uae/docs/README.PROGRAMMERS - annotate

Return to README.PROGRAMMERS CVS log

Up to [UNIX Amiga Emulator] / uae / docs

Annotation of uae/docs/README.PROGRAMMERS, revision 1.1.1.4

1.1 root 1: You can help to make this program better. If you fix bugs or implement new
2: features, I'd be grateful if you send me patches. For a list of interesting
3: projects, and for a brief summary on how UAE works, see below.
4:
5: A few guidelines for anyone who wants to help:
6: - Please contact me first before you implement major new features. Someone
7: else might be doing the same thing already. This has already happened :-(
8: Even if no one else is working on this feature, there might be alternative
9: and better/easier/more elegant ways to do it.
10: - If you have more than one Kickstart, try your code with each one.
11: - Patches are welcome in any form, but diff -u or diff -c output is preferred.
12: If I get whole source files, the first thing I do is to run diff on it. You
13: can save me some work here (and make my mailbox smaller).
14:
15: Some possible projects, in order of estimated difficulty:
1.1.1.2 root 16: - Add gamma correction
1.1.1.3 root 17: - If the serial port still isn't working (I've got no idea, I don't use it),
18: fix it.
1.1 root 19: - Someone with a 68020 data sheet might check whether all opcodes are
20: decoded correctly and whether all instructions really do what they are
21: supposed to do (I'm pretty sure it's OK by now, but you never know...).
1.1.1.3 root 22: - Add more 2.0 packets to filesys.c
1.1.1.2 root 23: - Multi-thread support is there now, it just needs someone to test it on a SMP
24: machine and to fix it so it improves speed instead of slowing the thing
25: down.
1.1 root 26: - Improve the Kickstart replacement to boot more demos.
27: - Snapshots as in CPE. Will need to collect all the variables containing
28: important information. Fairly easy, but boring. (Use core dumps instead :-)
1.1.1.2 root 29: _If_ someone attempts this, please be more clever than the various CPC
30: emulators and dump state only at one fixed point in the frame, preferrably
1.1.1.3 root 31: the vsync point. Also talk with Petter about this.
1.1 root 32: - Find out why uae.device has to be mounted manually with Kick 1.3.
33: The problem seems to be that we don't have a handler for it. I _think_ what
34: we need is the seglist of the standard filesystem handler. Problem is,
35: DOS hasn't been started when the devices are initialized and so we can't get
36: to the DosBase->RootNode->FileHandlerSeg pointer, and then there is the
37: confusing matter of BCPL GlobVecs and other weird stuff...
1.1.1.2 root 38: - Some incompatibilities might be fixed with user-modifiable fudge variables
39: the same way it's done in various C64 emulators.
40: - With the new display code, it would probably be easier than before to
41: implement ECS resolutions - however, a lot of places rely on the OCS timing
42: parameters and display sizes.
1.1 root 43: - Figure out a diskfile format that supports every possible non-standard
44: format.
45: - Implement 68551 MMU. I have docs now. Not among the most necessary things.
1.1.1.3 root 46: Should be done like exception 3 handling: add code to genamode in gencpu.c.
1.1 root 47: - Implement AGA support. Some bits and pieces exist.
48: - Reimplement Amiga OS. (Well-behaved) Amiga programs could then be made
49: to use the X Window System as a "public screen". Of course, not all the
1.1.1.3 root 50: OS would have to be re-done, only Intuition/GFX/Layers (which is enough).
51: [Started, look at gfxlib.c - not usable yet.]
1.1.1.2 root 52: - Find some extremely clever ways to optimize the smart update methods. Some
53: ideas:
54: a) Always use memcmpy() to check for bitplane differences. If no differences
55: are found, see if BPLxDELAY got modified, if so, scroll.
1.1.1.3 root 56: Problems:
1.1.1.2 root 57: * You'd still have to draw a few pixels around the DIW borders. Not very
58: hard.
59: * Scrolling with memcpy in video memory can be terribly slow (no, I
60: shouldn't have bought the cheaper video card with DRAMs)
61: * At least every 15 pixels a full update has to be done since the
62: bitplane pointers get updated after that. And that's with the slowest
63: scrolling - if the playfield scrolls faster, the benefit converges
64: against zero.
65: You could also do vertical scrolling tests, but similar problems arise -
66: where should one check? One line above/below? What about faster
67: scrolling? You could use the bitplane pointers as hints, but with
68: double/triple buffering this gets problematic, too.
69: On the whole, I don't think it would be worth the effort, even if it
70: works very well for a few games.
71: b) Well, there is no b). If I thought of something I forgot it while
72: writing a).
1.1 root 73: - Port it to Java and Emacs Lisp
74: - A formal proof of correctness would be nice.
75:
76:
1.1.1.3 root 77: Source file layout
78:
79: src/ contains (mostly) machine-independent C code.
80: include/ contains header files included by C code.
81: md-*/ CPU and compiler dependent files, linked to machdep by configure
82: od-*/ operating system dependent files, linked to osdep by configure
83: td-*/ thread library dependent files, linked to threaddep by configure
84: sd-*/ Sound code. sd-* is only for sound systems which are not OS specific
85: or for which no "od-*" directory exists. Linked to sounddep
86: targets/ Contains header files which contain some information about which
87: options a specific port of UAE understands.
88:
89:
90: Coding style
91:
92: As long as your code is hidden in a file buried in md-*/ or od-*/ where I
93: never have a look at it, you can probably get away with not following these
94: guidelines.
95:
96: * Do not include CR characters.
97: * Do not use GNU C extensions if you can't hide them in a macro or in a
98: system-specific file so that an alternative implementation is available
99: when GNU C is not used.
100: This applies to _all_ OS/CPU/compiler specific details. Basically, nothing
101: of that sort should appear in src/*.c (we're a bit away from that goal at
102: the moment, but it's getting better).
103: * Make sure your code does not make assumption about type sizes other than
104: the minimum widths allowed by C. If you need specific type sizes, use the
105: uae_u32 type and its friends.
106: * Set up your editor so that tab characters round up to the next position
107: where ((cursorx-1) % 8) == 0, i.e. 8 space tabs. Do not use 4 space tabs,
108: that makes the code awful to read on other machines and worse to edit.
109: * Lines can be up to 132 characters wide. Use SVGATextMode for the Linux
110: console, or use a windowing system in a high resolution.
111: * C++ comments are a no-no in C code.
112: * Indentation - look at some code in custom.c and try to follow it. Don't
113: use GNU 2-space-in-weird-places indentation, I find it awful. But _do_
114: follow the GNU rules for adding whitespace in expressions, and those for
115: breaking up multiple-line if statements.
116: Fixed indentation rules almost never make sense - break the rules if that
117: makes your code more readable.
118: Hint: Get jed from space.mit.edu, /pub/davis. It can indent your code
119: automatically. Put the following into your .jedrc, and it will come out
120: right:
121: C_INDENT = 4;
122: C_BRACE = 0;
123: C_BRA_NEWLINE = 0;
124: C_Colon_Offset = 1;
125: C_CONTINUED_OFFSET = 4;
126:
127:
1.1 root 128: How it works
129:
130: Let's start with the memory emulation. All addressable memory is split into
131: banks of 64K each. Each bank can define custom routines accessing bytes,
132: words, and longwords. All banks that really represent physical memory just
133: define these routines to write/read the specified amount of data to a chunk
1.1.1.3 root 134: of memory. This memory area is organized as an array of uae_u8, which means
1.1 root 135: that those parts of the emulator that want to access memory in a linear
1.1.1.3 root 136: fashion can get a (uae_u8 *) pointer and use it to circumvent the overhead of
1.1.1.2 root 137: the put_*() and get_*() calls. That is done, for example, in the
1.1 root 138: pfield_doline() function which handles screen refreshes.
139: Memory banks that represent hardware registers (such as the custom chip bank
140: at 0xDF0000) can trap reads/writes and take any necessary actions.
141:
142: To provide a good emulation of graphical effects, only one thing is vital:
143: Copper and playfield emulation have to be kept absolutely synchronous. If the
144: copper writes to (say) a color register in a specific cycle, the playfield
1.1.1.2 root 145: hardware needs to use the new information in the next word of data it
1.1 root 146: processes.
147: UAE 0.1 used to call routines like do_pfield() and do_copper() each time the
148: CPU emulator had finished an instruction. That was one of the reasons why it
149: was so slow. Recent versions try to draw complete scanlines in one piece. This
150: is possible if the copper does not write to any registers affecting the
151: display during that scanline. Therefore, drawing the line is deferred until
1.1.1.2 root 152: the last cycle of the line. However, sometimes a register which affects how
153: the screen will look is modified before the end of the line (think of copper
154: plasmas). That's what "struct decision thisline_decision" is for. It is
155: initialized at the start of each line. During the line, whenever a vital
156: register is changed, one of the decide_*() functions is called and may modify
157: thisline_decision. There are several independent decisions:
158: - which DIW should be used
159: - where does data fetch start/stop (or is the line in the border altogether)
160: - where should sprites be drawn (note: the same sprite can appear more than
161: once on one scanline, see Turrican I world 3 levels 1 and 3 for the best
162: example)
163: - what are the playfield pointers at the start of DDF. Related, what data do
164: they point to.
165: - what are the playfield modulos at the end of DDF
166: - coppermagic with the colors is remembered for later use
167: - so is copper magic with the bitplane delay values. I used to think there
168: was no useful application for modifying BPLCON1 while data is being
169: displayed, but Sanity demos can make Amiga emulator programmers look real
170: old.
171:
172: All of this is remembered while the raster line is processed by the hardware.
173: After the line (at hsync), all the decisions are made if they weren't made
174: before. At that point the line can be drawn by playfield_draw_line.
175: Additionally, all the decisions from the previous displayed frame are saved
176: and compared with the new ones, since often lines are not modified between
177: frames. This saves a lot of redrawing work.
1.1 root 178:
179: The CPU emulator no longer has to call all sorts of functions after each
180: instruction. Instead, it keeps a list of events that are scheduled (timer
181: interrupts, hsync and vsync events) and their "arrival time". Only the time
182: for the next event is checked after each CPU instruction. If it's higher than
183: the current cycle counter, the CPU can continue to execute.
184:
1.1.1.2 root 185: Things that can't be supported with the current "decision" model:
186: - Changes in lores/hires mode during one line. Dunno whether that was ever
187: used in reality.
188: - Changes to the bitplane DMA bit during one line. Hardly useful and not
1.1.1.3 root 189: likely to be used. [but there are at least two programs which do ugly
1.1.1.2 root 190: things like that, and there are some hacks in UAE that make those programs
191: work (Magic 12 Ray of Hope 2 is one of these demos)]
192: - Changes in bitplane data during one line. If programs do this kind of
193: thing, it's most likely accidental and the program is broken. Can happen
194: with programs that use the blitter incorrectly, like all the Andromeda
195: demos.
196: - others? (fill in if you can think of anything)
197:
198: All in all, it's unlikely that this causes compatibility problems. If it does,
199: fudge values could be introduced (although that sort of thing gets messy
200: quickly).
1.1 root 201:
202:
1.1.1.3 root 203: * Native code vs. 68k code
1.1.1.2 root 204:
1.1.1.3 root 205: It is possible to call native code from 68k code; autoconf.c has some routines
206: which make setting up a call trap very easy. However, it is not as easy to
207: call 68k code from native C code, at least not while Amiga Exec multitasking
208: is running. You ask why?
209:
210: Amiga process1 calls native function foo
211: Native function foo calls some 68k function and goes into 68k mode
212: Amiga context switch happens, process1 is put to sleep and process2 gets run.
213: Amiga process2 calls native function foo
214: Native function foo calls some 68k function and goes into 68k mode
215: Amiga context switch happens, process2 is put to sleep and process1 gets run.
216: Process 1 completes the 68k function called by foo and returns from 68k mode.
217:
218: There. Now we are in function foo again. When it called the 68k code, process2
219: was active. Now process1 is active, and the function we called in process2
220: hasn't completed yet. What a mess.
221:
222: To get around this, you need to do some stack magic. Code to do this exists,
223: but it must be adapted for each port, since setting up a different stack is
224: completely non-portable.
225:
226:
227: * How multithreading in filesys.c works
228:
229: AmigaOS is nice enough to start one processes for each mounted filesystem. All
230: of these run in the 68k emulation code, i.e. in the main UAE thread. This is
231: the reason why multithreading is desirable: if the main UAE thread blocks
232: waiting for I/O, the CPU emulation can't continue to run. Since the Amiga OS
233: is capable of multi-tasking, it is possible that other code could run until
234: the I/O operation is complete. The most important bit of code that can run is
235: the code that moves the mouse pointer - it's unpleasant if the pointer does
236: not follow mouse movement during disk/CD accesses.
237:
238: When a packet is received by the filesys.asm code, filesys_handler is called.
239: This function always runs in the main UAE thread.
240: - In the single-threaded case, this function performs the action that was
241: requested, then returns 0 to indicate "action completed, reply packet".
242: Nothing else is performed.
243: - In the multi-threaded case, filesys_handler figures out which unit the
244: packet was for and sends the packet to the UAE thread responsible for
245: handling this unit. filesys_handler returns 0 to indicate: queue the
246: packet. Also, one (at that point unused) field in the packet is set to
247: 0 to indicate that the action was not completed.
248:
249: The latter case is the interesting one. The thread that got the packet does
250: the following:
251: - perform the action as usual
252: - set the "command complete" field in the packet to -1.
253: - send a message to the AmigaOS (!) filesystem process. However, it can't do
254: that without some effort. We can't call 68k code from the emulator easily.
255: So we have to use an Amiga interrupt. The filesystem init code sets up an
256: Exec IntServer for the EXTER interrupt, and hsync_handler() checks
257: periodically whether the filesystem needs an interrupt and raises one if
258: necessary.
259: Only one dummy message is used per filesystem unit, which is allocated at
260: startup. This means that there must be some locking to prevent the unit
261: thread from sending the same message twice to the same port. To determine
262: whether the message is free, three counts are kept. "cmds_sent" is
263: incremented by the UAE thread whenever it has completed a command.
264: "cmds_acked" is set to the same value of cmds_sent at the point that the
265: interrupt handler got invoked and decided it must send a message. Finally,
266: cmds_complete is set to this value at the time the AmigaOS process receives
267: the dummy message. Whenever cmds_acked == cmds_complete, the dummy message
268: is free to be sent again.
269:
270: The EXTER interrupt basically walks through the units, looks at the cmds_*
271: fields and sends the dummy message to the Amiga filesystem process when
272: possible and necessary.
273:
274: When the Amiga filesystem process receives such a dummy message, it does the
275: following:
276: - increment cmds_complete as described above.
277: - walk through the queue of unprocessed commands and see which ones now have
278: a status of -1, indicating that they are finished. These are removed from
279: the queue and replied to.
280:
281:
282: * Calltraps at fixed locations
283:
284: F0FF00: return from 68k mode.
285: F0FF10: must have gotten lost somewhere ;)
286: F0FF20: used by filesys.c to store away some information from the startup
287: packet.
288: F0FF30: filesys_handler().
289: F0FF40: startup_handler(), handles only the startup packet for each
290: filesystem.
291: F0FF50: used by the EXTER interrupt which we set up for the filesystem.
292: F0FF60: used by the uaectrl/uae-control programs (see uaelib.c)
293: F0FF70: used by the task that gets set up for the mouse emulation.
1.1.1.2 root 294:
1.1.1.3 root 295: * How the compiler works
1.1.1.2 root 296:
297: .. yet to be written. To be decided, in fact.
1.1.1.3 root 298:
299:
300: Portability
301:
302: This section was out of date. I'll rewrite it.
303: Some day.

unix.superglobalmegacorp.com

This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.