Re: numbers

[Kai Li <Li-Kai@YALE.ARPA>]

    Both Symbolics and Apollo have  new  machines  in  the  works.   There  is  an
    instruction  pre-fetch  unit  which  may speed up 3600s 20% to 50%.  There are
    semi-custom VLSI things in the works also.  But even if the Tern can catch  up
    in  speed  with  the  3600,  the 3600 has a 27 bit cons cell address space, as
    compared to 19 bits in the 68010, and 22 AT BEST IFF they pin out the existant
    address  lines  in the 68020.  (am I wrong here?  Could the 68000 architecture
    support 32 bits of byte address, and thus 27 or 28 bits of cons cells???)

On an Apollo which uses either 68000 or 68010, a user process can have
24 bit address space which is 16Mbytes.  If anyone would like to
verify the information, please see

    Leach, Paul J., et al,
        "The Architecture of an Integrated Local Network,"
        Submitted to IEEE Journal on Selected Areas in Communications,
        Nov 1983.

If we really look for a large address space, then VAX VMS supports 32 bit
address space, though user program can actually use 30 bits.  The new
chip MicroVax-II benchmarks show that its speed is as fast as a 780,
depending on what you are doing.  In case anyone is interested and
has seen it, I forward the benchmark message.  "Ultrix-32m" is a fixed
version of Unix 4.2 maintained by DEC.

------- Forwarded Message

Date: Thu, 23 Aug 84 20:02:08 edt
To: baskett@decwrl, sysV, upo
Subject: MicroVAX-II and Ultrix-32m


        Ultrix-32m ran on the MicroVAX-II prototype in Hudson today at
4:30pm (8-23-84) on it's second boot attempt.

        People, it feels like a 780 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

        The following benchmark (obtained from Forest Baskett) was compiled
and run on THE machine. The SAME kernel and benchmark was also booted on one of
the UEG MicroVAX-I's in Merrimack later in the evening. The executable image
was then run (without recompiling) on a UEG VAX780 and VAX750.

        780             750             uVAX-I          uVAX-II
----------------------------------------------------------------------

perm    2480            4200            7460            2310

towers  2460            4420            8070            2390

queens  1010            1630            3010            1000

intmm   2250            4030            6650            3090

mm      2850            4810            8990            109420

puzzle  12140           22500           38270           12170

quick   1210            1990            3940            1290

bubble  1600            3150            6170            1770

tree    2920            4580            8560            2780

fft     4380            7830            16530           206170


The mm and fft are floating point intensive which explains the deviation.
What will it be like when we get the FPU chip in the board ??????????????


Ray Lanza

------- End of Forwarded Message
-------