Using Parallel Computing Toolbox to accelerate the Video and Image Processing Speed. Develop parallel code interactively

Transcription

1 Using Parallel Computing Toolbox to accelerate the Video and Image Processing Speed Presenter: Claire Chuang TeraSoft Inc. Agenda Develop parallel code interactively parallel applications for faster processing Data parallel applications for handling large data sets Schedule your programs to run Tips on developing parallel code 1

2 Serial V.S Parallel Parallel for-loops parfor i = 1 : n % do something with i end Mix task-parallel and serial code in the same function Run loops on a pool of MATLAB resources Iterations must be order-independent M-Lint analysis helps in converting existing for-loops into to parfor-loops 2

3 The Mechanics of parfor Loops a = zeros(10, 1) parfor i = 1:10 a(i) = i; end a a(i) = i; a(i) = i; a(i) = i; a(i) = i; Pool of MATLAB s Example: Color Segmentation Red Yellow >>edit colorsegmentation.m >>edit colorsegmentationsol.m Green Magenta 3

4 Large Data Sets (Data Parallel) spmd blocks (Data Parallel) spmd % single program across workers end Mix data-parallel and serial code in the same function Run on a pool of MATLAB resources Single Program runs simultaneously across workers Distributed arrays, message-passing Multiple Data spread across multiple workers Data stays on workers 4

5 A mental model for SPMD END x = 1 spmd y = x + 1 end y x 1 y = x + 1 x 1 y = x + 1 x 1 y = x + 1 x 1 y = x + 1 Pool of MATLAB s Demo: Face Recognition / Interactive Do we recognize this person? Compare this image against a database. Images in database are represented using six principal eigenfaces (component images). Image set must be handled in one bite. 5

6 Dataset of Faces Single snapshot used to build eigenfaces Data set also contains same individuals pulling different expressions used to test recognition algorithm 40 individuals in 10 poses in this dataset etc [ Face Data provided courtesy of AT&T Laboratories Cambridge ] Face Recognition Algorithm Sample faces processed into eigenface components Compute Mean Face Select 6 Principal Eigenfaces Subtract Compute Eigenvectors (Eigenfaces) Facial Signatures Face = 6 i= 1 a i e i M M M 6

7 Using the matlabpool SPMD END A MATLAB block communication Mostly Data Parallel TOOLBOXES PARFOR END BLOCKSETS A MATLAB block Independent iterations ( Parallel) MATLAB turns user code into parallel code. Agenda Develop parallel code interactively parallel applications for faster processing Data parallel applications for handling large data sets Schedule your programs to run Tips on developing parallel code 7

8 8Local s: Parallel Computing Toolbox Parallel Computing Toolbox Easily experiment with explicit parallelism on multicore machines Rapidly develop parallel applications on local computer Take full advantage of desktop power Separate computer cluster not required Distributed Applications Compute Cluster MATLAB Distributed Computing Server Client Machine Job CPU CPU TOOLBOXES Parallel Computing Toolbox Scheduler CPU BLOCKSETS CPU 8

9 Distributing s ( Parallel) Processes Time Time Programming Distributed Jobs ( Parallel) CPU multiplication Job CPU division Scheduler CPU addition CPU subtraction 9

10 Parallel Applications Compute cluster MATLAB Distributed Computing Engine Client Machine CPU Lab Job Lab CPU TOOLBOXES BLOCKSETS Parallel Computing Toolbox Scheduler Lab CPU Lab CPU 19 Programming Parallel Jobs (Data Parallel) We divide the work between the labs by having each lab calculate the integral of F over a subinterval [0, 1] as shown in the picture. >>edit parallel_pi.m 20 10

11 Third-Party Schedulers Client Machine Computer Cluster MATLAB Distributed Computing Server CPU TOOLBOXES Parallel Computing Toolbox Third-Party Scheduler Scheduler CPU CPU BLOCKSETS CPU Support for Schedulers Direct Support TORQUE Open API for others 11

12 Agenda Develop parallel code interactively parallel applications for faster processing Data parallel applications for handling large data sets Schedule your programs to run Tips on developing parallel code Factors to Consider for Speeding Up Your Code Share code and data with workers efficiently using FileDependencies or PathDependencies There is always an overhead to distribution Don t make a task too small Combine small repetitive function calls into one larger one Use the M-lint and parallel profiler (mpiprofile) to identify slow code Minimize I/O 12

13 Q & A Thanks!! 13