DoctorOctal wrote:The link you posted above is to an illegitimate, mangled copy of Ted’s original blog entry that describes what he’s done:
<
https://microcorelabs.wordpress.com/2020/11/12/ibm-pcjr-accelerator-mcl86jr/>
You can backtrack to earlier entries for more of the story.
He ended up designing a plug-in board with RAM as discussed in the thread on this forum. He made a few copies, and I have one.
I’ve done some preliminary testing, and it mostly seems to work well. Benchmarks show AT-like speed and applications run noticeably faster. I’ve run into some compatibility issues, however, that aren’t just because of the speed difference. For example, even in cycle-accurate mode, some benchmarks lock up the machine, and the built-in diagnostics are very weirdly screwed up. I think there may be some bugs to work out but haven’t yet had a chance to finish my testing and follow up with Ted.
Yes, there are defiantly some differences in the way the PCjr acts with my 8088 core versus the PC/XT... Everything I ran on the IBM PC/XT seems to run correctly, however when run in a PCjr most, but not everything, does... On the XT I had success running games, utilities, Lotus-123, all versions of Flight Simulator, every version of DOS, Windows(!!), and even the 8088MPH demo. 100% of the stuff I tried worked flawlessly... But some stuff on the Jr just refuses to cooperate 100%. It's true that the CTL-ALT_INST PCjr built in diagnosis seems to fail in a major way, so Im really hoping this could be a major clue where the problem is, but I haven't figured it out yet.
I do have one theory though... I wonder if there is something about the way my bus interface works (minimum mode) , or something else about my core that makes some software believe it is actually an IBM XT rather than a PCjr.. If so this software may start accessing components which don't exist in the PCjr or have been moved... One thing is that sometimes when running the MCL86jr in accelerated mode Norton's SI tool thinks the CPU is a V20 which, I believe, it is because of the prefetch queue.. I use the same sized queue; however maybe my implementation is slightly faster or slower that some software expects...
Anyway... Its the same MCL86 CPU core used in the XT and the PCjr with only the bus interface changed from Max to Min mode for the PCjr... I would love to know why some apps have issues; however most are ok with the MCL86jr and run fine both cycle accurate and accelerated. The code is open-sourced, so maybe someone can see something obviously wrong which would be great! The FPGA is also reprogrammable so if the bug is found the existing boards can be updated..
So, that's the scoop...
Thanks,
-Ted