56fd68e7dd
CCCL (CUDA C++ Core Libraries) provides: - CUB: device/block/warp-level GPU primitives (reduce, scan, sort, topk) - Thrust: high-level parallel algorithms (transform_reduce, sort, scan) - libcudacxx: CUDA C++ standard library (atomics, barriers, memory) - cudax: experimental features (memory resources, allocators) - Tuning policies: per-SM hardware-specific algorithm parameters Competition optimization vectors mapped to CCCL: - Output TPS (83% weight): warp_reduce, block_reduce, device_topk - Input TPS (14% weight): device_scan, block_load, prefetch - Cache TPS (3% weight): prefix caching strategy patterns - Memory (0.9 util): pooled/cached/buddy allocators Source: https://github.com/NVIDIA/cccl (shallow clone, HEAD only) License: Apache-2.0
337 lines
11 KiB
CMake
337 lines
11 KiB
CMake
set(
|
|
stf_test_sources
|
|
allocators/buddy_allocator.cu
|
|
context/context_resource_test.cu
|
|
context/simple_context_resource_test.cu
|
|
cpp/concurrency_test.cu
|
|
cpp/redundant_data.cu
|
|
cpp/redundant_data_different_modes.cu
|
|
cpp/scoped_graph_task.cu
|
|
cpp/task_get_stream.cu
|
|
cpp/test_pick_stream.cu
|
|
cpp/test_pick_stream_green_context.cu
|
|
cpp/user_streams.cu
|
|
cpp/wait_token.cu
|
|
cuda-samples/3_CUDA_Features/graphConditionalNodes/graphConditionalNodes.cu
|
|
dot/basic.cu
|
|
dot/graph_print_to_dot.cu
|
|
dot/sections.cu
|
|
dot/sections_2.cu
|
|
dot/sections_stackable.cu
|
|
dot/section_movable.cu
|
|
dot/with_events.cu
|
|
error_checks/ctx_mismatch.cu
|
|
error_checks/data_interface_mismatch.cu
|
|
error_checks/double_finalize.cu
|
|
error_checks/erase_frozen.cu
|
|
error_checks/misformed_tasks_dbl_end.cu
|
|
error_checks/misformed_tasks_dbl_start.cu
|
|
error_checks/non_managed_data.cu
|
|
error_checks/stackable_fence_in_nested.cu
|
|
error_checks/stackable_illegal_export.cu
|
|
error_checks/stackable_inconsistent_access_modes.cu
|
|
error_checks/stackable_wait_in_nested.cu
|
|
error_checks/uninitialized_data.cu
|
|
error_checks/write_frozen.cu
|
|
examples/cuda_kernels_driver.cu
|
|
examples/05-stencil-no-copy.cu
|
|
examples/05-stencil-places.cu
|
|
examples/05-stencil.cu
|
|
examples/05-stencil2d-places.cu
|
|
fhe/parse_arctyrex.cu
|
|
gnu/include_only.cpp
|
|
graph/concurrency_test.cu
|
|
graph/explicit_graph.cu
|
|
graph/explicit_graph_async.cu
|
|
graph/explicit_graph_while.cu
|
|
graph/explicit_graph_while-kernels.cu
|
|
graph/get_cache_stats.cu
|
|
graph/graph_cache_policy.cu
|
|
graph/graph_child_move_ownership.cu
|
|
graph/graph_ctx_low_level.cu
|
|
graph/static_graph_ctx.cu
|
|
hashtable/test.cu
|
|
interface/cuda_kernel_chain-add_deps.cu
|
|
interface/cuda_kernel_chain-add_deps_low_level.cu
|
|
interface/cuda_kernel_empty_args.cu
|
|
interface/move_operator.cu
|
|
local_stf/legacy_to_stf.cu
|
|
local_stf/legacy_to_stf_in_capture.cu
|
|
local_stf/logical_data_t_template.cu
|
|
local_stf/stackable.cu
|
|
local_stf/stackable2.cu
|
|
local_stf/stackable_add_deps.cu
|
|
local_stf/stackable_export.cu
|
|
local_stf/stackable_export2.cu
|
|
local_stf/stackable_move.cu
|
|
local_stf/stackable_nested_lib.cu
|
|
local_stf/stackable_nested_repeat.cu
|
|
local_stf/stackable_nested_while.cu
|
|
local_stf/stackable_nested.cu
|
|
local_stf/stackable_node_pool_growth.cu
|
|
local_stf/stream_ctx_lifetime_btb.cu
|
|
local_stf/stackable_read_only.cu
|
|
local_stf/stackable_threads.cu
|
|
local_stf/stackable_token.cu
|
|
local_stf/stackable_write_back.cu
|
|
local_stf/stackable_redundant_deps.cu
|
|
local_stf/stackable_tmp.cu
|
|
local_stf/task_dep_copy_move_semantics.cu
|
|
local_stf/threads_multiple_graphs.cu
|
|
places/affinity_gc.cu
|
|
places/cute_parallel_for.cu
|
|
places/managed.cu
|
|
places/managed_from_user.cu
|
|
places/non_current_device.cu
|
|
places/place_partition.cu
|
|
places/recursion.cu
|
|
places/execution_policy_kernel_launch_test.cu
|
|
reclaiming/graph.cu
|
|
reclaiming/graph_2.cu
|
|
reclaiming/graph_real_oom.cu
|
|
reclaiming/stream.cu
|
|
reductions/many_inc.cu
|
|
reductions/redux_test.cu
|
|
reductions/redux_test2.cu
|
|
reductions/slice2d_reduction.cu
|
|
reductions/slice_custom_op.cu
|
|
reductions/successive_reductions.cu
|
|
reductions/sum.cu
|
|
reductions/sum_array.cu
|
|
reductions/sum_multiple_places_no_refvalue.cu
|
|
slice/pinning.cu
|
|
stackable/graph_scope_test.cu
|
|
stencil/stencil-1D.cu
|
|
stress/empty_tasks.cu
|
|
stress/empty_tasks_alloc.cu
|
|
stress/kernel_chain.cu
|
|
stress/kernel_chain_fused.cu
|
|
stress/many_read.cu
|
|
stress/task_bench.cu
|
|
threads/axpy-threads-2.cu
|
|
# threads/axpy-threads.cu
|
|
utility/timing_with_fences.cu
|
|
utility/source_location_map.cu
|
|
)
|
|
|
|
set(
|
|
stf_test_codegen_sources
|
|
# algorithm/algorithm_with_read.cu
|
|
# algorithm/graph_algorithms.cu
|
|
# algorithm/in_graph_ctx.cu
|
|
# algorithm/nested.cu
|
|
allocators/adapter.cu
|
|
allocators/cap_tmp_buffers.cu
|
|
cpp/read_const.cu
|
|
cpp/reuse_computation.cu
|
|
cpp/reuse_computation_2.cu
|
|
error_checks/slice_check_bounds.cu
|
|
error_checks/unsatisfiable_spec.cu
|
|
examples/01-axpy-launch-ranges-cg.cu
|
|
examples/01-axpy-places.cu
|
|
# examples/09-nbody-algorithm.cu
|
|
examples/09-nbody-blocked.cu
|
|
examples/09-nbody.cu
|
|
freeze/constant_logical_data.cu
|
|
freeze/freeze.cu
|
|
freeze/freeze_rw.cu
|
|
freeze/freeze_untyped_rw.cu
|
|
freeze/freeze_write_back.cu
|
|
freeze/task_fence.cu
|
|
freeze/token.cu
|
|
graph/epoch.cu
|
|
graph/freeze_for_graph.cu
|
|
graph/freeze_nested_graphs.cu
|
|
graph/freeze_while_graphs.cu
|
|
graph/freeze_while_graphs_2.cu
|
|
graph/freeze_while_graphs_3.cu
|
|
# graph/graph_composition.cu
|
|
graph/graph_multiple_launch_resources.cu
|
|
graph/graph_tmp_data.cu
|
|
graph/many.cu
|
|
graph/multiple_graph_ctx.cu
|
|
green_context/axpy_gc.cu
|
|
green_context/cuda_graph.cu
|
|
green_context/gc_grid.cu
|
|
hash/ctx_hash.cu
|
|
hash/logical_data.cu
|
|
hashtable/fusion.cu
|
|
hashtable/fusion_reduction.cu
|
|
hashtable/parallel_for.cu
|
|
hashtable/parallel_for_shape.cu
|
|
interface/data_from_device.cu
|
|
interface/data_from_device_2.cu
|
|
interface/data_from_device_wb.cu
|
|
interface/graph_use_device_data.cu
|
|
interface/host_launch_deps.cu
|
|
interface/mix_stream_and_graph.cu
|
|
interface/mix_stream_and_graph_2.cu
|
|
interface/scal.cu
|
|
interface/scalar_div.cu
|
|
interface/scalar_interface.cu
|
|
interface/stream_add_callback.cu
|
|
local_stf/interop_cuda.cu
|
|
loop_dispatch/dispatch_on_streams.cu
|
|
loop_dispatch/loop_dispatch.cu
|
|
loop_dispatch/nested_loop_dispatch.cu
|
|
# loop_dispatch/stackable_loop_dispatch.cu # Disabled: double free or corruption (under investigation)
|
|
parallel_for/empty_shape_reduce.cu
|
|
parallel_for/fdtd.cu
|
|
parallel_for/parallel_for_all_devs.cu
|
|
parallel_for/parallel_for_box.cu
|
|
parallel_for/parallel_for_repeat.cu
|
|
parallel_for/test2_parallel_for_context.cu
|
|
parallel_for/tiled_loops.cu
|
|
parallel_for/parallel_for_host.cu
|
|
places/cuda_stream_place.cu
|
|
places/managed_from_shape.cu
|
|
reductions/reduce_sum.cu
|
|
reductions/successive_reductions_pfor.cu
|
|
reductions/sum_multiple_places.cu
|
|
reductions/write_back_after_redux.cu
|
|
stress/launch_overhead.cu
|
|
stress/launch_vs_parallelfor.cu
|
|
stress/parallel_for_overhead.cu
|
|
# threads/axpy-threads-pfor.cu # Currently has a difficult-to-reproduce concurrency problem
|
|
threads/axpy-threads-graph.cu
|
|
threads/axpy-threads-graph-capture.cu
|
|
)
|
|
|
|
# Examples using CUBLAS, CUSOLVER...
|
|
set(
|
|
stf_test_mathlib_sources
|
|
cuda-samples/0_Introduction/vectorAdd/vectorAdd_cudastf.cu
|
|
# Reduce compilation time by not adding this (useless) example
|
|
# cuda-samples/0_Introduction/vectorAdd/vectorAdd
|
|
cuda-samples/3_CUDA_Features/jacobiCudaGraphs/jacobi.cu
|
|
cuda-samples/3_CUDA_Features/jacobiCudaGraphs/jacobi_cudastf.cu
|
|
cuda-samples/4_CUDA_Libraries/conjugateGradientMultiDeviceCG/conjugateGradientMultiDeviceCG_custf.cu
|
|
cuda-samples/5_Domain_Specific/MonteCarloMultiGPU_cudastf/MonteCarloMultiGPU.cu
|
|
examples/07-cholesky-redux.cu
|
|
examples/07-cholesky-unified.cu
|
|
gnu/06-pdgemm.cpp
|
|
gnu/07-cholesky.cpp
|
|
)
|
|
|
|
set(
|
|
stf_unittested_headers
|
|
cuda/experimental/__stf/allocators/buddy_allocator.cuh
|
|
cuda/experimental/__stf/graph/graph_ctx.cuh
|
|
cuda/experimental/__stf/internal/async_resources_handle.cuh
|
|
cuda/experimental/__stf/internal/context.cuh
|
|
cuda/experimental/__stf/internal/execution_policy.cuh
|
|
cuda/experimental/__stf/internal/interpreted_execution_policy.cuh
|
|
cuda/experimental/__stf/internal/slice.cuh
|
|
cuda/experimental/__stf/internal/thread_hierarchy.cuh
|
|
cuda/experimental/__stf/internal/inner_shape.cuh
|
|
cuda/experimental/__stf/stream/stream_ctx.cuh
|
|
cuda/experimental/__stf/utility/cartesian_iterator.cuh
|
|
cuda/experimental/__stf/utility/cuda_safe_call.cuh
|
|
cuda/experimental/__stf/utility/dimensions.cuh
|
|
cuda/experimental/__stf/utility/handle.cuh
|
|
cuda/experimental/__stf/utility/hash.cuh
|
|
cuda/experimental/__stf/utility/memory.cuh
|
|
cuda/experimental/__stf/utility/scope_guard.cuh
|
|
cuda/experimental/__stf/stackable/stackable_ctx.cuh
|
|
cuda/experimental/__stf/utility/unittest.cuh
|
|
)
|
|
|
|
cccl_get_cudatoolkit()
|
|
|
|
## cudax_add_stf_test
|
|
#
|
|
# Add an stf test executable and register it with ctest.
|
|
#
|
|
# target_name_var: Variable name to overwrite with the name of the test
|
|
# target. Useful for adding target information after creation.
|
|
# source: The source file for the test.
|
|
#
|
|
# Additional args are passed to cudax_stf_configure_target.
|
|
function(cudax_add_stf_test target_name_var source)
|
|
get_filename_component(dir ${source} DIRECTORY)
|
|
get_filename_component(filename ${source} NAME_WE)
|
|
if (dir)
|
|
set(filename "${dir}/${filename}")
|
|
endif()
|
|
string(REPLACE "/" "." test_name "${filename}")
|
|
|
|
set(test_target cudax.test.stf.${test_name})
|
|
|
|
cccl_add_executable(${test_target} SOURCES ${source} ADD_CTEST)
|
|
cudax_stf_configure_target(${test_target} ${ARGN})
|
|
target_link_libraries(${test_target} PRIVATE cudax.compiler_interface)
|
|
|
|
set(${target_name_var} ${test_target} PARENT_SCOPE)
|
|
endfunction()
|
|
|
|
## cudax_add_stf_unittest_header
|
|
#
|
|
# Add an stf unittested header executable and register it with ctest.
|
|
#
|
|
# Unittested headers contain a set of tests that are enabled by including
|
|
# `unittest.cuh` and defining `UNITTESTED_FILE`.
|
|
#
|
|
# target_name_var: Variable name to overwrite with the name of the test
|
|
# target. Useful for adding target information after creation.
|
|
# source: The source file for the test.
|
|
#
|
|
# Additional args are passed to cudax_stf_configure_target.
|
|
function(cudax_add_stf_unittest_header target_name_var source)
|
|
get_filename_component(relative_path ${source} DIRECTORY)
|
|
get_filename_component(filename ${source} NAME_WE)
|
|
|
|
string(
|
|
REPLACE
|
|
"cuda/experimental/"
|
|
""
|
|
test_label
|
|
"${relative_path}/${filename}"
|
|
)
|
|
string(REPLACE "/" "." test_label "${test_label}")
|
|
|
|
set(test_target "cudax.test.stf.unittest_headers.${test_label}")
|
|
|
|
# Pass the full path to configure_file (this is configured from cudax/tests/stf/
|
|
get_filename_component(
|
|
source_full_path
|
|
../../../cudax/include/${source}
|
|
ABSOLUTE
|
|
)
|
|
set(source ${source_full_path})
|
|
|
|
set(ut_template "${cudax_SOURCE_DIR}/cmake/stf_header_unittest.in.cu")
|
|
set(ut_source "${cudax_BINARY_DIR}/unittest_headers/${test_target}.cu")
|
|
configure_file(${ut_template} ${ut_source} @ONLY)
|
|
|
|
cccl_add_executable(${test_target} SOURCES ${ut_source} ADD_CTEST)
|
|
cudax_stf_configure_target(${test_target} ${ARGN})
|
|
target_link_libraries(${test_target} PRIVATE cudax.compiler_interface)
|
|
|
|
set(${target_name_var} ${test_target} PARENT_SCOPE)
|
|
endfunction()
|
|
|
|
# Basic tests:
|
|
foreach (source IN LISTS stf_test_sources)
|
|
cudax_add_stf_test(test_target "${source}")
|
|
endforeach()
|
|
|
|
if (cudax_ENABLE_CUDASTF_CODE_GENERATION)
|
|
foreach (source IN LISTS stf_test_codegen_sources)
|
|
cudax_add_stf_test(test_target "${source}")
|
|
endforeach()
|
|
endif()
|
|
|
|
# Tests with mathlib deps:
|
|
if (cudax_ENABLE_CUDASTF_MATHLIBS)
|
|
foreach (source IN LISTS stf_test_mathlib_sources)
|
|
cudax_add_stf_test(test_target "${source}" LINK_MATHLIBS)
|
|
endforeach()
|
|
endif()
|
|
|
|
# Unittested headers
|
|
foreach (source IN LISTS stf_unittested_headers)
|
|
cudax_add_stf_unittest_header(test_target "${source}")
|
|
endforeach()
|
|
|
|
add_subdirectory(static_error_checks)
|