[ Date Index ] [ Thread Index ] [ <= Previous by date / thread ] [ Next by date / thread => ]
On 21/11/10 11:27, Philip Whateley wrote: > On Sun, 2010-11-21 at 10:08 +0000, Paul Sutton wrote: >> >> Following on from the abve post, the same page has a link to >> http://www.bbc.co.uk/news/uk-11789191 > Looks really modern to those of us who started out programming FORTRAN > IV on an ICL 1903S mainframe and then graduated (literally in my case) > to a Commodore PET 2001-8. The sad thing is I remember working with pre-FORTRAN 77 code at the Met Office, and they didn't employ me till the early 90's. Occasionally one would open up a FORTRAN source file, think "this looks odd", and realise it doesn't use any features not present in FORTRAN H. Now it is possible that the code had existed unchanged since before I was born, although I suspect more likely someone was still coding that way long after it was necessary or desirable. Oh and that diagram from the BBC vastly misrepresents the change in processor speeds since 1981. I suspect another case of the clock speed fallacy. The diagram shows a performance change of about 1500 fold, when in Instructions per second the 8088 at 4.7MHz was capable of ~1 MIP, and the Core i7 3.3GHz in comparison can do 150,000 MIP. So the increase is at least a factor of 100 greater than is suggested by the diagram. MIPS* under represents the improvement substantially, but it is the first figure I could find that is likely to mean anything at all between the two eras. That the Core i7 has substantially more CPU cache than RAM the 8088 could address, just underscores that simple comparisons are rather meaningless (hence "meaningless instructions per second" - but I'd hazard an educated guess that the Core i7 meaningless instructions can do a lot more per instruction than an 8088). -- The Mailing List for the Devon & Cornwall LUG http://mailman.dclug.org.uk/listinfo/list FAQ: http://www.dcglug.org.uk/listfaq