[email protected] (Tom Marchant) writes: > Bold assertion. Do you have any data to back that up?
360 allowed self-modifying instruction ... potentially instructions already in the pipeline. one of the claims for Amdahl "macro-code" was that it was essentially 370 with some tweaks ... including precluding self-modifying. even in some number of (non-cache, non-pipeline) 360s that did i-fetch a double word (or more at time) ... there was extra overhead in constantly checking whether there was storage alteration to address that was already fetched by the instruction unit (within the same double word). "harvard" architectures with split instruction and data caches w/o cache concistency .... get some processing performance by i-cache ignoring standard storage alterations. in the case of "store-in" data cache ... program loaders that may operate on instruction streams ... making alterations that appear in d-cache. before initiating exection ... the loader then executes special operation that explicitly forces d-cache alterated lines to main storage and invalidates possible corresponding i-cache lines. with the force of altered data (from d-cache) to storage and invalidation of corresponding addresses in i-cache ... subsequent instruction stream references to those addressed would be forced to fetch the contents from storage (explicit programming required to support alteration of potential instructions). "harvard" architecture with i-cache operation ignoring standard storage alterations ... is much easier to pipeline and easier to scaleup multiprocessor (not only being able to ignore storage alterations on the same processor ... but also able to ignore standard storage alterations on all the other processors). -- virtualization experience starting Jan1968, online at home since Mar1970 ---------------------------------------------------------------------- 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
