diff options
| author | vlefevre <vlefevre@280ebfd0-de03-0410-8827-d642c229c3f4> | 2016-11-04 10:32:55 +0000 |
|---|---|---|
| committer | vlefevre <vlefevre@280ebfd0-de03-0410-8827-d642c229c3f4> | 2016-11-04 10:32:55 +0000 |
| commit | 7db7cf7d0a77d173970a996f2825a804b997ab66 (patch) | |
| tree | 6f1fbdfd21fbd6ebf4113d0a74bb9c7eee6bb1a3 /doc/sum.txt | |
| parent | 325343888fbecc1bb1f93352daf277a10c7d715f (diff) | |
| download | mpfr-7db7cf7d0a77d173970a996f2825a804b997ab66.tar.gz | |
[doc/sum.txt] Minor improvement.
git-svn-id: svn://scm.gforge.inria.fr/svn/mpfr/trunk@10931 280ebfd0-de03-0410-8827-d642c229c3f4
Diffstat (limited to 'doc/sum.txt')
| -rw-r--r-- | doc/sum.txt | 5 |
1 files changed, 3 insertions, 2 deletions
diff --git a/doc/sum.txt b/doc/sum.txt index a36c453d9..4f2f788b2 100644 --- a/doc/sum.txt +++ b/doc/sum.txt @@ -307,11 +307,12 @@ output precision, denoted sq, for two reasons: accumulator is a multiple of GMP_NUMB_BITS. Since logn ≥ 2, we have dq ≥ 4 as wanted. +Accumulator precision: wq = cq + maxexp − minexp = cq + sq + dq. + + ┌──────────────────── wq ────────────────────┐ Accumulator: [--------]-----------------------------------] cq └─ maxexp sq + dq minexp ─┘ -Accumulator precision: wq = cq + maxexp − minexp = cq + sq + dq. - This idea is not sufficient because if the Table Maker's Dilemma occurs, the needed precision for the truncated sum will grow. In particular, one can easily reach a huge precision with a few small-precision inputs. For |