| commit | 21a85fef159f9942f636a43b14c64b481c2a05b2 | [log] [tgz] |
|---|---|---|
| author | Ruy Contributors <ruy-eng@google.com> | Wed Feb 15 11:37:59 2023 -0800 |
| committer | Copybara-Service <copybara-worker@google.com> | Wed Feb 15 11:38:30 2023 -0800 |
| tree | 302071265578f0f10baff83a8edb817a4a842a7e | |
| parent | 3168a5c8f4c447fd8cea94078121ee2e2cd87df0 [diff] |
Most callers directly depend on :cpuinfo, whose symbols are provided by :cpuinfo_impl, where linkstatic is set. When building tests (default to dynamic linking) with toolchains that set -fvisibility=hidden towards a static link to reduce code size, this would hide symbols and fail. This CL merges :cpuinfo, putting the symbols in the same build unit and solving this issue. -std=gnu99 is removed as it's not required to build this package. This allows the headers to build standalone, so textual_headers no longer need to be specified. :cpuinfo_mock now directly depends on :cpuinfo, so non-mock srcs are no longer needed and are removed from the target. A typo in a macro is fixed alongside. PiperOrigin-RevId: 509891384
This is not an officially supported Google product.
ruy is a matrix multiplication library. Its focus is to cover the matrix multiplication needs of neural network inference engines. Its initial user has been TensorFlow Lite, where it is used by default on the ARM CPU architecture.
ruy supports both floating-point and 8bit-integer-quantized matrices.
ruy is designed to achieve high performance not just on very large sizes, as is the focus of many established libraries, but on whatever are the actual sizes and shapes of matrices most critical in current TensorFlow Lite applications. This often means quite small sizes, e.g. 100x100 or even 50x50, and all sorts of rectangular shapes. It's not as fast as completely specialized code for each shape, but it aims to offer a good compromise of speed across all shapes and a small binary size.
Some documentation will eventually be available in the doc/ directory, see doc/README.md.