View Single Post
Old 21 October 2015, 22:04   #455
Join Date: Jan 2010
Location: Kansas
Posts: 1,284
Originally Posted by eXeler0 View Post
A tool to test compatibility may be nice have, sure.. but in a larger context.. is it even relevant? Its not like the original 68k family is without issues.. how many programs crashed on a 060 when it came out?
For most ppl it will be enough that "almost everything" runs, not being concerned if the CPU actually behaves like a genuine 68k.. whatever definition that might be.
The Apollo team uses test cases for testing the core. A superscalar core with caches needs significantly more testing and verification than a simpler core like the TG68. The early revisions of the 68060 had bugs which only occurred with a specific sequence of instructions for example. Programs which crashed on the 68060 were usually a result of buggy support software or the larger caches.

Originally Posted by wawa View Post
im not sure but i think matthey did some work in that direction. ill leave it for him to tell.
Gunnar wanted me to write some test cases but I never wrote any official test cases. I did point out some code which I thought would be good to test. I didn't know what test cases were already available and I never had access to them (although I believe the current test cases are extensive and are used regularly after core changes). I did contact Dave Alsup of the Fido 68k/CPU32 processor project hoping to get him interested in the Apollo project and maybe use some of the test tools they used. Dave was really nice but busy so we didn't get very far. I was also pushing for an eventual ASIC which would have given InnovASIC a reason to work with us but Gunnar didn't think it was realistic. I also modified the ADis 68k disassembler into a 68k code analyzer which gathered statistics used in development but that is not a testing and verification tool.

Originally Posted by wawa View Post
admittedly, what it doesnt run is kickstart 3.9 apparently, but they are actively working on compatibility. so while i wouldnt expect it to be 100% compatible to each of the 68k processor options, it may have a chance (maybe) to be as compatible to them as they were to each other, or better.
The lack of AmigaOS 3.9 compatibility may be Amiga specific and not a CPU problem. It also require a 68020 which adds the maximum complexity of any 68k integer CPU. It is a very good test for the Apollo core.

Last edited by matthey; 21 October 2015 at 22:34.
matthey is offline  
Page generated in 0.08656 seconds with 9 queries