A PPC native OS ?

[Karlos]

That's becaouse of

1) They had an absolutely awful interpretive 68K emulation that was black box based - all 68K code flowed throug the same instance of the interpreter

2) Within that emulation they had to factor in older 68000 code that used 24-bit addressing (ewww)

3) Their pre OS X systems wasn't exactly quick anyway. They realised how poor it was finally and jumped to bsd :-)

4) Large portions of the system code used fat binaries such that 68K stuff could run the same files as PPC. An upshot of this was that some of the critical emulated code was relying on emulated system calls too

This is much less of a problem for AmigaOS 4 because

1) It uses a JIT 68K emulation rather than interpretation

2) The emulation is thread based, so each 68K process has its own 'emulation' running

3) Much has been learned in recent years of how to achieve effective emulation

4) The OS is efficient

5) The starting PPC spec is quite a lot higher :-D

-edit-

According to the OS4 technical brief, the 68K emulation can outperform a 68060 in some cases even on a 603e 160MHz...

I guess we'll have to wait and see...

[Karlos]

Hmm, starting to get a funny reputation round here!

Nobody ever accused me of being efficient before! Many things, but not efficient :lol:

[Karlos]

xaccrocheur wrote:

I found no relevant documentation about this particular project (MacOS 2 PPC) and again less about AOS. An interesting data, even considering the very relevant differences you pointed in your post, would be the number of man-years needed, compared to the 25 Alan told us about AOS, witch is, correct me if I'm wrong not out yet ;-)

Where the heck can I read more about that ? Closed-source sometimes really sux

pX

Well, I actually registered as a developer for mac ages ago when I was first experimenting with cross platform coding (back when 68K macs were still the most popular user base but PPC was out). You got some insight from the developer info...

As for the OS4 stuff I mentioned, that's straight out of Hyperions technical brief they released in pdf form a few weeks/month back.

In my estimation, the biggest drag was rewriting exec and all of the ROM based resources, which were coded in asm and have BCPL pointer stuff kicking around for good measure. Not nice.

Porting the asm to C and testing, whilst not conceptually difficult, especially for coders with the experience that Hyperion's have, is still a big time drain.

Next up would be the new kernel features and integration into the existing architecture. Especially the JIT emulator.

The source for the x86 based JIT from WinUAE is available but obviously you need to re code the back end to emit PPC opcodes.

Also, given the GPL nature of that code, I guess Hyperion had to write their own from scratch.

As ans aside, I recall the WarpOS authors had a 68K emulation for WarpOS 5, but I never heard anything about it after their initial claimes that they got an emulated workbecnch 3.5 to boot up...

[Karlos]

Helgis75 wrote:
Be happy! Sing a song! Here we go.. :-)  :-)  :-)  :-D  8-)

If amiga.org ever starts an annual award ceremony, I'd like to nominate Helgis for his undying faith in AmigaOne/OS4... :-)