Date: Wed, 26 Jan 2011 17:22:58 +0530 From: "Jayachandran C." <c.jayachandran@gmail.com> To: Warner Losh <imp@bsdimp.com> Cc: freebsd-mips@freebsd.org Subject: Re: [PATCH] update sf_buf and uio for n64 Message-ID: <AANLkTimUXR6GZMMWbFw%2BC8X0ECTj=yRdMkEs3g4k0=bB@mail.gmail.com> In-Reply-To: <4D3F0B19.4060907@bsdimp.com> References: <AANLkTik%2BpMOxCaD70oZJPsmKa4mNCviZmzLhxe8wQR-y@mail.gmail.com> <4D3F0B19.4060907@bsdimp.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Jan 25, 2011 at 11:10 PM, Warner Losh <imp@bsdimp.com> wrote:
> On 01/25/2011 06:48, Jayachandran C. wrote:
>>
>> This is one of the remaining pieces in n64 work from Juli's octeon
>> branch. The attached patch updates the sf_buf code and uio_machdep.c
>> in n64 compilation to use direct mapping.
>>
>> Planning to check this in later this week, if there are no objections.
>>
>> JC.
>
> I really like the uio_machdep.c. =A0I think it is good and can go in with=
out
> further objection.
>
> I like the idea of the n64 direct mapping of the sbufs. =A0That's a good
> optimization. =A0I get nervous when I see some code implemented in one pl=
ace
> on one ifdef branch, and in a completely different file for another. =A0I=
s the
> gain in efficiency sufficient to justify this odd split? =A0If not, then
> having all the code in vm_machdep.c might make more sense. =A0If the gain=
is
> enough, then having a comment in vm_machdep.c pointing to sf_buf.h would =
be
> good to have as well... (that's assuming the simplifications suggested by
> Andrew Duane don't change the code distribution).
Not sure what you mean here, the implementation of sf_buf_kva() and
sf_buf_page() are in sf_buf.h, and the sf_buf_{init,alloc,free} are in
vm_machdep.c as before.
JC.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?AANLkTimUXR6GZMMWbFw%2BC8X0ECTj=yRdMkEs3g4k0=bB>