home *** CD-ROM | disk | FTP | other *** search
- Path: starlt.demon.co.uk!starlt
- From: AJ Heywood <starlt@starlt.demon.co.uk>
- Newsgroups: comp.lang.c++,comp.programming,comp.arch
- Subject: Re: Why are 32 bit better than 16 bit pgms?
- Date: Tue, 6 Feb 1996 21:26:24 +0000
- Organization: Chez_Nous
- Distribution: world
- Message-ID: <ky9$hCAAe8FxEwST@starlt.demon.co.uk>
- References: <4er4m4$78q@news1.ucsd.edu> <1996Feb5.163838.24531@amc.com>
- <1996Feb6.135808.12257@friend.kastle.com>
- NNTP-Posting-Host: starlt.demon.co.uk
- X-NNTP-Posting-Host: starlt.demon.co.uk
- MIME-Version: 1.0
- X-Newsreader: Turnpike Version 1.11 <+A$wb3xWjrCPkFMJ3Tqn0Pp3Dc>
-
- OK - so why are visual basic 32 bit commands slower then their
- 16 bit equivalents?
-
-
- In article <1996Feb6.135808.12257@friend.kastle.com>, Richard Krehbiel
- <rich@kastle.com> writes
- >curtis@amc.com (Curtis Green) wrote:
- >
- >>The "bits" quoted for the processor (16 bit, 32 bit, 64 bit, etc) is the
- >>size of the data path (data bus).
- >
- >[rest cut - I just want to gripe on this one point. :-) ]
- >
- >Why did Intel and IBM call the 8088 a 16 bit CPU? It has an 8 bit
- >bus.
- >
- >Why call the 386SX a 16 bit CPU when it can run any 386DX (32 bit)
- >code?
- >
- >Why don't they call the Pentium a 64 bit CPU since it has a 64 bit
- >data bus? Why doesn't DEC call the Alpha AXP a 128 bit CPU (or is it
- >256 - I'm not sure).
- >
- >What in the hell could you call the Motorola 68020/030 when for each
- >bus transaction it may choose 8, 16, or 32 bits as requested by the
- >peripheral being addressed?
- >
- >My humble opinion:
- >
- >"Bitness" has become a marketing term, i.e. meaningless (rather like
- >the term "RISC"). There are a dozen different pieces inside a CPU
- >these days each with it's own bit width. Outside the chip there's the
- >data bus and address bus width. Inside the chip you have the general
- >register width, special register widths (i.e. Segment register = 16
- >bits, FP register = 80 bits), ALU width (multiplied by the number of
- >ALUs if superscalar), internal-cache-to-register path width(s), etc.
- >Then there's the architectural virtual address size versus actual
- >virtual address size versus the physical address size versus the
- >implemented MMUs capabilities. (Intel advertised that the 386 could
- >address 64 Terabytes - which is baloney. They added up 13 bits of
- >segment plus 32 bits of offset. They somehow forgot that this "45
- >bit" virtual address is then funnelled into a 32 bit paged MMU.) You
- >probably also have to consider what mode the OS software's running
- >too.
- >
- >What really matters is how fast it'll run what software. You might as
- >well forget making any generalizations based on bit size.
- >
- >--
- >Richard Krehbiel, Kastle Systems, Arlington VA USA
- >rich@kastle.com (work) or richk@mnsinc.com (personal)
- >
-
- --
- AJ Heywood
-