Scalar Arithmetic Multiple Data: Customizable Precision for Deep Neural Networks
| Autor: | David Gregg, Michael J. Doyle, Andrew Anderson |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
010302 applied physics
FOS: Computer and information sciences Speedup Computer Science - Performance business.industry Computer science Computer Vision and Pattern Recognition (cs.CV) Computer Science - Computer Vision and Pattern Recognition 02 engineering and technology 01 natural sciences 020202 computer hardware & architecture Performance (cs.PF) Quantization (physics) Multiple data Software 0103 physical sciences 0202 electrical engineering electronic engineering information engineering Deep neural networks x86 Computer Science - Mathematical Software Arithmetic business Field-programmable gate array Implementation Mathematical Software (cs.MS) |
| Zdroj: | 2019 IEEE 26th Symposium on Computer Arithmetic (ARITH) ARITH |
| DOI: | 10.1109/arith.2019.00018 |
| Popis: | Quantization of weights and activations in Deep Neural Networks (DNNs) is a powerful technique for network compression, and has enjoyed significant attention and success. However, much of the inference-time benefit of quantization is accessible only through the use of customized hardware accelerators or by providing an FPGA implementation of quantized arithmetic. Building on prior work, we show how to construct arbitrary bit-precise signed and unsigned integer operations using a software technique which logically \emph{embeds} a vector architecture with custom bit-width lanes in universally available fixed-width scalar arithmetic. We evaluate our approach on a high-end Intel Haswell processor, and an embedded ARM processor. Our approach yields very fast implementations of bit-precise custom DNN operations, which often match or exceed the performance of operations quantized to the sizes supported in native arithmetic. At the strongest level of quantization, our approach yields a maximum speedup of $\thicksim6\times$ on the Intel platform, and $\thicksim10\times$ on the ARM platform versus quantization to native 8-bit integers. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|