On Wed, 19 Jul 2006 10:33:07 -0700, SUBSCRIBE IBM-MAIN tdell <T'[EMAIL PROTECTED]> wrote:
>I could bet the farm that's ECLipz , is code name for hardware emulation at >the micro/milli code level. The new architecture could be framed like >the CELL expecting one or more 390 CPU's married to a network of CELL >processors , which in turn is supported by several other RISC processors >to handle the various workloads and emulation. The major task of >instruction decoding, and memory fetching can be done by another processor >on behalf of the 390 CPU. That is a BIG portion of the heavy lifting >required to execute a program. That's a lot of machine cycles saved, that's No. Instruction decode and memory access (not just fetch) are already performed in parallel. >why RISC CPU's can boast they don't have those performance penalties to >contend with, since it's somewhat separated out of the normal decoding >behavior. Risc processors run at higher clock rates primapily because they are smaller. That means that they have shorter path lengths and less propogation delay. That doesn't mean that they get more work done in a given amount of time. The instructions are much simpler and there are a lot fewer of them. >Some 12 years ago,I initially believed the Linux having such a small kernel >in it's early life had a distinct advantage of other OS's, but I was >incorrect (and somewhat naive and misinformed) on how it would have >performed on RISC versus the INTEL (386/486) implementations. The numbers There is nothing about RISC vs. CISC that makes one or the other an inherently better performer because of the size of the kernel. >do bear that out but I don't have much of an interest in that area any more >since I back working exclusively with mainframes. I'm still a UNIX hobbyist >and run both AIX and Linux at home, utterly frustrating and fun at the same >time. If anyone out there know(Z) please tell me I'm still curious about >where this is all going. Be A RISC based line of 390 CPU's or POWER like >CPU with a 390 emulation. > slow. 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
