More Intel Appendix H secrets to be released (4M Pages)
>> [...] I'm ready to liberate another Appendix H topic: 4M Pages.
> If you want to see the _real_ stuff, you can take a look at any recent
> linux kernel. The linux-1.3.x series of kernels (from x >= 15 or so)
> use the 4MB page table feature automatically. This is real code that
> has seen real use (and which you can take a real look at)..
Okay. But Robert (and of course his unmentioned coworkers) has reverse-
engineered the whole stuff, before Linux came with 4MB paging. It's not
so hard to reverse-engineer this iPentium secret -- other's were harder
and took much more time.
> And no, I never signed any NDA: the details may be hidden away in some
> Appendix H, but the actual working of the chip was easy to guess at with
> only minimal trial-and-error from the hints given in any Pentium
> reference manual. It does work, and it does help performance somewhat.
...almost all the iPentium books I read during the last few years point
directly to those officially Intel documents and manuals, where you can
read the detailed and complete 4MB paging description -- except these 3
volume iPentium databook set itself.
If you read the CISC KING STORY carefully, then you know where you must
_BTW:_ Now, when I have contact to you, Linus, are you interested in some
sort of an API for accessing some Modelspecific Register related things
on iPentium processors? I am thinking of an API, which allows to create
Performance Monitoring programs, like my STAT.EXE (for DOS) from the 4P
package (see my footer), for Linux too. If you are interested, then you
can contact me via private eMail, please. Thanks in advance.
Ludwig-Kuehn-Str.15 FAX ++49-371-242091 http://www.x86.org/4p
D-09123 CHEMNITZ actual: 4p_v313.zip ftp://ftp.x86.org/x86
## CrossPoint v3.02 ##