Projects

Projects

Engineering projects, experiments, and systems work.

Vectorizing 64-bit Integer Division

Emulating 64-bit integer division using 32-bit SIMD lanes for a measurable speedup.

active
SIMDInteger arithmeticPerformance

Building PyTorch from Source

A reproducible workflow for building PyTorch from source with CUDA on WSL2.

maintained
PyTorchCUDABuild systems

Containment Domains (CDs)

A runtime system and programming model for expressing hierarchical resilience in MPI programs.

completed
ResilienceRuntime systemsHPCMPI

Extending CDs to CUDA Programs

Containing GPU errors on the device without host-side rollback intervention.

completed
GPUResilienceCUDA

Extending CDs to Task-based Execution

Modeling resilience of Legion programs to derive optimal recovery policies for tasks.

completed
ResilienceModelingLegion

GPU Snapshot

Architectural support for low-overhead checkpoint offloading in GPU-dense systems.

completed
GPUResilienceComputer architecture

Front-end Pipeline Optimization for Scripting Languages

Better profiling and function ordering to reduce I-TLB/cache misses and branch mispredictions.

completed
PerformanceCode layoutProfiling