Date: Mon, 25 Jun 2001 05:33:38 -0700 (PDT) From: Matt Dillon <dillon@earth.backplane.com> To: Bruce Evans <bde@zeta.org.au> Cc: Peter Wemm <peter@wemm.org>, Mikhail Teterin <mi@aldan.algebra.com>, jlemon@FreeBSD.org, cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: kernel size w/ optimized bzero() & patch set (was Re: Inline optimized bzero (was Re: cvs commit: src/sys/netinettcp_subr.c)) Message-ID: <200106251233.f5PCXc306427@earth.backplane.com> References: <Pine.BSF.4.21.0106252337370.7918-100000@besplex.bde.org>
next in thread | previous in thread | raw e-mail | index | archive | help
:#define bzero(p, n) ({ \
: if (__builtin_constant_p(n) && (n) <= X) \
: __builtin_memset((p), 0, (n)); \
: else \
: (bzero)((p), (n)); \
:})
:
:for X = 0, 4, 8, 12, 16, 32 and "infinity", the kernel sizes were:
:
: text data bss dec hex filename
:1962434 151436 349824 2463694 2597ce kernel.4
:1962442 151436 349824 2463702 2597d6 kernel.8
:1962446 151436 349824 2463706 2597da kernel.12
:1962466 151436 349824 2463726 2597ee kernel.0
:1962802 151436 349824 2464062 25993e kernel.16
:1962866 151436 349824 2464126 25997e kernel.20
:1963538 151436 349824 2464798 259c1e kernel.32
:1964098 151436 349824 2465358 259e4e kernel.infinity
:
:Summary: it's hard for the inline version to be smaller; even when it
:only needs to do one store-immediate operation, the kernel is only 32
:bytes smaller than the one using function calls which have to push
:2 args, do the call, and clean up. This is presumably due to increased
:register pressure for the inlined versions.
Yah. I adjusted my code so the 6*sizeof(int) and 7*sizeof(int) cases
did a loop rather then inline stores and the kernel size actually went up
by 40 bytes.
text data bss dec hex filename
1850705 159392 144536 2154633 20e089 kernel Normal bzero
1850833 159396 144536 2154765 20e10d kernel Most recent patch
1850873 159396 144536 2154805 20e135 kernel adjusted cases
Since the loop will normally be smaller then 7 stores, my guess is
that register and alignment pressures are to blame for the increased
size.
And my earlier assessment was wrong. 800+ bzero()'s, a difference of
128 bytes in the kernel, results in an average of 0.16 additional
instructions bytes per bzero, not 1 byte per bzero. I think this is
quite acceptable considering the positive effect.
In anycase, I think I'll stick to my most recent patch, it seems to
have the best overall effect with its ability to do up to 7 inline
stores (28 bytes) and loop-zero up to 64 bytes with only 0.16 additional
bytes per bzero on average. An inline can't get much better then that.
-Matt
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200106251233.f5PCXc306427>