topic try catch can be expensive in Intel® oneAPI DPC++/C++ Compiler

try catch can be expensive

Guracar__Ismayil — Fri, 13 Dec 2019 16:20:27 GMT

When I run the DPC++ compatibility tool on CUDA code I see that every one of my kernel launches in the generated cpp code is bracketed by a try-catch construct. It has been my experience that 'try' can be expensive in execution time (as well the experience of others, e.g. https://stackoverflow.com/questions/52312/what-is-the-real-overhead-of-try-catch-in-c). Is there a way to disable or consolidate the try-catch so that it doesn't occur at each kernel launch but perhaps after a block of launches. We might lose the ability to isolate errors to a particular launch, but that might be a reasonable tradeoff for performance in situations where large numbers of kernels must be launched.

Thanks

Ismayil

Hi,

AnubhavJ_intel — Mon, 16 Dec 2019 04:27:02 GMT

Hi,

Thanks for reaching out to us. We will get back to you soon.

Hi Ismayil,

Varsha_M_Intel — Tue, 17 Dec 2019 19:20:47 GMT

Hi Ismayil,

After migration to DPCPP, the user is required to review, complete, test and fine tune and optimize(as you may see fit) the output of the tool.

-Varsha

Re:try catch can be expensive

AbhishekD_Intel — Fri, 18 Jun 2021 10:01:07 GMT

Hi,

As we haven't heard back from you, we are considering that your issue has been resolved.

We will no longer monitor this thread. If you require any additional assistance from Intel, please start a new thread.

Any further interaction in this thread will be considered community only.

Warm Regards,

Abhishek