Re: [Librsb-users] confusing config error on cygwin
A shared memory parallel sparse matrix library including Sparse BLAS.
Brought to you by:
michelemartone
From: Michele M. <mic...@us...> - 2015-06-22 14:51:40
|
On 20150622@16:36, Marco Atzeri wrote: > On 6/22/2015 3:38 PM, Michele Martone wrote: > >On 20150622@15:26, Marco Atzeri wrote: > >>On 6/22/2015 1:57 PM, Michele Martone wrote: > >>>On 20150622@13:34, Marco Atzeri wrote: > >> > >>> > >>>If you have a suggestion about how to replicate the problem or how to > >>>fix it, I'll use it... > >> > >>I will look on it. > >Thanks. > > > > the build problem is due to the lack of usage of $(EXEEXT) in the > Makefile.am for the binary programs. > > See patch as example, I can still miss some places in it. Ok: --- I'll apply $(EXEEXT) where needed --- thanks ! > I was forced to disable the "scripts/readme-tests.sh" > as the rule was failing and creating a crap script. > We need to find a better way for it. If the generated script contains garbage is probably because one of the two: - the rule contains a non-standard expression - the tools involved (echo,grep,sed) behave out of standard > Unfortunately the real test fails immediately : > ... > ERROR 0xffffe000 : Memory hierarchy info failed to be detected. You > can bypass this by setting a meaningful > RSB_USER_SET_MEM_HIERARCHY_INFO environment variable. > ./scripts/readme-tests.sh: line 2: 8500 Segmentation fault > (core dumped) ./rsbench -oa -Ob -f ${srcdir}/A.mtx -qH -R -n1 -t100 > --verbose > Makefile:3052: recipe for target 'mtests' failed > > ----------------------------------------------------- > > Suggestions ? > Marco The following: export RSB_USER_SET_MEM_HIERARCHY_INFO="L2:4/64/512K,L1:8/64/32K" shall trick "ERROR 0xffffe000" not to occur. If this does not work I'll have to look more carefully into it. Please let me know. |