home *** CD-ROM | disk | FTP | other *** search
- Path: grafix.xs4all.nl!john.hendrikx
- Date: Sat, 03 Feb 96 17:35:15 GMT+1
- Newsgroups: comp.sys.amiga.programmer
- Distribution: world
- Subject: Re: Amiga doesn`t need Planar!
- MIME-Version: 1.0
- Content-Type: text/plain; charset=iso-8859-1
- Content-Transfer-Encoding: 8bit
- From: john.hendrikx@grafix.xs4all.nl (John Hendrikx)
- Message-ID: <john.hendrikx.4biy@grafix.xs4all.nl>
- Organization: Grafix Attack BBS Holland
-
- In a message of 02 Feb 96 Michael Van Elst wrote to All:
-
- >> Yes, but a planar system needs more memory accesses on average because of
- >> more severe alignment restrictions than Chunky,
-
- MVE> or less because of different needs of bits per pixel.
-
- Not really important as it is 8-bit, 16-bit, 24-bit or 32-bit in today's world,
- well except Amiga that is.
-
- Thank god C= didn't invent 'planar' sound hardware...
-
- >> In other words
- >> Planar can't take advantage of wider memory-busses as good as Chunky can.
-
- MVE> That's garbage. With a wider memory bus you increase alignment needs
- MVE> for both systems. And with your CPU-only system you can't use much
-
- True, but with a 8-bit bus the Chunky alignment needs are non-existant, while
- planar already has 8-pixel alignment with that bus. It gets worse for planar
- far faster as planar has a headstart in this.
-
- MVE> beyond 32bit. Of course that's not a real chunky problem.
-
- And I never said the system would be CPU only, would be quite ridiculous to do
- that. It is just an advantage of Chunky that the CPU can handle the display
- very well too.
-
- MVE> On the other hand it is pretty simple to use the wider bus to access
- MVE> completely different regions of the RAM. You can scale the planar
- MVE> hardware with the bus width a operate on multiple planes in parallel.
-
- Hmmm... so what you're saying is that if you had a 64-bit memory bus, and
- 'somehow' managed to rearrange memory in such a way that the same bytes of
- every 8 bitplanes are located next to each other you could modify them in just
- one access? Ie, if you plot a pixel in 8-bit you could do it in one access?
- That would be nice, but if you do that you lose the main advantage of planar,
- fast manipulation of single bitplanes. It would effectively turn your planar
- hardware into Chunky hardware, with almost the same advantages and
- disadvantages, except that to the CPU it LOOKS like Planar.
-
- Question is why would you ever want to do that? Not only does it require much
- wider memory busses to get good speed, you also lose the primary advantage of
- bitplanes (accessing them one plane at the time). On top of that you make it
- hard for the CPU to use this display as it is still planar. It doesn't make
- sense to do this unless you were doing it to be compatible or something with
- hardware banging programs.
-
- When I come to think of it this isn't possible at all unless you got some
- adaptive memory hardware or something. I mean, what would happen if you
- displayed 7 bitplanes this way? Would you have 1 byte unused after every 7 used
- bytes? What if I opened a 2nd screen? Could I use that 'extra' byte for a
- bitplane of a different screen? This would also kill Planar's advantage of
- having multiple bitplane pointers as they would need to be at specific
- locations in memory (kinda like a Interleaved bitmap). I don't think this is a
- reasonable possibility at all, in terms of cost and effort. You might as well
- go Chunky right away and save yourself the mess.
-
- >> As you can see the best planar gets is equal speed, but only under very,
- >> VERY specific circumstances.
-
- MVE> Why don't you compare wider rectangles ? The more data you have the
-
- We were talking about thick lines (which I changed to rectangles to make things
- easier), I couldn't make them any wider for just lines.
-
- MVE> smaller are the differences. Why don't you compare the cases where not
- MVE> 8 bit per pixel are needed ?
-
- Because those are irrelevant, 8 bits per pixel is the standard, if planar can't
- handle that fast than planar simply sucks.
-
- >> Chunky wins hands down in the last case however. So as I said, the line
- >> would have to be very thick to get good speeds on planar.
-
- MVE> Don't forget the overhead for the CPU to determine BYTE and WORD
- MVE> accesses, even more when you don't draw straight vertical lines.
-
- Who's talking about the CPU? It would be kinda unfair to compare CPU Chunky vs
- Hardware Planar (although even then Planar doesn't look too good). The fact
- that the CPU can 'help' with Chunky is just a bonus, there still is Chunky
- hardware of course.
-
- >> Drawing a horizontal line would of course be better on planar (although
- >> still equally fast in the best case), but as in that case the speed is
- >> about the same it is not really interesting.
-
- MVE> Sure. Just the worst cases are interesting for you.
-
- Actually only near horizontal lines are 'good cases'. As soon as the line is
- more than 45 degrees vertical the 'worst case' can be applied.
-
- >> The point is that if it takes more memory accesses to do a specific
- >> action than no matter what hardware you throw at it the number of memory
- >> accesses will remain higher.
-
- MVE> No. Try a 12bit/pixel display and don't forget to count all the
- MVE> overhead involved in the CPU to handle packed data.
-
- A fast CPU would laugh at this 'overhead' as memory accesses are so slow it can
- easily handle the extra calculations needed 'in between' memory accesses (like
- 040 C2P). And there is still the Chunky hardware which can do it for you...
-
- Anyway, I explained why I think wider memory busses wouldn't work as good with
- Planar in an other message to you.
-
- Grtz John
-
- -----------------------------------------------------------------------
- John.Hendrikx@grafix.xs4all.nl TextDemo/FastView/Etc... development
- -----------------------------------------------------------------------
- -- Via Xenolink 1.981, XenolinkUUCP 1.1
-