rollback the OpenCL advection kernel to previous version (standard + bilinear...
rollback the OpenCL advection kernel to previous version (standard + bilinear velocity interpolation in bilevel configuration). Need to improve the cache line and the interpolation order (see #35)
Showing
- hysop/backend/device/codegen/kernels/directional_advection.py 267 additions, 219 deletions...p/backend/device/codegen/kernels/directional_advection.py
- hysop/backend/device/opencl/autotunable_kernels/advection_dir.py 28 additions, 40 deletions...ackend/device/opencl/autotunable_kernels/advection_dir.py
- hysop/backend/device/opencl/operator/directional/advection_dir.py 6 additions, 3 deletions...ckend/device/opencl/operator/directional/advection_dir.py
- hysop/operator/base/advection_dir.py 16 additions, 46 deletionshysop/operator/base/advection_dir.py
Loading