Re: [Mplapack-devel] uninitialised value on iMlaenv1

[Maho N. <ma...@ri...>]

Hi Yamashita-san,
Thanks for your feedback.

gop    From: Makoto Yamashita <Mak...@is...>
Subject: [Mplapack-devel] uninitialised value on iMlaenv1
Date: Tue, 10 Feb 2009 16:06:01 +0900

> I found an unstable execution of Rgesv as the following message.
>
> ==22859== Conditional jump or move depends on uninitialised value(s)
> ==22859==    at 0x804B064: iMlaenv1(char const*, char const*, int, int, int, int) (iMlaenv.cpp:89)
> ==22859==    by 0x804B1DA: iMlaenv(int, char const*, char const*, int, int, int, int) (iMlaenv.cpp:240)
> ==22859==    by 0x804A9D3: Rgetrf(int, int, __gmp_expr<__gmpf_value, __gmpf_value>*, int, int*, int*) (Rgetrf.cpp:94)
> ==22859==    by 0x804A2D6: Rgesv(int, int, __gmp_expr<__gmpf_value, __gmpf_value>*, int, int*, __gmp_expr<__gmpf_value, __gmpf_value>*, int, int*) (Rgesv.cpp:91)
>
> The variable Mlaname might is not set correctly in Rgetrf.cpp.
>   nb = iMlaenv(1, "Rgetrf", " ", m, n, -1, -1);

no, prototype of iMlaenv is
int iMlaenv(int ispec, const char *name, const char *opts, int n1, int n2, int n3, int n4);
so I think this is correct...

Anyway if you still uncertain, please let me know.

Best,
-- Nakata Maho http://accc.riken.jp/maho/ , http://ja.openoffice.org/