diff options
author | vlefevre <vlefevre@280ebfd0-de03-0410-8827-d642c229c3f4> | 2016-09-05 08:50:16 +0000 |
---|---|---|
committer | vlefevre <vlefevre@280ebfd0-de03-0410-8827-d642c229c3f4> | 2016-09-05 08:50:16 +0000 |
commit | c96be5779880d5a9b482f9b6ecd133d57e66738d (patch) | |
tree | 73e6c4d0ac8a3b3480aef44d5bb352179950b7e4 /TODO | |
parent | 192e5406a88ed01405fafa5f5945f59f1f4bc687 (diff) | |
download | mpfr-c96be5779880d5a9b482f9b6ecd133d57e66738d.tar.gz |
[TODO] Updated item on tzeta:
* removed the old problem: the fact that the traces for the trunk and
for r9954 are different is normal (fixes in mpfr_can_round affecting
tgeneric.c), and the timings depend on GMP_CHECK_RANDOMIZE and seem
equivalent in average with "test_generic (..., 5);" in both cases;
* added the fact that tzeta has been much slower since r9848, at least
with the x86 32-bit ABI.
git-svn-id: svn://scm.gforge.inria.fr/svn/mpfr/trunk@10789 280ebfd0-de03-0410-8827-d642c229c3f4
Diffstat (limited to 'TODO')
-rw-r--r-- | TODO | 13 |
1 files changed, 2 insertions, 11 deletions
@@ -608,17 +608,8 @@ Table of contents: (The one done in r10573 allowed us to find a bug even without assertion checking.) -- after the fix of mpfr_can_round in r9883, tzeta has a different trace - from r9882, which may be expected as mpfr_can_round has an influence - in tgeneric.c, but a closer look is needed: check only the first - difference (with various GMP_CHECK_RANDOMIZE values), since after it, - the random numbers do not match. - With the 32-bit ABI, tzeta r9955 is even 2.17 times as slow as r9954; - the corresponding change: - - test_generic (2, 70, 5); - + test_generic (MPFR_PREC_MIN, 70, 5); - Find the real cause, whether a wrong result could be obtained before - the fix r9883, and whether the slowdown can be avoided. +- tzeta has been much slower due to r9848, at least with the x86 32-bit ABI; + see whether this was expected. ############################################################################## |