Blitter Vertical Fill - Page 2

BroFistMcfury · 04 June 2016, 18:49

I'm using an old version 2.3.3 but I just tried latest 3.2.2 and get the same result.

Toni Wilen · 04 June 2016, 18:59

It was already answered. It uses BLTCDAT which is uninitialized. (=garbage depends on KS version and what you did before you run the code)

BroFistMcfury · 04 June 2016, 19:10

Oh, so you're saying I should change the function to include BLTCDAT initialization like this?

BlitFillVertical:
bsr BlitWait
move.w d2,BLTAMOD(a5) ;set modulo A
move.w d2,BLTCMOD(a5) ;set modulo C
move.w d2,BLTDMOD(a5) ;set modulo D
move.w #0,BLTADAT(a5)
move.w #0,BLTCDAT(a5)
move.w #0,BLTDDAT(a5)
move.w #SRCA|SRCC|DEST|A_XOR_C,BLTCON0(a5)
move.w #0,BLTCON1(a5) ;fill mode
move.l #-1,BLTAFWM(a5)
move.l a0,BLTAPT(a5) ;source A
add.l d1,a0
move.l a0,BLTCPT(a5) ;source C
move.l a0,BLTDPT(a5) ;destination D
move.w d0,BLTSIZE(a5)
rts

BroFistMcfury · 06 June 2016, 11:25

My minterm is $5a by the way, which I believe to be correct.

Toni Wilen · 06 June 2016, 12:52

Something like that. Probably. I will only debug existing binary code

BroFistMcfury · 06 June 2016, 13:41

Heh - fair enough!

BroFistMcfury · 08 June 2016, 15:09

Found the problem!

A_XOR_C EQU NABC|ABNC | NANBC|ANBNC

Can you see the mistake?

StingRay · 08 June 2016, 17:31

Quote:

Originally Posted by BroFistMcfury

Found the problem!

A_XOR_C EQU NABC|ABNC | NANBC|ANBNC

Can you see the mistake?

I can.

BroFistMcfury · 09 June 2016, 20:29

Thanks StingRay for pointing out that the minterm wasn't compiled correctly because of the space between the constants!

...thanks to everyone else as well!

Leffmann · 09 June 2016, 23:30

If your assembler gives you a different result when adding a space in there, you need to find a better assembler

BroFistMcfury · 11 June 2016, 10:55

Too right!!!

sandruzzo · 12 June 2016, 06:31

what would be the advantage to have vertical fill?

BroFistMcfury · 12 June 2016, 10:30

As Dan said, you can do huge sin scrollers

dalton · 16 November 2017, 21:55

I'm trying to realize the following function using the blitter. It effectively fills one 16-bit column of a 28-byte wide bitmap. The height is 256 rows.

This loop does the job:

Code:

		eor.w d0,(a0)		; first row

		move.w #HEIGHT-2,d1
.loop
			move.w (a0),d0
			lea 28(a0),a0
			eor.w 	d0,(a0)

			dbf d1,.loop

So I wrote this piece of blitter code to do the same thing, and it simply doesn't do the job:

Code:

		move.w	#$0b5a,bltcon0(a1) ; D = A~C + ~AC
		move.w	#$0000,bltcon1(a1)

		move.w	#$ffff,bltafwm(a1)
		move.w	#$ffff,bltalwm(a1)

		move.w	#26,bltamod(a1)
		move.w	#26,bltcmod(a1)
		move.w	#26,bltdmod(a1)

		move.l	a0,bltapt(a1)
		lea	28(a0),a0
		move.l	a0,bltcpt(a1)
		move.l	a0,bltdpt(a1)

		move.w	#64*(HEIGHT-1)+1,bltsize(a1)

Prior to calling the routine, A1 is set to $dff00, and blitter done is asserted. It's a single threaded program and multi-tasking has been disabled. There are also other blitter operations in the program which works as intended. I've tried rearranging the order of blitter operations with no improvement.

I have some very similar code to clear the very same area, and that works just fine.

Code:

		move.w	#$01f0,bltcon0(a1) ; D = A~C + ~AC
		move.w	#$0000,bltcon1(a1)

		move.w	#$ffff,bltafwm(a1)
		move.w	#$ffff,bltalwm(a1)

		move.w	(a0),bltadat(a1)

		move.w	#26,bltdmod(a1)

		lea	28(a0),a0
		move.l	a0,bltdpt(a1)

		move.w	#64*(HEIGHT-1)+1,bltsize(a1)

Is there perhaps some special consideration that should be taken for one column blitter operations?

I'm banging my head against the wall here.

dalton · 16 November 2017, 22:31

I believe I figured it out. Due to pipelining in the blitter, the sources are fetched twice before the destination is written. Effectively making it impossible to do vertical fill on just one column.

16 November 2017, 21:55	#34
dalton tulou Join Date: Jun 2006 Location: Gothenburg / Sweden Posts: 88	I'm trying to realize the following function using the blitter. It effectively fills one 16-bit column of a 28-byte wide bitmap. The height is 256 rows. This loop does the job: Code: eor.w d0,(a0) ; first row move.w #HEIGHT-2,d1 .loop move.w (a0),d0 lea 28(a0),a0 eor.w d0,(a0) dbf d1,.loop So I wrote this piece of blitter code to do the same thing, and it simply doesn't do the job: Code: move.w #$0b5a,bltcon0(a1) ; D = A~C + ~AC move.w #$0000,bltcon1(a1) move.w #$ffff,bltafwm(a1) move.w #$ffff,bltalwm(a1) move.w #26,bltamod(a1) move.w #26,bltcmod(a1) move.w #26,bltdmod(a1) move.l a0,bltapt(a1) lea 28(a0),a0 move.l a0,bltcpt(a1) move.l a0,bltdpt(a1) move.w #64(HEIGHT-1)+1,bltsize(a1) Prior to calling the routine, A1 is set to $dff00, and blitter done is asserted. It's a single threaded program and multi-tasking has been disabled. There are also other blitter operations in the program which works as intended. I've tried rearranging the order of blitter operations with no improvement. I have some very similar code to clear the very same area, and that works just fine. Code: move.w #$01f0,bltcon0(a1) ; D = A~C + ~AC move.w #$0000,bltcon1(a1) move.w #$ffff,bltafwm(a1) move.w #$ffff,bltalwm(a1) move.w (a0),bltadat(a1) move.w #26,bltdmod(a1) lea 28(a0),a0 move.l a0,bltdpt(a1) move.w #64(HEIGHT-1)+1,bltsize(a1) Is there perhaps some special consideration that should be taken for one column blitter operations? I'm banging my head against the wall here.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Blitter fill timing	leonard	Coders. Asm / Hardware	42	01 September 2014 11:00
Clipping line for blitter fill	leonard	Coders. Asm / Hardware	12	27 April 2013 12:03
80 GB HD to fill!	fatboy	Amiga scene	16	20 July 2011 14:13
Sector fill pattern	absence	Coders. General	7	21 March 2009 21:50
Fill 'em	Tim Janssen	request.Old Rare Games	1	27 June 2003 09:25

04 June 2016, 18:49	#21
BroFistMcfury Registered User Join Date: May 2016 Location: London Posts: 14	I'm using an old version 2.3.3 but I just tried latest 3.2.2 and get the same result.

04 June 2016, 18:59	#22
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,504	It was already answered. It uses BLTCDAT which is uninitialized. (=garbage depends on KS version and what you did before you run the code)

04 June 2016, 19:10	#23
BroFistMcfury Registered User Join Date: May 2016 Location: London Posts: 14	Oh, so you're saying I should change the function to include BLTCDAT initialization like this? BlitFillVertical: bsr BlitWait move.w d2,BLTAMOD(a5) ;set modulo A move.w d2,BLTCMOD(a5) ;set modulo C move.w d2,BLTDMOD(a5) ;set modulo D move.w #0,BLTADAT(a5) move.w #0,BLTCDAT(a5) move.w #0,BLTDDAT(a5) move.w #SRCA\|SRCC\|DEST\|A_XOR_C,BLTCON0(a5) move.w #0,BLTCON1(a5) ;fill mode move.l #-1,BLTAFWM(a5) move.l a0,BLTAPT(a5) ;source A add.l d1,a0 move.l a0,BLTCPT(a5) ;source C move.l a0,BLTDPT(a5) ;destination D move.w d0,BLTSIZE(a5) rts

06 June 2016, 11:25	#24
BroFistMcfury Registered User Join Date: May 2016 Location: London Posts: 14	My minterm is $5a by the way, which I believe to be correct.

06 June 2016, 12:52	#25
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,504	Something like that. Probably. I will only debug existing binary code

06 June 2016, 13:41	#26
BroFistMcfury Registered User Join Date: May 2016 Location: London Posts: 14	Heh - fair enough!

08 June 2016, 15:09	#27
BroFistMcfury Registered User Join Date: May 2016 Location: London Posts: 14	Found the problem! A_XOR_C EQU NABC\|ABNC \| NANBC\|ANBNC Can you see the mistake?

09 June 2016, 20:29	#29
BroFistMcfury Registered User Join Date: May 2016 Location: London Posts: 14	Thanks StingRay for pointing out that the minterm wasn't compiled correctly because of the space between the constants! ...thanks to everyone else as well!

09 June 2016, 23:30	#30
Leffmann Join Date: Jul 2008 Location: Sweden Posts: 2,269	If your assembler gives you a different result when adding a space in there, you need to find a better assembler

11 June 2016, 10:55	#31
BroFistMcfury Registered User Join Date: May 2016 Location: London Posts: 14	Too right!!!

12 June 2016, 06:31	#32
sandruzzo Registered User Join Date: Feb 2011 Location: Italy/Rome Posts: 2,281	what would be the advantage to have vertical fill?

12 June 2016, 10:30	#33
BroFistMcfury Registered User Join Date: May 2016 Location: London Posts: 14	As Dan said, you can do huge sin scrollers

16 November 2017, 22:31	#35
dalton tulou Join Date: Jun 2006 Location: Gothenburg / Sweden Posts: 88	I believe I figured it out. Due to pipelining in the blitter, the sources are fetched twice before the destination is written. Effectively making it impossible to do vertical fill on just one column.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)