lotto@wjh12.harvard.edu (Jerry Lotto) (02/26/88)
In article <716@dsacg3.UUCP> ntm1458@dsacg3.UUCP (John Darby) writes: >From article <158@wjh12.harvard.edu>, by lotto@wjh12.harvard.edu (Jerry Lotto): >> it does not matter in this case. If we assume that there are 10 bits, >> addresses above 3ffh will "fold". If there are 12, addresses above >I thought the AT class machines like the Z-248 were 16 bit data path. They are. I/O port addressing can be done in two different ways (and maybe more) for a 286 architecture, I/O mapping and memory mapping. AT class machine use I/O mapping. I/O mapped devices reside in a private 64K I/O address space. This is decoded by different logic on the MB than the memory addresses. My recollection was that only the low order 10 bits were decoded and presented to the bus. Therefore, looking at port 7FFH will get you the value at 3FFH. John Levine had stated that there were 12 bits decoded, perhaps this is a PS/2ism. I arrived at the 10 bit number on a PC and an AT through experimentation, but I am not a studly hardware hacker and cannot read the schematic well enough to state why. Would someone in the know please comment on what particular feature of the MB drops the hi-order bits? >> 0fffh "fold". Therefore, com3 will be found at 220h in either case, as >What is "fold"? What I meant was that the hi order bits would be ignored. >> not sure about the GPIB adapter (it seems to do strange things with >What is a GPIB adapter? General Pupose Interface Bus, I guess that it was an HP standard. IBM sells a card that uses I/O space in this area, but it seems to use the bus directly and thus decode more than 10 or 12 bits. I don't own one so I cannot tell for certain. -- Gerald Lotto - Harvard Chemistry Dept. UUCP: {seismo,harpo,ihnp4,linus,allegra,ut-sally}!harvard!lotto ARPA: lotto@harvard.harvard.edu Newsgroups: comp.sys.ibm.pc,comp.sys.zenith,dla.pc Subject: Re: DOS 3.3 COM Ports Summary: Expires: References: <158@wjh12.harvard.edu> <716@dsacg3.UUCP> Sender: Reply-To: lotto@wjh12.UUCP (Jerry Lotto) Followup-To: Distribution: usa Organization: Harvard Univ. Chem. Dept. Keywords: In article <716@dsacg3.UUCP> ntm1458@dsacg3.UUCP (John Darby) writes: >From article <158@wjh12.harvard.edu>, by lotto@wjh12.harvard.edu (Jerry Lotto): >> it does not matter in this case. If we assume that there are 10 bits, >> addresses above 3ffh will "fold". If there are 12, addresses above >I thought the AT class machines like the Z-248 were 16 bit data path. They are. That is not what we are discussing. I/O port addressing can be done in two different ways (and maybe more) for a 286 architecture, I/O mapping and memory mapping. AT class machine use I/O mapping. I/O mapped devices reside in a private 64K I/O address space. This is decoded by different logic on the MB than the memory addresses. My recollection was that only the low order 10 bits were decoded and presented to the bus. Therefore, looking at port 7FFH will get you the value at 3FFH. John Levine had stated that there were 12 bits decoded. I am not a studly hardware hacker and cannot read the schematic well enough to prove things either way. Would someone in the know please comment? >> 0fffh "fold". Therefore, com3 will be found at 220h in either case, as >What is "fold"? What I meant was that the hi order bits would be ignored. >> not sure about the GPIB adapter (it seems to do strange things with >What is a GPIB adapter? General Pupose Interface Bus, I guess that it was an HP standard. IBM sells a card that uses I/O space in this area, but it seems to use the bus directly and thus decode more than 10 or 12 bits. I don't own one so I cannot tell for certain. -- Gerald Lotto - Harvard Chemistry Dept. UUCP: {seismo,harpo,ihnp4,linus,allegra,ut-sally}!harvard!lotto ARPA: lotto@harvard.harvard.edu