Search: 
GSRC Student Profile:

Hee-Seok Kim

n/a
(For more contact info, please log in.)

University of Illinois at Urbana‑Champaign
Advisor: Wen‑mei Hwu

GSRC theme:  architectures
Expected graduation:  May, 2014

Research Overview:  Thread Coarsening as an Optimization Technique for GPUs

ince fine-grained programming model is preferred for GPUs in order to launch thousands kernels in parallel, often times, a thread for GPUs underutilizes dedicated resources such as registers or instruction level parallelism. Thread coarsening merges workloads of two or more threads into one thread so that the hardware can run more efficiently. Also, this achieves elimination of redundant memory operations and computations as common values can be shared after coarsening.