View Single Post
Old 24 February 2017, 21:13   #10
matthey
Banned
 
Join Date: Jan 2010
Location: Kansas
Posts: 1,284
Quote:
Originally Posted by meynaf View Post
Possible solution : rewrite it completely.
This is the AROS way but it is slower than AOS3.x on same hardware. A rewrite in asm, on the contrary, would be faster than the original.
Program after program, rom library after rom library, it is time consuming and a daunting task, but nevertheless doable.
I have considered this also. It makes more sense for the low level stuff and drivers. It would probably be possible for the OS to be at least 20% smaller while primarily optimizing for performance. PeterK's icon.library has accomplished much more but it is has taken considerable concentration. There are much bigger high level libraries which would take much longer to recreate in assembler. It would make sense to do them in C. Heiroglyph had a good idea of allowing every C function to have an optional assembler function which could be used with a switch for easier debugging. I just wish there was a better lightweight Amiga 68k C compiler that did a reasonable job of 68k code output. VBCC had such promise but it seems to have stopped improving and perhaps even declined in the 68k backend. If you were going to pay someone to improve 68k code, paying for improved compiler backends may be more productive. Then again funding good mass produced cheap 68k hardware could spur interest and allow it to happen without funding compilers. People don't want to waste their time developing for something dead. Funding new software development or new hardware gives the chicken and the egg problem with which should be funded first. Certainly funding software and development for a few thousand people has failed (look at PPC efforts).

Quote:
Originally Posted by meynaf View Post
Now who can and is willing to do it, is another story.
Well. Give me enough money to pay my bills for several years and i'll do it
You, Don Adan, PeterK, me and maybe Cosmos (unpredictable) could probably cooperate and share code on a new 68k assembler AmiOS, AmigOS, AOS or whatever we would want to call it. Perhaps AROS could be used for the high level stuff with some optimizations. It would be helpful to use similar syntax and formatting which we would have to decide. Target would probably be 68020-68060. Decisions could be made by vote. Probably not much motivation without new mass produced hardware or cash though.
matthey is offline  
 
Page generated in 0.05133 seconds with 9 queries