The SC26 Test of Time Award will recognize a 2009 paper by Nathan Bell and Michael Garland on sparse matrix-vector multiplication, according to HPCwire. The paper was presented at SC09, a time when general-purpose GPU computing was still young and CUDA had only been introduced a few years earlier.
The award underscores how foundational the work became for GPU sparse computing. At the time, researchers were still learning how to effectively use GPUs for general-purpose tasks, and the paper helped establish methods that have remained relevant well over a decade later.
Because the source is a single announcement, there are no differing accounts to compare. The report focuses on the award itself and the historical context of the original work, rather than on technical details of the method or its subsequent applications.