Path: utzoo!utgpu!watmath!clyde!att!rutgers!tut.cis.ohio-state.edu!cwjcc!gatech!purdue!decwrl!ucbvax!hplabs!hpda!hpcuhb!hpesoc1!nicholso From: nicholso@hpesoc1.HP.COM (Ron Nicholson) Newsgroups: comp.arch Subject: Why no RISC clones? Message-ID: <5030006@hpesoc1.HP.COM> Date: 29 Nov 88 03:26:51 GMT Organization: Hewlett Packard, Cupertino Lines: 34 What I don't understand is this: how do RISC designers protect their investment. A large part of the investment seems to be in deciding what not to implement. This becomes public when the instruction set manual is published. There is a big business in 370 compatibles. Zilog got a big piece of the micro market by cloning the 8080 instruction set. All they had to change was the instruction set mnemonics to get around the Intel copyright. RISC processors have a lower design complexity than CISC processors in the same performance range, and they are architected to be independent of any particular implementation or technology. This should make them much easier to clone. So what's the deal. Is the instruction set a trade secret? Can some of the whiz-bang instructions only be done well by an IC layout which is copyrighted? Can an instruction be patented completely independently of how it is implemented? Is, perhaps, the collection of instructions patented? Maybe the architecture only works well with a compiler that is very difficult to implement. But does this make a difference in the shrink-wrap ABI type market? Maybe an ASCII dump of common instruction sequences spell "Copyright 19XX Acme-RISC Co." or the register set comes up with that message after a reset. If the instruction set is protected by patent or copyright, does that mean that a software company that writes an instruction set simulator is infringing? How about if that simulator is implemented in microcode or special hardware? ---- Ron Nicholson - Hewlett Packard - Cupertino, CA #include