diff options
Diffstat (limited to 'mpfr/BUGS')
-rw-r--r-- | mpfr/BUGS | 25 |
1 files changed, 21 insertions, 4 deletions
@@ -4,12 +4,29 @@ Known bugs: * The overflows/underflows are not yet implemented in all functions. +* problem with tsqrt on ia64-hpux (problem is that variable a in check4 has + a different value than in the call, may be a memory problem, since + it disappears when we remove the mpfr_set_str instruction, and + disappears without -O2) +mpfr_sqrt failed for a=4.15383748682786210282e+34, rnd_mode=GMP_RNDZ +expected 0.10010010111100001100100110000001011110111111001010000[00000000000]E53 +got 0.10110101000001001111001100110011111110011101111001100[00000000000]E58 + +* texp fails on ia64-hpux: mpfr_exp_2 and mpfr_exp3 for prec=601 + +* acosh is wrong near EMAX (idem asinh): + rounding mode GMP_RNDU: + 6096987078287.281250 ulp(s) for x=1.75062535692767808555e+308 + [mpfr: 7.097561817130842e2 libm: 7.10449328893644064919e+02] + +Potential bugs: + * Possible integer overflows on some machines. * mpfr_set_d may give wrong results on some architectures. -Potential bugs: +* Error analysis in mpfr_sin is incorrect. A test has been added to + show the bug. Check the other functions too... -* get_str.c: condition len != n + 1 is always satisfied at the end. - This means that there are too many reallocations (possibly useless - ones) and a possible bug. +* mpfr_hypot may fail for x very large, y very small and a very large + target precision. |