Johannes Dieterich

Optional offloading to AMD GPUs

2

This initial work via opt-in configure option enables offloading of some sgemm, dgemm, cgemm, zgemm operations to AMD GPUs via AMD's rocBLAS. It hence requires a working ROCm software stack...

Details: * Follow the CPU logic to support CONJUGATE_NO_TRANSPOSE where an intermediate buffer is allocated, the matrix is copied into it, and the imaginary part of the complex number is...

hcc in ROCm 1.8.2RC3 uses huge amount of memory to compile file

7

For full reproduction: Checkout [pytorch](https://github.com/ROCmSoftwarePlatform/pytorch), run `python3 tools/amd_build/build_pytorch_amd.py` and build pytorch. Compile only file that is problematic (on ubuntu 18.04 and w/ my own paths): ``` /opt/rocm/hcc/bin/clang-7.0 -cc1 -D__KALMAR_HC__=1 -D__HCC_HC__=1...

I am one of the lucky few that actually got the tablet back in the day. With the youyouta project, there seem to be more people getting tablets. It'd be...

[RFC] HIP backend for AMD GPU support

3

We would like to contribute a HIP backend to Faiss to support AMD GPUs. We have a working prototype that passes all unit tests on Navi hardware (6800XT, 7900XTX). The...

Update required cmake version to 3.24.

1

The CMakeLists.txt in faiss/gpu uses the $

ROCm support

24

* add hipify at configure time * ROCm specific code paths behind USE_ROCM guards * support for wavefront 32 (Navi) and 64 (MI) * use builtins to match inline PTX...