eInfoChips demos 25x migration speedup on Tesla

Cuda migration specialist claims ‘parallel programming is an art.'

eInfoChips (EIC) claims a 25x performance improvement for its Nvidia Tesla-based seismic imaging software. The solution was unveiled at Nvidia’s 2015 GPU Technology Conference in San Jose, California. EIC implemented a Kirchoff pre-stack depth migration algorithm on high-performance Nvidia Tesla K40 GPU accelerators. The 25x speedup was benchmarked over an Intel Core i7-2600 @3.40GHz system with 8GB memory. The port involved adding some 190 lines of Cuda to the 980 lines of C code in the kernel.

EIC claims expertise in parallelizing code for GPU deployment in automotive, medical, industrial automation and other verticals. CMO Parag Mehta explained, ‘Parallel programming is an art, mastered over multiple code migration projects. Our team of parallel programming experts has codified the leading practices with a mature checklist to achieve the maximum performance on GPU accelerators.’ More from EIC.

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.