/[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 44 - (show annotations)
Mon Oct 8 16:22:56 2007 UTC (16 years, 5 months ago) by dpavlin
File size: 24172 byte(s)
++ trunk/HISTORY	(local)
$Id: HISTORY,v 1.1632 2007/09/11 21:46:35 debug Exp $
20070616	Implementing the MIPS32/64 revision 2 "ror" instruction.
20070617	Adding a struct for each physpage which keeps track of which
		ranges within that page (base offset, length) that are
		continuously translatable. When running with native code
		generation enabled (-b), a range is added after each read-
		ahead loop.
		Experimenting with using the physical program counter sample
		data (implemented 20070608) together with the "translatable
		range" information, to figure out which physical address ranges
		would be worth translating to native code (if the number of
		samples falling within a range is above a certain threshold).
20070618	Adding automagic building of .index comment files for
		src/file/, src/promemul/, src src/useremul/ as well.
		Adding a "has been translated" bit to the ranges, so that only
		not-yet-translated ranges will be sampled.
20070619	Moving src/cpu.c and src/memory_rw.c into src/cpus/,
		src/device.c into src/devices/, and src/machine.c into
		src/machines/.
		Creating a skeleton cc/ld native backend module; beginning on
		the function which will detect cc command line, etc.
20070620	Continuing on the native code generation infrastructure.
20070621	Moving src/x11.c and src/console.c into a new src/console/
		subdir (for everything that is console or framebuffer related).
		Moving src/symbol*.c into a new src/symbol/, which should
		contain anything that is symbol handling related.
20070624	Making the program counter sampling threshold a "settings
		variable" (sampling_threshold), i.e. it can now be changed
		during runtime.
		Switching the RELEASE notes format from plain text to HTML.
		If the TMPDIR environment variable is set, it is used instead
		of "/tmp" for temporary files.
		Continuing on the cc/ld backend: simple .c code is generated,
		the compiler and linker are called, etc.
		Adding detection of host architecture to the configure script
		(again), and adding icache invalidation support (only
		implemented for Alpha hosts so far).
20070625	Simplifying the program counter sampling mechanism.
20070626	Removing the cc/ld native code generation stuff, program
		counter sampling, etc; it would not have worked well in the
		general case.
20070627	Removing everything related to native code generation.
20070629	Removing the (practically unusable) support for multiple
		emulations. (The single emulation allowed now still supports
		multiple simultaneous machines, as before.)
		Beginning on PCCTWO and M88K interrupts.
20070723	Adding a dummy skeleton for emulation of M32R processors.
20070901	Fixing a warning found by "gcc version 4.3.0 20070817
		(experimental)" on amd64.
20070905	Removing some more traces of the old "multiple emulations"
		code.
		Also looking in /usr/local/include and /usr/local/lib for
		X11 libs, when running configure.
20070909	Minor updates to the guest OS install instructions, in
		preparation for the NetBSD 4.0 release.
20070918	More testing of NetBSD 4.0 RC1.

1 $Id: TODO,v 1.556 2007/09/11 21:46:35 debug Exp $
2
3 Some things, in no specific order, that I'd like to fix:
4 (Some items in this list are perhaps already fixed.)
5
6 -------------------------------------------------------------------------------
7
8 A first pass of installation regression testing of NetBSD 4.0 RC1 in GXemul:
9
10 X pmax (including X Windows out-of-the-box)
11 X arc (1.6.2 -> 4.0! Yay!)
12 hpcmips
13 cobalt
14 evbmips
15 algor
16 sgimips
17 cats
18 evbarm
19 netwinder
20 prep nej, 2.0 är senaste som funkar :(
21 X macppc
22 pmppc
23 X dreamcast
24 X landisk
25
26 X = done and worked fine
27
28 -------------------------------------------------------------------------------
29
30 Perform a second regression test pass, when the actual NetBSD 4.0 release
31 has happened.
32
33 o) Test all guest OSes.
34 o) Update:
35 URLs
36 Versions
37
38 o) Make a new GXemul release: 0.4.6.1
39
40 -------------------------------------------------------------------------------
41
42 M88K:
43 o) FIP != NIP + 4, in rte! (Simulate delayed branch stuff.)
44 o) cpu_dyntrans.c: MEMORY_USER_ACCESS implementation for M88K!
45 o) xmem: Set transaction registers!
46 o) CMMUs:
47 o) Translation invalidations, could be optimized.
48 o) Move initialization from dev_mvme187 to somewhere
49 more reasonable?
50 o) Instruction trace by using bits of ??IP control regs.
51 o) Interrupts (these are machine dependent, though).
52 o) Implement devices etc. for one or more machine modes,
53 to get some guest OS running. OpenBSD/mvme88k on MVME187
54 seems to be the smartest path to follow for now.
55 o) VME bus device
56 o) PCC2
57 o) Cirrus Logic serial port controller
58 o) Instruction disassembly, and implementation:
59 o) See http://www.panggih.staff.ugm.ac.id/download/GCC/info/gcc.i5
60 for some strange cases of when "div" can fail (?)
61 o) Floating point stuff
62 o) "Graphics" instructions (M88110-specific)
63
64 MIPS:
65 o) Nicer MIPS status bits in register dumps.
66 o) Floating point exception correctness.
67 o) Fix this? Triggered by NetBSD/sgimips? Hm:
68 to_be_translated(): TODO: unimplemented instruction:
69 000000000065102c: 00200800 (d) rot_00 at,zr,0
70 o) Some more work on opcodes.
71 x) MIPS64 revision 2.
72 o) Find out which actual CPUs implement the rev2 ISA!
73 o) DINS, DINSM, DINSU etc
74 o) DROTR32 and similar MIPS64 rev 2 instructions,
75 which have a rotation bit which differs from
76 previous ISAs.
77 x) _MAYBE_ TX79 and R5900 actually differ in their
78 opcodes? Check this carefully!
79 o) Dyntrans: Count register updates are probably not 100% correct yet.
80 o) (Re)implement 128-bit loads/stores for R5900.
81 o) Coprocessor 1x (i.e. 3) should cause cp1 exceptions, not 3?
82 (See http://lists.gnu.org/archive/html/qemu-devel/2007-05/msg00005.html)
83 o) R4000 and others:
84 x) watchhi/watchlo exceptions, and other exception
85 handling details
86 o) MIPS 5K* have 42 physical address bits, not 40/44?
87 o) R10000 and others: (R12000, R14000 ?)
88 x) The code before the line
89 /* reg[COP0_PAGEMASK] = cpu->cd.mips.coproc[0]->tlbs[0].mask & PAGEMASK_MASK; */
90 in cpu_mips.c is not correct for R10000 according to
91 Lemote's Godson patches for GXemul. TODO: Go through all
92 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
93 and make sure everything works with R10000.
94 Then test with OpenBSD/sgi?
95 x) Entry LO mask (as above).
96 x) memory space, exceptions, ...
97 x) use cop0 framemask for tlb lookups
98 (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)
99
100 SuperH:
101 x) Auto-generation of loads/stores! This should get rid of at least
102 the endianness check in each load/store.
103 x) Experiment with whether or not correct ITLB emulation is
104 actually needed. (20070522: I'm turning it off today.)
105 x) SH4 interrupt controller:
106 x) MASKING should be possible!
107 x) SH4 DMA (0xffa00000)
108 x) SH4 UBC (0xff200000)
109 x) Store queues can copy 32 bytes at a time, there's no need to
110 copy individual 32-bit words. (Performance improvement.)
111 x) SH4 BSC (Bus State Controller)
112 x) Instruction tracing should include symbols for branch targets,
113 and so on, to make the output more human readable.
114 x) SH3-specific devices: Pretty much everything!
115 x) NetBSD/evbsh3, hpcsh! Linux?
116 x) Floating point speed!
117 x) Floating point exception correctness.
118 x) Exceptions for unaligned load/stores. OpenBSD/landisk uses
119 this mechanism for its reboot code (machine_reset).
120
121 Landisk SH4:
122 x) When NetBSD/landisk 4.0 has been released, make sure it works
123 in the emulator. (Update documentation, etc.)
124 NetBSD HEAD (as of April 2007) hangs during bootup, because it
125 turns on/off interrupts in an unfortunately synchronized way
126 with dyntrans. This needs to be fixed.
127
128 Dreamcast:
129 x) G2 DMA
130 x) LAN adapter (dev_mb8696x.c). NetBSD root-on-nfs.
131 x) PVR: Lots of stuff. See dev_pvr.c.
132 x) Better GDROM support
133 x) Modem
134 x) PCI bridge/bus?
135 x) Maple bus:
136 x) Correct controller input
137 x) Mouse input
138 x) Software emulation of BIOS calls:
139 x) GD-ROM emulation: Use the GDROM device.
140 x) Use the VGA font as a fake ROM font. (Better than
141 nothing.)
142 x) Make as many as possible of the KOS examples run!
143 x) More homebrew demos/games.
144 x) SPU: Sound emulation (ARM cpu).
145 x) VME processor emulation? "(Sanyo LC8670 "Potato")" according to
146 Wikipedia, LC86K87 according to Comstedt's page. See
147 http://www.maushammer.com/vmu.html for a good description of
148 the differences between LC86104C and the one used in the VME.
149
150 Alpha:
151 x) OSF1 PALcode, Virtual memory support.
152 x) PALcode replacement! PAL1E etc opcodes...?
153 x) Interrupt/exception/trap handling.
154 x) Floating point exception correctness.
155 x) More work on bootup memory and register contents.
156 x) More Alpha machine types, so it could work with
157 OpenBSD, FreeBSD, and Linux too?
158
159 SPARC (both the ISA and the machines):
160 o) Implement Adress space identifiers; load/stores etc.
161 o) Exception/trap/interrupt handling.
162 o) Save/restore register windows etc! Both v9 and pre-v9!
163 o) Finish the subcc and addcc flag computation code.
164 o) Add more registers (floating point, control regs etc)
165 o) Disassemly of some more instructions?
166 o) Are sll etc 32-bit sign-extending or zero-extending?
167 o) Floating point exception correctness.
168 o) SPARC v8, v7 etc?
169 o) More machine modes and devices.
170
171 POWER/PowerPC:
172 x) Fix DECR timer speed, so it matches the host.
173 x) NetBSD/prep 3.x triggers a possible bug in the emulator:
174 <wdc_exec_command(0xd005e514,0xd60cdd30,0,8,..)>
175 <ata_get_xfer(0,0xd60cdd30,0,8,..)>
176 <0x26c550(&ata_xfer_pool,2,0,8,..)>
177 <0x35c71c(0x3f27000,0,52,8,..)>
178 <ata_exec_xfer(0xd005e4c8,0x3f27000,0,13,..)>
179 <atastart(0xd005e4c8,0x3f27000,0,13,..)>
180 <__wdccommand_start(0xd005e4c8,0x3f27000,0,13,..)>
181 <bsw1(&prep_isa_io_space_tag,0x800001f6,0,176,..)>
182 [ wdc: write to SDH: 0xb0 (sectorsize 2, lba=1, drive 1, head 0) ]
183 <wdcwait(0xd005e4c8,72,64,0xbb8,..)>
184 <0x198120(0xd005e4c8,72,64,0xbb8,..)>
185 <bsr1(&prep_isa_io_space_tag,0,0,0xbb8,..)>
186 <delay(100,0,0,0xbb8,..)>
187 Note: <bsr1(&prep_isa_io_space_tag,0,0,0xbb8,..)>
188 x) PPC optimizations; instr combs
189 x) 64-bit stuff: either Linux on G5, or perhaps some hobbyist
190 version of AIX? (if there exists such a thing)
191 x) macppc: adb controller; keyboard (for framebuffer mode)
192 x) make OpenBSD/macppc work (PCI controller stuff)
193 x) Floating point exception correctness.
194 x) Alignment exceptions.
195
196 PReP:
197 x) Clock time! ("Bad battery blah blah")
198
199 Algor:
200 o) Other models than the P5064?
201 o) PCI interrupts... needed for stuff like the tlp NIC?
202
203 BeBox:
204 o) Interrupts. There seems to be a problem with WDC interrupts
205 "after a short while", although a few interrupts get through?
206 o) Perhaps find a copy of BeOS and try it?
207
208 Malta:
209 o) The Linux/Malta kernel at people.debian.org/~ths/qemu/malta/
210 almost works:
211 ./gxemul -x -o 'rd_start=0x80800000 rd_size=10000000 init=/bin/sh' -C 4KEc
212 -e malta 0x80800000:people.debian.org/~ths/qemu/malta/initrd.gz
213 people.debian.org/~ths/qemu/malta/vmlinux
214 (Remove "init=/bin/sh" to boot into the Debian installer.)
215 There are at least two things that need to be fixed:
216 1. PCI IDE; make Linux oops.
217 2. Implement the NIC.
218
219 HPCmips:
220 x) Mouse/pad support! :)
221 x) A NIC? (As a PCMCIA device?)
222
223 ARM:
224 o) See netwinder_reset() in NetBSD; the current "an internal error
225 occured" message after reboot/halt is too ugly.
226 o) Generic ARM "wait"-like instruction?
227 o) try to get netbsd/evbarm 3.x or 4.x running (iq80321)
228 o) make the xscale counter registers (ccnt) work
229 o) make the ata controller usable for FreeBSD!
230 o) Debian/cats crashes because of unimplemented coproc stuff.
231 fix this?
232
233 Test machines:
234 o) dev_fb block fill and copy
235 o) dev_fb draw characters (from the built-in font)?
236 o) dev_fb input device? mouse pointer coordinates and buttons
237 (allow changes in these to cause interrupts as well?)
238 o) Redefine the halt() function so that it stops "sometimes
239 soon", i.e. usage in demo code should be:
240 for (;;) {
241 halt();
242 }
243
244 Debugger:
245 o) How does SMP debugging work? Does it simply use "threads"?
246 What if the guest OS (running on an emulated SMP machine)
247 has a usertask running, with userland threads?
248 o) Try to make the debugger more modular and, if possible, reentrant!
249 o) Memory dumps should be able to dump both physical and
250 virtual emulated memory.
251 o) Evaluate expressions within []? That would allow stuff like
252 cpu[x] where x is an expression.
253 o) "pc = pc + 4" doesn't work! Bug. Should work. ("pc=pc+4" works.)
254 o) Settings:
255 x) Special handlers for Write!
256 +) MIPS coproc regs
257 +) Alpha/MIPS/SPARC zero registers
258 +) x86 64/32/16-bit registers
259 x) Value formatter for resulting output.
260 o) Call stack display (back-trace) of emulated programs.
261 o) Nicer looking output of register dumps, floating point registers,
262 etc. Warn about weird/invalid register contents.
263 o) Ctrl-C doesn't enter the debugger on some OSes (HP-UX?)...
264
265 Dyntrans:
266 x) NOTE: ARM etc. that load pc-relative constants, on writes to
267 pages that contain translations, the ENTIRE page must be
268 invalidated, not just the 1/32th that was code.
269 x) For 32-bit emulation modes, that have emulated TLBs: tlbindex
270 arrays of mapped pages? Things to think about:
271 x) Only 32-bit mode! (64-bit => too much code)
272 x) One array for global pages, and one array _PER ASID_,
273 for those archs that support that. On M88K, there should
274 be one array for userspace, and one for supervisor, etc.
275 x) Larger-than-4K-pages must fill several bits in the array.
276 x) No TLB search will be necessary.
277 x) Total host space used, for 4 KB pages: 1 MB per table,
278 i.e. 65 MB for 32-bit MIPS, 2 MB for M88K, if one byte
279 is used as the tlb index.
280 x) (The index is actually +1, so that 0 means no hit.)
281 x) "Merge" the cur_physpage and cur_ic_page variables/pointers to
282 one? I.e. change cur_ic_page to cur_physpage.ic_page or something.
283 x) Instruction combination collisions? How to avoid easily...
284 x) superh -- no hostpage for e.g. 0x8c000000. devices as ram!
285 x) Think about how to do both SHmedia and SHcompact in a reasonable
286 way! (Or AMD64 long/protected/real, for that matter.)
287 x) 68K emulation; think about how to do variable instruction
288 lengths across page boundaries.
289 x) Dyntrans with valgrind-inspired memory checker. (In memory_rw,
290 it would be reasonably simple to add; in each individual fast
291 load/store routine = a lot more work, and it would become
292 kludgy very fast.)
293 x) Dyntrans with SMP... lots of work to be done here.
294 x) Dyntrans with cache emulation... lots of work here as well.
295 x) Remove the concept of base RAM completely; it would be more
296 generic to allow RAM devices to be used "anywhere".
297 o) dev_mp doesn't work well with dyntrans yet
298 o) In general, IPIs, CAS, LL/SC etc must be made to work with dyntrans
299 x) Redesign/rethink the delay slot mechanism used for e.g. MIPS,
300 so that it caches a translation (that is, an instruction
301 word and the instr_call it was translated to the last
302 time), so that it doesn't need to do slow
303 to_be_translated for each end of page?
304 x) Program Counter statistics:
305 Per machine? What about SMP? All data to the same file?
306 A debugger command should be possible to use to enable/
307 disable statistics gathering.
308 Configuration file option!
309 x) Breakpoints:
310 o) Physical vs virtual addresses!
311 o) 32-bit vs 64-bit sign extension for MIPS, and others?
312 x) INVALIDATION should cause translations in _all_ cpus to be
313 invalidated, e.g. on a write to a write-protected page
314 (containing code)
315 x) 16-bit encodings? (MIPS16, ARM Thumb, etc)
316 x) Lots of other stuff: see src/cpus/README_DYNTRANS
317 x) Native code generation backends... think carefully about this.
318
319 Simple Valgrind-like checks?
320 o) Mark every address with bits which tell whether or not the address
321 has been written to.
322 o) What should happen when programs are loaded? Text/data, bss (zero
323 filled). But stack space and heap is uninitialized.
324 o) Uninitialized local variables:
325 A load from a place on the stack which has not previously
326 been stored to => warning. Increasing the stack pointer using
327 any available means should reset the memory to uninitialized.
328 o) If calls to malloc() and free() can be intercepted:
329 o) Access to a memory area after free() => warning.
330 o) Memory returned by malloc() is marked as not-initialized.
331 o) Non-passive, but good to have: Change the argument
332 given to malloc, to return a slightly larger memory
333 area, i.e. margin_before + size + margin_after,
334 and return the pointer + margin_before.
335 Any access to the margin_before or _after space results
336 in warnings. (free() must be modified to free the
337 actually allocated address.)
338
339 Better CD Image file support:
340 x) Support CD formats that contain more than 1 track, e.g.
341 CDI files (?). These can then contain a mixture of e.g. sound
342 and data tracks, and booting from an ISO filesystem path
343 would boot from [by default] the first data track.
344 (This would make sense for e.g. Dreamcast CD images, or
345 possibly other live-CD formats.)
346
347 Networking:
348 x) Redesign of the networking subsystem, at least the NAT translation
349 part. The current way of allowing raw ethernet frames to be
350 transfered to/from the emulator via UDP should probably be
351 extended to allow the frames to be transmitted other ways as
352 well.
353 x) Also adding support for connecting ttys (either to xterms, or to
354 pipes/sockets etc, or even to PPP->NAT or SLIP->NAT :-).
355 x) Documentation updates (!) are very important, making it easier to
356 use the (already existing) network emulation features.
357 x) Fix performance problems caused by only allowing a
358 single TCP packet to be unacked.
359 x) Don't hardcode offsets into packets!
360 x) Test with lower than 100 max tcp/udp connections,
361 to make sure that reuse works!
362 x) Make OpenBSD work better as a guest OS!
363 x) DHCP? Debian doesn't actually send DHCP packets, even
364 though it claims to? So it is hard to test.
365 x) Multiple networks per emulation, and let different
366 NICs in machines connect to different networks.
367 x) Support VDE (vde.sf.net)? Easiest/cleanest (before a
368 redesign of the network framework has been done) is
369 probably to connect it using the current (udp) solution.
370 x) Allow SLIP connections, possibly PPP, in addition to
371 ethernet?
372
373 Cache simulation:
374 o) Command line flags for:
375 o) CPU endianness?
376 o) Cache sizes? (multiple levels)
377 o) Separate from the CPU concept, so that multi-core CPUs sharing
378 e.g. a L2 cache can be simulated (?)
379 o) Instruction cache emulation is easiest (if separate from the
380 data cache); similar hack as the S;I; hack in cpu_dyntrans.c.
381 NOTE: if the architecture has a delay slot, then an instruction
382 slot can actually be executed as 2 instructions.
383 o) Data cache emulation = harder; each arch's load/store routines
384 must include support? running one instruction at a time and
385 having a cpu-dependant lookup function for each instruction
386 is another option (easier to implement, but very very slow).
387
388 Documentation:
389 x) Update the documentation regarding the testmachine interrupts.
390 x) Note about sandboxing/security:
391 Not all emulated instructions fail in the way they would
392 do on real hardware (e.g. a userspace program writing to
393 a system register might work in GXemul, but it would
394 fail on real hardware). Sandbox = contain from the
395 host OS. But the emulated programs will run "less
396 securely".
397 x) Try NetBSD/arc 4.x! (It seems to work with disk images!)
398 x) NetBSD/pmax 4 install instructions: xterm instead of vt100!
399 x) Rewrite the section about experimental devices, after the
400 framebuffer acceleration has been implemented, and demos
401 written. (Symbolic names instead of numbers; example
402 use cases, etc. Mention demo files that use the various
403 features?)
404 x) "a very simple linear framebuffer device (for graphics output)"
405 under "which machines does gxemul emulate" ==> better
406 description?
407
408 The Device subsystem:
409 x) allow devices to be moved and/or changed in size (down to a
410 minimum size, etc, or up to a max size); if there is a collision,
411 return false. It is up to the caller to handle this situation!
412 x) NOTE: Translations must be invalidated, both for
413 registering new devices, and for moving existing ones.
414 cpu->invalidate translation caches, for all CPUs that
415 are connected to a specific memory.
416
417 PCI:
418 x) Pretty much everything related to runtime configuration, device
419 slots, interrupts, etc must be redesigned/cleaned up. The current
420 code is very hardcoded and ugly.
421 o) Allow cards to be added/removed during runtime more easily.
422 o) Allow cards to be enabled/disabled (i/o ports, etc, like
423 NetBSD needs for disk controller detection).
424 o) Allow devices to be moved in memory during runtime.
425 o) Interrupts per PCI slot, etc. (A-D).
426 o) PCI interrupt controller logic... very hard to get right,
427 because these differ a lot from one machine to the next.
428 x) last write was ffffffff ==> fix this, it should be used
429 together with a mask to get the correct bits. also, not ALL
430 bits are size bits! (lowest 4 vs lowest 2?)
431 x) add support for address fixups
432 x) generalize the interrupt routing stuff (lines etc)
433
434 Clocks and timers:
435 x) Fix the PowerPC DECR interrupt speed! (MacPPC and PReP speed, etc.)
436 x) DON'T HARDCODE 100 HZ IN cpu_mips_coproc.c!
437 x) NetWinder timeofday is incorrect! Huh? grep -R for ta_rtc_read in
438 NetBSD sources; it doesn't seem to be initialized _AT ALL_?!
439 x) Cobalt TOD is incorrect!
440 x) Go through all other machines, one by one, and fix them.
441
442 Config file parser:
443 o) Rewrite it from scratch!
444 o) Usage of any expression available through the debugger
445 o) Allow interrupt controllers to be added! and interrupts
446 to be used in more ways than before
447 o) Support for running debugger commands (like the -c
448 command line option)
449
450 Floating point layer:
451 o) make it common enough to be used by _all_ emulation modes
452 o) implement correct error/exception handling and rounding modes
453 o) implement more helper functions (i.e. add, sub, mul...)
454 o) non-IEEE modes (i.e. x86)?
455
456 Userland emulation:
457 x) Try to prefix "/emul/mips/" or similar to all filenames,
458 and only if that fails, try the given filename.
459 Read this setting from an environment variable, and only
460 if there is none, fall back to hardcoded string.
461 x) File descriptor (0,1,2) assumptions? Find and fix these?
462 x) Dynamic linking!
463 x) Lots of stuff; freebsd, netbsd, linux, ... syscalls.
464 x) Initial register/stack contents (environment, command line args).
465 x) Return value (from main).
466 x) mmap emulation layer
467 x) errno emulation layer
468 x) ioctl emulation layer for all devices :-[
469 x) struct conversions for many syscalls
470
471 Sound:
472 x) generic sound framework
473 x) add one or more sound cards as devices; add a testmachine
474 sound card first?
475 x) Dreamcast sound? Generic PCI sound cards?
476
477 ASC SCSI controller:
478 x) NetBSD/arc 2.0 uses the ASC controller in a way which GXemul
479 cannot yet handle. (NetBSD 1.6.2 works ok.) (Possibly a problem
480 in NetBSD itself, http://mail-index.netbsd.org/source-changes/
481 2005/11/06/0024.html suggests that.)
482 NetBSD 4.x seems to work? :)
483
484 Caches / memory hierarchies: (this is mostly MIPS-specific)
485 o) src/memory*.c: Implement correct cache emulation for
486 all CPU types. (currently only R2000/R3000 is implemented)
487 (per CPU, multiple levels should be possible, associativity etc!)
488 o) R2000/R3000 isn't _100%_ correct, just almost correct :)
489 o) Move the -S (fill mem with random) functionality into the
490 memory.c subsystem, not machine.c or wherever it is now
491 o) ECC stuff, simulation of memory errors? (Machine dependent)
492 o) More than 4GB of emulated RAM, when run on a 32-bit host?
493 (using manual swap-out of blocks to disk, ugly)
494 o) A global command line option should be used to turn
495 cache emulation on or off. When off, caches should be
496 faked like they are right now. When on, caches and
497 memory latencies should be emulated as correctly as
498 possible.
499
500 File/disk/symbol handling:
501 o) Make sure that disks can be added/removed during runtime!
502 (Perhaps this needs a reasonably large re-write.)
503 o) Remove some of the complexity in file format guessing, for
504 Ultrix kernels that are actually disk images?
505 o) Remove temporary files (/tmp/gxemul.blahblah) if loading fails
506 for some reason (unrecognized file, etc).
507 o) Better handling of tape files
508 o) Read function argument count and types from binaries? (ELF?)
509 o) Better demangling of C++ names. Note: GNU's C++ differs from e.g.
510 Microsoft's C++, so multiple schemes must be possible. See
511 URL at top of src/symbol_demangle.c for more info.
512
513 Userland ABI emulation:
514 o) see src/useremul.c
515
516 Better framebuffer and X-windows functionality:
517 o) Do a complete rewrite of the framebuffer/console stuff, so that:
518 1) It does not rely on X11 specifically.
519 2) It is possible to interact with emulated framebuffers
520 and consoles "remotely", e.g. via a web page which
521 controls multiple virtualized machines.
522 3) It is possible to run on (hypothetical) non-X11
523 graphics systems.
524 o) Generalize the update_x1y1x2y2 stuff to an extend-region()
525 function...
526 o) -Yx sometimes causes crashes.
527 o) Simple device access to framebuffer_blockcopyfill() etc,
528 and text output (using the built-in fonts), for dev_fb.
529 o) CLEAN UP the ugly event code
530 o) Mouse clicks can be "missed" in the current system; this is
531 not good. They should be put on a stack of some kind.
532 o) More 2D and 3D framebuffer acceleration.
533 o) Non-resizable windows? Or choose scaledown depending
534 on size (and center the image, with a black border).
535 o) Different scaledown on different windows?
536 o) Non-integral scale-up? (E.g. 640x480 -> 1024x768)
537 o) Switch scaledown during runtime? (Ala CTRL-ALT-plus/minus)
538 o) Bug reported by Elijah Rutschman on MacOS with weird
539 keys (F5 = cursor down?).
540 o) Keyboard and mouse events:
541 x) Do this for more machines than just DECstation
542 x) more X11 cursor keycodes
543 x) Keys like CTRL, ALT, SHIFT do not get through
544 by themselves (these are necessary for example
545 to change the font of an xterm in X in the
546 emulator)
547 o) Generalize the framebuffer stuff by moving _ALL_ X11
548 specific code to a separate module.
549
550 -------------------------------------------------------------------------------
551

  ViewVC Help
Powered by ViewVC 1.1.26