View Single Post
Old 27 July 2019, 17:34   #679
Registered User

Join Date: Jul 2015
Location: The Netherlands
Posts: 1,287
Originally Posted by NorthWay View Post
It should _always_ save 1/4th if you can use the (rather safe) assumption that a chunky pixel with all bits clear means select source instead of bob.
This is only true if the chunky pixel is the same size as the bus, otherwise you are wasting bandwidth to do this. That was more or less the point I made in the explanation that followed the bit you quoted.

If you want something ZX Spectrum like with a mask around the bob then you would need as much bandwidth.
The optimal case however depends and memory width, blitter width, the number of parallell blitter units, and if you can have dynamic memory access sizing. I.e. no "zero" pixels in the bob data you read can simply convert the cookie-cut into a copy plus pointer advance. (Note that this does not work if you want to blit from a backup buffer to the viewed buffer, then we are back to 1/4th.) Then depending on circumstances you can juggle bus sizing to maximize the number of conversions.
I’m certain that planar blits can be made more efficient with tricks like the above as well. However, using the way that (soft) bobs actually were implemented in the real world means none of the above applies and my point remains valid.

As I’ve pointed out before, I don’t consider adding new hardware to one side of this comparison but not do the same for the other a fair way of comparing these things.
roondar is offline  
Page generated in 0.03986 seconds with 11 queries