Main > Software Forum

My official MAME output wip thread.

<< < (28/31) > >>

bdam:
Here's the data unaltered except for me zeroing out the bits we're not interested in and surrounding the values with pipe symbols.  The Min and Max values aren't particularly useful to show the reversal that's needed so I added a more descriptive example.
Min(0): 00000000|111111|000000000000000000
Max(63): 00000000|000000|000000000000000000
Other (5): 00000000|101000|000000000000000000

Good idea, I'll look around and see if I can find an existing function in MAME.

demeth:
http://graphics.stanford.edu/~seander/bithacks.html -- Study until you know them all by heart ;-)

bdam:
Yep, that's where I got my solution.  To be clear, the 'problem' is easily solved; it's just a matter of where we solve it.  In MAME or in the program that reads the output from MAME.

Howard_Casto:

--- Quote from: demeth on February 26, 2012, 11:59:20 am ---http://graphics.stanford.edu/~seander/bithacks.html -- Study until you know them all by heart ;-)


--- End quote ---

Yeah, like bdam said, the "problem" here isn't that we don't know how to fix it... it's that fixing it could slow down the mame driver and/or introduce sloppy code.  Mind you the speed decrease would almost definately be unnoticable, but he system 22 drivers aren't exactly speedy to begin with.

Personally I think that so long as it's only being reversed when the value is changing it should be ok. 

demeth:
Ah, that wasn't obvious (to me) from reading the related posts.

Looking at the driver in question, that kindof operation we're talking about shouldn't even register on the scale considering all the other things that's done in there indicative on the cleanup need for mame-3d overall,
heck, even just restrict keywording the A / B in matrix3d_Multiply should save that driver a whooole lot more (not considering the ~20-25ish% gain on using SIMD- intrisics..) than you could purposefully waste :P

Navigation

[0] Message Index

[#] Next page

[*] Previous page

Go to full version