algo: Optimize host code in GPU Hitfinder.

Introduces three optimizations to the host code of the STS Hitfinder:

• Allocate all buffers beforehand on GPU, do dynamic allocation per TS on CPU (allocating the pinned memory for GPU is very slow)
• Speedup sorting Digis by module by using a lookup table to map the sensor address to an index and parallelizing the copy
• Copy hits to pinned memory to speed up copy from the GPU

added Online Reconstruction labels

added 9 commits

• fcf0c97a - 1 commit from branch computing:master
• 307483a4 - Allocate memory beforehand on GPU in STS reco.
• e6d4e9ad - cleanup.
• 25b0adf4 - Flatten Digis in parallel.
• e6787dc4 - Optimize Digi sorting by sensors.
• 6172bda0 - Sort Digis in parallel into modules.
• 1b09376b - clang-format
• ef1db7f8 - cleanup.
• 144c7a29 - Handle exception from Hitfinder.

Compare with previous version

added 9 commits

• 77a3e4d9 - Allocate memory beforehand on GPU in STS reco.
• 3cb98478 - cleanup.
• 107dbdea - Flatten Digis in parallel.
• 4f9fb2ed - Optimize Digi sorting by sensors.
• 9f0953aa - Sort Digis in parallel into modules.
• e5db4a35 - clang-format
• 8ef40c18 - cleanup.
• 40c888ba - Handle exception from Hitfinder.
• 28dc9482 - clang-format.

Compare with previous version

added 1 commit

• bc457ede - fix compilation.

Compare with previous version

Dear @f.uhlig, @v.friese, @p.-a.loizeau,

you have been identified as code owner of at least one file which was changed with this merge request.

Please check the changes and approve them or request changes.

added CodeOwners label

requested review from @v.friese

assigned to @fweig

added 39 commits

• bc457ede...97569d6d - 29 commits from branch computing:master
• ed5bc9db - Allocate memory beforehand on GPU in STS reco.
• 0e1a4053 - cleanup.
• 14b0b93f - Flatten Digis in parallel.
• 2688a115 - Optimize Digi sorting by sensors.
• 3a4419b7 - Sort Digis in parallel into modules.
• 07e6efce - clang-format
• 42483702 - cleanup.
• fe3f183c - Handle exception from Hitfinder.
• e2341bb1 - clang-format.
• 722c8aea - fix compilation.

Compare with previous version

added 14 commits

• 7147c3e6 - 1 commit from branch computing:master
• 7147c3e6...a58fb809 - 3 earlier commits
• 1fd2c894 - Flatten Digis in parallel.
• cdd7d2c7 - Optimize Digi sorting by sensors.
• 17fd85ea - Sort Digis in parallel into modules.
• f6e34d7b - clang-format
• d91d4421 - cleanup.
• d06ca8ef - Handle exception from Hitfinder.
• ac499bb0 - clang-format.
• 977bc497 - fix compilation.
• 3c331a56 - algo: Add PartitionedSpan.
• 6c1b8aa0 - Return span to hits.

Compare with previous version

changed the description

added 18 commits

• 6c1b8aa0...51878292 - 4 commits from branch computing:master
• 51878292...37b09997 - 4 earlier commits
• 7518a58c - Optimize Digi sorting by sensors.
• 6455b8a5 - Sort Digis in parallel into modules.
• 64287af1 - clang-format
• daf320af - cleanup.
• 6b8b9c25 - Handle exception from Hitfinder.
• ca27330f - clang-format.
• 9850d050 - fix compilation.
• aa663a7e - algo: Add PartitionedSpan.
• b0260bbb - Return span to hits.
• c11f7fa5 - Terminate faster when monster events are encountered.

Compare with previous version

added 1 commit

• e99e7262 - clang-format

Compare with previous version

added 1 commit

• dd2bd55e - Only copy the actual number of Digis to GPU.

Compare with previous version

added 1 commit

• 792bc785 - Only copy the actual number of Digis to GPU.

Compare with previous version

added 21 commits

• 792bc785...a9c1a566 - 4 commits from branch computing:master
• a9c1a566...bb4ba3e8 - 7 earlier commits
• a260da9b - cleanup.
• 1629532f - Handle exception from Hitfinder.
• c38b7cd1 - clang-format.
• 429506f5 - fix compilation.
• d01c8a31 - algo: Add PartitionedSpan.
• 806e2ae3 - Return span to hits.
• 40d7c2d2 - Terminate faster when monster events are encountered.
• 1eed289d - clang-format
• 8df4ac85 - Only copy the actual number of Digis to GPU.
• 48d1baee - Fix clang errors.

Compare with previous version

approved this merge request

merged

mentioned in merge request !1411 (merged)