Horizontal Blanking

[sandruzzo]

Hi All

I would like to trigger an interrupt every time that the Horizontal Blanking occur,
How could I do that? It's a been long time since my last conding for Amiga, and I have to polish my skills a bit . I would like to do somenthing within the Horizontal Blankings' Time.

thanks a lot

[Leffmann]

Trigger the interrupt with a write to INTREQ in your Copper program.

[StingRay]

Originally Posted by sandruzzo

Hi All

I would like to trigger an interrupt every time that the Horizontal Blanking occur,

I suppose you mean the vertical blanking interval here? If so, patch the level 3 interrupt vector ($6c.w, take care of the VBR on 680x0 CPU's) and enable the the level 3 interrupt in INTENA (bit 5, VERTB).

[Thorham]

Originally Posted by StingRay

I suppose you mean the vertical blanking interval here? If so, patch the level 3 interrupt vector ($6c.w, take care of the VBR on 680x0 CPU's) and enable the the level 3 interrupt in INTENA (bit 5, VERTB).

And if not, then make a copper list that waits for the end (or start) of each scan line and trigger a copper interrupt after that.

[sandruzzo]

Thanks for your helps. I meant Horizontal not Vertical Blanking. I would like to do some test
in order to take advantage of this Horizontal blanking, Like copy speed. So I have to write a copperlist that wait all the lines (200 or 256), and after each wait, I have to a move that trigger this Interrupt, doesn't ?

thanks

[pmc]

You would have to wait for the first non visible horizontal wait pos on the right hand edge of each scanline and then trigger your interrupt by writing intreq.

[Toni Wilen]

What kind of setup? Use of "hblank" interrupts on non-fast RAM/<68020 CPU setup is waste of CPU time. It takes over 50 CPU cycles to enter and exit an interrupt.

[StingRay]

Originally Posted by sandruzzo

Thanks for your helps. I meant Horizontal not Vertical Blanking. I would like to do some test
in order to take advantage of this Horizontal blanking, Like copy speed.

I am still not sure that you mean horizontal blanking. (TC) C2P copyspeed routines (if that is what you're referring to with copyspeed) take advantage of the vertical blanking because chipmem access is faster then.

[sandruzzo]

StingRay,

Yes I meant horizontal blanking too. I would like to see what kind of outcome I'll see. Just to test...

[pmc]

Still don't get what you're trying to do personally - post your test code maybe?

[sandruzzo]

Just to see how quick or how slow could be to works with chip-mem when bitplane DMA aren't using it. I still have to write my code..

[StingRay]

Originally Posted by pmc

Still don't get what you're trying to do personally

You're not the only one. For me it doesn't make sense. But I'll wait and see.

[sandruzzo]

hehe! Nothing fancy. Just to play with interrupts.

[Thorham]

sandruzzo just wants to know how much speed can be gained if you also use the horizontal blank and not just the vertical one.

[StingRay]

I understood that. It still doesn't make sense to me!

[sandruzzo]

Thorham

I don't know how much speed can be gained. I want to see what would be difference about accessing during Hblank or during normal rasterline.

[Kalms]

You can do some estimations at least.

A line is approximately 228 buscycles. If you have a 320 pixels wide lores screen, then the bitplane DMA-on portion of that screen is 160 buscycles.

So you have 160 buscycles worth of DMA-on and 68 buscycles of DMA-off.

During the DMA-on portion, on a lores screen you will only see a performance degradation if you have more than 4 bitplanes active. On an A500, having 5 bitplanes active will steal 25% of the DMA slots. On an A1200, having 5/6/7/8 bitplanes active will steal 6.25%/12.5%/18.75%/25% of the DMA slots, respectively.

(If you are running in hires/shres then a lot more buscycles will be used up by bitplane DMA.)

Now the kicker is, how much time do you lose by triggering the interrupt? if you target another machine than standard a500, how do you adjust the amount of work being done during the interrupt to match the performance of the machine?

I never tried it myself because I assumed that the overhead of the hblank interrupt would outweigh the few buscycles saved. Also, I normally target faster machines and there it is difficult to make the code self-adjusting.

[mc6809e]

Originally Posted by sandruzzo

Hi All

I would like to trigger an interrupt every time that the Horizontal Blanking occur,
How could I do that? It's a been long time since my last conding for Amiga, and I have to polish my skills a bit . I would like to do somenthing within the Horizontal Blankings' Time.

thanks a lot

Are you attempting to get the CPU to use available odd chipram cycles at the beginning of a scan line? Maybe you're hoping to get the CPU to run concurrently with the blitter when no other DMA is occurring.

My advice would be to experiment with a copper list that triggers interrupts at various places along each horizontal line to discover when you get the greatest level of concurrency. The proper place will most likely vary with the instruction sequence you wish to execute. It's possible the best place to trigger an interrupt is somewhere in the middle of a scan line.

Other programmers are forced to interrupt on hblank or vblack. You have an Amiga. You can trigger an interrupt whenever you damn well please.

[sandruzzo]

Kalms , mc6809e

thanks for your reply. I just want to play a little bit with Amiga interrupt, like I neved did, and see if I can get a better blittler and cpu work when they have as much cycle as possible. but since you already did it...