hypotetical blitter question

[Rock'n Roll]

move.w #$0100,$dff040 ; BLTCON0: only channel D is activated

the blitter is a well designed hardware "chip", but I asked me for
a long time is the bit for output channel D (USED) useless? Because every
blitt needs an activated channel D, so this bit is always set to 1.

Are there blitter operations where channel D could be set to 0?
(From my side, the USED could be internaly hardwired set to 1,
if BLTCON0 is set to "normal" mode.)
And then this bit would be free for other things, maybe as a Shifter-
Select Bit (Bit8=0 - ASH, Bit8=1 CSH for a C-Shifter? )

2nd Question: There is no shifter for channel C, but there are
4 unused bits in BLTCON1 (bits 11 to 8 are set to 0). It could be
used for CSH3 - CSH0. Has a C-shifter never been missed?

[Jobbo]

One use is for collision detection. You run a blit with no stored results and just check BZERO bit in DMACONR for any collisions.

Not something I've tried but that seems like the main intent of the design.

[Rock'n Roll]

yes, of course, I don't remembered this! The example code could be found in lezione10h1.s in the ramjam course.

The other point: Maybe it makes no sense to discuss things that not came, but hypothetical a C-shifter would
give us more freedom. The Cookie-Cut D=AB+(NOT A)C is in this form, because only the channels A and B has a
shifter and channel C not. So, the channels are not interchangeable in any combination.

An other limitation what I don't like is:
Why is lea bitplane,d0 not possible? The command lea could also load a 32Bit-adress in a dataregister.
So we have always to write a move.l #bitplane,d0. (But that's the way it is.)

[a/b]

Because it's not logical. Data and address regs are not identical functionally, each group has a specific purpose. Also, lea opcode doesn't have room to specify 16 regs.

I would rather complain about eor not having <ea>,dx version, tst pc-relative modes being 020+, index scaling being 020+ (makes me cry every time).