Hi!
I think I am going to do a full GCC/Dev environment install on a linux machine or maybe just XAMPP it .vs. continuing messin' with MinGW. I also see that Watcom now has their C++ compiler out there for free. Been many years since I used Watcom but it used to produce better code than just about anything else at the time by a significant margin.
I fuddled a bit with VC++ 9's stuff last night and managed to produce a build that's a bit faster than the official release candidate, I just really dislike (dont understand why MS hasnt done something about it!) the runtime deployment issues of VC.
I have not looked at DOSBOX's actual code as of yet. Given its platform independence which clearly was a goal I assume there is nothing in it dealing with multiple core CPU's?
It would really be interesting to isolate cycle munching such as the video related I/O to use a core. Years (I mean years) back I coded a graphics kernel (brother and I) called BLEAM which creamed Fastgraf and numerous others in abilities/speed... Eventually became part of the Atari ST "ST Lightning" etc etc. Once we ported the assembler primitives to the 68K CPU's.
I worked for Atari, Sierra, EMI and others back when game codin' was fun. In fact our line segment code was given to Mike Abrash and ended up in Quake. Never saw a dime from it 😀
Course' back in the DOS Days much slews of code sat based atop the interrupt structure, not shielded from it as things are today with layers of crap atop more layers of crap.
While I dont have time to toy with such things right now if ya'll have some architectural diagrams etc. of DOSBOX (to save me from having to go, "Oh... thats how they did it!" (reverse engineering) I might enjoy playing about to partition "DB" for multi-core's. Without having a clue about DB's current architecture I have no idea how easy or difficult it might be. In reality if one looks at the Quad Core CPU's things could get very interesting!
Look forward to hearing from you.