[tree-ssa mudflap] ignore-reads speedup, bug fixes

law@redhat.com law@redhat.com
Tue Feb 3 08:38:00 GMT 2004


In message <20040130212846.GA12214@redhat.com>, "Frank Ch. Eigler" writes:
 >Hi -
 >
 >The following patch does several things.  The biggest is adding
 >a "don't instrument reads" suboption to the compiler, which basically
 >tells it not to instrument pointer *reads*.  For some baby benchmarks,
 >this change can speed up a fully instrumented program by a factor of two,
 >making it "only" a factor of two slower than an uninstrumented one.
On big stuff (say GCC itself) being able to ignore reads can easily be
the difference between something that runs in a minute or two or runs
for an hour or more.

jeff



More information about the Gcc-patches mailing list