Leif Rundberget (as quoted by Joel C. Ewing) wrote:

| Be careful with like terms between the PC(intel) world and the mainframe
| world.  When someone says they have a 64-bit Intel server (Intel,
| Solaris, AMD, etc.), it does not mean that the server can access an
| address 64-bits long, the 64-bits refers to the width of the bus.  So it
| can transfer 64-bits in parallel.

There is once widely used IBM terminology for this quantity. As far back as the System/360, a model 30 had a FETCH WIDTH of one eight-bit byte and the fastest models, as they came along, had fetch widths of eight bytes.

Higher fetch width is "good", but it can have odd side effects. There were System/360 models for which Load Halfword (LH) was slower than Load (L) not just because of sign propagation but also because LH fetched four or eight bytes from storage two or six of which then had to be discarded.

Analogously, there are z/Architecture situations in which eight-byte (64-bit) instructions are faster than their four-byte (32-bit) variants.

John Gilmore
Ashland, MA 01721-1817
USA

_________________________________________________________________
Don’t just search. Find. Check out the new MSN Search! http://search.msn.click-url.com/go/onm00200636ave/direct/01/

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html

Reply via email to