adjustments for CPU core changes

[picodrive.git] / cpu / Cyclone / Cyclone.txt

\r
      _____            __                                     \r
     / ___/__ __ ____ / /___   ___  ___   ___________________ \r
    / /__ / // // __// // _ \ / _ \/ -_) ___________________  \r
    \___/ \_, / \__//_/ \___//_//_/\__/ ___________________   \r
         /___/                                                \r
         ___________________  ____ ___   ___   ___   ___      \r
        ___________________  / __// _ \ / _ \ / _ \ / _ \     \r
       ___________________  / _ \/ _  // // // // // // /     \r
                            \___/\___/ \___/ \___/ \___/      \r
                                                              \r
___________________________________________________________________________\r
\r
  Cyclone 68000 (c) Copyright 2004 Dave.   Free for non-commercial use\r
\r
  Homepage: http://www.finalburn.com/\r
  Dave's e-mail: emudave(atsymbol)googlemail.com\r
  Replace (atsymbol) with @\r
\r
  Additional coding and bugfixes done by notaz, 2005-2007\r
  Homepage: http://mif.vu.lt/~grig2790/Cyclone/ , http://notaz.gp2x.de\r
  e-mail: notasas(atsymbol)gmail.com\r
___________________________________________________________________________\r
\r
\r
What is it?\r
-----------\r
\r
Cyclone 68000 is an emulator for the 68000 microprocessor, written in ARM 32-bit assembly.\r
It is aimed at chips such as ARM7 and ARM9 cores, StrongARM and XScale, to interpret 68000\r
code as fast as possible.\r
\r
Flags are mapped onto ARM flags whenever possible, which speeds up the processing of opcode.\r
\r
\r
What's New\r
----------\r
v0.0087 notaz\r
  - Reduced amount of code in opcode handlers by ~23% by doing the following:\r
    - Removed duplicate opcode handlers\r
    - Optimized code to use less ARM instructions\r
    - Merged some duplicate handler endings\r
  + Cyclone now does better job avoiding pipeline interlocks.\r
  + Replaced incorrect handler of DBT with proper one.\r
  + Changed "MOVEA (An)+ An" behaviour.\r
  + Fixed flag behaviour of ROXR, ASL, LSR and NBCD in certain situations.\r
    Hopefully got them right now.\r
  + Additional functionality added for MAME and other ports (see config.h).\r
\r
v0.0086 notaz\r
  + Cyclone now can be customized to better suit your project, see config.h .\r
  + Added an option to compress the jumptable at compile-time. Must call CycloneInit()\r
    at runtime to decompress it if enabled (see config.h).\r
  + Added missing CHK opcode handler (used by SeaQuest DSV).\r
  + Added missing TAS opcode handler (Gargoyles,Bubba N Stix,...). As in real genesis,\r
    memory write-back phase is ignored (but can be enabled in config.h if needed).\r
  + Added missing NBCD and TRAPV opcode handlers.\r
  + Added missing addressing mode for CMP/EOR.\r
  + Added some minor optimizations.\r
  - Removed 216 handlers for 2927 opcodes which were generated for invalid addressing modes.\r
  + Fixed flags for ASL, NEG, NEGX, DIVU, ADDX, SUBX, ROXR.\r
  + Bugs fixed in MOVEP, LINK, ADDQ, DIVS handlers.\r
  * Undocumented flags for CHK, ABCD, SBCD and NBCD are now emulated the same way as in Musashi.\r
  + Added Uninitialized Interrupt emulation.\r
  + Altered timing for about half of opcodes to match Musashi's.\r
\r
v0.0082 Reesy\r
  + Change cyclone to clear cycles before returning when halted\r
  + Added Irq call back function.  This allows emulators to be notified\r
    when cyclone has taken an interrupt allowing them to set internal flags\r
    which can help fix timing problems.\r
\r
v0.0081 notaz\r
  + .asm version was broken and did not compile with armasm. Fixed.\r
  + Finished implementing Stop opcode. Now it really stops the processor.\r
\r
v0.0080 notaz\r
  + Added real cmpm opcode, it was using eor handler before this.\r
    Fixes Dune and Sensible Soccer.\r
\r
v0.0078 notaz\r
  note: these bugs were actually found Reesy, I reimplemented these by\r
        using his changelog as a guide.\r
  + Fixed a problem with divu which was using long divisor instead of word.\r
    Fixes gear switching in Top Gear 2.\r
  + Fixed btst opcode, The bit to test should shifted a max of 31 or 7\r
    depending on if a register or memory location is being tested.\r
  + Fixed abcd,sbcd. They did bad decimal correction on invalid BCD numbers\r
    Score counters in Streets of Rage level end work now.\r
  + Changed flag handling of abcd,sbcd,addx,subx,asl,lsl,...\r
    Some ops did not have flag handling at all.\r
    Some ops must not change Z flag when result is zero, but they did.\r
    Shift ops must not change X if shift count is zero, but they did.\r
    There are probably still some flag problems left.\r
  + Patially implemented Stop and Reset opcodes - Fixes Thunderforce IV\r
\r
v0.0075 notaz\r
  + Added missing displacement addressing mode for movem (Fantastic Dizzy)\r
  + Added OSP <-> A7 swapping code in opcodes, which change privilege mode\r
  + Implemented privilege violation, line emulator and divide by zero exceptions\r
  + Added negx opcode (Shining Force works!)\r
  + Added overflow detection for divs/divu\r
\r
v0.0072 notaz\r
  note: I could only get v0.0069 cyclone, so I had to implement these myself using Dave's\r
        changelog as a guide.\r
  + Fixed a problem with divs - remainder should be negative when divident is negative\r
  + Added movep opcode (Sonic 3 works)\r
  + Fixed a problem with DBcc incorrectly decrementing if the condition is true (Shadow of the Beast)\r
\r
v0.0069\r
  + Added SBCD and the flags for ABCD/SBCD. Score and time now works in games such as\r
    Rolling Thunder 2, Ghouls 'N Ghosts\r
  + Fixed a problem with addx and subx with 8-bit and 16-bit values.\r
    Ghouls 'N' Ghosts now works!\r
\r
v0.0068\r
  + Added ABCD opcode (Streets of Rage works now!)\r
\r
v0.0067\r
  + Added dbCC (After Burner)\r
  + Added asr EA (Sonic 1 Boss/Labyrinth Zone)\r
  + Added andi/ori/eori ccr (Altered Beast)\r
  + Added trap (After Burner)\r
  + Added special case for move.b (a7)+ and -(a7), stepping by 2\r
    After Burner is playable! Eternal Champions shows more\r
  + Fixed lsr.b/w zero flag (Ghostbusters)\r
    Rolling Thunder 2 now works!\r
  + Fixed N flag for .b and .w arithmetic. Golden Axe works!\r
\r
v0.0066\r
  + Fixed a stupid typo for exg (orr r10,r10, not orr r10,r8), which caused alignment\r
    crashes on Strider\r
\r
v0.0065\r
  + Fixed a problem with immediate values - they weren't being shifted up correctly for some\r
    opcodes. Spiderman works, After Burner shows a bit of graphics.\r
  + Fixed a problem with EA:"110nnn" extension word. 32-bit offsets were being decoded as 8-bit\r
    offsets by mistake. Castlevania Bloodlines seems fine now.\r
  + Added exg opcode\r
  + Fixed asr opcode (Sonic jumping left is fixed)\r
  + Fixed a problem with the carry bit in rol.b (Marble Madness)\r
\r
v0.0064\r
  + Added rtr\r
  + Fixed addq/subq.l (all An opcodes are 32-bit) (Road Rash)\r
  + Fixed various little timings\r
\r
v0.0063\r
  + Added link/unlk opcodes\r
  + Fixed various little timings\r
  + Fixed a problem with dbCC opcode being emitted at set opcodes\r
  + Improved long register access, the EA fetch now does ldr r0,[r7,r0,lsl #2] whenever\r
     possible, saving 1 or 2 cycles on many opcodes, which should give a nice speed up.\r
  + May have fixed N flag on ext opcode?\r
  + Added dasm for link opcode.\r
\r
v0.0062\r
  * I was a bit too keen with the Arithmetic opcodes! Some of them should have been abcd,\r
    exg and addx. Removed the incorrect opcodes, pending re-adding them as abcd, exg and addx.\r
  + Changed unknown opcodes to act as nops.\r
    Not very technical, but fun - a few more games show more graphics ;)\r
\r
v0.0060\r
  + Fixed divu (EA intro)\r
  + Added sf (set false) opcode - SOR2\r
  * Todo: pea/link/unlk opcodes\r
\r
v0.0059: Added remainder to divide opcodes.\r
\r
\r
The new stuff\r
-------------\r
\r
Before using Cyclone, be sure to customize config.h to better suit your project. All options\r
are documented inside that file.\r
\r
IrqCallback has been changed a bit, unlike in previous version, it should not return anything.\r
If you need to change IRQ level, you can safely do that in your handler.\r
\r
Cyclone has changed quite a bit from the time when Dave stopped updating it, but the rest of\r
documentation still applies, so read it if you haven't done that yet. If you have, check the\r
"Accessing ..." parts.\r
\r
\r
ARM Register Usage\r
------------------\r
\r
See source code for up to date of register usage, however a summary is here:\r
\r
  r0-3: Temporary registers\r
  r4  : Current PC + Memory Base (i.e. pointer to next opcode)\r
  r5  : Cycles remaining\r
  r6  : Pointer to Opcode Jump table\r
  r7  : Pointer to Cpu Context\r
  r8  : Current Opcode\r
  r9  : Flags (NZCV) in highest four bits\r
 (r10 : Temporary source value or Memory Base)\r
 (r11 : Temporary register)\r
\r
\r
How to Compile\r
--------------\r
\r
Like Starscream and A68K, Cyclone uses a 'Core Creator' program which calculates and outputs\r
all possible 68000 Opcodes and a jump table into files called Cyclone.s and .asm\r
It then assembles these files into Cyclone.o and .obj\r
\r
Cyclone.o is the GCC assembled version and Cyclone.obj is the Microsoft assembled version.\r
\r
First unzip "Cyclone.zip" into a "Cyclone" directory.\r
If you are compiling for Windows CE, find ARMASM.EXE (the Microsoft ARM assembler) and\r
put it in the directory as well or put it on your path.\r
\r
Open up Cyclone.dsw in Visual Studio 6.0, compile and run the project.\r
Cyclone.obj and Cyclone.o will be created.\r
\r
\r
Compiling without Visual C++\r
----------------------------\r
If you aren't using Visual C++, it still shouldn't be too hard to compile, just get a C compiler,\r
compile all the CPPs and C file, link them into an EXE, and run the exe.\r
\r
  e.g. gcc Main.cpp OpAny.cpp OpArith.cpp OpBranch.cpp OpLogic.cpp OpMove.cpp Disa.c\r
  Main.exe\r
\r
\r
Adding to your project\r
----------------------\r
\r
To add Cyclone to you project, add Cyclone.o or obj, and include Cyclone.h\r
There is one structure: 'struct Cyclone', and one function: CycloneRun\r
\r
Don't worry if this seem very minimal - its all you need to run as many 68000s as you want.\r
It works with both C and C++.\r
\r
Byteswapped Memory\r
------------------\r
\r
If you have used Starscream, A68K or Turbo68K or similar emulators you'll be familiar with this!\r
\r
Any memory which the 68000 can access directly must be have every two bytes swapped around.\r
This is to speed up 16-bit memory accesses, because the 68000 has Big-Endian memory\r
and ARM has Little-Endian memory.\r
\r
Now you may think you only technically have to byteswap ROM, not RAM, because\r
16-bit RAM reads go through a memory handler and you could just return (mem[a]<<8) | mem[a+1].\r
\r
This would work, but remember some systems can execute code from RAM as well as ROM, and\r
that would fail.\r
So it's best to use byteswapped ROM and RAM if the 68000 can access it directly.\r
It's also faster for the memory handlers, because you can do this:\r
  \r
  return *(unsigned short *)(mem+a)\r
\r
\r
Declaring Memory handlers\r
-------------------------\r
\r
Before you can reset or execute 68000 opcodes you must first set up a set of memory handlers.\r
There are 7 functions you have to set up per CPU, like this:\r
\r
  static unsigned int   MyCheckPc(unsigned int pc)\r
  static unsigned char  MyRead8  (unsigned int a)\r
  static unsigned short MyRead16 (unsigned int a)\r
  static unsigned int   MyRead32 (unsigned int a)\r
  static void MyWrite8 (unsigned int a,unsigned char  d)\r
  static void MyWrite16(unsigned int a,unsigned short d)\r
  static void MyWrite32(unsigned int a,unsigned int   d)\r
\r
You can think of these functions representing the 68000's memory bus.\r
The Read and Write functions are called whenever the 68000 reads or writes memory.\r
For example you might set MyRead8 like this:\r
\r
  unsigned char MyRead8(unsigned int a)\r
  {\r
    a&=0xffffff; // Clip address to 24-bits\r
\r
    if (a<RomLength) return RomData[a^1]; // ^1 because the memory is byteswapped\r
    if (a>=0xe00000) return RamData[(a^1)&0xffff];\r
    return 0xff; // Out of range memory access\r
  }\r
\r
The other 5 read/write functions are similar. I'll describe the CheckPc function later on.\r
\r
Declaring a CPU Context\r
-----------------------\r
\r
To declare a CPU simple declare a struct Cyclone in your code. For example to declare\r
two 68000s:\r
\r
  struct Cyclone MyCpu;\r
  struct Cyclone MyCpu2;\r
\r
It's probably a good idea to initialise the memory to zero:\r
\r
  memset(&MyCpu, 0,sizeof(MyCpu));\r
  memset(&MyCpu2,0,sizeof(MyCpu2));\r
\r
Next point to your memory handlers:\r
\r
  MyCpu.checkpc=MyCheckPc;\r
  MyCpu.read8  =MyRead8;\r
  MyCpu.read16 =MyRead16;\r
  MyCpu.read32 =MyRead32;\r
  MyCpu.write8 =MyWrite8;\r
  MyCpu.write16=MyWrite16;\r
  MyCpu.write32=MyWrite32;\r
\r
You also need to point the fetch handlers - for most systems out there you can just\r
point them at the read handlers:\r
  MyCpu.fetch8  =MyRead8;\r
  MyCpu.fetch16 =MyRead16;\r
  MyCpu.fetch32 =MyRead32;\r
\r
( Why a different set of function pointers for fetch?\r
  Well there are some systems, the main one being CPS2, which return different data\r
  depending on whether the 'fetch' line on the 68000 bus is high or low.\r
  If this is the case, you can set up different functions for fetch reads.\r
  Generally though you don't need to. )\r
\r
Now you are nearly ready to reset the 68000, except you need one more function: checkpc().\r
\r
The checkpc() function\r
----------------------\r
\r
When Cyclone reads opcodes, it doesn't use a memory handler every time, this would be\r
far too slow, instead it uses a direct pointer to ARM memory.\r
For example if your Rom image was at 0x3000000 and the program counter was $206,\r
Cyclone's program counter would be 0x3000206.\r
\r
The difference between an ARM address and a 68000 address is also stored in a variable called\r
'membase'. In the above example it's 0x3000000. To retrieve the real PC, Cyclone just\r
subtracts 'membase'.\r
\r
When a long jump happens, Cyclone calls checkpc(). If the PC is in a different bank,\r
for example Ram instead of Rom, change 'membase', recalculate the new PC and return it:\r
\r
static int MyCheckPc(unsigned int pc)\r
{\r
  pc-=MyCpu.membase; // Get the real program counter\r
\r
  if (pc<RomLength) MyCpu.membase=(int)RomMem;          // Jump to Rom\r
  if (pc>=0xff0000) MyCpu.membase=(int)RamMem-0xff0000; // Jump to Ram\r
\r
  return MyCpu.membase+pc; // New program counter\r
}\r
\r
Notice that the membase is always ARM address minus 68000 address.\r
\r
The above example doesn't consider mirrored ram, but for an example of what to do see\r
PicoDrive (in Memory.cpp).\r
\r
\r
Almost there - Reset the 68000!\r
-------------------------------\r
\r
Next we need to Reset the 68000 to get the initial Program Counter and Stack Pointer. This\r
is obtained from addresses 000000 and 000004.\r
\r
Here is code which resets the 68000 (using your memory handlers):\r
\r
  MyCpu.srh=0x27; // Set supervisor mode\r
  MyCpu.a[7]=MyCpu.read32(0); // Get Stack Pointer\r
  MyCpu.membase=0;\r
  MyCpu.pc=MyCpu.checkpc(MyCpu.read32(4)); // Get Program Counter\r
\r
And that's ready to go.\r
\r
\r
Executing the 68000\r
-------------------\r
\r
To execute the 68000, set the 'cycles' variable to the number of cycles you wish to execute,\r
and then call CycloneRun with a pointer to the Cyclone structure.\r
\r
e.g.:\r
  // Execute 1000 cycles on the 68000:\r
  MyCpu.cycles=1000; CycloneRun(&MyCpu);\r
\r
For each opcode, the number of cycles it took is subtracted and the function returns when\r
it reaches 0.\r
\r
e.g.\r
  // Execute one instruction on the 68000:\r
  MyCpu.cycles=0; CycloneRun(&MyCpu);\r
  printf("  The opcode took %d cycles\n", -MyCpu.cycles);\r
\r
You should try to execute as many cycles as you can for maximum speed.\r
The number actually executed may be slightly more than requested, i.e. cycles may come\r
out with a small negative value:\r
\r
e.g.\r
  int todo=12000000/60; // 12Mhz, for one 60hz frame\r
  MyCpu.cycles=todo; CycloneRun(&MyCpu);\r
  printf("  Actually executed %d cycles\n", todo-MyCpu.cycles);\r
\r
To calculate the number of cycles executed, use this formula:\r
  Number of cycles requested - Cycle counter at the end\r
\r
\r
Interrupts\r
----------\r
\r
Causing an interrupt is very simple, simply set the irq variable in the Cyclone structure\r
to the IRQ number.\r
To lower the IRQ line, set it to zero.\r
\r
e.g:\r
  MyCpu.irq=6; // Interrupt level 6\r
  MyCpu.cycles=20000; CycloneRun(&MyCpu);\r
\r
Note that the interrupt is not actually processed until the next call to CycloneRun,\r
and the interrupt may not be taken until the 68000 interrupt mask is changed to allow it.\r
\r
( The IRQ isn't checked on exiting from a memory handler: I don't think this will cause\r
  me any trouble because I've never needed to trigger an interrupt from a memory handler,\r
  but if someone needs to, let me know...)\r
\r
\r
Accessing Cycle Counter\r
-----------------------\r
\r
The cycle counter in the Cyclone structure is not, by default, updated before\r
calling a memory handler, only at the end of an execution.\r
\r
*update*\r
Now this is configurable in config.h, there is no 'debug' variable.\r
\r
\r
Accessing Program Counter and registers\r
---------------------------------------\r
\r
You can read Cyclone's registers directly from the structure at any time (as far as I know).\r
\r
The Program Counter, should you need to read or write it, is stored with membase\r
added on. So use this formula to calculate the real 68000 program counter:\r
\r
  pc = MyCpu.pc - MyCpu.membase;\r
\r
The program counter is stored in r4 during execution, and isn't written back to the\r
structure until the end of execution, which means you can't read normally real it from\r
a memory handler.\r
\r
*update*\r
Now this is configurable in config.h, there is no 'debug' variable. You can even enable\r
access to SR if you need. However changing PC in memhandlers is still not safe, you should\r
better clear cycles, wait untill CycloneRun() returns and then do whatever you need.\r
\r
\r
Emulating more than one CPU\r
---------------------------\r
\r
Since everything is based on the structures, emulating more than one cpu at the same time\r
is just a matter of declaring more than one structures and timeslicing. You can emulate\r
as many 68000s as you want.\r
Just set up the memory handlers for each cpu and run each cpu for a certain number of cycles.\r
\r
e.g.\r
  // Execute 1000 cycles on 68000 #1:\r
  MyCpu.cycles=1000; CycloneRun(&MyCpu);\r
\r
  // Execute 1000 cycles on 68000 #2:\r
  MyCpu2.cycles=1000; CycloneRun(&MyCpu2);\r
\r
\r
Thanks to...\r
------------\r
\r
* All the previous code-generating assembler cpu core guys!\r
  Who are iirc... Neill Corlett, Neil Bradley, Mike Coates, Darren Olafson\r
    and Bart Trzynadlowski\r
\r
* Charles Macdonald, for researching just about every console ever\r
* MameDev+FBA, for keeping on going and going and going\r
\r
\r
-------------\r
\r
Dave  - 17th April 2004\r
notaz - 17th July 2006\r
\r
Homepage: http://www.finalburn.com/\r
Dave's e-mail: dev(atsymbol)finalburn.com\r
Replace (atsymbol) with @\r