Xref: utzoo comp.sys.intel:477 comp.sys.ibm.pc:17295 Path: utzoo!attcan!uunet!portal!cup.portal.com!mslater From: mslater@cup.portal.com Newsgroups: comp.sys.intel,comp.sys.ibm.pc Subject: Re: Raiders of the lost opcodes of the 80386 Message-ID: <7435@cup.portal.com> Date: 18 Jul 88 00:16:10 GMT References: <4071@korppi.tut.fi> <4814@killer.UUCP> Organization: The Portal System (TM) Lines: 34 XPortal-User-Id: 1.1001.4222 > correct opcodes for 386 bit string insert and extract instructions These instructions, which were documented in early 386 data sheets, are not supported. According to Intel's errata sheet: "Since the 80386 has unique and powerful 64-bit Double Shift instructions, and fast multi-bit shift and rotate instructions, the "Bit String Insert" and "Bit String Extract" instructions were removed. The insert/extract complex instructions did not provide an additional benefit that fully justified including them in 80386 silicon and all future compatible processors. A review concluded that the 80386 user obtains full performance in bit string manipulations using other powerful instructions such as 64-bit Double Shift, and other multi-bit shift/rotate instructions. These instructions support extremely fast manipulation of general unaligned bit strings of any length, by processing them in 32-bit chunks." The question that naturally arises is, so why were these instructions included in the architecture in the first place? I suppose we could be generous and take Intel's explanation at face value, and assume that the inclusion of these instructions was an overzealous action on the part of some architect that Intel later saw fit to rectify. On the other hand, I tend to be more cynical, and assume that these instructions didn't work, and that Intel decided to remove them from the spec rather than wait for another silicon pass. Intel has never gone out of their way to let people know that these instructions were dropped. Supposedly, everyone that purchased 386 chips got the errata sheet, and Intel notified the complier writers. And no one programs in assembly language any more, right? Michael Slater Editor and Publisher Microprocessor Report 550 California Ave., Suite 320, Palo Alto, CA 94306 415/494-2677 mslater@cup.portal.com sun!portal!cup.portal.com!mslater