distributed-ranges

distributed-ranges copied to clipboard

BenBrock

Sort performance optimisation with SYCL

1

mateuszpn

haichangsi

lslusarczyk

the dist_sort() function is too long to be clear, split it in reasonable way

mateuszpn

Add missing siugnatures, test them, then remove ifdef in below `sort` code of sort benchmarking. ``` #ifdef BENCH_SHP xhp::copy(local.begin(), local.end(), rng::begin(*a)); #else xhp::copy(local, rng::begin(*a)); #endif ```

haichangsi

use this function here: https://github.com/oneapi-src/distributed-ranges/pull/545#discussion_r1366677499

haichangsi

When running mhp-bench one can see the message `***WARNING*** Library was built as DEBUG. Timings may be affected.` even if compiled in our cmake based Release mode. This is because...

lslusarczyk

Multi-node tests on Borealis

1

lslusarczyk

(on cpu) Example with sycl version: ``` mpirun -n 112 ./shallow_water --sycl -f -t -n 1024 shallow_water: [...]/distributed-ranges/include/./dr/detail/sycl_utils.hpp:124: sycl::event dr::__detail::parallel_for(sycl::queue &, sycl::range, Fn &&) [Fn = (lambda at [...]/distributed-ranges/include/./dr/mhp/algorithms/md_for_each.hpp:142:29) &]:...

lslusarczyk