SYCL Throughput Fix, main branch (2026.05.06.)

[krasznaa]

I found an oversight from #1272. The detector description is not being copied correctly in traccc::sycl::full_chain_algorithm. 😦

Unfortunately this is not great news though. Because I was hoping that the crashes that I was getting very reproducibly with traccc_throughput_mt_sycl, would shed a light on the very rare and random crashes that I was seeing with traccc_throughput_mt_cuda yesterday...

Event processing   [==========>                                       ] 21% [00m:05s] terminate called after throwing an instance of 'vecmem::cuda::runtime_error'
terminate called recursively
  what():  /shared/measurements_2026_05_05/build_main/_deps/vecmem-src/cuda/src/utils/cuda/async_copy.cpp:76 Failed to execute: cudaEventSynchronize(m_event) (an illegal memory access was encountered)
../traccc/extras/traccc_odd_throughput_mt_profiler.sh: line 138: 102737 Aborted                 (core dumped) ${TRACCC_EXECUTABLE} --read-bfield-from-file --bfield-file=geometries/odd/odd-bfield.cvf --input-directory="${TRACCC_INPUT_DIR}/geant4_${EVTDIR}/" --input-events=500 --cpu-threads=${NTHREAD} --cold-run-events=$((5*${NTHREAD})) --processed-events=$((${TRACCC_EVT_COUNT[${EVTDIR}]}*${NTHREAD})) --track-candidates-range=5:100 --seedfinder-vertex-range=-150:150 --max-num-branches-per-seed=2 --initial-links-per-seed=10 --deterministic=1 --log-file="${TRACCC_CSV_FILE}"

As I told @stephenswat today, these may be a bit hard to debug unfortunately. As I had to run on a lot of events to see a couple of crashes. 🤔

Oh, well. At least the easy problem is fixed with this PR.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
50.0% Duplication on New Code

See analysis details on SonarQube Cloud