fls(N), ffs(N) and fls64(N) can be optimised on x86_64. Currently they use a
CMOV instruction after the BSR/BSF to set the destination register to -1 if the
value to be scanned was 0 (in which case BSR/BSF set the Z flag).
Instead, according to the AMD64 specification, we can make use of the fact that
BSR/BSF doesn't modify its output register if its input is 0. By preloading
the output with -1 and incrementing the result, we achieve the desired result
without the need for a conditional check.
The Intel x86_64 specification, however, says that the result of BSR/BSF in
such a case is undefined. That said, when queried, one of the Intel CPU
architects said that the behaviour on all Intel CPUs is that:
(1) with BSRQ/BSFQ, the 64-bit destination register is written with its
original value if the source is 0, thus, in essence, giving the effect we
want. And,
(2) with BSRL/BSFL, the lower half of the 64-bit destination register is
written with its original value if the source is 0, and the upper half is
cleared, thus giving us the effect we want (we return a 4-byte int).
Further, it was indicated that they (Intel) are unlikely to get away with
changing the behaviour.
It might be possible to optimise the 32-bit versions of these functions, but
there's a lot more variation, and so the effective non-destructive property of
BSRL/BSRF cannot be relied on.
[ hpa: specifically, some 486 chips are known to NOT have this property. ]
I have benchmarked these functions on my Core2 Duo test machine using the
following program:
#include <stdlib.h>
#include <stdio.h>
#ifndef __x86_64__
#error
#endif
#define PAGE_SHIFT 12
typedef unsigned long long __u64, u64;
typedef unsigned int __u32, u32;
#define noinline __attribute__((noinline))
static __always_inline int fls64(__u64 x)
{
long bitpos = -1;
asm("bsrq %1,%0"
: "+r" (bitpos)
: "rm" (x));
return bitpos + 1;
}
static inline unsigned long __fls(unsigned long word)
{
asm("bsr %1,%0"
: "=r" (word)
: "rm" (word));
return word;
}
static __always_inline int old_fls64(__u64 x)
{
if (x == 0)
return 0;
return __fls(x) + 1;
}
static noinline // __attribute__((const))
int old_get_order(unsigned long size)
{
int order;
size = (size - 1) >> (PAGE_SHIFT - 1);
order = -1;
do {
size >>= 1;
order++;
} while (size);
return order;
}
static inline __attribute__((const))
int get_order_old_fls64(unsigned long size)
{
int order;
size--;
size >>= PAGE_SHIFT;
order = old_fls64(size);
return order;
}
static inline __attribute__((const))
int get_order(unsigned long size)
{
int order;
size--;
size >>= PAGE_SHIFT;
order = fls64(size);
return order;
}
unsigned long prevent_optimise_out;
static noinline unsigned long test_old_get_order(void)
{
unsigned long n, total = 0;
long rep, loop;
for (rep = 1000000; rep > 0; rep--) {
for (loop = 0; loop <= 16384; loop += 4) {
n = 1UL << loop;
total += old_get_order(n);
}
}
return total;
}
static noinline unsigned long test_get_order_old_fls64(void)
{
unsigned long n, total = 0;
long rep, loop;
for (rep = 1000000; rep > 0; rep--) {
for (loop = 0; loop <= 16384; loop += 4) {
n = 1UL << loop;
total += get_order_old_fls64(n);
}
}
return total;
}
static noinline unsigned long test_get_order(void)
{
unsigned long n, total = 0;
long rep, loop;
for (rep = 1000000; rep > 0; rep--) {
for (loop = 0; loop <= 16384; loop += 4) {
n = 1UL << loop;
total += get_order(n);
}
}
return total;
}
int main(int argc, char **argv)
{
unsigned long total;
switch (argc) {
case 1: total = test_old_get_order(); break;
case 2: total = test_get_order_old_fls64(); break;
default: total = test_get_order(); break;
}
prevent_optimise_out = total;
return 0;
}
This allows me to test the use of the old fls64() implementation and the new
fls64() implementation and also to contrast these to the out-of-line loop-based
implementation of get_order(). The results were:
warthog>time ./get_order
real 1m37.191s
user 1m36.313s
sys 0m0.861s
warthog>time ./get_order x
real 0m16.892s
user 0m16.586s
sys 0m0.287s
warthog>time ./get_order x x
real 0m7.731s
user 0m7.727s
sys 0m0.002s
Using the current upstream fls64() as a basis for an inlined get_order() [the
second result above] is much faster than using the current out-of-line
loop-based get_order() [the first result above].
Using my optimised inline fls64()-based get_order() [the third result above]
is even faster still.
[ hpa: changed the selection of 32 vs 64 bits to use CONFIG_X86_64
instead of comparing BITS_PER_LONG, updated comments, rebased manually
on top of 83d99df7c4 x86, bitops: Move fls64.h inside __KERNEL__ ]
Signed-off-by: David Howells <dhowells@redhat.com>
Link: http://lkml.kernel.org/r/20111213145654.14362.39868.stgit@warthog.procyon.org.uk
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
We would include <asm-generic/bitops/fls64.h> even without __KERNEL__,
but that doesn't make sense, as:
1. That file provides fls64(), but the corresponding function fls() is
not exported to user space.
2. The implementation of fls64.h uses kernel-only symbols.
3. fls64.h is not exported to user space.
This appears to have been a bug introduced in checkin:
d57594c203 bitops: use __fls for fls64 on 64-bit archs
Cc: Stephen Hemminger <shemminger@vyatta.com>
Cc: Alexander van Heukelum <heukelum@mailshack.com>
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Link: http://lkml.kernel.org/r/4EEA77E1.6050009@zytor.com
The majority of architectures implement ext2 atomic bitops as
test_and_{set,clear}_bit() without spinlock.
This adds this type of generic implementation in ext2-atomic-setbit.h and
use it wherever possible.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Suggested-by: Andreas Dilger <adilger@dilger.ca>
Suggested-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
minix bit operations are only used by minix filesystem and useless by
other modules. Because byte order of inode and block bitmaps is different
on each architecture like below:
m68k:
big-endian 16bit indexed bitmaps
h8300, microblaze, s390, sparc, m68knommu:
big-endian 32 or 64bit indexed bitmaps
m32r, mips, sh, xtensa:
big-endian 32 or 64bit indexed bitmaps for big-endian mode
little-endian bitmaps for little-endian mode
Others:
little-endian bitmaps
In order to move minix bit operations from asm/bitops.h to architecture
independent code in minix filesystem, this provides two config options.
CONFIG_MINIX_FS_BIG_ENDIAN_16BIT_INDEXED is only selected by m68k.
CONFIG_MINIX_FS_NATIVE_ENDIAN is selected by the architectures which use
native byte order bitmaps (h8300, microblaze, s390, sparc, m68knommu,
m32r, mips, sh, xtensa). The architectures which always use little-endian
bitmaps do not select these options.
Finally, we can remove minix bit operations from asm/bitops.h for all
architectures.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Greg Ungerer <gerg@uclinux.org>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Andreas Schwab <schwab@linux-m68k.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Michal Simek <monstr@monstr.eu>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Hirokazu Takata <takata@linux-m32r.org>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Acked-by: Paul Mundt <lethal@linux-sh.org>
Cc: Chris Zankel <chris@zankel.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
As the result of conversions, there are no users of ext2 non-atomic bit
operations except for ext2 filesystem itself. Now we can put them into
architecture independent code in ext2 filesystem, and remove from
asm/bitops.h for all architectures.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Introduce little-endian bit operations to the big-endian architectures
which do not have native little-endian bit operations and the
little-endian architectures. (alpha, avr32, blackfin, cris, frv, h8300,
ia64, m32r, mips, mn10300, parisc, sh, sparc, tile, x86, xtensa)
These architectures can just include generic implementation
(asm-generic/bitops/le.h).
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Mikael Starvik <starvik@axis.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Matthew Wilcox <willy@debian.org>
Cc: Grant Grundler <grundler@parisc-linux.org>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Kazumoto Kojima <kkojima@rr.iij4u.or.jp>
Cc: Hirokazu Takata <takata@linux-m32r.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Chris Zankel <chris@zankel.net>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Hans-Christian Egtvedt <hans-christian.egtvedt@atmel.com>
Acked-by: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
asm-generic/bitops/find.h has the extern declarations of find_next_bit()
and find_next_zero_bit() and the macro definitions of find_first_bit()
and find_first_zero_bit(). It is only usable by the architectures which
enables CONFIG_GENERIC_FIND_NEXT_BIT and disables
CONFIG_GENERIC_FIND_FIRST_BIT.
x86 and tile enable both CONFIG_GENERIC_FIND_NEXT_BIT and
CONFIG_GENERIC_FIND_FIRST_BIT. These architectures cannot include
asm-generic/bitops/find.h in their asm/bitops.h. So ifdefed extern
declarations of find_first_bit and find_first_zero_bit() are put in
linux/bitops.h.
This makes asm-generic/bitops/find.h usable by these architectures
and use it. Also this change is needed for the forthcoming duplicated
extern declarations cleanup.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: x86@kernel.org
Cc: Chris Metcalf <cmetcalf@tilera.com>
While debugging bit_spin_lock() hang, it was tracked down to gcc-4.4
misoptimization of non-inlined constant_test_bit() due to non-volatile
addr when 'const volatile unsigned long *addr' cast to 'unsigned long *'
with subsequent unconditional jump to pause (and not to the test) leading
to hang.
Compiling with gcc-4.3 or disabling CONFIG_OPTIMIZE_INLINING yields inlined
constant_test_bit() and correct jump, thus working around the kernel bug.
Other arches than asm-x86 may implement this slightly differently;
2.6.29 mitigates the misoptimization by changing the function prototype
(commit c4295fbb60) but probably fixing the issue
itself is better.
Signed-off-by: Alexander Chumachenko <ledest@gmail.com>
Signed-off-by: Michael Shigorin <mike@osdn.org.ua>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Add support for the hardware version of the Hamming weight function,
popcnt, present in CPUs which advertize it under CPUID, Function
0x0000_0001_ECX[23]. On CPUs which don't support it, we fallback to the
default lib/hweight.c sw versions.
A synthetic benchmark comparing popcnt with __sw_hweight64 showed almost
a 3x speedup on a F10h machine.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
LKML-Reference: <20100318112015.GC11152@aftab>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Impact: reduce kernel image size
Hugh Dickins noticed that older gcc versions when the kernel
is built for code size didn't inline some of the bitops.
Mark all complex x86 bitops that have more than a single
asm statement or two as always inline to avoid this problem.
Probably should be done for other architectures too.
Ingo then found a better fix that only requires
a single line change, but it unfortunately only
works on gcc 4.3.
On older gccs the original patch still makes a ~0.3% defconfig
difference with CONFIG_OPTIMIZE_INLINING=y.
With gcc 4.1 and a defconfig like build:
6116998 1138540 883788 8139326 7c323e vmlinux-oi-with-patch
6137043 1138540 883788 8159371 7c808b vmlinux-optimize-inlining
~20k / 0.3% difference.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Ingo noticed that using signed arithmetic seems to confuse the gcc
inliner, and make it potentially decide that it's all too complicated.
(Yeah, yeah, it's a constant. It's always positive. Still..)
Based-on: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Impact: Minor optimization.
Implement change_bit with immediate bit count as "lock xorb". This is
similar to "lock orb" and "lock andb" for set_bit and clear_bit
functions.
Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Change header guards named "ASM_X86__*" to "_ASM_X86_*" since:
a. the double underscore is ugly and pointless.
b. no leading underscore violates namespace constraints.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
