Skip to content

SYCL: Turn on AOT in CI and Update MKL #6007

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

Merged
merged 4 commits into from
Jul 9, 2025
+15 −13
Merged

SYCL: Turn on AOT in CI and Update MKL #6007

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
f7c56b7
SYCL CI: Turn on AOT
WeiqunZhang Jul 8, 2025
8e7ce57
Update mkl calls
WeiqunZhang Jul 9, 2025
625be8e
Split the big kernel in the implicit solver into three.
WeiqunZhang Jul 9, 2025
9e459e3
Intel CI: Remove some workarounds
WeiqunZhang Jul 9, 2025
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
8 changes: 4 additions & 4 deletions .github/workflows/dependencies/dpcpp.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,16 +29,16 @@ df -h
# https://github.com/BLAST-WarpX/warpx/pull/1566#issuecomment-790934878

# try apt install up to five times, to avoid connection splits
# FIXME install latest version of IntelLLVM, Intel MKL
# after conflicts with openPMD are resolved
Copy link
Member Author

@WeiqunZhang WeiqunZhang Jul 9, 2025

Choose a reason for hiding this comment

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

@ax3l I updated the code that uses MKL FFT in this PR. What were the conflicts with openPMD?

Copy link
Member Author

@WeiqunZhang WeiqunZhang Jul 9, 2025

Choose a reason for hiding this comment

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

Okay it was added in #5419. Since there is no error now, it is safe to remove it.

Copy link
Member Author

@WeiqunZhang WeiqunZhang Jul 9, 2025

Choose a reason for hiding this comment

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

We can also try to remove -O1 -NDEBUG, which was added in #3478 to an address oneAPI 2022.2.0 issue.

status=1
for itry in {1..5}
do
sudo apt-get install -y --no-install-recommends \
build-essential \
cmake \
intel-oneapi-compiler-dpcpp-cpp=2024.2.1-1079 \
intel-oneapi-mkl-devel=2024.2.1-103 \
intel-oneapi-compiler-dpcpp-cpp \
intel-oneapi-mkl-devel \
intel-ocloc \
libigc-dev \
g++ gfortran \
libopenmpi-dev \
openmpi-bin \
Expand Down
7 changes: 3 additions & 4 deletions .github/workflows/intel.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,15 +53,13 @@ jobs:
set +e
source /opt/intel/oneapi/setvars.sh
set -e
export PATH=$PATH:/opt/intel/oneapi/compiler/2024.2/bin # FIXME
export CXX=$(which icpx)
export CC=$(which icx)

python3 -m pip install --upgrade pip
python3 -m pip install --upgrade build packaging setuptools[core] wheel

cmake -S . -B build_sp \
-DCMAKE_CXX_FLAGS_RELEASE="-O1 -DNDEBUG" \
-DCMAKE_VERBOSE_MAKEFILE=ON \
-DWarpX_EB=OFF \
-DWarpX_PYTHON=ON \
Expand Down Expand Up @@ -119,16 +117,17 @@ jobs:
set +e
source /opt/intel/oneapi/setvars.sh
set -e
export PATH=$PATH:/opt/intel/oneapi/compiler/2024.2/bin # FIXME
export CXX=$(which icpx)
export CC=$(which icx)
export CXXFLAGS="-fsycl ${CXXFLAGS}"

cmake -S . -B build_sp \
-DCMAKE_CXX_FLAGS_RELEASE="-O1 -DNDEBUG" \
-DBUILD_SHARED_LIBS=ON \
-DCMAKE_VERBOSE_MAKEFILE=ON \
-DWarpX_COMPUTE=SYCL \
-DAMReX_SYCL_AOT=ON \
-DAMReX_INTEL_ARCH=pvc \
-DAMReX_PARALLEL_LINK_JOBS=4 \
-DWarpX_EB=ON \
-DWarpX_FFT=ON \
-DWarpX_PYTHON=ON \
Expand Down
6 changes: 4 additions & 2 deletions Source/FieldSolver/ImplicitSolvers/ImplicitSolver.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -280,7 +280,8 @@ void ImplicitSolver::ComputeJfromMassMatrices ()
}

Jx(i,j,k,n) = Jx0(i,j,k,n) + SxxdEx + SxydEy + SxzdEz;
},
});
amrex::ParallelFor(
Jby, ncomps, [=] AMREX_GPU_DEVICE (int i, int j, int k, int n)
{
const int idx[3] = {i, j, k};
Expand Down Expand Up @@ -342,7 +343,8 @@ void ImplicitSolver::ComputeJfromMassMatrices ()
}

Jy(i,j,k,n) = Jy0(i,j,k,n) + SyxdEx + SyydEy + SyzdEz;
},
});
amrex::ParallelFor(
Jbz, ncomps, [=] AMREX_GPU_DEVICE (int i, int j, int k, int n)
{
const int idx[3] = {i, j, k};
Expand Down
2 changes: 1 addition & 1 deletion Source/ablastr/math/fft/AnyFFT.H
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@
# endif
# include <hip/hip_complex.h>
# elif defined(AMREX_USE_SYCL)
# include <oneapi/mkl/dfti.hpp>
# include <oneapi/mkl/dft.hpp>
# else
# include <fftw3.h>
# endif
Expand Down
5 changes: 3 additions & 2 deletions Source/ablastr/math/fft/WrapMklFFT.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,7 @@
#include "ablastr/utils/TextMsg.H"
#include "ablastr/profiler/ProfilerWrapper.H"

#include <complex>
#include <cstdint>

namespace ablastr::math::anyfft
Expand Down Expand Up @@ -53,9 +54,9 @@ namespace ablastr::math::anyfft
}

fft_plan.m_plan->set_value(oneapi::mkl::dft::config_param::PLACEMENT,
DFTI_NOT_INPLACE);
oneapi::mkl::dft::config_value::NOT_INPLACE);
fft_plan.m_plan->set_value(oneapi::mkl::dft::config_param::FWD_STRIDES,
strides.data());
strides);
fft_plan.m_plan->commit(amrex::Gpu::Device::streamQueue());

// Store meta-data in fft_plan
Expand Down
Loading