At 11:12 AM 5/15/2002 +0300, SlowByte wrote: >Many thanks, this really rocks on my machine with 128MB of RAM :) I think >compiles are about a minute faster now :)
Thats good to hear. >One thing that worries me is the size of the parrot executable, right now it's >about 350k and growing. When compressed with UPX, it's 55k, but still, we >should >avoid executable bloat :) There are places where this can be fixed, one of the main culprits appears to be the pre-dereferencing. It basically duplicates everything in the core for the sake of that optimization. Don't ask me to fix it, I'm not sure what "fix" means here since it is a good idea, I'm just not sure whether the implementation is worth the bloat. My thinking is that the prederef option should be a compile time option, not a duplicate runtime. While I'm on this, this also goes for computed goto, but in a lesser degree. CG doesn't duplicate all the tables that the prederef core does. We still have 3 cores compiled in on machines that support CG, any way you look at it, but disablign CG only nets a bit of size. For now I think thats ok, because testing is the important thing and this can be a pre-production release switch. -Melvin
