文件列表:

CMakeLists.txt (2535, 2023-12-21)
LICENSE (35149, 2023-12-21)
include/ (0, 2023-12-21)
include/CMakeLists.txt (51, 2023-12-21)
include/Saxpy/ (0, 2023-12-21)
include/Saxpy/CMakeLists.txt (215, 2023-12-21)
include/Saxpy/saxpy.h (720, 2023-12-21)
include/Utilities/ (0, 2023-12-21)
include/Utilities/CMakeLists.txt (290, 2023-12-21)
include/Utilities/debug.h (2448, 2023-12-21)
include/Utilities/device.h (280, 2023-12-21)
include/Utilities/streamtimer.h (644, 2023-12-21)
src/ (0, 2023-12-21)
src/CMakeLists.txt (84, 2023-12-21)
src/ExclusiveScan/ (0, 2023-12-21)
src/ExclusiveScan/CMakeLists.txt (225, 2023-12-21)
src/ExclusiveScan/exclusivescan.cu (8228, 2023-12-21)
src/Saxpy/ (0, 2023-12-21)
src/Saxpy/CMakeLists.txt (363, 2023-12-21)
src/Saxpy/saxpy.cu (2786, 2023-12-21)
src/Saxpy/saxpy.test.cpp (7132, 2023-12-21)
src/Saxpy/settings.h (433, 2023-12-21)
src/Utilities/ (0, 2023-12-21)
src/Utilities/CMakeLists.txt (316, 2023-12-21)
src/Utilities/debug.cpp (2397, 2023-12-21)
src/Utilities/device.cpp (869, 2023-12-21)
src/Utilities/streamtimer.cpp (1160, 2023-12-21)

# CUDA Projects A place for my personal CUDA tinkering to reside. The layout of the repository adheres to [The Pitchfork Layout](https://api.csswg.org/bikeshed/?force=1&url=https://raw.githubusercontent.com/vector-of-bool/pitchfork/develop/data/spec.bs), specifically using [seperate header placement](https://api.csswg.org/bikeshed/?force=1&url=https://raw.githubusercontent.com/vector-of-bool/pitchfork/develop/data/spec.bs#src.header-placement) and [merged test placement](https://api.csswg.org/bikeshed/?force=1&url=https://raw.githubusercontent.com/vector-of-bool/pitchfork/develop/data/spec.bs#src.tests). ### Dependencies ### * [CMake](https://cmake.org/download/) * [CUDA Toolkit](https://developer.nvidia.com/cuda-downloads) ### Perform Out-Of-Source Build ### From the repository root: ``` mkdir build && cd build && cmake .. ``` For example, to build *Tests* for *Release* on Windows: ``` msbuild.exe Tests.vcxproj /property:Platform=x64 /property:Configuration=Release ``` To then run the *Tests* executable: ``` Release\Tests.exe ``` --- ## Exclusive Scan ## *Work In Progress*. Currently toying around with my exclusive scan implementation using the CUDA *cooperative groups* API. Given a binary associative operator $\oplus$ with the identity element $\mathbb{i}$ and a contiguous array of *n* elements $\mathbf{A}_n = [a_1, a_2, \ldots, a_n]$: $$\textrm{Exclusive Scan}(\mathbf{A}_n) = [\mathbb{i}, a_1, a_1 \oplus a_2, a_1 \oplus a_2 \oplus a_3, \ldots, a_1 \oplus a_2 \ldots \oplus a_n-1]$$ --- ## Saxpy ## The canonical *single-precision ax + y kernel*: $$\mathbf{\overline{z}} = \alpha\mathbf{\overline{x}} + \mathbf{\overline{y}} \quad \textrm{where} \quad \mathbf{\overline{x}}, \mathbf{\overline{y}}, \mathbf{\overline{z}} \in \mathbb{R}^{n}, \textrm{ } \alpha \in \mathbb{R}$$ This is used as a prototype kernel for code and file organization. --- ## Utilities ## CUDA and host utility functions that are useful when working with the CUDA programming model. ---

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：