Also check out the Icking Archive http://icking-music-archive.org/index.php which has many many many scores (I believe about 16k). Main difference with Mutopia is that they do not necessarily come with sources. V.

Let me give you a system from high order finite elements and see your iterative method crash and burn. And SOR has not been the standard choice for the last 30 years. Some version of Conjugate Gradient methods (such as GMRES or BiCGstab) is what almost everyone uses.

I'll have to read the paper to see precisely how widely applicable it is, but SDD systems are not quite as interesting as they claim. Nor is solving linear systems with them. Netflix and Google &c do not solve linear systems with their SDD matrices: they need the dominant eigenvector. To find that you run the power method, which is matrix-vector multipication. Much simpler. CFD and such applications they mention do give rise to SDD systems, but only in the case of diffusion, no convection. Also, diagonal dominance comes from using 2nd order methods (finite element or finite difference) only. Higher degrees of approximation gives SPD (symmetric positive definite) systems. Now, if they could crack that problem it would be truly interesting. V.

The guy only seems to record. That's the easy part. He says that he makes master mixes without compression or EQ. True, mastering engineers like to have that. However, the main reason he does that is because those tools simply don't exist on Linux. If you want to do your own mastering, where are the Waves plugins? The IKmultimedia Fairchild and Pultec emulations? Can you even hook up a UAD card to Linux? Same story for software synthesizers. With Logic you get more synths (and effects) than you'll use in a lifetime. Where is the physical modeling synth on Linux? An analog emulation of the quality of ES2? Or of Pro53, which is Mac/PC? Where are the innovative synths such as from Camel Audio? Doing my music on Linux, I'd feel like having only a recorder to score a symphony. V.