Re: Softfloat on i386

[David Holland <dholland-pkgtech%netbsd.org@localhost>]

On Wed, Dec 31, 2025 at 08:12:41PM -0500, Greg Troxel wrote:
 > > I'd describe both of these as retrocomputing. There's no reason to
 > > junk a running 20-year-old x86 if you have a mission for it that fits
 > > in its RAM. But there aren't many such missions, or so many
 > > still-working machines, and likely the power bill justifies replacing
 > > them anyway. :-|
 > 
 > More or less, but I sort "retrocomputing" if people are playing with
 > things because it is cool that they are old, and not retrocomputing if
 > they are using something because it can do what they want to do, even if
 > it is old.   I am running 2x RPI3 right now, and they are old and slow
 > and really not sensible, but it's easier to keep them running than replace.

Fair enough. But it also doesn't really matter, because
'retrocomputing' is just a sticker we put on certain actions to help
assess how we feel about them.

 > > So I guess the question is: which parts of this audience are we really
 > > intending to cater to, and how much of it is pre-sse2 hardware? Or
 > > should we be running multiple sets of (maybe smaller) builds?
 > >
 > > (Obviously support or not support of something just in golang is not
 > > itself a reason to rearrange the set of builds we have or the
 > > subcategories of x86 they're intended for. But looking at that
 > > categorization periodically to see if it's still sensible is probably
 > > worthwhile.)
 > 
 > Maybe a poll is in order.  

Yeah. I don't know what it "should" be or have any particular opinion,
other than a vague sense that in the absence of other criteria, maybe
we should put our x86 build resources towards more amd64 builds and
fewer i386 builds than we have been.

 > Your email more or less convinces me that
 > defaulting to building for CPUs w/o sse2 should be default.
 > 
 > I think my leaning to "assume sse2 as normal and people can rebuild"
 > comes from so many things that just don't work at all w/o sse2,
 > e.g. those that need 64-bit atomic operations, it seems.

Conversely, I'm kinda thinking maybe it makes sense to partition the
i386 target on this and do two sets of builds.

Do we know what fraction of packages (or of packages that will build
and run at all on i386, etc.) are affected by this distinction? I have
the vague sense that it's more and more.

 > Overall, I suspect almost everyone using i386 can rebuild a few
 > packages.  So I say let's not worry that much until actual people raise
 > actual complaints.

Right, which brings us back to the original topic :-)

-- 
David A. Holland
dholland%netbsd.org@localhost