C++17 parallel algorithms and HIPSTDPAR – AMD lab notes
This post discusses how to leverage C++17 parallel algorithms on AMD GPUs with HIPSTDPAR
This post discusses how to leverage C++17 parallel algorithms on AMD GPUs with HIPSTDPAR
Sparse matrix vector multiplication (SpMV) is a core computational kernel of nearly every implicit sparse linear algebra solver. This is the first post in the series covering SpMV.
In the fourth and final part of Finite Difference Laplacian blog series we cover scaling studies and cache size limitations
This presentation is a practical implementation of a solution aimed at making the most of every sample by caching the estimated radiance into a cache hierarchy used for both sampling and filtering.
GPUOpen Matrix Compendium: This page shows a selection of matrices in the coordinate system expected by DirectX® 9.
The GPUOpen Matrix Compendium covers how matrices are used in 3D graphics and implementations in host code and shading languages. It’s a growing guide, so keep checking back!
GPUOpen Matrix Compendium: This page shows a selection of matrices in a pre-multiplication, right-handed coordinate system.
GPUOpen Matrix Compendium: This page shows a selection of matrices in the coordinate system expected by OpenGL®.
GPUOpen Matrix Compendium: This page shows a selection of matrices in a post-multiplication, left-handed coordinate system.
This presentation introduces a novel algorithm for PC and console developers to efficiently generate sparse distance fields in real-time.
In this post we introduce two common optimizations that can be applied to the kernel to reduce data movement and bring us closer to the new peak: loop tiling to explicitly reduce memory loads and re-order the memory access pattern to improve caching.
In this blog series, we share the lessons learned from tuning a wide range of scientific applications, libraries, and frameworks for AMD GPUs.