This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: gcc compile-time performance
From: Daniel Egger <degger@fhm.edu>
Date: 18 May 2002 20:02:02 +0200
Actually one of the problems is that the work is shifted from the CPU to
the compiler by creating more complicated monster CPU's, be that
superscalar hyperpipelines VLIW cpus or something else.
Support for these kinds of processors is not the reason
GCC is so slow. In fact, the performance of scheduling and
packing is quite phenomal with the DFA state machine
infrastructure done by Vladimir Makarov.
Most of the slowdowns are from front-end specific issues, GGC, and
stupid algorithms in the optimization passes. Those can all be
cured, some of which quite easily.