Help with compile

[JimDrew]

Can you someone look at the attached code and tell me exactly what I need to compile it on an Amiga?

I know absolutely *nothing* about C code, except what it looks like when you disassemble the assembly code.

I need to compile the tester program, which is an instruction set tester that was written in C. I need to run this on an Amiga (and Amiga emulators) to test the instruction set accuracy. Thank you!!

Jim

[demolition]

From looking inside the makefile, it looks to me like it is designed to be compiled with GCC on Windows or OSX.

[JimDrew]

Yes, but the docs in the github elude to the fact that it was created on an Amiga... the samples were for sure. I would assume that it can be compiled to run on an Amiga changing the make file. I installed gcc on the Amiga and tried to compile it, but got all kinds of errors about missing .h files. My goal is to turn this into 68K assembly that can be used on Amiga and Mac emulators to test the instruction set accuracy.

[kamelito]

do you mean that they never were accurate in your past products?

[rkidd7952]

Hi Jim,

Did you use gnu make along with gcc to compile? From a quick look a the source, it doesn't appear to use any exotic system include files, but it uses a gnu make extension to build the compiler options that specify where to find the .h files in the source tree.

I don't know much (anything, really) about the state of gcc/g++ on the Amiga, but the makefile refers to version 4.7 and lists a couple language options that may not be available on the Amiga (-std=gnu99, -std=gnu++0x). I'm not sure if the source actually uses new language features that require these options.

There are only a few source files in the base executable, so you could try compiling by hand rather than chasing down the correct version of make. Here are the compiler commands the makefile invokes when I run make DEBUG=1 on my mac.

g++ -I. -Ilib/ -O0 -g -Wall -c 68k/68k.cpp -o obj/68k.o
g++ -I. -Ilib/ -O0 -g -Wall -c 68k/exception.cpp -o obj/68kexception.o
g++ -I. -Ilib/ -O0 -g -Wall -c main.cpp -o obj/main.o
g++ -I. -Ilib/ -O0 -g -Wall -c results.cpp -o obj/results.o
g++ -I. -Ilib/ -O0 -g -Wall -c tester.cpp -o obj/tester.o
g++ -o out/portable68k obj/68k.o obj/68kexception.o obj/main.o obj/results.o obj/tester.o

I edited the makefile to remove the -std=gnu* options. I got a couple compiler warnings (signed vs unsigned comparison, recommended parentheses around an expression), but nothing too dire. I suspect the -std=gnu* options aren't needed, but I haven't tried to run the program to find out.

Those commands use unix syntax to refer to the current directory, so you'll need to change -I. appropriately. I think that'd be -I"" or -I "" on the Amiga.

Robert

[JimDrew]

Originally Posted by kamelito

do you mean that they never were accurate in your past products?

Ha! No, that's not the case. I never saw any issues with the 68040 core I did for the PC based Mac emulation. I would like to check WinUAE and FPGA based Amiga emulations for instruction execution and timing accuracy - vs. a real Amiga.

Robert, can you attach the changes you made. I know absolutely nothing about C code. I couldn't write a 'hello world' program in C if my life depended on it. I could do it in a dozen different assembly languages, but not C. I have no use for portability so C has never been a requirement for me.

[alkis]

I've cross-compiled it, but I don't think it runs well. (on emulated A1200)
It ends up with "Errors: 352".

If I compile it for linux, I get no error(s) on the output. "Errors: 0"
Is that supposed to happen?

Oh, and it's not C code. It's c++

[JimDrew]

Are these compiler errors or is the results from running it? If from running it, can you attach the Amiga executable so I can disassemble it and check it? Thanks!

C/C++/C#/C!/CC = it all equals some form for C.

[ajk]

The "Errors: xxx" is output from the compiled program after running all the individual tests.

[alkis]

Those were results from running it.

It's in the zone for you. (portable68k.zip)

Needs ixemul.library and a pretty huge stack.
I was getting crashes with 50k stack. So I went 'stack 2000000' (2mil) and got it running.

[JimDrew]

Thanks... I will check it out!

Yeah, there is definitely something wrong with the compiled code. A real Amiga shows 352 errors as well.

[Hewitson]

Originally Posted by JimDrew

I couldn't write a 'hello world' program in C if my life depended on it.

Writing simple C programs isn't difficult.

#include <stdio.h>

main() {
printf("Hello World!\n");
}

[JimDrew]

#include system.gs

Main:  lea HelloString(pc),a0
       CALLFUNC OutputString
       rts

HelloString .db "Hello World!",LF

That makes way more sense to me... and it's smaller code and faster!


I am still stuck trying to get the portable68k code to work on an Amiga. Anyone have any ideas? Thanks!!

[meynaf]

I am wondering how a compiled program can test cpu compatibility... Hum... Isn't this the task for pure asm instead ?

[demolition]

I haven't looked into the source, but it could be using some inline asm at critical points?

[phx]

The 68k subdirectory contains a simple 68k emulator.

So I doubt the program makes any sense for the original poster, as just the emulator will be tested.

[demolition]

In that case, you might as well run the program on a PC rather than try to make it run on Amiga?

[JimDrew]

The goal is to remove the "tester" portion so that it will run on a real 68K, not through the emulator. A bunch of us are after a means to test 68k CPU cores (software and FPGA based). This is the only thing I can find that will test all of the 68K instructions. If I can get this compiled to run on an Amiga, I can disassemble it and turn it into a 68K hunk that can be inc'd into any test code.

[demolition]

But testing an emulator and the real thing are very different tasks. With the emulator, the testing software can control the time stepping in the emulator and log all required data from the outside. It needs to run much faster than the system it is testing. This is somewhat equivalent to hooking up a digital analyzer to a real Amiga and logging all relevant signals.

I don't see how it is possible to reliably test the system you're running on, at least not if you're going for clock accurate timings.

[JimDrew]

We don't care about clock cycle accuracy. We are after something that will test every instruction with every possible addressing mode, and compare the results to what a real 68K would have. There are several bugs in the TG68K code for example with certain instructions being used in certain ways.

There are tons of testers for Z80, 6502, 6809, etc. but nothing we can find (other than this program) that will test the 68K series.