a code reduction function addition to cam_xpt
Brooks Davis
brooks at one-eyed-alien.net
Mon Nov 20 23:48:20 UTC 2006
On Mon, Nov 20, 2006 at 11:18:43PM +0100, Attilio Rao wrote:
> 2006/11/20, Brooks Davis <brooks at one-eyed-alien.net>:
> >On Mon, Nov 20, 2006 at 01:31:01AM +0100, Attilio Rao wrote:
> >> 2006/11/20, mjacob at freebsd.org <mjacob at freebsd.org>:
> >> >There are *far* too many:
> >> >
> >> > xpt_print_path(path);
> >> > printf("foo\n");
> >> >
> >> >constructs. How about we just join them?
> >> >
> >> >==== //depot/projects/newisp/cam/cam_xpt.c#12 -
> >> >/home/FreeBSD/p4/newisp/cam/cam_xpt.c ====
> >> >@@ -63,6 +63,7 @@
> >> > #include <cam/scsi/scsi_all.h>
> >> > #include <cam/scsi/scsi_message.h>
> >> > #include <cam/scsi/scsi_pass.h>
> >> >+#include <machine/stdarg.h> /* for xpt_print below */
> >> > #include "opt_cam.h"
> >> >
> >> > /* Datastructures internal to the xpt layer */
> >> >@@ -4160,6 +4161,16 @@
> >> > }
> >> > }
> >> >
> >> >+void
> >> >+xpt_print(struct cam_path *path, const char *fmt, ...)
> >> >+{
> >> >+ va_list ap;
> >> >+ xpt_print_path(path);
> >> >+ va_start(ap, fmt);
> >> >+ vprintf(fmt, ap);
> >> >+ va_end(ap);
> >> >+}
> >> >+
> >> >==== //depot/projects/newisp/cam/cam_xpt.h#1 -
> >> >/home/FreeBSD/p4/newisp/cam/cam_xpt.h ====
> >> >@@ -62,6 +62,7 @@
> >> > int xpt_path_comp(struct cam_path *path1,
> >> > struct cam_path *path2);
> >> > void xpt_print_path(struct cam_path *path);
> >> >+void xpt_print(struct cam_path *path, const char
> >*fmt,
> >> >...);
> >> > int xpt_path_string(struct cam_path *path, char
> >*str,
> >> > size_t str_len);
> >> > path_id_t xpt_path_path_id(struct cam_path *path);
> >>
> >> Would not be better a preprocessing stub?
> >>
> >> something like:
> >>
> >> #define XPT_PRINT(path, fmt, ...) do {
> >> \
> >> xpt_print_path(path);
> >> \
> >> printf(fmt, __VA_ARGS__);
> >> \
> >> } while (0)
> >
> >Why? What is gained? FWIW, when I added if_printif it reduced kernel
> >size by several KB. If there's a similar effect here we should take
> >advantage of it.
>
> It is simply faster (one function calling less), even if probabilly
> this could be mitigated with -fomit-frame-pointer (IMHO, this is not
> as over used as if_printf...).
There's not possibly way you could measure a meaningful difference in
performance except in a case where the system was unusable because it
spent all its time blocked waiting for the console to update.
-- Brooks
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-scsi/attachments/20061120/3997d519/attachment.pgp
More information about the freebsd-scsi
mailing list