> Just wondering if  someone could guestimate the porting effort required to
> support the Toshiba TX49-series 64-bit core. We may be using a TX49 cpu in
> our next project, but are looking at alternatives to the Linux offerings.

TX49 core looks rather ''well natured''; it has standard 5 stage pipe,
48 entry R4000 TLB and fully featured FP unit, plus some features from
R5000 implementation.  The largest technical difference is, TX49 has
4way cache for I and D.  This makes a bit harder to port ''cache logic
ignorant'' softwares on the processor. However, TX39 series has 2way
caches (mostly), and NetBSD can handle QED derivative processors too,
so cache issue should not be fatal.

> Also, does anyone have any idea of how well regular NetBSD runs in a
> realtime application, like interrupt and task switch latencies?

Having no real measurement done, the kernel behaves not-so-smartly in
terms of quick responseness and lightweightness.  Indeed, newer NetBSD
releases run more sluggish than used to be.

Tohru Nishimura