Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!snorkelwacker.mit.edu!paperboy!meissner From: meissner@osf.org (Michael Meissner) Newsgroups: comp.lang.perl Subject: Re: Perl 4.0 on Ultrix DECStation Message-ID: Date: 25 Feb 91 19:15:35 GMT References: <1991Feb23.041953.29490@uvaarpa.Virginia.EDU> Sender: news@OSF.ORG Organization: Open Software Foundation Lines: 57 In-reply-to: ted@evi.com's message of 23 Feb 91 04:19:53 GMT In article <1991Feb23.041953.29490@uvaarpa.Virginia.EDU> ted@evi.com (Ted Stefanik) writes: | Important Conditions: | | OS: ULTRIX 3.1 (UWS V2.1 Rev 14) | Hardware: DECStation 2100 | Compiler: cc (MIPS ccom1.31) | Optimization: -O -Olimit 3000 | Parser: bison -v | Malloc: perl's malloc | Exceptions: Disabled WAITPID, didn't optimize perly.o | Configure: All defaults except those listed above. | The "test -x" failure causes slight messiness during Configure. | | Test Failure: | | op/eval........FAILED on test 9 | op/s...........FAILED on test 40 | lib/big........../lib/bigint.pl did not return a true value at ./lib/big.t | line 2. | FAILED on test 0 | | Solution: Disabled optimization on perl.c. Then got another Test Failure: | | lib/big........FAILED on test 172 | Failed 1 test. | *** Error code 9 (ignored) | Out of memory! | | Solution: Disabled optimization on cmd.c. | | Now 4.0 Beta passes all tests. | | Notes: | 1) I found that it is fine to allow optimization when compiling perly.c. | 2) The lib/big stuff always failed on my Perl 3.0 versions too, but | since I don't use the bigint package, I never bothered tracing it | down to a C optimizer bug. Another way to 'solve' this problem is to use my patches to GCC (which provide full debug support and better omit frame pointer code). In normal circumstances, GCC produces slightly slower code than the MIPS compiler, but it seems to not have the nasty optimizer bugs that PERL has found in some of the MIPS compiler revisions. Of course the fact that I use PERL as a regression suite for new revisions of GCC with full optimization (-O, -fomit-frame-pointer, -fstrength-reduce, and -finline-functions) has nothing to do with it :-) At present, 1.37.1 patches are available for anonymous FTP on foobar.colorado.edu, and the 1.39 patches should be there shortly. Send me mail if you want to be on my beta list. -- Michael Meissner email: meissner@osf.org phone: 617-621-8861 Open Software Foundation, 11 Cambridge Center, Cambridge, MA, 02142 Considering the flames and intolerance, shouldn't USENET be spelled ABUSENET?