Mail Archives: pgcc/2001/01/09/17:18:26
On Tue, Jan 09, 2001 at 07:46:37PM +0000, Nick Kurshev <nickols_k AT mail DOT ru> wrote:
> I have compiled new pgcc-2.95.2.1 with following keys (three times):
> -march=k6 -mcpu=k6 -mmmx -fschedule-insns2 -O3 -fomit-frame-pointer -pipe -ffast-math
Don't do that, mmx makes your programs run slower on k6, and of coiurse
the kernel will not work with that (just as it will not work with floating
point).
> Also I have compiled BASH with new compiler with same keys as kernel, but anew compiled bash
> was not able to handle command line arguments propertly. As effect kernel at boot time could not execute
> any scripts in /etc/init.d, command "make install" does not work e.t.c.
Do not use experimental options. pgcc handles bash fine without -mmx.
> And last observation: New compiler (pgcc-2.95.2.1) is stopped too frequently ( by signals of different kind:
> such as: 4, 9, 11). As rule, gcc received signal #11 or #9 only when memory is exhausted. But in this case
> free memory was ~50Mb. Also it does not remove wrongly generated .o file after receiving such signals. (but
> gcc does it).
This indicates an overclocked machine or broken hardware (e.g. memory).
pgcc reacts in exactly the same way to signals as gcc does. If you have
any signals that you cannot reproduce this is a hardware problem. pgcc
generates tighter code and it is well possible that you gcc-compiled
programs run more-or-less reliably on overclocked chips/memoory but pgcc
ones don't.
--
-----==- |
----==-- _ |
---==---(_)__ __ ____ __ Marc Lehmann +--
--==---/ / _ \/ // /\ \/ / pcg AT opengroup DOT org |e|
-=====/_/_//_/\_,_/ /_/\_\ XX11-RIPE --+
The choice of a GNU generation |
|
- Raw text -