A x86 emulator test suite / emulator tester? I don't think one exists.

I've seen them for 6502, 6809, 8080, 8085, Z80, 68K, PPC... I don't remember one for x86.

This looks interesting:
https://code.google.com/p/corkami/source/brow … es32.asm?r=1569

[EDIT] Qemu may have a partial x86 test suite, try them?

Is there a thread that lists the best tools for benchmarking each operating system?

Maybe? But it's not this one. This thread is actually about torture-testing x86 CPU emulators, being as it is properly placed within the "PC Emulation" subforum 😀

Thanks for the link. The Code looks interesting and as a start, I will pick it up and expand it for my needs.

Thanks for your help and your answers.

Best Regards, donkeyulli

No problem!

I'd recommend that if you are building your own x86 emulator test suite that you look to other CPU emulator test suites for broad inspiration:
I have copies of Gwenole Beauchesne's GPL-licensed M68K-Tester and PPCTorture / PPC-Tester and their test results, though that might be still floating out there.
http://web.archive.org/web/20090417202939/htt … jects/68ktester
http://web.archive.org/web/20090131234602/htt … jects/ppctester

ZEXALL (Z80): http://mdfs.net/Software/Z80/Exerciser/

8080/8085: http://www.idb.me.uk/sunhillow/8080.html

6502: http://visual6502.org/wiki/index.php?title=6502TestPrograms

I'm sure there's others. I still need to investigate if Qemu has one as rumored.

You might want to try Topbench.

Stiletto wrote:

I'm sure there's others. I still need to investigate if Qemu has one as rumored.

Got it! Yes, QEMU does/did. Gimme a little bit, gonna do a write-up.

Is there an existing available suite of tests I can use while developing an x86 emulator to check my CPU emulation? (ASM, I assume?) I didn't see any, much less a 'standard' set. I assume the various emulator projects have some, but not sure if it's alright to snag them for my own purposes - assuming I can locate them!

Ideally there'd be tests for each of the cpu generations, I assume, and 16 and 32 bit versions. (Yes, yes, and 64 bit, but can't say I'd use them...)

If it doesn't exist, I wonder if it would be usefully enough to start creating? I had a start 10+ years ago when I was working on a Dos emulator, but it was far from complete, very specific, and DosBox's grand success rendered my work happily irrelevant and I moved on to other stuff. I'm curious about 16 bit Win31 stuff and thought about starting a new emulator project for that, but would love to have a set of reference tests to work from if they exist. 😀

Thanks for reminding me, I'd forgotten about this thread.

The answer in general is yes, Qemu has one.
http://git.qemu.org/?p=qemu.git;a=tree;f=test … 0754976;hb=HEAD

A MAME/MESS developer teammate of mine had this to say about his 01/18/2012 compiles on 7/5/2012:

Firewave wrote:

I did some new compiles based on the qemu-1.0 test-i386 source with DJGPP. The archive contains a version with all tests enabled (as far as possible on Windows), only 486 ops, only 386 ops and a 386 version without any FPU. The modified sourcefile is also included.

It's not a complete test, but it is quite helpful to test i386-based emulation cores.

Since qemu 1.0, there were minor additions and adjustments. It will build on Linux, but there is no way in the official test to disable feature of smaller CPUs, that's why I did the changes. Also it won't work on Windows and that's the main reason I changed it.

It's possible this situation has changed since July 2012 though, so check the qemu git.

On the other hand, maybe you can collaborate with donkeyulli to create a new one.

Of course. 😀 Thank you very much, good stuff to dive into.

I may pester Firewave, perhaps he's done more recent builds. 😀

(FYI Marvin people: this utility is a bit pointless to run in hardware...)

I apologize for bumping a rather dead thread, but have some questions over this "test" utility

It seems to spew out a bunch of information, but do we have a official reference to what the values should be?
Or should I just go an run it on real hardware and compare the results.

I ask because I ran this on Bochs, QEMU, DOSBox, PCem, and my machine and got some variations, mainly floating point stuff, between all of them (Besides Bochs and My Machine, same results)

gamax92 wrote:

Or should I just go an run it on real hardware and compare the results.

Yup. You'll even get different results on different CPUs. Examples are flags behavior and SH[RL]D.

I tried running the attachment by Stiletto on my emulator, UniPCemu. It seems to crash most of the time, due to an IRQ0 firing due to interrupts being available while it's loading the executable? Is that supposed to happen with CWSDPMI?

(sorry if I'm necroposting here, but it's related)