Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
8 years ago

Report Loop Analysis NDrange threads

Hello everyone Im struggling to find what means the Details tab in Loop Analysis section of the report, saying nd-range: thread capacity = 303 example 2Dimension Kernel: ___Kernel(...
Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
8 years ago

--- Quote Start ---

Thread-id-dependent branching will have a negative performance impact, because run time will be dominated by threads that perform the most amount of work. Furthermore, having such branches in your code will prevent you from being able to use the SIMD attribute to increase the performance of your kernel. However, the runtime scheduler will try to minimize the number of pipeline stalls/bubbles and maximize performance in any case.

--- Quote End ---

Thanks again for the help HRZ! You are a big help here on OpenCL section :)