Support 64 or 80 bit long double for i960

Jeffrey A Law law@cygnus.com
Tue Jan 19 16:20:00 GMT 1999


  In message < 199901192004.PAA11170@tootie.to.cygnus.com >you write:
  > 
  > 
  > The following patch provides implementation long double as 64 or 80
  > bit floating numbers.
  > 
  > Vlad Makarov
  > 
  > 
  >  1999-01-19  Vladimir N. Makarov  <vmakarov@cygnus.com>
  >  
  > 	* invoke.texi (-mlong-double-64): New option description.
  > 
  >  1999-01-19  Jim Wilson  <wilson@cygnus.com>
  >  
  >  	* libgcc2.c: Change all uses of LONG_DOUBLE_TYPE_SIZE to
  >  	LIBGCC2_LONG_DOUBLE_TYPE_SIZE.
  >  	(LIBGCC2_LONG_DOUBLE_TYPE_SIZE): New.  Set to LONG_DOUBLE_TYPE_SIZE
  >  	if not defined.
  >  	* i960/i960.h (MULTILIB_DEFAULTS): Define to mnumerics.
  >  	(CPP_SPECS): Add -mlong-double-64 support.
  >  	(TARGET_FLAG_LONG_DOUBLE_64, TARGET_LONG_DOUBLE_64): New.
  >  	(TARGET_SWITCHES): Add -mlong-double-64 support.
  >  	(LONG_DOUBLE_TYPE_SIZE): Likewise.
  >  	(LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Define.
  >  	* i960/vx960-coff.h (MULTILIB_DEFAULTS): Define to msoft-float.
  >  	(CPP_SPECS): Add -mlong-double-64 support.
  >  	* i960/t-960bare (MULTILIB_OPTIONS): Add mlong-double-64.
  >  	(MULTILIB_DIRNAMES): Add ld64.
  >  	* i960/t-vxworks960 (MULTILIB_OPTIONS, MULTILIB_DIRNAMES): Likewise.
I don't think this is complete.  You should do some greps on Cygnus's gcc
sources for i960-80bit.  I think there are some machine independent changes
you need for this stuff to work.

jeff



More information about the Gcc-patches mailing list