forked from kostrzewa/tmLQCD
-
Notifications
You must be signed in to change notification settings - Fork 0
/
BUGS
36 lines (26 loc) · 1.46 KB
/
BUGS
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
16.04.2010: order in propagators for the heavy doublet inverted in
file -> fixed
14.05.2009: bug in reread functionality fixed.
30.05.2008: scidac checksum for gauge fields not correctly working
different result for serial and MPI run
--> fixed (Remi 02.03.2009)
problem was a wrong computation of rank for I/O writing functions only
when MPI was used
16.04.2007: enable-gaugecopy and disable-newdiracop is buggy, at least
in the 1-dim parallel case. -> fixed, wrong #ifdef in xchange_gauge
around MPI_Waitall
18.1.2007: phmc does not work with halfspinor, seg fault in Hopping_Matrix.
--> fixed
For x86_64 there seems to be a compiler bug in gcc 3.3.3. The code
is not working with the default sse2 optimisation done by the compiler
and now switched of with mfpmath=387. It has to be checked, whether this
is really a compiler bug, a strange interplay between our sse2 units and
the compiler units or a bug in the code (17.08.2004) --> fixed
bug in the serial version of the code fixed. (16.08.2004)
there are now two exchange routines for the spinor fields, since at
least on the iwarp pc cluster the version with MPI_Type_vector is not
working for small local lattices. Maybe this is a bug in the MPI
implementation, but maybe not ...
Seems to be a bug in the MPI driver for infiniband. (17.08.2004)
write and read of spinor fields not yet tested --> tested
--disable-mpi will be configured correctly, but the code is not correct. --> fixed