文件列表:

.clang-format (356, 2023-10-28)
BENCHMARKS.md (41639, 2023-10-28)
BUILDING.md (1730, 2023-10-28)
CMakeLists.txt (3177, 2023-10-28)
CMakePresets.json (3594, 2023-10-28)
CODE_OF_CONDUCT.md (129, 2023-10-28)
CONTRIBUTING.md (485, 2023-10-28)
HACKING.md (4379, 2023-10-28)
LICENSE.md (34904, 2023-10-28)
cmake/ (0, 2023-10-28)
cmake/coverage.cmake (837, 2023-10-28)
cmake/cpm.cmake (30878, 2023-10-28)
cmake/dev-mode.cmake (686, 2023-10-28)
cmake/docs-ci.cmake (3351, 2023-10-28)
cmake/docs.cmake (1292, 2023-10-28)
cmake/folders.cmake (837, 2023-10-28)
cmake/install-rules.cmake (140, 2023-10-28)
cmake/lint-targets.cmake (842, 2023-10-28)
cmake/lint.cmake (1305, 2023-10-28)
cmake/open-cpp-coverage.cmake.example (1170, 2023-10-28)
cmake/prelude.cmake (315, 2023-10-28)
cmake/project-is-top-level.cmake (154, 2023-10-28)
cmake/spell-targets.cmake (581, 2023-10-28)
cmake/spell.cmake (657, 2023-10-28)
cmake/variables.cmake (1059, 2023-10-28)
cmake/windows-set-path.cmake (842, 2023-10-28)
intermediate-benchmarks.md (33276, 2023-10-28)
run.bat (1362, 2023-10-28)
run.sh (780, 2023-10-28)
src/ (0, 2023-10-28)
src/main.cpp (68183, 2023-10-28)
src/nanobench.h (127891, 2023-10-28)
src/shims/ (0, 2023-10-28)
src/shims/linux/ (0, 2023-10-28)
src/shims/linux/sal.h (224696, 2023-10-28)
... ...

# Vector Math Benchmark This project benchmarks several open source vector math libraries against one another to establish a baseline for performance. Currently, it tests [GLM](https://github.com/g-truc/glm), [DirectXMath](https://github.com/microsoft/DirectXMath), [SimpleMath from DirectXTK](https://github.com/microsoft/DirectXTK/blob/main/Inc/SimpleMath.h), [this fork of Sony's Vectormath](https://github.com/glampert/vectormath), [move::vectormath](https://github.com/move-engine/move-vectormath), and [Realtime Math](https://github.com/nfrechette/rtm). It tests the performance of all libraries under SSE4.2, AVX, and AVX2. Note that this repository was created with `cmake-init`, and much of what is here is boilerplate related to it. The only really important things in the repository are `src/main.cpp`, which includes all of the benchmarking code, and [BENCHMARKS.md](BENCHMARKS.md), which contains the latest benchmarking results. # Results TL;DR: Based on the current benchmarks (reflected in intermediate-benchmarks.md), it seems that the generally highest performing configuration across both AMD and Intel is Realtime Math, though it lacks many features compared to the other libraries. `move::vectormath` seems to have generally similar performance to RTM (which makes sense given that it's built on top of RTM), though similarly suffers from a lack of features at the moment (though it does have more features than RTM, as well as including more game-centric extensions for RTM based on DXM/GLM, depending on the operation in question). DXM with SSE4.2 seems to be the fastest non-RTM library, though it trades blows with Vectormath for matrix operations and GLM for vector operations. See the [BENCHMARKS](BENCHMARKS.md) document for more details. # Building and installing See the [BUILDING](BUILDING.md) document. # Contributing See the [CONTRIBUTING](CONTRIBUTING.md) document. # Licensing This project is licensed under GPLv3 to encourage users to contribute their additions back to the main project.

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：