This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
help: real_convert aborts when targetting new cpu
- From: Tupshin Harper <tupshin at tupshin dot com>
- To: gcc at gcc dot gnu dot org
- Date: Wed, 26 Feb 2003 22:13:08 -0800
- Subject: help: real_convert aborts when targetting new cpu
I'm trying to target a processor with uncommon size/boundary settings, e.g.
#define BITS_PER_UNIT 32
#define UNITS_PER_WORD 1
#define PARM_BOUNDARY 32
#define STACK_BOUNDARY 32
#define FUNCTION_BOUNDARY 32
#define BIGGEST_ALIGNMENT 32
#define EMPTY_FIELD_BOUNDARY 32
this is similar the c4x port.
It builds, but when run, cc1 aborts in real_convert(real.c) as should
happen when fmt is null. It appears that real_convert is being called
with mode=VOIDMode, and there is no real_format_for_mode that matches.
I'm not sure why mode is being set to VOIDMode. Does anybody have any
suggestions for likely things to check? c4x is the only platform that is
laid out anything like this, and it does strange and mysterious things i
don't understand yet.
-Tupshin