From e643c954932209f5ae19d732c77bd116ecba8313 Mon Sep 17 00:00:00 2001 From: zimmerma Date: Sun, 28 Oct 2007 12:18:04 +0000 Subject: removed item on mpfr_sinh_cosh (done) added more info on item on efficiency of mpfr_sin git-svn-id: svn://scm.gforge.inria.fr/svn/mpfr/trunk@4926 280ebfd0-de03-0410-8827-d642c229c3f4 --- TODO | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) (limited to 'TODO') diff --git a/TODO b/TODO index e4f248bee..f4690347c 100644 --- a/TODO +++ b/TODO @@ -267,12 +267,14 @@ Table of contents: mpfr_random (x); for (i = 0; i < k; i++) mpfr_atan (y, x, GMP_RNDN); for precision 300 and k=1000, we get 1070ms, and 500ms only for p=400! - improve mpfr_sin on values like ~pi (do not compute sin from cos, because - of the cancellation). For instance, reduce the input to [0,pi/4], and - define auxiliary functions for which the argument is assumed to be already - reduced (so that the sin function can avoid unnecessary computations by - calling the auxiliary cos function instead of the full cos function). -- combined mpfr_sinh_cosh() [Geoff Bailey, 20 Apr 2005, - and Kaveh R. Ghazi, 17 Jan 2007] + of the cancellation). For instance, reduce the input modulo pi/2 in + [-pi/4,pi/4], and define auxiliary functions for which the argument is + assumed to be already reduced (so that the sin function can avoid + unnecessary computations by calling the auxiliary cos function instead of + the full cos function). This will require a native code for sin, for + example using the reduction sin(3x)=3sin(x)-4sin(x)^3. + See http://websympa.loria.fr/wwsympa/arc/mpfr/2007-08/msg00001.html and + the following messages. - improve generic.c to work for number of terms <> 2^k - rewrite mpfr_greater_p... as native code. - inline mpfr_neg? Problems with NAN flags: -- cgit v1.2.1