BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:The Lift Project: Performance Portable GPU Code generation via Rew
 rite Rules - Michel Steuwer (University of Edinburgh)
DTSTART:20161209T110000Z
DTEND:20161209T123000Z
UID:TALK69538@talks.cam.ac.uk
CONTACT:Liang Wang
DESCRIPTION:Computers have become increasingly complex with the emergence 
 of heterogeneous hardware combining multicore CPUs and GPUs. These paralle
 l systems exhibit tremendous computational power at the cost of increased 
 programming effort resulting in a tension between performance and code por
 tability. Typically\, code is either tuned in a low-level imperative langu
 age using hardware-specific optimizations to achieve maximum performance o
 r is written in a high-level\, possibly functional\, language to achieve p
 ortability at the expense of performance.\n \nIn this talk\, I will presen
 t our novel approach Lift which aims for combining high-level programming\
 , code portability\, and high-performance. Starting from a high-level func
 tional program we apply semantic preserving rewrite rules to transform it 
 into a low-level functional representation\, close to the OpenCL programmi
 ng model\, from which eventually OpenCL code is generated. Our rewrite rul
 es define a space of possible implementations which we automatically explo
 re for generating hardware-specific OpenCL implementations.\n \nI will dis
 cuss the design and implementation of the Lift language and compiler showi
 ng the advantages of a functional internal compiler representation for hig
 h-performance code generation. I will show experimental results of the hig
 h-performance OpenCL code generated with Lift. Our experiments show how ha
 rdware specific implementations are automatically derived from simple func
 tional high-level algorithmic expressions offering performance on a par wi
 th highly tuned code for multicore CPUs and GPUs written by experts.\n\nBi
 o: Michel Steuwer is a Postdoctoral Research Associate in the CArD group a
 t the University of Edinburgh. He received his PhD in 2015 from the Univer
 sity of Muenster in Germany. His research interests span all areas of para
 llel programming from languages and programming models to their implementa
 tion in compilers and libraries as well as their execution at runtime. Rec
 ently\, his research has particularly focused on structured parallel progr
 amming models\, heterogeneous and GPU computing\, and novel compilation te
 chniques. He is the principle designer of the skeleton library SkelCL (git
 hub.com/skelcl) and the Lift language and compiler (github.com/lift-projec
 t).
LOCATION:FW11\, Computer Laboratory\, William Gates Building
END:VEVENT
END:VCALENDAR