VILLASframework / fpga

Archived: VILLASfpga has been integrated into the VILLASnode repo

Home Page:https://fein-aachen.org/en/projects/villas-fpga/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Add basic GPU/CUDA integration as a shared library - [merged]

stv0g opened this issue · comments

In GitLab by @daniel-k on May 15, 2018, 18:22

Merges feature/gpu -> develop

This adds basic support for integrating a GPU via CUDA into the system. Various fixes and new features throughout the whole codebase were neccessary (refer to the commits) for a nice integration. In the current state, GPU-master DMA is possible to host RAM and FPGA BRAM as well as FPGA-master DMA to GPU memory.

TODO:

  • Currently, CUDA is required to build. I will make compilation of the library optional during compile-time in the next few days
  • This is based on !8, so merge !8 first

In GitLab by @daniel-k on May 16, 2018, 10:58

added 1 commit

  • 4e4f93d - cmake: only build GPU library if CUDA is present

Compare with previous version

In GitLab by @daniel-k on May 16, 2018, 10:59

marked the checklist item Currently, CUDA is required to build. I will make compilation of the library optional during compile-time in the next few days as completed

In GitLab by @daniel-k on May 16, 2018, 10:59

marked the checklist item This is based on !8, so merge !8 first as completed

In GitLab by @daniel-k on May 16, 2018, 11:01

Now that !8 is merged and 4e4f93d implements conditional build of the GPU library, I think we can merge. I'll remove the WIP.

In GitLab by @daniel-k on May 16, 2018, 11:02

unmarked as a Work In Progress

merged

mentioned in commit 4cbba96