vkCompViz: Universal C++ Library for GPU-Based Experiments

vkCompViz: Universal C++ Library for GPU-Based Experiments

Článek recenzovaný mimo WoS a Scopus

GPUs are gaining popularity due to their massive computational parallelism and usage in interactive graphics or machine learning applications. Prototyping GPU-executed experiments is often time-consuming due to the complexity of GPU-related APIs. Multiplatform and multi-vendor support is also not guaranteed with all existing APIs. The vkCompViz C++ library offers a simple way to write a GPGPU (general-purpose computing on graphics processing units) program. Only the paths to files with GPU code (shader/kernel), paths to input images, or a buffer of arbitrary input data need to be provided by the host application. The library is capable of running a sequence of compute shaders, processing the input data, and storing the result or presenting it in a window. Memory usage and computational time can also be automatically measured.

GPUs are gaining popularity due to their massive computational parallelism and usage in interactive graphics or machine learning applications. Prototyping GPU-executed experiments is often time-consuming due to the complexity of GPU-related APIs. Multiplatform and multi-vendor support is also not guaranteed with all existing APIs. The vkCompViz C++ library offers a simple way to write a GPGPU (general-purpose computing on graphics processing units) program. Only the paths to files with GPU code (shader/kernel), paths to input images, or a buffer of arbitrary input data need to be provided by the host application. The library is capable of running a sequence of compute shaders, processing the input data, and storing the result or presenting it in a window. Memory usage and computational time can also be automatically measured.

GPU, Vulkan, GPGPU

GPU, Vulkan, GPGPU

CHLUBNA, T.

17.01.2026

Journal of open source software

11

117

Spojené státy americké

5

https://joss.theoj.org/papers/10.21105/joss.08871