Maggie Programming Forum
<< MAGGIE PROGRAMMING FORUM >>
<< WITH MR. PINK / RESERVOIR GODS >>
Greetings coding obsessed people and welcome to another "Maggie Programming Forum" which gives you, the reading public, the chance to ask us, the writing public, technical questions.
Our first query comes from Mr. Ed Clevland of Liverpool.
"About a month ago I bought myself a Jaguar controller and tried to use it on my STE. Basically your Powerpad routine will not work properly on this machine. The routine as supplied in Maggie will not accept some keypad/fire presses and will not do diagonals. I started mucking about with it myself and eventually got it almost fully working. The problem seems to be connected with timing. With your routine you write the mask to $ff9202 once and then do two reads for the data. I've found that for the STE you need to write the mask, read one set of data, then write the mask again and read the other set of data otherwise the second lot of information is lost. I assume because the Falcons' processor is faster than the STE's it can 'grab' the data before it vanishes, and so you wouldn't see this problem.Even so, it still won't work properly for me. I can't get diagonals to work on it, but sometimes 'up/right' will, and pressing button A cuts out the directional data. Even stranger, I re-wrote the routine using the information you gave in your Maggie article and put it into a program to display the values from the joypad registers in decimal. The diagonals worked, but only in that specific program. If I cut and past this routine out of that program into another, the diagonals stop working.
I assume it's all to do with timing, there must be something you have to sync with to get proper data out of the joypad, like the way the microwire works on the DMA sound chip. With this you have to write a mask to one register and then wait until it returns to its original value then you can write more data to it. I'm sure there must be a similar method for the joypad."
Thanks for the enquiry Ed. My first advice to you is: GET A NEW PRINTER! Or at least a new cartridge as the type is so blank we thought at first that someone had sent us a blank sheet of paper through the post!
When I wrote my powerpad routine I was trying to make them as optimal as possible, doing the minimum amount of writes to hardware. The routine works fine on my Falcon, and also runs perfectly on my STE. Perhaps you have a dodgy powerpad or port?
You are right about the timing problems though. You will definitely get timing errors, if you are not careful, when interfacing with the extended joystick ports. I find that if you access then at the start of the VBL the results you get are fine, otherwise they can be unpredictable. Perhaps this is due to a fault in my routines, but VBL syncing seems to iron out all the errors. I suggest that the reading the powerpads should be your first routine after waiting for the vertical blank:
e.g. bsr wait_vbl
bsr read_powerpadsThis will cause the minimum of errors. You may get problems if your read_powerpads routine is interrupted by an interrupt, by the time this returns you may have lost valuable data. The way to stop this is blank the status register at the start of your powerpad routine:
read_powerpads:
move.w sr,-(a7) ;save status register
ori.w #$0700,sr ;interrupts off
... ;(read powerpad code)
move.w (a7)+,sr ;restore status register
rts ;returnYou still may get problems if the interrupt occurs after the wait vbl and before the read powerpads, or if it occurs before the VBL reaches to bottom of the screen, but carries on through part of the next frame. The way to solve this is to call your read powerpads routine from the Vertical Blank itself. You should have a VBL handler that sits on $70.w, updates a counter and calls any VBL specific routines (NB it is not a good idea to put your whole game on the VBL like some people do!!). You still have the problem of a higher priority interrupt interrupting the vbl routine. This can be stopped by blanking the status register at the start of the vbl. You need not worry about saving the contents of the old sr as this is automatically stored on the stack when the interrupt occurs and popped off by the RTE instruction.
my_vbl:
ori.w #$0700,sr ;interrupts off
bsr read_powerpads ;call read powerpads routine
addq.l #1,$466.w ;update vbl counter
rtsI hope this solves all your problems.
Last issue I discussed the new technique of hardware fading in true colour mode by messing around with the videl registers. This was an effect invented by Creator of Aggression and is now common place in Falcon demos. I mentioned that the drawback that I was having was that the border colour increases as I implement the fade down as the background palette colour has to be increased.
Longshot from Swedish coding team Impulse writes to Maggie to give us some advice.
"True Colour fading - in Maggie 19 you wrote about this cool trick. But you say that 'the border colour increases as the main picture decreases'. I used this fading technique last year in Angest, Bugs From Outer Space and now in Corsair, and it's working beautifully. Perhaps you are referring to a little raster bar in the top border? If so, sit back and have a cup of tea, hehe...This can be eliminated by clearing the VBE together with the HBE. That is, "move.w #0,$ff82a6". I don't know whether this is such a nice thing to do, but I can't think of any other way to eliminate that colour bar at the top. While I discovered this, I also discovered some nice things to do with these registers... :)If there are any queries, check out the (horrible) source code found on the disk".
Thanks Longshot for your advice and code. The routine we are interested in is the hardware fade, and Longshot has supplied us with two different versions of this, one using standard logical shifts and the other using the new 68030 bitfield instructions.
hfade_tc_rgb:
move.w #0,$ff8286 ; horizontal border end
move.w #0,$ff82a6 ; vertical border end
movem.l d1-d2/a0,-(sp)
lea $ff9800,a0 ; RRRRRr.. GGGGGg.. ........ BBBBBb..
move.w d0,d1 ; green intensity (0-63)
lsl.w #2,d1
move.w d0,d2 ; red intensity (0-63)
lsl.w #2,d2
lsl.w #8,d2
add.w d2,d1
move.w d1,(a0)
move.w d0,d1 ; blue intensity (0-63)
lsl.w #2,d1
move.w d1,2(a0)
movem.l (sp)+,d1-d2/a0
rts
; This is a shorter way to set the register, using the 030 instruction
set:
hfade_tc_rgb2:
move.w #0,$ff8286 ; horizontal border end
move.w #0,$ff82a6 ; vertical border end
movem.l d1-d2,-(sp)
bfins d0,d1{0:6} ; red intensity
bfins d0,d1{8:6} ; green intensity
bfins d0,d1{24:6} ; blue intensity
move.l d1,$ffff9800.w ; RRRRRr.. GGGGGg.. ........ BBBBBb..
; 0 8 16 24
movem.l (sp)+,d1-d2
rtsThanks again Longshot. The fading, as you say, is very smooth, although we still get a rather bright border as the background colour approaches $fcfc00fc.
If you have any programming queries (in any language) or other technical questions, drop us a line and we'll try to sort you out:
Leon O'Reilly ( Mr. Pink )
Maggie Programming Forum
Cwm Isaf
Abermule
Montgomery
Powys
SY15 6JL
Pterm until next time!
