Port-newsmips archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Booting! (Was: not boot for you! (Was: Re: HD size & NWS-3710))



Q about how fast R3000@25MHz really is;

>> Now, I do have a question:  it seems the CPU is the same as the one 
>> used in a DEC5000, like my 5000/25.  How fast is the CPU?
>
> It's R3000/25MHz and IMHO it is not so fast.
> I haven't tried dhrystone (or other benchmarks),
> but it looks as same as SPARCstation1+ (L64801/25MHz),
> and slower than HP 9000/382 (M68040/25MHz).

R3000@25MHz is beaten by 486DX2-66, and rugs behind
any 68040.   More worse, the NetBSD implementation of virtual
memory subsystem for MIPS processors is _so poor_ that it
spoils much of the processor's potential.  The deficiency can be
easily observed at /etc/rc.d/ processing.  In technical summaries,
1. TLB handling is very poor especially on solving COW condition.
    The impact; slow command invocation
2. nonsenseness on address space tearing down.
    The impact; MIPS processor is designed to have near zero CPU
    CPU cycle waste when process address space is going to be
    trashed away.
3. Absolute mistake in the way of ZFOD page creation.
    The impact; ZFOD operation does _produce_ virtual aliases condition
    for virtually indexed/physically tagged cache.  It hurts most of R4000
    descendents.
Redesigning pmap.c for MIPS processor will improve fast responsiveness.

Toru Nishimura/ALKYL Technology




Home | Main Index | Thread Index | Old Index