openMSX "Accuracy"

Paladin (747)

29-10-2018, 08:58

First off, I have returned to MSX again, after stopped coding z80-assembly (demos) in the early nineties. One of the few in Norway, I think :-) Anyways, coming to see the scene is alive and breathing, is fantastic. I assume that emulators are a big part of the reason MSX is still going strong, and I applaude the work you guys are doing. openMSX is really a great product!

Can someone explain the difference between openMSX' "Very High" and "Cycle" in this comparison?
http://emulation.gametechwiki.com/index.php/MSX_emulators

By Grauw

Ascended (10768)

29-10-2018, 09:10

Idk, makes no sense to me.

openMSX is cycle accurate, the most accurate MSX emulator I know.

That MAME is in that list as "recommended" for MSX emulation is... interesting.

By Bengalack

Paladin (747)

29-10-2018, 09:24

Sounds good. I'll go with openMSX, and see if I can brush off some dust from the old skillz

By ARTRAG

Enlighted (6935)

29-10-2018, 09:35

The table at the link is misleading, at least about accuracy
Openmsx is way more accurate than bluemsx
The reason why you could prefer bluemsx is its interface and its integrated tools
But for coding demos you can't rely on it.

By ren

Paragon (1934)

29-10-2018, 10:17

+ blueMSX isn't in active development anymore.
A couple of years ago, when blueMSX was still actively maintained, both emulators were quite competitive & helped each other to improve.
openMSX has had quite some improvements & (exotic) emu fixes since that blueMSX misses.

Also, I believe, blueMSX's save-state feature is not 100% reliable. openMSX's should be, and has the awesome (& dangerous ;-)) reversebar™
Next to what ARTRAG is saying, I believe some people still prefer blueMSX's sound/music reproduction.

Wb btw, I reckon you've heard of the awesome Syntax Infinity, IO & Baltak Rampage demos?

By Bengalack

Paladin (747)

29-10-2018, 11:03

I haven't. But I probably should have. I can check it out now.

I stopped programming Z80 around 1992, I think. My friend Arne and I made some demos behind the name of Northwind up here in Norway, but we were not much in contact with the European scene. We just swapped demos with some Swedish, Danish and Finnish people. I do remember a group called ANMA though

Recently I made some dumps from the demos, using SCART->HDMI->HDMI-capture I did it as to preserve the work, as I assumed MSX was dead. I was wrong Here are some of them (and pardon the infantile scroll-texts, that's just how it was when you were 16 in the early nineties):

Example from MSX2 and FM-Pak
https://www.jottacloud.com/s/00762b2ee95d5b14642bf8b308b6577...

Example from MSX2 and PSG
https://www.jottacloud.com/s/007ee4eee21f7f5450cb0e001f530cf...

Old work on MSX1, with ripped PSG-music
https://www.jottacloud.com/s/007a74a10837b1d46be946c7b91f881...

By NYYRIKKI

Enlighted (6067)

29-10-2018, 14:42

They all come back... Sooner or later they all come back

I remember very well that I downloaded "Fear of an FM-planet" from Funet some time in 1997 or so and I was quite impressed to learn it was made in Norway. The whole Norway "MSX scene" has been quite a mystery to me ever since. I don't believe MSX2 computers have been sold there and yet there was this kind of stuff coming out in "early days" (pretty much meaning time before Internet was an affordable hobby)... I hope you could light up this mystery to me a bit. How big was the user base? You also probably have some insight information about the copy parties etc. ? Was the MSX2 scene built around home upgraded SVI-738's or did you import some Phillips machines from Netherlands?

To avoid this being just a pure topic hijack, I must say that the comparison results can be pretty much explained from the fact that the results are from 2005... During last 13 years the openMSX has improved constantly while the BlueMSX project has been just idling. Although BlueMSX is outdated, it is still alive among users due to the fact that it is very easy to approach compared to openMSX that really feels & smells like developer tool instead of end user product. From technical side it usually does everything you need and much much more, but you need to be quite a hacker to master the power you are given.

From accuracy point of view I find it quite unlikely that you will encounter any problems with openMSX even as a developer. Every now and then some developer may find some weird quirk that is not emulated correctly, but these are extremely rarely anymore in MSX1/MSX2/MSX2+ standard features and generally they are pretty much fixed as soon as they are properly documented.

By TomH

Champion (361)

29-10-2018, 15:23

I have an editing account at emulation.gametechwiki.com which at least means I've tried to ask about and to divine the intended meanings of the various columns. Though I'm far from being the authority.

My subjective basic impressions that may or may not be correct are:

nobody wants to claim an emulator has perfect accuracy, because there's never a formal specification to ratify emulators against, so probably none of them is, and definitely none of them provably is;
as a rule of thumb, they use 'very high' literally to mean that — it's a qualitative assessment;
whereas 'cycle' tends to be repetition of an implementation detail that has been conflated with a qualitative assessment;
the recommended column is to do with emulation quality only tangentially, such as that affects user experience. Broad host platform support, features like save states and debuggers, etc, all factor in. I think it's about the probability that you'd recommend it to somebody with no prior knowledge of the person's use case. Oh, and they seem to love it if you have a libretro 'core'.

I've been able to find no logic whatsoever to the ordering of those tables. Things that are recommended tend to float to the top, but not always, and otherwise it's a bit of a free-for-all.

I also explicitly queried and, unlike Wikipedia, they're fine with you providing the Wiki content for your own emulator. I mean, as long as you're sensible, obviously. I put my own CLK into that table, and I put the X in the recommended column. I also added the http://emulation.gametechwiki.com/index.php/Accuracy#Subcycle_accuracy subsection in an attempt to help to disabuse people of the idea that "cycle accurate" means "perfectly accurate" (albeit from a "cycles are still arbitrary discrete quantums" point of view than an implementation detail versus implementation quality argument).

The other thing about that wiki is that a lot of its traffic seems to be people exploring platforms they've no inherent familiarity with via emulation. That's why I added the 'generations' subsection. It'd probably be a good thing to add to the table too.

By MsxKun

Paragon (1124)

29-10-2018, 21:13

Well, I got some simple code that works on an emulated Casio PV-7 while it crashes on a real Casio PV-7.
And it's really simple and must be my fault that it crashes. But funny the emulated one works when I shouldn't.
When I have time to check and do a quick tiny test and figure where's the problem, then I could report to openMSX team.

And that, for example, sprite moving in steps of 2 pixels (for example) is not smooth as it is on a real machine.
But other than that, openMSX works great 99% of the time and it works on linux too, and when you get used to it, the debugger can do lots of things and the console is a big help too.

But about accuracy, never trust 100% ANY emulator. Unless, maybe, if it works native on a computer, without any OS messing under it.

By Manuel

Ascended (19462)

29-10-2018, 22:34

The OS has no influence, except for timing smoothness.

By MsxKun

Paragon (1124)

29-10-2018, 23:08

Manuel wrote:

The OS has no influence, except for timing smoothness.

The most important, probably.