Abstract:  Today’s multi-core  era  places significant  demands on  an optimizing compiler,  which must parallelize  programs, exploit memory hierarchy, and  leverage the ever-increasing  SIMD (Single Instruction Multiple Data) capabilities of modern processors. Such a compiler must be able to  apply complex sequences of loop  transformations to target the  available  hardware  resources.    The  polyhedral  model  is  an algebraic  representation of  programs  that allows  to construct  and search  for complex  sequences of  optimizations.  This  model  is now mature and reaches production compilers  such as IBM XL, GCC (GRAPHITE framework) or  more recently LLVM  (Polly framework). In this  talk, I will introduce the polyhedral model,  from theory to practice, as well as every component of a polyhedral compilation framework, from initial program  analysis to  final code  generation. I  will  discuss various optimization   techniques   in   this   framework   (model-based   and iterative-based) and present its main restrictions and open problems.

This talk is intended for  a general computer science audience with an interest  on  compilation  and/or  program  optimization;  very  basic knowledge of linear algebra (matrix-vector multiply) being a plus.

>>more

 
buy viagra online | buy viagra pills | Viagra Product Information | free porn | cheap viagra | order discount viagra | free viagra sample | 100 mg viagra | Viagra For Sale | Purchase viagra