qintar@agora.UUCP (Jim Seymour) (11/23/89)
Forgive me if this is not the appropriate newsgroup to ask this, but...
We need some information regarding a bug/feature in the 80286.
(It may be old hat to some of you, but it's new to us).
During a rep string operation that exceeds the limit of a segment, the
registers at the time of the exception have been changed and the
instruction can not be restarted where it left off. In the case of
a rep movsb (forward direction) to an expand-down segment, CX has been
decremented, and SI/DI have been incremented *before* the exception is
processed. Thus, the exception handler (when it tries to recover)
always misses the first byte.
On the 80386, everything works fine. Does anybody have any information
(or even a confirmation) on this? *Any* help would be greatly
appreciated.
You may contact me via e-mail at qintar@agora.UUCP, or by other means
as below. Please do not post a reply as I do not read this newsgroup.
(My work phone is 503/647-0849).
-Jim Seymour ...tektronix!tessi!agora!qintar
=================================================================
Cipher Systems, Inc. Home:
P.O. Box 329 1308 S.E. Division
North Plains, OR 97133 Portland, OR 97202