Just a quick question. Is Blt*Mask*BitMapRastPort() hardware accelerated?

If not, does MorphOS have something similar that is h/w accelerated?

ChrisH,

Nope. I think the best alternative method use BltBitMapAlpha() or BltBitMapRastPortAlpha(). They are hardware accelerated where supported (most gfx cards). WritePixelArrayAlpha() is only altivec accelerated.

@Itix

That's odd... why these "Mask" functions don't use accelerated "Alpha" ones internally?

Crumb,
Quote:

This call has not been updated for quite a while because its most popular use for it has been the legacy Workbench icon display in the past and there has been not much use for it beyond that scope. Ambient uses the alpha (or even non-alpha) blitting calls.

@Crumb
I imagine that h/w accelerating Mask functions would not be efficient - every time you called BltMaskBitMapRastPort(), it would have to convert the supplied 1-bit mask bitmap (which could have changed) into an 8-bit (or more) alpha channel bitmap.

Since masks are stored in non-video ram, and alpha bitmaps are stored in video ram, that would involve the CPU reading every pixel of the mask bitmap & making a write to the alpha bitmap. If you were going to do that, it wouldn't be much slower for the CPU to just write the actual (unmasked) pixels to the bitmap target of BltMaskBitMapRastPort(). i.e. what it is already doing.

@itix
Thanks for the pointer. It looks like it is impossible for BltBitMapAlpha() to use a separate 8-bit alpha-channel-only bitmap? Although I can't even see a way to allocate such a bitmap using CyberGraphics anyway.

[ Edited by ChrisH 14.03.2012 - 09:50 ]

Right, alpha channel must be part of source bitmap. If you are operating with true colour bitmaps it is not problem but rules out using LUT or hicolor bitmaps.

CGX supports using alpha from destination bitmap but I dont know if this operation is HW accelerated.