In a message dated 10/9/99 12:06:01 PM Eastern Daylight Time,
[log in to unmask] writes:

<<
 Maybe we could have achieved this level of sessions if we had fewer people
 running everything from within HPDesk...or if we'd bought SPLash! All of
 our terminal io from native mode programs go through a CM switch to
 SPL/assembler routines Jeff Kell wrote long ago.  However, if we'd done
 that, we probably still have the 960 and not the 969/120.:-)

 >
 > The secret on the 960 was to use job-management software to
 > single-thread Quiz, even after it was compiled with PDQ.  At least,
 > that's what we think helped.

 IIRC, PDQ is CM, but in it at least a batch report would mostly stay in CM
 (or is it OCTCOMPed?).
  >>

We OCTCOMPed everything immediately.
I started converting programs to native mode as soon as the 960 went into
service, in 1991.  The TRANSACT programs were easy, but the FORTRAN
all had to be converted from FORTRAN/66 to FORTRAN/77.

Yes, PDQ is CM, but even without OCTCOMPing, PDQ produces an
incredible improvement in performance.  QUIZ reports are not allowed
into production unless they are compiled with PDQ, and the users do not
have access to QUIZ interactively.  (They use INFORM and REPORT.)

Cecile Chi