GPU

Triton is a language and compiler for writing highly efficient custom Deep-Learning primitives. The aim of Triton is to provide an open-source environment for expressing tensor math workloads that offers high flexibility, developer productivity and end to end performance.

We implemented GVProf, the first value profiler that locates value redundancy problems in applications running on GPU-based clusters. Our experiments show that GVProf incurs acceptable overhead and scales to large executions. GVProf provides useful insights to guide performance optimization. Under the guidance of GVProf, we optimized several HPC and machine learning workloads, obtaining speedups up to 1.93x.

Advancements in deep learning are often associated with increasing model sizes. Training and deploying large models require …

Presented a talk about Triton and requested feedback from Intel engineers

Presented a talk about Triton and called for contributions to improving the language

GPUs are widely used in today’s computing platforms to accelerate applications in various domains. However, scarce GPU memory resources …

Presented triton programming language and its next step

Presented triton programming language and a deep learning profiler

Presented a talk about our value profiling tool at ASPLOS'22

Finite-difference methods based on high-order stencils are commonly used for modeling of seismic wave propagation, weather forecasting, …