2x Cpu Benchmarks

[Squidge]

Ok, I wrote some benchmarking stuff today (calculating prime numbers using square roots) to give the 2x a bit of a performance test, and see how the two processors compare.

Running the test on the 920, it completes in 17.2 seconds.
Running the test on the 940, with the 920 twiddling it's thumbs, it completes in 17.1 seconds.
Running the test on both processors (interleaving, so each does exactly half the work load), it completes in 8.9 seconds.

I think this proves that the second processor is all but useless

Now, all we need to do is spread big loops that don't rely on results from a previous loop, over both processors using a similar technique.

I've still got to test memory bandwidth - accessing memory with the 920, 940, and then both together. So keep your ears open for the results of that test

 

[Mr. Anderson]

Running the test on both processors (interleaving, so each does exactly half the work load), it completes in 8.9 seconds.

WOW! Just WOW! That ends in an efficiency of more than 190% when both cpu's are used correctly. incredible. However this is the best case scenario but it really proves how much potential a dual core programmed aplication has.

Great work! I guess the doubters will be silenced for a while

 

[Mr Doctor]

Great work! I guess the doubters will be silenced for a while

​

What kind of perfect world do you live in?

 

[Minkoff]

So then, I suppose an improvement of alot more than 2 or 3 FPS could be achieved by using the second processor.
Thanks for the info, Squidge.

 

[declaration]

yeah, thanks for the info Squidge.

Its really good of you to keep us all informed of your experiments.

 

[sand_man]

Well I hope the second cpu is somewhat beneficial because we paid for it :s

 

[Mr Doctor]

Well I hope the second cpu is somewhat beneficial because we paid for it :s

​

You got ripped off! I got mine free with my GP2x B)

 

[x68000]

Well I hope the second cpu is somewhat beneficial because we paid for it :s

​

we really have no right to expect something from the second core whatsoever, unless of course you program something yourself which takes advantage of it.

@Squidge, do you think that the second core could be implemented in any way to benefit existing emus, or are we only going to see it in new emus built from the ground up?

 

[Mr. Anderson]

Great work! I guess the doubters will be silenced for a while

​

What kind of perfect world do you live in?

​

I said for a while, I didn't say for how long. A while could be a week, a day, an hour

 

[Eolair]

Great work! I guess the doubters will be silenced for a while

​

What kind of perfect world do you live in?

​

I said for a while, I didn't say for how long. A while could be a week, a day, an hour

​

1001 PSP fanbois curl up into balls in their showers under steaming water with an expression of horror on their faces (think shower scene from "Ace Ventura: Pet Detective").

1001 PSP comes back to the forum in the morning with 1001 reasons why the test _must_ be invalid.

I'd say pretty soon

 

[DijiTao]

Personally I'd be very curious to see how the same test will run on a PSP. The PSP also has two cores, so the exact same type of results should be attenable. (Results of running on Core "A", Results of running on Core "B" only, Result of running on both cores.)

 

[TyBO!]

Personally I'd be very curious to see how the same test will run on a PSP. The PSP also has two cores, so the exact same type of results should be attenable. (Results of running on Core "A", Results of running on Core "B" only, Result of running on both cores.)

​

Are you sure?

I'm fairly certian the PSP has a single 333 MHz processor, that comes underclocked at 222 MHz.

 

[elinscheid]

Personally I'd be very curious to see how the same test will run on a PSP. The PSP also has two cores, so the exact same type of results should be attenable. (Results of running on Core "A", Results of running on Core "B" only, Result of running on both cores.)

​

Are you sure?

I'm fairly certian the PSP has a single 333 MHz processor, that comes underclocked at 222 MHz.

​

Skeezix has a pretty cool blog and he talked about the PSP specs:
http://www.codejedi.com/cgi-bin/blog.cgi/e....blog?seemore=y

 

[mittens]

Well, you mean the GPU core of PSP?
I don't think it is fully programmable, but I have no idea :unsure:

BTW, be aware everybody, this is just a 'TECHDEMO', isn't it?
(It is not playable and where is sound?)

EDIT: sorry, I read the article, PSP HAS two cores, awesome. :rolleyes: but I still much much prefer my GP2X (it has strong 2 arms :lol: )

 

[Chip]

BTW at be aware everybody at this is just a 'TECHDEMO' at isn't it?
(It is not playable and where is sound?)

​

What!? You mean I can't actually play the "calculate prime numbers" game with sound? That 2nd core really IS useless.

Let me know when I can get some full speed 60fps skip0 square root action going.

 

[Tobriand]

EDIT: sorry, I read the article, PSP HAS two cores, awesome. :rolleyes: but I still much much prefer my GP2X (it has strong 2 arms :lol: )

​

I'm fairly sure the ARM9 series of chips aren't strongarm, unfotunately... I looked into stuff that was made for RiscOS that was Open Source some years back for the GP32 (after all, RiscOS is for ARMs, so it stands to reason that proggies optimized for it ought to port across fairly easily, unless they relied on bitsies in the OS itself). Strongarm chips seemed - from what I could gather - to be the next step up from the ARM9 series; presumably less for the mobile market, but now I'm completely guessing.

But I don't think they're the same (there was an open source snes emu that said you'd need "at least a strongarm processor to run [it] at approaching full speed" - and I'm fairly sure it wasn't in ASM, else Yoyo and LDchen (iirhnc) wouldn't have had to write one for OS9x from the ground up).

On another note, it's *very* cool to see the potential for the second CPU's usage. Now, at least, we have *some* figures to hand to wave in the avatars of those who continue to deny its ability to do anything - although not necessarily in the avatars of those who say it's useless for emus; I assume the same thing is true today as it was yesterday - i.e. that unless an emu is written with the idea of a dual core in mind, it's not going to speed things up *that* much.

As ever, I don't know what I'm talking about when it comes to coding though.

 

[mittens]

@Tobriand:
Thanks for such a nice info about ARM processor (seriously)
Anyway I used term 'strong' for comparing with PSP not the Strongarm processor though.

Hmm, how much they are faster than Arm9? I've only used Intel's before. (Ah, Z80 also)

 

[Tobriand]

Actually, taking another look in the light of the GP2x's increased CPU whack on the GP32, it seems as though the speed difference between them is an inconsequential 6MHz... there's an article http://en.wikipedia.org/wiki/StrongARM there about the Strongarm processor, and some general ARM core specs http://en.wikipedia.org/wiki/ARM_architecture there.

 

[reallynotnick]

*looks left and right* What no DaveC?

Haha just kidding dave, we all love you.

 

[PSyMastR]

*looks left and right* What no DaveC?

Haha just kidding dave, we all love you.

​

Dave! Where are you? I want a comment. I also want an essay by Magus. What happened to this place?