Skip to content

Conversation

@hominhquan
Copy link
Contributor

  • Add --enable-dma option in configure script
  • DMA-specific control-trees for GEMM and TRSM families
  • Reference DMA backend implementation based on bli_pthread and memcpy
  • Vendor DMA library to be added/declared in bli_dma_vendor_type_defs.h

Minh Quan Ho added 5 commits October 20, 2021 17:07
Details:
- Using `--enable-debug=<address|thread>` will accordingly add
`-fsanitize=<address|thread>` to cflags and
`-fsanitize=<address|thread> -static-libasan` to lflags. Useful for debug.
Details:
- Add barrier to sync all threads when reading the local mem_t struct in the stack of the chief thread
- Reference: flame@dfc1267
Details:
- Add --[enable|disable]-dma option in configure.
- New error codes, bli_info.c etc.
- Add reference implementation of DMA backend based on bli_pthread
- Add DMA-related headers in frame/base/dma/
Details:
- Add new control-trees for GEMM family, guarded by BLIS_ENABLE_DMA
- Additional fields in packm's params to handle post-packing DMA-prefetching
Details:
- New control-trees for TRSM
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant