Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753564AbYLIEDU (ORCPT ); Mon, 8 Dec 2008 23:03:20 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751527AbYLIEDN (ORCPT ); Mon, 8 Dec 2008 23:03:13 -0500 Received: from ozlabs.org ([203.10.76.45]:45356 "EHLO ozlabs.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751511AbYLIEDM (ORCPT ); Mon, 8 Dec 2008 23:03:12 -0500 From: Rusty Russell To: Eric Dumazet Subject: Re: [PACH] smp: uninline num_online_cpus() & num_possible_cpus() Date: Tue, 9 Dec 2008 14:33:03 +1030 User-Agent: KMail/1.10.1 (Linux/2.6.27-9-generic; KDE/4.1.2; i686; ; ) Cc: Andrew Morton , linux kernel References: <493965F8.6060407@cosmosbay.com> In-Reply-To: <493965F8.6060407@cosmosbay.com> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200812091433.03670.rusty@rustcorp.com.au> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 6918 Lines: 193 On Saturday 06 December 2008 04:03:44 Eric Dumazet wrote: > num_online_cpus() and num_possible_cpus() are not performance > critical and are quite large. > > Unlining them shrinks kernel text size by 7523 bytes on x86, > if NR_CPUS>32 Hi Eric! Slight misdiagnosis, I think. One base problem is addressed in fixing the bitmap operators (see "[PATCH] bitmap: test for constant as well as small size for inline versions" on lkml Message-Id: <200811160907.07140.rusty@rustcorp.com.au>). This is already in linux-next, and I've pasted it below. Worse, you used the obsolete cpumask operators :) Thanks! Rusty. bitmap: test for constant as well as small size for inline versions bitmap_zero et al have a fastpath for nbits <= BITS_PER_LONG, but this should really only apply where the nbits is known at compile time. This only saves about 1200 bytes on an allyesconfig kernel, but with cpumasks going variable that number will increase. text data bss dec hex filename 35327852 5035607 6782976 47146435 2cf65c3 vmlinux-before 35326640 5035607 6782976 47145223 2cf6107 vmlinux-after Signed-off-by: Rusty Russell --- include/linux/bitmap.h | 35 +++++++++++++++++++---------------- 1 file changed, 19 insertions(+), 16 deletions(-) diff -r b4540ad329c1 include/linux/bitmap.h --- a/include/linux/bitmap.h Thu Nov 06 13:00:51 2008 +1100 +++ b/include/linux/bitmap.h Thu Nov 06 14:34:07 2008 +1100 @@ -137,9 +137,12 @@ extern void bitmap_copy_le(void *dst, co (1UL<<((nbits) % BITS_PER_LONG))-1 : ~0UL \ ) +#define small_const_nbits(nbits) \ + (__builtin_constant_p(nbits) && (nbits) <= BITS_PER_LONG) + static inline void bitmap_zero(unsigned long *dst, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) *dst = 0UL; else { int len = BITS_TO_LONGS(nbits) * sizeof(unsigned long); @@ -150,7 +153,7 @@ static inline void bitmap_fill(unsigned static inline void bitmap_fill(unsigned long *dst, int nbits) { size_t nlongs = BITS_TO_LONGS(nbits); - if (nlongs > 1) { + if (!small_const_nbits(nbits)) { int len = (nlongs - 1) * sizeof(unsigned long); memset(dst, 0xff, len); } @@ -160,7 +163,7 @@ static inline void bitmap_copy(unsigned static inline void bitmap_copy(unsigned long *dst, const unsigned long *src, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) *dst = *src; else { int len = BITS_TO_LONGS(nbits) * sizeof(unsigned long); @@ -171,7 +174,7 @@ static inline void bitmap_and(unsigned l static inline void bitmap_and(unsigned long *dst, const unsigned long *src1, const unsigned long *src2, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) *dst = *src1 & *src2; else __bitmap_and(dst, src1, src2, nbits); @@ -180,7 +183,7 @@ static inline void bitmap_or(unsigned lo static inline void bitmap_or(unsigned long *dst, const unsigned long *src1, const unsigned long *src2, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) *dst = *src1 | *src2; else __bitmap_or(dst, src1, src2, nbits); @@ -189,7 +192,7 @@ static inline void bitmap_xor(unsigned l static inline void bitmap_xor(unsigned long *dst, const unsigned long *src1, const unsigned long *src2, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) *dst = *src1 ^ *src2; else __bitmap_xor(dst, src1, src2, nbits); @@ -198,7 +201,7 @@ static inline void bitmap_andnot(unsigne static inline void bitmap_andnot(unsigned long *dst, const unsigned long *src1, const unsigned long *src2, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) *dst = *src1 & ~(*src2); else __bitmap_andnot(dst, src1, src2, nbits); @@ -207,7 +210,7 @@ static inline void bitmap_complement(uns static inline void bitmap_complement(unsigned long *dst, const unsigned long *src, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) *dst = ~(*src) & BITMAP_LAST_WORD_MASK(nbits); else __bitmap_complement(dst, src, nbits); @@ -216,7 +219,7 @@ static inline int bitmap_equal(const uns static inline int bitmap_equal(const unsigned long *src1, const unsigned long *src2, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) return ! ((*src1 ^ *src2) & BITMAP_LAST_WORD_MASK(nbits)); else return __bitmap_equal(src1, src2, nbits); @@ -225,7 +228,7 @@ static inline int bitmap_intersects(cons static inline int bitmap_intersects(const unsigned long *src1, const unsigned long *src2, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) return ((*src1 & *src2) & BITMAP_LAST_WORD_MASK(nbits)) != 0; else return __bitmap_intersects(src1, src2, nbits); @@ -234,7 +237,7 @@ static inline int bitmap_subset(const un static inline int bitmap_subset(const unsigned long *src1, const unsigned long *src2, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) return ! ((*src1 & ~(*src2)) & BITMAP_LAST_WORD_MASK(nbits)); else return __bitmap_subset(src1, src2, nbits); @@ -242,7 +245,7 @@ static inline int bitmap_subset(const un static inline int bitmap_empty(const unsigned long *src, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) return ! (*src & BITMAP_LAST_WORD_MASK(nbits)); else return __bitmap_empty(src, nbits); @@ -250,7 +253,7 @@ static inline int bitmap_empty(const uns static inline int bitmap_full(const unsigned long *src, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) return ! (~(*src) & BITMAP_LAST_WORD_MASK(nbits)); else return __bitmap_full(src, nbits); @@ -258,7 +261,7 @@ static inline int bitmap_full(const unsi static inline int bitmap_weight(const unsigned long *src, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) return hweight_long(*src & BITMAP_LAST_WORD_MASK(nbits)); return __bitmap_weight(src, nbits); } @@ -266,7 +269,7 @@ static inline void bitmap_shift_right(un static inline void bitmap_shift_right(unsigned long *dst, const unsigned long *src, int n, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) *dst = *src >> n; else __bitmap_shift_right(dst, src, n, nbits); @@ -275,7 +278,7 @@ static inline void bitmap_shift_left(uns static inline void bitmap_shift_left(unsigned long *dst, const unsigned long *src, int n, int nbits) { - if (nbits <= BITS_PER_LONG) + if (small_const_nbits(nbits)) *dst = (*src << n) & BITMAP_LAST_WORD_MASK(nbits); else __bitmap_shift_left(dst, src, n, nbits); -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/