/[gxemul]/trunk/TODO
This is repository of my old source code which isn't updated any more. Go to git.rot13.org for current projects!
ViewVC logotype

Contents of /trunk/TODO

Parent Directory Parent Directory | Revision Log Revision Log


Revision 36 - (show annotations)
Mon Oct 8 16:21:34 2007 UTC (13 years, 1 month ago) by dpavlin
File size: 24543 byte(s)
++ trunk/HISTORY	(local)
$Id: HISTORY,v 1.1497 2007/03/18 03:41:36 debug Exp $
20070224	Minor update to the initialization of the ns16550 in
		machine_walnut.c, to allow that machine type to boot with the
		new interrupt system (although it is still a dummy machine).
		Adding a wdc at 0x14000000 to machine_landisk.c, and fixing
		the SCIF serial interrupts of the SH4 cpu enough to get
		NetBSD/landisk booting from a disk image :-)  Adding a
		preliminary install instruction skeleton to guestoses.html.
20070306	Adding SH-IPL+G PROM emulation, and also passing the "end"
		symbol in r5 on bootup, for Landisk emulation. This is enough
		to get OpenBSD/landisk to install :)  Adding a preliminary
		install instruction skeleton to the documentation. SuperH
		emulation is still shaky, though :-/
20070307	Fixed a strangeness in memory_sh.c (read/write was never
		returned for any page). (Unknown whether this fixes any actual
		problems, though.)
20070308	dev_ram.c fix: invalidate code translations on writes to
		RAM, emulated as separate devices. Linux/dreamcast gets
		further in the boot process than before, but still bugs out
		in userland.
		Fixing bugs in the "stc.l gbr,@-rN" and "ldc.l @rN+,gbr" SuperH 
		instructions (they should NOT check the MD bit), allowing the
		Linux/dreamcast Live CD to reach userland correctly :-)
20070310	Changing the cpu name "Alpha" in src/useremul.c to "21364" to
		unbreak userland syscall emulation of FreeBSD/Alpha binaries.
20070314	Applying a patch from Michael Yaroslavtsev which fixes the
		previous Linux lib64 patch to the configure script.
20070315	Adding a (dummy) sun4v machine type, and SPARC T1 cpu type.
20070316	Creating a new directory, src/disk, and moving diskimage.c
		to it. Separating out bootblock loading stuff from emul.c into
		new files in src/disk.
		Adding some more SPARC registers.
20070318	Preparing/testing for a minirelease, 0.4.4.1.

==============  RELEASE 0.4.4.1  ==============


1 $Id: TODO,v 1.464 2007/03/16 14:13:59 debug Exp $
2
3 -------------------------------------------------------------------------------
4
5 Landisk SH4:
6 HITTA OCH FIXA SH4-STABILITETSPROBLEMEN!
7 SH4 RTC, läs värdmaskinens klocka.
8 När NetBSD 4.0 och OpenBSD 4.1 kommit ut, uppdatera
9 installationsinstruktionerna i dokumentationen, flytta upp till
10 supportat OS (guestos & index).
11 Lägg till landisk i emulated machines i intro.html.
12 Sätt machine_landisk stable = 1, om/när allt funkar.
13
14 Add a special function to the testmachine devices, console perhaps?
15 which prints a zero-terminated string like "[[ %s ]]\n" when given
16 a virtual memory address (or a physical? or perhaps both should be possible).
17 o) NOT to the console, but to the emulator's controlling terminal.
18 DEBUG output, not console output!
19 o) Very useful for debugging. DOCUMENT it!
20
21 Fix the PowerPC DECR interrupt speed! (MacPPC and PReP speed, etc.)
22
23 Prova att installera GNU Linux/DECstation _UTAN_
24 slow_serial_interrupts_hack_for_linux! Det kanske funkar med
25 nya interruptsystemet?
26
27 -------------------------------------------------------------------------------
28
29 Possible (relatively large) work packages to concentrate on in the future:
30
31 x) SMP:
32 Get SMP working again. It is pretty much broken since I started
33 the conversion from the old bintrans system to the new dyntrans system.
34 Add better Test machine demos for SMP in the demos directory.
35 Either multiple dev_mp (one per cpu), or the single dev_mp must
36 _really_ keep all per-cpu values separate.
37
38 x) Alpha and SPARC emulation:
39 These are the next two CPU/machine families that I want to emulate.
40
41 x) Network:
42 Redesign of the networking subsystem, at least the NAT translation
43 part. The current way of allowing raw ethernet frames to be
44 transfered to/from the emulator via UDP should probably be extended
45 to allow the frames to be transmitted other ways as well.
46 Also adding support for connecting ttys (either to xterms, or to
47 pipes/sockets etc, or even to PPP->NAT or SLIP->NAT :-).
48 Documentation updates (!) are very important, making it easier to
49 use the (already existing) network emulation features.
50
51 x) PCI:
52 Pretty much everything related to runtime configuration, device
53 slots, interrupts, whatever. The current code is very hardcoded
54 and ugly.
55 o) Allow cards to be added/removed during runtime more easily.
56 o) Allow cards to be enabled/disabled (i/o ports, etc, like
57 NetBSD needs for disk controller detection).
58 o) Allow devices to be moved in memory during runtime.
59 o) Interrupts per PCI slot, etc. (A-D).
60 o) PCI interrupt controller logic... very hard to get right,
61 because these differ a lot from one machine to the next.
62
63 x) Debugging:
64 Think more about SMP debugging, etc. Right now, the
65 debugger is a mess. Also, a better connection to GDB would be
66 very nice to have.
67
68 x) Userland emulation:
69 Primary goals would be NetBSD and Linux syscall emulation.
70
71 x) Variable-length ISA support:
72 The support for emulation of variable length ISAs (such as
73 M68K, AMD64/i386, or AVR32) is not really there yet. It would
74 be very fun to work on this, to get it right.
75
76 x) Native code generation backends.
77 Not necessary to have the emulator working, but speed is
78 always nice to have.
79
80 x) Remove all "addr mult" device support!
81 Long term goal. Replace the current hacks with address
82 multiplication _per device_ with layered devices, so that the
83 end device (wdc, ns16550, whatever) only supports one access
84 style, and the layers in between convert to 16/32/64-bit
85 word alignment.
86
87 And of course, there are _LOTS_ of minor TODOs spread out throughout
88 the source code, which must be fixed sooner or later.
89
90 -------------------------------------------------------------------------------
91
92 Some other things, in random order, that I'd like to fix: (Some items in
93 this list are probably out-to-date by now.)
94
95 Dyntrans:
96 x) Instruction combination collisions? How to avoid easily...
97 x) Think about how to do both SHmedia and SHcompact in a reasonable
98 way! (Or AMD64 long/protected/real, for that matter.)
99 x) 68K emulation; think about how to do variable instruction
100 lengths across page boundaries.
101 x) Dyntrans with valgrind-inspired memory checker. (In memory_rw,
102 it would be reasonably simple to add; in each individual fast
103 load/store routine = a lot more work, and it would become
104 kludgy very fast.)
105 x) Dyntrans with SMP... lots of work to be done here.
106 x) Dyntrans with cache emulation... lots of work here as well.
107 x) Remove the concept of base RAM completely; it would be more
108 generic to allow RAM devices to be used "anywhere".
109 o) dev_mp doesn't work well with dyntrans yet
110 o) In general, IPIs, CAS, LL/SC etc must be made to work with dyntrans
111 x) Redesign/rethink the delay slot mechanism used for e.g. MIPS,
112 so that it caches a translation (that is, an instruction
113 word and the instr_call it was translated to the last
114 time), so that it doesn't need to do slow
115 to_be_translated for each end of page?
116 x) Program Counter statistics:
117 Per machine? What about SMP? All data to the same file?
118 A debugger command should be possible to use to enable/
119 disable statistics gathering.
120 Configuration file option!
121 x) Breakpoints:
122 o) Physical vs virtual addresses!
123 o) 32-bit vs 64-bit sign extension for MIPS, and others?
124 x) INVALIDATION should cause translations in _all_ cpus to be
125 invalidated, e.g. on a write to a write-protected page
126 (containing code)
127 x) 16-bit encodings? (MIPS16, ARM Thumb, 32-bit SH on SH64)
128 x) Lots of other stuff: see src/cpus/README_DYNTRANS
129 x) true recompilation backend? think carefully about this.
130 o) abstract syntax for emitting opcopdes
131 o) convert into native code only after an entire
132 block has been translated? probably best.
133 o) x86/amd64 code generator can be very similar... perhaps
134 o) branches to already translated code blocks can
135 link the blocks together
136 o) load/store are the most important.
137
138 Simple Valgrind-like checks?
139 o) Mark every address with bits which tell whether or not the address
140 has been written to.
141 o) What should happen when programs are loaded? Text/data, bss (zero
142 filled). But stack space and heap is uninitialized.
143 o) Uninitialized local variables:
144 A load from a place on the stack which has not previously
145 been stored to => warning. Increasing the stack pointer using
146 any available means should reset the memory to uninitialized.
147 o) If calls to malloc() and free() can be intercepted:
148 o) Access to a memory area after free() => warning.
149 o) Memory returned by malloc() is marked as not-initialized.
150 o) Non-passive, but good to have: Change the argument
151 given to malloc, to return a slightly larger memory
152 area, i.e. margin_before + size + margin_after,
153 and return the pointer + margin_before.
154 Any access to the margin_before or _after space results
155 in warnings. (free() must be modified to free the
156 actually allocated address.)
157
158 MIPS:
159 o) Nicer MIPS status bits in register dumps.
160 o) Alignment exceptions.
161 o) Floating point exception correctness.
162 o) Fix this? Triggered by NetBSD/sgimips? Hm:
163 to_be_translated(): TODO: unimplemented instruction:
164 000000000065102c: 00200800 (d) rot_00 at,zr,0
165 o) Some more work on opcodes.
166 x) MIPS64 revision 2.
167 o) Find out which actual CPUs implement the rev2 ISA!
168 o) DROTR32 and similar MIPS64 rev 2 instructions,
169 which have a rotation bit which differs from
170 previous ISAs.
171 o) EI and DI instructions for MIPS64/32 rev 2.
172 NOTE: These are _NOT_ the same as for R5900!
173 x) _MAYBE_ TX79 and R5900 actually differ in their
174 opcodes? Check this carefully!
175 o) Dyntrans: Count register updates are probably not 100% correct yet.
176 o) Refactor code for performance and readability/maintainability.
177 o) (Re)implement 128-bit loads/stores for R5900.
178 o) R4000 and others:
179 x) watchhi/watchlo exceptions, and other exception
180 handling details
181 o) R10000 and others: (R12000, R14000 ?)
182 x) The code before the line
183 /* reg[COP0_PAGEMASK] = cpu->cd.mips.coproc[0]->tlbs[0].mask & PAGEMASK_MASK; */
184 in cpu_mips.c is not correct for R10000 according to
185 Lemote's Godson patches for GXemul. TODO: Go through all
186 register definitions according to http://techpubs.sgi.com/library/tpl/cgi-bin/getdoc.cgi/hdwr/bks/SGI_Developer/books/R10K_UM/sgi_html/t5.Ver.2.0.book_263.html#HEADING334
187 and make sure everything works with R10000.
188 Then test with OpenBSD/sgi?
189 x) Entry LO mask (as above).
190 x) memory space, exceptions, ...
191 x) use cop0 framemask for tlb lookups
192 (http://techpubs.sgi.com/library/tpl/cgi-bin/getdoc.cgi/hdwr/bks/SGI_Developer/books/R10K_UM/sgi_html/t5.Ver.2.0.book_284.html)
193
194 SuperH:
195 x) SH4 interrupt controller:
196 x) Implement correct priorities of interrupts
197 x) SH4 DMA (0xffa00000)
198 x) SH4 UBC (0xff200000)
199 x) Store queues can copy 32 bytes at a time, there's no need to
200 copy individual 32-bit words. (Performance improvement.)
201 x) SH4 BSC (Bus State Controller)
202 x) SH4 RTC: Read the host's clock.
203 x) Instruction tracing should include symbols for branch targets,
204 and so on, to make the output more human readable.
205 x) NetBSD/evbsh3, dreamcast, mmeye, hpcsh! Linux?
206 x) Replace pc-relative loads with immediate load, if within the
207 same page. (Similar to the same optimization for ARM.)
208 x) Floating point speed!
209 x) Floating point exception correctness.
210 x) Think carefully about how to implement SH5/SH64 (for evbsh5).
211
212 Dreamcast:
213 x) G2 DMA
214 x) LAN adapter (dev_mb8696x.c). NetBSD root-on-nfs.
215 x) PVR: Lots of stuff. See dev_pvr.c.
216 x) GDROM
217 x) Modem
218 x) PCI bridge/bus?
219 x) Maple bus:
220 x) Correct controller input
221 x) Mouse input
222 x) Software emulation of BIOS calls:
223 x) GD-ROM emulation: Use the GDROM device.
224 x) Use the VGA font as a fake ROM font. (Better than
225 nothing.)
226 x) Linux/dreamcast? (The gentoo kernel currently crashes.)
227 x) Make as many as possible of the KOS examples run!
228 x) More homebrew demos/games.
229 x) SPU: Sound emulation (ARM cpu).
230 x) VME processor emulation? "(Sanyo LC8670 "Potato")" according to
231 Wikipedia, LC86K87 according to Comstedt's page. See
232 http://www.maushammer.com/vmu.html for a good description of
233 the differences between LC86104C and the one used in the VME.
234
235 Transputer:
236 x) Implement support for Helios binaries.
237 x) Stack and register contents at startup?
238 x) Figure out how to boot an entire Helios distribution.
239 x) Implement all instructions. :)
240
241 RCA1802/RCA1805, CHIP8:
242 x) CHIP8 -> RCA180x conversion
243 x) Think about how to do dual-mode, variable-instr-length
244 ISAs, and switch between modes.
245 x) 1805 "extended" opcode -> trigger CHIP8 emulation?
246 That is, all calls 0NNN could point to 0x68 opcodes,
247 which, if running on a 1802 in CHIP8-emulation-mode,
248 would be manually interpreted.
249 x) Better solution:
250 CHIP8 calls to 00xx => handle at high level,
251 calls to 0xxx in general = call 180X machine code
252 (0000 = reboot?)
253 x) 1802 info: http://www.nyx.net/~lturner/public_html/Cosmac.html
254 and: http://www.elf-emulation.com/1802.html
255 x) 1805 extended opcodes: Implement at least disassembly support!
256 x) Keyboard input.
257 x) Sound (beep only).
258 x) Slow-down to correct speed? Wikipedia: "it was usually operated
259 at 3.58 MHz/2 to suit the requirements of the 1861 chip which
260 gave a speed of a little over 100,000 instructions per second"
261 (Note that _CHIP8_ emulation would then be even slower.)
262 x) SCHIP48 (Super) emulation:
263 Some more opcodes, 128x64 framebuffer, larger
264 sprites and fonts.
265
266 Alpha:
267 x) OSF1 PALcode, Virtual memory support.
268 x) PALcode replacement! PAL1E etc opcodes...?
269 x) Interrupt/exception/trap handling.
270 x) Floating point exception correctness.
271 x) More work on bootup memory and register contents.
272 x) More Alpha machine types, so it could work with
273 OpenBSD, FreeBSD, and Linux too?
274
275 SPARC:
276 o) Implement Adress space identifiers; load/stores etc.
277 o) Save/restore register windows etc! Both v9 and pre-v9!
278 o) Finish the subcc and addcc flag computation code.
279 o) Add more registers (floating point, control regs etc)
280 o) Exception/trap handling.
281 o) Disassemly of some more instructions?
282 o) Are sll etc 32-bit sign-extending or zero-extending?
283 o) Finish the GDB register stuff.
284 x) Floating point exception correctness.
285 o) SPARC v8, v7 etc?
286
287 Debugger:
288 o) How does SMP debugging work? Does it simply use "threads"?
289 What if the guest OS (running on an emulated SMP machine)
290 has a usertask running, with userland threads?
291 o) Try to make the debugger more modular and, if possible, reentrant!
292 o) Remove the emul command? (But show network info if showing
293 machines?)
294 o) Evaluate expressions within []? That would allow stuff like
295 cpu[x] where x is an expression.
296 o) Settings:
297 x) Special handlers for Write!
298 +) MIPS coproc regs
299 +) Alpha/MIPS/SPARC zero registers
300 +) x86 64/32/16-bit registers
301 x) Value formatter for resulting output.
302 o) see src/debugger.c for more
303
304 POWER/PowerPC:
305 x) Fix DECR timer speed, so it matches the host.
306 x) NetBSD/prep 3.x triggers a possible bug in the emulator:
307 <wdc_exec_command(0xd005e514,0xd60cdd30,0,8,..)>
308 <ata_get_xfer(0,0xd60cdd30,0,8,..)>
309 <0x26c550(&ata_xfer_pool,2,0,8,..)>
310 <0x35c71c(0x3f27000,0,52,8,..)>
311 <ata_exec_xfer(0xd005e4c8,0x3f27000,0,13,..)>
312 <atastart(0xd005e4c8,0x3f27000,0,13,..)>
313 <__wdccommand_start(0xd005e4c8,0x3f27000,0,13,..)>
314 <bsw1(&prep_isa_io_space_tag,0x800001f6,0,176,..)>
315 [ wdc: write to SDH: 0xb0 (sectorsize 2, lba=1, drive 1, head 0) ]
316 <wdcwait(0xd005e4c8,72,64,0xbb8,..)>
317 <0x198120(0xd005e4c8,72,64,0xbb8,..)>
318 <bsr1(&prep_isa_io_space_tag,0,0,0xbb8,..)>
319 <delay(100,0,0,0xbb8,..)>
320 Note: <bsr1(&prep_isa_io_space_tag,0,0,0xbb8,..)>
321 x) PPC optimizations; instr combs
322 x) 64-bit stuff: either Linux on G5, or perhaps some hobbyist
323 version of AIX? (if there exists such a thing)
324 x) macppc: adb controller; keyboard (for framebuffer mode)
325 x) make OpenBSD/macppc work (PCI controller stuff)
326 x) Floating point exception correctness.
327 x) Alignment exceptions.
328
329 PReP:
330 Clock time! ("Bad battery blah blah")
331
332 Algor:
333 o) Other models than the P5064?
334 o) PCI interrupts... needed for stuff like the tlp NIC?
335
336 BeBox:
337 o) Interrupts. There seems to be a problem with WDC interrupts
338 "after a short while", although a few interrupts get through?
339 o) Perhaps find a copy of BeOS and try it?
340
341 HPCmips:
342 x) Mouse/pad support! :)
343 x) A NIC? (As a PCMCIA device?)
344
345 AVR:
346 o) Everything.
347
348 ARM:
349 o) See netwinder_reset() in NetBSD; the current "an internal error
350 occured" message after reboot/halt is too ugly.
351 o) ARM "wait"-like instruction?
352 o) try to get netbsd/evbarm 3.x or 4.x running (iq80321)
353 o) make the xscale counter registers (ccnt) work
354 o) make the ata controller usable for FreeBSD!
355 o) Zaurus emulation:
356 x) OpenBSD/zaurus
357 x) NetBSD/zaurus? See the following URL:
358 http://mail-index.netbsd.org/port-arm/2006/11/19/0000.html
359 o) Debian/cats crashes because of unimplemented coproc stuff.
360 fix this?
361
362 Test machines:
363 o) dev_fb block fill and copy
364 o) dev_fb draw characters (from the built-in font)?
365 o) dev_fb input device? mouse pointer coordinates and buttons
366 (allow changes in these to cause interrupts as well?)
367 o) Redefine the halt() function so that it stops "sometimes
368 soon", i.e. usage in demo code should be:
369 for (;;) {
370 halt();
371 }
372
373 Better CD Image file support:
374 x) Support CD formats that contain more than 1 track, e.g.
375 CDI files (?). These can then contain a mixture of e.g. sound
376 and data tracks, and booting from an ISO filesystem path
377 would boot from [by default] the first data track.
378 (This would make sense for e.g. Dreamcast CD images, or
379 possibly other live-CD formats.)
380
381 Networking:
382 x) Fix performance problems caused by only allowing a
383 single TCP packet to be unacked.
384 x) Don't hardcode offsets into packets!
385 x) Test with lower than 100 max tcp/udp connections,
386 to make sure that reuse works!
387 x) Make OpenBSD work better as a guest OS!
388 x) DHCP? Debian doesn't actually send DHCP packets, even
389 though it claims to? So it is hard to test.
390 x) Multiple networks per emulation, and let different
391 NICs in machines connect to different networks.
392 x) Support VDE (vde.sf.net)? Easiest/cleanest (before a
393 redesign of the network framework has been done) is
394 probably to connect it using the current (udp) solution.
395 x) Allow SLIP connections, possibly PPP, in addition to
396 ethernet?
397
398 Cache simulation:
399 o) Command line flags for:
400 o) CPU endianness?
401 o) Cache sizes? (multiple levels)
402 o) Separate from the CPU concept, so that multi-core CPUs sharing
403 e.g. a L2 cache can be simulated (?)
404 o) Instruction cache emulation is easiest (if separate from the
405 data cache); similar hack as the S;I; hack in cpu_dyntrans.c.
406 NOTE: if the architecture has a delay slot, then an instruction
407 slot can actually be executed as 2 instructions.
408 o) Data cache emulation = harder; each arch's load/store routines
409 must include support? running one instruction at a time and
410 having a cpu-dependant lookup function for each instruction
411 is another option (easier to implement, but very very slow).
412
413 Documentation:
414 x) Note about sandboxing/security:
415 Not all emulated instructions fail in the way they would
416 do on real hardware (e.g. a userspace program writing to
417 a system register might work in GXemul, but it would
418 fail on real hardware). Sandbox = contain from the
419 host OS. But the emulated programs will run "less
420 securely".
421 x) Try NetBSD/arc 4.x! (It seems to work with disk images!)
422 x) NetBSD/pmax 4 install instructions: xterm instead of vt100!
423 x) BETTER DEVICE EXAMPLES!
424 o) Move away from technical.html to somewhere new.
425 o) DEVICE_TICK
426 o) Implement example devices using interrupts, dyntrans
427 memory access, etc.?
428 x) Document the dyntrans core?
429 x) Rewrite the section about experimental devices, after the
430 framebuffer acceleration has been implemented, and demos
431 written. (Symbolic names instead of numbers; example
432 use cases, etc. Mention demo files that use the various
433 features?)
434 x) "a very simple linear framebuffer device (for graphics output)"
435 under "which machines does gxemul emulate" ==> better
436 description?
437 x) Better description on how to set up a cross compiler?
438 Example for MIPS64.
439 o) Automagic documentation generation?
440 x) machines, cpus, devices.
441 x) REMEMBER that several machines/devices can be in
442 the same source file!
443 o) Try to rewrite the install instructions for those machines
444 that use 3MAX into using CATS or hpcmips? (To remove the need
445 to use a raw ffs partition, using up all of the disk image.)
446
447 More generic out_of_memory error reporting, and check everywhere!
448 Causes: OpenBSD has low default limits for normal users.
449 Host is 32-bit? (32-bit hosts are limited to 4 GB or less
450 of userspace memory.)
451 You are actually low on RAM. (As trivial as this might sound,
452 Unix systems usually allow processes to allocate virtual
453 memory beyond the amount of RAM in the machine.)
454
455 The Device subsystem:
456 x) allow devices to be moved and/or changed in size (down to a
457 minimum size, etc, or up to a max size); if there is a collision,
458 return false. It is up to the caller to handle this situation!
459 x) NOTE: Translations must be invalidated, both for
460 registering new devices, and for moving existing ones.
461 cpu->invalidate translation caches, for all CPUs that
462 are connected to a specific memory.
463 x) keep track of interrupts and busses? actually, allowing any device
464 to be a bus might be a nice idea.
465 x) turn interrupt controllers into devices? :-)
466 x) refactor various clocks/nvram/cmos into one device?
467
468 PCI:
469 x) last write was ffffffff ==> fix this, it should be used
470 together with a mask to get the correct bits. also, not ALL
471 bits are size bits! (lowest 4 vs lowest 2?)
472 x) add support for address fixups
473 x) generalize the interrupt routing stuff (lines etc)
474
475 Clocks and timers:
476 x) DON'T HARDCODE 100 HZ IN cpu_mips_coproc.c!
477 x) Test the 8253? Right now it doesn't seem to be used?
478 x) NetWinder timeofday is incorrect!
479 x) Cobalt TOD is incorrect!
480 x) Go through all other machines, one by one, and fix them.
481
482 Config file parser:
483 o) Rewrite it from scratch!
484 o) Usage of any expression available through the debugger
485 o) Allow interrupt controllers to be added! and interrupts
486 to be used in more ways than before
487 o) Support for running debugger commands (like the -c
488 command line option)
489
490 Floating point layer:
491 o) make it common enough to be used by _all_ emulation modes
492 o) implement correct error/exception handling and rounding modes
493 o) implement more helper functions (i.e. add, sub, mul...)
494 o) non-IEEE modes (i.e. x86)?
495
496 Userland emulation:
497 x) Lots of stuff; freebsd and netbsd (and linux?) syscalls.
498 x) Dynamic linking? Hm.
499
500 Sound:
501 x) generic sound framework
502 x) add one or more sound cards as devices; add a testmachine
503 sound card first?
504 x) Dreamcast sound? Generic PCI sound cards?
505
506 ASC SCSI controller:
507 x) NetBSD/arc 2.0 uses the ASC controller in a way which GXemul
508 cannot yet handle. (NetBSD 1.6.2 works ok.) (Possibly a problem
509 in NetBSD itself, http://mail-index.netbsd.org/source-changes/
510 2005/11/06/0024.html suggests that.)
511 NetBSD 4.x seems to work? :)
512
513 Caches / memory hierarchies: (this is mostly MIPS-specific)
514 o) src/memory*.c: Implement correct cache emulation for
515 all CPU types. (currently only R2000/R3000 is implemented)
516 (per CPU, multiple levels should be possible, associativity etc!)
517 o) R2000/R3000 isn't _100%_ correct, just almost correct :)
518 o) Move the -S (fill mem with random) functionality into the
519 memory.c subsystem, not machine.c or wherever it is now
520 o) ECC stuff, simulation of memory errors? (Machine dependent)
521 o) More than 4GB of emulated RAM, when run on a 32-bit host?
522 (using manual swap-out of blocks to disk, ugly)
523 o) A global command line option should be used to turn
524 cache emulation on or off. When off, caches should be
525 faked like they are right now. When on, caches and
526 memory latencies should be emulated as correctly as
527 possible.
528
529 File/disk/symbol handling:
530 o) Make sure that disks can be added/removed during runtime!
531 (Perhaps this needs a reasonably large re-write.)
532 o) Remove some of the complexity in file format guessing, for
533 Ultrix kernels that are actually disk images?
534 o) Better handling of tape files
535 o) Read function argument count and types from binaries? (ELF?)
536 o) Better demangling of C++ names. Note: GNU's C++ differs from e.g.
537 Microsoft's C++, so multiple schemes must be possible. See
538 URL at top of src/symbol_demangle.c for more info.
539
540 Userland ABI emulation:
541 o) see src/useremul.c
542
543 Better framebuffer and X-windows functionality:
544 o) Generalize the update_x1y1x2y2 stuff to an extend-region()
545 function...
546 o) -Yx sometimes causes crashes.
547 o) Simple device access to framebuffer_blockcopyfill() etc,
548 and text output (using the built-in fonts), for dev_fb.
549 o) CLEAN UP the ugly event code
550 o) Mouse clicks can be "missed" in the current system; this is
551 not good. They should be put on a stack of some kind.
552 o) More 2D and 3D framebuffer acceleration.
553 o) Non-resizable windows? Or choose scaledown depending
554 on size (and center the image, with a black border).
555 o) Different scaledown on different windows?
556 o) Non-integral scale-up? (E.g. 640x480 -> 1024x768)
557 o) Switch scaledown during runtime? (Ala CTRL-ALT-plus/minus)
558 o) Bug reported by Elijah Rutschman on MacOS with weird
559 keys (F5 = cursor down?).
560 o) Keyboard and mouse events:
561 x) Do this for more machines than just DECstation
562 x) more X11 cursor keycodes
563 x) Keys like CTRL, ALT, SHIFT do not get through
564 by themselves (these are necessary for example
565 to change the font of an xterm in X in the
566 emulator)
567 o) Generalize the framebuffer stuff by moving _ALL_ X11
568 specific code to src/x11.c!
569

  ViewVC Help
Powered by ViewVC 1.1.26