the relentless march of progress

From various channels, early this morning:

* bryner is really disturbed by redhat's gcc 3.3.3 (fc2) generating worse code than their 3.3.2 in fc1

< vlad> vladimir@tornado[1099]% rpm -q gcc34
< vlad> gcc34-3.4.0-1
< vlad> who needs 3.3.3?

< shaver> use gcc34?
< bryner> it's even worse

2 comments to “the relentless march of progress”

  1. Mike Hoye
    entered 4 July 2004 @ 7:24 pm

    …crappy reflexes…

    For those of us who are happy when something actually compiles, how do you actually determine this? My kung-fu is not powerful, but I’ve written a basic compiler, and as I understand it the optimizations are the really hard bit. Presumably this is where you notice it, with the new compiler creating poor performance?

  2. entered 4 July 2004 @ 9:36 pm

    Well, one thing you do is collect metrics about code size and page-load time that leave you hanging on a 10ms or 10KB change like it was the three-hour IMAX-filmed final episode of “Survivor: Friends”. And then you have people like bryner who spend a lot of time thinking about compiler output and how to trick gcc into producing code that is fast and small. I can barely read x86 assembly, so I avoid that sort of duty.