1 |
$Id: TODO,v 1.166 2005/08/16 05:44:33 debug Exp $ |
$Id: TODO,v 1.410 2006/11/06 05:32:38 debug Exp $ |
2 |
|
|
3 |
=============================================================================== |
This file is my list of things I want to work on in the future. It is in |
4 |
|
random order, and some parts of it are probably out-to-date by now. |
5 |
|
|
|
High priority stuff: |
|
6 |
|
|
7 |
|
Dyntrans: |
8 |
MIPS bintrans: |
x) Instruction combination collisions? How to avoid easily... |
9 |
x) call/return address cache? |
x) Think about how to do both SHmedia and SHcompact in a reasonable |
10 |
|
way! (Or AMD64 long/protected/real, for that matter.) |
11 |
dyntrans: |
x) 68K emulation; think about how to do variable instruction |
12 |
x) memory write protection for ARM, but NOT for Alpha (because |
lengths across page boundaries. |
13 |
it has the IMB instruction... hm) |
x) Dyntrans with valgrind-inspired memory checker. (In memory_rw, |
14 |
|
it would be reasonably simple to add; in each individual fast |
15 |
x) call/return address cache |
load/store routine = a lot more work, and it would become |
16 |
|
kludgy very fast.) |
17 |
x) instr_call sequence analysis support? (Useful for |
x) Dyntrans with SMP... lots of work to be done here. |
18 |
handtuning combinations.) |
x) Dyntrans with cache emulation... lots of work here as well. |
19 |
|
o) dev_mp doesn't work well with dyntrans yet |
20 |
x) opcode statistics support? |
o) In general, IPIs, CAS, LL/SC etc must be made to work with dyntrans |
21 |
TODO: is instr_call statistics enough? |
x) Redesign/rethink the delay slot mechanism used for e.g. MIPS, |
22 |
|
so that it caches a translation (that is, an instruction |
23 |
x) support for archs that allow transparent unaligned load/stores |
word and the instr_call it was translated to the last |
24 |
|
time), so that it doesn't need to do slow |
25 |
x) SMP: detect when an instruction such as ll/sc or cas is used, |
to_be_translated for each end of page? |
26 |
and "synchronize" approximately the number of executed instructions |
x) Program Counter statistics: |
27 |
(or cycles) across all CPUs. |
Per machine? What about SMP? All data to the same file? |
28 |
|
A debugger command should be possible to use to enable/ |
29 |
x) support for variable-length instructions (x86, m68k, ...) |
disable statistics gathering. |
30 |
Perhaps: don't increase the next_ic between every |
Configuration file option! |
31 |
instruction, but let each instruction's handler do |
x) Breakpoints: |
32 |
that for itself. |
o) Physical vs virtual addresses! |
33 |
Problem: what about instructions crossing a (virtual) |
o) 32-bit vs 64-bit sign extension for MIPS, and others? |
34 |
page boundary? They cannot be translated once |
x) INVALIDATION should cause translations in _all_ cpus to be |
35 |
and for all :( and must be interpreted slowly! |
invalidated, e.g. on a write to a write-protected page |
36 |
|
(containing code) |
37 |
x) support for THUMB or MIPS16 (arm, mips) |
x) 16-bit encodings? (MIPS16, ARM Thumb, 32-bit SH on SH64) |
38 |
|
x) Lots of other stuff: see src/cpus/README_DYNTRANS |
39 |
x) support for Delay slots! (mips, sparc, hppa) |
x) true recompilation backend? think carefully about this, |
40 |
|
experiment in a separate project (not in GXemul) |
41 |
x) Alpha: hahaha, zapnot and inserts/extracts don't |
o) First test would be to just implement a simple |
42 |
compile into very nice code :-| fix this |
instruction such as MIPS' addiu or lui, on AMD64 |
43 |
|
hosts... |
44 |
x) 64-bit virtual memory translation tables (PPC, etc) |
|
45 |
|
Simple Valgrind-like checks? |
46 |
x) x86: convert to dyntrans. LOTS of stuff to consider. |
o) Mark every address with bits which tell whether or not the address |
47 |
|
has been written to. |
48 |
|
o) What should happen when programs are loaded? Text/data, bss (zero |
49 |
=============================================================================== |
filled). But stack space and heap is uninitialized. |
50 |
|
o) Uninitialized local variables: |
51 |
Lower priority, but still important: |
A load from a place on the stack which has not previously |
52 |
|
been stored to => warning. Increasing the stack pointer using |
53 |
Redesign the entire "mainbus" concept: |
any available means should reset the memory to uninitialized. |
54 |
o) Easily configurable interrupt routing in SMP systems. |
o) If calls to malloc() and free() can be intercepted: |
55 |
o) Specific clock/bus speeds, cpu speeds etc. |
o) Access to a memory area after free() => warning. |
56 |
o) Synchronization over network? |
o) Memory returned by malloc() is marked as not-initialized. |
57 |
|
o) Non-passive, but good to have: Change the argument |
58 |
Caches / memory hierarchies: (this is mostly MIPS-specific) |
given to malloc, to return a slightly larger memory |
59 |
o) MIPS coproc.c: bits in config registers should reflect |
area, i.e. margin_before + size + margin_after, |
60 |
correct cache sizes for _all_ CPU types. (currently only |
and return the pointer + margin_before. |
61 |
implemented for R4000, R1x000, and a few others) |
Any access to the margin_before or _after space results |
62 |
o) src/memory*.c: Implement correct cache emulation for |
in warnings. (free() must be modified to free the |
63 |
all CPU types. (currently only R2000/R3000 is implemented) |
actually allocated address.) |
64 |
(per CPU, multiple levels should be possible, |
|
65 |
associativity etc!) |
MIPS: |
66 |
o) R2000/R3000 isn't _100%_ correct, just almost correct :) |
o) Nicer MIPS status bits in register dumps. |
67 |
o) Move the -S (fill mem with random) functionality into the |
o) Alignment exceptions. |
68 |
memory.c subsystem, not machine.c or wherever it is now |
o) Floating point exception correctness. |
69 |
o) ECC stuff, simulation of memory errors? (Machine dependant) |
o) Fix this? Triggered by NetBSD/sgimips? Hm: |
70 |
o) More than 4GB of emulated RAM, when run on a 32-bit host? |
to_be_translated(): TODO: unimplemented instruction: |
71 |
(using manual swap-out of blocks to disk, ugly) |
000000000065102c: 00200800 (d) rot_00 at,zr,0 |
72 |
o) A global command line option should be used to turn |
o) Some more work on opcodes. |
73 |
cache emulation on or off. When off, caches should be |
x) MIPS64 revision 2. |
74 |
faked like they are right now. When on, caches and |
o) Find out which actual CPUs implement the rev2 ISA! |
75 |
memory latencies should be emulated as correctly as |
o) DROTR32 and similar MIPS64 rev 2 instructions, |
76 |
possible. |
which have a rotation bit which differs from |
77 |
|
previous ISAs. |
78 |
Network layer: |
o) EI and DI instructions for MIPS64/32 rev 2. |
79 |
o) Multiple networks per emulation, and let different |
NOTE: These are _NOT_ the same as for R5900! |
80 |
NICs in machines connect to different networks. |
x) _MAYBE_ TX79 and R5900 actually differ in their |
81 |
o) many other issues: see src/net.c |
opcodes? Check this carefully! |
82 |
|
o) Dyntrans: Count register updates are probably not 100% correct yet. |
83 |
MIPS CPU emulation: |
o) Refactor code for performance and readability/maintainability. |
84 |
o) i386 bintrans backend: movn etc, slt[u] for |
o) (Re)implement 128-bit loads/stores for R5900. |
85 |
64-bit mode, 64-bit shifts etc |
o) R4000 and others: |
86 |
o) Instructions: |
x) watchhi/watchlo exceptions, and other exception |
87 |
o) All ISAs: |
handling details |
88 |
o) Floating point exception handling, and |
o) R10000 and others: (R12000, R14000 ?) |
89 |
add more instructions. |
x) memory space, exceptions, ... |
90 |
o) Finish the MIPS16 translator, and test it! |
x) use cop0 framemask for tlb lookups |
91 |
o) MIPS ISA I, II, III, IV |
(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) |
92 |
o) MIPS V (SIMD vector stuff?) |
|
93 |
o) MDMX (MIPS Digital Media Extension) |
SuperH: |
94 |
o) MIPS 3D |
x) DMA (0xffa00000) |
95 |
o) MIPS MT (Multi-thread stuff) (What's this?) |
x) Instruction tracing should include symbols for branch targets, |
96 |
o) Warn about mis-used bit fields (ie bits that |
and so on... |
97 |
should be all zeroes, warn about if they are not)! |
x) SH4 interrupt controller: |
98 |
Both for coprocessor registers and for instruction |
x) Implement correct priorities of interrupts |
99 |
opcodes. |
x) SH4 BSC (Bus State Controller) |
100 |
o) the special2 stuff is a mess right now |
x) NetBSD/evbsh3, dreamcast, mmeye, hpcsh! Linux? |
101 |
o) warn and/or cause exceptions for unimplemented |
x) Replace pc-relative loads with immediate load, if within the |
102 |
instructions (depending on CPU type) |
same page. (Similar to the same optimization for ARM.) |
103 |
o) R2000/R3000: |
x) Floating point exception correctness. |
104 |
x) R3000 "tri-byte stores". (What's this?) |
x) Floating point speed! |
105 |
o) R4000 and others: |
x) Think carefully about how to implement SH5/SH64 (for evbsh5). |
106 |
x) watchhi/watchlo exceptions, and other exception |
|
107 |
handling details |
Dreamcast: |
108 |
o) R10000 and others: (R12000, R14000 ?) |
x) CD image bootup: |
109 |
x) memory space, exceptions, ... |
0) Find IP.BIN, and load it to 0x8c008000. |
110 |
x) use cop0 framemask for tlb lookups |
1) Run code at 0x8c008300 (SEGA license code). |
111 |
(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) |
2) When the license code runs a "boot menu" syscall, |
112 |
o) Implement load delays? Warnings on interlocks. |
load the 1ST_READ.BIN file (unscrambled?) to 0x8c010000. |
113 |
o) Implement all coprocessor 0 bits / functions. |
3) Run code at 0x8c00b800 (Bootstrap 1). This will in turn |
114 |
x) coproc 0 selectors! (R4000 ?) |
jump to 0x8c00e000 (Bootstrap 2), and then jump to |
115 |
o) R4300 (nintendo64, no mmu?), R5900 (playstation2, weird |
0x8c010000, to start the program. |
116 |
TLB/cache? 128-bit GPRs, new instructions), |
(Try with e.g. Comstedt's Serial IP Slave, to make sure it |
117 |
4K (note: NOT R4000), 5K (note: NOT R5000), |
works as expected.) |
118 |
R6000 (ISA II), R8000 |
x) LAN adapter. |
119 |
o) Multi-cpu stuff: |
x) PVR: Lots of stuff. See dev_pvr.c. |
120 |
+) Interrupt routing (ie devices vs mainbus, or |
x) Maple bus: |
121 |
connect each device to a fixed cpu) |
x) Correct controller input |
122 |
+) SGI's NUMA architecture. Study |
x) Mouse input |
123 |
x) Linux sources |
x) PROM/BIOS calls: |
124 |
x) SGI's specs on NUMA address space |
x) GD-ROM emulation |
125 |
+) Ultrix? NetBSD doesn't do SMP on MIPS yet :-( |
x) NetBSD/dreamcast: Root on nfs? |
126 |
+) Own experiments with ycx2. |
x) Linux/dreamcast? (The gentoo kernel currently crashes.) |
127 |
|
x) More homebrew demos/games. |
128 |
File/disk handling: |
x) Sound emulation (ARM cpu). |
129 |
o) Better handling of tape files |
x) VME processor emulation? |
130 |
|
|
131 |
Debugger: |
Transputer: |
132 |
o) Read function argument count and types from binaries? (ELF?) |
x) Implement support for Helios binaries. |
133 |
o) Demangle C++ names. |
x) Stack and register contents at startup? |
134 |
|
x) Figure out how to boot an entire Helios distribution. |
135 |
Userland ABI emulation: |
x) Implement all instructions. :) |
136 |
o) see src/useremul.c |
|
137 |
|
RCA1802/RCA1805, CHIP8: |
138 |
Terminal based interactive debugger: |
x) CHIP8 -> RCA180x conversion |
139 |
o) see src/debugger.c |
x) Think about how to do dual-mode, variable-instr-length |
140 |
|
ISAs, and switch between modes. |
141 |
Terminal/console stuff: |
x) 1805 "extended" opcode -> trigger CHIP8 emulation? |
142 |
o) allow emulated serial ports to be connected to the outside |
That is, all calls 0NNN could point to 0x68 opcodes, |
143 |
world in a more generic way, or even to other emulated |
which, if running on a 1802 in CHIP8-emulation-mode, |
144 |
machines(!) |
would be manually interpreted. |
145 |
|
x) Better solution: |
146 |
Regression tests. (Needs to be totally rewritten, the old framework |
CHIP8 calls to 00xx => handle at high level, |
147 |
was removed because it was useless.) |
calls to 0xxx in general = call 180X machine code |
148 |
|
(0000 = reboot?) |
149 |
Save state of the whole emulated machine, to be able to load it back |
x) 1802 info: http://www.nyx.net/~lturner/public_html/Cosmac.html |
150 |
in later? (Memory, all device's states, all registers and |
and: http://www.elf-emulation.com/1802.html |
151 |
so on. Like taking a snapshot. (SimOS seems to do this, |
x) 1805 extended opcodes: Implement at least disassembly support! |
152 |
according to its website.)) |
x) Keyboard input. |
153 |
|
x) Sound (beep only). |
154 |
Better X-windows functionality: |
x) Slow-down to correct speed? Wikipedia: "it was usually operated |
155 |
o) CLEAN UP the ugly event code |
at 3.58 MHz/2 to suit the requirements of the 1861 chip which |
156 |
o) Mouse clicks can be "missed" in the current system; this is |
gave a speed of a little over 100,000 instructions per second" |
157 |
not good. They should be put on a stack of some kind. |
(Note that _CHIP8_ emulation would then be even slower.) |
158 |
o) More 2D and 3D framebuffer acceleration. |
x) SCHIP48 (Super) emulation: |
159 |
o) Non-resizable windows? Or choose scaledown depending |
Some more opcodes, 128x64 framebuffer, larger |
160 |
on size (and center the image, with a black border). |
sprites and fonts. |
161 |
o) Different scaledown on different windows? |
|
162 |
o) Switch scaledown during runtime? (Ala CTRL-ALT-plus/minus) |
Alpha: |
163 |
o) Keyboard and mouse events: |
x) OSF1 PALcode, Virtual memory support. |
164 |
x) Do this for more machines than just DECstation |
x) PALcode replacement! PAL1E etc opcodes...? |
165 |
x) more X11 cursor keycodes |
x) Interrupt/exception/trap handling. |
166 |
x) Keys like CTRL, ALT, SHIFT do not get through |
x) Floating point exception correctness. |
167 |
by themselves (these are necessary for example |
x) More work on bootup memory and register contents. |
168 |
to change the font of an xterm in X in the |
x) More Alpha machine types, so it could work with |
169 |
emulator) |
OpenBSD, FreeBSD, and Linux too? |
170 |
o) Generalize the framebuffer stuff by moving _ALL_ X11 |
|
171 |
specific code to src/x11.c! |
SPARC: |
172 |
|
o) Implement Adress space identifiers; load/stores etc. |
173 |
Statistics: (this could be interesting) |
o) Save/restore register windows etc! |
174 |
o) Save to file and show graphics. It should be possible to |
o) Finish the subcc and addcc flag computation code. |
175 |
run gxemul after a simulation to just show the graphics, |
o) Add more registers (floating point, control regs etc) |
176 |
or convert to a .ppm or .tga or similar. |
o) Exception/trap handling. |
177 |
o) memory accesses (to measure cache efficiency and |
o) Disassemly of some more instructions? |
178 |
page coloring efficiency) |
o) Are sll etc 32-bit sign-extending or zero-extending? |
179 |
o) nr of simultaneous ASIDs in use in the TLB, for MIPS |
o) Finish the GDB register stuff. |
180 |
o) percentage of time spent in different "states", such as |
x) Floating point exception correctness. |
181 |
running userland code, kernel code, or idling (for CPUs |
o) SPARC v8, v7 etc? |
182 |
that have such an instruction, or whenever the PC is |
|
183 |
inside a specific idle-function (address range)). |
Debugger: |
184 |
Possible additional state (for example on R3000): caches |
o) How does SMP debugging work? Does it simply use "threads"? |
185 |
disabled. |
What if the guest OS (running on an emulated SMP machine) |
186 |
o) position of read/write on (SCSI) disks |
has a usertask running, with userland threads? |
187 |
|
o) Try to make the debugger more modular and, if possible, reentrant! |
188 |
|
o) Remove the emul command? (But show network info if showing |
189 |
|
machines?) |
190 |
|
o) Settings: |
191 |
|
x) Special handlers for Write! |
192 |
|
+) MIPS coproc regs |
193 |
|
+) Alpha/MIPS/SPARC zero registers |
194 |
|
+) x86 64/32/16-bit registers |
195 |
|
x) Value formatter for resulting output. |
196 |
|
o) see src/debugger.c for more |
197 |
|
|
198 |
|
POWER/PowerPC: |
199 |
|
x) find and fix the bug which causes NetBSD/macppc to fail after |
200 |
|
an install! |
201 |
|
x) NetBSD/prep 3.x triggers a possible bug in the emulator: |
202 |
|
<wdc_exec_command(0xd005e514,0xd60cdd30,0,8,..)> |
203 |
|
<ata_get_xfer(0,0xd60cdd30,0,8,..)> |
204 |
|
<0x26c550(&ata_xfer_pool,2,0,8,..)> |
205 |
|
<0x35c71c(0x3f27000,0,52,8,..)> |
206 |
|
<ata_exec_xfer(0xd005e4c8,0x3f27000,0,13,..)> |
207 |
|
<atastart(0xd005e4c8,0x3f27000,0,13,..)> |
208 |
|
<__wdccommand_start(0xd005e4c8,0x3f27000,0,13,..)> |
209 |
|
<bsw1(&prep_isa_io_space_tag,0x800001f6,0,176,..)> |
210 |
|
[ wdc: write to SDH: 0xb0 (sectorsize 2, lba=1, drive 1, head 0) ] |
211 |
|
<wdcwait(0xd005e4c8,72,64,0xbb8,..)> |
212 |
|
<0x198120(0xd005e4c8,72,64,0xbb8,..)> |
213 |
|
<bsr1(&prep_isa_io_space_tag,0,0,0xbb8,..)> |
214 |
|
<delay(100,0,0,0xbb8,..)> |
215 |
|
Note: <bsr1(&prep_isa_io_space_tag,0,0,0xbb8,..)> |
216 |
|
x) PPC optimizations; instr combs |
217 |
|
x) 64-bit stuff: either Linux on G5, or perhaps some hobbyist |
218 |
|
version of AIX? (if there exists such a thing) |
219 |
|
x) macppc: adb controller; keyboard (for framebuffer mode) |
220 |
|
x) make OpenBSD/macppc work (PCI controller stuff) |
221 |
|
x) Floating point exception correctness. |
222 |
|
x) Alignment exceptions. |
223 |
|
|
224 |
|
Algor: |
225 |
|
o) Other models than the P5064? |
226 |
|
o) PCI interrupts... needed for stuff like the tlp NIC? |
227 |
|
|
228 |
|
HPCmips: |
229 |
|
x) Mouse/pad support! :) |
230 |
|
x) A NIC? (As a PCMCIA device?) |
231 |
|
|
232 |
|
AVR: |
233 |
|
o) Everything. |
234 |
|
|
235 |
|
AVR32: |
236 |
|
o) Everything. It would be good if there was NetBSD/avr32 to |
237 |
|
experiment with... |
238 |
|
|
239 |
|
ARM: |
240 |
|
o) See netwinder_reset() in NetBSD; the current "an internal error |
241 |
|
occured" message after reboot/halt is too ugly. |
242 |
|
o) ARM "wait"-like instruction? |
243 |
|
o) try to get netbsd/evbarm 3.x running (iq80321) |
244 |
|
o) make the xscale counter registers (ccnt) work |
245 |
|
o) make the ata controller usable for FreeBSD! |
246 |
|
o) zaurus for openbsd... |
247 |
|
o) debian/cats crashes because of unimplemented coproc stuff. |
248 |
|
fix this? |
249 |
|
|
250 |
|
Test machines: |
251 |
|
+ dev_fb block fill and copy |
252 |
|
+ dev_fb draw characters (from the built-in font)? |
253 |
|
+ dev_fb input device? mouse pointer coordinates and buttons |
254 |
|
(allow changes in these to cause interrupts as well?) |
255 |
|
+ Redefine the halt() function so that it stops "sometimes |
256 |
|
soon", i.e. usage in demo code should be: |
257 |
|
for (;;) { |
258 |
|
halt(); |
259 |
|
} |
260 |
|
|
261 |
|
Better CD Image file support: |
262 |
|
x) Support CD formats that contain more than 1 track, e.g. |
263 |
|
CDI files (?). These can then contain a mixture of e.g. sound |
264 |
|
and data tracks, and booting from an ISO filesystem path |
265 |
|
would boot from [by default] the first data track. |
266 |
|
(This would make sense for e.g. Dreamcast CD images, or |
267 |
|
possibly other live-CD formats.) |
268 |
|
|
269 |
|
Networking: |
270 |
|
x) Fix performance problems caused by only allowing a |
271 |
|
single TCP packet to be unacked. |
272 |
|
x) Don't hardcode offsets into packets! |
273 |
|
x) Test with lower than 100 max tcp/udp connections, |
274 |
|
to make sure that reuse works! |
275 |
|
x) Make OpenBSD work better as a guest OS! |
276 |
|
x) DHCP? Debian doesn't actually send DHCP packets, even |
277 |
|
though it claims to? So it is hard to test. |
278 |
|
x) Multiple networks per emulation, and let different |
279 |
|
NICs in machines connect to different networks. |
280 |
|
x) Support VDE (vde.sf.net)? Easiest/cleanest (before a |
281 |
|
redesign of the network framework has been done) is |
282 |
|
probably to connect it using the current (udp) solution. |
283 |
|
x) Allow SLIP connections, possibly PPP, in addition to |
284 |
|
ethernet? |
285 |
|
|
286 |
|
Cache simulation: |
287 |
|
o) Command line flags for: |
288 |
|
o) CPU endianness? |
289 |
|
o) Cache sizes? (multiple levels) |
290 |
|
o) Separate from the CPU concept, so that multi-core CPUs sharing |
291 |
|
e.g. a L2 cache can be simulated (?) |
292 |
|
o) Instruction cache emulation is easiest (if separate from the |
293 |
|
data cache); similar hack as the S;I; hack in cpu_dyntrans.c. |
294 |
|
NOTE: if the architecture has a delay slot, then an instruction |
295 |
|
slot can actually be executed as 2 instructions. |
296 |
|
o) Data cache emulation = harder; each arch's load/store routines |
297 |
|
must include support? running one instruction at a time and |
298 |
|
having a cpu-dependant lookup function for each instruction |
299 |
|
is another option (easier to implement, but very very slow). |
300 |
|
|
301 |
|
Documentation: |
302 |
|
x) Note about sandboxing/security: |
303 |
|
Not all emulated instructions fail in the way they would |
304 |
|
do on real hardware (e.g. a userspace program writing to |
305 |
|
a system register might work in GXemul, but it would |
306 |
|
fail on real hardware). Sandbox = contain from the |
307 |
|
host OS. But the emulated programs will run "less |
308 |
|
securely". |
309 |
|
x) Try NetBSD/arc 4.x! (It seems to work with disk images!) |
310 |
|
x) NetBSD/pmax 4 install instructions: xterm instead of vt100! |
311 |
|
x) DEVICE_TICK in technical.html |
312 |
|
x) Rewrite the section about experimental devices, after the |
313 |
|
framebuffer acceleration has been implemented, and demos |
314 |
|
written. (Symbolic names instead of numbers; example |
315 |
|
use cases, etc. Mention demo files that use the various |
316 |
|
features?) |
317 |
|
x) "a very simple linear framebuffer device (for graphics output)" |
318 |
|
under "which machines does gxemul emulate" ==> better |
319 |
|
description? |
320 |
|
x) Better description on how to set up a cross compiler? |
321 |
|
Example for MIPS64. |
322 |
|
o) Automagic documentation generation? |
323 |
|
x) machines, cpus, devices. |
324 |
|
x) REMEMBER that several machines/devices can be in |
325 |
|
the same source file! |
326 |
|
o) Try to rewrite the install instructions for those machines |
327 |
|
that use 3MAX into using CATS or hpcmips? (To remove the need |
328 |
|
to use a raw ffs partition, using up all of the disk image.) |
329 |
|
|
330 |
|
More generic out_of_memory error reporting, and check everywhere! |
331 |
|
Causes: OpenBSD has low default limits for normal users. |
332 |
|
Host is 32-bit? (32-bit hosts are limited to 4 GB or less |
333 |
|
of userspace memory.) |
334 |
|
You are actually low on RAM. (As trivial as this might sound, |
335 |
|
Unix systems usually allow processes to allocate virtual |
336 |
|
memory beyond the amount of RAM in the machine.) |
337 |
|
|
338 |
|
The Device subsystem: |
339 |
|
x) allow devices to be moved and/or changed in size (down to a |
340 |
|
minimum size, etc, or up to a max size); if there is a collision, |
341 |
|
return false. It is up to the caller to handle this situation! |
342 |
|
x) NOTE: Translations must be invalidated, both for |
343 |
|
registering new devices, and for moving existing ones. |
344 |
|
cpu->invalidate translation caches, for all CPUs that |
345 |
|
are connected to a specific memory. |
346 |
|
x) keep track of interrupts and busses? actually, allowing any device |
347 |
|
to be a bus might be a nice idea. |
348 |
|
x) turn interrupt controllers into devices? :-) |
349 |
|
x) refactor various clocks/nvram/cmos into one device? |
350 |
|
|
351 |
|
PCI: |
352 |
|
x) last write was ffffffff ==> fix this, it should be used |
353 |
|
together with a mask to get the correct bits. also, not ALL |
354 |
|
bits are size bits! (lowest 4 vs lowest 2?) |
355 |
|
x) add support for address fixups |
356 |
|
x) generalize the interrupt routing stuff (lines etc) |
357 |
|
|
358 |
|
Clocks and timers: |
359 |
|
x) DON'T HARDCODE 100 HZ IN cpu_mips_coproc.c! |
360 |
|
x) Test the 8253? Right now it doesn't seem to be used? |
361 |
|
x) NetWinder timeofday is incorrect! |
362 |
|
x) Cobalt TOD is incorrect! |
363 |
|
x) Go through all other machines, one by one, and fix them. |
364 |
|
|
365 |
|
Busses: |
366 |
|
o) Redesign the entire "mainbus" concept! |
367 |
|
x) Busses should be placed in a hierarchical tree (?) |
368 |
|
x) Specific clock/bus speeds, cpu speeds etc. |
369 |
|
o) Interrupt routing subsystem: |
370 |
|
x) IF POSSIBLE, try to make the new system work with the |
371 |
|
current system, but print annoying warning messages. :) |
372 |
|
Think carefully about this. |
373 |
|
x) Registry for all available interrupts. |
374 |
|
+) Each interrupt controller (including CPU cores |
375 |
|
that can handle interrupts) should register its |
376 |
|
interrupts, e.g. |
377 |
|
cpu[0].irq[3] |
378 |
|
cpu[0].irq[3].pcmcia_slot[1] |
379 |
|
cpu[0].irq.pci[3] |
380 |
|
+) Note: MIPS cpus have multiple irqs in the core, |
381 |
|
while some other CPUs only have one (irq[0] |
382 |
|
or just irq). |
383 |
|
x) Users should use interrupt _names_ instead of integers |
384 |
|
when attaching to an interrupt controller, but when |
385 |
|
asserting/deasserting irq lines, small integers must |
386 |
|
still be used (for obvious performance reasons). |
387 |
|
Figure out a way to do this nicely! |
388 |
|
x) Any users need to say whether they need the interrupt line |
389 |
|
exclusively or allow shared access. |
390 |
|
x) Must work with everything from native IRQs to |
391 |
|
TurboChannel/PCI/ISA/ADB/PCMCIA/... |
392 |
|
x) Must work with SMP emulation! |
393 |
|
x) Make it with device_add(). How does the end user find |
394 |
|
out the name of an interrupt controller/line in e.g. |
395 |
|
a configuration file? |
396 |
|
o) Synchronization over network? or at least in dyntrans within |
397 |
|
one emulated machine |
398 |
|
o) Convert to real busses: TurboChannel, PCMCIA, ADB |
399 |
|
|
400 |
|
Config file parser: |
401 |
|
o) Rewrite it from scratch! |
402 |
|
o) Usage of any expression available through the debugger |
403 |
|
o) Support for running debugger commands (like the -c |
404 |
|
command line option) |
405 |
|
|
406 |
|
Floating point layer: |
407 |
|
o) make it common enough to be used by _all_ emulation modes |
408 |
|
o) implement correct error/exception handling and rounding modes |
409 |
|
o) implement more helper functions (i.e. add, sub, mul...) |
410 |
|
o) non-IEEE modes (i.e. x86)? |
411 |
|
|
412 |
|
Userland emulation: |
413 |
|
x) Lots of stuff; freebsd and netbsd (and linux?) syscalls. |
414 |
|
x) Dynamic linking? Hm. |
415 |
|
|
416 |
|
Sound: |
417 |
|
x) generic sound framework |
418 |
|
x) add one or more sound cards as devices; add a testmachine |
419 |
|
sound card first? |
420 |
|
|
421 |
|
ASC SCSI controller: |
422 |
|
x) NetBSD/arc 2.0 uses the ASC controller in a way which GXemul |
423 |
|
cannot yet handle. (NetBSD 1.6.2 works ok.) (Possibly a problem |
424 |
|
in NetBSD itself, http://mail-index.netbsd.org/source-changes/ |
425 |
|
2005/11/06/0024.html suggests that.) |
426 |
|
NetBSD 4.x seems to work? :) |
427 |
|
|
428 |
|
Caches / memory hierarchies: (this is mostly MIPS-specific) |
429 |
|
o) src/memory*.c: Implement correct cache emulation for |
430 |
|
all CPU types. (currently only R2000/R3000 is implemented) |
431 |
|
(per CPU, multiple levels should be possible, associativity etc!) |
432 |
|
o) R2000/R3000 isn't _100%_ correct, just almost correct :) |
433 |
|
o) Move the -S (fill mem with random) functionality into the |
434 |
|
memory.c subsystem, not machine.c or wherever it is now |
435 |
|
o) ECC stuff, simulation of memory errors? (Machine dependent) |
436 |
|
o) More than 4GB of emulated RAM, when run on a 32-bit host? |
437 |
|
(using manual swap-out of blocks to disk, ugly) |
438 |
|
o) A global command line option should be used to turn |
439 |
|
cache emulation on or off. When off, caches should be |
440 |
|
faked like they are right now. When on, caches and |
441 |
|
memory latencies should be emulated as correctly as |
442 |
|
possible. |
443 |
|
|
444 |
|
File/disk/symbol handling: |
445 |
|
o) Remove some of the complexity in file format guessing, for |
446 |
|
Ultrix kernels that are actually disk images? |
447 |
|
o) Better handling of tape files |
448 |
|
o) Read function argument count and types from binaries? (ELF?) |
449 |
|
o) Better demangling of C++ names. Note: GNU's C++ differs from e.g. |
450 |
|
Microsoft's C++, so multiple schemes must be possible. See |
451 |
|
URL at top of src/symbol_demangle.c for more info. |
452 |
|
|
453 |
|
Userland ABI emulation: |
454 |
|
o) see src/useremul.c |
455 |
|
|
456 |
|
Terminal/console: |
457 |
|
o) allow emulated serial ports to be connected to the outside |
458 |
|
world in a more generic way, or even to other emulated |
459 |
|
machines(?) |
460 |
|
|
461 |
|
Save state of the whole emulated machine, to be able to load it back |
462 |
|
in later? (Memory, all device's states, all registers and |
463 |
|
so on. Like taking a snapshot. (SimOS seems to do this, |
464 |
|
according to its website.)) |
465 |
|
|
466 |
|
Better framebuffer and X-windows functionality: |
467 |
|
o) Generalize the update_x1y1x2y2 stuff to an extend-region() |
468 |
|
function... |
469 |
|
o) -Yx sometimes causes crashes. |
470 |
|
o) Simple device access to framebuffer_blockcopyfill() etc, |
471 |
|
and text output (using the built-in fonts), for dev_fb. |
472 |
|
o) CLEAN UP the ugly event code |
473 |
|
o) Mouse clicks can be "missed" in the current system; this is |
474 |
|
not good. They should be put on a stack of some kind. |
475 |
|
o) More 2D and 3D framebuffer acceleration. |
476 |
|
o) Non-resizable windows? Or choose scaledown depending |
477 |
|
on size (and center the image, with a black border). |
478 |
|
o) Different scaledown on different windows? |
479 |
|
o) Non-integral scale-up? (E.g. 640x480 -> 1024x768) |
480 |
|
o) Switch scaledown during runtime? (Ala CTRL-ALT-plus/minus) |
481 |
|
o) Bug reported by Elijah Rutschman on MacOS with weird |
482 |
|
keys (F5 = cursor down?). |
483 |
|
o) Keyboard and mouse events: |
484 |
|
x) Do this for more machines than just DECstation |
485 |
|
x) more X11 cursor keycodes |
486 |
|
x) Keys like CTRL, ALT, SHIFT do not get through |
487 |
|
by themselves (these are necessary for example |
488 |
|
to change the font of an xterm in X in the |
489 |
|
emulator) |
490 |
|
o) Generalize the framebuffer stuff by moving _ALL_ X11 |
491 |
|
specific code to src/x11.c! |
492 |
|
|