Page 2 of 6

Re: Slow STFM blitter oddness

Posted: 31 Dec 2019 19:09
by pixelpusher
From a software perspective it could either be GB6 (it's obviously not written in C, so hard to say what the runtime does) or it could be related to a slower ROM access on the STFM (the "VDI scroll test" code seems to call some dialog [GEM?] functions, which then - I guess - trigger some scrolling internally).

I'd guess there's more going on in this GB6 test than just a simple VDI call. There was once a great tool - SysMon - that allowed tracing/loggin all OS calls an app (and its runtime) does ...

The Blit test in GB6 on the other hand is really using the VDI raster copy operation - just like it does in GEM_TEST.

Re: Slow STFM blitter oddness

Posted: 31 Dec 2019 19:15
by exxos
Its not ROM access, because the same "Fault" happens when using my booster which has its own ROM decoding logic.

Re: Slow STFM blitter oddness

Posted: 31 Dec 2019 19:37
by pixelpusher
exxos wrote: 31 Dec 2019 19:15 Its not ROM access, because the same "Fault" happens when using my booster which has its own ROM decoding logic.
Then chances are that the (Basic?) runtime used by GB6 does more than just calling VDI functions - use SysMon, if you have it .... I'd also check the VDI text output with with GEM_TEST to see if it confirms this odd speed drop (and if it does, then double-check with NVDI installed too).

Getting the same speed in the blit test (using the VDI raster op) but a lower speed in the text output doesn't make sense from a software pov.

Re: Slow STFM blitter oddness

Posted: 31 Dec 2019 19:43
by exxos
pixelpusher wrote: 31 Dec 2019 19:37 Then chances are that the (Basic?) runtime used by GB6 does more than just calling VDI functions - use SysMon, if you have it .... I'd also check the VDI text output with with GEM_TEST to see if it confirms this odd speed drop (and if it does, then double-check with NVDI installed too).

Getting the same speed in the blit test (using the VDI raster op) but a lower speed in the text output doesn't make sense from a software pov.
Well, I am not a programmer guy, so no idea about software debugging :( Even so, GB6 runs fine on the STE with blitter.

Its not like ROM access speed is the issue. In fact ROM is running 50MHz on my booster, faster than STE's 32MHz booster, and still tests are slower. If RAM access was a issue, then all tests would be slower (blitting test etc). So direct RAM copy of data seems to be working fine.

In both cases, STFM with blitter TOS206, vs STE with blitter TOS206, gives different results.


EDIT:

Does anyone have a MEGA ST running TOS206 with blitter they can try GB6 on ?

EDIT2:

Actually as PaulJ said the H4 is basically OK.. then it must be some STFM related hardware bug...

Re: Slow STFM blitter oddness

Posted: 31 Dec 2019 19:53
by PhilC
I can possibly test it on a Mega tomorrow as out at the mo.

Remember that the Mega usually records 99% on tests due to the slightly slower oscillator.

Re: Slow STFM blitter oddness

Posted: 31 Dec 2019 19:54
by exxos
PhilC wrote: 31 Dec 2019 19:53 I can possibly test it on a Mega tomorrow as out at the mo.
Does it have TOS206 though as that is what is under testing...

Re: Slow STFM blitter oddness

Posted: 31 Dec 2019 19:55
by PhilC
It has flashy clock in it, will that do?

Re: Slow STFM blitter oddness

Posted: 31 Dec 2019 19:56
by exxos
PhilC wrote: 31 Dec 2019 19:55 It has flashy clock in it, will that do?
Yep :)

Re: Slow STFM blitter oddness

Posted: 31 Dec 2019 20:57
by exxos
@PaulJ Do you have a STFM you could run TOS206 and blitter on ?

Is it the MM9092 blitter ?

I've got sidetracked into a STE vs STFM issue.. but as the H4 looks basically fine... its looking like some hardware issue on the original STFM which somehow we have Inadvertently fixed on the H4... Either that or my STFM I am using has something bad on it.

Re: Slow STFM blitter oddness

Posted: 31 Dec 2019 22:14
by PaulJ
exxos wrote: 31 Dec 2019 19:54
PhilC wrote: 31 Dec 2019 19:53 I can possibly test it on a Mega tomorrow as out at the mo.
Does it have TOS206 though as that is what is under testing...
Numbers look ok since this mega 4 is running @ 32.04 instead of 32.08. Also numbers suck like your seeing if acc's and programs are in startup. This is with XBOOT only, no acc's.

IMG_2415.jpg
IMG_2415.jpg (845.13 KiB) Viewed 4249 times