Fix instabilities in Velo CPU ci-tests (v3)

Locally it doesn't seem to fluctuate. To be tested in the ci. Some minor changes are expected due to differences between the vectorized CPU code and the GPU code. Throughput is expected to drop a bit only for the CPU test.

Drop the specialized vectorized velo seeding function for CPU.

Closes Moore#607 (closed) Moore#769 (closed)

FYI @dovombru @cagapopo @jonrob @cmarinbe

changed target branch from master to 2024-patches

added 1 commit

• 497a5569 - Remove vectorized CPU code, use same code as GPU to minimize differences

Compare with previous version

added RTA label

/ci-test

added ci-test-triggered label

• [2024-06-15 19:28] Validation started with lhcb-2024-patches-mr#777
• [2024-06-18 10:33] Validation started with lhcb-2024-patches-mr#798
• [2024-06-22 18:26] Validation started with lhcb-2024-patches-mr#859
• [2024-06-23 04:30] Validation started with lhcb-2024-patches-mr#861

mentioned in commit 4c613781

mentioned in commit Moore@cef28e3c

mentioned in merge request Moore!3577 (merged)

resolved all threads

/ci-test Moore!3577 (merged)

Thanks a lot for digging into this @ahennequ. Do I understand correctly that you suspect the vectorized CPU implementation of search by triplet to have caused the fluctuations? Might be worth adding to the description of the MR that this implementation is being dropped.

changed the description

added 31 commits

• 497a5569...8cb48390 - 30 commits from branch 2024-patches
• a3d16437 - Remove vectorized CPU code, use same code as GPU to minimize differences

Compare with previous version

marked this merge request as ready

mentioned in issue Moore#769 (closed)

added 7 commits

• a3d16437...3d3b5a8b - 6 commits from branch 2024-patches
• 36063715 - Remove vectorized CPU code, use same code as GPU to minimize differences

Compare with previous version

/ci-test Moore!3577 (merged)

mentioned in commit df787d5d

mentioned in merge request !1690 (merged)