We present a new efficient OpenCL-based Accelerator for large scale
Convolutional Neural Networks called Fast Inference on FPGAs for Convolution
Neural Network (FFCNN). FFCNN is based on a deeply pipelined OpenCL kernels
architecture. As pointed out before, high-level synthesis tools such as the
OpenCL framework can easily port codes originally designed for CPUs and GPUs to
FPGAs, but it is still difficult to make OpenCL codes run efficiently on FPGAs.
This work aims to propose an efficient FPGA implementation of OpenCL
High-Performance Computing Applications. To do so, a Data reuse and task
mapping techniques are also presented to improve design efficiency. In
addition, the following motivations were taken into account when developing
FFCNN: 1) FFCNN has been designed to be easily implemented on Intel OpenCL SDK
based FPGA design flow. 2) In FFFCN, different techniques have been integrated
to improve the memory band with and throughput. A performance analysis is
conducted on two deep CNN for Large-Scale Images classification. The obtained
results, and the comparison with other works designed to accelerate the same
types of architectures, show the efficiency and the competitiveness of the
proposed accelerator design by significantly improved performance and resource
utilization.

本文介绍了一种基于 OpenCL 的卷积神经网络加速器设计，称为 FFCNN，它包括数据重用和任务映射技术，这些技术可以在大规模图像分类中提高性能和资源利用率。

FFCNN：基于 FPGA 的卷积神经网络推理快速加速

FFCNN: Fast FPGA based Acceleration for Convolution neural network inference

The use of local memory is important to improve the performance of OpenCL
programs. However, its use may not always benefit performance, depending on
various application characteristics, and there is no simple heuristic for
deciding when to use it. We develop a machine learning model to decide if the
optimization is beneficial or not. We train the model with millions of
synthetic benchmarks and show that it can predict if the optimization should be
applied for a single array, in both synthetic and real benchmarks, with high
accuracy.

本研究利用机器学习模型，在数百万个合成基准测试以及真实基准测试中，高精度地预测和决定是否应该使用本地存储器以优化单个数组的性能。