Date: Mon, 20 Nov 2006 17:48:09 -0600 From: Brooks Davis <brooks@one-eyed-alien.net> To: Attilio Rao <attilio@freebsd.org> Cc: freebsd-scsi@freebsd.org, mjacob@freebsd.org Subject: Re: a code reduction function addition to cam_xpt Message-ID: <20061120234809.GB5155@lor.one-eyed-alien.net> In-Reply-To: <3bbf2fe10611201418m15d50703m37d9d5620e5c832d@mail.gmail.com> References: <20061119161631.L44297@ns1.feral.com> <3bbf2fe10611191631h6883b862uf8088533913a7bc6@mail.gmail.com> <20061120221153.GA5155@lor.one-eyed-alien.net> <3bbf2fe10611201418m15d50703m37d9d5620e5c832d@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
[-- Attachment #1 --]
On Mon, Nov 20, 2006 at 11:18:43PM +0100, Attilio Rao wrote:
> 2006/11/20, Brooks Davis <brooks@one-eyed-alien.net>:
> >On Mon, Nov 20, 2006 at 01:31:01AM +0100, Attilio Rao wrote:
> >> 2006/11/20, mjacob@freebsd.org <mjacob@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
[-- Attachment #2 --]
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (FreeBSD)
iD8DBQFFYj65XY6L6fI4GtQRAn+lAKDkgNWSQwKZaR8la5NCOG/Z/L1HygCbB7FL
ySV96+J2Gr3AyAJ8QjFDlXg=
=BkUa
-----END PGP SIGNATURE-----
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20061120234809.GB5155>