Same Old Same Old
Dear Intel,
Please make the Intel Pentium 4 sane.
Love, Ross.
Intel Pentium 4's are weird. I do a pretty-much null change (from looking at the assembler a jump table was re-ordered) and performance on the P4 is 30% better, but the same binaries on a P3 have the expected identical performance. I then optimised away some redundant code (again, looking at the assembler shows many instances of large chunks removed) and the program actually slows down. This is starting to make optimising for speed not a Fun Game. At least the ever-handy Meld is showing its worth again, quickly and easily displaying the differences in two assembler files.
monitor-calibration-tool is going well, I've rewritten most of it and the crack is slowly being removed. Hopefully I'll be able to get another release out next week, which will be the first release to be useful (unless you are really weird and want to calibrate your screen for gamma 1.0)
Finally, it appears that IBM don't want my Java Reflection article, the corporate gits. I blame SCO for this, I'm not sure how but I think blaming SCO is a positive action here. If my second choice doesn't want to pay for it, then an absolutely fabulous article (if I do say so myself) will be online here next week or so.
