Polynomial Interpolation and Extrapolation Marija Stanojevic SI 20100011

  • Slides: 14
Download presentation
Polynomial Interpolation and Extrapolation Marija Stanojevic SI 2010/0011

Polynomial Interpolation and Extrapolation Marija Stanojevic SI 2010/0011

Computing with dataflow engines (DFEs) Dataflow Technology Computing with control flow processors 2

Computing with dataflow engines (DFEs) Dataflow Technology Computing with control flow processors 2

Maxeler Technology MAXELER DATAFLOW COMPUTING One result per clock cycle Engine is written in

Maxeler Technology MAXELER DATAFLOW COMPUTING One result per clock cycle Engine is written in Java and Host is written in C. Minimal frequency f achieves maximal performance, thus for a given power budget, we get Maximum Performance Computing (MPC)! 3 3 /16

Neville's algorithm Given a set of n+1 data points (xi, yi) where no two

Neville's algorithm Given a set of n+1 data points (xi, yi) where no two xi are the same, the interpolating polynomial is the polynomial p of degree at most n with the property: p(xi) = yi for all i = 0, …, n This polynomial exists and it is unique. Neville's algorithm evaluates this polynomial at some point and gives value y and error estimate dy It is easy to work with algorithm recursively as shown in figure in the right bottom corner 4

Neville's algorithm Algorithm is based on formula: It can be easily become recursive algorithm

Neville's algorithm Algorithm is based on formula: It can be easily become recursive algorithm by next transformations: 5

Maxeler code - Java 6

Maxeler code - Java 6

Maxeler code - Java 7

Maxeler code - Java 7

Maxeler code - C 8

Maxeler code - C 8

Successful building and running 9

Successful building and running 9

Results X and Y coordinates of points were in range from 0 to 10

Results X and Y coordinates of points were in range from 0 to 10 Algorithm complexity is N 2 Interpolation is done for: – 4 points with time of execution on Maxeler: 0. 01 s – 100 points with time of execution on Maxeler: 0. 14 s – 1000 points with time of execution on Maxeler: 4. 68 s – 2000 points with time of execution on Maxeler: 31. 09 s 10

Results 11

Results 11

Conclusion This project implements Neville's interpolation algorithm on Maxeler Whole project code can be

Conclusion This project implements Neville's interpolation algorithm on Maxeler Whole project code can be found on: https: //github. com/mstanojevic 118/Maxeler. Project There are two versions of C code for Maxeler; File with Nice. Output in name gives nicer, but less extensive output Maxeler is practical for big amount of data, because it's fast and energy saving 12

References Milutinovic V. , Rakocevic G. , Stojanovic S. , Sustran Z. , Mencer

References Milutinovic V. , Rakocevic G. , Stojanovic S. , Sustran Z. , Mencer O. , Pell O. , Flynn M. , Korolija N. , Data. Flow Exa. Scale Super. Computing: Revisiting the Paradigm and the Algorithms, https: //www. bsc. es/sites/default/files/public/ mare_nostrum/hpc-events/hpcac 2012 -12_maxeler. pdf Selected Max. Compiler Examples, Sustran Z. , Stojanovic S. , home. etf. rs/~vm/tutorial/Berlin/11. %20 Maxeler-examples. pptx Press W. , Teukolsky S. , Vetterling W. , Flannery B. , Numerical Receipes in C, Cambridge University Press, 2002 Neville's algorithm, https: //en. wikipedia. org/wiki/Neville%27 s_algorithm 13

Thank you for your attention! 14

Thank you for your attention! 14