8085 simulator in C

[activearun.83]

Hi everyone,

Can anyone give some idea about developing an 8085 simulator
program in C? What techniques can be handled to develop this?

 

[Walter Roberson]

Can anyone give some idea about developing an 8085 simulator
program in C? What techniques can be handled to develop this?

-Personally-, I wouldn't bother unless it was an assignment or
a pet learning project (and even then, I'd reconsider whether
8085 was the right choice of processors.)

If you wanted an 8085 simulator, I'd look around and try out
http://sourceforge.net/projects/sim85
or Vaneet Singla's 8085 simulator (available in a wide variety of places)

 

[user923005]

Hi everyone,

Can anyone give some idea about developing an 8085 simulator
program in C? What techniques can be handled to develop this?

http://www.google.com/search?hl=en&q=8085+simulator+program+in+C&btnG=Google+Search

 

[mstorkamp]

-Personally-, I wouldn't bother unless it was an assignment or
a pet learning project (and even then, I'd reconsider whether
8085 was the right choice of processors.)

I don't think 8085 is such a bad choice. Very simple instruction set.
I did this one weekend about 20 years ago just for the heck of it. I
had an old single board PC I wanted to emulate on a VAX.

The OP wanted suggestions on techniques to do this. As I recall, the
upper nibble of the opcode pretty much tells you what the instruction
will do, so I made an array of 16 functions and used the upper nibble
to index into the array. Then I had 16 individual functions to handle
the operations. The lower nibble is mostly for addressing, so that too
can index into an array for the correct register. Setting the flags
properly was the trickiest part.

 

[Ben Pfaff]

The OP wanted suggestions on techniques to do this. As I recall, the
upper nibble of the opcode pretty much tells you what the instruction
will do, so I made an array of 16 functions and used the upper nibble
to index into the array. Then I had 16 individual functions to handle
the operations. The lower nibble is mostly for addressing, so that too
can index into an array for the correct register. Setting the flags
properly was the trickiest part.

This sounds like a fun instruction set on which to practice
writing a (static or dynamic) binary translator. (Of course,
there's no way to actually run the translated code without
invoking undefined behavior, from a comp.lang.c point of view.)

 

[Chris Hills]

I don't think 8085 is such a bad choice.

It is if you want o buy one... I am certain the are obsolete these
days,.

Very simple instruction set.
I did this one weekend about 20 years ago just for the heck of it. I
had an old single board PC I wanted to emulate on a VAX.

Exactly....20 years ago It might be technically nice but getting the HW
is another matter. As for the tools.... i will dust of the compiler on
the 5.25" floppies.....

 

[mstorkamp]

wrote:

It is if you want o buy one... I am certain the are obsolete these
days,.


Exactly....20 years ago It might be technically nice but getting the HW
is another matter. As for the tools.... i will dust of the compiler on
the 5.25" floppies.....

You can still get the Z80 at DigiKey. Same instruction set, even
easier to connect up (the address and data bus are not multiplexed
like on the 8085). But then the OP just wants to emulate it, not build
one (but I would suggest doing so anyway). And since he wants to do it
in C, if he has trouble along the way his questions will likely be on
topic for this group. As to the rest of this discussion...

 

[Jean-Marc Bourguet]

You can still get the Z80 at DigiKey. Same instruction set, even
easier to connect up (the address and data bus are not multiplexed
like on the 8085).

I think you are confusing 8080 and 8085.

 

[mstorkamp]

I think you are confusing 8080 and 8085.

Nope, the 8080 used an external clock generator (8224) and system
controller (8228) that were built into the 8085. There are a couple of
undefined opcodes in the 8080/8085 that the Z80 defined to extend the
architecture of the 8080/8085. It doubled the number of registers,
demuxed the IO, and added a few new registers. However, as long as you
don't use those undefined opcodes, it is 100% software compatible
(only the timing is different).

 

[Kenneth Brody]

This sounds like a fun instruction set on which to practice
writing a (static or dynamic) binary translator. (Of course,
there's no way to actually run the translated code without
invoking undefined behavior, from a comp.lang.c point of view.)

How so? You have a struct holding the emulator's CPU registers, and a
64K unsigned char array (or two 32K arrays, since I seem to recall the
standard not requiring support for 64K arrays) representing the memory.

Now, emulating the hardware which might be surrounding the CPU is
another story. (Though you're probably more likely to run into
"implementation defined" or "unspecified" behavior than actual UB.)



--
+-------------------------+--------------------+-----------------------+
| Kenneth J. Brody | www.hvcomputer.com | #include |
| kenbrody/at\spamcop.net | www.fptech.com | <std_disclaimer.h> |
+-------------------------+--------------------+-----------------------+
Don't e-mail me at: <mailto:[email protected]>

 

[Walter Roberson]

Ben Pfaff wrote:

How so? You have a struct holding the emulator's CPU registers, and a
64K unsigned char array (or two 32K arrays, since I seem to recall the
standard not requiring support for 64K arrays) representing the memory.

Ben is talking about binary *translator* -- i.e., that the 8085 binary
code be examined and converted to code native to the hosting system
and run directly as native code instead of by emulating each 8085
instruction. And he is correct that you can't actually run the
translated code without undefined behaviour, as C does not offer
any mechanism to compute native code and set that code into execution.

For example, C does *not* offer

unsigned char object_code[SomeSizeOrOther];
object_code[SomeLocationOrOther] = SomeNativeOpcodeOrOther;
((void(*)()) object_code)(); /* undefined behaviour! */

There is no C facility to convert an object pointer into a function
pointer.

 

[Scott Fluhrer]

Nope, the 8080 used an external clock generator (8224) and system
controller (8228) that were built into the 8085. There are a couple of
undefined opcodes in the 8080/8085 that the Z80 defined to extend the
architecture of the 8080/8085. It doubled the number of registers,
demuxed the IO, and added a few new registers. However, as long as you
don't use those undefined opcodes, it is 100% software compatible
(only the timing is different).

Oops, only 99.99% software compatible -- if you did an 8 bit add or
subtract, the 8080/8085 would save the parity of the result into the 'parity
flag', while the Z80 would save whether the operation overflowed into the
parity flag. That generally came up only if you were deliberately writing
code that would work differently on the different processors, but it is a
difference.

 

[Kenneth Brody]

Ben is talking about binary *translator* -- i.e., that the 8085 binary
code be examined and converted to code native to the hosting system
and run directly as native code instead of by emulating each 8085
instruction.

Ah. I was thinking "emulator" rather than "translator". Of
course, something that generates (portable) C source code which
is to be compiled elsewhere could qualify as a "translator".

And he is correct that you can't actually run the translated code
without undefined behaviour, as C does not offer any mechanism to
compute native code and set that code into execution.

For example, C does *not* offer

unsigned char object_code[SomeSizeOrOther];
object_code[SomeLocationOrOther] = SomeNativeOpcodeOrOther;
((void(*)()) object_code)(); /* undefined behaviour! */

There is no C facility to convert an object pointer into a function
pointer.

Is this actually UB, or is it "implementation defined"?

Also, isn't a compiler allowed to define the behavior which is UB
according to the standard? In other words, a compiler _could_
explicitly allow such a construct, with the caveat (of course)
that the pointer must point to valid machine code, to allow
things similar to Java's JIT[1] compiler.


[1] JIT == "Just In Time". As I understand it, the Java bytecode is
compiled into native machine code as the program runs.

--
+-------------------------+--------------------+-----------------------+
| Kenneth J. Brody | www.hvcomputer.com | #include |
| kenbrody/at\spamcop.net | www.fptech.com | <std_disclaimer.h> |
+-------------------------+--------------------+-----------------------+
Don't e-mail me at: <mailto:[email protected]>

 

[mstorkamp]

He is better off simulating the 8080 instruction set, which can
then be expanded into the Z80 set, which in turn can be expanded
into the 64180 set. The 8085 is a dead end.

I think you are confusing 8085 and 8086/8088.

 

[Scott Fluhrer]

Ben is talking about binary *translator* -- i.e., that the 8085 binary
code be examined and converted to code native to the hosting system
and run directly as native code instead of by emulating each 8085
instruction.

Ah. I was thinking "emulator" rather than "translator". Of
course, something that generates (portable) C source code which
is to be compiled elsewhere could qualify as a "translator".

And he is correct that you can't actually run the translated code
without undefined behaviour, as C does not offer any mechanism to
compute native code and set that code into execution.

For example, C does *not* offer

unsigned char object_code[SomeSizeOrOther];
object_code[SomeLocationOrOther] = SomeNativeOpcodeOrOther;
((void(*)()) object_code)(); /* undefined behaviour! */

There is no C facility to convert an object pointer into a function
pointer.

Is this actually UB, or is it "implementation defined"?

Undefined Behavior. "Implementation defined" means that the implementation
is required to document it, and the Standard places no such requirement.

Also, isn't a compiler allowed to define the behavior which is UB
according to the standard?

Of course. Undefined Behavior means the Standard places no requirements,
and so the implementation is free to do whatever it wants. If it wants to
define the behavior in a particular way, and additionally to document what
this behavior is, that's just fine.

In other words, a compiler _could_
explicitly allow such a construct, with the caveat (of course)
that the pointer must point to valid machine code, to allow
things similar to Java's JIT[1] compiler.


[1] JIT == "Just In Time". As I understand it, the Java bytecode is
compiled into native machine code as the program runs.

--
+-------------------------+--------------------+-----------------------+
| Kenneth J. Brody | www.hvcomputer.com | #include |
| kenbrody/at\spamcop.net | www.fptech.com | <std_disclaimer.h> |
+-------------------------+--------------------+-----------------------+
Don't e-mail me at: <mailto:[email protected]>

 

[Bill Leary]

You can still get the Z80 at DigiKey. Same instruction set, even
easier to connect up (the address and data bus are not multiplexed
like on the 8085).

Not the same instruction set. The common denominator is the 8080. The Z80
adds a lot of new instructions. The 8085 added a few new instructions.
Some of those new ones for the 8085 use the same op-codes as new ones for
the Z80, but do entirely different things. There were also a few flags
which responded differently on the 8085 than they did on the 8080 or Z80.
When we wanted a single executable to run on two or three of these devices,
we used to write "detector code" which nudged at these flags and
instructions to figure out which machine the code was running on.

If one want's to use other processors "based" at 8080 there's also the
64180, the Z180 (Z80's with MMUs and a few other new features) and the
processor in the original Game Boy, which included some, but not all, of the
Z80 additional instructions and, as I recall, a few of it's own, and I seem
to remember an Intel 8080 or 8085 based CPU intended specifically for the
embedded market which had a few other additional instructions specifically
intended to cut down on external I/O hardware.

- Bill

 

[Bart van Ingen Schenau]

For example, C does *not* offer

unsigned char object_code[SomeSizeOrOther];
object_code[SomeLocationOrOther] = SomeNativeOpcodeOrOther;
((void(*)()) object_code)(); /* undefined behaviour! */

There is no C facility to convert an object pointer into a function
pointer.

Is this actually UB, or is it "implementation defined"?

The C standard does not define the behaviour, therefor it is UB.
This does not preclude other standards (like POSIX), or the compiler
writers themselves to specify what behaviour will happen.

UB merely means that the C standard does not restrict the behaviour of
the construct. Therefor, you should not use such constructs in code
that must be portable to all possible C implementations, as you can't
know what the behaviour will be.

In practice do compilers also implement other standards, like POSIX, or
have people come to rely on certain behaviours which all may indicate a
certain behaviour for constructs that are UB under the C standard
alone.

Due to the topicality of this group, we only look to the C standard to
see what behaviour we will get, so we will mark thing as undesirable
(due to UB) which are fairly common in other environments.

Also, isn't a compiler allowed to define the behavior which is UB
according to the standard? In other words, a compiler _could_
explicitly allow such a construct, with the caveat (of course)
that the pointer must point to valid machine code, to allow
things similar to Java's JIT[1] compiler.

Yes, and other standards may even require that the compiler does so.

Bart v Ingen Schenau

 

[CINVESTAV]

hi

i'm interested in your idea. i would like to know if could deloveped the program in c. i have some ideas and i wonder if you could help me