» Thursday, 6 September A.D. 2007
sourcerer birthday
I celebrated my one-year anniversary as a Sourcerer yesterday. A more or less complete list of all that I've done (leaving out some of the grotty details so you non-computer types won't be too bored reading it...):
- Made compilers on Windows and Linux generate identical object code for an x86 Linux target;
- Wrote a debugging sprite that lets GDB talk to hardware boards through a third-party protocol library;
- Reined in GCC's garbage collector from blowing itself out of the water in certain cases;
- Wrote a disassembler, simulator, and partial binutils port for an 8-bit microcontroller;
- Wrote a GDB backend for the same;
- Wrote an optimization pass for GCC;
- Forward-ported patches and did testing work on a major toolchain revision targeting a real-time operating system;
- Taught GCC to generate ABI-compliant code for a PowerPC variant;
- Improved code generation for the same, including enabling vectorization when possible;
- Debugged GCC's register allocator for convoluted cases on an m68k variant;
- Modified the build system at work to incorporate software licensing bits;
- Modified GCC to avoid generating code using the floating-point registers unless absolutely necessary (controlled by a command line switch);
- Served as release manager for two major toolchain releases for PowerPC processors;
- Started assisting with GCC's new link-time optimizer frontend.
I'm excited about the year ahead, especially the last item, as it offers plenty of opportunities to get my hands dirty with GCC hacking. The testing and debugging work has also been very valuable, but I know there are still a lot of places I can improve--particularly in my approach to debugging and making effective use of computer cycles (e.g. running builds/tests overnight so the results are sitting in front of me when I connect in the morning). I don't think I'd want to be doing anything else.
posted by Nate @ 5:28PM