yuval@taux01.UUCP (11/30/87)
When the bus-lock prefix is asserted (either explicitly, or as an automatic result of XCHG), how many interlocked bus-cycle are possible before the locking expires? in particular, can a misaligned LOCK STOSD instruction cause the 386 to write twice to memory in interlocked fashion? how many bus-transactions are interlocked on a misaligned LOCK MOVSD?
Bob-B@cup.portal.com (12/01/87)
>When the bus-lock prefix is asserted (either explicitly, or as an >automatic result of XCHG), how many interlocked bus-cycle are >possible before the locking expires? in particular, can a misaligned >LOCK STOSD instruction cause the 386 to write twice to memory in >interlocked fashion? how many bus-transactions are interlocked on a >misaligned LOCK MOVSD? God only knows. :-) In theory, though, a LOCK STOSD or LOCK MOVSD would cause an '86 processor to lock the buss for the entire length of the buss transaction. That is, if STOSD were to execute with one write, the LOCK would be just for that period. But, if the STOSD was to a misaligned place, the LOCK would be just extended ... it would cover both writes. ----- done help page forward back skip add_article unsubscribe transfer rotate oommand: SCREEN DUMP ANSI-BBS FDX 1200 N81 LOG CLOSED PRT OFF CR CR