(Downloads - 0)
For more info about our services contact : help@bestpfe.com
Table of contents
1 The LHCb experiment
1.1 Introduction
1.2 The Large Hadron Collider
1.3 The LHCb detector
1.3.1 Vertex Locator
1.3.2 Upstream Tracker
1.3.3 Scintillating Fibre Tracker
1.3.4 Ring Imaging Cherenkov Detectors
1.3.5 Calorimeters
1.3.6 Muon stations
1.4 The High Level Trigger
1.4.1 Event Building
1.4.2 HLT1
1.4.3 HLT2
1.4.4 LHCb Software Framework
1.5 Conclusion
2 Parallelism on CPU
2.1 Introduction
2.2 CPU Architecture
2.2.1 Multi-core architectures
2.2.2 Cache Level Hierarchy
2.2.3 Data Layout
2.3 Single Instruction Multiple Data
2.3.1 Instruction sets
2.3.2 SIMD speedup and frequency scaling
2.4 The SIMDWrappers library
2.4.1 Design objectives
2.4.2 Comparison with other SIMD libraries
2.4.3 Instruction emulation
2.5 Conclusion
3 Parallelism on GPU
3.1 Introduction
3.2 From arcade video games to HPC
3.3 CUDA programming model
3.4 Grid-stride loops
3.5 Shared memory optimisations
3.6 Warp-level programming
3.7 Conclusion
4 Connected Component Analysis
4.1 Introduction
4.2 Connected Component Labeling and Analysis
4.2.1 One component at a time
4.2.2 Multi-pass iterative algorithms
4.2.3 Direct two-pass algorithms
4.2.4 Mask topology: blocks and segments
4.3 HA4: Hybrid pixel/segment CCL for GPU
4.3.1 Strip labeling
4.3.2 Border Merging
4.3.3 CCL – Final labeling
4.3.4 CCA and Feature Computation
4.3.5 Processing two pixels per thread
4.3.6 Experimental Evaluation
4.4 FLSL: Faster LSL for GPU
4.4.1 Full segments (FLSL)
4.4.2 On-The-Fly feature merge (OTF)
4.4.3 Conict detection (CD)
4.4.4 Number of updates and conicts
4.4.5 Experimental Evaluation
4.5 SIMD Rosenfeld
4.5.1 SIMD Union-Find
4.5.2 SIMD Rosenfeld pixel algorithm
4.5.3 SIMD Rosenfeld sub-segment algorithm
4.5.4 Multi-thread SIMD algorithms
4.5.5 Experimental Evaluation
4.6 SparseCCL
4.6.1 General parameterizable ordered SparseCCL
4.6.2 Acceleration structure for un-ordered pixels
4.6.3 Case study: specialization for LHCb VELO Upgrade
4.6.4 Experimental Evaluation
4.7 Conclusion
5 VELO reconstruction algorithm
5.1 Introduction
5.2 Tracking algorithms
5.3 Evolution of the VELO detector and algorithms
5.3.1 Reconstruction in the Run 1 and 2 VELO detector
5.3.2 Reconstruction of the upgraded VELO detector
5.4 SIMD Velo reconstruction
5.4.1 Structure of the algorithm
5.4.2 Seeding tracks
5.4.3 Extending tracks
5.4.4 Numerical precision
5.5 Benchmarks
5.5.1 Throughput
5.5.2 Reconstruction physics eciency
5.6 Conclusion
6 Scalability of the LHCb software
6.1 Introduction
6.2 Evaluation of HLT1 on CPUs
6.3 Evaluation of HLT2 on CPUs
6.4 Conclusion
Conclusion
