|
Re: MAME 0.182
01/26/17 03:39 AM
|
|
|
> That is completely work in progress and a proof of concept. It will most likely never > be included in MAME directly because of the hassle around static linking which would > completely break the separation of abstraction (netlist description) and > implementation.
I really don't understand why compiling and linking the optimised solvers into MAME is worse than building the netlist descriptions into MAME, which we already do.
> The design is that by using a hash the code looks whether a suitable solver exists in > the shared library. If not, well, netlist uses the normal code. If a special solver > exists - specifically for the matrix size and occupation distribution - it will take > it from the shared library. Easy as that - if we don't have it you won't get it. But > it's completely optional and proof of concept.
Yes, but the fact that it looks for and potentially loads a shared library from the working directory is a security issue in itself. Why can't it look for a solver specifically for the matrix size and occupation distribution in the executable itself?
|
|