Method for compiling a parallel thread execution program for general execution
作者： Vinod Grover , Andrew Kerr , Sean Lee
DOI:
关键词:
摘要: A technique is disclosed for executing a compiled parallel application on general purpose processor. The comprises thread execution code, which includes single-instruction multiple-data (SIMD) constructs, as well references to intrinsic functions conventionally available in graphics processing unit. code transformed into an intermediate representation, vector instruction constructs. SIMD constructs are mapped instructions within the representation. Intrinsic corresponding emulated runtime implementations. advantageously enables applications unit be executed central configured support instructions.
lens.org 参考文章(46)
Christopher G. Demetriou, Matthew N. Papakipos, Systems and methods for caching compute kernels for an application running on a parallel-processing computer system ,(2007)
Wei-Tek Tsai, Viet N. Ngo, Outer loop vectorization ,(1994)
Partha Pal Tirumalai, Spiros Kalogeropulos, Pipelined loop parallelization with pre-computations ,(2011)
Junhong Sun, Lin Chen, Chihong Zhang, Guofang Jiao, Shader compile system and method ,(2007)
Nalini Vasudevan, Christopher J. Hughes, Jayashankar Bharadwaj, Sara S. Baghsorkhi, Milind B. Girkar, Mark J. Charney, Victor W. Lee, Daehyun Kim, Albert Hartono, Robert Valentine, Loop vectorization methods and apparatus ,(2012)
Jeffry E. Gonion, Systems and methods for compiler-based vectorization of non-leaf code ,(2010)
Jeffrey Hammes, Lisa Krause, Jon Steidel, Daniel Poznanovic, System and method for partitioning control-dataflow graph representations ,(2003)
Lakshminarayanan Renganarayana, Uday Kumar Bondhugula, Yuan Zhao, John Kevin P. O'Brien, Alexandre E. Eichenberger, Reducing parallelism of computer source code ,(2010)