Abstract “Recurrent neural networks”(RNNs) are powerful artificial intelligence models that
have shown remarkable effectiveness in several tasks such as music generation, speech …

Applications extensively use data objects with a regular and fixed layout, which leads to the
recurrence of access patterns over memory regions. Spatial data prefetching techniques …

Data prefetching, ie, the act of predicting an application's future memory accesses and
fetching those that are not in the on-chip caches, is a well-known and widely used approach …

We empirically evaluate an undervolting technique, ie, underscaling the circuit supply
voltage below the nominal level, to improve the power-efficiency of Convolutional Neural …

Transfer learning in natural language processing (NLP) uses increasingly large models that
tackle challenging problems. Consequently, these applications are driving the requirements …

Optimizing deep learning models is generally performed in two steps:(i) high-level graph
optimizations such as kernel fusion and (ii) low level kernel optimizations such as those …

Graphics Processing Units (GPUs) are widely-used accelerators for data-parallel
applications. In many GPU applications, GPU memory bandwidth bottlenecks performance …

The Register File (RF) in GPUs is a critical structure that maintains the state for thousands of
threads that support the GPU processing model. The RF organization substantially affects …

Graphics Processing Units (GPUs) are widely used as the accelerator of choice for
applications with massively data-parallel tasks. However, recent studies show that GPUs …

As modern GPU workloads grow in size and complexity, there is an ever-increasing demand
for GPU computational power. Emerging workloads contain hundreds or thousands of GPU …