Episode 5.3 - Optimization of Vectorization: Data Structures
Table of Contents:
00:07 - Optimization of Vectorization
00:29 - Please leave you question and comments
00:55 - Unit-stride access to data
01:02 - Load/Store operations on vector registers
01:27 - Scattered data vs contiguous structure
01:48 - Example: Coulomb's Law
01:52 - Problem statement
02:11 - Potential for vectorization
02:30 - Naive code implementation
02:44 - Code explanation
03:01 - Inafficient vectorization
03:37 - SoA vs AoS
03:56 - Advantage of unit-stride access in SoA
04:13 - Performance results
04:19 - Explaining results
04:46 - Baseline performance
04:51 - Optimized performance
05:04 - Performance with relaxed precision
05:17 - Important optimization, but may be difficult to implement
05:43 - New applications: think about data structures beforehand!
05:51 - Final words
Видео Episode 5.3 - Optimization of Vectorization: Data Structures канала Vadim Karpusenko
00:07 - Optimization of Vectorization
00:29 - Please leave you question and comments
00:55 - Unit-stride access to data
01:02 - Load/Store operations on vector registers
01:27 - Scattered data vs contiguous structure
01:48 - Example: Coulomb's Law
01:52 - Problem statement
02:11 - Potential for vectorization
02:30 - Naive code implementation
02:44 - Code explanation
03:01 - Inafficient vectorization
03:37 - SoA vs AoS
03:56 - Advantage of unit-stride access in SoA
04:13 - Performance results
04:19 - Explaining results
04:46 - Baseline performance
04:51 - Optimized performance
05:04 - Performance with relaxed precision
05:17 - Important optimization, but may be difficult to implement
05:43 - New applications: think about data structures beforehand!
05:51 - Final words
Видео Episode 5.3 - Optimization of Vectorization: Data Structures канала Vadim Karpusenko
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
Tree/Graph of my current friends likes and interests, branched by categoriesEpisode 5.1 - Optimization roadmapEpisode 3.6 - Shared Virtual MemoryEpisode 4.2 - Automatic Vectorization and Array NotationEpisode 2.3 - Vector Instruction Support in Intel ArchitecturesTech Talk: Linus Torvalds on git (russian) part 6Episode 2.6 - Knights Landing, the Next Manycore ArchitectureLecture - strip-mining for vectorizationTrip from Sunnyvale to San FranciscoEpisode 5.10 - Do you have enough parallelism in your codedihedrals of GB88Episode 4.4 - Thread Parallelism and OpenMPEpisode 3.3 - Native MPI ApplicationsTech Talk: Linus Torvalds on git (russian) part 3GB88 protein (PDB: 2JWU) 2 beta - alpha - 2 betaEpisode 5.19 - Closing wordsEpisode 3.7 - Asynchronous OffloadAtlanta aquariumThe Appalachian TrailEpisode 2.2 - Details of Intel MIC Architecture