commit | 8155854bfecacedaf1879eba2a0f1a2223b42a75 | [log] [tgz] |
---|---|---|
author | Frank Barchard <fbarchard@google.com> | Tue Feb 11 16:35:26 2020 -0800 |
committer | XNNPACK Team <xnnpack-github-robot@google.com> | Tue Feb 11 16:35:56 2020 -0800 |
tree | 40c0aa29c07052e3bd04575631c1bdaeb2017e72 | |
parent | 79ade18026f844af50574866c4f6aaf24a0bd974 [diff] |
Direct branch to source remainder handler for GEMM/IGEMM. All ld64, ld128 and A53 microkernels move source remainder to a branch forward (label 5) and branch directly to the handler if there is less than main loop channels. For 2 channels if there is a remainder, it is 1 channel so additional check is not required. For 4 channels if there is a remainder, and it is not 2 channels, it is 1 channel so additional check is not required. Standardize on label 4 for clamp code, 5 for remainder. Should be a small performance improvement for 1 channel, by branching directly to the remainder code, and case where there is no remainder the branch is not taken. PiperOrigin-RevId: 294549181
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 MediaPipe, TensorFlow Lite, and TensorFlow.js.
XNNPACK implements the following neural network operators:
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.
The table below presents single-threaded performance of XNNPACK library on three generations of MobileNet models and three generations of Pixel phones.
Model | Pixel, ms | Pixel 2, ms | Pixel 3a, ms |
---|---|---|---|
MobileNet v1 1.0X | 81 | 89 | 88 |
MobileNet v2 1.0X | 48 | 55 | 54 |
MobileNet v3 Large | 40 | 44 | 44 |
MobileNet v3 Small | 12 | 14 | 14 |
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.
Model | Pixel, ms | Pixel 2, ms | Pixel 3a, ms |
---|---|---|---|
MobileNet v1 1.0X | 45 | 27 | 46 |
MobileNet v2 1.0X | 28 | 18 | 28 |
MobileNet v3 Large | 23 | 16 | 24 |
MobileNet v3 Small | 7 | 6 | 8 |
Benchmarked on January 9, 2020 with end2end_bench --benchmark_min_time=5
on an Android/ARM64 build (bazel build -c opt --config android_arm64 :end2end_bench
) and neural network models with randomized weights and inputs.
The table below presents multi-threaded performance of XNNPACK library on three generations of MobileNet models and three generations of Raspberry Pi boards.
Model | RPi 2 (BCM2836), ms | RPi 3+ (BCM2837B0), ms | RPi 4 (BCM2711), ms |
---|---|---|---|
MobileNet v1 1.0X | 380 | 115 | 76 |
MobileNet v2 1.0X | 217 | 80 | 45 |
MobileNet v3 Large | 180 | 67 | 41 |
MobileNet v3 Small | 57 | 23 | 15 |
Benchmarked on January 9, 2020 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.
XNNPACK is a based on QNNPACK library. Unlike QNNPACK, XNNPACK focuses entirely on floating-point operators, and its API is no longer compatible with QNNPACK.