/[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 26 - (show annotations)
Mon Oct 8 16:20:10 2007 UTC (13 years ago) by dpavlin
File size: 13416 byte(s)
++ trunk/HISTORY	(local)
$Id: HISTORY,v 1.1264 2006/06/25 11:08:04 debug Exp $
20060624	Replacing the error-prone machine type initialization stuff
		with something more reasonable.
		Finally removing the old "cpu_run" kludge; moving around stuff
		in machine.c and emul.c to better suit the dyntrans system.
		Various minor dyntrans cleanups (renaming translate_address to
		translate_v2p, and experimenting with template physpages).
20060625	Removing the speed hack which separated the vph entries into
		two halves (code vs data); things seem a lot more stable now.
		Minor performance hack: R2000/R3000 cache isolation now only
		clears address translations when going into isolation, not
		when going out of it.
		Fixing the MIPS interrupt problems by letting mtc0 immediately
		cause interrupts.

==============  RELEASE 0.4.0.1  ==============


1 $Id: TODO,v 1.298 2006/06/25 11:08:04 debug Exp $
2
3 Hm. This file is in random order, and not all parts of it are up-to-date.
4
5 --------------
6
7 x) FIX THE NON-R3000 TRANSLATION CACHE INVALIDATION BOTTLENECKS!
8 x) Find a way to get rid of the cpu_create_or_reset_tc in the
9 R2000/R3000 cache isolation code. (NetBSD works without it,
10 but not Ultrix and Linux yet.)
11 x) Formalize the statistics gathering stuff for dyntrans...
12 x) ... and use it to optimize MIPS dyntrans stuff.
13 x) Clock framework? Go through all clock devices, make sure they
14 return correct data, and run at correct speeds!
15 x) Optimizations, continuing on 64-bit issues etc with dyntrans
16 x) Dyntrans with SMP... lots of work to be done here.
17 x) Dyntrans with cache emulation... lots of work here as well.
18 x) Actually use the settings object, better debugger stuff, etc.
19 x) Wait for new releases of NetBSD, and test with those.
20
21 --------------
22
23 SMP:
24 o) dev_mp doesn't work well with dyntrans yet
25 o) In general, IPIs, CAS, LL/SC etc must be made to work with dyntrans
26
27 MIPS:
28 o) Fix invalidate_asid so it works well for non-R3000 too!
29 x) [Re]add an interrupt-asserted bit for MIPS, to speed up
30 interrupt handling slightly?
31 +) Print a warning on the first reserved instruction.
32 +) Some more work on opcodes.
33 x) MIPS64 revision 2.
34 x) _MAYBE_ TX79 and R5900 actually differ in their
35 opcodes? Check this carefully!
36 o) Dyntrans: Count register updates are probably not 100% correct yet.
37 o) Dyntrans: SMP correctness
38 o) Refactor code for performance and readability/maintainability.
39 o) Instruction combinations? Possible candidates (but profile first!):
40 o) multiple loads/stores in a row
41 o) strlen, memset loops etc
42 o) compare + branch
43 o) DROTR32 and similar MIPS64 rev 2 instructions, which have
44 a rotation bit which differs from previous ISAs.
45 o) EI and DI instructions for MIPS64/32 rev 2. NOTE: These are
46 _NOT_ the same as for R5900!
47 o) R4000 and others:
48 x) watchhi/watchlo exceptions, and other exception
49 handling details
50 o) R10000 and others: (R12000, R14000 ?)
51 x) memory space, exceptions, ...
52 x) use cop0 framemask for tlb lookups
53 (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)
54
55 Dyntrans:
56 x) Move the mips_init_64bit_dummy_tables() etc calls into
57 src/cpu.c, for all 64-bit cpus?
58 x) 64-bit "phystranslation" lookup as in 32-bit mode? Would probably
59 help performance a bit.
60 x) Common fatal_abort() function, which drops into the debugger
61 without continuing.
62 x) INVALIDATION should cause translations in _all_ cpus to be
63 invalidated, e.g. on a write to a write-protected page
64 (containing code)
65 x) better (formally defined) instr call statistics (-s command
66 line option?), multiple different types? (virtual pc, physical pc)
67 x) Call/return hints?
68 x) 16-bit encodings? (MIPS16, ARM Thumb, SH3, ...)
69 x) H8?
70 x) Lots of other stuff: see src/cpus/README_DYNTRANS
71 x) true recompilation backend? think carefully about this,
72 experiment in a separate project (not in GXemul)
73 x) Remove the dyntrans_alignment_check functionality; although
74 it gives slightly higher peformance sometimes, it increases
75 the complexity of the code too much!
76
77 Alpha:
78 o) Virtual memory (tlbs etc)
79 o) Get {NetBSD,OpenBSD,Linux}/alpha booting. :)
80
81 SPARC:
82 o) Add all registers (floating point, control regs etc)
83 o) Save/restore register windows etc!
84 o) Disassemly of some more instructions.
85 o) Are sll etc 32-bit sign-extending or zero-extending?
86 o) Finish the cmp (subcc) flag computation code.
87 o) Finish the GDB register stuff.
88
89 Debugger:
90 o) How does SMP debugging work? Does it simply use "threads"?
91 What if the guest OS (running on an emulated SMP machine)
92 has a usertask running, with userland threads?
93 o) Try to make the debugger more modular and, if possible, reentrant!
94 o) Remove the emul command? (But show network info if showing
95 machines?)
96 o) Generalize the expression evaluator. (debugger_expr.c?)
97 settable variables ("show nr of instructions on average")
98 emul[x] defaults to current emul
99 machine[x] defaults to current machine
100 cpu[x] defaults to currently focused cpu
101 registers cpu arch dependent (#-prefix)
102 symbols @-prefix
103 numeric constants decimal, hex, and octal ($-prefix)
104 boolean yes,no, true,false
105 operators (+ - * / % & | ^ !)
106 parentheses for grouping subexpressions
107 NOTE: the change from % to # for register prefix!
108 examples:
109 emul[0].machine[2].cpu[0].pc
110 machine[test2].cpu[1].ra = main
111 settings.show_trace_tree = yes
112
113 Settings:
114 o) Remove a setting.
115 o) Read/write a setting given a name. (Read as
116 string and/or int64_t simultaneously?)
117
118 Help command should have subsections! One for "expressions",
119 mirrored in the documentation, but the internal help should
120 be the one that should be considered correct.
121 o) see src/debugger.c for more
122
123 POWER/PowerPC:
124 x) PPC optimizations; instr combs
125 x) 64-bit stuff
126 x) find and fix the bug which causes NetBSD/macppc to fail after
127 an install!
128 x) macppc: adb controller; keyboard (for framebuffer mode)
129 x) make OpenBSD/macppc work (PCI controller stuff)
130
131 Algor:
132 PCI and ISA and LOCAL interrupts! --> wdc could start working
133 Add interrupt controller in dev_algor.c.
134
135 ARM:
136 o) try to get netbsd/evbarm 3.x running (iq80321)
137 o) make the xscale counter registers (ccnt) work
138 o) make the ata controller usable for FreeBSD!
139 o) zaurus for openbsd...
140 o) debian/cats crashes because of unimplemented coproc stuff.
141 fix this?
142
143 Cache simulation:
144 o) Separate from the CPU concept, so that multi-core CPUs sharing
145 e.g. a L2 cache can be simulated (?)
146 o) Instruction cache emulation is easiest (if separate from the
147 data cache); similar hack as the S;I; hack in cpu_dyntrans.c.
148 NOTE: if the architecture has a delay slot, then an instruction
149 slot can actually be executed as 2 instructions.
150 o) Data cache emulation = harder; each arch's load/store routines
151 must include support? running one instruction at a time and
152 having a cpu-dependant lookup function for each instruction
153 is another option (easier to implement, but very very slow).
154
155 Documentation:
156 o) machines, cpus, devices.
157 o) Automagic documentation generation:
158 x) REMEMBER that several machines/devices can be in
159 the same source file!
160 o) Try to rewrite the install instructions for those machines
161 that use 3MAX into using CATS? (To remove the need to a raw
162 ffs partition using up all of the disk image.)
163
164 More generic out_of_memory error reporting, and check everywhere!
165 Causes: OpenBSD has low default limits for normal users.
166 Host is 32-bit? (32-bit hosts are limited to 4 GB or less
167 of userspace memory.)
168 You are actually low on RAM. (As trivial as this might sound,
169 Unix systems usually allow processes to allocate virtual
170 memory beyond the amount of RAM in the machine.)
171
172 Breakpoints: 32-bit vs 64-bit sign extension for MIPS, warnings, etc.
173 Use the debugger's symbolic name stuff. (which will have to be
174 extended soon to support stuff like "2*x + symbol + y" etc. cool
175 stuff)
176
177 Sprite (guest OS for DECstation emulation)
178 x) Timing problems during bootup?
179
180 The Device subsystem:
181 x) allow devices to be moved and/or changed in size (down to a
182 minimum size, etc, or up to a max size)
183 x) keep track of interrupts and busses? actually, allowing any device
184 to be a bus might be a nice idea.
185 x) turn interrupt controllers into devices? :-)
186 x) refactor various clocks/nvram/cmos into one device?
187
188 Clocks:
189 x) General framework for automagic clock adjustment for _all_
190 kinds of clocks and timers. (Which should be possible to turn
191 off, of course, like the way DECstation emulation works now.)
192
193 PCI:
194 x) last write was ffffffff ==> fix this, it should be used
195 together with a mask to get the correct bits. also, not ALL
196 bits are size bits! (lowest 4 vs lowest 2?)
197 x) add support for address fixups
198 x) generalize the interrupt routing stuff (lines etc). this should
199 be per machine? or per bus, that's better
200 x) add a "pcn" NIC (AMD PCnet32 Lance 79c970 (PCI 1022:2000)),
201 could be useful for several machine modes (Malta, Algor, evbarm,
202 hp700?, macppc, etc.)
203
204 Network layer:
205 o) DHCP (for Debian and BSD installers :-)
206 o) increase performance
207 o) don't rely on NetBSD-ish usage
208 o) Multiple networks per emulation, and let different
209 NICs in machines connect to different networks.
210 o) many other issues: see src/net.c
211
212 Busses:
213 o) Redesign the entire "mainbus" concept!
214 o) Busses should be placed in a hierarchical tree!
215 o) Easily configurable interrupt routing in SMP systems.
216 o) Specific clock/bus speeds, cpu speeds etc.
217 o) Synchronization over network? or at least in dyntrans within
218 one emulated machine
219 o) dev->bus: TurboChannel, PCMCIA, ADB?
220
221 Config file parser:
222 o) Rewrite it from scratch!
223 o) Usage of any expression available through the debugger
224 o) Support for running debugger commands (like the -c
225 command line option)
226
227 Floating point layer:
228 o) make it common enough to be used by _all_ emulation modes
229 o) implement more stuff
230 o) non-IEEE modes (i.e. x86)?
231
232 Userland emulation:
233 x) Lots of stuff; freebsd and netbsd (and linux?) syscalls.
234 x) Dynamic linking? Hm.
235
236 Sound:
237 x) generic sound framework
238 x) add one or more sound cards as devices
239
240 ASC SCSI controller:
241 x) NetBSD/arc 2.0 uses the ASC controller in a way which GXemul
242 cannot yet handle. (NetBSD 1.6.2 works ok.) (Possibly a problem
243 in NetBSD itself, http://mail-index.netbsd.org/source-changes/
244 2005/11/06/0024.html suggests that.)
245
246 Caches / memory hierarchies: (this is mostly MIPS-specific)
247 o) MIPS coproc.c: bits in config registers should reflect
248 correct cache sizes for _all_ CPU types. (currently only
249 implemented for R4000, R1x000, and a few others)
250 o) src/memory*.c: Implement correct cache emulation for
251 all CPU types. (currently only R2000/R3000 is implemented)
252 (per CPU, multiple levels should be possible,
253 associativity etc!)
254 o) R2000/R3000 isn't _100%_ correct, just almost correct :)
255 o) Move the -S (fill mem with random) functionality into the
256 memory.c subsystem, not machine.c or wherever it is now
257 o) ECC stuff, simulation of memory errors? (Machine dependent)
258 o) More than 4GB of emulated RAM, when run on a 32-bit host?
259 (using manual swap-out of blocks to disk, ugly)
260 o) A global command line option should be used to turn
261 cache emulation on or off. When off, caches should be
262 faked like they are right now. When on, caches and
263 memory latencies should be emulated as correctly as
264 possible.
265
266 File/disk/symbol handling:
267 o) Better handling of tape files
268 o) Read function argument count and types from binaries? (ELF?)
269 o) Better demangling of C++ names. Note: GNU's C++ differs from e.g.
270 Microsoft's C++, so multiple schemes must be possible. See
271 URL at top of src/symbol_demangle.c for more info.
272
273 Userland ABI emulation:
274 o) see src/useremul.c
275
276 Terminal/console:
277 o) allow emulated serial ports to be connected to the outside
278 world in a more generic way, or even to other emulated
279 machines(?)
280
281 Save state of the whole emulated machine, to be able to load it back
282 in later? (Memory, all device's states, all registers and
283 so on. Like taking a snapshot. (SimOS seems to do this,
284 according to its website.))
285
286 Better framebuffer and X-windows functionality:
287 o) -Yx sometimes causes crashes.
288 o) Simple device access to framebuffer_blockcopyfill() etc,
289 and text output (using the built-in fonts), for dev_fb.
290 o) CLEAN UP the ugly event code
291 o) Mouse clicks can be "missed" in the current system; this is
292 not good. They should be put on a stack of some kind.
293 o) More 2D and 3D framebuffer acceleration.
294 o) Non-resizable windows? Or choose scaledown depending
295 on size (and center the image, with a black border).
296 o) Different scaledown on different windows?
297 o) Switch scaledown during runtime? (Ala CTRL-ALT-plus/minus)
298 o) Bug reported by Elijah Rutschman on MacOS with weird
299 keys (F5 = cursor down?).
300 o) Keyboard and mouse events:
301 x) Do this for more machines than just DECstation
302 x) more X11 cursor keycodes
303 x) Keys like CTRL, ALT, SHIFT do not get through
304 by themselves (these are necessary for example
305 to change the font of an xterm in X in the
306 emulator)
307 o) Generalize the framebuffer stuff by moving _ALL_ X11
308 specific code to src/x11.c!
309
310 Statistics: (this could be interesting)
311 o) Save to file and show graphics. It should be possible to
312 run gxemul after a simulation to just show the graphics,
313 or convert to a .ppm or .tga or similar.
314 o) memory accesses (to measure cache efficiency and
315 page coloring efficiency)
316 o) nr of simultaneous ASIDs in use in the TLB, for MIPS
317 o) percentage of time spent in different "states", such as
318 running userland code, kernel code, or idling (for CPUs
319 that have such an instruction, or whenever the PC is
320 inside a specific idle-function (address range)).
321 Possible additional state (for example on R3000): caches
322 disabled.
323 o) position of read/write on (SCSI) disks
324

  ViewVC Help
Powered by ViewVC 1.1.26