-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Tuesday 22 January 2002 23:59, Hal Daume III wrote:
I'm open to all ideas...
Here is one: a matlab like matrix based language that is compiled into efficient parallel shared-nothing code. You can go for dense or sparse matrices, and allow imperative features or not. The compiler decides all data distribution/replication. The general case is too difficult but you could take a very limited language, ie array expressions with a few operations like A <- A^T * (B + C * C ) An easier one would use HPF style hints from the programmer for data distribution. distribute A checker-board B <- f (A * A^T) The compiler would align B with A, ie infer B's distribution from A's looking at the statement above. Thanks, - -- Eray Ozkural (exa) <erayo@cs.bilkent.edu.tr> Comp. Sci. Dept., Bilkent University, Ankara www: http://www.cs.bilkent.edu.tr/~erayo GPG public key fingerprint: 360C 852F 88B0 A745 F31B EA0F 7C07 AE16 874D 539C -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.6 (GNU/Linux) Comment: For info see http://www.gnupg.org iD8DBQE8TiR4fAeuFodNU5wRAmsdAJ9eAsQ9gj6g4hAqGljonvL34ZX5hACgpvlr wEnS4v5eljrvR5J02FfOi8g= =LJbf -----END PGP SIGNATURE-----