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

Daint: change cuda module #1553

Merged
merged 7 commits into from
Sep 25, 2020
Merged

Daint: change cuda module #1553

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
33daf66
Update daint.sh
havogt Sep 24, 2020
236858c
try to make Clang happy
havogt Sep 24, 2020
86302bc
Update loops.hpp
havogt Sep 24, 2020
e54f50d
Update daint_cray.sh
havogt Sep 25, 2020
0b54b26
disable hugepage test for cray
havogt Sep 25, 2020
675310b
update references
havogt Sep 25, 2020
7858015
Merge remote-tracking branch 'upstream/cuda-on-daint' into cuda-on-daint
havogt Sep 25, 2020
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
4 changes: 4 additions & 0 deletions include/gridtools/common/hugepage_alloc.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,10 @@ namespace gridtools {
* reduce cache set conflicts.
*/
inline void *hugepage_alloc(std::size_t size) {
#ifdef __cray__
// see issue https://github.com/GridTools/gridtools/issues/1557
#warning "hugepage_alloc on Cray might be suboptimal"
#endif
static std::atomic<std::size_t> s_offset(64);
auto offset = s_offset.load(std::memory_order_relaxed);
auto next_offset = offset;
Expand Down
12 changes: 4 additions & 8 deletions include/gridtools/stencil/cpu_ifirst/loops.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,13 +28,7 @@ namespace gridtools {
namespace loops_impl_ {
template <class Stage, class Ptr, class Strides>
GT_FORCE_INLINE void i_loop(int_t size, Stage stage, Ptr &ptr, Strides const &strides) {
#ifdef NDEBUG
// TODO(anstaf & fthaler):
// Maybe we have to re-run tests with different combinations of pragmas on different compilers,
// the current set of pragmas is at the border of legality for the present code, so maybe we can find a better option.
#pragma ivdep
#pragma omp simd
#endif
for (int_t i = 0; i < size; ++i) {
using namespace literals;
stage(ptr, strides);
Expand Down Expand Up @@ -109,7 +103,8 @@ namespace gridtools {
int_t i_blocks = info.i_blocks();
int_t j_blocks = info.j_blocks();
int_t k_size = grid.k_size();
thread_pool::parallel_for_loop(ThreadPool(),
thread_pool::parallel_for_loop(
ThreadPool(),
[&](auto i, auto k, auto j) {
tuple_util::for_each([block = info.block(i, j, k)](auto &&loop) { loop(block); }, loops);
},
Expand Down Expand Up @@ -157,7 +152,8 @@ namespace gridtools {
template <class ThreadPool, class Grid, class Loops>
void run_loops(std::false_type, Grid const &grid, Loops loops) {
execinfo info(ThreadPool(), grid);
thread_pool::parallel_for_loop(ThreadPool(),
thread_pool::parallel_for_loop(
ThreadPool(),
[&](auto i, auto j) {
tuple_util::for_each([block = info.block(i, j)](auto &&loop) { loop(block); }, loops);
},
Expand Down
2 changes: 1 addition & 1 deletion jenkins/envs/daint.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,7 @@ function module() {
}

module load daint-gpu
module load cudatoolkit/10.1.105_3.27-7.0.1.1_4.1__ga311ce7
module load cudatoolkit
module rm CMake
module load /users/jenkins/easybuild/daint/haswell/modules/all/CMake/3.14.5

Expand Down
2 changes: 0 additions & 2 deletions jenkins/envs/daint_cray.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,5 +11,3 @@ export GTCMAKE_CMAKE_CUDA_HOST_COMPILER="$CXX"

export CUDAHOSTCXX="$CXX"
export CTEST_PARALLEL_LEVEL=1
export CXXFLAGS='-fno-cray-gpu -fno-cray'
export CFLAGS='-fno-cray-gpu -fno-cray-mallopt -fno-cray'
Loading