Sorry for the delay in summarizing. The original question was
I've built the gnu fileutils both with SC1.0 cc and acc, and found
that the executables produced by acc are always larger than those
produced by cc. Doesn't acc optimize as well as cc? I used the -fast
option in both cases.
I forgot to say that I compared the sizes after striping, so there are
no symbol table influences.
I got about 10 replies, most of them saying that optimization is
concerned with speed, not size, so even if they're bigger they may be
faster. But there may be more than this, however. Here are 3 possible
reasons:
1. Casper Dik <casper@fwi.uva.nl> says: "acc always links with -lansi.
This library is static and can account for a 16K increase in code size."
2. Keith.Bierman@Eng.Sun.COM (Keith Bierman fpgroup) says that
>acc will duplicate some libc code to correct "suboptimal codes in lc".
>To avoid this, we can use
> acc $youroptions $yourcodes -lc
>acc will put in a -lm for you.
3. rjcronin@uop.com (Robert J. Cronin) got a big increase in size:
>I have some small utilities that compile to about 20 kbytes
>under cc, but are about 130 kbytes under acc. It appears that acc
>wants to bring in libm, which results in the size increase, where as cc
>is happy without libm.
Here's some more info on performance differences:
Richard.Czech@gmd.de (Richard Czech) says:
>I compiled the dhrystone benchmark using both compilers. The acc
>compiled executable was 20% faster than the cc compiled executable.
On the other hand, David Fetrow <fetrow@biostat.washington.edu> says:
>I have one example program (all of 20 lines) that can be made much
>faster under cc than gcc, and gcc was faster than acc.
Thanks to all who responded,
Carlos
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:07:49 CDT