GEMBENCH 6

[exxos]

Current benchmark files.

CT6066 - CT60 @ 66mhz
DEFAULT - STE TOS 206 8mhz stock
STE20608 - same as above
STE20616 - STE TOS 206 @ 16mhz
STE20632 - STE TOS 206 @ 32mhz
FALCON30 - Stock falcon
STFMT28 - STFM running T28 booster (thanks Sophie)
STFM1048 - STFM TOS 104 @ 8mhz (stock machine)
PAK030 - STacy with PAK 030 booster (thanks Darklord)
CT6090SV - CT60 @ 90mhz with Supder Videl (Thanks Fredrik)

These will be included in the next build 23.

[GadgetUK164]

Thanks for continuing to update Gembench =D

[DarkLord]

Ditto. Tis a fine effort indeed.

[ThorstenOtto]

> I've been working on Build 15 for some days now and fixed a lot of bugs.

Is that available somewhere? I can't find any links. I've tried GB606B13.zip, but all i get is

"Arithmetic overflow at line 2385 in file E:\GB6\GUI14.BAS"

Then the program waits for a keypress and exits.

[exxos]

> I've been working on Build 15 for some days now and fixed a lot of bugs.

Is that available somewhere? I can't find any links. I've tried GB606B13.zip, but all i get is

"Arithmetic overflow at line 2385 in file E:\GB6\GUI14.BAS"

Then the program waits for a keypress and exits.

I'm on build 23.. But its not live yet, build 22 is, but might still have the issue you mention above.

http://www.exxoshost.co.uk/atari/last/GB6/index.htm

[ThorstenOtto]

> I'm on build 23.. But its not live yet, build 22 is, but might still have the issue you mention above.

Yes it has the same issue (line number is slightly different, 2349 for build 22). Just for info, i'm trying to run this on an emulator, not real hardware (SToNX in this case), so you can expect some numbers to get quite large. I had similar problems already with other benchmarks, either displaying negative numbers or overwriting the text fields in the GEM dialogs

[exxos]

> I'm on build 23.. But its not live yet, build 22 is, but might still have the issue you mention above.

Yes it has the same issue (line number is slightly different, 2349 for build 22). Just for info, i'm trying to run this on an emulator, not real hardware (SToNX in this case), so you can expect some numbers to get quite large. I had similar problems already with other benchmarks, either displaying negative numbers or overwriting the text fields in the GEM dialogs

This issue should be fixed in the next build. Someone else was running a emulator and scores over 32768% caused the overflow error. In build 23, it will hit 99999%, anything higher will just say "error". I hope the new build 23 will be ready sometime over the next few days.

[troed]

Hi exxos,

FYI, it seems Gembench 6 doesn't pick up cache enabled on MSTE, as well as not running the FPU test even though it detects the FPU. Gembench 4.03 works fine. See screenshots in this thread: http://www.atari-forum.com/viewtopic.php?f=15&t=30068

(The MSTE in question has a 68882, it's possible the FPU test works fine on 68881 - that I don't know)

/Troed

[exxos]

GB6 doesn't read the cache status, it just turns it on, and sets the cache on if found. d.m.l wrote that code, so will have to wait for him to cache up there GB6 should say cache on when it loads. Also use the cache functions in GB6 not any external desk acc's as it will likely confuse the cache status. I have a feeling Doug mentioned something about the MSTE cache, might need fixing, but neither of us have a MSTE (IIRC)

FPU,if it finds the FPU, it should run the test. Though don't forget there is no FPU test files yet, so it will not show a result, even though it should actually run the test. If you use a Falcon benchmark file then you should see a FPU value, as I assume I did the benchmark with the FPU installed.

I need to install monster in my STFM and run the test to create a new benchmark file, though someone else mentioned this a few weeks ago. I just don't have time to setup a STFM and do the test currently. There is no alt-ram benchmark in any files yet either, might be in the CT60 benchmark file.

[troed]

FPU,if it finds the FPU, it should run the test.

"FPU NOT FOUND" comes up when it should, even though (as you can see in the screenshot) it has identified an I/O mapped FPU

I'd be happy to beta test any changes, of course.

/Troed

[exxos]

FPU,if it finds the FPU, it should run the test.

"FPU NOT FOUND" comes up when it should, even though (as you can see in the screenshot) it has identified an I/O mapped FPU

I'd be happy to beta test any changes, of course.

/Troed

Strange, If its found the FPU it should run the code. I will take a look at the code quickly and see if I can find anything, thanks for the report.

[exxos]

I'd be happy to beta test any changes, of course.
/Troed

GB6.zip

There ya go, give that PRG a try.

[troed]

I'd be happy to beta test any changes, of course.
/Troed

GB6.zip

There ya go, give that PRG a try.

Well, now I'm sure it tries to run the test. 11 (edit: verified, it's 11) bombs though It _might_ be that the test expects a 68881, not 68882. Worth verifying - it's a newly written test, right?

This explanation might give a clue:

"A line 1010/1111 Emulator Error occurs when the processor tries to execute an invalid instruction. The code 1010/1111 is not really relevant (the code depends on the invalid instruction you tried to execute)."


http://www.atari-forum.com/viewtopic.ph ... 96#p121251

(The FPU test in Gembench 4.03 works fine)

/Troed

[exxos]

I have a feeling IO MAPPED was not programmed by Doug in the ASM routines. Which might have been why I bypassed the test to stop the crashing. IIRC doug hasn't done any IO Mapped programing ,so really it would need someone to add one to GB6's assembly code. I know myself and doug talked a lot about the code, but it was some months ago so can't remember exactly. IIRC we thought it wasn't worth spending the time on doing the code, probably because MSTE's don't come with a FPU as standard ?

[troed]

I have a feeling IO MAPPED was not programmed by Doug in the ASM routines. Which might have been why I bypassed the test to stop the crashing. IIRC doug hasn't done any IO Mapped programing ,so really it would need someone to add one to GB6's assembly code. I know myself and doug talked a lot about the code, but it was some months ago so can't remember exactly. IIRC we thought it wasn't worth spending the time on doing the code, probably because MSTE's don't come with a FPU as standard ?

Sounds reasonable, and true, only upgraded Mega ST (SFP004 expansion board) and Mega STEs (68881/68882) are I/O mapped I think. The latter come with an empty socket for the FPU so it's somewhat less of a barrier than an expansion board.

/Troed

[rpineau]

FPU on 68000 is always I/O Mapped (SFP004 and MegaSTE). The FPU test in GB6 are for 68020 and up with direct FPU access (via FPU instruction).
To test the FPU on 68000, specific code using the mapped I/O would be required.
Rodolphe

[DarkLord]

My STacy that we used for testing GB6 has a 68882 in it, via the Pak/3 board.

[frank.lukas]

MagiXMac Set ...

MAGIXMAC.GB6.zip

[exxos]

MagiXMac Set ...

MAGIXMAC.GB6.zip

Is that run a on a PC/emulator type thing ?

[frank.lukas]

Magic for Apple OSX

[ragnar76]

is there any guideline to run the tests? i've rediscovered tos2win (beside hatari, steem and stemulator gold) and it runs with the beta from above. results are in the attachment

Edit: I think something is broken. I get a lot of errors in the gui but inside the exported file it looks correct. also the results of gfx, cpu and average are all at 12000%

[exxos]

GB6 isn't designed to run on emualtors, pretty much the fastest machine it was based on was the CT60 at 100mhz. Likely you have errors as your %'s are in like 6 figures

I think in a later build anything over 5 figures just comes up "ERROR" now.

Latest builds are posted here http://www.exxoshost.co.uk/atari/last/GB6/index.htm but not had time yet to release a new version. There is only minor changes in the next build anyway.

[ragnar76]

hmkay.... have to dig out my ct63 falcon with 100mhz and report the results

[BlankVector]

I have just tested GemBench 6 (GB606 B22) on EmuTOS for Amiga and it does not work. I don't know if it is expected or not.

For memories, EmuTOS for Amiga is the same as EmuTOS for Atari, except that:
- direct access to Atari hardware is not supported (and will never work as hardware is missing)
- BIOS/XBIOS calls specific to Atari hardware are not supported, other ones (most of them) are supported
- only ST-High video mode is supported

What happens with GemBench6 :
- the program starts and shows its initial dialog box
- GEM Dialog Box: the dialog disappears, then it hangs.
- VDI tests: the window opens, then it hangs.
- GEM Windows: works fine.
- Integer Division: hangs
- RAM Access: hangs.
- ROM Access: hangs.
- Blitting: displays a rectangle filled with pattern, then hangs.
- VDI Enquire : hangs.
- Alt-RAM : hangs.

[exxos]

I have just tested GemBench 6 (GB606 B22) on EmuTOS for Amiga and it does not work. I don't know if it is expected or not.

Its a Atari program to run on a Atari machine with Atari TOS. Likely will end in disaster trying to run it on basically a alien machine.