Date: Fri, 17 May 1996 18:48:19 +0200 (MET DST) From: Peter Mutsaers <plm@simplex.nl> To: nate@sri.MT.net Cc: freebsd-questions@FreeBSD.org Subject: Re: make world with gcc version 2.7.2 or pgcc 2.7.2.9? Message-ID: <199605171648.SAA04818@plm.simplex.nl> In-Reply-To: <199605170437.WAA25870@rocky.sri.MT.net> (message from Nate Williams on Thu, 16 May 1996 22:37:54 -0600)
next in thread | previous in thread | raw e-mail | index | archive | help
>> On Thu, 16 May 1996 22:37:54 -0600 you wrote:
>> Did anyone try to rebuild everything, including the kernel, with gcc
>> version 2.7.2 or pgcc, and with -O2 or higher?
NW> I believe Kaleb Keithley rebuilt X with one of them and it didn't work.
NW> This implies to me it's a compiler bug, as there are quite a few known
NW> optimizer bugs in gcc 2.7.2 for the x86, and at least one generic
NW> optimizer bug that was posted on gnu.gcc.bug.
The generic bug was, I think, shown by this example:
#include <stdio.h>
int A[3];
unsigned int b = 3;
void printit(void)
{
int i;
for(i = 0; i < b; i++)
fprintf(stdout, "A[%d] = %d\n", i, A[i]);
}
int main()
{
int i;
for(i = 0; i < b; i++) *(A+i) = i - 3;
printit();
return 0;
}
gcc 2.7.2 doesn't show correct results. gcc 2.7.2.9 (the current pgcc
release) has this fixed however. It also has fixed the x86 bug that
made it necessary when using -O2 or higher to add -fno-strength-reduce.
NW> 'probably bogus' and faster code. Since there are known bugs,
NW> given a system the size of FreeBSD the possibility of the bug
NW> being tickled *somewhere* is high. And, given that the speedups
NW> for stock gcc aren't stupendous (unlike pgcc which can be quite
NW> high, but the code generation is suspect in many cases) it's not
NW> worth it.
Hmm. Do the pentium patches introduce new bugs?
--
______________________________________________________________________
Peter Mutsaers | Abcoude (Utrecht), | "Quod licet bovis,
plm@simplex.nl | the Netherlands | non licet Jovi."
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199605171648.SAA04818>