Forum Discussion
Hi @RN1,
Good day, just following up on the previous clarification.
By any chances did you managed to look into the it?
Best Wishes
BB
Good day.
I answered in the other thread.
We did more tests and nothing, the performance is not good and the results are the same as those attached in the other post.
We compared "System viewers->Graph viewer (beta)" in openCL with "Views->System viewer" in oneAPI. OpenCL is doing coalescing to local accesses, while oneAPI it isn't. It uses many small units. We assume that regarding the DSPs something is bad as well, but it is complex to see since the generated graph is huge. Maybe it is not reducing/trimming the tree, but we did nothing different from OpenCL, so, we need support in how to optimize these local accesses to achieve at least similar performance than in OpenCL (nothing is explained in the optimization guides).
Following this thread, Do you know why interoperability OpenCL-oneAPI does not work with the fpga?