color bit in the least significant bit of the right child pointer, in
order to reduce red-black tree linkage overhead by ~2X as compared to
sys/tree.h.
Use the new red-black tree implementation in malloc, which drops
memory usage by ~0.5 or ~1%, for 32- and 64-bit systems, respectively.
after similar calls related to struct pwd in libutil/pw_util.c:
- gr_equal()
Perform a deep comparison of two struct grp's. It does a thorough, yet
unoptimized comparison of all the members regardless of order.
- gr_make()
Create a string (see group(5)) from a struct grp.
- gr_dup()
Duplicate a struct grp. Returns a value that is a single contiguous
block of memory.
- gr_scan()
Create a struct grp from a string (as produced by gr_make()).
MFC after: 3 weeks
There were no checks for left and right precisions at all, and
a check for field width had integer overflow bug.
Reported by: Maksymilian Arciemowicz
Security: http://securityreason.com/achievement_securityalert/53
Submitted by: Maxim Dounin <mdounin@mdounin.ru>
MFC after: 3 days
__sFILE. This was supposed to be done in 6.0. Some notes:
- Where possible I restored the various lines to their pre-__sFILEX state.
- Retire INITEXTRA() and just initialize the wchar bits (orientation and
mbstate) explicitly instead. The various places that used INITEXTRA
didn't need the locking fields or _up initialized. (Some places needed
_up to exist and not be off the end of a NULL or garbage pointer, but
they didn't require it to be initialized to a specific value.)
- For now, stdio.h "knows" that pthread_t is a 'struct pthread *' to
avoid namespace pollution of including all the pthread types in stdio.h.
Once we remove all the inlines and make __sFILE private it can go back
to using pthread_t, etc.
- This does not remove any of the inlines currently and does not change
any of the public ABI of 'FILE'.
MFC after: 1 month
Reviewed by: peter
deals with the usual __opendir2() calls, and the rest part with an interface
translator to expose fdopendir(3) functionality. Manual page was obtained from
kib@'s work for *at(2) system calls.
counted in the width specification in scanf.
This is not a security problem, since this function is only used to
parse a user's configuration file.
Submitted by: Joerg Sonnenberger
Obtained from: dragonflybsd
MFC after: 1 week
1. Previously, printing the number 1.0 could produce 0x1p+0, 0x2p-1,
0x4p-2, or 0x8p-3, depending on what happened to be convenient. This
meant that printing a value as a double and printing the same value
as a long double could produce different (but equivalent) results.
The change is to always make the leading digit a 1, unless the
number is 0. This solves the aforementioned problem and has
several other advantages.
2. Use the FPU to do rounding. This is far simpler and more portable
than manipulating the bits, and it fixes an obsure round-to-even
bug. It also raises the exceptions now required by IEEE 754R.
The drawbacks are that it is usually slightly slower, and it makes
printf less effective as a debugging tool when the FPU is hosed
(e.g., due to a buggy softfloat implementation).
3. On i386, twiddle the rounding precision so that (2) works properly
for long doubles.
4. Make several simplifications that are now possible due to (2).
5. Split __hldtoa() into a separate file.
Thanks to remko for access to a sparc64 box for testing.
flags appropriately. The next step is to make it raise a SIGFPE if
any exceptions are unmasked.
Thanks to remko for access to a sparc64 box for testing.
- fma(x, y, z) returns z, not NaN, if z is infinite, x and y are finite,
x*y overflows, and x*y and z have opposite signs.
- fma(x, y, z) doesn't generate an overflow, underflow, or inexact exception
if z is NaN or infinite, as per IEEE 754R.
- If the rounding mode is set to FE_DOWNWARD, fma(1.0, 0.0, -0.0) is -0.0,
not +0.0.
returns errno, because errno can be mucked by user's signal handler and
most of pthread api heavily depends on errno to be correct, this change
should improve stability of the thread library.
This makes little difference in float precision, but in double
precision gives a speedup of about 30% on amd64 (A64 CPU) and i386
(A64). This depends on fabs[f]() being inline and efficient. The
bit fiddling (or any use of SET_HIGH_WORD(), which libm does too
much because it was best on old 32-bit machines) always causes
packing overheads and sometimes causes stalls in the packing, since
it operates on only part of a variable in the double precision case.
It apparently did cause stalls in a critical path here.
fabs(x+0.0)+fabs(y+0.0) when mixing NaNs. This improves
consistency of the result by making it harder for the compiler to reorder
the operands. (FP addition is not necessarily commutative because the
order of operands makes a difference on some machines iff the operands are
both NaNs.)
__xdrrec_getrec has returned TRUE, then we have a complete request in
the buffer - calling xdrrec_skiprecord is not necessary. In particular,
if there is another record already buffered on the stream,
xdrrec_skiprecord will discard both this request and the next
one, causing the call to xdr_callmsg to fail and the stream to be
closed.
Sponsored by: Isilon Systems
live in libm, while modf() lives in libc due to historical
mistakes. I'm claiming in the manpage that they all live in libm,
since programmers should not rely on the mistake.
user-mode lock manager, build a kernel with the NFSLOCKD option and
add '-k' to 'rpc_lockd_flags' in rc.conf.
Highlights include:
* Thread-safe kernel RPC client - many threads can use the same RPC
client handle safely with replies being de-multiplexed at the socket
upcall (typically driven directly by the NIC interrupt) and handed
off to whichever thread matches the reply. For UDP sockets, many RPC
clients can share the same socket. This allows the use of a single
privileged UDP port number to talk to an arbitrary number of remote
hosts.
* Single-threaded kernel RPC server. Adding support for multi-threaded
server would be relatively straightforward and would follow
approximately the Solaris KPI. A single thread should be sufficient
for the NLM since it should rarely block in normal operation.
* Kernel mode NLM server supporting cancel requests and granted
callbacks. I've tested the NLM server reasonably extensively - it
passes both my own tests and the NFS Connectathon locking tests
running on Solaris, Mac OS X and Ubuntu Linux.
* Userland NLM client supported. While the NLM server doesn't have
support for the local NFS client's locking needs, it does have to
field async replies and granted callbacks from remote NLMs that the
local client has contacted. We relay these replies to the userland
rpc.lockd over a local domain RPC socket.
* Robust deadlock detection for the local lock manager. In particular
it will detect deadlocks caused by a lock request that covers more
than one blocking request. As required by the NLM protocol, all
deadlock detection happens synchronously - a user is guaranteed that
if a lock request isn't rejected immediately, the lock will
eventually be granted. The old system allowed for a 'deferred
deadlock' condition where a blocked lock request could wake up and
find that some other deadlock-causing lock owner had beaten them to
the lock.
* Since both local and remote locks are managed by the same kernel
locking code, local and remote processes can safely use file locks
for mutual exclusion. Local processes have no fairness advantage
compared to remote processes when contending to lock a region that
has just been unlocked - the local lock manager enforces a strict
first-come first-served model for both local and remote lockers.
Sponsored by: Isilon Systems
PR: 95247 107555 115524 116679
MFC after: 2 weeks
section header entry if the application is not taking charge of ELF
object layout.
Update (c) years, and bump the manual page's date.
Submitted by: kaiw
(NAP, GN and PANU). No reason to not to support them.
Separate SDP parameters data structures for the BNEP based profiles.
Generalize Service Availability SDP parameter creation.
Requested by: Iain Hibbert < plunky at rya-online dot net >
MFC after: 3 days
_thr_suspend_check() which messes sigmask saved in thread structure.
- Don't suspend a thread has force_exit set.
- In pthread_exit(), if there is a suspension flag set, wake up waiting-
thread after setting PS_DEAD, this causes waiting-thread to break loop
in suspend_common().
of the array length needed to store all the directory entries.
Although BSD has historically guaranteed that st_size is the size
of the directory file, POSIX does not, and more to the point, some
recent filesystems such as ZFS use st_size to mean something else.
The fix is to not stat the directory at all, set the initial
array size to 32 entries, and realloc it in powers of 2 if that
proves insufficient.
PR: 113668
from the private archive_write structure and fix up all writers to use
the format fields in the base "archive" structure. This error made it
impossible to query the format after setting up a writer because the
write format was stored in an inaccessible place.
"file" is described by multiple "lines" each possibly containing
multiple "keywords." Incorporate some additions from Joerg Sonnenberger
to handle linked files and correctly deal with backing files on disk.
Disable the use of PaxHeader.<pid> for the fake pax extension pathname
until I can make the name here settable. Otherwise, tests that try
to compare output to static pre-generated reference files break.
(including pathname, gname, uname) be stored in UTF-8. This usually
doesn't cause problems on FreeBSD because the "C" locale on FreeBSD
can convert any byte to Unicode/wchar_t and from there to UTF-8. In
other locales (including the "C" locale on Linux which is really
ASCII), you can get into trouble with pathnames that cannot be
converted to UTF-8.
Libarchive's pax writer truncated pathnames and other strings at the
first nonconvertible character. (ouch!) Other archivers have worked
around this by storing unconvertible pathnames as raw binary, a
practice which has been sanctioned by the Austin group. However,
libarchive's pax reader would segfault reading headers that weren't
proper UTF-8. (ouch!) Since bsdtar defaults to pax format, this
affects bsdtar rather heavily.
To correctly support the new "hdrcharset" header that is going into
SUS and to handle conversion failures in general, libarchive's pax reader
and writer have been overhauled fairly extensively. They used to do
most of the pax header processing using wchar_t (Unicode); they now do
most of it using char so that common logic applies to either UTF-8 or
"binary" strings.
As a bonus, a number of extraneous conversions to/from wchar_t have
been eliminated, which should speed things up just a tad.
Thanks to: Bjoern Jacke for originally reporting this to me
Thanks to: Joerg Sonnenberger for noting a bad typo in my first draft of this
Thanks to: Gunnar Ritter for getting the standard fixed
MFC after: 5 days
rely on a deprecated value to set the default. This is also
related to a longer-term goal of setting the default block
size based on format and possibly other factors, which makes
it a bad idea to tie this to a published constant.
new interface. Mark the functions that are going away in
libarchive 3.0.
In particular, archive_version_string() now computes the
string rather than assuming that it will be created by the
build infrastructure. Eventually, this will allow some
simplification of the build infrastructure.
* There are now only two public version identifiers: "number" is
a single integer that combines Major/minor/release in a single
value of the form Mmmmrrr. This is easy to compare against for
checking feature support. "string" is a displayable text string
of the form "libarchive M.mm.rr".
* The number is present both as a macro (version of the installed header)
and a function (version of the shared library). The string form
is available only as a function.
* Retain the older version definitions for now, but mark them all
as deprecated, to disappear in libarchive 3.0 (whenever that happens).
* Rework the various deprecation conditionals to use ARCHIVE_VERSION_NUMBER.
An ancillary goal is to reduce the number of @...@ substitutions that
are required. Someday, I might even be able to avoid build-time
processing of archive.h entirely.
Remove the entirely pointless symbolic constant
and sizeof(unsigned char). (The constant
here is doubly wrong, since not only does
it obscure a basic format constant, it was
never intended to be a tar-specific value,
so could conceivably be changed at some point
in the future.)
filename table whose size is less than 65536 bytes.
The original intention was to not consume the filename table, so the
client will have a chance to look at it. To achieve that, the library
call decompressor->read_ahead to read(look ahead) but do not call
decompressor->consume to consume the data, thus a limit was raised
since read_ahead call can only look ahead at most BUFFER_SIZE(65536)
bytes at the moment, and you can not "look any further" before you
consume what you already "saw".
This commit will turn GNU/SVR4 filename table into "archive format
data", i.e., filename table will be consumed by libarchive, so the
65536-bytes limit will be gone, but client can no longer have access
to the content of filename table.
'ar' support test suite is changed accordingly. BSD ar(1) is not
affected by this change since it doesn't look at the filename table.
Reported by: erwin
Discussed with: jkoshy, kientzle
Reviewed by: jkoshy, kientzle
Approved by: jkoshy(mentor), kientzle
uudecode into the main test driver and invoking it just-in-time
within the various tests.
Also, incorporate a number of improvements to the main test support
code that have proven useful on other projects where I've used this
framework.
(left over from when the unified read/write structure was copied
to form separate read and write structures) and eliminate the
pointless initialization of a couple of the unused fields.
Solaris and AIX.
fcntl(fd, F_DUP2FD, arg) and dup2(fd, arg) are functionnaly equivalent.
Document it.
Add some regression tests (identical to the dup2(2) regression tests).
PR: 120233
Submitted by: Jukka Ukkonen
Approved by: rwaston (mentor)
MFC after: 1 month
Significant changes:
- rev. 1.11: Use PRId64 instead of a cast to long long and %lld to print
an int64_t.
- rev. 1.12: Fix a bug that humanize_number() produces "1000" where it
should be "1.0G" or "1.0M". The bug reported by Greg Troxel.
PR: 118461
PR: 102694
Approved by: rwatson (mentor)
Obtained from: NetBSD
MFC after: 1 month
that there might be starvations, but because we have already locked the
thread, the cpuset settings will always be done before the new thread
does real-world work.
we set scheduling parameters and cpu binding fully in userland, and
because default scheduling policy is SCHED_RR (time-sharing), we set
default sched_inherit to PTHREAD_SCHED_INHERIT, this saves a system
call.
however if current thread is executing cancellation handler, signal
SIGCANCEL may have already been blocked, this is unexpected, unblock the
signal in new thread if this happens.
MFC after: 1 week
and assignment.
- Add a reference to a struct cpuset in each thread that is inherited from
the thread that created it.
- Release the reference when the thread is destroyed.
- Add prototypes for syscalls and macros for manipulating cpusets in
sys/cpuset.h
- Add syscalls to create, get, and set new numbered cpusets:
cpuset(), cpuset_{get,set}id()
- Add syscalls for getting and setting affinity masks for cpusets or
individual threads: cpuid_{get,set}affinity()
- Add types for the 'level' and 'which' parameters for the cpuset. This
will permit expansion of the api to cover cpu masks for other objects
identifiable with an id_t integer. For example, IRQs and Jails may be
coming soon.
- The root set 0 contains all valid cpus. All thread initially belong to
cpuset 1. This permits migrating all threads off of certain cpus to
reserve them for special applications.
Sponsored by: Nokia
Discussed with: arch, rwatson, brooks, davidxu, deischen
Reviewed by: antoine
e_rem_pio2.c:
This case goes up to about 2**20pi/2, but the comment about it said that
it goes up to about 2**19pi/2.
It went too far above 2**pi/2, giving a multiplier fn with 21 significant
bits in some cases. This would be harmful except for a numerical
accident. It happens that the terms of the approximation to pi/2,
when rounded to 33 bits so that multiplications by 20-bit fn's are
exact, happen to be rounded to 32 bits so multiplications by 21-bit
fn's are exact too, so the bug only complicates the error analysis (we
might lose a bit of accuracy but have bits to spare).
e_rem_pio2f.c:
The bogus comment in e_rem_pio2.c was copied and the code was changed
to be bug-for-bug compatible with it, except the limit was made 90
ulps smaller than necessary. The approximation to pi/2 was not
modified except for discarding some of it.
The same rough error analysis that justifies the limit of 2**20pi/2
for double precision only justifies a limit of 2**18pi/2 for float
precision. We depended on exhaustive testing to check the magic numbers
for float precision. More exaustive testing shows that we can go up
to 2**28pi/2 using a 53+25 bit approximation to pi/2 for float precision,
with a the maximum error for cosf() and sinf() unchanged at 0.5009
ulps despite the maximum error in rem_pio2f being ~0.25 ulps. Implement
this.
This reduces the size of a statically-linked binary by approximately 100KB
in a trivial "return (0)" test application. readelf -S was used to verify
that the .text section was reduced and that using strlen() saved a few
more bytes over using sizeof(). Since the section of code is only called
when environ is corrupt (program bug), I went with fewer bytes over fewer
cycles.
I made minor edits to the submitted patch to make the output resemble
warnx().
Submitted by: kib bz
Approved by: wes (mentor)
MFC after: 5 days
them. Thus, any fd whose value is greater than SHRT_MAX is handled
incorrectly (the short value is sign-extended when converted to an int).
An unpleasant side effect is that if fopen() opens a file and gets a
backing fd that is greater than SHRT_MAX, fclose() will fail and the file
descriptor will be leaked. Better handle this by fixing fopen(), fdopen(),
and freopen() to fail attempts to use a fd greater than SHRT_MAX with
EMFILE.
At some point in the future we should look at expanding the file descriptor
in FILE to an int, but that is a bit complicated due to ABI issues.
MFC after: 1 week
Discussed on: arch
Reviewed by: wollman
{SHRT_MAX}, so {STREAM_MAX} should be no greater than that. (This
does not exactly meet the letter of POSIX but comes reasonably close
to it in spirit.)
MFC after: 14 days
gives an average speedup of about 12 cycles or 17% for
9pi/4 < |x| <= 2**19pi/2 and a smaller speedup for larger x, and a
small speeddown for |x| <= 9pi/4 (only 1-2 cycles average, but that
is 4%).
Inlining this is less likely to bust caches than inlining the float
version since it is much smaller (about 220 bytes text and rodata) and
has many fewer branches. However, the float version was already large
due to its manual inlining of the branches and also the polynomial
evaluations.
__kernel_rem_pio2(). This simplifies analysis of aliasing and thus
results in better code for the usual case where __kernel_rem_pio2()
is not called. In particular, when __ieee854_rem_pio2[f]() is inlined,
it normally results in y[] being returned in registers. I couldn't
get this to work using the restrict qualifier.
In float precision, this saves 2-3% in most cases on amd64 and i386
(A64) despite it not being inlined in float precision yet. In double
precision, this has high variance, with an average gain of 2% for
amd64 and 0.7% for i386 (but a much larger gain for usual cases) and
some losses.
this function and its callers cosf(), sinf() and tanf() don't waste time
converting values from doubles to floats and back for |x| > 9pi/4.
All these functions were optimized a few years ago to mostly use doubles
internally and across the __kernel*() interfaces but not across the
__ieee754_rem_pio2f() interface.
This saves about 40 cycles in cosf(), sinf() and tanf() for |x| > 9pi/4
on amd64 (A64), and about 20 cycles on i386 (A64) (except for cosf()
and sinf() in the upper range). 40 cycles is about 35% for |x| < 9pi/4
<= 2**19pi/2 and about 5% for |x| > 2**19pi/2. The saving is much
larger on amd64 than on i386 since the conversions are not easy to
optimize except on i386 where some of them are automatic and others
are optimized invalidly. amd64 is still about 10% slower in cosf()
and tanf() in the lower range due to conversion overhead.
This also gives a tiny speedup for |x| <= 9pi/4 on amd64 (by simplifying
the code). It also avoids compiler bugs and/or additional slowness
in the conversions on (not yet supported) machines where double_t !=
double.
e_rem_pio2.c:
Float and double precision didn't work because init_jk[] was 1 too small.
It needs to be 2 larger than you might expect, and 1 larger than it was
for these precisions, since its test for recomputing needs a margin of
47 bits (almost 2 24-bit units).
init_jk[] seems to be barely enough for extended and quad precisions.
This hasn't been completely verified. Callers now get about 24 bits
of extra precision for float, and about 19 for double, but only about
8 for extended and quad. 8 is not enough for callers that want to
produce extra-precision results, but current callers have rounding
errors of at least 0.8 ulps, so another 1/2**8 ulps of error from the
reduction won't affect them much.
Add a comment about some of the magic for init_jk[].
e_rem_pio2.c:
Double precision worked in practice because of a compensating off-by-1
error here. Extended precision was asked for, and it executed exactly
the same code as the unbroken double precision.
e_rem_pio2f.c:
Float precision worked in practice because of a compensating off-by-1
error here. Double precision was asked for, and was almost needed,
since the cosf() and sinf() callers want to produce extra-precision
results, at least internally so that their error is only 0.5009 ulps.
However, the extra precision provided by unbroken float precision is
enough, and the double-precision code has extra overheads, so the
off-by-1 error cost about 5% in efficiency on amd64 and i386.
variations (e500 currently), this provides a gcc-level FPU emulation and is an
alternative approach to the recently introduced kernel-level emulation
(FPU_EMU).
Approved by: cognet (mentor)
MFp4: e500
fabs(), a conditional branch, and sign adjustments of 3 variables for
x < 0 when the branch is taken. In double precision, even when the
branch is perfectly predicted, this saves about 10 cycles or 10% on
amd64 (A64) and i386 (A64) for the negative half of the range, but
makes little difference for the positive half of the range. In float
precision, it also saves about 4 cycles for the positive half of the
range on i386, and many more cycles in both halves on amd64 (28 in the
negative half and 11 in the positive half for tanf), but the amd64
times for float precision are anomalously slow so the larger
improvement is only a side effect.
Previous commits arranged for the x < 0 case to be handled simply:
- one part of the rounding method uses the magic number 0x1.8p52
instead of the usual 0x1.0p52. The latter is required for large |x|,
but it doesn't work for negative x and we don't need it for large |x|.
- another part of the rounding method no longer needs to add `half'.
It would have needed to add -half for negative x.
- removing the "quick check no cancellation" in the double precision
case removed the need to take the absolute value of the quadrant
number.
Add my noncopyright in e_rem_pio2.c
FP-to-FP method to round to an integer on all arches, and convert this
to an int using FP-to-integer conversion iff irint() is not available.
This is cleaner and works well on at least ia64, where it saves 20-30
cycles or about 10% on average for 9Pi/4 < |x| <= 32pi/2 (should be
similar up to 2**19pi/2, but I only tested the smaller range).
After the previous commit to e_rem_pio2.c removed the "quick check no
cancellation" non-optimization, the result of the FP-to-integer
conversion is not needed so early, so using irint() became a much
smaller optimization than when it was committed.
An earlier commit message said that cos, cosf, sin and sinf were equally
fast on amd64 and i386 except for cos and sin on i386. Actually, cos
and sin on amd64 are equally fast to cosf and sinf on i386 (~88 cycles),
while cosf and sinf on amd64 are not quite equally slow to cos and sin
on i386 (average 115 cycles with more variance).
9pi/2 < |x| < 32pi/2 since it is only a small or negative optimation
and it gets in the way of further optimizations. It did one more
branch to avoid some integer operations and to use a different
dependency on previous results. The branches are fairly predictable
so they are usually not a problem, so whether this is a good
optimization depends mainly on the timing for the previous results,
which is very machine-dependent. On amd64 (A64), this "optimization"
is a pessimization of about 1 cycle or 1%; on ia64, it is an
optimization of about 2 cycles or 1%; on i386 (A64), it is an
optimization of about 5 cycles or 4%; on i386 (Celeron P2) it is an
optimization of about 4 cycles or 3% for cos but a pessimization of
about 5 cycles for sin and 1 cycle for tan. I think the new i386
(A64) slowness is due to an pipeline stall due to an avoidable
load-store mismatch (so the old timing was better), and the i386
(Celeron) variance is due to its branch predictor not being too good.
the the double to int conversion operation which is very slow on these
arches. Assume that the current rounding mode is the default of
round-to-nearest and use rounding operations in this mode instead of
faking this mode using the round-towards-zero mode for conversion to
int. Round the double to an integer as a double first and as an int
second since the double result is needed much earler.
Double rounding isn't a problem since we only need a rough approximation.
We didn't support other current rounding modes and produce much larger
errors than before if called in a non-default mode.
This saves an average about 10 cycles on amd64 (A64) and about 25 on
i386 (A64) for x in the above range. In some cases the saving is over
25%. Most cases with |x| < 1000pi now take about 88 cycles for cos
and sin (with certain CFLAGS, etc.), except on i386 where cos and sin
(but not cosf and sinf) are much slower at 111 and 121 cycles respectivly
due to the compiler only optimizing well for float precision. A64
hardware cos and sin are slower at 105 cycles on i386 and 110 cycles
on amd64.
the same as lrint() except it returns int instead of long. Though the
extern lrint() is fairly fast on these arches, it still takes about
12 cycles longer than the inline version, and 12 cycles is a lot in
applications where [li]rint() is used to avoid slow conversions that
are only a couple of times slower.
This is only for internal use. The libm versions of *rint*() should
also be inline, but that would take would take more header engineering.
Implementing irint() instead of lrint() also avoids a conflict with
the extern declaration of the latter.
on i386 (A64), 5 cycles on amd64 (A64), and 3 cycles on ia64). gcc
tends to generate very bad code for accessing floating point values
as bits except when the integer accesses have the same width as the
floating point values, and direct accesses to bit-fields (as is common
only for long double precision) always gives such accesses. Use the
expsign access method, which is good for 80-bit long doubles and
hopefully no worse for 128-bit long doubles. Now the generated code
is less bad. There is still unnecessary copying of the arg on amd64
and i386 and mysterious extra slowness on amd64.
pi/4 <= |x| <= 3pi/4. Use the same branch ladder as for float precision.
Remove the optimization for |x| near pi/2 and don't do it near the
multiples of pi/2 in the newly optimized range, since it requires
fairly large code to handle only relativley few cases. Ifdef out
optimization for |x| <= pi/4 since this case can't occur because it
is done in callers.
On amd64 (A64), for cos() and sin() with uniformly distributed args,
no cache misses, some parallelism in the caller, and good but not great
CC and CFLAGS, etc., this saves about 40 cycles or 38% in the newly
optimized range, or about 27% on average across the range |x| <= 2pi
(~65 cycles for most args, while the A64 hardware fcos and fsin take
~75 cycles for half the args and 125 cycles for the other half). The
speedup for tan() is much smaller, especially relatively. The speedup
on i386 (A64) is slightly smaller, especially relatively. i386 is
still much slower than amd64 here (unlike in the float case where it
is slightly faster).
saves an average of about 8 cycles or 5% on A64 (amd64 and i386 --
more in cycles but about the same percentage on i386, and more with
old versions of gcc) with good CFLAGS and some parallelism in the
caller. As usual, it takes a couple more multiplications so it will
be slower on old machines.
Convert to __FBSDID().
Maybe. In the meantime, my workarounds for trying to coax UTC without
timegm() are getting uglier and uglier. Apparently, some systems
don't support setenv()/unsetenv(), so you can't set the TZ env var and
hope thereby to coax mktime() into generating UTC. Without that, I
don't see a really good alternative to just giving up and converting to
localtime with mktime(). (I suppose I should research the Perl library
approach for computing an inverse function to gmtime(); that might
actually be simpler than this growing list of hacks.)
now returns a value, which supports such convenient
constructs as:
if (assert(NULL != foo())) {
}
Also be careful to setlocale("C") for each new test to
avoid locale pollution.
Also a couple of minor portability enhancements.
* If the platform can't restore char nodes, block nodes, or fifos,
don't try and just return error.
* Include O_BINARY in most open() calls (define O_BINARY to 0 if the
platform doesn't provide a definition already)
* Refactor the ownership restore to more cleanly support platforms
that don't have any form of {l,f,}chown() call.
* Comment a lingering issue with older Unix-like systems that allow
root to hose the filesystem. I don't (yet) have a good solution for
this, but I expect it will require adding more redundant stat()
calls. <sigh>
MFC after: 14 days
optimization of about 10% for cos(x), sin(x) and tan(x) on
|x| < 2**19*pi/2. We didn't do this before because __ieee754__rem_pio2()
is too large and complicated for gcc-3.3 to inline very well. We don't
do this for float precision because it interferes with optimization
of the usual (?) case (|x| < 9pi/4) which is manually inlined for float
precision only.
This has some rough edges:
- some static data is duplicated unnecessarily. There isn't much after
the recent move of large tables to k_rem_pio2.c, and some static data
is duplicated to good affect (all the data static const, so that the
compiler can evaluate expressions like 2*pio2 at compile time and
generate even more static data for the constant for this).
- extern inline is used (for the same reason as in previous inlining of
k_cosf.c etc.), but C99 apparently doesn't allow extern inline
functions with static data, and gcc will eventually warn about this.
Convert to __FBSDID().
Indent __ieee754_rem_pio2()'s declaration consistently (its style was
made inconsistent with fdlibm a while ago, so complete this).
Fix __ieee754_rem_pio2()'s return type to match its prototype. Someone
changed too many ints to int32_t's when fixing the assumption that all
ints are int32_t's.
reallocation, when junk filling is enabled. Junk filling must occur
prior to shrinking, since any deallocated trailing pages are immediately
available for use by other threads.
Reported by: Mats Palmgren <mats.palmgren@bredband.net>
allocation patterns, number of CPUs, and MALLOC_OPTIONS settings indicate
that lazy deallocation has the potential to worsen throughput dramatically.
Performance degradation occurs when multiple threads try to clear the lazy
free cache simultaneously. Various experiments to avoid this bottleneck
failed to completely solve this problem, while adding yet more complexity.
Bruce for putting lots of effort into these; getting them right isn't
easy, and they went through many iterations.
Submitted by: Steve Kargl <sgk@apl.washington.edu> with revisions from bde
is a violation of RFC 1034 [STD 13], it is accepted by certain name servers
as well as other popular operating systems' resolver library.
Bugs are mine.
Obtained from: ume
MFC after: 2 weeks
of disk names, where you must free each pointer, as well as the array
by hand. [1]
- Destaticize "disks" in Disk_Names, it has no reasons to be static.
PR: kern/96077 [1]
PR: kern/114110 [1]
MFC after: 1 month
Approved by: rwatson (mentor)
|x| or |y| and b is |y| or |x|) when mixing NaN arg(s).
hypot*() had its own foot shooting for mixing NaNs -- it swaps the
args so that |x| in bits is largest, but does this before quieting
signaling NaNs, so on amd64 (where the result of adding NaNs depends
on the order) it gets inconsistent results if setting the quiet bit
makes a difference, just like a similar ia64 and i387 hardware comparison.
The usual fix (see e_powf.c 1.13 for more details) of mixing using
(a+0.0)+-(b+0.0) doesn't work on amd64 if the args are swapped (since
the rder makes a difference with SSE). Fortunately, the original args
are unchanged and don't need to be swapped when we let the hardware
decide the mixing after quieting them, but we need to take their
absolute value.
hypotf() doesn't seem to have any real bugs masked by this non-bug.
On amd64, its maximum error in 2^32 trials on amd64 is now 0.8422 ulps,
and on i386 the maximum error is unchanged and about the same, except
with certain CFLAGS it magically drops to 0.5 (perfect rounding).
Convert to __FBSDID().
be into 12+24 bits of precision for extra-precision multiplication,
but was into 13+24 bits. On i386 with -O1 the bug was hidden by
accidental extra precision, but on amd64, in 2^32 trials the bug
caused about 200000 errors of more than 1 ulp, with a maximum error
of about 80 ulps. Now the maximum error in 2^32 trials on amd64
is 0.8573 ulps. It is still 0.8316 ulps on i386 with -O1.
The nearby decomposition of 1/ln2 and the decomposition of 2/(3ln2) in
the double precision version seem to be sub-optimal but not broken.
This uses 2 tricks to improve consistency so that more serious problems
aren't hidden in simple regression tests by noise for the NaNs:
- for a signaling NaN, adding 0.0 generates the invalid exception and
converts to a quiet NaN, and doesn't have too many effects for other
types of args (it converts -0 to +0 in some rounding modes, but that
hopefully doesn't change the result after adding the NaN arg). This
avoids some inconsistencies on i386 and ia64. On these arches, the
result of an operation on 2 NaNs is apparently the largest or the
smallest of the NaNs as bits (consistently largest or smallest for
each arch, but the opposite). I forget which way the comparison
goes and if the sign bit affects it. The quiet bit is is handled
poorly by not always setting it before the comparision or ignoring
it. Thus if one of the args was originally a signaling NaN and the
other was originally a quiet NaN, then the result depends too much
on whether the signaling NaN has been quieted at this point, which
in turn depends on optimizations and promotions. E.g., passing float
signaling NaNs to double functions must quiet them on conversion;
on i387, loading a signaling NaN of type float or double (but not
long double) into a register involves a conversion, so it quiets
signaling NaNs, so if the addition has 2 register operands than it
only sees quiet NaNs, but if the addition has a memory operand then
it sees a signaling NaN iff it is in the memory operand.
- subtraction instead of addition is used to avoid a dubious optimization
in old versions of gcc. For SSE operations, mixing of NaNs apparently
always gives the target operand. This is not as good as the i387
and ia64 behaviour. It doesn't mix NaNs at all, and makes addition
not quite commutative. Old versions of gcc sometimes rewrite x+y
to y+x and thus give different results (in bits) for NaNs. gcc-3.3.3
rewrites x+y to y+x for one of pow() and powf() but not the other,
so starting from float NaN args x and y, powf(x, y) was almost always
different from pow(x, y).
These tricks won't give consistency of 2-arg float and double functions
with long double ones on amd64, since long double ones use the i387
which has different semantics from SSE.
Convert to __FBSDID().
and trunc() to the corresponding long double functions. This is not
just an optimization for these arches. The full long double functions
have a wrong value for `huge', and the arches without full long doubles
depended on it being wrong.
This has the side effect of confusing gcc-4.2.1's optimizer into more
often doing the right thing. When it does the wrong thing here, it
seems to be mainly making too many copies of x with dependency chains.
This effect is tiny on amd64, but in some cases on i386 it is enormous.
E.g., on i386 (A64) with -O1, the current version of exp2() should
take about 50 cycles, but took 83 cycles before this change and 66
cycles after this change. exp2f() with -O1 only speeded up from 51
to 47 cycles. (exp2f() should take about 40 cycles, on an Athlon in
either i386 or amd64 mode, and now takes 42 on amd64). exp2l() with
-O1 slowed down from 155 cycles to 123 for some args; this is unimportant
since the i386 exp2l() is a fake; the wrong thing for it seems to
involve branch misprediction.
faster on all machines tested (old Celeron (P2), A64 (amd64 and i386)
and ia64) except on ia64 when compiled with -O1. It takes 2 more
multiplications, so it will be slower on old machines. The speedup
is about 8 cycles = 17% on A64 (amd64 and i386) with best CFLAGS
and some parallelism in the caller.
Move the evaluation of 2**k up a bit so that it doesn't compete too
much with the new polynomial evaluation. Unlike the previous
optimization, this rearrangement cannot change the result, so compilers
and CPU schedulers can do it, but they don't do it quite right yet.
This saves a whole 1 or 2 cycles on A64.
when the result is +-0. IEEE754 requires (in all rounding modes) that
if the result is +-0 then its sign is the same as that of the first
arg, but in round-towards-minus-infinity mode an uncorrected implementation
detail always reversed the sign. (The detail is that x-x with x's
sign positive gives -0 in this mode only, but the algorithm assumed
that x-x always has positive sign for such x.)
remquo() and remquof() seem to need the same fix, but I cannot test them
yet.
Use long doubles when mixing NaN args. This trick improves consistency
of results on at least amd64, so that more serious problems like the
above aren't hidden in simple regression tests by noise for the NaNs.
On amd64, hardware remainder should be used since it is about 10 times
faster than software remainder and is already used for remquo(), but
it involves using the i387 even for floats and doubles, and the i387
does NaN mixing which is better than but inconsistent with SSE NaN mixing.
Software remainder() would probably have been inconsistent with
software remainderl() for the same reason if the latter existed.
Signaling NaNs cause further inconsistencies on at least ia64 and i386.
Use __FBSDID().
