Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 8 Nov 2015 17:02:14 +0100
From:      Marius Strobl <marius@alchemy.franken.de>
To:        Poul-Henning Kamp <phk@phk.freebsd.dk>
Cc:        Warner Losh <imp@bsdimp.com>, sbruno@freebsd.org, freebsd-arch <freebsd-arch@freebsd.org>, sparc64@freebsd.org
Subject:   Re: Sparc64 doesn't care about you, and you shouldn't care about Sparc64
Message-ID:  <20151108160214.GA31931@alchemy.franken.de>
In-Reply-To: <93593.1446679609@critter.freebsd.dk>
References:  <563A5893.1030607@freebsd.org> <2AAC0EF3-528B-476F-BA9C-CDC3004465D0@bsdimp.com> <93593.1446679609@critter.freebsd.dk>

next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Nov 04, 2015 at 11:26:49PM +0000, Poul-Henning Kamp wrote:
> --------
> In message <2AAC0EF3-528B-476F-BA9C-CDC3004465D0@bsdimp.com>, Warner Losh write
> s:
> 
> >I concur.  I think sparc64 has had a nice run, but it's time to 
> >recognize that the run is nearing its end.
> 
> The main reason we wantd to have sparc64 in the fold was that it
> was the opposite sex than i386, and thus helped find endianess bugs.
> 
> The secondary reason was that it was 64 bit vs. i386's 32 bit.
> 

Maybe that was the original motivation. However, in my perception,
the main benefit of sparc64 for the entire tree over the last couple
of years was to be a real magnet for alignment bugs in MI code,
mainly network related things (with powerpc/powerpc64 being second
place in that regard). It did that job so well that I repeatedly
wondered myself: Who on earth actually is using arm and mips? I
mean, sure, Juniper has its own IP stack but f. e. at the time
alignment bugs in netgraph(4) got reported on sparc64, I really
would have expected at least some people to use a MIPS-based router
board for running a PPPoE session in order to terminate their DSL
lines.
That's even true as of today; f. e., there are still alignment bugs
left to be fixed in dummynet(4) (which apparently has been written
without platforms with strict alignment requirements in mind and,
thus, is somewhat a PITA to properly fix). These bugs get reported
for sparc64 from time to time but I've never seen a single one in
the context of arm, mips or powerpc/powerpc64).

Marius




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20151108160214.GA31931>