Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

CUDA variant optimised beyond k-caching #104

Merged
merged 4 commits into from
Dec 20, 2024
Merged

CUDA variant optimised beyond k-caching #104

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
6b975a2
CUDA variant optimised beyond k-caching, e.g., pipelined global-to-sh…
MichaelSt98 Dec 11, 2024
7fb7401
describe new even more optimized CUDA C variant in README
MichaelSt98 Dec 19, 2024
add9bbc
improve on CMake usage for CUDA C variants
MichaelSt98 Dec 19, 2024
6beb1d9
specify/set CUDA architecture to 80 for GitHub toolchain file
MichaelSt98 Dec 19, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,6 +54,7 @@ ecbuild_add_option( FEATURE CUDA
CONDITION CMAKE_CUDA_COMPILER )
if( HAVE_CUDA )
enable_language( CUDA )
find_package(CUDAToolkit)
endif()

ecbuild_add_option( FEATURE HIP
Expand Down
6 changes: 5 additions & 1 deletion README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,7 +80,11 @@ Balthasar Reuter ([email protected])
of CLOUDSC including host side hoisted temporary local variables.
- **dwarf-cloudsc-cuda-k-caching**: GPU-enabled, further optimized CUDA
C version of CLOUDSC including loop fusion and temporary local
array demotion.
array demotion.
- **dwarf-cloudsc-cuda-opt**: GPU-enabled, further optimized beyond
k-caching CUDA C version that buffers some variables and
uses pipelined global-to-shared memory copies that are overlapped
with compute (TMA loads).
- **dwarf-cloudsc-gpu-scc-field**: GPU-enabled and optimized version of
CLOUDSC that uses the SCC loop layout, and uses [FIELD API](https://github.com/ecmwf-ifs/field_api) (a Fortran library purpose-built for IFS data-structures that facilitates the
creation and management of field objects in scientific code) to perform device offload
Expand Down
4 changes: 4 additions & 0 deletions arch/toolchains/github-ubuntu-nvhpc.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,6 +32,10 @@ set( OpenACC_Fortran_FLAGS "-acc=gpu -mp" CACHE STRING "" )
# COMMON FLAGS
####################################################################

if(NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
set(CMAKE_CUDA_ARCHITECTURES 80)
endif()

set(ECBUILD_Fortran_FLAGS "-fpic")
set(ECBUILD_Fortran_FLAGS "${ECBUILD_Fortran_FLAGS} -Mframe")
set(ECBUILD_Fortran_FLAGS "${ECBUILD_Fortran_FLAGS} -Mbyteswapio")
Expand Down
100 changes: 65 additions & 35 deletions src/cloudsc_cuda/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,17 @@ if( HAVE_CLOUDSC_C_CUDA )
enable_language(CUDA)
enable_language(CXX)

if (NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
set( CLOUDSC_CUDA_FLAGS "-lineinfo -maxrregcount=128" )
else()
set( CLOUDSC_CUDA_FLAGS "-lineinfo -maxrregcount=128 -gencode arch=compute_${CMAKE_CUDA_ARCHITECTURES},code=sm_${CMAKE_CUDA_ARCHITECTURES}>" )
endif()
if(CMAKE_BUILD_TYPE STREQUAL "Debug")
set( CLOUDSC_CUDA_OPT_FLAGS "-O0 -g -G" )
else()
set( CLOUDSC_CUDA_OPT_FLAGS "--ptxas-options=-O3 -O3 -use_fast_math" )
endif()

###### SCC-CUDA ####
ecbuild_add_library(
TARGET dwarf-cloudsc-c-cuda-lib
Expand All @@ -41,22 +52,12 @@ if( HAVE_CLOUDSC_C_CUDA )
$<${HAVE_HDF5}:hdf5::hdf5>
$<${HAVE_SERIALBOX}:Serialbox::Serialbox_C>
$<${HAVE_OMP}:OpenMP::OpenMP_C>
CUDA::cudart
DEFINITIONS
${CLOUDSC_DEFINITIONS}
)

target_include_directories(
dwarf-cloudsc-c-cuda-lib
PUBLIC
${CMAKE_CUDA_TOOLKIT_INCLUDE_DIRECTORIES}
)
if (NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
target_compile_options(dwarf-cloudsc-c-cuda-lib PRIVATE $<$<COMPILE_LANGUAGE:CUDA>>)
else()
target_compile_options(dwarf-cloudsc-c-cuda-lib PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:
--ptxas-options=-O3 -use_fast_math -maxrregcount=128 -gencode arch=compute_${CMAKE_CUDA_ARCHITECTURES},code=sm_${CMAKE_CUDA_ARCHITECTURES}>)
endif()

target_compile_options(dwarf-cloudsc-c-cuda-lib PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:SHELL:${CLOUDSC_CUDA_OPT_FLAGS} ${CLOUDSC_CUDA_FLAGS}>)
set_target_properties( dwarf-cloudsc-c-cuda-lib PROPERTIES CUDA_SEPARABLE_COMPILATION ON)

ecbuild_add_executable(
Expand Down Expand Up @@ -99,22 +100,12 @@ if( HAVE_CLOUDSC_C_CUDA )
$<${HAVE_HDF5}:hdf5::hdf5>
$<${HAVE_SERIALBOX}:Serialbox::Serialbox_C>
$<${HAVE_OMP}:OpenMP::OpenMP_C>
CUDA::cudart
DEFINITIONS
${CLOUDSC_DEFINITIONS}
)

target_include_directories(
dwarf-cloudsc-c-cuda-hoist-lib
PUBLIC
${CMAKE_CUDA_TOOLKIT_INCLUDE_DIRECTORIES}
)
if (NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
target_compile_options(dwarf-cloudsc-c-cuda-hoist-lib PRIVATE $<$<COMPILE_LANGUAGE:CUDA>>)
else()
target_compile_options(dwarf-cloudsc-c-cuda-hoist-lib PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:
--ptxas-options=-O3 -use_fast_math -maxrregcount=128 -gencode arch=compute_${CMAKE_CUDA_ARCHITECTURES},code=sm_${CMAKE_CUDA_ARCHITECTURES}>)
endif()

target_compile_options(dwarf-cloudsc-c-cuda-hoist-lib PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:SHELL:${CLOUDSC_CUDA_OPT_FLAGS} ${CLOUDSC_CUDA_FLAGS}>)
set_target_properties( dwarf-cloudsc-c-cuda-hoist-lib PROPERTIES CUDA_SEPARABLE_COMPILATION ON)

ecbuild_add_executable(
Expand Down Expand Up @@ -157,21 +148,12 @@ if( HAVE_CLOUDSC_C_CUDA )
$<${HAVE_HDF5}:hdf5::hdf5>
$<${HAVE_SERIALBOX}:Serialbox::Serialbox_C>
$<${HAVE_OMP}:OpenMP::OpenMP_C>
CUDA::cudart
DEFINITIONS
${CLOUDSC_DEFINITIONS}
)

target_include_directories(
dwarf-cloudsc-c-cuda-k-caching-lib
PUBLIC
${CMAKE_CUDA_TOOLKIT_INCLUDE_DIRECTORIES}
)
if (NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
target_compile_options(dwarf-cloudsc-c-cuda-k-caching-lib PRIVATE $<$<COMPILE_LANGUAGE:CUDA>>)
else()
target_compile_options(dwarf-cloudsc-c-cuda-k-caching-lib PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:
--ptxas-options=-O3 -use_fast_math -maxrregcount=128 -gencode arch=compute_${CMAKE_CUDA_ARCHITECTURES},code=sm_${CMAKE_CUDA_ARCHITECTURES}>)
endif()
target_compile_options(dwarf-cloudsc-c-cuda-k-caching-lib PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:SHELL:${CLOUDSC_CUDA_OPT_FLAGS} ${CLOUDSC_CUDA_FLAGS}>)
set_target_properties( dwarf-cloudsc-c-cuda-k-caching-lib PROPERTIES CUDA_SEPARABLE_COMPILATION ON)

ecbuild_add_executable(
Expand All @@ -191,6 +173,54 @@ if( HAVE_CLOUDSC_C_CUDA )
)
###

###### SCC-CUDA-OPT ####
ecbuild_add_library(
TARGET dwarf-cloudsc-c-cuda-opt-lib
INSTALL_HEADERS LISTED
SOURCES
cloudsc/yoecldp_c.h
cloudsc/load_state.h
cloudsc/load_state.cu
cloudsc/cloudsc_c_opt.h
cloudsc/cloudsc_c_opt.cu
cloudsc/cloudsc_driver_opt.h
cloudsc/cloudsc_driver_opt.cu
cloudsc/cloudsc_validate.h
cloudsc/cloudsc_validate.cu
cloudsc/mycpu.h
cloudsc/mycpu.cu
PUBLIC_INCLUDES
$<INSTALL_INTERFACE:include>
$<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/cloudsc>
PUBLIC_LIBS
$<${HAVE_HDF5}:hdf5::hdf5>
$<${HAVE_SERIALBOX}:Serialbox::Serialbox_C>
$<${HAVE_OMP}:OpenMP::OpenMP_C>
CUDA::cudart
DEFINITIONS
${CLOUDSC_DEFINITIONS}
)

target_compile_options(dwarf-cloudsc-c-cuda-opt-lib PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:SHELL:${CLOUDSC_CUDA_OPT_FLAGS} ${CLOUDSC_CUDA_FLAGS}>)
set_target_properties( dwarf-cloudsc-c-cuda-opt-lib PROPERTIES CUDA_SEPARABLE_COMPILATION ON)

ecbuild_add_executable(
TARGET dwarf-cloudsc-c-cuda-opt
SOURCES dwarf_cloudsc.cpp
LIBS dwarf-cloudsc-c-cuda-opt-lib
)

target_link_libraries(dwarf-cloudsc-c-cuda-opt dwarf-cloudsc-c-cuda-opt-lib)

ecbuild_add_test(
TARGET dwarf-cloudsc-c-cuda-opt-serial
COMMAND bin/dwarf-cloudsc-c-cuda-opt
ARGS 1 1000 128
WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/../../..
OMP 1
)
###

# Create symlink for the input data
if( HAVE_SERIALBOX )
execute_process(COMMAND ${CMAKE_COMMAND} -E create_symlink
Expand Down
Loading