in one case, airy_bound accidentally chose the worst formula instead of the best, always giving inf near arg(z) = pi/3

This commit is contained in:
Fredrik Johansson 2016-03-16 01:21:56 +01:00
parent d6097d77d7
commit e357fa38e2

View file

@ -212,7 +212,7 @@ acb_hypgeom_airy_bound(mag_t ai, mag_t aip, mag_t bi, mag_t bip, const acb_t z)
/* multiply by exp(-2 pi i / 3) in upper half plane
and by exp(2 pi i / 3) in lower half plane, to stay close
to positive reals */
if (arf_sgn(zim) <= 0)
if (arf_sgn(zim) >= 0)
acb_conj(z1, z1);
acb_mul(z1, z1, z, wp);