Scientific/high-performance computing

• Linux in High-Performance Computing, Linux in HEP. Avalon and Beowulf are parallel machines running Linux. I've also seen Extreme Linux, but it's a bloody silly difficult-to-use graphical site, and if there's anything of value there it's very well hidden.
• TOP500: listing the fastest 500 supercomputers.
• egcs was `an experimental step in the development of GCC, the GNU C compiler'. In early 1999, GNU announced that egcs would become the GNU C compiler, and be renamed GCC.
• UltraPenguin is a version of Linux for Sun Ultras. I don't have any pointers, yet.
• See also the Linux section of my machines page.

Numerical analysis

• William Kahan is at UCB, and was one of those responsible for the design of IEEE arithmetic. There are several papers and lecture notes here on numerical analysis, floating point representations, and why a Pentium or an old Mac is more accurate than a Sun!
• `Basic Issues in Floating Point Arithmetic and Error Analysis': lecture note from Jim Demmel, UCB.
• There's a good deal of incidental chatter (some of it particularly well-informed) about numerical analysis and floating-point performance in the newsgroup comp.lang.fortran.
• Sun Microsystems' Numerical Computation Guide, which covers the IEEE Standard in detail.
• An article Handling floating-point exceptions in numeric programs by John R Hauser, and a software implementation of IEEE called SoftFloat, by the same author.
• Floating Point topics, including FAQ (not sure how authoritative it is, though)
• IEEE-754 info from LORIA
• Note about SIGFPE on alpha Linux, and a corresponding note about the alpha's IEEE FP support
• Floating point on Linux:
  • Comments and materials on proposals (to X3J11) for better floating point proposals for numerical calculations in C.
  • wmexcep: code for handling floating point exceptions