Messages in this thread | | | From | (Alan Cox) | Subject | Re: Linux on Merced | Date | Fri, 31 Oct 1997 16:45:22 +0000 (GMT) |
| |
> I am sorry, Alan, but this is not exactly true. Under an NDA agreement, > you cannot disclose any work based on previously proprietary > information. _After_ Intel has released the specs, and _if_ a GCC port
Welcome to my kill file. Read what I said
> reverse engineering. You have to start from _explicitly_ publicly > available information and work in absolute clean-room conditions. Intel
Please go and read a book on trade secret law while you are at it
> usually helps) into killing the Alpha architecture and selling Intel > their semiconductor operations (for below cost) to get out of this legal > tangle. Do you need more proof?
Bollocks - you dont even understand all the issues involved
> Dell and Compaq are releasing NT machines based on SMP _and_ clustering > to reach into the kind of parallel performance available from Linux.
Not really. You cant buy the software for the cost of the Linux setup including hardware.
> And Intel has just bought a company that had developed special hardware > for 8-CPU SMP boxes. You are certainly aware of that as you are the > major developer for Linux SMP, apart from Linux himself.
Lots of folks have been doing this - but SMP doesnt scale - thats why SGI have their Onyx stuff
> Alpha Linux port, and now this architecture is dead. The other > architectures have a restricted user base and incomplete > implementations. Perhaps the Linux Community should start thinking in > strategic terms about CPU choice?
Merced is irrelevant to any issues until its publically available, unless you want to go and do a lot of deep digging into HP's patents and other clues on the thing, steal a chip and write the code.
You can certainly plan ahead by reading up on VLIW and writing a gcc scheduler for a theoretical VLIW box and testing it.. Why not - go ahead do something you think will be useful.
Alan
| |