Super PI for Atari

[moulinaie]

Hi,

 

I started writing a SUPER PI program that calculates the decimals of PI. The SUPER PI for PC has no available sources, so I took, as a start, a Quick Basic program. I can't tell if it uses the same method..!

 

Here are some first results for 16.000 decimals of PI:

 

TT 32MHz + TT Ram : 211.03 seconds

TT 48MHz + TT Ram: 150.21 seconds

Falcon CT060 100MHz: 19.53 seconds.

 

I didn't run it for 1 million decimals, but from my first tests I would say that:

 

1.000.000 / 16.000 = 62.5 times

So the time would be (62.5)² = 3906 times longer.

 

For the CT060, 21H 11mn 29 sec.

 

I didn't implement yet the text saving of the result, I think it is part of the test, so the times will be greater when finished.

 

Guillaume.

[GroovyBee]

I assume that you aren't displaying any sort of progress bar on the screen. If you are, you might want to disable it because that can take quite a bit of time too.

[moulinaie]

I assume that you aren't displaying any sort of progress bar on the screen. If you are, you might want to disable it because that can take quite a bit of time too.

 

it's a simple TOS program in text mode with really nothing looking like an interface.

 

:-)

Guillaume.

[GroovyBee]

Even the TOS mode text display is very slow .

[moulinaie]

yes you're right.

 

but nothing is displayed.

 

guillaume

[moulinaie]

Hi,

 

I need your help to fill my tests!

The Super PI program is ready to be used, you'll find it on my page (TEST 1)

http://gtello.pagesperso-orange.fr/kronos_soft.htm

 

Just run it and send me back the infos to add some lines to my table. We'll keep the 16384 digits calculation as a base.

 

Thanks to everyone!

 

(you need a 68030 or better).

 

Guillaume.

[Cyprian]

Time for calculation for 16384 digits: 2.20 s, under Aranym 0.9.10 JIT on Win7 / i5-2520M @ 2.5GHz:

 

 

http://260ste.appspot.com/tmp/Aranym_pi_out.zip

[moulinaie]

Time for calculation for 16384 digits: 2.20 s, under Aranym 0.9.10 JIT on Win7 / i5-2520M @ 2.5GHz:

 

 

http://260ste.appspo...anym_pi_out.zip

 

Yeah, not so bad!

It's like a Falcon CT060 at 940 MHz..!!

 

I add this to the list.

 

I know that someone has a Stacy+PAK030, that would be cool to know its bench.

 

Guillaume.

[+DarkLord]

That would be me. I'm working on this Guillaume. I've not got my keyboard hooked up at the

moment so I can't enter values. I'll try to get around to that by the weekend.

 

I did run TEST 6 : FPU test (68030+68882)

 

The time slices were:

 

first test - 4279 1/100th sec

 

2nd test - 3860 1/100th sec

 

 

Does that seem about right for a 32mhz 68030 with NO FastRam, and 68882 FPU (rated

at 50mhz, but honestly, I'm not sure what its running at - 32mhz as well?)

 

Thanks.

[moulinaie]

That would be me. I'm working on this Guillaume. I've not got my keyboard hooked up at the

moment so I can't enter values. I'll try to get around to that by the weekend.

 

I did run TEST 6 : FPU test (68030+68882)

 

The time slices were:

 

first test - 4279 1/100th sec

 

2nd test - 3860 1/100th sec

 

 

Does that seem about right for a 32mhz 68030 with NO FastRam, and 68882 FPU (rated

at 50mhz, but honestly, I'm not sure what its running at - 32mhz as well?)

 

Thanks.

 

You have the same results as a TT 32MHz with fast ram.

So if your Stacy don't have Fast Ram, you can be sure that the frequency is greater than 32MHz.

 

Guillaume.

[moulinaie]

That would be me. I'm working on this Guillaume. I've not got my keyboard hooked up at the

moment so I can't enter values. I'll try to get around to that by the weekend.

 

I did run TEST 6 : FPU test (68030+68882)

 

The time slices were:

 

first test - 4279 1/100th sec

 

2nd test - 3860 1/100th sec

 

Thanks.

 

I put your result on my page.

If you don't have the keyboard, I can send you a copy of SuperPI with a fixed number of digits (16384), so you won't have to type the value.

 

Guillaume.

[+DarkLord]

 

I put your result on my page.

If you don't have the keyboard, I can send you a copy of SuperPI with a fixed number of digits (16384), so you won't have to type the value.

 

Guillaume.

 

That would work. Just send it by PM if you like.

 

Its not that I don't have my keyboard, I've just got the whole thing apart where I'm still doing

some mods on it, as well as getting the PAK 68/3 board fully operational again.

 

Thanks!

[Cyprian]

Guillaume,

 

just to give you complete overview of that system, I've done the rest of your test:

 

Aranym 0.9.10 JIT on Win7 / i5-2520M @ 2.5GHz:

MiNT 1-17-cur/ XaAES v0.998 800x608x32@72

ST-RAM: 14MB; TT-RAM 64MB

 

 

TEST 2 : Creating an animation with M_PLAYER (68030+):

Total Time: 12.4 sec

Average: 8.9 frame/sec

 

 

TEST 3 : Replaying an animation w/o synchro with M_PLAYER (68030+):

Total Time: 0.3 sec

Average: 358 frame/sec

 

 

TEST 4 : raytracing with POV3 (68000+):

Total Time pov030: 6 sec

Total Time povste: 3 sec

http://260ste.appspo...p/stats_pov.zip

 

 

TEST 5 : Video conversion with MP_STE (68000+, STLow/STHigh):

this one freeze desktop

 

 

TEST 6 : FPU test (68030+68882):

Time 1: 131 - 2,18 sec

Time 2: 42 - 0,70 sec

Edited February 17, 2012 by Cyprian_K

[+DarkLord]

Okay, I ran the Super PI test.

 

Attached is a screen shot with the results:

 

 

Looks like my PAK 68/3 equipped STacy with a 6882 is staying

somewhere between TT's with no FastRAM and those with

FastRAM.

 

Thanks!

[+DarkLord]

And here is the results from the POV test:

 

woodbox.pov Statistics, Resolution 64 x 48

----------------------------------------------------------------------------

Pixels: 3072 Samples: 3072 Smpls/Pxl: 1.00

Rays: 7090 Saved: 286 Max Level: 5/5

----------------------------------------------------------------------------

Ray->Shape Intersection Tests Succeeded Percentage

----------------------------------------------------------------------------

Box 126575 16450 13.00

Cone/Cylinder 18069 3786 20.95

CSG Intersection 9197 5923 64.40

CSG Union 9197 6276 68.24

Plane 28760 11091 38.56

Sphere 165376 8246 4.99

Bounding Object 78252 9197 11.75

----------------------------------------------------------------------------

Calls to Noise: 3446 Calls to DNoise: 37875

----------------------------------------------------------------------------

Shadow Ray Tests: 197865 Succeeded: 4957

Reflected Rays: 2893

Refracted Rays: 1125

----------------------------------------------------------------------------

Smallest Alloc: 10 bytes Largest: 8200

Peak memory used: 293073 bytes

----------------------------------------------------------------------------

Time For Parse: 0 hours 0 minutes 13.0 seconds (13 seconds)

Time For Trace: 0 hours 3 minutes 35.0 seconds (215 seconds)

Total Time: 0 hours 3 minutes 48.0 seconds (228 seconds)

 

I'll attach STATS.OUT too, just in case.

 

STATS.OUT.zip

 

Thanks!

[+DarkLord]

TEST 5 : Video conversion with MP_STE (68000+, STLow/STHigh)

 

Okay, here are the results from this test.

 

First one is in ST-HIGH:

 

 

and the next one is ST-LOW:

 

 

Thanks.

[moulinaie]

Looks like my PAK 68/3 equipped STacy with a 6882 is staying

somewhere between TT's with no FastRAM and those with

FastRAM.

 

Yes exactly! So this card gives a lot of power to the Stacy in spite of its RAM running at 8MHz (16MHz for the ST Ram into a TT).

For the SuperPI test, PAK = TT

 

For the MP_STE conversion, a lot of memory access is performed and a lot of disk access too. That could explain that the Stacy is a bit slower than the TT in this case.

 

Again, for the POV test, a huge use of the FPU is done, is the 68882 clocked at 50MHz too?? If it's slower (32MHz or 25MHz) that could explain the difference with the TT.

 

But hey, you've got an incredible machine! You can go to my test page, I've added your results to the tables (BTW what TOS version have you got? how many MB of RAM?)

Will you put a better screen than the original? (you could include a PAL to VGA converter into the unused battery bloc to have at the same time ST LOW and ST HIGH on the same monitor.

 

Guillaume.

[+DarkLord]

 

Yes exactly! So this card gives a lot of power to the Stacy in spite of its RAM running at 8MHz (16MHz for the ST Ram into a TT).

For the SuperPI test, PAK = TT

 

For the MP_STE conversion, a lot of memory access is performed and a lot of disk access too. That could explain that the Stacy is a bit slower than the TT in this case.

 

Again, for the POV test, a huge use of the FPU is done, is the 68882 clocked at 50MHz too?? If it's slower (32MHz or 25MHz) that could explain the difference with the TT.

 

If I understand the workings of the PAK board, the CPU and FPU both take the signal from the crystal oscillator and run at that

speed. So for example, my 68030 is rated at 40mhz and the FPU is rated at 50mhz, but they both follow the master signal from

the crystal oscillator chip on the board, which is 32mhz. IF I've got this right.

 

But hey, you've got an incredible machine! You can go to my test page, I've added your results to the tables (BTW what TOS version have you got? how many MB of RAM?)

 

Thanks for that.

 

Umm, I'm running the modified English PAK TOS, version 3.06. Its what is required for the board to auto-boot correctly. I also have TOS 1.04 that I can use in compatibility mode (I flip a switch hooked to a jumper on the board and it will boot in either one). The compatibility mode actually does make it boot back up into a standard, stock, 8mhz ST again. Nice for software that misbehaves with the PAK. (that means games, mostly). For the purposes of your tests though, I ran all of them under PAK TOS v3.06.

 

I've got 2 megs of RAM. I've got a 4 meg board that I ordered from Best Electronics here, but for some reason it only reports 2 megs. I can see the RAM is on the board, so I don't know what is going on with that. I'll have to return it to Best Electronics for repair or replacement.

 

Will you put a better screen than the original? (you could include a PAL to VGA converter into the unused battery bloc to have at the same time ST LOW and ST HIGH on the same monitor.

 

Guillaume.

 

I'd love to replace the stock mono screen that the STacy has, with something better, and in color. I'm not sure about how to do that since its so hard to find a screen that will work with the ST's odd synched RGB signal. Oh, and on top of that, I'm here in the USA, so its NTSC and not PAL. SCART sure seems nice with PAL machines - I wish we had its equal here.

 

What I'm doing right now is using a scan converter between the STacy's external color monitor port on the back plane, and an old VGA monitor I had lying around. I'm just doing that while I've got it apart, doing all the upgrades and mods. Its also a nice way to play color games from the STacy, if you have an extra monitor handy. I do this at work when its slow.

 

Also, I removed the unused battery compartment from the STacy some time ago, to make room for the power supply which I had to move for the PAK board. So while there is still a little room left in the battery compartment area, above the power supply, its not much, so I'm not sure what I could fit in there.

 

There are a couple of message threads here at AtariAge and at AtariForum if you wanted to see what I've done and am doing to my STacy. Warning - they are quite long sometimes!

 

http://www.atariage.com/forums/topic/191506-first-pak-board-bootup-attempt/

 

http://www.atari-forum.com/viewtopic.php?f=15&t=22327

 

Thanks!

[moulinaie]

Guillaume,

 

just to give you complete overview of that system, I've done the rest of your test:

 

Aranym 0.9.10 JIT on Win7 / i5-2520M @ 2.5GHz:

MiNT 1-17-cur/ XaAES v0.998 800x608x32@72

ST-RAM: 14MB; TT-RAM 64MB

 

Hi Cyprian !!

 

I was wondering how to integrate those results to my page as there is really a great difference between emulators and the original machines.

Maybe should I create a new paragraph about emulators to compare Aranym, Hatari etc... on different machines, with or without JIT.

 

Guillaume.

[+DarkLord]

Well, just for the fun of it, I unplugged my PAK 68/3 board from my STacy

and put Jim Allen's/Fast Technology's T25 accelerator board back in, just

to run a few of these tests for comparison.

 

TEST 5 : VideTEST 5 : Video conversion with MP_STE (68000+, STLow/STHigh)

 

STLow gave this:

 

 

and STHigh gave this:

 

 

Thanks!

[+DarkLord]

Hmm, no go on this test:

 

TEST 4 : raytracing with POV3 (68000+)

 

I tried to run this with the T25 in my STacy and 2 megs of RAM, but it

kept giving an error about where it loaded up the woodmaps, saying

there wasn't enough memory...