Presentation
Using Integrated Processor Graphics to Accelerate Concurrent Data and Index Structures
Author
Event Type
ACM Student Research Competition
Poster
TP
EX
TimeTuesday, November 13th8:30am - 5pm
LocationC2/3/4 Ballroom
DescriptionWith the advent of computing systems with on-die integrated processor graphics (iGPU), new programming challenges have emerged from these heterogeneous systems. We proposed different data and index structure algorithms that can benefit from the Intel's iGPU architecture and the C for Media (CM) programming model. We aim that certain data structures can run on the iGPU more efficiently than the CPU cores, achieving important performance gains and energy savings. To the best of our knowledge, this is the first attempt to use iGPU for running workloads on concurrent data and index structures. Experimental results show speedups of up to 4x on concurrent data structures and 11x on index structures when comparing with state-of-the-art CPU implementations. Energy savings of up to 300% are also obtained when running these algorithms on iGPU.
Archive
