Client: Intel
Responsibilities: Research & development, tutorial design & presentation
This was part of a bigger goal of evaluating potential OpenCL 2.0 feature additions. Due to the nature of the kind of computations, the work needed to be broken into two kernels, each with different workload configurations.
While OpenCL follows through on the intent to be portable across GPU manufacturers, we found the programs/kernels required additional tuning to keep optimal performance due to architectural and compiler differences.
Technologies:
- OpenCL
- C