After looking at some of the issues associated with a direct calls to object code, I've decided to take the approach of spawning an instance of the target application, then use the vpi interface to communicate with a shared memory or message queue. Has anyone been down this route before that is willing to share their experiences?
Advantages of this approach "Clean" VPI interface to work with Less development time Disadvantages More modules to compile and keep track of netlist, ngspice-vpi.o vhdl executable, ghdl-vpi.o, (greater end user complexity) Use of less portable data transport engine OS data transport model probably slower than direct function call. Thanks for any insights Mike On Mon, 2010-10-11 at 19:18 +0200, [email protected] wrote: > Hi Mike ! > > On Mon, 11 Oct 2010 10:57:44 -0500, Mike Crowe <[email protected]> > wrote: > > Hello > > I am trying to develop a mixed mode simulator. To that end I have > > implemented a vpi style interface into ngspice. The next step is to > > make calls from the ngspice-vpi code to GHDL complied code. > > Where can we find your work ? That would be interesting to see > and add to "external resources" of the GHDL page (if we ever > have write access to the wiki) > > > I see that whygee as done some work documenting the call structure to a > > GHDL compiled function and has provided some nice example code > > i'm happy that it's useful to others, it was some tough work > but i see it was worth it :-) > > > I also understand from Tristan that the call structure is to be > > considered unstable. > > it's "undocumented" so i understand this as well. > > > Has there been any other work done on creating a "C" wrapper that would > > help isolate the instability of the ABI that anyone is willing to share? > > adding a new layer is probably not a good and sane idea. > this layer would add its own quirks and make everything more complex > overall... > > How I see it : currently it's undocumented and yet it works, and > i bet it will continue so for a while. > During this "while", we can work together on this list to define > a definitive ABI. I don't know if Tristan is willing to share > his thoughts on this matter but that seems necessary. > He makes the final decisions so let's provide him with feedback. > > > > Have you whygee, or anyone else, done any further work on this? > > I am now pushing in other directions, but i keep an interested and > concerned > eye on this subject (and many more). > > > Alternatively, is it possible to use the GHDL vpi interface to get and > > set values within the GHDL code? This would get rid of the ABI > > instability all together. > > I don'k know, i stick to the C ABI at this moment. > I'll probably have a look soon because i have new needs. > > > Thanks in advance > keep us informed :-) > > > Mike > > yg > > _______________________________________________ > Ghdl-discuss mailing list > [email protected] > https://mail.gna.org/listinfo/ghdl-discuss _______________________________________________ Ghdl-discuss mailing list [email protected] https://mail.gna.org/listinfo/ghdl-discuss
