Target level of Emulation/Ported games for Pandora 2..?

[Foxgod]

If i had to guess i would say it will open the door for stuff like yabase and nulldce to run at playable speeds (here's the if) ...if... it gets ported. (Yabase has a public demo on android that looks almost playable on dual core a9's @ 1.2ghz) i can see nds as playable (already a port for both OS's currently) for some games but i dont think full speed on all of them for sure, i am hoping it will be similar in speed to the current N64 on pandora speed.

Yabause has been ported, unless I'm very much mistaken, and it runs like a dog. This is (again, from memory) using a pretty highly optimised dynarec core from Ari64 - I'd not like to stick my neck out with foxgod's though and say it will get much better. I don't think Yabause will ever be playable on the P1.


D.

Thats not what i said, but technically it do will be playable on the P1 someday, but by then we will be having a P3, like i said, Yabause updates slowly, SSF runs much better, but its windows only, and it requires hardware features that can only be found in a pentium 4 and newer.


But specs wise it can do well on a slower machine.

 

[craigix]

You should at least start reading all of the posts in the threads you read because this has been explained several times already. Cortex-A15 has a physical address extension that supports 40-bits (you need 33 physical address bits to support 8GB.. how you could guess 44 or 48 I really have no idea..) OMAP5 is known to support up to 8GB and Exynos 5 may as well.


The problem is that physical address extensions are a kludge and not the same as having true support for the memory, in reality it'll be hard for users to utilize 8GB.. but Craig saw that the SoC he wants supports it and since he wants to use the "best" hardware possible, well.. Fortunately for him people eat this shit up, probably lots will think that 8GB will make their emus faster.

I don't think the average Pandora user thinks that more RAM = better emulator performance. I'm reasonably sure we don't sell to the PSP crowd.


I'm also pretty sure people won't think 8GB of RAM is 'shit' when almost every phone/tablet has it. I don't know why you speak like this about new hardware, you seem to have a similar downer on HD screens, but they are coming, tech moves on constantly, and why should it not? (I don't actually want to debate this before you write an essay).


However, (**I cannot prove this in any way, I may just be making this up, etc.**) 8GB will be possible to use in an up coming Android which supports these newer SoCs (maybe as a swap, maybe more, I don't have details*).


*Allegedly. May never happen etc. I made this up. etc. claims based on nothing.

 

[Foxgod]

You should at least start reading all of the posts in the threads you read because this has been explained several times already. Cortex-A15 has a physical address extension that supports 40-bits (you need 33 physical address bits to support 8GB.. how you could guess 44 or 48 I really have no idea..) OMAP5 is known to support up to 8GB and Exynos 5 may as well.


The problem is that physical address extensions are a kludge and not the same as having true support for the memory, in reality it'll be hard for users to utilize 8GB.. but Craig saw that the SoC he wants supports it and since he wants to use the "best" hardware possible, well.. Fortunately for him people eat this shit up, probably lots will think that 8GB will make their emus faster.

I don't think the average Pandora user thinks that more RAM = better emulator performance. I'm reasonably sure we don't sell to the PSP crowd.


I'm also pretty sure people won't think 8GB of RAM is 'shit' when almost every phone/tablet has it. I don't know why you speak like this about new hardware, you seem to have a similar downer on HD screens, but they are coming, tech moves on constantly, and why should it not? (I don't actually want to debate this before you write an essay).


However, (**I cannot prove this in any way, I may just be making this up, etc.**) 8GB will be possible to use in an up coming Android which supports these newer SoCs (maybe as a swap, maybe more, I don't have details*).


*Allegedly. May never happen etc. I made this up. etc. claims based on nothing.

Things indeed better be as future proof as possible.


Might as well try and make the run of the Pandora 2 as long as possible, only way to do this is indeed by pumping it with specs.


64K memory adressing will become a standard in handhelds faster then people realise right now.


Those who dont dig it, are in the same boat as the 8 bit camp vs the Microsoft 16 bit camp back in the 80ies.


I can already see the complaints about the pandora not supporting 64bit android only a year and a half after the pandora 2 its release....

 

[ZXDunny]

If i had to guess i would say it will open the door for stuff like yabase and nulldce to run at playable speeds (here's the if) ...if... it gets ported. (Yabase has a public demo on android that looks almost playable on dual core a9's @ 1.2ghz) i can see nds as playable (already a port for both OS's currently) for some games but i dont think full speed on all of them for sure, i am hoping it will be similar in speed to the current N64 on pandora speed.

Yabause has been ported, unless I'm very much mistaken, and it runs like a dog. This is (again, from memory) using a pretty highly optimised dynarec core from Ari64 - I'd not like to stick my neck out with foxgod's though and say it will get much better. I don't think Yabause will ever be playable on the P1.

Thats not what i said, but technically it do will be playable on the P1 someday

Jesus, you're definitely not a coder, are you? You really don't have the slightest clue, do you?


Colour me gobsmacked.


D.

 

[Exophase]

You haven't been reading his posts much have you..


Hopefully one day software and hardware development will be fueled by the power of optimism and unicorns instead of these bleak, bleak technical realities.

 

[mcobit]

It is funny, how Foxgod say,s that the windows emu, that runs so much better only runs with pentium4 and up... You know, every p4 is more powerful as a p1. So why should it ever run good on a Pandora1???


also the rate of updates doesn't equal the performans of the emulator...


Just too funny

 

[Foxgod]

It is funny, how Foxgod say,s that the windows emu, that runs so much better only runs with pentium4 and up... You know, every p4 is more powerful as a p1. So why should it ever run good on a Pandora1???


also the rate of updates doesn't equal the performans of the emulator...


Just too funny

Learn to read plz, it requires a p4 because it depends on features that can only be found in p4 and up, got nothing to do with clock speed.

 

[mcobit]

So how do you know, that it will run on lower clockspeed and features than p4?

 

[Foxgod]

If i had to guess i would say it will open the door for stuff like yabase and nulldce to run at playable speeds (here's the if) ...if... it gets ported. (Yabase has a public demo on android that looks almost playable on dual core a9's @ 1.2ghz) i can see nds as playable (already a port for both OS's currently) for some games but i dont think full speed on all of them for sure, i am hoping it will be similar in speed to the current N64 on pandora speed.

Yabause has been ported, unless I'm very much mistaken, and it runs like a dog. This is (again, from memory) using a pretty highly optimised dynarec core from Ari64 - I'd not like to stick my neck out with foxgod's though and say it will get much better. I don't think Yabause will ever be playable on the P1.

Thats not what i said, but technically it do will be playable on the P1 someday

Jesus, you're definitely not a coder, are you? You really don't have the slightest clue, do you?


Colour me gobsmacked.


D.

Thanks for asking, but i completed a java study, and will bring it to practise soon when i transfer from the bussines to the IT dep at work soon.


What i said is that the P1 is powerful enough, but the software is far from mature enough.

 

[mcobit]

If i had to guess i would say it will open the door for stuff like yabase and nulldce to run at playable speeds (here's the if) ...if... it gets ported. (Yabase has a public demo on android that looks almost playable on dual core a9's @ 1.2ghz) i can see nds as playable (already a port for both OS's currently) for some games but i dont think full speed on all of them for sure, i am hoping it will be similar in speed to the current N64 on pandora speed.

Yabause has been ported, unless I'm very much mistaken, and it runs like a dog. This is (again, from memory) using a pretty highly optimised dynarec core from Ari64 - I'd not like to stick my neck out with foxgod's though and say it will get much better. I don't think Yabause will ever be playable on the P1.

Thats not what i said, but technically it do will be playable on the P1 someday

Jesus, you're definitely not a coder, are you? You really don't have the slightest clue, do you?


Colour me gobsmacked.


D.

Thanks for asking, but i completed a java study, and will bring it to practise soon when i transfer from the bussines to the IT dep at work soon.


What i said is that the P1 is powerful enough, but the software is far from mature enough.

Yabause has a lot of developement time on x86 and recently got a very good arm dynarec. Why isn't it mature?

 

[Foxgod]

So how do you know, that it will run on lower clockspeed and features than p4?

Why would it not?


SSF doesnt depend on the p4 for its clockspeed, it depends on it because of SSE2 which can only be found in P4 and up.


You can still run it on a slower athlon that features SSE2.

If i had to guess i would say it will open the door for stuff like yabase and nulldce to run at playable speeds (here's the if) ...if... it gets ported. (Yabase has a public demo on android that looks almost playable on dual core a9's @ 1.2ghz) i can see nds as playable (already a port for both OS's currently) for some games but i dont think full speed on all of them for sure, i am hoping it will be similar in speed to the current N64 on pandora speed.

Yabause has been ported, unless I'm very much mistaken, and it runs like a dog. This is (again, from memory) using a pretty highly optimised dynarec core from Ari64 - I'd not like to stick my neck out with foxgod's though and say it will get much better. I don't think Yabause will ever be playable on the P1.

Thats not what i said, but technically it do will be playable on the P1 someday

Jesus, you're definitely not a coder, are you? You really don't have the slightest clue, do you?


Colour me gobsmacked.


D.

Thanks for asking, but i completed a java study, and will bring it to practise soon when i transfer from the bussines to the IT dep at work soon.


What i said is that the P1 is powerful enough, but the software is far from mature enough.

Yabause has a lot of developement time on x86 and recently got a very good arm dynarec. Why isn't it mature?

BEcause it doesnt emulate 100% of the saturn yet, first comes implementation, then comes optimisation, not the other way around.

 

[mcobit]

So how do you know, that it will run on lower clockspeed and features than p4?

Why would it not?


SSF doesnt depend on the p4 for its clockspeed, it depends on it because of SSE2 which can only be found in P4 and up.


You can still run it on a slower athlon that features SSE2.

If it need sse2, you would need really good neon code on Pandora.


Did you run it on an Athlon with 500MHz?

 

[Foxgod]

So how do you know, that it will run on lower clockspeed and features than p4?

Why would it not?


SSF doesnt depend on the p4 for its clockspeed, it depends on it because of SSE2 which can only be found in P4 and up.


You can still run it on a slower athlon that features SSE2.

If it need sse2, you would need really good neon code on Pandora.


Did you run it on an Athlon with 500MHz?

Did you?


If it doesnt work well then it needs more optimisation, which wont be done for a while, the implemetation isnt complete yet.

 

[mcobit]

So how do you know, that it will run on lower clockspeed and features than p4?

Why would it not?


SSF doesnt depend on the p4 for its clockspeed, it depends on it because of SSE2 which can only be found in P4 and up.


You can still run it on a slower athlon that features SSE2.

If it need sse2, you would need really good neon code on Pandora.


Did you run it on an Athlon with 500MHz?

Did you?

No, but I don't make assumptions, that it will run fullspeed like this.

 

[ZXDunny]

If i had to guess i would say it will open the door for stuff like yabase and nulldce to run at playable speeds (here's the if) ...if... it gets ported. (Yabase has a public demo on android that looks almost playable on dual core a9's @ 1.2ghz) i can see nds as playable (already a port for both OS's currently) for some games but i dont think full speed on all of them for sure, i am hoping it will be similar in speed to the current N64 on pandora speed.

Yabause has been ported, unless I'm very much mistaken, and it runs like a dog. This is (again, from memory) using a pretty highly optimised dynarec core from Ari64 - I'd not like to stick my neck out with foxgod's though and say it will get much better. I don't think Yabause will ever be playable on the P1.

Thats not what i said, but technically it do will be playable on the P1 someday

Jesus, you're definitely not a coder, are you? You really don't have the slightest clue, do you?


Colour me gobsmacked.

Thanks for asking, but i completed a java study, and will bring it to practise soon when i transfer from the bussines to the IT dep at work soon.

Yeah, I stand by what I said. Java will bring you untold richness of knowledge of the internals of computer hardware right enough.

BEcause it doesnt emulate 100% of the saturn yet, first comes implementation, then comes optimisation, not the other way around.

Good god. You think that as they add more emulated subsystems to the program it'll get faster?


Keep going, this is comedy gold!


D.

 

[Foxgod]

So how do you know, that it will run on lower clockspeed and features than p4?

Why would it not?


SSF doesnt depend on the p4 for its clockspeed, it depends on it because of SSE2 which can only be found in P4 and up.


You can still run it on a slower athlon that features SSE2.

If it need sse2, you would need really good neon code on Pandora.


Did you run it on an Athlon with 500MHz?

Did you?

No, but I don't make assumptions, that it will run fullspeed like this.

How could it if the emulator isnt done yet, thats why i said saturn wont run well on p1 right now, but will someday, but by then we will be on the Pandora 3, so theres no point in trying for the P1.

If i had to guess i would say it will open the door for stuff like yabase and nulldce to run at playable speeds (here's the if) ...if... it gets ported. (Yabase has a public demo on android that looks almost playable on dual core a9's @ 1.2ghz) i can see nds as playable (already a port for both OS's currently) for some games but i dont think full speed on all of them for sure, i am hoping it will be similar in speed to the current N64 on pandora speed.

Yabause has been ported, unless I'm very much mistaken, and it runs like a dog. This is (again, from memory) using a pretty highly optimised dynarec core from Ari64 - I'd not like to stick my neck out with foxgod's though and say it will get much better. I don't think Yabause will ever be playable on the P1.

Thats not what i said, but technically it do will be playable on the P1 someday

Jesus, you're definitely not a coder, are you? You really don't have the slightest clue, do you?


Colour me gobsmacked.

Thanks for asking, but i completed a java study, and will bring it to practise soon when i transfer from the bussines to the IT dep at work soon.

Yeah, I stand by what I said. Java will bring you untold richness of knowledge of the internals of computer hardware right enough.

BEcause it doesnt emulate 100% of the saturn yet, first comes implementation, then comes optimisation, not the other way around.

Good god. You think that as they add more emulated subsystems to the program it'll get faster?


Keep going, this is comedy gold!


D.

No, what i said is that once its 100% implemented, they can start code optimisation, then itl run faster on slower systems, you just dont read do you>?


I am done with you...

 

[ZXDunny]

You haven't been reading his posts much have you..

Not particularly - this is the first time I've seen him actually try to tell people who can code at the hardware level that they're wrong because of his sunny disposition!


Foxgod, you're top of my moron pile!


D.

 

[mcobit]

OMG, I'm done, too. You won't learn or understand the points risen here.


But I agree with ZXDunny: Go on...

 

[ZXDunny]

Jesus, you're definitely not a coder, are you? You really don't have the slightest clue, do you?

Thanks for asking, but i completed a java study, and will bring it to practise soon when i transfer from the bussines to the IT dep at work soon.

Yeah, I stand by what I said. Java will bring you untold richness of knowledge of the internals of computer hardware right enough.

BEcause it doesnt emulate 100% of the saturn yet, first comes implementation, then comes optimisation, not the other way around.


...


No, what i said is that once its 100% implemented, they can start code optimisation, then itl run faster on slower systems, you just dont read do you>?


I am done with you...

You just said:

Thats not what i said, but technically it [yabause] do will be playable on the P1 someday


But specs wise it can do well on a slower machine.

Which I take to mean that you're confident that when it's even slower due to more complete emulation, optimisation will make it even faster than the point it's at now. You're a fecking idiot is what you are. I dunno, maybe they could really go for gold and add a dynarec!


I'll share an anecdote with you. Some time ago, I (and a friend or two) wrote a Spectrum emulator. Now, an incomplete emulation ran on a 286 processor. It wasn't very accurate, and left out a lot of features that a few coders back in the day used to employ, but it got the job done for the majority of software. Our emulator added a lot more. It was as close to a perfect emulation as we had knowledge for at the time.


It emulated a minimum of 2 CPUs, 3 in the case of the 128k Spectrum. We added better graphics support for multicolour displays, we added accurate sound, we added previously unknown z80 CPU behaviour. The result? It ran barely on my 1.6GHz PC. So we optimised - this was the days before dynarecs were commonplace so the optimisation was things like better pipelining, using assembler in places where the compiler made sub-optimal code. We switched to more efficient algorithms, coded to the cache in our data structures, moved to hardware accelerated graphics etc. Went all out, short of recompiling.


The result? It ran quite well on a 300MHz CPU.


More complete emulation, as fast as it could be made, was nowhere near as fast as it needed to be to run on a 286 (and couldn't, in any case). Granted ours ran in windows rather than DOS, but even so - the better emulation that ran a bigger percentage of known software required far more raw CPU time.


And no amount of optimisation will offset the speed required for more complex in-depth emulation. Synchronisation between the two (or three) CPUs meant that multithreading was useless for all but actual display presentation.


You're wrong, no matter how you slice it.


D.

 

[Exophase]

On Yabause: IMO it COULD be a lot faster if really optimized VDP1/2 renderers were written. Ari64 agrees, this is probably what's holding back speed the most, not the CPUs and DSPs. Although if the DSPs (especially the one used for geometry transform)


But good luck getting people to work on that sort of thing. The way the 3D renders isn't even very well understood/documented.