mz80.txt

著名ARC模拟器源码,包括多个平台

Multi-Z80 32 Bit emulator
Copyright 1996, 1997, 1998, 1999, 2000 - Neil Bradley, All rights reserved

			    MZ80 License agreement
			    -----------------------

(MZ80 Refers to both the assembly code emitted by makez80.c and makez80.c
itself)

MZ80 May be distributed in unmodified form to any medium.

MZ80 May not be sold, or sold as a part of a commercial package without
the express written permission of Neil Bradley ([e-mail]). This
includes shareware.

Modified versions of MZ80 may not be publicly redistributed without author
approval ([e-mail]). This includes distributing via a publicly
accessible LAN. You may make your own source modifications and distribute
MZ80 in source or object form, but if you make modifications to MZ80
then it should be noted in the top as a comment in makez80.c.

MZ80 Licensing for commercial applications is available. Please email
[e-mail] for details.

Synthcom Systems, Inc, and Neil Bradley will not be held responsible for
any damage done by the use of MZ80. It is purely "as-is".

If you use MZ80 in a freeware application, credit in the following text:

"Multi-Z80 CPU emulator by Neil Bradley ([e-mail])"

must accompany the freeware application within the application itself or
in the documentation.

Legal stuff aside:

If you find problems with MZ80, please email the author so they can get
resolved. If you find a bug and fix it, please also email the author so
that those bug fixes can be propogated to the installed base of MZ80
users. If you find performance improvements or problems with MZ80, please
email the author with your changes/suggestions and they will be rolled in
with subsequent releases of MZ80.

The whole idea of this emulator is to have the fastest available 32 bit
Multi-Z80 emulator for the x86, giving maximum performance.
 
                         MZ80 Contact information
			  -------------------------

Author      : Neil Bradley ([e-mail])
Distribution: ftp://ftp.synthcom.com/pub/emulators/cpu/makez80.zip (latest)

You can join the cpuemu mailing list on Synthcom for discussion of Neil
Bradley's Z80 (and other) CPU emulators. Send a message to 
"[e-mail]" with "subscribe" in the message body. The
traffic is fairly low, and is used as a general discussion and announcement
for aforementioned emulators.


			     MZ80 Documentation
			     -------------------

MZ80 Is a full featured Z80 emulator coded in 32 bit assembly. It runs well
over a hundred games, in addition to it supporting many undocumented Z80
instructions required to run some of the Midway MCR games, Galaga, and
countless other wonderful Z80 based arcade games.

MZ80 Contains a makez80.c program that must be compiled. It is the program
that emits the assembly code that NASM will compile. This minimizes the
possibility of bugs creeping in to MZ80 for the different addressing modes
for each instruction. It requires NASM 0.97 or greater.

The goal of MZ80 is to have a high performance Z80 emulator that is capable
of running multiple emulations concurrently at full speed, even on lower-end
machines (486/33). MZ80 Harnesses the striking similarities of both the Z80
and the x86 instruction sets to take advantage of flag handling which greatly
reduces the time required to emulate a processor, so no extra time is spent
computing things that are already available in the native x86 processor,
allowing it to perform leaps and bounds over comparable C based Z80 emulators
on the same platform.

MZ80 Is designed exclusively for use with NASM, the Netwide Assembler. This
gives the ultimate in flexibility, as NASM can emit object files that work
with Watcom, Microsoft Visual C++ (4.0-current), DJGPP, Borland C++, and
gcc under FreeBSD or Linux. MZ80 Has been tested with each one of these
compilers and is known to work properly on each.


			    What's in the package
			    ---------------------

MZ80.TXT               - This text file

MAKEZ80.C              - Multi Z80 32 Bit emulator emitter program

MZ80.H                 - C Header file for MZ80 functions


			  What's new in this release
			  --------------------------

Revision 3.3:

	* Undocumented opcodes added to the C emitter
	* Bug fix to the C emission that properly handles shared RAM regions
	  (I.E. with handlers that are NULL)
	* Now using 32 bit registers to do register/memory access. Slight
	  speed increase (assembly version only)

Revision 3.2:
	
	* R Register emulation now accurate with a real Z80
	* mz80int() Called when interrupts are disabled causes the
	  z80intPending flag to be set, and an interrupt will be caused after
	  the execution of EI and the next instruction. See "IMPORTANT NOTE
	  ABOUT INTERRUPTS" below
	* The instruction after EI executes fully before interrupt status is
	  checked. (as does a real Z80)


Revision 3.1:

	* Fixed bug in memory dereference when handler was set to NULL (keeps
	  system from crashing or faulting)
	* Removed the only stricmp() from the entire file and replaced it
	  with strcmp() so that stdlibs without it will compile
	* Changed cyclesRemaining > 0 to cyclesRemaining >= 0 to be compatible
	  with the ASM core
	* Removed additional sub [dwCyclesRemaining], 5 at the beginning of
	  mz80exec() (ASM Core only). Increases timing accuracy.
	* NMIs And INTs add additional time to dwElapsedTicks as it should
	* mz80ReleaseTimeslice() Sets remaining clocks to 0 instead of 1


Revision 3.0:

	* All instructions validated against a real Z80. Used an ISA card
	  with a Z80 on it to validate flag handling, instruction handling,
	  timing, and other goodies. The only thing not implemented/emulated
	  is flag bit 3 & 5 emulation. Believed to be 100% bug free!
	* 80% Speed improvement over version 2.7 of mz80
	* z80stb.c Removed. Use -c to emit a C version of mz80! API compatible!
	  Note that this is mostly, but not fully, debugged, so consider the
	  C version a beta! It's at least healthier than z80stb.c was. The C 
	  version does not include the undocumented Z80 instructions.
	* mz80nmi() No longer trashes registers it uses when using -cs
	* IN/OUT Instructions work properly when using -16
	* IN A, (xxh) uses A as high 8 bits of I/O fetch address when using -16
	* IM 0/IM 1 Description in documentation fixed
	* Sizes of all context registers increased to 32 bits - for speed!
	* IFF1/IFF2 Now properly emulated
	* JR Instruction offset can fetch from $ffff and properly wrap
	* LDIR/LDDR Instruction now won't go to completion - instead it will
	  run until BC=0 or the # of cycles to execute have expired. These
	  instructions used to run to completion - even beyond the # of cycles
	  left to execute
	* INI/IND/INIR/INDR countdown bug fixed - it was decrementing B twice
	  for each IN! Whoops!
	* If you specify NULL as a handler address to a memory region, mz80 will
	  use vpData as a pointer to where that block of data resides. Quite
	  useful for multiprocessor emulations that share the same memory.
	* EDI Now keeps track of cycle counting for faster execution
	* Modified memory region scanning code to use 32 bit registers instead
	  of their 16 bit counterparts
	* Get/SetContext() uses rep movsd/movsb. Insignificant overall, but
	  why waste the time?
	* Debugging routines added. See the "DEBUGGING" section below for more
	  information. NOTE: The debugging routines are not yet available in
	  the C emission.
	* Timing done slightly differently now. Mz80 now executes one 
	  instruction past the timing given on input. For example, mz80exec(0)
	  will cause a single instruction to be executed (thusly -ss was
	  removed).

Revision 2.7:

	* Fixed OTIR/OTDR/INIR/INDR instructions so their 16 bit counterparts
	  work properly
	* Emulation core 30-70% faster overall than 2.6 due to optimization to
	  the timing routines
	* Replaced word reads/writes with a special word write routine rather
	  than the standard calling to read/write byte functions
	* z80stb.c (the C equivalent of mz80) compiles properly now
	* Fixed OS/2 text/segment issue
	* Fixed bug in set/getCPU context that ensures that ES=DS and avoids
	  crashes. Caused crashes under OS/2 and other OS's

Revision 2.6:

	* Emulator core 5-30% faster overall. Some 16 and 8 bit instructions
	  sped up when using their 32 bit equivalents.
	* Fix to -l so that proper labels without leading and trailing 
	  underscores so Linux/FreeBSD compiles will work properly
	* Single step now executes the # of instructions passed in to z80exec()
	  instead of just 1 as it had in prior releases. This is only active
	  when the -ss option is used.
	* The -nt option was added. This will cause the timing information to
	  not be added in, speeding up execution. Warning: Only do this if your
	  emulated target does not require instruction timing!
	* Updated documentation errors
	* C Version of mz80 (mz80.c) that is API compliant is distributed with
	  the archive (With kind permission of Edward Massey).

Revision 2.5:

	* Fixed an unconditional flag being cleared in the ddcbxx instructions.
	  It caused Donkey Kong's barrels to not roll.

Revision 2.4:

	* Fixed improper HALT handling (didn't advance the PTR when it should)
	* Fixed SRL (IX+$xx) instruction so that carry wasn't trashed
	* Fixed single stepping problems with it giving too much time to 
	  any given instruction
	* Fixed half carry flag handling with 16 bit SBC and ADD instructions
	* Fixed DAA emulation so that parity flags weren't getting trashed

Revision 2.3:

	* Fixed many stack handling bugs
	* Timing problems fixed. The prior version was causing massive 
	  overruns on maximum timeslices with some insutructions.

Revision 2.2:

	* Fixed a bug in CPI/CPD/CPIR/CPDR that mishandled flags
	* All known bugs are out of mz80 now
	* Added the -cs option to route all stack operations through the
	  handlers (required for games like Galaga)

Revision 2.1:

	* Fixed a bug in CPI/CPD/CPIR/CPDR that caused intermittent lockups.
	  Also fixed a bug that caused erratic behavior in several video games.
	* Added INI/IND/INIR/INDR instruction group
	* Added OUTI/OUTD/OTIR/OTDR instruction group

Revision 1.0:

	* First release! The whole thing is new!


ASSEMBLING FOR USE WITH WATCOM C/C++
------------------------------------

Watcom, by default, uses register calling conventions, as does MZ80. To
create a proper emulator for Watcom:

	makez80 MZ80.asm -x86

From here:

	nasm -f win32 MZ80.asm

Link the MZ80.obj with your Watcom linker.


ASSEMBLING FOR USE WITH MICROSOFT VISUAL C++ AND BORLAND C++
--------------------------------------------------------------------

Visual C++ and Borland C++ use stack calling conventions by default. To
create a proper emulator for these compilers:

	makez80 MZ80.asm -s -x86

For Visual C++ or Borland C++:

	nasm -f win32 MZ80.asm

Link with your standard Visual C++ or Borland C++.


ASSEMBLING FOR USE WITH DJGPP, GCC/FREEBSD, OR GCC/LINUX
--------------------------------------------------------------------

DJGPP Uses stack calling conventions:

	makez80 MZ80.asm -s -x86

To assemble:

	nasm -f coff MZ80.asm

Link with your standard DJGPP linker. The same holds true for GCC under
FreeBSD or Linux. If you're using GCC, use the -l option to generate "plain"
labels so that gcc's linker will properly link things.


MAKEZ80 COMMAND LINE OPTIONS
----------------------------

-s	- Use stack calling conventions (DJGPP, MSVC, Borland, etc...)

-cs	- Force all stack operations to go through the Read/Write memory handlers.
	  This slows things down, but is useful when needed.

-16	- Treat all I/O input and output as 16 bit (BC)

-l	- Create 'plain' labels - ones without leading and trailing underscores

-nt	- Do not generate timing code - this speeds the emulator up, but the
	  downside is that no timing info is available.

-c	- Emit a C mz80 emulator (API Compatible with the assembly version - 
	  handy for porters!)

-x86	- Emit an assembly (x86) mz80 emulator

-os2	- Generate OS/2 compatible segmentation


IMPORTANT NOTE ABOUT INTERRUPTS
-------------------------------

A minor change was made between the 3.1 and 3.2 versions of makez80 in the
way that interrupts were handled.

On a real Z80, the !INT line is a level triggered interrupt, meaning that if
the interrupt line is held low, the Z80 will continue to take interrupts 
immediately after the instruction after the EI instruction is executed until
the interrupt line is high again.

In 3.1, if an interrupt came in and interrupts were disabled, the interrupt
would never be "latched" for later execution. The Z80 does not have any
internal latching capabilities, however external hardware often does hold
the interrupt line low until the interrupt is executed, in effect, a latch.

I've only found one video game so far that requires the "raising/lowering"
of the interrupt line (Ataxx). In the games that I've tried, it has improved
performance, in some cases drastically, and in others not at all. This can
be accounted for by interrupts being taken now, where they were being dropped
in prior mz80 releases.

mz80 Emulates the most commonly used scenario. Now when mz80int() is executed
and a nonzero value is returned (indicating interrupts were disabled), it
will set z80intPending, and the interrupt will be taken after execution of
one instruction beyond the EI instruction.

So now, if mz80int() returns a nonzero value, that means an interrupt is
latched. If clearing this latch is desired or the old behavior of 3.1 is 
desired, make a call to the mz80ClearPendingInterrupt() call. It's a 2 
instruction call that has extremely small overhead and will not affect 
performance in any measurable way.

In any case, MZ80 will now execute one instruction after EI regardless of
how much time is available to avoid the possibility of an interrupt request
coming in directly after the EI instruction. 


STEPS TO EMULATION
------------------

NOTE: -16 Is a command line option that will treat all I/O as 16 bit. That
is, in an instruction like "IN AL, (C)", the addressed passed to the I/O
handler will be BC instead of just C. Bear this in mind when considering your
emulated platform.

There are a few steps you want to go through to get proper emulation, and a
few guidelines must be followed.

1) Create a MZ80CONTEXT

2) Create your virtual 64K memory space using whatever means of obtaining
   memory you need to do.

3) Set mz80Base in your context to be the base of your 64K memory space

4) Load up your image to be emulated within that 64K address space.

5) Set z80IoRead and z80IoWrite to their appropriate structure arrays. Here's
   an example:

struct z80PortRead ReadPorts[] =
{
	{0x10,	0x1f,	SoundChip1Read},
	{0x20,	0x2f,	SoundChip2Read}
	{(UINT32) -1,     (UINT32) -1, NULL}
};

When an IN instruction occurs, mz80 will probe this table looking for a
handler to the address of the "IN" instruction. If it is found in the list,
it's up to the handler to return the proper value. Otherwise, a value of
0ffh is returned internally if no handler for that I/O address is found. In
the case above, SoundChip1Read is called when the I/O address is between 0x10-
0x1f. A similar structure is used for I/O writes as well (OUT):

struct z80PortWrite WritePorts[] =
{
	{0x20,	0x2f,	SoundChip2Write},
	{0x30,	0x36,	VideoCtrlWrite},
	{(UINT32) -1, 	(UINT32) -1, NULL}
}

Of course, this does the opposite that the z80PortRead struct, and instead
looks for a handler to hand some data to. If it doesn't find an appropriate
handler, nothing happens.

6) Set mz80MemoryRead & mz80MemoryWrite to their appropriate structure
   arrays. Here is an example:

struct MemoryWriteByte GameWrite[] =
{