Nintendo 64 Emulation


Hi... I know there has been much...MUCH speculation about the pandora`s ability to run some kind of N64 emulation but I was wondering... does anyone but me still remember CORN (for windoze back in the 90`s when no one had much idea how to emulate a N64)?

I know your all gonna shout at me and say its closed sourse windoze specific HLE code optimised for Mario64 and although it runs other stuff most of it is glitchy BUT, the point im gonna try n make is it ran on my Pentium 650 with 128 megs of ram and a voodoo card at a blistering pace (Full I think, well im sure!) so If someone was to have the skills, the time and the desire to have a go from scratch or port and radically optimise one of the more portable N64 Emulators then I really think they could be on to something...

I wish I had the skills, GLBasic is about my limit, lol (Flame Shields are set to maximum :blink: )

**OOPs Ive edited this** I didnt realise this thread had gone so far from topic. perhaps noone cares what im saying LOL**
 
Philly said:
I know your all gonna shout at me and say its closed sourse windoze specific HLE code optimised for Mario64 and although it runs other stuff most of it is glitchy BUT, the point im gonna try n make is it ran on my Pentium 650 with 128 megs of ram and a voodoo card at a blistering pace (Full I think, well im sure!) so If someone was to have the skills, the time and the desire to have a go from scratch or port and radically optimise one of the more portable N64 Emulators then I really think they could be on to something...

Not a matter of skills. CORN is a static recompiler, not HLE *edit* or, it does use HLE as well, but it's the static recompiler that is the bottleneck for compatibility*/edit*. That is why its game compatibility is so low. It would be a huge time investment to make a static recompiler work with the whole library of 64 games. It's faster *edit*for the person authoring the emulator*/edit* just to HLE the GPU subroutines since most games used the same SDK for that anyway.
 
Last edited by a moderator:
I only read the first page, so I don't know if anybody's mentioned this yet.

Back in the day, people said that PSP devs would never get an n64 emulator running at more than .1 fps. But even right now, I have fully playable n64 emulator called daedalusX64 on my PSP. On SM64, I quite often get complete full speed. Gives that the Pandora is 3 times faster than the PSP, (Both overclocked) I should think an optimized port of this open source project would yield three times better performance. so on SM64, perhaps 90 fps at good times, and 30 fps (full speed) when there's horrible, terrible, huge . I could live with that, couldn't you?

I know that comparing clock speed over different platforms is bad, but even if it's just twice as powerful, we still get full speed.
 
midna25 said:
I only read the first page, so I don't know if anybody's mentioned this yet.

Back in the day, people said that PSP devs would never get an n64 emulator running at more than .1 fps. But even right now, I have fully playable n64 emulator called daedalusX64 on my PSP. On SM64, I quite often get complete full speed. Gives that the Pandora is 3 times faster than the PSP, (Both overclocked) I should think an optimized port of this open source project would yield three times better performance. so on SM64, perhaps 90 fps at good times, and 30 fps (full speed) when there's horrible, terrible, huge . I could live with that, couldn't you?

I know that comparing clock speed over different platforms is bad, but even if it's just twice as powerful, we still get full speed.

Technically, I think full-speed is 60fps. That's not even considering the fact that MANY emulators first emulate just enough functions to get Mario 64 running, optimize those (so that it runs quickly), then add other functions. There might not be much compatibility (over the whole library of N64 games), and other games (and more complex, later games) might bring the emu to a screeching halt. The fact that it runs SM64 at 30fps doesn't tell us much.
 
Last edited by a moderator:
midna25: some mistakes in your calculation:
- PSP and N64 are both mips which probably simplifies the recompiler a good bit
- The Pandora is not 3 times faster than the PSP

Philly: there is UltraHLE which is open-source from what I remember

Can't you guys just wait until someone ports something and then judge wether its possible or not?
 
I'd love to write an N64 emulator, or port Mupen64Plus, but I really don't think I have the skillset at the moment. Mupen64Plus might be worth looking into. It's open source and has multiple interpreters. An ARM dynarec one would need to be written for maximum speed, but I think possibly the biggest hurdle would be porting the OGL code to OGLES.

But again, I haven't looked at it much.
 
JayFoxRox said:
midna25: some mistakes in your calculation:
- PSP and N64 are both mips which probably simplifies the recompiler a good bit
- The Pandora is not 3 times faster than the PSP

Philly: there is UltraHLE which is open-source from what I remember

Can't you guys just wait until someone ports something and then judge wether its possible or not?

I was unaware that n64 and PSP are more compatible with each other. IS that true?

Anyway, I was rounding up.

PSP factory clock speed = 222mhz
Pandora factory clock speed = 600

PSP overclocked speed = 333mhz
Pandora overclocked speed = (around) 900mhz

So 2.7027027 times faster. (Not getting into different architectures)
 
Last edited by a moderator:
Vorporeal said:
Technically, I think full-speed is 60fps. That's not even considering the fact that MANY emulators first emulate just enough functions to get Mario 64 running, optimize those (so that it runs quickly), then add other functions. There might not be much compatibility (over the whole library of N64 games), and other games (and more complex, later games) might bring the emu to a screeching halt. The fact that it runs SM64 at 30fps doesn't tell us much.

Shit. Accodental double post. I was trying to get a quote on this post.

Anyway, to my knowledge F-Zero is the only game for the n64 which runs at 60fps. That's on the original system. SM64 ran at 30fps on my desktop's n64 emulator. It was quite full speed. Emulated on my PSP at the full speed parts, it doesn't seem a bit choppy.

Also, SM64 isn't the only game it runs. There are a good amount of others, even if not all of them achieve the same speeds as SM64.

This'll tell you a little more about DaedalusX64. This isn't recent either. There has been even more progress. The video makes it seem slower too.

http://www.youtube.com/watch?v=joy3Oh8JRLQ
 
Last edited by a moderator:
midna25 said:
JayFoxRox said:
midna25: some mistakes in your calculation:
- PSP and N64 are both mips which probably simplifies the recompiler a good bit
- The Pandora is not 3 times faster than the PSP

Philly: there is UltraHLE which is open-source from what I remember

Can't you guys just wait until someone ports something and then judge wether its possible or not?

I was unaware that n64 and PSP are more compatible with each other. IS that true?

Anyway, I was rounding up.

PSP factory clock speed = 222mhz
Pandora factory clock speed = 600

PSP overclocked speed = 333mhz
Pandora overclocked speed = (around) 900mhz

So 2.7027027 times faster. (Not getting into different architectures)
i'm sorry, but you're going to have to get into different architectures if you're comparing the speed. mhz doesn't prove anything. when the pandora comes out, we'll see how many times faster it is than the psp.
 
Last edited by a moderator:
midna25 said:
PSP factory clock speed = 222mhz
Pandora factory clock speed = 600

PSP overclocked speed = 333mhz
Pandora overclocked speed = (around) 900mhz
Technically 333 is the PSPs rated clock speed, 222 is underclocked for battery conservation. So the clock for clock, the Pandora isn't even twice as fast as the PSP at the officially rated clock speed.
However, unlike the Pandora's processor, which has been found stable at clock speeds 50% higher than rated (though that's not guaranteed) the PSP pretty much crashes at 366.
That's all pretty much irrelevant, though.
 
Last edited by a moderator:
midna25 said:
This'll tell you a little more about DaedalusX64. This isn't recent either.
Most people are up to speed on Daedalus. And this whole topic, really. ;)

There are those of us who are desperately holding on to the hope that the original author of Daedalus is working on a Pandora port in secret. We've been assured repeatedly that it's not the case, but N64 desires are stubborn ones. :D It would be the ideal scenario. Until something is announced though, there's really nothing we can discuss that hasn't been discussed over and over again.
 
Last edited by a moderator:
borgqueenx said:
But no dev has come yet, so i dont think its likely that this is gonna happen.
To bad though, n64 is on my top 3 of emu's i want, together with psp and ds games.

I think you are banking on the wrong device :)
 
Last edited by a moderator:
WizardStan said:
midna25 said:
PSP factory clock speed = 222mhz
Pandora factory clock speed = 600

PSP overclocked speed = 333mhz
Pandora overclocked speed = (around) 900mhz
Technically 333 is the PSPs rated clock speed, 222 is underclocked for battery conservation. So the clock for clock, the Pandora isn't even twice as fast as the PSP at the officially rated clock speed.
However, unlike the Pandora's processor, which has been found stable at clock speeds 50% higher than rated (though that's not guaranteed) the PSP pretty much crashes at 366.
That's all pretty much irrelevant, though.
I think it is more interesting to know, if the Pandora or the PSP is fater-per-MHz. I mean, is the Pandora at 333MHz faster than a PSP at 333MHz? I don't know exactly how this could be measured, maybe in MIPS??? ^^"""" Pure Per-MHz-Power could be an interesting value to compare such different Machines maybe...
 
Last edited by a moderator:
fusion_power said:
I think it is more interesting to know, if the Pandora or the PSP is fater-per-MHz. I mean, is the Pandora at 333MHz faster than a PSP at 333MHz? I don't know exactly how this could be measured, maybe in MIPS??? ^^"""" Pure Per-MHz-Power could be an interesting value to compare such different Machines maybe...
Faster per mhz doing what? ints or floats? memory reads/writes? instruction throughput? There are too many metrics to say "pandora is xx% faster than PSP at the same clock speed."
 
Last edited by a moderator:
Comparing the 2 processors is a very hard task to do. Due to the difference in architecture even if you try to run the same code on both and compare the results it is still not a perfect test. They do not have the same instruction set which will make the code be run different on each machine after compiling. Even if a piece of code is written to try and be as similar as possible in assembly for each there will still be issues of op codes that are only there for legacy support/simplicity that would hinder the performance. We are just going to have to wait and see what happens when that pandora comes out and people start porting code to it.
 
dpoarch said:
Even if a piece of code is written to try and be as similar as possible in assembly for each there will still be issues of op codes that are only there for legacy
support/simplicity that would hinder the performance.

This isn't x86. Neither architecture has opcodes that shouldn't ever be used because sequences of other opcodes are always faster.

Underclocking the Pandora for a comparison is no good because the memory ratios go down with it when you do that. They're both running 333MHz DDR SDRAM, although I don't know what the bus width of PSP's is - I wouldn't be surprised if it were 16 instead of 32 because the memory bandwidth and especially latency on the device is quite poor. There might be other architectural flaws causing this. The big difference Pandora will make is its L2 cache. This is also probably why iPhone 3GS is much faster than iPhone 3G without being clocked that much higher. The CPU is also theoretically capable of achieving far more instructions per cycle because of its superscalar architecture. Cortex-A8 is really way ahead of Allegrex in integer performance, but in floating point performance its VFPU can probably give NEON a run for its money.
 
Last edited by a moderator:
I think what everyone wants is just some piece of mind, someone starting a project that may even go nowhere, but just a dev declaring intent. Exophase, would you do the honours :p
 
Luftwaffles said:
I think what everyone wants is just some piece of mind, someone starting a project that may even go nowhere, but just a dev declaring intent. Exophase, would you do the honours :p
"Exophase is bad. Nothing good will ever come of him."
 
Last edited by a moderator:
Back
Top