(?)superpi_ss3

[x714n]

anyone try this? it give better result for prescott, but is this legal?
with superpi mod i get 37.953s
http://www.xtremesystems.org/forums/attachment.php?attachmentid=30662&stc=1

with superpi i get 37s
http://www.xtremesystems.org/forums/attachment.php?attachmentid=30660&stc=1

and with superpi_ss3 i get 34s :confused:
http://www.xtremesystems.org/forums/attachment.php?attachmentid=30661&stc=1

[Charles Wirth]

Most people will say yes and some dont care.

There are too many versions on pi now and it is quite confusing to what we are looking at in screen shots without a version or checksum. I like the msec version myself.

[[XC] moddolicous]

I think most people now use use the cheat free one, so I think the other ones are obsolete. Its like the older cpu-z versions.

[Kunaak]

with there being too many versions of Super Pi out there now, with all the patches and english/japan versions too, its impossible to get a fair and even score, cause some peope use the patchs, some dont, some use a faster japanese version, most don't... theres just no even/fair ways to get a super pi score these days.

thats why the cheat free version seems to be "the" version to use now.
no patches, no faster japan version, just raw CPU power.

[impar]

Greetings!

And why not a updated "no cheat" version using the SSE3 feature current CPUs have?

[metro.cl]

any link to download? i wanna try this on c2d :)

[impar]

Greetings!

I downloaded from here:
http://www.overclock.net/downloads/28044-definitive-super-pi-thread.html

Considering all CPUs now support SSE3...

[spdycpu]

Both the SSE2 and SSE3 binaries look faked to me. I've done a little bit of assembly programming and what I see there looks like it doesn't do anything. Just a few instructions to keep it from running on a non-sse2 or sse3 CPU, who knows what they changed to make the scores magically better. It isn't SSE2/3, at least the way it was implemented here. Another thing is the emms instruction is never used which sort of gives this away as fake. If you start manipulating xmm registers without clearing everything with emms first bad things will happen, so always issue an emms before any SIMD code and after any SIMD code. Read more about the emms instruction here: http://www.tacc.utexas.edu/services/userguides/intel/c_ug/linux335.htm

Disassembling the SuperPI SSE3 executable shows the only interaction with xmm registers is this:


4312c9: 0f 17 0f movhps %xmm1,(%edi)


Here is the SSE2 SuperPI:


420142: f2 0f 10 04 24 movsd (%esp),%xmm0
420147: f2 0f 2c c0 cvttsd2si %xmm0,%eax
4312c9: 0f 17 0f movhps %xmm1,(%edi)


Again, no emms in either one of them either.

If you want to see what an assembly dump of a PI program REALLY using SSE2/3 looks like, check out some stuff from QuickPI v4.0a 32 bit (hopefully Steve Pagliarulo has no problems with me putting this up, if so PM me and I'll remove it): http://chess.homelinux.com/qpi40a-xmm.txt
QPI 4.0a issues 2 emms instructions as they should be used, one before SIMD calculations and one after you're done with SIMD calculations (that isn't included in the assembly dump though).

This brings me to another point I've been wondering. Why in the world do people still use SuperPI? It is ridiculously slow and unoptimized. An analogy would be something like this. Imagine a quarter mile strip where people raced 4 wheel buggies on a dirt strip w/ potholes. The buggies aren't too affected by the road because they're slow as hell anyway. Imagine that we've got NHRA dragsters that can do 300+ mph now days, *BUT*, we're still using the same junk road thats full of dirt, potholes, pebbles, etc. Using programs like QPI and Pifast would be analogous to switching to a paved race way. You'll get the most out of your dragster.

The only question now is if Steve Pagliarulo (author of QuickPI) would be willing to make a benchmark style GUI with anti-cheat protection for his program. This should make the transition from SuperPI to QuickPI effortless for everybody.

[impar]

Greetings!
This brings me to another point I've been wondering. Why in the world do people still use SuperPI? It is ridiculously slow and unoptimized.Agree. Thats why I ressurected the thread.

Thanks for clearing the SuperPI SSE3 subject.

The only question now is if Steve Pagliarulo (author of QuickPI) would be willing to make a benchmark style GUI with anti-cheat protection for his program. This should make the transition from SuperPI to QuickPI effortless for everybody.Yep. The current form of QuickPI is... "antiquated".

[G H Z]

This brings me to another point I've been wondering. Why in the world do people still use SuperPI?

You already know why spdy, no one has come along with a better (more popular) creation for calculating PI. Why don't you get together with Steve Pagliarulo and write one? The anit-cheat protection your mentioning would be great considering there have been so many concerns regarding Super PI.

[spdycpu]

You already know why spdy, no one has come along with a better (more popular) creation for calculating PI. Why don't you get together with Steve Pagliarulo and write one? The anit-cheat protection your mentioning would be great considering there have been so many concerns regarding Super PI.

It seems like people can't be bothered to type in a few things these days, everything has to be point and click. You can still use QuickPI among the "trusted" people, those you know who won't cheat. It isn't a huge problem to type a little, is it? I can see how point/click is easier, sure, but that doesn't make QuickPI useless either.

As far as doing GUI programming with him, I don't see that happening. Not only have I never talked to him about this before, I don't do any sort of GUI programming. I just do math/simd related stuff (which he can already do extremely well).

[G H Z]

I don't have a problem running it but be realistic. You asked about why people use Super PI, well, a big part of that is it's easy - no command line stuff.

Anyway it was just a thought.

[spaglia]

The only question now is if Steve Pagliarulo (author of QuickPI) would be willing to make a benchmark style GUI with anti-cheat protection for his program. This should make the transition from SuperPI to QuickPI effortless for everybody.

Hey Guys, Thanks for taking an interest in QuickPi. There is already an anti-cheat in the program (use the binary option). You can then post binary files and the Reader Program (QRD) will verify the binary. The binaries also don't necessarily have to have all the digits, just include their checksum.

As far as the GUI goes, I would not call the program 'antiquated' although a GUI would be a great improvement. I already have some GUI ideas in mind but they have taken a back-seat to the other program features.

Steve

[dinos22]

i think the reason most people use SuperPI mod v1.5 is because we can all easily compare results :)

if people start using a bunch of different versions it will be a pain the back side :(

[impar]

Greetings!

Not really.
Software evolves.