[rescue] Re: Being jobless
Francisco Javier Mesa-Martinez
lefa at ucsc.edu
Tue Jul 29 22:05:11 CDT 2003
> No, that it generates bad code for RISC platforms. While the core
> (on the other side of the microcode) of many x86 processors is RISC,
> IA32 is not RISC.
It generates OK code IMHO with respect to MIPS and SPARC, it is NOT as
good as the vendor provided tools, but hey you can not beat the price. And
it is a known shortcomming that their RISC scheduler is not as good as the
one for the CISC backend. But that is part of the deal with RISC designs
anyway... they tend to produce complex compilers :). A lot of the moaning
and bitching about GCC is a bit unjustified IMHO... but if one really
needs to extract every ounce of performace out of a SPARC or MIPS, they
should be using the vendor tools anyways... If one is to use a large
Origin, one should use the MIPS C with the autoparallelization packages,
or they should be shot IMHO :).
It will be interesting to see if the RISC scheduler improvements made by
apple and IBM can be returned to the main GCC distribution, last time I
heard there were some legal issues that prevented the GCC team from
accepting apple's changes/improvements into the main gcc distribtion.
> > Again, I thought the comment was about GNU C not being a good compiler.
> It isn't a good compiler. It sucks on the majority of the platforms it
> runs on. If it runs stellar on one platform, that's hardly indicative
> of a good product.
THis is very unfair really, it is a very portable compiler. And as such it
is the most succesful compiler out there. It is an amazing feat to have a
compiler so portable that can produce relatively decent code on so many
Trying to be the devil's advocate I guess.....
More information about the rescue