Package: wnpp Severity: wishlist Owner: Cordell Bloor <c...@slerp.xyz> X-Debbugs-Cc: debian-devel@lists.debian.org, c...@slerp.xyz, debian...@lists.debian.org
* Package name : rocblas Version : 5.3.0 Upstream Author : Advanced Micro Devices, Inc. * URL : https://github.com/ROCmSoftwarePlatform/rocBLAS * License : Expat (MIT/X) Programming Lang: C, C++, HIP Description : ROCm library for basic linear algebra rocBLAS is a library for performing basic linear algebra routines on AMD GPUs. It provides implementations of BLAS operations such as dot product (DOT), general matrix multiplication (GEMM), triangular matrix multiplication (TRMM), and many others. It is built on the AMD ROCm platform and optimized for discrete AMD GPUs. The rocBLAS library is implemented in the HIP programming language and is supplemented by tuned assembly kernels. rocBLAS provides a C99 interface and supports operations on matrices of half-, single- or double-precision floating point numbers. The library provides variants for its routines that efficiently operate on batches of matrices and that support a variety of data layouts. rocBLAS is a key package required to leverage AMD GPU hardware for scientific computing and AI. Many applications in that domain benefit significantly from fast matrix multiplication. The rocBLAS library is a dependency of PyTorch, Tensorflow, Tasmanian, pika and others. It's also required by other components of the ROCm platform such as hipBLAS, rocSOLVER, hipSOLVER, rocALUTION, MIOpen, and MIGraphX.
