Basically wot are graphics card interupts?
Ive seen that some people have had to disable them to get their GFX cards to run.
So wot are the benifits?
Does having them off greatly reduce performance?
Will they work with cards that they dont work with now in the future?

Cheers to anyone who clarifies this for me!

Curtis

Last edited by Curty on 07-Sep-2005 at 09:41 PM.

_________________
A1XE G4 800mhz,512mb,Radeon9200,SonyDRU-510A and 800a,ESI Juli@, Sil680, WiFi.

@Curty
Well.....*deep breath*, not sure how this applies to graphics cards, so hopefully olegil or someone else will jump in here, but hardware interrupts are basically a means to avoid polling of devices.

Instead of having software sit in a busy loop trying to read from a device, a device instead generates an Interrupt Request, which notifies the CPU (via a signal on the bus, in hardware) and is then passed off to an IRQ handler.

This is _somewhat_ like the difference between synchronous and asynchronous IO in programs- one can either wait to get something back from a device (like a serial port for example, but could be anything), or can do an asynchronous call which is then notified when data is available for reading. Ok, maybe a bad explanation, but software polling consumes CPU time 'doing nothing' while waiting for a signal from a device, interrupts don't.

I'm not sure what a graphics interrupt might do- notify the system soem GPU task is completed?

Anyone ablke to clarify this without more of my guessing here?

_________________
Are we not done with the same silly arguments and flames yet??!

@wegster

BUMP!

_________________
A1XE G4 800mhz,512mb,Radeon9200,SonyDRU-510A and 800a,ESI Juli@, Sil680, WiFi.

The interrupt occurs when the visible frame has been drawn.

_________________
: AmigaOneXE (unmod.) 750FX/512 MB +stuff & AmigaOS 4.(0|1)
: A1200/68060&96MB/SCSI/EM1200-Voodoo3 & OS 3.5
: A500/1MB
: Pegasos (ff) 512 MB & MorphOS
Praise seitan.

@Thematic

Quote:

The interrupt occurs when the visible frame has been drawn.

Some products offer many more signal options, such as beam is at top of frame, bottom of frame, at the end of a horizontal sweep, and there's a bunch more that I don't even remember what they all are now. Some graphics chips can also pass on interrupts for companion devices such as TV tuners, MPEG encode/decoders, etc.

_________________
All glory to the Hypnotoad!

@All
Thanks for the answers,
But it some of my questions still lay unanswered.
I still would like to know if you will experience any noticeable loss of performance if interupt are turned off?
cheers

_________________
A1XE G4 800mhz,512mb,Radeon9200,SonyDRU-510A and 800a,ESI Juli@, Sil680, WiFi.

@Curty

Quote:

Does having them off greatly reduce performance?

yes

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga

Ah, that clears things up.....NOT !

Curty, what do you say we string him up by his thumbs until he talks.

_________________
Hate tends to make you look stupid...

@sundown

Me thinks we should string him up by his nutts!

_________________
A1XE G4 800mhz,512mb,Radeon9200,SonyDRU-510A and 800a,ESI Juli@, Sil680, WiFi.

@NutsAboutAmiga

Can you elaborate on this anymore?

_________________
A1XE G4 800mhz,512mb,Radeon9200,SonyDRU-510A and 800a,ESI Juli@, Sil680, WiFi.

@billt

quoting Thematic:"The interrupt occurs when the visible frame has been drawn."

During this "particular" type of interrupt which used to be known as VBI
(vertical blank interrupt), is the available time frame to run code generated
by the interrupt the same regardless of graphics card? Does code that "overruns"
the vertical blanking period still cause graphic problems?

Also, does mouse movement still adversely affect interrupt code run during the
horizontal blanking period (horizontal sweep) in some cases?

This used to be the case with "Digiview" which shoved a new palette into the
color registers at the end of each scan line, while still trying to read the
mouse. Remember "dynamic hires mode" and the distortion that appeared when you
moved the mouse during the display of that mode?

Best Wishes,
#6

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

@Curty

Quote:

I still would like to know if you will experience any noticeable loss of performance if interupt are turned off?

Interrupts are used to schedule things on an as-needed basis. You don't need to keep looking in on things to dind out if something is needed, the graphics chip will use an interrupt to let the system know when it needs something, so less wasted resources in overhead.

Also, there are some things which use for example a vertical blanking interrupt as a timer. For some things like knowing when to update part of a screen's framebuffer or swap buffers in a double or triple buffering display. Without interrupts, you're polling to find out when to do things or just guessing, but either is more overhead then when using interrupts.

_________________
All glory to the Hypnotoad!

@number6

Quote:

is the available time frame to run code generated by the interrupt the same regardless of graphics card?

Depends on what the code does. Is it driver code, or application/game code? Also, the available time will change with different refresh rates. The faster refresh rate, the more vblanks per second, and the shorter time you have to run code during each vblank.

Quote:

Does code that "overruns" the vertical blanking period still cause graphic problems?

This could happen. If your code goes halfway into the next frame beaming onto the display, you might see something change half-way down or something like that of the new image is different enough.

Quote:

Also, does mouse movement still adversely affect interrupt code run during the horizontal blanking period (horizontal sweep) in some cases?

I don't know anything about that. I never had a Digiview.

_________________
All glory to the Hypnotoad!

@billt

Thank you for your informative answer.
I used to program interrupts (VBI & scanline) on 8bit computers and I had always
wondered if there was a similarity to their nature regarding modern day graphic
systems.
I have seen code "overrun" before. We used to have VBI timer as well that would
determine beginning of VBI time. Another vector could be examined to determine
end time. iirc though, we had to run the code to be absolutely sure that there
was no "overrun". Back then, we could "shorten" the screen display area to gain
more VBI time.
Your explanation was quite clear about refresh rates. I can see how that would
affect available VBI time.
Oh, the comment about "mouse movement" during display of "Digiview" dynamic
hi-res mode was more or less rhetorical. It DID affect the screen. Color registers
would "lose" their information temporarily and tremendous graphic disturbance
took place. I only asked the question because I was curious if you had seen this
particular effect. It was due to the miniscule amount of time available in
horizontal blank period. Since the mouse still had to be checked in main code,
while the mouse was stationary everything was fine. But when it detected a mouse
movement and had to process that...too much code I think and it "overran" the
horizontal interrupts. My opinion anyway.

Thanks again,
#6

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

@number6

So do you think that people that are unable to have interupts on now, will, be able in the future to have interupts on with say the final os4 version?
cheers

_________________
A1XE G4 800mhz,512mb,Radeon9200,SonyDRU-510A and 800a,ESI Juli@, Sil680, WiFi.

@Curty

Hi,

I am just a member here. I was merely relating some personal experience and my
personal understanding about some types of interrupts to Bill.
I wanted to know if there were similarities in the ways interrupts are handled
(compared to my -old- knowledge). I was also curious to know if the basic nature
of potential problems, like code overrun, was still applicable with current
interrupt implementation. Bill's answers were very educational for me.
I suggest that Bill answer your question, though, since he has FAR more knowlege
than I on this subject.

Best Wishes,
#6

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*