On Wed, 23 Jan 2008 12:50:54 -0500, Craddock, Chris <[EMAIL PROTECTED]> wrote:
>> BTW, do the contents of the ECB other than the WAIT and POST bits >> matter to WAIT? IOW, instead of "XC ECB,ECB" (or CS) would it suffice >> to do as little as "MVI ECB,x'3F'", or would it disrupt the fabric >> of the universe? (Don't ask "why"; mere scientific interest.) > >No the universe will be fine. The only bits that matter to WAIT or POST >are X'80' and X'40' which I sardonically observe to be not exactly a >scientific test on any storage that notionally belongs to a user. But I >daresay I have already flogged all the dust off that deceased equine. > >Anyhow, to answer your original question; Yes. Provided you clear both >of those bits "somehow" (in fact, in any way you like) you will get the >result you expect. The storage block concurrency rules ensure you will >see your change, or the other guy's change, but not bits of both. Um... Surely XC is a classic example of an instruction that cannot be safely used to concurrently update a storage location. Even though XC of a location with itself is block concurrent within a doubleword, that ensures only that all the bytes are seen to be fetched or stored at once. There can be two fetches and one store, as observed by other CPUs. This is fine if there's only one XC ever clearing the ECB (other tasks just look), but if two tasks can both use XC at the same time (or XC and any other non-interlocked store instruction), the other store reference could occur between the two fetch refs of the XC, and the exclusive or could be done using different values, yielding a non-zero result. Yeah, yeah - in the real world this ain't gonna happen, since any known processor is going to be optimized to just set the location to zeros. But it's still not a correct technique if more than one task could be setting the ECB content. Tony H. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
