4x Code Performance with SIMD
Автор: Keifer
Загружено: 2025-05-24
Просмотров: 60498
Dives into the significant performance gains of using SIMD instructions via auto-vectorization with a use case inspired by "Bunnymark" benchmarks. Data layout and compiler flags have a major impact on software performance, so it's crucial to understand their impact.
raylib bunnymark example: https://www.raylib.com/examples/textu...
Compiler Explorer Links for video examples:
Array of Structures: https://godbolt.org/z/WcqjnqoMq
Large Structure: https://godbolt.org/z/bzxd5r8M5
Structure of Arrays: https://godbolt.org/z/as1MeTzvM
Hybrid (Unaligned): https://godbolt.org/z/zhW5eYGWj
Hybrid (Aligned): https://godbolt.org/z/36K6a3ehE
Larger exploratory project: https://github.com/KeithJH/kinematics...
Music: Untitled by @keiferjh ( • BGM from "4x Code Performance with SIMD" )
Chapter Timestamps:
00:00 - Bunnymark
00:59 - Auto-vectorization
02:20 - Array of Structures (AoS)
04:04 - Compiler Flags (-O2)
05:11 - Compiler Flags (-O3)
06:11 - Compiler Flags (-march)
08:19 - Large Structure
09:11 - Structure of Arrays (SoA)
09:41 - Hybrid (AoSoA)
10:46 - Alignment
11:23 - Summary
Доступные форматы для скачивания:
Скачать видео mp4
-
Информация по загрузке:
