By "bypass cache" this means to use non-temporal stores to avoid polluting
the cache. On x86 for instance some SSE instructions are non-allocating in
the cache, and on SPARC we have block store ASIs and non-allocating (on
some implementations of v9) prefetch instructions.
Typically cache-bypass stores have the side effect of displacing any lines
from the cache which happen to be valid in CPU caches as the copy is
performed (forcing them into the invalid state). When moving a lot of data
this is preferred because the assumption is that the data won't be
accessed by this CPU for awhile, if at all, so just get it out to memory
and leave it there rather than leaving it in CPU caches.
- Eric
Matty wrote:
Howdy,
While reading through the source code for read():
http://cvs.opensolaris.org/source/xref/on/usr/src/uts/common/syscall/rw.c
I came across the following:
/*
168 * Only use bypass caches when the count is large enough
169 */
170 if (bcount < copyout_min_size)
171 auio.uio_extflg = UIO_COPY_CACHED;
172 else
173 auio.uio_extflg = UIO_COPY_DEFAULT;
[ ..... ]
Could someone shed some light on what a bypass cache, and where the
magic value 128k (value of copyout_min_size) comes from?
Thanks,
- Ryan
--
UNIX Administrator
http://daemons.net/~matty
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
--
Eric Lowe Solaris Kernel Development Austin, Texas
Sun Microsystems. We make the net work. x40577/+1(512)366-9080
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
