home *** CD-ROM | disk | FTP | other *** search
- Path: nntp.teleport.com!sschaem
- From: sschaem@teleport.com (Stephan Schaem)
- Newsgroups: comp.sys.amiga.programmer
- Subject: Re: FWD: Fate of 68080
- Date: 31 Jan 1996 12:31:45 GMT
- Organization: Teleport - Portland's Public Access (503) 220-1016
- Message-ID: <4ennfh$jcc@maureen.teleport.com>
- References: <4e7rhi$4fo@maureen.teleport.com> <4ealme$8fi@hades.datashopper.dk>
- NNTP-Posting-Host: kelly.teleport.com
- X-Newsreader: TIN [version 1.2 PL2]
-
- Michael Berg (mberg@datashopper.dk) wrote:
- : In a message of 25 Jan 96 Stephan Schaem wrote to :
-
- : SS> A 133mhz pentium is near 2 time slower then a 133mhz 604.
-
- : If the 604 is a RISC processor, then I am sure glad I don't have a 604 in
- : my machine. What a drag!
-
- the test didn't mesearu an instruction... but an operation...
-
- : (For the slower readers -- RISC means Reduced Instruction Set, which in
- : turn means the processor has to execute many more instructions than a CISC
- : architecture to achieve the same thing. I don't recall the typical
- : RISC/CISC factor, but it's certainly more than 2).
- : __
-
- Look at the PPC instruction set and addressing mode... it compare very
- well the the cisc the 68000 is.. risc mainly is having fixed size
- instruction, then of course only including instruction that really impact
- performace.
-
- In general its not more then 2.... This is not a proff or anything but I
- compiled a loop under a few CPU to see, and in 68020+ it took 13 inst 6register, in pa risc 14 inst 6 register,
- ppc 18 inst 6 register, 486 16 inst 7 reg... (The loop was interesting to me
- because it was somethng that would take maybe 80% of the CPU time at some
- stage of the program, and I was looking for a new platform :)
-
- (BTW the loop used on the 68020, cisc instruction like :
- move.size (offset,Saddress,index*type),(Daddress)
-
- Well, the pa risc seem to do pretty good with its 'reduced' inst set :)
- only 1 extra instruction.
-
- And actually risc can code in less instruction...
-
- C = A + B;
-
- This is coded in 1 instruction on most risc... 2 on most cisc
-
- Some risc can do this in 1 inst (The 68060 would need 3)
-
- C = (A + B) * 2;
-
- BTW
-
- C = (0 + B) is equal to C = B
- C = -C is equal to C = (0 - C)
-
- So risc dont need to code a register move or neg instruction. they reduce
- but are not slower (In that example). Also having lots of register
- is a risc 'trademark' , and if they are 64bit like many of the
- latest risc its another advantage.
-
- risc are not as bad as you might think...
-
- BTW, I said that the instruction ratio is not more then 2, (could
- actually be less :) ... but the code size dont reflect this.
- If you average 680x0 code take ~3 byte per instruction, risc 4 byte.
-
- Stephan
-