Page 1 of 1

Should "lagless MAME" (aka Shmupmame) be allowed?

Posted: Fri Nov 26, 2010 12:13 am
by Zhorik
The question has come up on "The official awesome replay thread" about whether use of "lagless MAME" (aka Shmupmame) should be allowed for MARP submissions. Here's the thread: viewtopic.php?f=2&t=11642&p=52474#p52474

The recording in question is: http://replay.marpirc.net//inp/e/3/c/pr ... _win99.zip

More information about Shmupmame can be found here: http://shmups.system11.org/viewtopic.php?f=1&t=30659

MARP rule 3b states, "Since April 15, 2007, use of WolfMAME is required if you take first place in any game", so clearly Shmupmame is not allowed for first place scores. However in general for non-first place scores, other versions of MAME are currently allowed.

The potential problem is that according to Kale, "if you use a "lagless" version you are using a version on which you remove stuff from the original HW ... original Dodonpachi Arcade version has that lag and it's called sprite buffering. Sprite tables written to the work RAM are then copied thru a DMA, that DMA isn't instant, it usually takes a latency of 20 to 40 ms."

If Kale's explanation is correct, than I don't think Shmupmame should be allowed for MARP submissions.

First, it means that Shmupmame is "fixing a problem" that is actually being correctly emulated in MAME and is intentionally "breaking" the emulation and making it not behave like the arcade game. The rationale for doing this is summarized by Sikraiken as "OS overhead (and peripheral latency) add lag. With sprite/layer buffering removed, the lag is more accurate to what the actual PCB has since the actual PCB doesn't have to worry about OS overhead and peripheral latency." I understand the desire by players to try and make play "feel" as much like the arcade as possible, but IMO breaking MAME's correct emulation of the sprite/layer buffering is not the correct approach, at least for MARP.

Second, it means that Shmupmame users are at an advantage to players using any other version of MAME since they have less lag than anybody playing in any "normal" version of MAME or WolfMAME. I don't think that is fair, even for non-first place submissions.

There's been some discussion on this on "The official awesome replay thread", but I'm hoping we can move that discussion here.

-George

Re: Should "lagless MAME" (aka Shmupmame) be allowed?

Posted: Fri Nov 26, 2010 12:25 am
by Haze
As I stated in the other thread it definitely shouldn't be allowed, it's effectively breaking the emulation, and even as MARP stands old versions with broken emulaion have their scores 0'd.

It's presenting an advantage to players using this build, and even an advantage to players when compared to the PCB in situations where sprite->bg sync are important for hiding secrets and such.

Re: Should "lagless MAME" (aka Shmupmame) be allowed?

Posted: Fri Nov 26, 2010 8:12 am
by mahlemiut
I find a lack of sync between backgrounds and sprites irritating, to be honest. I know some systems do have some graphic elements out of sync with one another, but when the goal is accuracy, it has to be how the original arcade was.

I remember when Metamoqester was first added... I think it was the Titicaca stage, had horrible lag between the parts of the background. Wouldn't want to be prone to seasickness playing that, and I can't see why anyone would want to subject themselves to that just for the sake of seeing sprites about 17ms earlier... Thank god that was fixed.

Re: Should "lagless MAME" (aka Shmupmame) be allowed?

Posted: Fri Nov 26, 2010 8:56 am
by Haze
yeah it can be irritating (personally I do find it irritating) but on say 'gratia' in older MAME versions without sprite buffering you know exactly which objects you can get hidden powerups from because they lag a frame behind the actual tilemaps. It's a HUGE advantage if playing that game, and completely not fair (although I believe it's broken in other ways on current versions) I'm not sure if they de-lag m32.c, I don't care to find out, but that is a prime example of where such hacks will make the game obviously 'broken' and easier to play.

as I said elsewhere. I think going forward MARP should only allow official MAME (including SDL, which is on the other major platforms) and WoflMAME. The former for convenience, the later to 'verify' thing better.

Unofficial builds can have all manner of hacks hidden away which might change how the game plays, even in very subtle ways, supporting them makes little sense IMHO.

Re: Should "lagless MAME" (aka Shmupmame) be allowed?

Posted: Fri Nov 26, 2010 11:53 am
by Kale
My POV is already on the other thread, so I vote against lagless MAME as well here.

Re: Should "lagless MAME" (aka Shmupmame) be allowed?

Posted: Fri Nov 26, 2010 4:26 pm
by MJS
In my opinion, rule 3b is enough.

Disallowing the use of certain mame versions will just lead to players like Prometheus to not share great replays like DoDonpachi's with us.

Instead, I think it's more constructive to allow submission and contact them (as Chad and Phil already do by writing in the description box) to tell them about WolfMAME, or if they intentionally used another version, to understand why.

About Shmupmame, I don't beleive it gives any significant advantage. Your inputs responds 16, maybe 32 milliseconds faster... so? To have a good run at a game you need more than reflexes. If you really are a player where that kind of delay makes a difference, then go buy the real game and go for the world record! :mrgreen:

Ok, sorry for the jokes. To sum up, I think Shmupmame does dirty hacks (to gain small benefit) and I wouldn't recommend anyone using it. However, I think we should allow replays done with it, as with any other version of mame out there as long as the playback works. As Chad once said, MARP is a repository of replays. In the case of people using hacked versions like this one, convince them not to use them.