ryuon
Brownian dynamics
検証 run - 粒子編。
昨日の結果を眺める。
rej 版でも見てみる。やはり dt をかなり小さくしないとダメだな。
JGdP の場合も、dt が大きいとFBの計算でこける。
semi-implicit predictor-corrector スキームを自分なりに実装。
Jendrejack et al は1次スキーム。 mid-point や Ball-Melrose 並の semi-implicit を組もう、ということ。
Somasi et al や Hsieh et al のような、非線形方程式解法のトリックは 今のところ、なしで。
ARPACK problem
In a few days, I was stuck in some problem on ARPACK.
Some check in libstokes using ARPACK failed on openSUSE
(although the same code is just fine on FreeBSD and SUSE-Linux).
By comparing the behaviors between openSUSE and FreeBSD step-by-step,
it is found that the problem is coming from "dlarnv",
the random number generator in LAPACK, which returns
"nan" in some element.
At first I suspect some memory-leakage in my side, but it looks fine.
Then, I googled and found the site "ARPACK/LAPACK Symbol Duplication"@tiker. After applying the patch for SECOND,
all checks are fine on openSUSE. Thanks for the pactch.Hopefully, this posting will save somebody some trouble.
That's true.
Now I guess that some trouble of ARPACK on FreeBSD before
is also related to this.
12/14/2007: the problem is just gone (?)