| 1 |
$Id: TODO,v 1.125 2005/03/06 08:21:10 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 |
|
Algor: PCI and ISA and LOCAL interrupts! --> wdc could start working |
| 6 |
|
Add interrupt controller in dev_algor.c. |
| 7 |
|
|
| 8 |
Caches / memory hierarchies: (this is mostly MIPS-specific) |
Dyntrans: |
| 9 |
o) MIPS coproc.c: bits in config registers should reflect |
TOP Priority: |
| 10 |
correct cache sizes for _all_ CPU types. (currently only |
x) 64-bit stuff: good generic virtual memory translation, and |
| 11 |
implemented for R4000, R1x000, and a few others) |
other structures. not all 64 bits need to be used |
| 12 |
o) src/memory*.c: Implement correct cache emulation for |
(e.g. Alpha) |
| 13 |
all CPU types. (currently only R2000/R3000 is implemented) |
x) Delay slots! |
| 14 |
(per CPU, multiple levels should be possible, |
x) Old MIPS -> dyntrans! |
| 15 |
associativity etc!) |
Other stuff: |
| 16 |
o) R2000/R3000 isn't _100%_ correct, just almost correct :) |
x) INVALIDATION should cause translations in _all_ cpus to be |
| 17 |
o) Move the -S (fill mem with random) functionality into the |
invalidated, e.g. on a write to a write-protected page |
| 18 |
memory.c subsystem, not machine.c or wherever it is now |
(containing code) |
| 19 |
o) ECC stuff, simulation of memory errors? (Machine dependant) |
x) Think about sharing code between CPUs of the same |
| 20 |
o) More than 4GB of emulated RAM, when run on a 32-bit host? |
arch/bitlength/endianness (the translation caches). |
| 21 |
(using manual swap-out of blocks to disk, ugly) |
(Eg. a little-endian R4000 and a little-endian R5000 could |
| 22 |
o) A global command line option should be used to turn |
share code, but not an R3000 and an R4000.) |
| 23 |
cache emulation on or off. When off, caches should be |
x) better (meaningful) instr call statistics |
| 24 |
faked like they are right now. When on, caches and |
x) Call/return hints? |
| 25 |
memory latencies should be emulated as correctly as |
x) 16-bit encodings? (MIPS16, ARM Thumb, SH3, ...) |
| 26 |
possible. |
x) PPC optimizations; instr combs |
| 27 |
|
x) Alpha |
| 28 |
Network layer: |
x) SPARC |
| 29 |
o) Multiple networks per emulation, and let different |
x) PPC (64-bit stuff) |
| 30 |
NICs in machines connect to different networks. |
x) Lots of other stuff: see src/cpus/README_DYNTRANS |
| 31 |
o) Network across multiple hosts. |
x) true recompilation backend? think carefully about this, |
| 32 |
o) Fixed MAC addresses for NICs, fixed IPv4 addresses if |
experiment in a separate project (not in GXemul) |
| 33 |
DHCP or RARP is used. |
|
| 34 |
o) many other issues: see src/net.c |
More generic out_of_memory error reporting, and check everywhere! |
| 35 |
|
Causes: OpenBSD has low default limits for normal users. |
| 36 |
Configure script: |
Host is 32-bit? (32-bit hosts are limited to 4 GB or less |
| 37 |
o) Only enable Alpha prefetch on pca56 etc, not on ev4, ev5. |
of userspace memory.) |
| 38 |
o) Use getopts? |
You are actually low on RAM. (As trivial as this might sound, |
| 39 |
o) X11 libs and headers via command line option? |
Unix systems usually allow processes to allocate virtual |
| 40 |
o) Verify that the configure script, building and running |
memory beyond the amount of RAM in the machine.) |
| 41 |
the emulator works on some platforms: |
|
| 42 |
+) Irix, AIX, QNX |
Breakpoints: 32-bit vs 64-bit sign extension for MIPS, warnings, etc. |
| 43 |
+) MacOS X (I've not tried this personally yet) |
Use the debugger's symbolic name stuff. (which will have to be |
| 44 |
+) Ultrix using gcc? (Inside the emulator) |
extended soon to support stuff like "2*x + symbol + y" etc. cool |
| 45 |
|
stuff) |
| 46 |
CPU emulation: |
|
| 47 |
o) Binary translation: |
Sprite (guest OS for DECstation emulation) |
| 48 |
+) use an Intermediate Representation! |
x) Timing problems during bootup? |
| 49 |
+) recursive translation? |
|
| 50 |
+) basic blocks instead of one-instruction? :) |
The Device subsystem: |
| 51 |
+) on Alpha: don't assume pca56-like byte load/store? |
x) allow devices to be moved and/or changed in size (down to a |
| 52 |
+) write backends for UltraSparc and MIPS |
minimum size, etc, or up to a max size) |
| 53 |
+) see src/bintrans.c for more info |
x) keep track of interrupts and busses? actually, allowing any device |
| 54 |
|
to be a bus might be a nice idea |
| 55 |
MIPS CPU emulation: |
x) refactor various clocks/nvram/cmos into one device? |
| 56 |
o) Instructions: |
|
| 57 |
o) Regression tests (see tests/README for more info): |
ARM: |
| 58 |
o) Floating point exception handling, and |
o) add ID for "i80321 600MHz rev 2 (XScale core)" |
| 59 |
add more instructions. |
o) make the ata controller usable for FreeBSD! |
| 60 |
o) Finish the MIPS16 translator, and test it! |
x) zaurus for openbsd... |
| 61 |
o) MIPS ISA I, II, III, IV |
|
| 62 |
o) MIPS V (SIMD vector stuff?) |
PCI: |
| 63 |
o) MDMX (MIPS Digital Media Extension) |
x) add support for address fixups |
| 64 |
o) MIPS 3D |
x) generalize the interrupt routing stuff (lines etc). this should |
| 65 |
o) MIPS MT (Multi-thread stuff) (What's this?) |
be per machine? or per bus, that's better |
| 66 |
o) Warn about mis-used bit fields (ie bits that |
|
| 67 |
should be all zeroes, warn about if they are not)! |
MacPPC: |
| 68 |
Both for coprocessor registers and for instruction |
x) dev_zs / interrupts? |
| 69 |
opcodes. |
x) adb controller; keyboard |
| 70 |
o) the special2 stuff is a mess right now |
x) make OpenBSD/macppc work (PCI controller stuff) |
| 71 |
o) warn and/or cause exceptions for unimplemented |
|
| 72 |
instructions (depending on CPU type) |
Network layer: |
| 73 |
o) R2000/R3000: |
o) DHCP (for Debian and BSD installers :-) |
| 74 |
x) R3000 "tri-byte stores". (What's this?) |
o) increase performance |
| 75 |
o) R4000 and others: |
o) don't rely on NetBSD-ish usage |
| 76 |
x) watchhi/watchlo exceptions, and other exception |
o) Multiple networks per emulation, and let different |
| 77 |
handling details |
NICs in machines connect to different networks. |
| 78 |
o) R10000 and others: (R12000, R14000 ?) |
o) many other issues: see src/net.c |
| 79 |
x) memory space, exceptions, ... |
|
| 80 |
x) use cop0 framemask for tlb lookups |
Busses: |
| 81 |
(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) |
o) Redesign the entire "mainbus" concept! |
| 82 |
o) Implement load delays? Warnings on interlocks. |
o) Busses should be placed in a hierarchical tree! |
| 83 |
o) Implement all coprocessor 0 bits / functions. |
o) Easily configurable interrupt routing in SMP systems. |
| 84 |
x) coproc 0 selectors! (R4000 ?) |
o) Specific clock/bus speeds, cpu speeds etc. |
| 85 |
o) R4300 (nintendo64, no mmu?), R5900 (playstation2, weird |
o) Synchronization over network? or at least in dyntrans within |
| 86 |
TLB/cache? 128-bit GPRs, new instructions), |
one emulated machine |
| 87 |
4K (note: NOT R4000), 5K (note: NOT R5000), |
o) dev->bus: TurboChannel, PCMCIA, ADB? |
| 88 |
R6000 (ISA II), R8000 |
|
| 89 |
o) Multi-cpu stuff: |
Config file parser: |
| 90 |
+) SGI's NUMA architecture. Study |
o) Refresh/rewrite it :) |
| 91 |
x) Linux sources |
o) Usage of any expression available through the debugger |
| 92 |
x) SGI's specs on NUMA address space |
o) Expressions such as "add device" would be nice to be able |
| 93 |
+) Ultrix? NetBSD doesn't do SMP on MIPS yet :-( |
to do on the command line manually. |
| 94 |
+) Own experiments with ycx2. |
o) Allow machine() entries even if there is no emul() entry (but |
| 95 |
|
then don't allow any emul() entries at all). |
| 96 |
Emulation of specific machines and devices: |
o) Support for running debugger commands (like the -c |
| 97 |
o) Clean up stuff to make it possible to emulate multiple |
command line option) |
| 98 |
(different) machines simultaneously. |
|
| 99 |
o) Use same clock for all emulations and machines. |
Floating point layer: |
| 100 |
o) Clean up the device stuff (registering of devices etc) |
o) make it common enough to be used by _all_ emulation modes |
| 101 |
o) Various SCSI and IDE controllers |
o) non-IEEE modes (i.e. x86)? |
| 102 |
o) PS/2-style keyboard controller (for several machines) |
|
| 103 |
o) Generic busses; |
Userland emulation: |
| 104 |
x) PCI: i/o and interrupts |
x) Lots of stuff; freebsd and netbsd (and linux?) syscalls. |
| 105 |
x) QBus-22 (DECsystem 5500, 5400?) |
x) Dynamic linking? Hm. |
| 106 |
o) DECstations (pmax): |
|
| 107 |
x) ioasic |
Sound: |
| 108 |
x) framebuffers: |
x) generic sound framework |
| 109 |
+) better cursor support, overlays?, |
x) add one or more sound cards as devices |
| 110 |
+) 2D/3D acceleration, PX[G] |
|
| 111 |
x) status words / control words, make this more |
Caches / memory hierarchies: (this is mostly MIPS-specific) |
| 112 |
portable/cleaner |
o) MIPS coproc.c: bits in config registers should reflect |
| 113 |
x) scsi controller(s): sii, DMA for asc |
correct cache sizes for _all_ CPU types. (currently only |
| 114 |
x) serial controllers: ssc, scc (and more work on dc?) |
implemented for R4000, R1x000, and a few others) |
| 115 |
x) nvram on decstation 5000/125: when using X11, |
o) src/memory*.c: Implement correct cache emulation for |
| 116 |
set console=g or similar |
all CPU types. (currently only R2000/R3000 is implemented) |
| 117 |
x) DECstation 5840? "xbi-based SMP" |
(per CPU, multiple levels should be possible, |
| 118 |
o) SGI and ARC machines (sgimips, arc): |
associativity etc!) |
| 119 |
x) IP32 ("O2"): (Lots of stuff) |
o) R2000/R3000 isn't _100%_ correct, just almost correct :) |
| 120 |
+) mec (ethernet) |
o) Move the -S (fill mem with random) functionality into the |
| 121 |
+) pci |
memory.c subsystem, not machine.c or wherever it is now |
| 122 |
+) ahc (scsi) |
o) ECC stuff, simulation of memory errors? (Machine dependent) |
| 123 |
+) ps2 kbd |
o) More than 4GB of emulated RAM, when run on a 32-bit host? |
| 124 |
+) memory controller |
(using manual swap-out of blocks to disk, ugly) |
| 125 |
+) framebuffer/graphics |
o) A global command line option should be used to turn |
| 126 |
+) caches |
cache emulation on or off. When off, caches should be |
| 127 |
x) more ARCBIOS stuff |
faked like they are right now. When on, caches and |
| 128 |
x) Memory/interrupt controllers |
memory latencies should be emulated as correctly as |
| 129 |
x) IP30 (Linux with graphics support?) |
possible. |
| 130 |
x) SMP / NUMA? (SGI and various ARC machines) |
|
| 131 |
o) Cobalt: PCI and interrupt system, ethernet, |
MIPS CPU emulation: (note: this is for the OLD mips stuff) |
| 132 |
harddisk controller(s) |
o) i386 bintrans backend: movn etc, slt[u] for |
| 133 |
o) Playstation 2: |
64-bit mode, 64-bit shifts etc |
| 134 |
Hardware: OHCI usb controller, keyboard, IDE, ... |
o) Instructions: |
| 135 |
Has an R3000A as a subsystem (!) |
o) All ISAs: |
| 136 |
o) hpcmips: framebuffer(s) and harddisk controller, |
o) Floating point exception handling, and |
| 137 |
among other things |
add more instructions. |
| 138 |
o) newsmips: |
o) Finish the MIPS16 translator, and test it! |
| 139 |
o) NetBSD/newsmips, get it to detect a "real" |
o) MIPS ISA I, II, III, IV |
| 140 |
model, right now everything is 100% bogus |
o) MIPS V (SIMD vector stuff?) |
| 141 |
o) Less interesting platforms: |
o) MDMX (MIPS Digital Media Extension) |
| 142 |
o) mipsco? (NetBSD) |
o) MIPS 3D |
| 143 |
o) wgrisc? (big endian R3000, in OpenBSD's attic) |
o) MIPS MT (Multi-thread stuff) (What's this?) |
| 144 |
o) other embedded / evaluation MIPS boards |
o) Warn about mis-used bit fields (ie bits that |
| 145 |
o) Playstation 1? R3000A. Weird hardware? |
should be all zeroes, warn about if they are not)! |
| 146 |
No FPU, but something as coproc 2. |
Both for coprocessor registers and for instruction |
| 147 |
o) Nintendo 64, http://www.nintendo.com/systems/n64/n64_specs.jsp |
opcodes. |
| 148 |
(R4300, 4MB RAM, really weird memory map) |
o) the special2 stuff is a mess right now |
| 149 |
|
o) warn and/or cause exceptions for unimplemented |
| 150 |
File/disk handling: |
instructions (depending on CPU type) |
| 151 |
o) Read function argument count and types from binaries? (ELF?) |
o) R2000/R3000: |
| 152 |
o) ELF: separate LE/BE (MIPS instruction format) from LSB/MSB |
x) R3000 "tri-byte stores". (What's this?) |
| 153 |
(ELF structure format)? |
o) R4000 and others: |
| 154 |
o) Better handling of tape files |
x) watchhi/watchlo exceptions, and other exception |
| 155 |
|
handling details |
| 156 |
Userland ABI emulation: |
o) R10000 and others: (R12000, R14000 ?) |
| 157 |
o) see src/useremul.c |
x) memory space, exceptions, ... |
| 158 |
|
x) use cop0 framemask for tlb lookups |
| 159 |
Terminal based interactive debugger: |
(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) see src/debugger.c |
o) Implement load delays? Warnings on interlocks. |
| 161 |
|
o) Implement all coprocessor 0 bits / functions. |
| 162 |
GDB interface? (Maybe not necessary; the built-in debugger feels |
x) coproc 0 selectors! (R4000 ?) |
| 163 |
more useful.) |
o) R4300 (nintendo64, no mmu?), R5900 (playstation2, weird |
| 164 |
|
TLB/cache? 128-bit GPRs, new instructions), |
| 165 |
GUI, interactive debugger? GTK+? (The GUI _must be optional_!) |
4K (note: NOT R4000), 5K (note: NOT R5000), |
| 166 |
o) Breakpoints (complex expressions? combinations of register |
R6000 (ISA II), R8000 |
| 167 |
states, memory accesses and so on) |
o) Multi-cpu stuff: |
| 168 |
o) Instruction trace |
+) Interrupt routing (ie devices vs mainbus, or |
| 169 |
o) Disassembly of RAM (_NOT_ the same as instruction trace) |
connect each device to a fixed cpu) |
| 170 |
o) Function call trace |
+) SGI's NUMA architecture. Study |
| 171 |
o) Inspection of CPU registers/state, coprocessor |
x) Linux sources |
| 172 |
registers/state, any device registers/state/contents |
x) SGI's specs on NUMA address space |
| 173 |
o) Framebuffers in windows / tabs. |
+) Ultrix? NetBSD doesn't do SMP on MIPS yet :-( |
| 174 |
o) Contents of RAM |
+) Own experiments with ycx2. |
| 175 |
o) "Start new emulation" should allow amount of |
|
| 176 |
RAM and nr of cpus to easily be specified. |
File/disk/symbol handling: |
| 177 |
Multiple simultaneous machines should be possible. |
o) Better handling of tape files |
| 178 |
o) Disk images. |
o) Read function argument count and types from binaries? (ELF?) |
| 179 |
|
o) Demangle C++ names. |
| 180 |
Regression tests: |
|
| 181 |
o) see tests/ |
Debugger: |
| 182 |
|
o) see src/debugger.c for more |
| 183 |
Save state of the whole emulated machine, to be able to load it back |
|
| 184 |
in later? (Memory, all device's states, all registers and |
Userland ABI emulation: |
| 185 |
so on. Like taking a snapshot. (SimOS seems to do this, |
o) see src/useremul.c |
| 186 |
according to its website.)) |
|
| 187 |
|
Terminal/console: |
| 188 |
Better X-windows functionality: |
o) allow emulated serial ports to be connected to the outside |
| 189 |
o) CLEAN UP the ugly event code |
world in a more generic way, or even to other emulated |
| 190 |
o) Mouse clicks can be "missed" in the current system; this is |
machines(?) |
| 191 |
not good. They should be put on a stack of some kind. |
|
| 192 |
o) More 2D and 3D framebuffer acceleration. |
Save state of the whole emulated machine, to be able to load it back |
| 193 |
o) Non-resizable windows? Or choose scaledown depending |
in later? (Memory, all device's states, all registers and |
| 194 |
on size (and center the image, with a black border). |
so on. Like taking a snapshot. (SimOS seems to do this, |
| 195 |
o) Different scaledown on different windows? |
according to its website.)) |
| 196 |
o) Switch scaledown during runtime? (Ala CTRL-ALT-plus/minus) |
|
| 197 |
o) Keyboard and mouse events: |
Better framebuffer and X-windows functionality: |
| 198 |
x) Do this for more machines than just DECstation |
o) -Yx sometimes causes crashes. |
| 199 |
x) more X11 cursor keycodes |
o) Simple device access to framebuffer_blockcopyfill() etc, |
| 200 |
x) Keys like CTRL, ALT, SHIFT do not get through |
and text output (using the built-in fonts), for dev_fb. |
| 201 |
by themselves (these are necessary for example |
o) CLEAN UP the ugly event code |
| 202 |
to change the font of an xterm in X in the |
o) Mouse clicks can be "missed" in the current system; this is |
| 203 |
emulator) |
not good. They should be put on a stack of some kind. |
| 204 |
o) Generalize the framebuffer stuff by moving _ALL_ X11 |
o) More 2D and 3D framebuffer acceleration. |
| 205 |
specific code to src/x11.c! |
o) Non-resizable windows? Or choose scaledown depending |
| 206 |
|
on size (and center the image, with a black border). |
| 207 |
Statistics: (this could be interesting) |
o) Different scaledown on different windows? |
| 208 |
o) Save to file and show graphics. It should be possible to |
o) Switch scaledown during runtime? (Ala CTRL-ALT-plus/minus) |
| 209 |
run gxemul after a simulation to just show the graphics, |
o) Keyboard and mouse events: |
| 210 |
or convert to a .ppm or .tga or similar. |
x) Do this for more machines than just DECstation |
| 211 |
o) memory accesses (to measure cache efficiency and |
x) more X11 cursor keycodes |
| 212 |
page coloring efficiency) |
x) Keys like CTRL, ALT, SHIFT do not get through |
| 213 |
o) nr of simultaneous ASIDs in use in the TLB, for MIPS |
by themselves (these are necessary for example |
| 214 |
o) percentage of time spent in different "states", such as |
to change the font of an xterm in X in the |
| 215 |
running userland code, kernel code, or idling (for CPUs |
emulator) |
| 216 |
that have such an instruction, or whenever the PC is |
o) Generalize the framebuffer stuff by moving _ALL_ X11 |
| 217 |
inside a specific idle-function (address range)). |
specific code to src/x11.c! |
| 218 |
Possible additional state (for example on R3000): caches |
|
| 219 |
disabled. |
Statistics: (this could be interesting) |
| 220 |
o) position of read/write on (SCSI) disks |
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 |
|
|
Parent Directory
| 
Revision Log
| 
Patch