Hi there ive hacked a few lines into asco for it to support gnucap. as it seems, there is no asco list and the asco deveoper is busy with other things. somebody might be interested anyway.
ive tried to implement a serious pipeline based interface between asco and gnucap. it was fun to implement, but i even expect some speedup, from not having expand netlists and find operating points over and over again -- depending on the use-case. the inverter example shows that it works. (i have not yet bothered to make pipelining work for the other simulators, but that should be quite an easy thing to do.) get a clone of my asco repo typing git clone git://tool.em.cs.uni-frankfurt.de/git/asco --branch gnucap the current status of README.gnucap is attached to this mail. regards felix PS: there are mails missing in the list archives. must be something wrong with http://lists.gnu.org/archive/html/gnucap-devel/
communucation with gnucap: - a gnucap process is spawned and <file>.gc is fed unchanged into it. - parameters then are passed to gnucap (no fancy #foo# substitution) - measurements are executed. if a measurement needs a simulation, it will be executed (if not already done) a simulation is defined in .cfg under # Analysis # with a line foo:bar:arg1:arg2:[..arg9:] the template in extract/foo will be used as the simulation command. here #ARGX# will be replaced by the arguments foo, arg1, ... arg9. a measurement is a measurement as before except it references a simulation. for example, command:as:before:foo: does a measurement after foo has been run. a slightly modified gnucap source is available. just git clone git://tool.em.cs.uni-frankfurt.de/git/gnucap --branch asco BUGS: - missing trailing ":" in cfg may cause trouble. - asco-test does not work (yet?) - gnucap backend does not support no pipeline - probes need to be announces manually - needs modified gnucap to run, because: - nop-command needed to trick +-continuation - echo command needed to advertise end of measurements and fflush - io_trace must not report to stdout - benchmark needed - how does .mark influence dcop-speed? - how does pipe influence speed? - does MPI work? - update asco.pdf (source not available?)
_______________________________________________ Gnucap-devel mailing list [email protected] https://lists.gnu.org/mailman/listinfo/gnucap-devel
