• Home
  • History
  • Annotate
Name Date Size #Lines LOC

..--

cmake/23-Nov-2023-625475

doc/23-Nov-2023-536452

example/23-Nov-2023-417346

ruy/23-Nov-2023-39,52728,085

third_party/23-Nov-2023-393342

.gitignoreD23-Nov-2023297 3124

.gitmodulesD23-Nov-2023192 76

Android.bpD23-Nov-20232.6 KiB9891

BUILDD23-Nov-202399 85

CMakeLists.txtD23-Nov-20233.7 KiB9179

CONTRIBUTING.mdD23-Nov-20231.1 KiB2920

LICENSED23-Nov-202311.1 KiB203169

README.mdD23-Nov-20231 KiB2617

WORKSPACED23-Nov-20231.5 KiB4538

README.md

1# The ruy matrix multiplication library
2
3This is not an officially supported Google product.
4
5ruy is a matrix multiplication library. Its focus is to cover the matrix
6multiplication needs of neural network inference engines. Its initial user has
7been TensorFlow Lite, where it is used by default on the ARM CPU architecture.
8
9ruy supports both floating-point and 8bit-integer-quantized matrices.
10
11## Efficiency
12
13ruy is designed to achieve high performance not just on very large sizes, as
14is the focus of many established libraries, but on whatever are the actual sizes
15and shapes of matrices most critical in current TensorFlow Lite applications.
16This often means quite small sizes, e.g. 100x100 or even 50x50, and all sorts of
17rectangular shapes. It's not as fast as completely specialized code for each
18shape, but it aims to offer a good compromise of speed across all shapes and a
19small binary size.
20
21## Documentation
22
23Some documentation will eventually be available in the doc/ directory, see
24[doc/README.md](doc/README.md).
25
26