On Mon, 1 Dec 2008 11:20:22 -0600, John McKown wrote: > >On the current machines, there are two classes of instructions. The "simple" >instructions (like SR and LA and so forth) are "hard wired". The more >difficult instructions (such as MVCLE) are "millicoded".
I don't know about MVCLE, but on the z10 there are 668 instructions implemented entirely in hardware, according to the Charles Webb presentation. That would have to include a lot of "difficult" instructions. According to the same document, there are are a total of 894 instructions, leaving 226 implemented in millicode, or with millicode assists. There are about 80 new instructions in the -6 edition of the POO, published in February, 2008. >So, if you can >replace a millicoded instruction with a small number of "hard wired" >instructions, then you should be better off, speed wise. Presumably. >However, once >again, IBM does not document which instructions are "hard coded" and which >are millicode. I have heard that IBM has, at times, "fixed" a "broken" hard >coded instruction by replacing it with a millicoded instruction installed by >the CE. Don't know how often that happens, but it is one of the benefits of millicode. I found this article very interesting: http://researchweb.watson.ibm.com/journal/rd/483/heller.html -- Tom Marchant ---------------------------------------------------------------------- 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
