Archana,
I modified the freeverb low pass feedback comb filter kernel to work with our asio host and verified that the wait time is still about 5 seconds per sample block. I decided instead of spending a lot of effort trying to port in the monolithic full freeverb kernel that I wrote, that I would experiment with reducing the amount of times the lbcf kernel was called per sample block. The thing is, even when it is called once per sample block, the time per block remained at about 500ms, which is still clearly unacceptable. It is highly doubtful that the monolithic kernel, which requires MANY more buffers to work, will provide acceptable latency.
So the next step is to find out what is going wrong, why the lbcf kernel is taking so long on its own. Is the kernel actually working? Is the gpu tied up on a particular instruction? Does it even run on the gpu?
Also, it would be good if someone could follow up with Dr. Femiani or the 'Russ' person he mentioned about the whole gpu not running kernels thing.
RS
RS
I'll send you what I have on the next kernel when I get home. You'll have to write the setup code yourself. the stuff I was working on was in an
Sorry, it was the new asio host project, forgot the exact name and location but I thought I left it open.
RS