Star-P, Lecture Slides - Assembly Programming, Slides of Assembly Language Programming

Download Star-P, Lecture Slides - Assembly Programming and more Slides Assembly Language Programming in PDF only on Docsity!

Prof. Alan Edelman, MIT.

6.189 IAP 2007 MIT

6.189 IAP 2007

Lecture 13

Star-P

The Inside Story behind InteractiveSupercomputing's Star-P Platform for HighPerformance Computing for MATLAB(r)Alan EdelmanMassachusetts Institute of Technology

Professor of Applied MathematicsComputer Science and AI Laboratories

Interactive Supercomputing

Chief Science Officer

Company

-^

Background:– Started in 1995, Founded in 2004– Parallel Computing Harder than most realize– Technology: Star-P software platform supporting automatic

parallelization and interactive execution of desktop technicalapplications on parallel servers

• Platform: Clients: MATLAB, MATHEMATICA, PYTHON– Platform: Engines, your code, etc. -^

Value:– Modern Client/Server Parallel Computation– OPEN PLATFORM– Can plug in existing parallel and serial software seamlessly– Years of experience

4

Client/Server Parallel ComputingThe Client (a math lab) is the browser Slide contains trademarks owned by private corporations. Use of logos or registered trademarks does not imply endorsement

Web vs traditional•^

Bank/financial

-^

Email

-^

Travel

-^

Photos

-^

MIT Grades

-^

Your ParallelComputing!

Familiar Desktop Tools

Star-P Client

• Connects to server• Redirects library calls• Optimizes serial code

Star-P Computation Engine

Data-Parallel Computations

Task-Parallel Computations

OpenConnect Library API Link

Data-Parallel Computations

• Global array

syntax

• Operations on

large distributeddata sets

• World-class

parallel libraries

ppeval syntax (parallel function)

Answer does not depend on distribution:

Parallel computers need shapes toenter from all sides.

-^

a=rand(500,500,200p);*

-^

[u,s,v]=ppeval(‘svd’,a); % default svd on z-dim

-^

a=rand(500,500p,200);*

-^

[u,s,v]=ppeval(‘svd’,a); % default svd on z-dimanyway

P1P2P3P

P1 P2 P3 P

Task-Parallel Computations

• Multiple independent

calculations

• Simple, intuitive

w/Star-P’s abstraction

• Plug in popular

computation engines

Star-P OpenConnect Library API

• Leverage data-

and task-parallellibraries, solvers

• Commercial and

open source

• Enable access

through desktopVHLLs

Hardware Accelerators

• Embed compute-

intensive algorithms

• FPGAs, GPUs, etc.• Library functions,

called from desktopapps

High-speed I/O

• Native parallel I/O• Direct transfer

between disk andserver CPUs

• Eliminate

client/server datatransfer

• No need to manually

break up files

Classroom Homework• The Buffon Needle Problem^ function z=Buffon(a,b,l, trials)r=rand(trials,3);x=ar(:,1)+lcos(2pir(:,3)); y=br(:,2)+lsin(2pir(:,3));inside = (x >= 0) & (y>=0) & (x <= a) & (y <= b);buffonpi=(2l(a+b) - l^2)/ (ab(1-sum(inside)/trials));

a b

Buffon(1,1,1.5,1000*p)