i already tried that, with the precompiled version. makes the same error.
i am using ubuntu 9.04 64 bit.
svn checkout is from today evening.
Yes, chains are generated from the commandline with the commandstring
from cgi
the devno=X seems not to work.
if i try that with 2 instances in different folders than the chain/s is
the half for each instance.
with one instance i have around 148/s with 2 instances on 2 different
cards i only get 70 in one instance and the second instance has to rest...
Sascha Krissler schrieb:
Funny thing...
With selfcompiled and with precompiled version i get with the test data
this result:
make test
./c --algorithm A51 --condition rounds:rounds=32 --implementation
sharedmem --device cuda --roundfunc
xor:condition=distinguished_point::bits=15:generator=lfsr::tablesize=32::advance=0
--work increment --consume print --logger verbose generate --chains 1024
--chainlength 3072 > testdata
Initialize implementation sharedmem...
Run 512 operations...
../tmto/device/cuda/host_side_methods.hpp(74)[void
tmto::device::cuda::cudaSynchronize()]: cuda error: unspecified launch
failure
make: *** [test] Fehler 1
with data from the cgi it seems to work.
do you mean that chains are generated with the command line from the cgi?
it looks like your compilation failed.
are you on 32bit or 64bit? if on 32bit, then you probably did not use HOST_FLAGS
in Makefile.local/Makefile (depending on your svn revision)
try 'make test' with a precompiled version and make sure you rename it
to 'c' in the same directory that you type 'make test'
______________________________________________________
GRATIS für alle WEB.DE-Nutzer: Die maxdome Movie-FLAT!
Jetzt freischalten unter http://movieflat.web.de
_______________________________________________
A51 mailing list
[email protected]
http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51
_______________________________________________
A51 mailing list
[email protected]
http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51
