MiSTer FPGA Forum

Chris23235 wrote: ↑Mon Feb 15, 2021 7:54 pm The Mame emulation of the Sega Saturn hardware (SSV) is a low level emulation and it runs fine on PCs for years now.

Mame is high level. Mame never been about accuracy.

darksakul wrote: ↑Mon Feb 15, 2021 3:57 pm The one issue with that line of logic is but because something that was thought impossible was made to happen, does not mean it apply to every scenario. etc.........

Guess we'll see in a year if we got a good working PS1 and Saturn core.

darksakul wrote: ↑Mon Feb 15, 2021 11:38 pm

Chris23235 wrote: ↑Mon Feb 15, 2021 7:54 pm The Mame emulation of the Sega Saturn hardware (SSV) is a low level emulation and it runs fine on PCs for years now.

Mame is high level. Mame never been about accuracy.

Mame is low level it has always been low level emulation and Mame always strifed for accuracy. I don't know where this "Mame [has] never been about accuracy" comes from, the Mame contributors are known for their love for accuracy for years now up to the point of implementing the emulation in a way that makes it unplayable performance wise.

Cycle accuracy is something completely different. Not every low level emulation is cycle accurate. These are different things.
Low Level emulation means (very simplified) emulating the individual instructions of the source platform's chips instead of interpreting them in a way that makes use of features in the chips of the emulation target platform. This has nothing to do with cycle accuracy.

Cycle accurate emulation is rare, this is true but a low level emulation of the Saturn doesn't have to be cycle accurate.

Mame was never accurate, you need a powerful CPU for software accuracy and Mame can run on a toaster powered by a potato

darksakul wrote: ↑Tue Feb 16, 2021 8:37 pm Mame was never accurate, you need a powerful CPU for software accuracy and Mame can run on a toaster powered by a potato

It most certainly doesn't. The versions of MAME ran on low powered hardware like ARM are 15 years old now.

MAME strives for accuracy but human readable C code has always been just as much as a priority. So compromises have been made to say the least and it is what it is. Anyway, to have a good experience all around with current versions of MAME you need a CPU with a Passmark single threaded performance score of around 2500 or more.

And with such a CPU many of the demanding games won't run fullspeed. Many games in Mame don't run fullspeed on any machine available because of the high demands of the low level emulation. If you want to see how demanding Mame can get have a look at Bio Freaks.

with a i7-8700 (three years old cpu) you have a solid 100% (max 150%) on bio freaks with mame

Good to know there must have been a lot of work done on the driver since I last tried the game last year on my Ryzen 1700 (Passmark score 2000). It was so far away from fullspeed then that I didn't thought they achieve fullspeed anytime soon. But maybe it was just an Intel vs. AMD thing, I saw some games run better on Intel CPUs in Mame.

breiztiger wrote: ↑Wed Feb 17, 2021 9:40 am with a i7-8700 (three years old cpu) you have a solid 100% (max 150%) on bio freaks with mame

That CPU isn´t precisely a toaster powered by a potato

breiztiger wrote: ↑Wed Feb 17, 2021 9:40 am with a i7-8700 (three years old cpu) you have a solid 100% (max 150%) on bio freaks with mame

This is still a very fast CPU. I actually don't even own a computer with a CPU this fast.

Mame strives for accuracy, but in actuality it's far from it

It is among the most accurate software emulators and the whole community owes a lot to the work that went into Mame. E.g. when everybody was fine with decrypted cps2 roms, people from Mame implemented the proper encrypted roms, when everybody was fine with the wrong enemy behavior in Bubble Bobble Mame corrected it. The list goes on.

Mame is great for what it is and for the people that like it = FPGA is great for what it is and for the people that like it

darksakul wrote: ↑Thu Feb 18, 2021 3:50 am Mame strives for accuracy, but in actuality it's far from it

Why don't you lend a hand to make it better then? You seem to know all about it's issues..

retrorepair wrote: ↑Thu Feb 18, 2021 11:54 am

darksakul wrote: ↑Thu Feb 18, 2021 3:50 am Mame strives for accuracy, but in actuality it's far from it

Why don't you lend a hand to make it better then? You seem to know all about it's issues..

Like the Linux kernel, MAME rejects almost all patches that come from outside their small group. MAME has grown so monolithic that patching the sound to work in one game will break behavior in 50 others. MAME merging with MESS made things 10x worse. Fixing accuracy in old games can turn into a major project.

dmckean wrote: ↑Thu Feb 18, 2021 5:47 pm

retrorepair wrote: ↑Thu Feb 18, 2021 11:54 am

darksakul wrote: ↑Thu Feb 18, 2021 3:50 am Mame strives for accuracy, but in actuality it's far from it

Why don't you lend a hand to make it better then? You seem to know all about it's issues..

Like the Linux kernel, MAME rejects almost all patches that come from outside their small group. MAME has grown so monolithic that patching the sound to work in one game will break behavior in 50 others. MAME merging with MESS made things 10x worse. Fixing accuracy in old games can turn into a major project.

I totally disagree. I've even got code in MAME. In fact, it spurred further improvements from others which was also accepted.

If your code doesn't make sense then it'll probably be rejected. If not, anyone can submit. Not sure where you got that idea from.

They've rejected my fixes in the past. Afraid that fixing something axiomatic might break a game (but no tests or proof that breakage occurred).
This was years ago, but it poisoned me on contributing.

retrorepair wrote: ↑Thu Feb 18, 2021 6:30 pm

dmckean wrote: ↑Thu Feb 18, 2021 5:47 pm

retrorepair wrote: ↑Thu Feb 18, 2021 11:54 am

Why don't you lend a hand to make it better then? You seem to know all about it's issues..

Like the Linux kernel, MAME rejects almost all patches that come from outside their small group. MAME has grown so monolithic that patching the sound to work in one game will break behavior in 50 others. MAME merging with MESS made things 10x worse. Fixing accuracy in old games can turn into a major project.

I totally disagree. I've even got code in MAME. In fact, it spurred further improvements from others which was also accepted.

If your code doesn't make sense then it'll probably be rejected. If not, anyone can submit. Not sure where you got that idea from.

To be fair, "if your code doesn't make sense" is going to depend upon the perspective of the person who is reviewing your code, and that can often be skewed

dshadoff wrote: ↑Thu Feb 18, 2021 7:37 pm They've rejected my fixes in the past. Afraid that fixing something axiomatic might break a game (but no tests or proof that breakage occurred).
This was years ago, but it poisoned me on contributing.

This must have been before they were up on Github, I didn't see your username up there in any pull requests or issues, which is very probable, it's such an old project and you've been involved in this scene a long time too I was curious to see what happened with that.

OK guys, can we take the MAME fight somewhere else?

darksakul wrote: ↑Tue Feb 16, 2021 8:37 pm Mame was never accurate, you need a powerful CPU for software accuracy and Mame can run on a toaster powered by a potato

You can't generalize MAME though. Each different driver in MAME could be an accurate or not so accurate representation of the game being emulated. And there are certainly tradeoffs. Older and less complex hardware might be emulated in a way that is faster but not particularly accurate. But depending on the behavior of the game there may not be any noticeable impact from doing it that way. But if you did emulate more accurately you may get a huge jump in the requirements of the host system.

Either way, more options for playing and preserving these old games is always a good thing.

I'm certainly looking forward to seeing how this Saturn development goes.

the_importer wrote: ↑Fri Feb 19, 2021 4:48 am OK guys, can we take the MAME fight somewhere else?

I second that, especially since most of the talk is clueless. Go learn MAME first and then, if you have concern, talk about it on MAMEdev or just fork it and talk about your fork in your fork forum.

This is a topic about the Saturn and a core development of it.

A new update!!

Everybody needs to jump onto this guys patreon quick! Incredible progress!

Ryuoken wrote: ↑Fri Mar 26, 2021 12:20 pm A new update!!

Though i might sounds skeptic, but why aren't we seeing him load / boot the core from the dev's fpga?

MostroW wrote: ↑Fri Mar 26, 2021 1:32 pm

Ryuoken wrote: ↑Fri Mar 26, 2021 12:20 pm A new update!!

Though i might sounds skeptic, but why aren't we seeing him load / boot the core from the dev's fpga?

I see absolutely no reason to be skeptical. This guy has been responsible for some of the MiSTer's best cores.

hiddenbyleaves wrote: ↑Fri Mar 26, 2021 1:29 pm Everybody needs to jump onto this guys patreon quick! Incredible progress!

I agree, I am supporting him, it would be incredible to have this core.

MostroW wrote: ↑Fri Mar 26, 2021 1:32 pm

Ryuoken wrote: ↑Fri Mar 26, 2021 12:20 pm A new update!!

Though i might sounds skeptic, but why aren't we seeing him load / boot the core from the dev's fpga?

Not sure what you mean by"the dev's fpga", maybe you could elaborate. Do you mean that you don't see the de10-nano in the shot, or do you mean you don't see the OSD?

Core development for the MiSTer does not require the core to be integrated into the framework prior to development beginning. Core development also doesn't have to begin on the de10-nano (PS1 core is being developed currently on a xilinx arty7 a100t board, because ultraembedded is used to it and xilinx's toolchain, and the team will figure out porting to the mister later).

He brings up the OSD menu at one point in the video and loads a bios.

Oh you're right. Either way, I think it's useful that people recognize that FPGA core development doesn't necessarily require the MiSTer framework at the beginning.