home *** CD-ROM | disk | FTP | other *** search
- Path: grafix.xs4all.nl!john.hendrikx
- Date: Sat, 30 Mar 96 23:04:01 GMT+1
- Newsgroups: comp.sys.amiga.programmer
- Distribution: world
- Subject: Re: AB3D II beats Quake....
- 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.4ph5@grafix.xs4all.nl>
- Organization: Private
-
- In a message of 28 Mar 96 Stephan Schaem wrote to All:
-
- >> On CISC, it's not possible, because opcode are not 32 bit aligned. This
- >> means that before decoding intstruction i, you must decode instructions 0
- >> to i-1.
-
- SS> Thats not a problem really... x86 nowdays have a risc core and decode
- SS> the x86 'language'. I heard that maybe 18% of the P6 is actually
- SS> x86 related the rest is just risc design.
-
- Actually I heard that the P6 just decodes EVERYTHING which might be an x86
- instruction and if it later turns out that it actually wasn't a real
- instruction (because an earlier instruction was longer than 1 byte) it just
- discards the results of the fake instructions. That's wasting an incredible
- amount of power.
-
- >> This way RISC can also implement powerful branch prediction, which tend
- >> to add no overhead whether the branch is taken or not. Such prediction
- >> technology are not usable in CISC ; using them would mean adding thousand
- >> of transistors that could be used to speed up other instructions.
-
- SS> The P6 seem to show that cisc with alot of effort can perform pretty
- SS> well.
-
- Sure, but I bet it costs Intel more than 10 times as much money to get the P6
- to perform as well as the PPC604. Just think of what the PPC604 could have
- been with 10 times as large a budget. Also I think integrating a huge cache on
- the chip had a LOT more to do with the current performance of the P6 (and of
- course the usual overinflated Intel specmarks).
-
- >> Again, I don't agree. The problem is not the size of the opcode, but the
- >> time needed to execute it. Allowing 1 byte opcode means you won't be able
- >> to do pipelining and predecoding of the instructions flow. I don't think
- >> any chip firm today would go that way, ie using 1 byte opcode.
-
- SS> Its hard to say what would be the best instruction size/format...
-
- Maybe, but I think there are definitely very good reasons not to use 16 bit or
- 8 bit instructions sizes anymore.
-
- >> >Intel is not dumb, they said 3 years ago what I understood nowadays.
- >> >Time for other people to understand it as well. >
- >> Intel is producing mass CPU, not clever CPU. I'm much more interested in
- >> work and advices from HP, MIPS, ...
-
- SS> Intel also design advance risc that even SGI used for high end
- SS> geometry engine. HP also use intel risc in mass quatity. Intel
- SS> is not stupid and has ALOT of resource to take crap design like
- SS> the x86 and turn it around to be a performer.
-
- Performer? Why not divide the 'performance' by the price-tag and compare it
- with other chips.
-
- >> One of the big problem with the x86, is the poor number of register and
- >> the way they have to be used. Really, having 32 or 64 regs (PPC) greatly
- >> helps speeding up execution (as an ASM supporter, I think you will agree
- >> on the importance of the number of registers).
-
- C compilers like lots of registers as well, especially if they are general
- purpose registers. It not only makes good compilers easier to write (there are
- a lot less rules to be taken into account), it probably also makes for faster
- compile times (less special cases to check and optimize).
-
- Grtz John
- -- Via Xenolink 1.985B5, XenolinkUUCP 1.1
-