[Cc to the gcc mailing-list] On 2009-09-25 02:18:55 +0200, Vincent Lefevre wrote: > Also, as EXP_BITS is the full (biased) exponent size, it seems that > the real.c comment is buggy (27 -> 26).
Looking at the history: Index: real.h =================================================================== --- real.h (revision 107860) +++ real.h (revision 107861) @@ -35,7 +35,7 @@ }; #define SIGNIFICAND_BITS (128 + HOST_BITS_PER_LONG) -#define EXP_BITS (32 - 5) +#define EXP_BITS (32 - 6) #define MAX_EXP ((1 << (EXP_BITS - 1)) - 1) [...] but real.c wasn't updated: it still has denormal number fits in 17 exponent bits; we store 27. that comes from r83133. -- Vincent Lefèvre <vinc...@vinc17.org> - Web: <http://www.vinc17.org/> 100% accessible validated (X)HTML - Blog: <http://www.vinc17.org/blog/> Work: CR INRIA - computer arithmetic / Arénaire project (LIP, ENS-Lyon)
