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.

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

[Refactor] CI job generator: define in the version.py which backend combinations should be tested #2099

Merged
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
19 changes: 17 additions & 2 deletions script/job_generator/alpaka_filter.py
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,12 @@

from alpaka_job_coverage.globals import * # pylint: disable=wildcard-import,unused-wildcard-import
from alpaka_globals import * # pylint: disable=wildcard-import,unused-wildcard-import
from alpaka_job_coverage.util import row_check_name, row_check_version, is_in_row
from alpaka_job_coverage.util import (
row_check_name,
row_check_version,
is_in_row,
row_check_backend_version,
)


def alpaka_post_filter(row: List) -> bool:
Expand All @@ -22,7 +27,7 @@ def alpaka_post_filter(row: List) -> bool:
for clang_cuda_version in ["15", "16"]:
if row_check_version(row, HOST_COMPILER, "==", clang_cuda_version):
return False

# Debug builds with nvcc <= 11.6 produce compiler errors
if (
is_in_row(row, BUILD_TYPE)
Expand All @@ -40,4 +45,14 @@ def alpaka_post_filter(row: List) -> bool:
):
return False

# OpenMP is not supported for clang as cuda compiler
# https://github.com/alpaka-group/alpaka/issues/639
if row_check_name(row, DEVICE_COMPILER, "==", CLANG_CUDA) and (
row_check_backend_version(row, ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE, "==", ON_VER)
or row_check_backend_version(
row, ALPAKA_ACC_CPU_B_SEQ_T_OMP2_ENABLE, "==", ON_VER
)
):
return False

return True
45 changes: 24 additions & 21 deletions script/job_generator/generate_job_yaml.py
Original file line number Diff line number Diff line change
Expand Up @@ -218,6 +218,23 @@ def verify_image(
return "".join(verified_container_url)


@typechecked
def append_backend_variables(
variables: Dict[str, str], job: Dict[str, Tuple[str, str]]
):
"""Searches for enabled back-ends in the job parameters and appends the back-end
variable to variables to enable it in the CI job.

Args:
variables (Dict[str, str]): variables of the CI job
job (Dict[str, Tuple[str, str]]): job with parameters from the pair-wise
generator.
"""
for backend in BACKENDS_LIST:
if backend in job:
variables[backend] = "ON"


@typechecked
def job_variables(job: Dict[str, Tuple[str, str]]) -> Dict[str, str]:
"""Add variables to the job depending of the job dict.
Expand Down Expand Up @@ -270,14 +287,12 @@ def job_variables(job: Dict[str, Tuple[str, str]]) -> Dict[str, str]:
variables["ALPAKA_CI_CMAKE_VER"] = job[CMAKE][VERSION]
variables["ALPAKA_BOOST_VERSION"] = job[BOOST][VERSION]

# TODO(SimeonEhrig) at the moment, not all backends are available in
# versions.sw_versions[BACKENDS], therefore we have to set each backend explicit
# Later, we can iterate over versions.sw_versions[BACKENDS] and check, if a Backend needs to be
# enabled
# all back-ends are disabled by default
# back-ends are conditionally enabled depending on the job parameters
variables[ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE] = "OFF"
variables[ALPAKA_ACC_CPU_B_SEQ_T_FIBERS_ENABLE] = "OFF"
variables[ALPAKA_ACC_CPU_B_SEQ_T_OMP2_ENABLE] = "OFF"
variables[ALPAKA_ACC_CPU_B_SEQ_T_SEQ_ENABLE] = "ON"
variables[ALPAKA_ACC_CPU_B_SEQ_T_SEQ_ENABLE] = "OFF"
variables[ALPAKA_ACC_CPU_B_SEQ_T_THREADS_ENABLE] = "OFF"
variables[ALPAKA_ACC_CPU_B_TBB_T_SEQ_ENABLE] = "OFF"
variables[ALPAKA_ACC_GPU_CUDA_ENABLE] = "OFF"
Expand All @@ -290,8 +305,9 @@ def job_variables(job: Dict[str, Tuple[str, str]]) -> Dict[str, str]:
else:
variables["ALPAKA_TEST_MDSPAN"] = "OFF"

append_backend_variables(variables, job)

if job[DEVICE_COMPILER][NAME] == HIPCC:
variables[ALPAKA_ACC_GPU_HIP_ENABLE] = "ON"
variables["CC"] = "clang"
variables["CXX"] = "clang++"
variables["GPU_TARGETS"] = "${CI_GPU_ARCH}"
Expand Down Expand Up @@ -322,7 +338,6 @@ def job_variables(job: Dict[str, Tuple[str, str]]) -> Dict[str, str]:
ALPAKA_ACC_GPU_CUDA_ENABLE in job
and job[ALPAKA_ACC_GPU_CUDA_ENABLE][VERSION] != OFF_VER
):
variables[ALPAKA_ACC_GPU_CUDA_ENABLE] = "ON"
variables["ALPAKA_CI_STDLIB"] = "libstdc++"
variables["CMAKE_CUDA_ARCHITECTURES"] = job[SM_LEVEL][VERSION]
variables["ALPAKA_CI_CUDA_VERSION"] = job[ALPAKA_ACC_GPU_CUDA_ENABLE][VERSION]
Expand All @@ -336,16 +351,11 @@ def job_variables(job: Dict[str, Tuple[str, str]]) -> Dict[str, str]:
variables["CC"] = "gcc"
variables["CXX"] = "g++"
variables["ALPAKA_CI_GCC_VER"] = job[HOST_COMPILER][VERSION]
variables[ALPAKA_ACC_CPU_B_SEQ_T_THREADS_ENABLE] = "ON"
variables[ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE] = "ON"
variables[ALPAKA_ACC_CPU_B_SEQ_T_OMP2_ENABLE] = "ON"
# configuration, if Clang is the CUDA host compiler
elif job[HOST_COMPILER][NAME] == CLANG:
variables["CC"] = "clang"
variables["CXX"] = "clang++"
variables["ALPAKA_CI_CLANG_VER"] = job[HOST_COMPILER][VERSION]
variables[ALPAKA_ACC_CPU_B_SEQ_T_THREADS_ENABLE] = "ON"
variables[ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE] = "ON"
else:
raise RuntimeError(
"generate_job_yaml.job_variables(): unknown CUDA host compiler: "
Expand All @@ -357,9 +367,6 @@ def job_variables(job: Dict[str, Tuple[str, str]]) -> Dict[str, str]:
variables["CXX"] = "clang++"
variables["ALPAKA_CI_CLANG_VER"] = job[DEVICE_COMPILER][VERSION]
variables["CMAKE_CUDA_COMPILER"] = "clang++"
variables[ALPAKA_ACC_CPU_B_SEQ_T_THREADS_ENABLE] = "ON"
variables[ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE] = "OFF"
variables[ALPAKA_ACC_CPU_B_SEQ_T_OMP2_ENABLE] = "OFF"

# oneAPI configuration
if job[DEVICE_COMPILER][NAME] == ICPX:
Expand All @@ -371,12 +378,8 @@ def job_variables(job: Dict[str, Tuple[str, str]]) -> Dict[str, str]:
variables["ALPAKA_CI_CLANG_VER"] = "16"
variables["ALPAKA_CI_STDLIB"] = "libstdc++"
variables["ALPAKA_CI_ONEAPI_VERSION"] = job[DEVICE_COMPILER][VERSION]
variables[ALPAKA_ACC_SYCL_ENABLE] = "ON"
variables["alpaka_SYCL_ONEAPI_CPU"] = "ON"
variables["alpaka_SYCL_ONEAPI_CPU_ISA"] = "avx2"
variables[ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE] = "OFF" # Turn off OpenMP back-ends until Intel fixes https://github.com/intel/llvm/issues/10711
variables[ALPAKA_ACC_CPU_B_SEQ_T_OMP2_ENABLE] = "OFF"
variables[ALPAKA_ACC_CPU_B_TBB_T_SEQ_ENABLE] = "ON"

return variables

Expand Down Expand Up @@ -410,7 +413,7 @@ def job_tags(job: Dict[str, Tuple[str, str]]) -> List[str]:
and job[ALPAKA_ACC_SYCL_ENABLE][VERSION] != OFF_VER
):
return ["x86_64", "cpuonly"]

# fallback
return ["x86_64", "cpuonly"]

Expand Down Expand Up @@ -448,7 +451,7 @@ def global_variables() -> Dict[str, str]:
variables["ALPAKA_CI_BOOST_LIB_DIR"] = "$HOME/boost_libs"
variables["ALPAKA_CI_CUDA_DIR"] = "$HOME/cuda"
variables["ALPAKA_CI_HIP_ROOT_DIR"] = "$HOME/hip"
variables["alpaka_ENABLE_WERROR"] ="ON"
variables["alpaka_ENABLE_WERROR"] = "ON"

return variables

Expand Down
2 changes: 1 addition & 1 deletion script/job_generator/requirements.txt
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
alpaka-job-coverage == 1.5.2
alpaka-job-coverage == 1.5.3
allpairspy == 2.5.0
typeguard < 3.0.0
pyaml
Expand Down
76 changes: 59 additions & 17 deletions script/job_generator/versions.py
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,6 @@
from alpaka_globals import * # pylint: disable=wildcard-import,unused-wildcard-import


# TODO: only an example
sw_versions: Dict[str, List[str]] = {
GCC: ["9", "10", "11", "12", "13"],
CLANG: ["9", "10", "11", "12", "13", "14", "15", "16"],
Expand All @@ -30,11 +29,40 @@
],
HIPCC: ["5.0", "5.1", "5.2", "5.3", "5.4", "5.5"],
ICPX: ["2023.1.0", "2023.2.0"],
# Contains all enabled back-ends.
# There are special cases for ALPAKA_ACC_GPU_CUDA_ENABLE and ALPAKA_ACC_GPU_HIP_ENABLE
# which have to be combined with nvcc and hipcc versions.
BACKENDS: [
# ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE,
ALPAKA_ACC_GPU_CUDA_ENABLE,
ALPAKA_ACC_GPU_HIP_ENABLE,
ALPAKA_ACC_SYCL_ENABLE,
# ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE,
# nvcc
[
ALPAKA_ACC_CPU_B_SEQ_T_SEQ_ENABLE,
ALPAKA_ACC_CPU_B_SEQ_T_THREADS_ENABLE,
ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE,
ALPAKA_ACC_CPU_B_SEQ_T_OMP2_ENABLE,
ALPAKA_ACC_GPU_CUDA_ENABLE,
],
# clang-cuda
# OpenMP is not supported for clang as cuda compiler
# https://github.com/alpaka-group/alpaka/issues/639
# therefore a extra combination is required
[
ALPAKA_ACC_CPU_B_SEQ_T_SEQ_ENABLE,
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

How can I see that this is for clang-cuda except by the comment? So I am wondering how the generator know that for clang-cuda this configuration should be used and not the first entry in BACKENDS?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You cannot see it directly in code, because it is decided by filter rules. So clang-cuda requires ALPAKA_ACC_GPU_CUDA_ENABLE and does not allow the OMP2 backends:

# OpenMP is not supported for clang as cuda compiler
# https://github.com/alpaka-group/alpaka/issues/639
if row_check_name(row, DEVICE_COMPILER, "==", CLANG_CUDA) and (
row_check_backend_version(row, ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE, "==", ON_VER)
or row_check_backend_version(
row, ALPAKA_ACC_CPU_B_SEQ_T_OMP2_ENABLE, "==", ON_VER
)
):
return False

You can check if a compiler backend passes the filter with ajc-validate tool (except the alpaka filter, I need to implement support for this 🤔 ).

ALPAKA_ACC_CPU_B_SEQ_T_THREADS_ENABLE,
ALPAKA_ACC_GPU_CUDA_ENABLE,
],
# hip
[
ALPAKA_ACC_CPU_B_SEQ_T_SEQ_ENABLE,
ALPAKA_ACC_GPU_HIP_ENABLE,
],
# sycl (oneAPI)
# Turn off OpenMP back-ends until Intel fixes https://github.com/intel/llvm/issues/10711
[
ALPAKA_ACC_CPU_B_SEQ_T_SEQ_ENABLE,
ALPAKA_ACC_CPU_B_TBB_T_SEQ_ENABLE,
ALPAKA_ACC_SYCL_ENABLE,
],
],
UBUNTU: ["20.04"],
CMAKE: ["3.22.6", "3.23.5", "3.24.4", "3.25.3", "3.26.4"],
Expand Down Expand Up @@ -85,24 +113,38 @@ def get_compiler_versions(clang_cuda: bool = True) -> List[Tuple[str, str]]:

@typechecked
def get_backend_matrix() -> List[List[Tuple[str, str]]]:
"""Generate backend list, where only backend is active on the same time.
"""Generates a back-end list which contains different combinations of enabled back-ends.

Returns:
List[List[Tuple[str, str]]]: The backend list.
"""
combination_matrix: List[List[Tuple[str, str]]] = []

# TODO(SimeonEhrig) only working for HIP in the moment
if HIPCC in sw_versions:
for rocm_version in sw_versions[HIPCC]:
combination_matrix.append([(ALPAKA_ACC_GPU_HIP_ENABLE, rocm_version)])

if NVCC in sw_versions:
for cuda_version in sw_versions[NVCC]:
combination_matrix.append([(ALPAKA_ACC_GPU_CUDA_ENABLE, cuda_version)])

if ICPX in sw_versions:
combination_matrix.append([(ALPAKA_ACC_SYCL_ENABLE, ON_VER)])
for backend_list in sw_versions[BACKENDS]:
combination: List[Tuple[str, str]] = []
# all back-ends except ALPAKA_ACC_GPU_HIP_ENABLE and
# ALPAKA_ACC_GPU_CUDA_ENABLE can be enabled without having a version number attached to them
for backend_name in backend_list:
if backend_name not in [
ALPAKA_ACC_GPU_HIP_ENABLE,
ALPAKA_ACC_GPU_CUDA_ENABLE,
]:
combination.append((backend_name, ON_VER))

# add a back-end entry for each CUDA SDK version
if ALPAKA_ACC_GPU_CUDA_ENABLE in backend_list:
for cuda_version in sw_versions[NVCC]:
combination_copy = combination[:]
combination_copy.append((ALPAKA_ACC_GPU_CUDA_ENABLE, cuda_version))
combination_matrix.append(combination_copy)
# add a back-end entry for each HIP SDK version
elif ALPAKA_ACC_GPU_HIP_ENABLE in backend_list:
for rocm_version in sw_versions[HIPCC]:
combination_copy = combination[:]
combination_copy.append((ALPAKA_ACC_GPU_HIP_ENABLE, rocm_version))
combination_matrix.append(combination_copy)
else:
combination_matrix.append(combination)

return combination_matrix

Expand Down