Google Git
Sign in
chromium / external / github.com / google / XNNPACK / 415698f51d7dd3e93f4f119edd30cbc402a45cb2
commit415698f51d7dd3e93f4f119edd30cbc402a45cb2[log] [tgz]
authorZhi An Ng <zhin@google.com>Mon Jul 25 18:13:46 2022
committerXNNPACK Team <xnnpack-github-robot@google.com>Mon Jul 25 18:22:05 2022
treec1496e3195fecc3ce8f193db2467b6446e20cae6
parent2247560904f5366d6d370bb080cfc2dbe9f57598 [diff]
Add an option to create weights cache specifying the capacity of the underlying buffer

By default, weights cache is 1MB. If the size of all weights used in a model exceeds 1MB, every operation that requires packing of weights will grow the weights cache by a small amount (size of the weights needed by the operator). This leads to unnecessary memory movements. We can increase the default size of the weights cache, but this penalizes models which don't use or have smaller weights. So, this new weights cache creation method takes a size, which is the initial capacity of the weights cache. The weights cache is guaranteed to not grow as long as the size of weights used in the weights cache is <= this initial size (though it can still grow once during soft finalization).

PiperOrigin-RevId: 463131786
5 files changed
tree: c1496e3195fecc3ce8f193db2467b6446e20cae6
  1. .github/
  2. bench/
  3. cmake/
  4. eval/
  5. include/
  6. models/
  7. scripts/
  8. src/
  9. test/
  10. third_party/
  11. tools/
  12. .bazelrc
  13. .clang-format
  14. .gitignore
  15. BUILD.bazel
  16. build_defs.bzl
  17. CMakeLists.txt
  18. CONTRIBUTING.md
  19. emscripten.bzl
  20. LICENSE
  21. preamble.js.lds
  22. README.md
  23. WORKSPACE
README.md

XNNPACK

XNNPACK is a highly optimized library of floating-point neural network inference operators for ARM, WebAssembly, and x86 platforms. XNNPACK is not intended for direct use by deep learning practitioners and researchers; instead it provides low-level performance primitives for accelerating high-level machine learning frameworks, such as TensorFlow Lite, TensorFlow.js, PyTorch, and MediaPipe.

Supported Architectures

  • ARM64 on Android, Linux, macOS, and iOS (including WatchOS and tvOS)
  • ARMv7 (with NEON) on Android
  • ARMv6 (with VFPv2) on Linux
  • x86 and x86-64 (up to AVX512) on Windows, Linux, macOS, Android, and iOS simulator
  • WebAssembly MVP
  • WebAssembly SIMD
  • RISC-V (RV32GV and RV64GC)

Operator Coverage

XNNPACK implements the following neural network operators:

  • 2D Convolution (including grouped and depthwise)
  • 2D Deconvolution (AKA Transposed Convolution)
  • 2D Average Pooling
  • 2D Max Pooling
  • 2D ArgMax Pooling (Max Pooling + indices)
  • 2D Unpooling
  • 2D Bilinear Resize
  • 2D Depth-to-Space (AKA Pixel Shuffle)
  • Add (including broadcasting, two inputs only)
  • Subtract (including broadcasting)
  • Divide (including broadcasting)
  • Maximum (including broadcasting)
  • Minimum (including broadcasting)
  • Multiply (including broadcasting)
  • Squared Difference (including broadcasting)
  • Global Average Pooling
  • Channel Shuffle
  • Fully Connected
  • Abs (absolute value)
  • Bankers' Rounding (rounding to nearest, ties to even)
  • Ceiling (rounding to integer above)
  • Clamp (includes ReLU and ReLU6)
  • Convert (includes fixed-point and half-precision quantization and dequantization)
  • Copy
  • ELU
  • Floor (rounding to integer below)
  • HardSwish
  • Leaky ReLU
  • Negate
  • Sigmoid
  • Softmax
  • Square
  • Transpose
  • Truncation (rounding to integer towards zero)
  • PReLU

All operators in XNNPACK support NHWC layout, but additionally allow custom stride along the Channel dimension. Thus, operators can consume a subset of channels in the input tensor, and produce a subset of channels in the output tensor, providing a zero-cost Channel Split and Channel Concatenation operations.

Performance

Mobile phones

The table below presents single-threaded performance of XNNPACK library on three generations of MobileNet models and three generations of Pixel phones.

ModelPixel, msPixel 2, msPixel 3a, ms
FP32 MobileNet v1 1.0X828688
FP32 MobileNet v2 1.0X495355
FP32 MobileNet v3 Large394244
FP32 MobileNet v3 Small121414

The following table presents multi-threaded (using as many threads as there are big cores) performance of XNNPACK library on three generations of MobileNet models and three generations of Pixel phones.

ModelPixel, msPixel 2, msPixel 3a, ms
FP32 MobileNet v1 1.0X432746
FP32 MobileNet v2 1.0X261828
FP32 MobileNet v3 Large221624
FP32 MobileNet v3 Small768

Benchmarked on March 27, 2020 with end2end_bench --benchmark_min_time=5 on an Android/ARM64 build with Android NDK r21 (bazel build -c opt --config android_arm64 :end2end_bench) and neural network models with randomized weights and inputs.

Raspberry Pi

The table below presents multi-threaded performance of XNNPACK library on three generations of MobileNet models and three generations of Raspberry Pi boards.

ModelRPi Zero W (BCM2835), msRPi 2 (BCM2836), msRPi 3+ (BCM2837B0), msRPi 4 (BCM2711), msRPi 4 (BCM2711, ARM64), ms
FP32 MobileNet v1 1.0X39193021147277
FP32 MobileNet v2 1.0X1987191794146
FP32 MobileNet v3 Large1658161673840
FP32 MobileNet v3 Small47450221315
INT8 MobileNet v1 1.0X2589128462924
INT8 MobileNet v2 1.0X149582302017

Benchmarked on Feb 8, 2022 with end2end-bench --benchmark_min_time=5 on a Raspbian Buster build with CMake (./scripts/build-local.sh) and neural network models with randomized weights and inputs. INT8 inference was evaluated on per-channel quantization schema.

Publications

Ecosystem

Machine Learning Frameworks

Acknowledgements

XNNPACK is a based on QNNPACK library. Over time its codebase diverged a lot, and XNNPACK API is no longer compatible with QNNPACK.