Subject: Re: Procedure for gcc maiming?
To: Thor Lancelot Simon <tls@panix.com>
From: Zdenek Salvet <salvet@nyx.dcs.muni.cz>
List: current-users
Date: 04/25/1994 17:29:08
> Can someone in the know give me a short lesson on what need be done to
> properly maim a gcc distribution to fit in the NetBSD source tree?  Which
> symbols need be defined to what and other niceties would also be a help, along
> with where to define them, since the standard GNU mechanism for this seems to
> have vanished.  
> 
> This information passed along, I'll be able to make the iGCC 2.4.0
> (Intel-worked-over GCC 2.4) distribution available with NetBSD modifications
> as needed.  On my Pentium, it makes some FP-intensive code run up to 60%
> faster than the stock GCC 2.4.5, and Intel claim up to 15% improvement on the
> 486 as well.
  
  I don't recommend to include iGCC 2.4.0 into NetBSD as primary compiler;
  I got internal errors several times when building Xfree server.

( e.g. 
   cfbply1rctG.c: In function `cfbFillPoly1RectGeneral':
   cfbply1rctG.c:305: internal error--insn does not satisfy its constraints:
   (insn:HI 449 1143 450 (parallel[ 
               (set (reg:SI 0 %eax)
                   (div:SI (reg:SI 0 %eax)
                       (reg:SI 1 %edx)))
               (set (reg:SI 1 %edx)
                   (mod:SI (reg:SI 0 %eax)
                       (reg:SI 1 %edx)))
           ] ) 104 {divmodsi4} (insn_list 1143 (nil))
       (nil))
   gcc: Internal compiler error: program cc1 got fatal signal 6
                                                                    )

  On the other hand, it may be good for experiments (necessarily backed 
  with stock GCC 2.4.5 )

                                    Zdenek Salvet
                                    salvet@nyx.dcs.muni.cz

------------------------------------------------------------------------------