/[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 30 - (show annotations)
Mon Oct 8 16:20:40 2007 UTC (13 years ago) by dpavlin
File size: 17305 byte(s)
++ trunk/HISTORY	(local)
$Id: HISTORY,v 1.1325 2006/08/15 15:38:37 debug Exp $
20060723	More Transputer instructions (pfix, nfix, opr, mint, ldl, ldlp,
		eqc, rev, ajw, stl, stlf, sthf, sub, ldnl, ldnlp, ldpi, move,
		wcnt, add, bcnt).
		Adding more SPARC instructions (andcc, addcc, bl, rdpr).
		Progress on the igsfb framebuffer used by NetBSD/netwinder.
		Enabling 8-bit fills in dev_fb.
		NetBSD/netwinder 3.0.1 can now run from a disk image :-)
20060724	Cleanup/performance fix for 64-bit virtual translation table
		updates (by removing the "timestamp" stuff). A full NetBSD/pmax
		3.0.1 install for R4400 has dropped from 667 seconds to 584 :)
		Fixing the igsfb "almost vga" color (it is 24-bit, not 18-bit).
		Adding some MIPS instruction combinations (3*lw, and 3*addu).
		The 8048 keyboard now turns off interrupt enable between the
		KBR_ACK and the KBR_RSTDONE, to work better with Linux 2.6.
		Not causing PPC DEC interrupts if PPC_NO_DEC is set for a
		specific CPU; NetBSD/bebox gets slightly further than before.
		Adding some more SPARC instructions: branches, udiv.
20060725	Refreshing dev_pckbc.c a little.
		Cleanups for the SH emulation mode, and adding the first
		"compact" (16-bit) instructions: various simple movs, nop,
		shll, stc, or, ldc.
20060726	Adding dummy "pcn" (AMD PCnet NIC) PCI glue.
20060727	Various cleanups; removing stuff from cpu.h, such as
		running_translated (not really meaningful anymore), and
		page flags (breaking into the debugger clears all translations
		anyway).
		Minor MIPS instruction combination updates.
20060807	Expanding the 3*sw and 3*lw MIPS instruction combinations to
		work with 2* and 4* too, resulting in a minor performance gain.
		Implementing a usleep hack for the RM52xx/MIPS32/MIPS64 "wait"
		instruction (when emulating 1 cpu).
20060808	Experimenting with some more MIPS instruction combinations.
		Implementing support for showing a (hardcoded 12x22) text
		cursor in igsfb.
20060809	Simplifying the NetBSD/evbmips (Malta) install instructions
		somewhat (by using a NetBSD/pmax ramdisk install kernel).
20060812	Experimenting more with the MIPS 'wait' instruction.
		PCI configuration register writes can now be handled, which
		allow PCI IDE controllers to work with NetBSD/Malta 3.0.1 and
		NetBSD/cobalt 3.0.1. (Previously only NetBSD 2.1 worked.)
20060813	Updating dev_gt.c based on numbers from Alec Voropay, to enable
		Linux 2.6 to use PCI on Malta.
		Continuing on Algor interrupt stuff.
20060814	Adding support for routing ISA interrupts to two different
		interrupts, making it possible to run NetBSD/algor :-)
20060814-15	Testing for the release.

==============  RELEASE 0.4.2  ==============


1 $Id: TODO,v 1.343 2006/08/14 18:46:30 debug Exp $
2
3 Hm. This file is in random order, and not all parts of it are up-to-date.
4
5
6 Implementation:
7 x) ARM "wait"-like instruction.
8 x) CLOCK FRAMEWORK!
9 x) Mouse support for NetBSD/pmax 4.x!
10 x) See netwinder_reset() in NetBSD; the current "an internal error
11 occured" message after reboot/halt is too ugly.
12 x) 64-bit ranges in src/cpus/memory_mips_v2p.c
13 x) Revert the dyntrans page template experiment? Hm.
14 x) Refactor the cpu type detection/initialization/listing.
15 Macro, which can be used as long as the cpu definitions
16 contain a 'name'?
17 x) Testmachine includes:
18 + dev_fb block fill and copy
19 + dev_fb draw characters (from the built-in font)?
20 + dev_fb input device? mouse pointer coordinates and buttons
21 (allow changes in these to cause interrupts as well?)
22 + Redefine the halt() function so that it stops "sometimes
23 soon", i.e. usage in demo code should be:
24 for (;;) {
25 halt();
26 }
27 x) Continue on SPARC emulation
28 + Enable it in the configure script as soon as it can
29 run all the demo programs.
30 x) Continue on Alpha emulation (virtual memory, etc). Cleanup.
31 x) Nicer MIPS status bits in register dumps.
32 x) Alignment exceptions (MIPS, PPC, ARM?, ...)
33 x) Rewrite the networking stack; make OpenBSD work better as a guest
34 OS, fix the performance problems, make Linux work with DHCP, etc.
35 Support VDE (vde.sf.net)? Allow SLIP connections, possibly PPP,
36 in addition to ethernet?
37 x) Implement more ethernet NICs.
38 x) IOP (I2O) device?
39
40 Documentation:
41 x) "Install netbsd/pmax first" => only use the install kernel?
42 x) Rewrite the section about experimental devices, after the
43 framebuffer acceleration has been implemented, and demos
44 written. (Symbolic names instead of numbers; example
45 use cases, etc. Mention demo files that use the various
46 features?)
47 x) "a very simple linear framebuffer device (for graphics output)"
48 under "which machines does gxemul emulate" ==> better
49 description?
50 x) Better description on how to set up a cross compiler?
51 Example for MIPS64.
52
53 Long-term design:
54 x) Instruction combination collisions? How to avoid easily...
55 x) Think about how to do both SHmedia and SHcompact in a reasonable
56 way!
57 o) Actually use the settings object, better debugger stuff, etc!
58 o) Debugger command for enabling/disabling instruction statistics
59 during runtime. machine.statistics = on|off
60 x) MAINBUS REDESIGN!
61 x) PCI redesign... I need to read up on how PCI actually works :)
62 x) Clock framework! Go through all clock devices, make sure they
63 return correct data, and run at correct speeds!
64 x) Dyntrans with valgrind-inspired memory checker. (In memory_rw,
65 it would be reasonably simple to add; in each individual fast
66 load/store routine = a lot more work, and it would become
67 kludgy very fast.)
68 x) Dyntrans with SMP... lots of work to be done here.
69 x) Dyntrans with cache emulation... lots of work here as well.
70 x) Reimplement the config file parser from scratch.
71
72 -------------------------------------------------------------------------------
73
74 Simple Valgrind-like checks?
75 o) Mark every address with bits which tell whether or not the address
76 has been written to.
77 o) What should happen when programs are loaded? Text/data, bss (zero
78 filled). But stack space and heap is uninitialized.
79 o) Uninitialized local variables:
80 A load from a place on the stack which has not previously
81 been stored to => warning. Increasing the stack pointer using
82 any available means should reset the memory to uninitialized.
83 o) If calls to malloc() and free() can be intercepted:
84 o) Access to a memory area after free() => warning.
85 o) Memory returned by malloc() is marked as not-initialized.
86 o) Non-passive, but good to have: Change the argument
87 given to malloc, to return a slightly larger memory
88 area, i.e. margin_before + size + margin_after,
89 and return the pointer + margin_before.
90 Any access to the margin_before or _after space results
91 in warnings. (free() must be modified to free the
92 actually allocated address.)
93
94 SMP:
95 o) dev_mp doesn't work well with dyntrans yet
96 o) In general, IPIs, CAS, LL/SC etc must be made to work with dyntrans
97
98 MIPS:
99 +) Some more work on opcodes.
100 x) The "wait" instruction. How to implement this functionality?
101 (SMP, non-MIPS, interrupt correctness, host idling, ...)
102 x) MIPS64 revision 2.
103 o) Find out which actual CPUs implement the rev2 ISA!
104 x) _MAYBE_ TX79 and R5900 actually differ in their
105 opcodes? Check this carefully!
106 o) Dyntrans: Count register updates are probably not 100% correct yet.
107 o) Refactor code for performance and readability/maintainability.
108 o) DROTR32 and similar MIPS64 rev 2 instructions, which have
109 a rotation bit which differs from previous ISAs.
110 o) EI and DI instructions for MIPS64/32 rev 2. NOTE: These are
111 _NOT_ the same as for R5900!
112 o) (Re)implement 128-bit loads/stores for R5900.
113 o) R4000 and others:
114 x) watchhi/watchlo exceptions, and other exception
115 handling details
116 o) R10000 and others: (R12000, R14000 ?)
117 x) memory space, exceptions, ...
118 x) use cop0 framemask for tlb lookups
119 (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)
120
121 Dyntrans:
122 x) Redesign/rethink the delay slot mechanism used for e.g. MIPS,
123 so that it caches a translation (that is, an instruction
124 word and the instr_call it was translated to the last
125 time), so that it doesn't need to do slow
126 to_be_translated for each end of page?
127 x) Program Counter statistics:
128 Per machine? What about SMP? All data to the same file?
129 A debugger command should be possible to use to enable/
130 disable statistics gathering.
131 Configuration file option!
132 x) Common fatal_abort() function, which drops into the debugger
133 without continuing.
134 x) INVALIDATION should cause translations in _all_ cpus to be
135 invalidated, e.g. on a write to a write-protected page
136 (containing code)
137 x) 16-bit encodings? (MIPS16, ARM Thumb, SH3, ...)
138 x) Lots of other stuff: see src/cpus/README_DYNTRANS
139 x) true recompilation backend? think carefully about this,
140 experiment in a separate project (not in GXemul)
141 o) First test would be to just implement a simple
142 instruction such as MIPS' addiu or lui, on AMD64
143 hosts...
144 x) Idle loop detection? (Depends on target.) Could be turned
145 into usleep(1) or similar on the host... except when doing
146 e.g. SMP emulation. Then it becomes trickier.
147
148 Transputer:
149 x) Implement support for Helios binaries.
150 x) Stack and register contents at startup?
151 x) Figure out how to boot an entire Helios distribution.
152 x) Implement all instructions. :)
153
154 Alpha:
155 o) Virtual memory (tlbs etc)
156 o) Get {NetBSD,OpenBSD,Linux}/alpha booting. :)
157
158 SPARC:
159 o) Load/stores to alternate address spaces!
160 o) Save/restore register windows etc!
161 o) Finish the subcc and addcc flag computation code.
162 o) Add more registers (floating point, control regs etc)
163 o) Disassemly of some more instructions?
164 o) Are sll etc 32-bit sign-extending or zero-extending?
165 o) Finish the GDB register stuff.
166 o) SPARC v8, v7 etc?
167
168 Debugger:
169 o) How does SMP debugging work? Does it simply use "threads"?
170 What if the guest OS (running on an emulated SMP machine)
171 has a usertask running, with userland threads?
172 o) Try to make the debugger more modular and, if possible, reentrant!
173 o) Remove the emul command? (But show network info if showing
174 machines?)
175 o) Generalize the expression evaluator. (debugger_expr.c?)
176 settable variables ("show nr of instructions on average")
177 emul[x] defaults to current emul
178 machine[x] defaults to current machine
179 cpu[x] defaults to currently focused cpu
180 registers cpu arch dependent (#-prefix)
181 symbols @-prefix
182 numeric constants decimal, hex, and octal ($-prefix)
183 boolean yes,no, true,false
184 operators (+ - * / % & | ^ !)
185 parentheses for grouping subexpressions
186 NOTE: the change from % to # for register prefix!
187 examples:
188 emul[0].machine[2].cpu[0].pc
189 machine[test2].cpu[1].ra = main
190 settings.show_trace_tree = yes
191
192 Settings:
193 o) Remove a setting.
194 o) Read/write a setting given a name. (Read as
195 string and/or int64_t simultaneously?)
196 o) Warnings when exiting the emulator, if the
197 settings have not been removed exactly in
198 the same way as they were added? This would
199 improve code cleanliness in the long term.
200 (I.e. require a corresponding _destroy()
201 function for all _new functions... machine_
202 cpu_ etc.)
203
204 Help command should have subsections! One for "expressions",
205 mirrored in the documentation, but the internal help should
206 be the one that should be considered correct.
207 o) see src/debugger.c for more
208
209 POWER/PowerPC:
210 x) PPC optimizations; instr combs
211 x) 64-bit stuff: either Linux on G5, or perhaps some hobbyist
212 version of AIX? (if there exists such a thing)
213 x) find and fix the bug which causes NetBSD/macppc to fail after
214 an install!
215 x) macppc: adb controller; keyboard (for framebuffer mode)
216 x) make OpenBSD/macppc work (PCI controller stuff)
217
218 Algor:
219 PCI interrupts... needed or stuff like the tlp NIC?
220
221 ARM:
222 o) try to get netbsd/evbarm 3.x running (iq80321)
223 o) make the xscale counter registers (ccnt) work
224 o) make the ata controller usable for FreeBSD!
225 o) zaurus for openbsd...
226 o) debian/cats crashes because of unimplemented coproc stuff.
227 fix this?
228
229 Cache simulation:
230 o) Command line flags for:
231 o) CPU endianness?
232 o) Cache sizes? (multiple levels)
233 o) Separate from the CPU concept, so that multi-core CPUs sharing
234 e.g. a L2 cache can be simulated (?)
235 o) Instruction cache emulation is easiest (if separate from the
236 data cache); similar hack as the S;I; hack in cpu_dyntrans.c.
237 NOTE: if the architecture has a delay slot, then an instruction
238 slot can actually be executed as 2 instructions.
239 o) Data cache emulation = harder; each arch's load/store routines
240 must include support? running one instruction at a time and
241 having a cpu-dependant lookup function for each instruction
242 is another option (easier to implement, but very very slow).
243
244 Documentation:
245 o) machines, cpus, devices.
246 o) Automagic documentation generation:
247 x) REMEMBER that several machines/devices can be in
248 the same source file!
249 o) Try to rewrite the install instructions for those machines
250 that use 3MAX into using CATS? (To remove the need to a raw
251 ffs partition using up all of the disk image.)
252
253 More generic out_of_memory error reporting, and check everywhere!
254 Causes: OpenBSD has low default limits for normal users.
255 Host is 32-bit? (32-bit hosts are limited to 4 GB or less
256 of userspace memory.)
257 You are actually low on RAM. (As trivial as this might sound,
258 Unix systems usually allow processes to allocate virtual
259 memory beyond the amount of RAM in the machine.)
260
261 Breakpoints: 32-bit vs 64-bit sign extension for MIPS, warnings, etc.
262 Use the debugger's symbolic name stuff. (which will have to be
263 extended soon to support stuff like "2*x + symbol + y" etc. cool
264 stuff)
265
266 The Device subsystem:
267 x) allow devices to be moved and/or changed in size (down to a
268 minimum size, etc, or up to a max size)
269 x) keep track of interrupts and busses? actually, allowing any device
270 to be a bus might be a nice idea.
271 x) turn interrupt controllers into devices? :-)
272 x) refactor various clocks/nvram/cmos into one device?
273
274 Clocks:
275 x) General framework for automagic clock adjustment for _all_
276 kinds of clocks and timers. (Which should be possible to turn
277 off, of course, like the way DECstation emulation works now.)
278
279 PCI:
280 x) last write was ffffffff ==> fix this, it should be used
281 together with a mask to get the correct bits. also, not ALL
282 bits are size bits! (lowest 4 vs lowest 2?)
283 x) add support for address fixups
284 x) generalize the interrupt routing stuff (lines etc). this should
285 be per machine? or per bus, that's better
286 x) add a "pcn" NIC (AMD PCnet32 Lance 79c970 (PCI 1022:2000)),
287 could be useful for several machine modes (Malta, Algor, evbarm,
288 hp700?, macppc, etc.)
289
290 Network layer:
291 o) DHCP (for Debian and BSD installers :-)
292 o) increase performance
293 o) don't rely on NetBSD-ish usage
294 o) Multiple networks per emulation, and let different
295 NICs in machines connect to different networks.
296 o) many other issues: see src/net.c
297
298 Busses:
299 o) Redesign the entire "mainbus" concept!
300 o) Busses should be placed in a hierarchical tree!
301 o) Easily configurable interrupt routing in SMP systems.
302 o) Specific clock/bus speeds, cpu speeds etc.
303 o) Synchronization over network? or at least in dyntrans within
304 one emulated machine
305 o) dev->bus: TurboChannel, PCMCIA, ADB?
306
307 Config file parser:
308 o) Rewrite it from scratch!
309 o) Usage of any expression available through the debugger
310 o) Support for running debugger commands (like the -c
311 command line option)
312
313 Floating point layer:
314 o) make it common enough to be used by _all_ emulation modes
315 o) implement more stuff
316 o) non-IEEE modes (i.e. x86)?
317
318 Userland emulation:
319 x) Lots of stuff; freebsd and netbsd (and linux?) syscalls.
320 x) Dynamic linking? Hm.
321
322 Sound:
323 x) generic sound framework
324 x) add one or more sound cards as devices
325
326 ASC SCSI controller:
327 x) NetBSD/arc 2.0 uses the ASC controller in a way which GXemul
328 cannot yet handle. (NetBSD 1.6.2 works ok.) (Possibly a problem
329 in NetBSD itself, http://mail-index.netbsd.org/source-changes/
330 2005/11/06/0024.html suggests that.)
331
332 Caches / memory hierarchies: (this is mostly MIPS-specific)
333 o) src/memory*.c: Implement correct cache emulation for
334 all CPU types. (currently only R2000/R3000 is implemented)
335 (per CPU, multiple levels should be possible, associativity etc!)
336 o) R2000/R3000 isn't _100%_ correct, just almost correct :)
337 o) Move the -S (fill mem with random) functionality into the
338 memory.c subsystem, not machine.c or wherever it is now
339 o) ECC stuff, simulation of memory errors? (Machine dependent)
340 o) More than 4GB of emulated RAM, when run on a 32-bit host?
341 (using manual swap-out of blocks to disk, ugly)
342 o) A global command line option should be used to turn
343 cache emulation on or off. When off, caches should be
344 faked like they are right now. When on, caches and
345 memory latencies should be emulated as correctly as
346 possible.
347
348 File/disk/symbol handling:
349 o) Remove some of the complexity in file format guessing, for
350 Ultrix kernels that are actually disk images?
351 o) Better handling of tape files
352 o) Read function argument count and types from binaries? (ELF?)
353 o) Better demangling of C++ names. Note: GNU's C++ differs from e.g.
354 Microsoft's C++, so multiple schemes must be possible. See
355 URL at top of src/symbol_demangle.c for more info.
356
357 Userland ABI emulation:
358 o) see src/useremul.c
359
360 Terminal/console:
361 o) allow emulated serial ports to be connected to the outside
362 world in a more generic way, or even to other emulated
363 machines(?)
364
365 Save state of the whole emulated machine, to be able to load it back
366 in later? (Memory, all device's states, all registers and
367 so on. Like taking a snapshot. (SimOS seems to do this,
368 according to its website.))
369
370 Better framebuffer and X-windows functionality:
371 o) -Yx sometimes causes crashes.
372 o) Simple device access to framebuffer_blockcopyfill() etc,
373 and text output (using the built-in fonts), for dev_fb.
374 o) CLEAN UP the ugly event code
375 o) Mouse clicks can be "missed" in the current system; this is
376 not good. They should be put on a stack of some kind.
377 o) More 2D and 3D framebuffer acceleration.
378 o) Non-resizable windows? Or choose scaledown depending
379 on size (and center the image, with a black border).
380 o) Different scaledown on different windows?
381 o) Switch scaledown during runtime? (Ala CTRL-ALT-plus/minus)
382 o) Bug reported by Elijah Rutschman on MacOS with weird
383 keys (F5 = cursor down?).
384 o) Keyboard and mouse events:
385 x) Do this for more machines than just DECstation
386 x) more X11 cursor keycodes
387 x) Keys like CTRL, ALT, SHIFT do not get through
388 by themselves (these are necessary for example
389 to change the font of an xterm in X in the
390 emulator)
391 o) Generalize the framebuffer stuff by moving _ALL_ X11
392 specific code to src/x11.c!
393
394 Statistics: (this could be interesting)
395 o) Save to file and show graphics. It should be possible to
396 run gxemul after a simulation to just show the graphics,
397 or convert to a .ppm or .tga or similar.
398 o) memory accesses (to measure cache efficiency and
399 page coloring efficiency)
400 o) nr of simultaneous ASIDs in use in the TLB, for MIPS
401 o) percentage of time spent in different "states", such as
402 running userland code, kernel code, or idling (for CPUs
403 that have such an instruction, or whenever the PC is
404 inside a specific idle-function (address range)).
405 Possible additional state (for example on R3000): caches
406 disabled.
407 o) position of read/write on (SCSI) disks
408

  ViewVC Help
Powered by ViewVC 1.1.26