The next 700 accelerated layers: From mathematical expressions of network computation graphs to accelerated GPU kernels, automatically

Keyphrases

• High Performance 100%
• Mathematical Expression 100%
• GPU Kernel 100%
• Network Computation 100%
• Computational Graph 100%
• Just-in-time 33%
• Hardware Accelerator 33%
• Optimization Algorithm 33%
• Memory Allocation 33%
• Performance Penalty 33%
• Evolutionary Algorithms 33%
• CUDA Kernel 33%
• Compiler-based 33%
• Optimizing Compilers 33%
• Engineering Cost 33%
• Fully-automatic 33%
• Deep Learning 33%
• Productivity Gap 33%
• Tuned Library 33%
• Automatic Synthesis 33%
• Engineering Performance 33%
• Tensor Algebra 33%
• Domain-specific Languages 33%
• Linear Optimization 33%
• CuDNN 33%
• Deep Learning Framework 33%
• PyTorch 33%
• Call Library 33%
• Transparent Integration 33%
• Tensor Notation 33%
• Polyhedral Framework 33%
• Deployment Distributions 33%

Computer Science

• Computation Graph 100%
• Graphics Processing Unit 100%
• Deep Learning Method 100%
• hardware acceleration 50%
• Optimization Algorithm 50%
• Memory Allocation 50%
• Performance Penalty 50%
• Just-in-Time 50%
• Learning Framework 50%
• Optimizing Compiler 50%
• Linear Optimization 50%
• Automatic Synthesis 50%
• Domain-specific languages 50%
• Library Performance 50%
• Evolutionary Algorithm 50%