CLOUDSC low-level GPU (transpilation) via Loki (CUF/CUDA)#328

Merged

reuterbal merged 36 commits intomainfrom nams-cloudsc-low-level-GPU

Sep 6, 2024

+2,890-843

Commits on Sep 6, 2024

introducing c-like backends: cppgen, cudagen

MichaelSt98
authored and
reuterbal
committed
Fix case-sensitivity problem in data offload

MichaelSt98
authored and
reuterbal
committed
Introduce 'LowerBlockIndexTransformation', 'LowerBlockLoopTransformation'

MichaelSt98
authored and
reuterbal
committed
Add low level GPU offload pipelines

MichaelSt98
authored and
reuterbal
committed
Introduce SCCLowLevel* transformation

MichaelSt98
authored and
reuterbal
committed
Add some SCC base/vector functionality

MichaelSt98
authored and
reuterbal
committed
Fortran C transpilation: add functionality and distinguish between C and CUDA gen

MichaelSt98
authored and
reuterbal
committed
Introduce new modes and alter how some modes are dispatched (CUF, CUDA)

MichaelSt98
authored and
reuterbal
committed
Please linter

MichaelSt98
authored and
reuterbal
committed
Introduce a 'CaseInsensitiveDefaultDict'

MichaelSt98
authored and
reuterbal
committed
Some fixes related to multiple (nested) kernels and elemental routines

MichaelSt98
authored and
reuterbal
committed
Adapt tests for scc_cuf and block_index related transformations

MichaelSt98
authored and
reuterbal
committed
cleanup/remove some unnecessary comments

MichaelSt98
authored and
reuterbal
committed
Small fix for lower loop trafo and improve/extend corresponding tests

MichaelSt98
authored and
reuterbal
committed
scc_cuf cleanup, removal of some unnecessary comments etc

MichaelSt98
authored and
reuterbal
committed
remove 'cuf-dynamic' and the 'dynamic memory allocation on the device'-option in general

MichaelSt98
authored and
reuterbal
committed
cgen/cudagen case-sensitivity for function declaration

MichaelSt98
authored and
reuterbal
committed
cuda backend, case-sensitivity for calls

MichaelSt98
authored and
reuterbal
committed
smaller fixes: convert kwargs to args within fortran_c transpilation, check for intent, ...

MichaelSt98
authored and
reuterbal
committed
append/extend/improve f2c transpilation tests

MichaelSt98
authored and
reuterbal
committed
final clean-up, add documentation, etc

MichaelSt98
authored and
reuterbal
committed
improve docstring for CaseInsensitiveDict

MichaelSt98
authored and
reuterbal
committed
improved on c-like backends

MichaelSt98
authored and
reuterbal
committed
'add_explicit_array_dimensions' and 'remove_explicit_array_dimensions' utilities within 'array_indexing'

MichaelSt98
authored and
reuterbal
committed
Improve on/refactor 'LowerBlockIndexTransformation' and 'LowerBlockLoopTransformation'

MichaelSt98
authored and
reuterbal
committed
remove debugging leftovers, add documentation, improve minor implementation details, ...

MichaelSt98
authored and
reuterbal
committed
fix includes for successors for F2C

MichaelSt98
authored and
reuterbal
committed
add tmp_path for tests

MichaelSt98
authored and
reuterbal
committed
fix call signature changes for lowering block loop transformation

MichaelSt98
authored and
reuterbal
committed
add missing import

MichaelSt98
authored and
reuterbal
committed
fix for elemental routines and single variable declaration not only grouped by shape

MichaelSt98
authored and
reuterbal
committed
missing tmp_path within tests (again)

MichaelSt98
authored and
reuterbal
committed
specify/fix intent for new argument which allows to remove unnecessary check for intent being None

MichaelSt98
authored and
reuterbal
committed
remove debug leftovers, small improvements, ...

MichaelSt98
authored and
reuterbal
committed
Proper handling of omp pragmas for LowerBlockLoop trafo

MichaelSt98
authored and
reuterbal
committed
Introduce new modes idem-lower, idem-lower-loop, cuda-parametrise and cuda-hoist

MichaelSt98
authored and
reuterbal
committed