OT -- INTC, SUNW, risc, vliw,...
Scott I agree with everything you said in your post. I think Merced will have a great future and yes this will be a threat to Microsoft (macroscum?). But in the long term, it will also be threat to Intel. This is because even though many of the original RISC companies will not be making theirown processors in the short term, the skill set and potential to make RISC processors is very much alive. Once all computers start to run on RISC engine (with a x86 translation unit on the front end), you will have other companies produce theirown RISC processors+x86 translation to compete with Intel. So Intel's monopoly willfade away. At least this has been the premiss in my mind for the last 5 years.
The other factor that can speed up (or delay) this process is the extent to which Merced and its follow ups come close to pure VLIW chips. Generating machine code for VLIW chips is a very different task than other processors. For a CISC chip, you do p-code optimization and then generate the machine code. For a RISC chip, you do p-code optimization, then use your knowledge of the cpu architecture to further optimize the machine code you generate (note you only need an overview of the cpu). For VLIW chips, you need intimate knowledge of the cpu before you can generate optimal code. This means that a code than runs on a 300 MHz chip will run poorly on the 400 MHz version of the same chip and may not even run at all. To overcome this problem, comercial programs for VLIW machines are distributed as an interim p-code and small compiler. The installation program will essentially generate the final executable code for your particular machine.
How will this affect Intel? Well, if it becomes standard practice to compile the code as part of the installation, then eventually the need for the x86 translator will fade away as people will ship code for the RISC backend in the first place. Andy Grove does have reasons to be paranoid, but not for another 5~10 years.
Sun Tzu |
