FreeBSD doesn't seem interested in having low-quality implementations of
these functions:
cacoshl, cacosl, casinhl, casinl, catanhl, catanl, ccoshl, ccosl, cexpl,
clog, clogf, clogl, cpow, cpowf, cpowl, csinhl, csinl, ctanhl, ctanl.
And they still haven't got round to writing good implementations, so for
now let's just take the NetBSD ones so we have the full set.
Bug: http://b/27555792
Change-Id: I6b72003cf749b1043f006377a01fffe5e1d659bc
Instead of adding the used common files to each arch that doesn't
override a file, make the architectures list the files they override.
Also updates the Android.bp file to match the Android.mk file.
Change-Id: I7d6a9c2e1c6b6cc9430aa818a89a2ccf52f62c98
Even though the control registers being read/written in fenv.c only
have 32 bits that are used, the instructions take a 64 bit register.
Make sure the inline assembler in the macros use 64 bit values.
Verified that before the change and after the change, the disassembly
is exactly the same.
In addition, add -Werror to the cflags.
Change-Id: I6603779327488c23e3aab13300edf2e02b101916
Don't enable the inlines when building libm itself. Otherwise clang gets
upset by seeing both an inline and a non-inline definition.
This reverts commit c5deb0f883.
Change-Id: If7abdb351f5a5549d6a331b33af408e8fcfa9868
Also remove cruft meant to support long-obsolete compilers. More
benchmarks.
Bug: http://b/23195789
Change-Id: Ief538e41e77a77e8013b2f4f359584e8df2c47d8
Clang/llvm x86_64 configuration of long double is still
incompatible with gcc.
https://llvm.org/bugs/show_bug.cgi?id=23897
BUG: 21899249
Change-Id: I47fd6d0ce2daa84ae0035e7885a4d90e4c3056f7
Add hardware implementations for sqrt, ceil, floor and trunc for
x86 and x86_64. These routines, and in particular sqrt are much
faster than the BSD C language versions of these functions.
Fixed whitespace errors.
Revised x86 versions with respect to alignment.
Rebased for Android 5.0
Change-Id: I86bdb520ce5e589b0cf63778f353fbd3263c8f0e
Author: James Rose <james.rose@intel.com>
Signed-off-by: James Rose <james.rose@intel.com>
We build libm with -fvisibility=hidden, so we weren't exporting any
of the <complex.h> functions.
We also weren't building many of the functions anyway.
We were also missing the complex inverse trigonometric functions.
And because we didn't even have perfunctory "call each function once"
tests, we didn't notice that we weren't exporting any symbols, so this
patch adds at least that level of testing.
Change-Id: Ibcf2843f507126c51d134cc5fc8d67747e033a0d
Upstream has implemented lgammal/lgammal_r for ld128, and fixed the
sign problem we reported with all the lgamma*_r functions and -0.
Bug: 17471883
Change-Id: Ibb175d9cab67efae75f1010796fd44c9ba6ce4fc
This way it's a lot harder for us to screw up (since we should always
be including <sys/cdefs.h> anyway).
Bug: 14659579
Change-Id: I23070fff3296b0d1c683bb5e3a6e214146327d53
<features.h> is supposed to take user-settable stuff like _GNU_SOURCE
and _BSD_SOURCE and turn them into __USE_GNU and __USE_BSD for use in
the C library headers. Instead, bionic used to unconditionally define
_BSD_SOURCE and _GNU_SOURCE, and then test _GNU_SOURCE in the header
files (which makes no sense whatsoever).
Bug: 14659579
Change-Id: Ice4cf21a364ea2e559071dc8329e995277d5b987
Since this was not done earlier, there are binary compatibility concerns
that prevent us from being able to apply this to LP32.
Bug: 11156955
Change-Id: Ie717c3ae4b81c749548a45a993c834e109700b27
This patch fixes the ARM64 ABI for libm. fenv_t is now split in 32bit status
and 32bit control. This mirrors the AArch64 FPU control and status
registers (FPCR, FPSR).
The patch also refactors the libm implementation for ARM64 into a finer
grained control over the FPU registers.
Bionic-benchmarks has been expanded with 3 more benchmarks for floating
point operations. The new libm implementation for ARM64 performs better
over all the math benchmarks available.
Change-Id: I2a7f81d6b4e55c91f8a63a4c69614fc8b1bcf2db
Signed-off-by: Serban Constantinescu <serban.constantinescu@arm.com>
I've reported the wcsftime bug upstream, but we really just want to use -D
to ensure the buggy code isn't built. (I've also brought our strftime a bit
closer to upstream now we have the right define.)
I don't think upstream is likely to fix all their sign-compare and
uninitialized warnings, so let's just silence them.
As for libm, again upstream isn't likely to fix all their warnings, and
silencing those made the ones that were our fault stand out. I've fixed
our <math.h> to fix the warnings caused by our lack of definitions for
the non-imprecise long-double functions. I checked the C99 standard, and
all these functions are there.
Change-Id: Iee8e1182c1db375058fb2c451eceb212bab47a37
- promoted IEEEld2bits to fpmath since most of the where the same for
diffrent archs
- removed _fpmath
- reinstated weak_references
- moved isfinite and isnormal to libc
- clean up fake_long_doubles
- clean up some useless ifdefs
- added missing nexttoward* tests
Bug: 14134235
Change-Id: I95639c4885653fe47fd7dc0570ee5bb3389bbc6b
Also move isinf and isnan into libc like everyone else.
Also move fpclassify to libc like the BSDs (but unlike glibc). We need
this to be able to upgrade our float/double/long double parsing to gdtoa.
Also add some missing aliases. We now have all of:
isnan, __isnan, isnanf, __isnanf, isnanl, __isnanl,
isinf, __isinf, isinff, __isinff, isinfl, __isinfl,
__fpclassify, __fpclassifyd, __fpclassifyf, __fpclassifyl.
Bug: 13469877
Change-Id: I407ffbac06c765a6c5fffda8106c37d7db04f27d
* reworked amd64/_fpmath.h and arm64/_fpmath.h to support 128-bit long
doubles.
* improved tests to cover long double cases
* made modfl an alias for LP32
Tests pass on x86, x86_64, arm, arm64 and mips.
Bug: 12921273
Change-Id: Ibe39acde57972447a8950fa45b1e702acc68ebeb
The most relevant changes are the ones in s_erff.c which should improve
the overall speed of exececution and the maximum error in ULP.
Change-Id: I7bd867f902db418db67f02cd58578624357b1ee8
Factored out common declarations to include/fenv.h and pushed
the implementation to .c files.
Bug: 11050744
Change-Id: I446b13cc4bc599d328343a8d392b07de280f6304
libc/libm support for MIPS64 targets
Change-Id: I8271941d418612a286be55495f0e95822f90004f
Signed-off-by: Chris Dearman <chris.dearman@imgtec.com>
Signed-off-by: Raghu Gandham <raghu.gandham@imgtec.com>
Use arch-specific LOCAL_ variables to build libm for both
TARGET_ARCH and TARGET_2ND_ARCH.
Bug: 11654773
Change-Id: I6da794ba722bb68e7484d8869c6eb0425b7d17cb
The aarch64 toolchain doesn't support them, and we were already implementing
most of the fake long-double functions differently anyway.
Change-Id: I4a2f2df81972ee9c34ddfe96cec81b126506d881
This patch adds intial support for AArch64 to bionic's libm.
Change-Id: I9ae0f895bbdd7fe67815e6ca1ead627581163a27
Signed-off-by: Serban Constantinescu <serban.constantinescu@arm.com>
Also fix the signature of usleep, and the definition of useconds_t which
should be unsigned, as the 'u' in its name implies.
This patch also cleans up the existing FreeBSD hacks by moving the libm
stuff from <sys/cdefs.h> to a libm-private header, and adding comments
about the hacks we use to build FreeBSD source.
Change-Id: Ibe5067a380502df94a0a3a7901969b35411085b6
This was originally motivated by noticing that we were setting the
wrong bits for the well-known tls entries. That was a harmless bug
because none of the well-known tls entries has a destructor, but
it's best not to leave land mines lying around.
Also add some missing POSIX constants, a new test, and fix
pthread_key_create's return value when we hit the limit.
Change-Id: Ife26ea2f4b40865308e8410ec803b20bcc3e0ed1
This doesn't usually matter because there's a weak reference; you'd
miss this if you stopped using s_logb.c though.
Change-Id: I912fceae327a378031cd24a64aefa9dfd84f7f90
This brings us up to date with FreeBSD HEAD, fixes various bugs, unifies
the set of functions we support on ARM, MIPS, and x86, fixes "long double",
adds ISO C99 support, and adds basic unit tests.
It turns out that our "long double" functions have always been broken
for non-normal numbers. This patch fixes that by not using the upstream
implementations and just forwarding to the regular "double" implementation
instead (since "long double" on Android is just "double" anyway, which is
what BSD doesn't support).
All the tests pass on ARM, MIPS, and x86, plus glibc on x86-64.
Bug: 3169850
Bug: 8012787
Bug: https://code.google.com/p/android/issues/detail?id=6697
Change-Id: If0c343030959c24bfc50d4d21c9530052c581837
We have two copies of fenv.h for every architecture, one of which
isn't used. We also have unused makefiles and files for architectures
we don't support.
This patch removes all the obviously useless files.
Bug: http://code.google.com/p/android/issues/detail?id=38196
Change-Id: I1919b6621ba513aa24aa947a34815bc51191487c
This should help prevent broken builds next time I'm messing with
assembler/compiler/linker flags...
Change-Id: I30f15a3ce3c3f3c60cad7bc59aaba9f42d792224
Remove the hand-collated ones, and switch to a script that pulls the
copyright headers out of every file and collects the unique ones.
Change-Id: Ied3b98b3f56241df97166c410ff81de4e0157c9d
Currently we will get the wrong result as follows:
remquof(0x7bb33336, 0x63000000) = -671088640, 0x00000000
remquo(0xbff0000000000003, 0x3ff0000000000003) = 1, 0x8000000000000000
remquo(0x9120000000000001, 0x0000000000000005) = -1288490188, 0x0000000000000004
while the correct one should be:
remquof(0x7bb33336, 0x63000000) = 1476395008, 0x00000000
remquo(0xbff0000000000003, 0x3ff0000000000003) = -1, 0x8000000000000000
remquo(0x9120000000000001, 0x0000000000000005) = -1288490189, 0x0000000000000001
Fixed in this patch.
Change-Id: I540b348cd10a539f3b39b1753945c893c4c7ec46
Signed-off-by: Jingwei Zhang <jingwei.zhang@intel.com>
Signed-off-by: Jin Wei <wei.a.jin@intel.com>
Signed-off-by: Jack Ren <jack.ren@intel.com>
Signed-off-by: Bruce Beare <bruce.j.beare@intel.com>
from http://svnweb.freebsd.org/base?view=revision&revision=176101
"
Oops, fix the fix in rev.1.10. logb() and logbf() were broken on
denormals, and logb() remained broken after 1.10 because the fix for
logbf() was incompletely translated.
Convert to __FBSDID().
"
Change-Id: I54f33648db7c421b06eee1ea8e63c57a179fae0d
Signed-off-by: Jack Ren <jack.ren@intel.com>
Signed-off-by: Jin Wei <wei.a.jin@intel.com>
Signed-off-by: Bruce Beare <bruce.j.beare@intel.com>
We don't have a toolchain anymore, we don't have working original
kernel headers, and nobody is maintaining this so there is really
no point in keeping this here. Details of the patch:
- removed code paths from Android.mk files related to the SuperH
architecture ("sh")
- removed libc/arch-sh, linker/arch-sh, libc/kernel/arch-sh
- simplified libc/SYSCALLS.TXT
- simplified the scripts in libc/tools/ and libc/kernel/tools
Change-Id: I26b0e1422bdc347489e4573e2fbec0e402f75560
Signed-off-by: David 'Digit' Turner <digit@android.com>
http://code.google.com/p/android/issues/detail?id=19276
GCC would remove inline asm due to lack of knowledge
of FPU register changes.
Change-Id: I9f9e8623fa6580843b7cd8178439ace8c2db2d51
Signed-off-by: Mark D Horn <mark.d.horn@intel.com>
Signed-off-by: Bruce Beare <bruce.j.beare@intel.com>
Author: Jingwei Zhang <jingwei.zhang@intel.com>
sincos() functions would be turned into infinite calls to
itself if sincos optimization is applied to itself. See
gcc bugzilla http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46926
arm-linux-androideabi-4.4.3 toolchain does not have this problem
because sincos optimization is entirely disabled. Starting
from arm-linux-androideabi-4.6 toolchain, we enable sincos optimization
in gcc.
This patch simply enforce -O0 on this function to minimize the
change.
Change-Id: I0fc00b5f1dd71c0a024943bdedfed29b0d195e82
This patch fixes a known bug in bionic libm
due to aliasing issues in gcc 4.2 and 4.4; more
specifically in frexpf.
The function frexpf is used to extract the
mantissa and exponent from a double precision number.
The bug has already been reported here:
https://code.google.com/p/android/issues/detail?id=6697
Change-Id: I2e1f2e0a45906642d2225b9d150ed391d2bf331c
Signed-off-by: Rodrigo Obregon <robregon@ti.com>
Original compiler error:
target arm C: libm <= bionic/libm/src/e_j0f.c
bionic/libm/src/e_j0f.c: In function 'j0f':
bionic/libm/src/e_j0f.c:66: warning: comparison between signed and unsigned integer expressions
bionic/libm/src/e_j0f.c: In function 'y0f':
bionic/libm/src/e_j0f.c:140: warning: comparison between signed and unsigned integer expressions
target arm C: libm <= bionic/libm/src/e_j1.c
It's subtle but ix is masked with 0x7f000000 so it can never ever have a value
greater than 0x80000000. So I switched to using the unmasked hx and added a
cast as a reward to the compiler for being right.
I checked the original routines that e_j0f.c was ported from (in e_j0.c) and
the double's don't use 0x80000000 so this issue didn't exist there.
Let that be a warning to those that just slap on casts to shut up the compiler,
sometimes it's sniffed out a bug for you. :-)
Similar fixes in the other functions.
Change-Id: I7a776e5d4721fc3a9e3bd89179b67e9af3a2ebfa
Warning from compiler:
target arm C: libm <= bionic/libm/src/e_atan2.c
bionic/libm/src/e_atan2.c: In function 'atan2':
bionic/libm/src/e_atan2.c:71: warning: suggest parentheses around arithmetic in operand of '|'
target arm C: libm <= bionic/libm/src/e_atan2f.c
