone the resource or flush out the rendering state until the data is no longer referenced. Using GLES 3 with glMapBufferRange with the unsynchronized bit should avoid most of those architectural issues of course, so if you can post a reproducer we will be happy to investigate.

    Thanks for reporting,
    Pete

Reply
  • Hi sonicadvance1,

    Is there any chance you could post a reproducing test case so we can clearly see what issue you are running in to?

    The devil is in the detail for most cases of this type of performance bottlenecks, as they usually stem from OpenGL ES's synchronous view of the world interacting badly with the reality of a pipelined rendering architecture. For example, if you try to modify data which is still referenced by a pending rendering operation in the command stream we either need to clone the resource or flush out the rendering state until the data is no longer referenced. Using GLES 3 with glMapBufferRange with the unsynchronized bit should avoid most of those architectural issues of course, so if you can post a reproducer we will be happy to investigate.

    Thanks for reporting,
    Pete

Children