home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!pmafire!news.dell.com!swrinde!zaphod.mps.ohio-state.edu!cis.ohio-state.edu!ucbvax!cgl!cgl.ucsf.edu!gregc
- From: gregc@socrates.ucsf.edu (Greg Couch%CGL)
- Newsgroups: comp.sys.sgi
- Subject: Re: 3.10 Based C++/Fortran 77
- Message-ID: <gregc.715642054@cgl.ucsf.edu>
- Date: 4 Sep 92 21:27:34 GMT
- References: <32258@adm.brl.mil> <p5aa664@twilight.wpd.sgi.com> <1992Sep2.171736.10178@spatial.com> <pdcguvk@twilight.wpd.sgi.com>
- Sender: news@cgl.ucsf.edu (USENET News System)
- Organization: UCSF Computer Graphics Lab
- Lines: 26
-
- wsj@triton.wpd.sgi.com (Bill Johnson) writes:
-
- >In general, you'll find the 3.10 compilers to be significantly more reliable
- >than the 2.20 compilers. There is no easy way to run both simultaneously.
- >Mixing and matching 2.20 pieces with 3.10 pieces will not work, in general.
-
- We have found the 3.10 optimizer (-O, -O2) to be less reliable than the
- 2.20 optimizer. In particular, watch out for loops where the increment is
- done somewhere within the loop. For example (bug already filed with customer
- support),
-
- result = 0;
- len = strlen(s);
- while (len > 0)
- if (s[--len] == '*')
- result = 1;
- return result;
-
- returns 1 even when the string doesn't have an asterix in it.
-
- Unfortunately, we have inherited 10's of thousands of lines of code, that used
- to work, with the above idiom scattered everywhere. And rather than examine
- each line, we now compile it -O1 with the consequential slower performance.
-
- - Greg Couch
- gregc@cgl.ucsf.edu
-