There was an error fetching the commit references. Please try again later.
Pass field information (shape,stride) as single elements instead of arr
- small (length < 5) arrays with shape and stride information had to be memcpy'd to the GPU before every kernel call - instead of passing the information as arrays, the single elements are passed - leads to more function arguments, but simplifies GPU kernel calls -> changes in all backends required
Showing
- astnodes.py 59 additions, 64 deletionsastnodes.py
- backends/cbackend.py 1 addition, 1 deletionbackends/cbackend.py
- backends/dot.py 4 additions, 3 deletionsbackends/dot.py
- boundaries/boundaryhandling.py 5 additions, 4 deletionsboundaries/boundaryhandling.py
- cpu/cpujit.py 44 additions, 58 deletionscpu/cpujit.py
- cpu/kernelcreation.py 1 addition, 1 deletioncpu/kernelcreation.py
- data_types.py 1 addition, 2 deletionsdata_types.py
- datahandling/parallel_datahandling.py 3 additions, 2 deletionsdatahandling/parallel_datahandling.py
- field.py 6 additions, 14 deletionsfield.py
- gpucuda/cudajit.py 53 additions, 54 deletionsgpucuda/cudajit.py
- kerncraft_coupling/generate_benchmark.py 6 additions, 6 deletionskerncraft_coupling/generate_benchmark.py
- kerncraft_coupling/kerncraft_interface.py 4 additions, 4 deletionskerncraft_coupling/kerncraft_interface.py
- kernelparameters.py 92 additions, 0 deletionskernelparameters.py
- llvm/llvm.py 8 additions, 4 deletionsllvm/llvm.py
- llvm/llvmjit.py 25 additions, 27 deletionsllvm/llvmjit.py
- sympyextensions.py 5 additions, 0 deletionssympyextensions.py
- transformations.py 7 additions, 13 deletionstransformations.py
Please register or sign in to comment