1 |
$Id: TODO,v 1.139 2005/06/23 06:55:43 debug Exp $ |
$Id: TODO,v 1.209 2006/02/18 21:03:09 debug Exp $ |
2 |
|
|
3 |
Here's a short list of what I have left to do on GXemul. This file is a mess. |
Hm. This file is in random order, and not all parts of it are up-to-date. |
|
Order is random. Don't trust it to be up-to-date. |
|
4 |
|
|
5 |
Caches / memory hierarchies: (this is mostly MIPS-specific) |
Algor: PCI and ISA and LOCAL interrupts! --> wdc could start working |
6 |
o) MIPS coproc.c: bits in config registers should reflect |
Add interrupt controller in dev_algor.c. |
7 |
correct cache sizes for _all_ CPU types. (currently only |
|
8 |
implemented for R4000, R1x000, and a few others) |
Dyntrans: |
9 |
o) src/memory*.c: Implement correct cache emulation for |
TOP Priority: |
10 |
all CPU types. (currently only R2000/R3000 is implemented) |
x) 64-bit stuff: good generic virtual memory translation, and |
11 |
(per CPU, multiple levels should be possible, |
other structures. not all 64 bits need to be used |
12 |
associativity etc!) |
(e.g. Alpha) |
13 |
o) R2000/R3000 isn't _100%_ correct, just almost correct :) |
x) Delay slots! |
14 |
o) Move the -S (fill mem with random) functionality into the |
x) Old MIPS -> dyntrans! |
15 |
memory.c subsystem, not machine.c or wherever it is now |
Other stuff: |
16 |
o) ECC stuff, simulation of memory errors? (Machine dependant) |
x) INVALIDATION should cause translations in _all_ cpus to be |
17 |
o) More than 4GB of emulated RAM, when run on a 32-bit host? |
invalidated, e.g. on a write to a write-protected page |
18 |
(using manual swap-out of blocks to disk, ugly) |
(containing code) |
19 |
o) A global command line option should be used to turn |
x) Think about sharing code between CPUs of the same |
20 |
cache emulation on or off. When off, caches should be |
arch/bitlength/endianness (the translation caches). |
21 |
faked like they are right now. When on, caches and |
(Eg. a little-endian R4000 and a little-endian R5000 could |
22 |
memory latencies should be emulated as correctly as |
share code, but not an R3000 and an R4000.) |
23 |
possible. |
x) better (meaningful) instr call statistics |
24 |
|
x) Call/return hints? |
25 |
Network layer: |
x) 16-bit encodings? (MIPS16, ARM Thumb, SH3, ...) |
26 |
o) Multiple networks per emulation, and let different |
x) PPC optimizations; instr combs |
27 |
NICs in machines connect to different networks. |
x) Alpha |
28 |
o) many other issues: see src/net.c |
x) SPARC |
29 |
|
x) PPC (64-bit stuff) |
30 |
CPU emulation: |
x) Lots of other stuff: see src/cpus/README_DYNTRANS |
31 |
o) Binary translation: |
x) true recompilation backend? think carefully about this, |
32 |
+) use an Intermediate Representation! |
experiment in a separate project (not in GXemul) |
33 |
+) recursive translation? |
|
34 |
+) basic blocks instead of one-instruction? :) |
More generic out_of_memory error reporting, and check everywhere! |
35 |
+) on Alpha: don't assume pca56-like byte load/store? |
Causes: OpenBSD has low default limits for normal users. |
36 |
+) write backends for UltraSparc and MIPS |
Host is 32-bit? (32-bit hosts are limited to 4 GB or less |
37 |
+) see src/bintrans.c for more info |
of userspace memory.) |
38 |
o) Dynamic-but-not-binary translation? |
You are actually low on RAM. (As trivial as this might sound, |
39 |
o) All the non-MIPS modes need a bit of work. |
Unix systems usually allow processes to allocate virtual |
40 |
|
memory beyond the amount of RAM in the machine.) |
41 |
MIPS CPU emulation: |
|
42 |
o) Instructions: |
Breakpoints: 32-bit vs 64-bit sign extension for MIPS, warnings, etc. |
43 |
o) Regression tests (see tests/README for more info): |
Use the debugger's symbolic name stuff. (which will have to be |
44 |
o) Floating point exception handling, and |
extended soon to support stuff like "2*x + symbol + y" etc. cool |
45 |
add more instructions. |
stuff) |
46 |
o) Finish the MIPS16 translator, and test it! |
|
47 |
o) MIPS ISA I, II, III, IV |
Sprite (guest OS for DECstation emulation) |
48 |
o) MIPS V (SIMD vector stuff?) |
x) Timing problems during bootup? |
49 |
o) MDMX (MIPS Digital Media Extension) |
|
50 |
o) MIPS 3D |
The Device subsystem: |
51 |
o) MIPS MT (Multi-thread stuff) (What's this?) |
x) allow devices to be moved and/or changed in size (down to a |
52 |
o) Warn about mis-used bit fields (ie bits that |
minimum size, etc, or up to a max size) |
53 |
should be all zeroes, warn about if they are not)! |
x) keep track of interrupts and busses? actually, allowing any device |
54 |
Both for coprocessor registers and for instruction |
to be a bus might be a nice idea |
55 |
opcodes. |
x) refactor various clocks/nvram/cmos into one device? |
56 |
o) the special2 stuff is a mess right now |
|
57 |
o) warn and/or cause exceptions for unimplemented |
ARM: |
58 |
instructions (depending on CPU type) |
o) add ID for "i80321 600MHz rev 2 (XScale core)" |
59 |
o) R2000/R3000: |
o) make the ata controller usable for FreeBSD! |
60 |
x) R3000 "tri-byte stores". (What's this?) |
x) zaurus for openbsd... |
61 |
o) R4000 and others: |
|
62 |
x) watchhi/watchlo exceptions, and other exception |
PCI: |
63 |
handling details |
x) add support for address fixups |
64 |
o) R10000 and others: (R12000, R14000 ?) |
x) generalize the interrupt routing stuff (lines etc). this should |
65 |
x) memory space, exceptions, ... |
be per machine? or per bus, that's better |
66 |
x) use cop0 framemask for tlb lookups |
|
67 |
(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) |
MacPPC: |
68 |
o) Implement load delays? Warnings on interlocks. |
x) dev_zs / interrupts? |
69 |
o) Implement all coprocessor 0 bits / functions. |
x) adb controller; keyboard |
70 |
x) coproc 0 selectors! (R4000 ?) |
x) make OpenBSD/macppc work (PCI controller stuff) |
71 |
o) R4300 (nintendo64, no mmu?), R5900 (playstation2, weird |
|
72 |
TLB/cache? 128-bit GPRs, new instructions), |
Network layer: |
73 |
4K (note: NOT R4000), 5K (note: NOT R5000), |
o) DHCP (for Debian and BSD installers :-) |
74 |
R6000 (ISA II), R8000 |
o) increase performance |
75 |
o) Multi-cpu stuff: |
o) don't rely on NetBSD-ish usage |
76 |
+) SGI's NUMA architecture. Study |
o) Multiple networks per emulation, and let different |
77 |
x) Linux sources |
NICs in machines connect to different networks. |
78 |
x) SGI's specs on NUMA address space |
o) many other issues: see src/net.c |
79 |
+) Ultrix? NetBSD doesn't do SMP on MIPS yet :-( |
|
80 |
+) Own experiments with ycx2. |
Busses: |
81 |
|
o) Redesign the entire "mainbus" concept! |
82 |
Emulation of specific machines and devices: |
o) Busses should be placed in a hierarchical tree! |
83 |
o) Use same clock for all emulations and machines. |
o) Easily configurable interrupt routing in SMP systems. |
84 |
o) Clean up the device stuff (registering of devices etc) |
o) Specific clock/bus speeds, cpu speeds etc. |
85 |
o) Various SCSI and IDE controllers |
o) Synchronization over network? or at least in dyntrans within |
86 |
o) PS/2-style keyboard controller (for several machines) |
one emulated machine |
87 |
o) Generic busses; |
o) dev->bus: TurboChannel, PCMCIA, ADB? |
88 |
x) PCI: i/o and interrupts |
|
89 |
x) QBus-22 (DECsystem 5500, 5400?) |
Config file parser: |
90 |
o) DECstations (pmax): |
o) Refresh/rewrite it :) |
91 |
x) ioasic |
o) Usage of any expression available through the debugger |
92 |
x) framebuffers: |
o) Expressions such as "add device" would be nice to be able |
93 |
+) better cursor support, overlays?, |
to do on the command line manually. |
94 |
+) 2D/3D acceleration, PX[G] |
o) Allow machine() entries even if there is no emul() entry (but |
95 |
x) status words / control words, make this more |
then don't allow any emul() entries at all). |
96 |
portable/cleaner |
o) Support for running debugger commands (like the -c |
97 |
x) scsi controller(s): sii, DMA for asc |
command line option) |
98 |
x) serial controllers: ssc, scc (and more work on dc?) |
|
99 |
x) nvram on decstation 5000/125: when using X11, |
Floating point layer: |
100 |
set console=g or similar |
o) make it common enough to be used by _all_ emulation modes |
101 |
x) DECstation 5840? "xbi-based SMP" |
o) non-IEEE modes (i.e. x86)? |
102 |
o) SGI and ARC machines (sgimips, arc): |
|
103 |
x) IP32 ("O2"): (Lots of stuff) |
Userland emulation: |
104 |
+) mec (ethernet) |
x) Lots of stuff; freebsd and netbsd (and linux?) syscalls. |
105 |
+) pci |
x) Dynamic linking? Hm. |
106 |
+) ahc (scsi) |
|
107 |
+) ps2 kbd |
Sound: |
108 |
+) memory controller |
x) generic sound framework |
109 |
+) framebuffer/graphics |
x) add one or more sound cards as devices |
110 |
+) caches |
|
111 |
x) more ARCBIOS stuff |
Caches / memory hierarchies: (this is mostly MIPS-specific) |
112 |
x) Memory/interrupt controllers |
o) MIPS coproc.c: bits in config registers should reflect |
113 |
x) IP30 (Linux with graphics support?) |
correct cache sizes for _all_ CPU types. (currently only |
114 |
x) SMP / NUMA? (SGI and various ARC machines) |
implemented for R4000, R1x000, and a few others) |
115 |
o) Playstation 2: |
o) src/memory*.c: Implement correct cache emulation for |
116 |
Hardware: OHCI usb controller, keyboard, ... |
all CPU types. (currently only R2000/R3000 is implemented) |
117 |
o) Less interesting platforms: |
(per CPU, multiple levels should be possible, |
118 |
o) newsmips: |
associativity etc!) |
119 |
o) NetBSD/newsmips, get it to detect a "real" |
o) R2000/R3000 isn't _100%_ correct, just almost correct :) |
120 |
model, right now everything is 100% bogus |
o) Move the -S (fill mem with random) functionality into the |
121 |
o) mipsco? (NetBSD) |
memory.c subsystem, not machine.c or wherever it is now |
122 |
o) wgrisc? (big endian R3000, in OpenBSD's attic) |
o) ECC stuff, simulation of memory errors? (Machine dependent) |
123 |
o) other embedded / evaluation MIPS boards |
o) More than 4GB of emulated RAM, when run on a 32-bit host? |
124 |
o) Playstation 1? R3000A. Weird hardware? |
(using manual swap-out of blocks to disk, ugly) |
125 |
No FPU, but something as coproc 2. |
o) A global command line option should be used to turn |
126 |
o) Nintendo 64, http://www.nintendo.com/systems/n64/n64_specs.jsp |
cache emulation on or off. When off, caches should be |
127 |
(R4300, 4MB RAM, really weird memory map) |
faked like they are right now. When on, caches and |
128 |
|
memory latencies should be emulated as correctly as |
129 |
File/disk handling: |
possible. |
130 |
o) Read function argument count and types from binaries? (ELF?) |
|
131 |
o) Better handling of tape files |
MIPS CPU emulation: (note: this is for the OLD mips stuff) |
132 |
|
o) i386 bintrans backend: movn etc, slt[u] for |
133 |
Userland ABI emulation: |
64-bit mode, 64-bit shifts etc |
134 |
o) see src/useremul.c |
o) Instructions: |
135 |
|
o) All ISAs: |
136 |
Terminal based interactive debugger: |
o) Floating point exception handling, and |
137 |
o) see src/debugger.c |
add more instructions. |
138 |
|
o) Finish the MIPS16 translator, and test it! |
139 |
Terminal/console stuff: |
o) MIPS ISA I, II, III, IV |
140 |
o) allow emulated serial ports to be connected to the outside |
o) MIPS V (SIMD vector stuff?) |
141 |
world in a more generic way, or even to other emulated |
o) MDMX (MIPS Digital Media Extension) |
142 |
machines(!) |
o) MIPS 3D |
143 |
|
o) MIPS MT (Multi-thread stuff) (What's this?) |
144 |
Regression tests: (Should be totally rewritten.) |
o) Warn about mis-used bit fields (ie bits that |
145 |
o) see tests/ |
should be all zeroes, warn about if they are not)! |
146 |
|
Both for coprocessor registers and for instruction |
147 |
Save state of the whole emulated machine, to be able to load it back |
opcodes. |
148 |
in later? (Memory, all device's states, all registers and |
o) the special2 stuff is a mess right now |
149 |
so on. Like taking a snapshot. (SimOS seems to do this, |
o) warn and/or cause exceptions for unimplemented |
150 |
according to its website.)) |
instructions (depending on CPU type) |
151 |
|
o) R2000/R3000: |
152 |
Better X-windows functionality: |
x) R3000 "tri-byte stores". (What's this?) |
153 |
o) CLEAN UP the ugly event code |
o) R4000 and others: |
154 |
o) Mouse clicks can be "missed" in the current system; this is |
x) watchhi/watchlo exceptions, and other exception |
155 |
not good. They should be put on a stack of some kind. |
handling details |
156 |
o) More 2D and 3D framebuffer acceleration. |
o) R10000 and others: (R12000, R14000 ?) |
157 |
o) Non-resizable windows? Or choose scaledown depending |
x) memory space, exceptions, ... |
158 |
on size (and center the image, with a black border). |
x) use cop0 framemask for tlb lookups |
159 |
o) Different scaledown on different windows? |
(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) |
160 |
o) Switch scaledown during runtime? (Ala CTRL-ALT-plus/minus) |
o) Implement load delays? Warnings on interlocks. |
161 |
o) Keyboard and mouse events: |
o) Implement all coprocessor 0 bits / functions. |
162 |
x) Do this for more machines than just DECstation |
x) coproc 0 selectors! (R4000 ?) |
163 |
x) more X11 cursor keycodes |
o) R4300 (nintendo64, no mmu?), R5900 (playstation2, weird |
164 |
x) Keys like CTRL, ALT, SHIFT do not get through |
TLB/cache? 128-bit GPRs, new instructions), |
165 |
by themselves (these are necessary for example |
4K (note: NOT R4000), 5K (note: NOT R5000), |
166 |
to change the font of an xterm in X in the |
R6000 (ISA II), R8000 |
167 |
emulator) |
o) Multi-cpu stuff: |
168 |
o) Generalize the framebuffer stuff by moving _ALL_ X11 |
+) Interrupt routing (ie devices vs mainbus, or |
169 |
specific code to src/x11.c! |
connect each device to a fixed cpu) |
170 |
|
+) SGI's NUMA architecture. Study |
171 |
Statistics: (this could be interesting) |
x) Linux sources |
172 |
o) Save to file and show graphics. It should be possible to |
x) SGI's specs on NUMA address space |
173 |
run gxemul after a simulation to just show the graphics, |
+) Ultrix? NetBSD doesn't do SMP on MIPS yet :-( |
174 |
or convert to a .ppm or .tga or similar. |
+) Own experiments with ycx2. |
175 |
o) memory accesses (to measure cache efficiency and |
|
176 |
page coloring efficiency) |
File/disk/symbol handling: |
177 |
o) nr of simultaneous ASIDs in use in the TLB, for MIPS |
o) Better handling of tape files |
178 |
o) percentage of time spent in different "states", such as |
o) Read function argument count and types from binaries? (ELF?) |
179 |
running userland code, kernel code, or idling (for CPUs |
o) Demangle C++ names. |
180 |
that have such an instruction, or whenever the PC is |
|
181 |
inside a specific idle-function (address range)). |
Debugger: |
182 |
Possible additional state (for example on R3000): caches |
o) see src/debugger.c for more |
183 |
disabled. |
|
184 |
o) position of read/write on (SCSI) disks |
Userland ABI emulation: |
185 |
|
o) see src/useremul.c |
186 |
|
|
187 |
|
Terminal/console: |
188 |
|
o) allow emulated serial ports to be connected to the outside |
189 |
|
world in a more generic way, or even to other emulated |
190 |
|
machines(?) |
191 |
|
|
192 |
|
Save state of the whole emulated machine, to be able to load it back |
193 |
|
in later? (Memory, all device's states, all registers and |
194 |
|
so on. Like taking a snapshot. (SimOS seems to do this, |
195 |
|
according to its website.)) |
196 |
|
|
197 |
|
Better framebuffer and X-windows functionality: |
198 |
|
o) -Yx sometimes causes crashes. |
199 |
|
o) Simple device access to framebuffer_blockcopyfill() etc, |
200 |
|
and text output (using the built-in fonts), for dev_fb. |
201 |
|
o) CLEAN UP the ugly event code |
202 |
|
o) Mouse clicks can be "missed" in the current system; this is |
203 |
|
not good. They should be put on a stack of some kind. |
204 |
|
o) More 2D and 3D framebuffer acceleration. |
205 |
|
o) Non-resizable windows? Or choose scaledown depending |
206 |
|
on size (and center the image, with a black border). |
207 |
|
o) Different scaledown on different windows? |
208 |
|
o) Switch scaledown during runtime? (Ala CTRL-ALT-plus/minus) |
209 |
|
o) Keyboard and mouse events: |
210 |
|
x) Do this for more machines than just DECstation |
211 |
|
x) more X11 cursor keycodes |
212 |
|
x) Keys like CTRL, ALT, SHIFT do not get through |
213 |
|
by themselves (these are necessary for example |
214 |
|
to change the font of an xterm in X in the |
215 |
|
emulator) |
216 |
|
o) Generalize the framebuffer stuff by moving _ALL_ X11 |
217 |
|
specific code to src/x11.c! |
218 |
|
|
219 |
|
Statistics: (this could be interesting) |
220 |
|
o) Save to file and show graphics. It should be possible to |
221 |
|
run gxemul after a simulation to just show the graphics, |
222 |
|
or convert to a .ppm or .tga or similar. |
223 |
|
o) memory accesses (to measure cache efficiency and |
224 |
|
page coloring efficiency) |
225 |
|
o) nr of simultaneous ASIDs in use in the TLB, for MIPS |
226 |
|
o) percentage of time spent in different "states", such as |
227 |
|
running userland code, kernel code, or idling (for CPUs |
228 |
|
that have such an instruction, or whenever the PC is |
229 |
|
inside a specific idle-function (address range)). |
230 |
|
Possible additional state (for example on R3000): caches |
231 |
|
disabled. |
232 |
|
o) position of read/write on (SCSI) disks |
233 |
|
|