Google GitGoogle Git
Sign in
android/platform/external/ruy/a8907c5
commit
a8907c5d49ed3cdba6b45e17ac2c0e8f4c2122b8
[log]
author
Ruy Contributors <ruy-eng@google.com>
Mon May 06 11:58:51 2019 -0400
committer
Benoit Jacob <benoitjacob@google.com>
Tue Mar 10 16:30:10 2020 -0400
tree
ecd79dd8375c251bd1bc4ef5630b94d97cd71dce
parent
cb9587f01b6d778e8c5e59b56bfea2035e3da743 [diff]
PR #28392: [aarch64] make TensorFlow build for aarch64 linux natively with bazel

Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/28392

Tested on Coral Dev Board running Mendel Linux chef release
```
bazel --host_jvm_args=-Xms128m --host_jvm_args=-Xmx2048m \
build --config opt --local_resources 1024,1,1 \
//tensorflow/tools/pip_package/build_pip_package
```

Copybara import of the project:

  - aaf0f0ea1e027d7a4d1fd40efa6bb2a5cfd5996d make TensorFlow build for aarch64 linux by Koan-Sin Tan <koansin.tan@gmail.com>
  - 9bb784b389b1c09f5bb390638a6e701c86fb8b6c Merge aaf0f0ea1e027d7a4d1fd40efa6bb2a5cfd5996d into 95563... by freedom" Koan-Sin Tan <koansin.tan@gmail.com>

COPYBARA_INTEGRATE_REVIEW=https://github.com/tensorflow/tensorflow/pull/28392 from freedomtan:build_for_the_coral_dev_board aaf0f0ea1e027d7a4d1fd40efa6bb2a5cfd5996d
PiperOrigin-RevId: 246834423
8 files changed
tree: ecd79dd8375c251bd1bc4ef5630b94d97cd71dce
  1. allocator.cc
  2. allocator.h
  3. allocator_test.cc
  4. benchmark.cc
  5. block_map.cc
  6. block_map.h
  7. blocking_counter.cc
  8. blocking_counter.h
  9. BUILD
  10. check_macros.h
  11. common.h
  12. context.cc
  13. context.h
  14. detect_dotprod.cc
  15. detect_dotprod.h
  16. dispatch.h
  17. example.cc
  18. impl.h
  19. internal_matrix.h
  20. kernel.cc
  21. kernel.h
  22. matrix.h
  23. opt_set.h
  24. pack.cc
  25. pack.h
  26. path.h
  27. pmu.cc
  28. pmu.h
  29. README.md
  30. ruy.h
  31. ruy_test.bzl
  32. ruy_test_ext.bzl
  33. ruy_visibility.bzl
  34. size_util.h
  35. spec.h
  36. test.h
  37. test_fast.cc
  38. test_slow.cc
  39. test_special_specs.cc
  40. thread_pool.cc
  41. thread_pool.h
  42. time.h
  43. trace.cc
  44. trace.h
  45. tune.cc
  46. tune.h
  47. tune_test.cc
  48. tune_tool.cc
README.md

ruy is not BLAS

ruy is a matrix multiplication library. Its focus is to cover the matrix multiplication needs of TensorFlow Lite.

ruy supports both floating-point (like Eigen) and quantized (like gemmlowp).

Status

ruy is very new, immature code. It has quite good test coverage, but the code is in flux, lacks comments, needs more cleanup, and there are no design docs at the moment.

We hope to improve on all that and integrate ruy into TensorFlow Lite, at first as a non-default path for ARM A64 only, over the next few weeks [April 2019].

Efficiency

ruy is designed to achieve maximal 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.

ruy is currently only optimized for ARM A64; other architectures have only slow reference code at the moment.

ruy is currently optimized only for the following combination of storage orders: LHS = row-major, RHS = column-major, destination = column-major. All other combinations of storage orders fall back to slow reference code at the moment.