Commit d1ce4a0b authored by Robert Maynard's avatar Robert Maynard

Fix the default launch sizes for Tesla hardware.

The 8x8x8 is a better launch strategy for most VTK-m kernels.
The current problem is that a couple of VTK-m kernels use a
high number of registers and this number of threads combines to
require too many registers.

What we should do in the longer run is have more controls over
kernel launches on a per kernel basis. This will require VTK-m
to extract the number of registers being used by each kernel
parent 7fdf5ec2
......@@ -68,8 +68,8 @@ struct VTKM_CONT_EXPORT ScheduleParameterBuilder
{ 32, 128, 8, { 16, 16, 1 }, 32, { 8, 8, 4 } } }, //VTK-m default for less than pascal
{ GPU_STRATA::PASCAL, { 32, 128, 8, { 16, 16, 1 }, 32, { 8, 8, 4 } } }, //VTK-m default for pascal
{ GPU_STRATA::VOLTA, { 32, 128, 8, { 16, 16, 1 }, 32, { 8, 8, 4 } } }, //VTK-m default for volta
{ GPU_STRATA::PASCAL_HPC, { 32, 256, 16, { 16, 16, 1 }, 32, { 8, 8, 8 } } }, //P100
{ GPU_STRATA::VOLTA_HPC, { 32, 256, 16, { 16, 16, 1 }, 32, { 8, 8, 8 } } }, //V100
{ GPU_STRATA::PASCAL_HPC, { 32, 256, 16, { 16, 16, 1 }, 64, { 8, 8, 4 } } }, //P100
{ GPU_STRATA::VOLTA_HPC, { 32, 256, 16, { 16, 16, 1 }, 64, { 8, 8, 4 } } }, //V100
}
, Compute(nullptr)
{
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment