Hi Denis,

30 Dec 98, Denis Tonn of 1:153/908 wrote to Eddy Thilleman:

DT> Thunking can be either expensive (in CPU overhead) or cheap
DT> depending on how it is done.

OS/2 v1.x versions used the 'expensive' conversion (heavy CPU overhead)?

DT> BTW: It is a fallacy to assume that 32 bit programs are always
DT> "faster" than 16 bit programs.

I assume this is a general remark because I don't remember saying that
32-bit programs are faster than 16-bit programs.

DT> The only Intel CPU this is true for is the Pentium Pro.

Only because the Pentium Pro handles 16-bit code poorly and 32-bit code efficient?

DT> As long as the program code fits into a 64K segment and the
DT> data fits into 2 (or 4 on a 386) 64K data segments the

What about on processors above the 386, like the 486 and all pentiums flavors?

DT> The 32 bit program does not need to reload a selector to
DT> access more than 256K.

A selector is a CPU register and acts (for the CPU/operating system) like a
pointer works (for an app) ?

DT> Yes. The tiling of the LDT depends on this, and the "minimum"
DT> conversion work is based on the tiling.

I see.

Cheers   -=Eddy=-   (eddy.thilleman{at}net.hcc.nl)

...  * <- Tribble  _ <- Tribble after unfortunate steam roller accident
--- MBM v4.14