Content deleted Content added
→Open hardware implementations: added the MIAOW block diagram Tags: Mobile edit Mobile web edit Advanced mobile edit |
→Open hardware SIMT processors: add ref to Hotchips for MIAOWGPU Tags: Mobile edit Mobile web edit Advanced mobile edit |
||
Line 53:
NVIDIA GPUs have a concept of the thread group called as "warp" composed of 32 hardware threads executed in lock-step. The equivalent in AMD GPUs is [[Graphics Core Next#Compute units|"wavefront"]], although it is composed of 64 hardware threads. In OpenCL, it is called as "sub-group" for the abstract term of warp and wavefront. CUDA also has the warp shuffle instructions which make parallel data exchange in the thread group faster,<ref>[https://developer.nvidia.com/blog/faster-parallel-reductions-kepler/ Faster Parallel Reductions on Kepler | NVIDIA Technical Blog]</ref> and OpenCL allows a similar feature support by an extension cl_khr_subgroups.<ref>[https://registry.khronos.org/OpenCL/sdk/3.0/docs/man/html/cl_khr_subgroups.html cl_khr_subgroups(3) Manual Page]</ref>
== Open hardware
=== MIAOW GPU ===
[[File:MIAOW_GPU_diagram.png|thumb|MIAOW GPU and associated Computation Unit block diagram.<ref>https://github.com/VerticalResearchGroup/miaow/wiki/Architecture</ref> ]]▼
▲[[File:MIAOW_GPU_diagram.png|thumb|MIAOW GPU and associated Computation Unit block diagram.<ref>https://github.com/VerticalResearchGroup/miaow/wiki/Architecture</ref> ]]
The MIAOW Project by the Vertical Research Group is an implementation of AMDGPU "Southern Islands".<ref> https://research.cs.wisc.edu/vertical/wiki/index.php/Main/Projects#miaow </ref> An overview of the internal architecture and design goals was presented at Hotchips.<ref>https://pages.cs.wisc.edu/~vinay/pubs/MIAOW-HotChips.pdf</ref>
== See also ==
| |||