4 ms·
I wish I could find the part written by J.E. Robertson. I took several graduate level courses from him as an undergraduate in the mid '60s. He became the worl
by DonGateley 12y ago
I wish I could find the part written by J.E. Robertson. I took several graduate level courses from him as an undergraduate in the mid '60s. He became the world expert on division. He found that if you recoded to the redundant base -1, 0, 1 then number encodings could be found that had lots of zeros which he could optimize over and speed the division up. There were other sequential configurations he did special things with as well. You wouldn't believe the amount of theory he deduced from studying the redundant encoding problem and its application. The result was SRT division. He was the R and I can't remember who S and T were.
SRT division was implemented in the Illiac II.
- pbsd 12y agoSRT is named after Sweeney [1], Robertson [2], Tocher [3], which came up with the idea independently of each other. There is also a 1956 paper by Nadler said to have a similar algorithm, but it seems impossible to find. [1] http://www.computerhistory.org/collections/catalog/102632302 http://www.computerhistory.org/collections/catalog/102632302 [2] https://archive.org/details/newclassofdigita82robe https://archive.org/details/newclassofdigita82robe [3] http://qjmam.oxfordjournals.org/content/11/3/364.full.pdf http://qjmam.oxfordjournals.org/content/11/3/364.full.pdf
- DonGateley 12y agoThanks. [3] is essentially the outline of the first course I took from him. It didn't occur me that SRT division might have any recent relevance, ancient history and all that, but Google tells me that the Pentium floating point bug was in fact a buggy lookup table for SRT division and that it is still apparently in use in microprocessors today. I wish he could know that.