The Next 700 Accelerated Layers
Citations Over TimeTop 10% of 2019 papers
Abstract
Deep learning frameworks automate the deployment, distribution, synchronization, memory allocation, and hardware acceleration of models represented as graphs of computational operators. These operators wrap high-performance libraries such as cuDNN or NNPACK. When the computation does not match any predefined library call, custom operators must be implemented, often at high engineering cost and performance penalty, limiting the pace of innovation. To address this productivity gap, we propose and evaluate: (1) a domain-specific language with a tensor notation close to the mathematics of deep learning; (2) a Just-In-Time optimizing compiler based on the polyhedral framework; (3) carefully coordinated linear optimization and evolutionary algorithms to synthesize high-performance CUDA kernels; (4) the transparent integration of our flow into PyTorch and Caffe2, providing the fully automatic synthesis of high-performance GPU kernels from simple tensor algebra. The performance is comparable to, and often exceeds the performance of, highly tuned libraries.
Related Papers
- → Evaluation of CUDA for X-Ray Imaging System(2012)4 cited
- → On the Cost of a General GPU Framework: The Strange Case of CUDA 4.0 vs. CUDA 5.0(2012)3 cited
- High speed molecular dynamics simulation approach based on CUDA(2013)
- REALIZATION OF ISLAND MODEL GENETIC ALGORITHM ON NVIDIA CUDA ARCHITECTURES(2016)
- → Poster: On the Cost of a General GPU Framework: The Strange Case of CUDA 4.0 vs. CUDA 5.0(2012)