home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Linux Cubed Series 7: Sunsite
/
Linux Cubed Series 7 - Sunsite Vol 1.iso
/
system
/
emulator
/
uae-0.000
/
uae-0
/
uae-0.6.0
/
README.PROGRAMMERS
< prev
next >
Wrap
Text File
|
1996-06-12
|
9KB
|
156 lines
You can help to make this program better. If you fix bugs or implement new
features, I'd be grateful if you send me patches. For a list of interesting
projects, and for a brief summary on how UAE works, see below.
A few guidelines for anyone who wants to help:
- Please contact me first before you implement major new features. Someone
else might be doing the same thing already. This has already happened :-(
Even if no one else is working on this feature, there might be alternative
and better/easier/more elegant ways to do it.
- Some coding guidelines.
* Avoid GNU C extensions by all means. They make your code non-portable.
* Avoid GNU indentation style by all means. It makes your code unreadable.
* Try to indent your code nicely. There are editors like JED, which is
available from space.mit.edu:/pub/davis and highly recommendable, that do
this for you automatically.
* Use eight space tabs, four space tabs make a mess of the code.
- If you have access to more than one Unix system, try compiling/running your
code on all of these. Remember, UAE is supposed to run on the DEC Alpha: so
don't assume sizeof(char*) == sizeof(int) == sizeof(long)
- If you have more than one Kickstart, try your code with each one.
- Patches are welcome in any form, but diff -u or diff -c output is preferred.
If I get whole source files, the first thing I do is to run diff on it. You
can save me some work here (and make my mailbox smaller).
Some possible projects, in order of estimated difficulty:
- Someone with a 68000 data sheet might check whether all opcodes are
decoded correctly and whether all instructions really do what they are
supposed to do (I'm pretty sure it's OK by now, but you never know...).
- Modify UAE to run in two threads for systems that have more than one CPU.
I've thought of a way how to do it, but I don't have a multiprocessor board
to try this. Here's how it works: Start with smart update method #1. This
method nicely bundles all the data necessary for drawing a scanline into two
structures: linedescr and line_data. Instead of actually drawing the line
in the main task, signal the second one to use that information. Synchronize
the two at the end of a frame to avoid lossage.
If I haven't overlooked a big problem, it should be trivial.
Sound output could also be moved into a separate task, but the gain will be
much smaller.
- Implement all 68020 instructions/addressing modes.
- Improve the Kickstart replacement to boot more demos.
- Write an Amiga program that communicates with UAE and provides a user-
interface similar to the X Windows one, but from within the emulation.
- Calculating the flags after each instruction is time consuming and usually
completely unnecessary. Modify gencpu to generate instructions that don't
set the flags, and have some code in the central loop to decide which
set of instructions to use. This means additional overhead, of course,
which might eat any improvements. [Tried it, hardly a difference, but
maybe I did it in a stupid way]
- Translate Kickstart ROM (or any software) to C code with a slightly
modified gencpu, and link it to the rest of the emulator. Definitely
possible, maybe illegal. Better don't do that for now.
- Make a modified gencpu that can generate x86 assembly instead of C. This
might give a nice speedup. However, the CPU emulation already is the
fastest part in UAE - screen updates are much more time consuming.
[Partly done, some __asm__ magic for flag calculations. Bad for performance
are all the unnecessary register push/pops that the compiler generates.]
- Improve sound emulation (Find bugs).
- Snapshots as in CPE. Will need to collect all the variables containing
important information. Fairly easy, but boring. (Use core dumps instead :-)
- Find out why uae.device doesn't work with Kick 1.3. Not really important,
no one in his right mind uses it anyway now there is the unixfs.device
- Make unixfs.device bootable.
- The playfield hardware is the only important part of the Amiga that is
not really well documented in the HRM. Write some test programs that
do all sorts of weird things with the copper (like turning off bitplane
DMA during a line, and turning it directly on again: VERY interesting
result) and try to emulate this perfectly.
- Figure out a diskfile format that supports every possible non-standard
format.
- Translate basic blocks of m68k instructions to intermediate code that is
interpreted instead of the actual code. Optimize it of course.
(Not sure whether it would really be faster. But ARDI's Executor does
a splendid job on translation. This should really help even if the target
code isn't native).
- Implement 68551 MMU. I have docs now. Not among the most necessary things.
- Implement 68882 FPU. Not among the most necessary items, either. (Some docs)
- Implement AGA support. Maybe some easy parts can be tried first (like 8
bitplane support). I have sufficient documentation by now.
- Reimplement Amiga OS. (Well-behaved) Amiga programs could then be made
to use the X Window System as a "public screen". Of course, not all the
OS would have to be re-done, only Intuition/GFX. [Started, look at gfxlib.c]
- Translate instructions on the fly to native code, like Executor does.
I'll NEVER try that one myself.
- Find some extremely clever ways to optimize the smart update methods. Maybe
try to support scrolling. Maybe find a way to update only parts of a line
(similar to LOW_BANDWIDTH). All such methods would probably be terribly
complicated, and not easy to get right without sacrificing compatibility.
Maybe it's not such a good idea.
- Port it to Java
- A formal proof of correctness would be nice.
Things I don't really want to do:
- Full ECS support, with braindamage like Productivity and SuperHires.
How it works
Let's start with the memory emulation. All addressable memory is split into
banks of 64K each. Each bank can define custom routines accessing bytes,
words, and longwords. All banks that really represent physical memory just
define these routines to write/read the specified amount of data to a chunk
of memory. This memory area is organized as an array of WORDs, which means
that those parts of the emulator that want to access memory in a linear
fashion can get a (WORD *) pointer and use it to circumvent the overhead of
the put_word() and get_word() calls. That is done, for example, in the
pfield_doline() function which handles screen refreshes.
Memory banks that represent hardware registers (such as the custom chip bank
at 0xDF0000) can trap reads/writes and take any necessary actions.
In some places, this scheme is abused: The uae.device and unixfs.device are
stored in a segment at 0xF00000 containing a ROMtag structure, so it is
recognized at bootup. Since this is a ROM area, writes shouldn't occur
normally and are therefore used to trap into emulation routines for these
devices.
To provide a good emulation of graphical effects, only one thing is vital:
Copper and playfield emulation have to be kept absolutely synchronous. If the
copper writes to (say) a color register in a specific cycle, the playfield
hardware needs to use the new information in the next word of data it
processes.
UAE 0.1 used to call routines like do_pfield() and do_copper() each time the
CPU emulator had finished an instruction. That was one of the reasons why it
was so slow. Recent versions try to draw complete scanlines in one piece. This
is possible if the copper does not write to any registers affecting the
display during that scanline. Therefore, drawing the line is deferred until
the last cycle of the line. If the copper writes to a hardware register before
that, the function pfield_may_need_update() is called and this one determines
whether it should fall back to the cycle-for-cycle approach. This is very
rarely needed, mainly for copper-plasma effects and such, and the general case
is much faster.
The CPU emulator no longer has to call all sorts of functions after each
instruction. Instead, it keeps a list of events that are scheduled (timer
interrupts, hsync and vsync events) and their "arrival time". Only the time
for the next event is checked after each CPU instruction. If it's higher than
the current cycle counter, the CPU can continue to execute.
Portability
The main thing you need to worry about when porting UAE to a new platform is
the OS dependent source file handling all the graphics output. Currently,
there are xwin.c, svga.c, mac.c, dos-grx.c, dos-null.c, bebox.cpp and
NeXTwin.m. Rewriting one of these to use the features of your operating
system should be fairly easy.
You might need to worry a little about datatypes. UAE requires that your C
compiler supports 8 bit, 16 bit and 32 bit integers, otherwise a MC68000
emulation would be not very easy to get right. Some typedefs to hide the
actual types used can be found in amiga.h. The CPU emulation is not the only
place that makes some potentially non-portable assumptions: The graphics
code in custom.c, mainly the pfield_doline() function and its friends, may
need some work if you have a really weird architecture.
The only thing that's left are some Unixoid assumptions, mainly in
filesys.c, but also in debug.c. Put in a few #ifdefs and modify the Makefile
if necessary/possible, or make up a dummy unixfs-null.c file that contains
stubs.
Apart from all that, it's fairly portable...