Foxi Project

[Fizyk]

I've been trying to diagnose why my external code execution fails... so far no luck...

I'm just guessing, but you probably have problems with the way CPU compiles code. I told you once on IRC "just jump to the code", and while that is exactly what you should do, a simple jmp won't suffice.

Let's say you have a CPU with external code connected to the MemBus of the main CPU. Then your external code is at address 65536. You could do "jmp 65536" and your external code would start executing, but... it will fail the first time it encounters an instruction, that used a label.

The problem is that when CPU compiles code, it translates all labels to addresses, but it assumes, that you program starts from address 0, and when you connect it as external code, it starts from address 65536! There are two ways to make it work:
1. Use segmentation.
2. Use "offset" macro.

With using segmentation, you make the CPU execute the code, like it was starting from address 0. Instead of jumping to 65536 (which is 0:65536), you jump to 65536:0 using jmpf:

jmpf 0,65536 //offset first, then segment

This changes CS to 65536 and resets IP to 0. If your external program then encounters "jmp label", it changes IP to label, but leaves CS unchanged, so it will really jump to 65536+label - exactly where you want it to jump. If you had used "jmp 65536", CS would still be 0 and it would have jumped somewhere into your main CPU.

The way with "offset" macro is very simple. Just put "offset 65536" on the beginning of your external code - this will make the compiler know, that the code is shifted by 65536 bytes and all label addresses will be translated accordingly.

The second way has a drawback. While you can still run the external program as a standalone code when using the first way, you cannot do that with the second way, unless you remove the "offset" macro from the code.

I hope this helps you

[Black Phoenix]

I'm just guessing, but you probably have problems with the way CPU compiles code. I told you once on IRC "just jump to the code", and while that is exactly what you should do, a simple jmp won't suffice.

Let's say you have a CPU with external code connected to the MemBus of the main CPU. Then your external code is at address 65536. You could do "jmp 65536" and your external code would start executing, but... it will fail the first time it encounters an instruction, that used a label.

The problem is that when CPU compiles code, it translates all labels to addresses, but it assumes, that you program starts from address 0, and when you connect it as external code, it starts from address 65536! There are two ways to make it work:
1. Use segmentation.
2. Use "offset" macro.

With using segmentation, you make the CPU execute the code, like it was starting from address 0. Instead of jumping to 65536 (which is 0:65536), you jump to 65536:0 using jmpf:

jmpf 0,65536 //offset first, then segment

This changes CS to 65536 and resets IP to 0. If your external program then encounters "jmp label", it changes IP to label, but leaves CS unchanged, so it will really jump to 65536+label - exactly where you want it to jump. If you had used "jmp 65536", CS would still be 0 and it would have jumped somewhere into your main CPU.

The way with "offset" macro is very simple. Just put "offset 65536" on the beginning of your external code - this will make the compiler know, that the code is shifted by 65536 bytes and all label addresses will be translated accordingly.

The second way has a drawback. While you can still run the external program as a standalone code when using the first way, you cannot do that with the second way, unless you remove the "offset" macro from the code.

I hope this helps you

You can use address bus to auto-translate all addresses up by 65536, effetively allowing second way

[Fizyk]

Oh, I didn't know about the possibility with the address bus.

Also, I forgot to mention one thing: with the first way, it will still fail at memory references. Every #variable will refer to the main CPU, because variable names are also labels, and in this way label addresses are translated with the assumption that code begins with 0. This can be fixed by changing DS to 65536, as #variable really is DS:#variable - so when you change DS, this reference will be shifted by 65536 bytes and everything will be ok. The way with "offset" macro doesn't have this problem - offset already shifts the variables too for you.

[Black Phoenix]

Oh, I didn't know about the possibility with the address bus.

Also, I forgot to mention one thing: with the first way, it will still fail at memory references. Every #variable will refer to the main CPU, because variable names are also labels, and in this way label addresses are translated with the assumption that code begins with 0. This can be fixed by changing DS to 65536, as #variable really is DS:#variable - so when you change DS, this reference will be shifted by 65536 bytes and everything will be ok. The way with "offset" macro doesn't have this problem - offset already shifts the variables too for you.

Negative address bus offsets are your friends!

[-=Fox=-]

I've tried :

jmpf 0,65536

It didn't seem to do that job... I've tried several variants of it as well.

I didn't think about offsets, I'm not sure what I should try... So obviously I'm a little confused... I need concrete examples that work to really understand it.

So could I get a very small example of WORKING external execution code?

AND! (questions)

Does the external CPU (with code) need to have anything done to it, to better have its code run.

And does the code have to be formatted in a certain way, should their be defines or alloc's in external the code?

[Divran]

Are you using a graphics tablet to get the aim coordinates? I suggest using :toLocal() and :keyUse() for more epicness. I made an E2 that works just like a graphics tablet that way

[Fizyk]

So could I get a very small example of WORKING external execution code?

Main CPU:

mov ds,65536
jmpf 0,65536

External CPU:
Any code that you could execute in a single CPU

And of course that's with the assumption, that the external CPU is connected directly to the MemBus of the main CPU.

Does the external CPU (with code) need to have anything done to it, to better have its code run.

And does the code have to be formatted in a certain way, should their be defines or alloc's in external the code?

No and no.

[-=Fox=-]

AH! Sweet Fizyk, that makes sense.

I'll be updating the First post with a more updated version of the dupe *soon*

[-=Fox=-]

Hey! Just letting everyone know I'm still alive.

I've had MAJOR setbacks in internet access and getting on Gmod so hang on to your pants.

I have a new version of Foxi and I've got the first peripheral for it (a Modem!) and a program to work the new peripheral (which has a few bugs) but at any rate I'll continue to work the bugs out and TRY to get a release out to you guys, I know I haven't been active but its a limitation that I have to live with for the moment.

Hang in there!

[krazyfool]

I look forward to its epicness ^__^