Count Read/Write sizes and flops in opencl_module_hmc (Feature #235)
The R/W sizes and flops still have to be counted for the kernels in opencl_module_hmc.
In 7b46c60d I added flop sizes for the kernels in opencl_module_hmc.
I am not sure at one point: In the gaussian kernels, one creates a lot of gaussian normal pairs. I did not count these operations in yet, although they for sure are calculationwise the most time consuming in these kernels.
What do you think, (how) should one include them?
- Assignee changed from Christopher Pinke to Matthias Bach
- % Done changed from 0 to 100
- Status changed from In Progress to Feedback