Date: Fri, 6 Dec 2013 13:07:57 +1100 (EST) From: Bruce Evans <brde@optusnet.com.au> To: Bruce Evans <brde@optusnet.com.au> Cc: freebsd-numerics@FreeBSD.org, Filipe Maia <filipe.c.maia@gmail.com>, Steve Kargl <sgk@troutmask.apl.washington.edu> Subject: Re: dead code in lgamma_r[f]? Message-ID: <20131206130353.U1549@besplex.bde.org> In-Reply-To: <20131206114426.I1329@besplex.bde.org> References: <20131205173700.GA64575@troutmask.apl.washington.edu> <20131205182324.GA65135@troutmask.apl.washington.edu> <CAN5hRiV6N1arc4RDv=9JbRiXp-J9o3WzAbeZSOpAxks2ZeG%2B_w@mail.gmail.com> <20131205195225.GA65732@troutmask.apl.washington.edu> <20131206102724.W1033@besplex.bde.org> <20131206114426.I1329@besplex.bde.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 6 Dec 2013, Bruce Evans wrote: > On Fri, 6 Dec 2013, Bruce Evans wrote: > >> On Thu, 5 Dec 2013, Steve Kargl wrote: >> ... >>> If we again look at the code from __ieee754_lgamma_r(), we see >>> that sin_pi() is called if ix < 0x43300000, so by the time we >>> arrive at the 'if(ix<0x43300000)' statement we already know that >>> the condition is true. >> >> No, only for negative integers. hx<0 classifies negative values, and >> then ix>=0x43300000 classifies numbers that are so large negative that >> they must be integers, and the result is a sort of pole error. We >> are just filtering out this case, perhaps as an optimization. > > Oops, sin_pi() is only called for negative integers, so your change > seems to be correct. Just add a comment about the limited domain > of sin_pi() (it already has one saying that "x is assumed negative". > > With the limited range, we can improve more things. floor() can be > replaced by adding and subtracting 0x1p52 or 0x1.8p52 like we do for > trig and exp functions, followed by an adjustment to get floor() if > necessary. We can use irint(z) instead of bit magic or (int)z to > extract the parity bit of z z == y case. Note that fdlibm can't use > (int)z since this can overflow, and overflow is often not benign in > practice. irint(z) can overflow too, but it is easy to specify irint() > as having benign behaviour. Oops^2. s/easy/hard/. When irint(z) overflows on i386, it gives an overflow exception and a result of INT_MIN instead of benign truncation. To avoid these problems, it would have to extract the low bits like lgamma() does now, but that would slow down the usual case of a severely limited range. Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20131206130353.U1549>