Commit Graph

130 Commits

Author SHA1 Message Date
Dmitriy Ivanov
4d0d31475f Add versions to libm.so symbols
Bug: http://b/20139821
Change-Id: Ifdfcac62d59eb8f0985990aa8512046ecb87449e
(cherry picked from commit 636f5ddddc)
2015-06-17 13:50:37 -07:00
Dmitriy Ivanov
69a5fb951d Apply work around b/19059885 to x86
Bug: http://b/19059885
Bug: http://b/21203348
Change-Id: Ic375e9f877d68de8f866d17362879a7dde638465
2015-05-15 18:39:16 -07:00
Elliott Hughes
97139e6adc Merge "Update NOTICE files." into mnc-dev 2015-05-12 17:29:25 +00:00
Elliott Hughes
f0bd92db89 Update NOTICE files.
Change-Id: I4e170118abf452dd87ffe79d343ba1e2b53955de
2015-05-12 10:09:01 -07:00
Dmitriy Ivanov
f7009f450e Switch libc/libcstdc++/libm to hash-style=both
Bug: http://b/19059885
Change-Id: Ie42703d91f291fd689a1cf9c35bcab5c487b9a84
(cherry picked from commit 88e777d6f4)
2015-05-08 12:17:21 -07:00
Christopher Ferris
6e86146b85 Fix typo in alias symbol.
Change-Id: I41d2c29bd4a9ba382bbe3440541c8e1506fc5809
2015-03-14 12:17:45 -07:00
Christopher Ferris
995b813e91 For libm, use a macro for aliasing symbols.
Change-Id: Ibd42ebc387c2bf3eba9aa96091770915b4b34184
2015-03-13 23:57:15 -07:00
Dmitriy Ivanov
d7e091efc4 Merge "Switch libc and libm to sysv-only hash style" 2015-03-13 21:04:08 +00:00
Dmitriy Ivanov
2e16d2cf1e Switch libc and libm to sysv-only hash style
Bug: 19059885
Change-Id: I60a23dc5f9c756994d566818332ca42b305b4a05
2015-03-13 19:48:03 +00:00
Christopher Ferris
f074b531d2 Merge "Add the optimized implementation of 18 math functions for x86 and x86_64 respectively" 2015-03-13 17:18:21 +00:00
Goran Jakovljevic
1156508265 Generate gnu hash for arm only
Fixes build failure caused by:
https://android-review.googlesource.com/#/c/139660

As previously discussed [1], MIPS does not support GNU-style ELF hashes.

[1] https://android-review.googlesource.com/#/c/49282/2/tests/Android.mk

Change-Id: I9a7966eebfd1ef0a587a20b71faefde38e84ab62
2015-03-10 10:01:50 -07:00
Dmitriy Ivanov
4a55c46d30 Generate sysv hash for libc, libstdc++ and libm
Bug: 19059885
Change-Id: I35e2a9cd0bb3914d9d0c82d163bfaf0fff844667
2015-03-09 19:38:56 -07:00
Jingwei Zhang
5d4f0e6a26 Add the optimized implementation of 18 math functions for x86 and x86_64 respectively
Change-Id: I31bf601448a9427f825517f3a0ff24de47f49bfa
Signed-off-by: Jingwei Zhang <jingwei.zhang@intel.com>
Signed-off-by: Mingwei Shi <mingwei.shi@intel.com>
2015-03-09 13:19:08 -07:00
James Rose
45789b63db libm: Add hardware sqrt, ceil, floor and trunc for x86 & x86_64
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>
2015-03-06 16:48:30 -08:00
Shu Zhang
c78fa26d47 libm: arm: add arm specific sqrt and sqrtf
Add arm specific optimized sqrt and sqrtf.

Change-Id: I8ea417fc98a800c8cf4f47a0c6668c51d927b1dc
2015-03-04 14:07:35 -08:00
Amaury Le Leyzour
32936c895d Add aarch64 instructions wherever possible
fma, sqrt and various rounding functions have aarch64 instruction equivalent

Change-Id: I1284f31b9f78f914281e5563b8d44db8362b627d
2015-03-04 12:18:52 -08:00
Shu Zhang
3a629af0ad libm: arm: Add arm specific floor() optimization
Add arm specific floor() implementation which avoids VMSR and VMRS
instructions.

Change-Id: Ibd4cd7147aa2f98c9b5bbaf74948843ea619dba4
2015-02-12 18:27:58 -08:00
Duane Sand
3d535d2312 [MIPS] Fix Mips64 build of libm
Change-Id: Ib742b408d4f21d5ec8d455a3fea7c08165c7f321
2015-02-11 12:51:42 -08:00
Christopher Ferris
d40d1a08cd Cleanup the libm makefile.
Change-Id: I3d95f75da24ee3b064a9cce67586b0e6f32272d8
2015-02-09 14:09:49 -08:00
Elliott Hughes
998f2d5a50 Remove more BSD cruft from the public headers.
Change-Id: I1e3097f745b526db448bf72fbc05fb2ba23929a0
2014-12-19 19:30:11 -08:00
Elliott Hughes
6b2b5856d9 Update bionic NOTICE files.
Change-Id: Ic280872ef3137febd9d4a87d5c7662e264692071
2014-12-18 16:27:30 -08:00
Dan Albert
da194780b1 Disable ASAN for more of our libraries.
Since ASAN depends on these libraries, we need to exclude them from
being instrumented.

Change-Id: I62005b6d04ee7b2283c4bd54ace5304386c67358
2014-12-18 08:03:49 -08:00
Elliott Hughes
1b66f6e9d7 Move a private x86-64 header file out of the limelight.
Currently this is on everyone's include path and is getting exported
via the NDK.

Change-Id: Ia417839fb65c7e0afddcd9fae3438665dae8e752
2014-11-26 19:31:19 -08:00
Elliott Hughes
b8ee16f1dc Fix our <complex.h> support.
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
2014-11-06 11:16:55 -08:00
Dan Albert
e91d9cd4ad Don't use an STL for libm.
The default is none right now, but this will be needed when the
default becomes libc++.

Change-Id: Ib022adac2b9dc08fbfae830908128043cfffb91e
2014-10-16 07:44:47 -07:00
Elliott Hughes
14538ca75e Catch up on upstream's round/roundf/roundl.
Not sure how we missed these, but better late than never...

Change-Id: Ib08d1bb6e340a1907cbeb1cbe220e33f70642bdc
2014-10-10 10:21:43 -07:00
Elliott Hughes
488268b134 Sync libm with upstream.
Change-Id: I3b4e2c9c6ce6c5934f270a51ce5eb9154c5805d5
2014-10-09 16:20:37 -07:00
Dan Albert
12e3f22c7c Add coverage flag for libm.
Change-Id: I195d38c5e1f50d1cc6d4353dc8857f11b803b7a3
2014-09-30 16:56:43 -07:00
Elliott Hughes
755318548d Fix lgamma_r/lgammaf_r/lgammal_r for -0.
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
2014-09-18 11:23:58 -07:00
Elliott Hughes
c729d4f237 Get new coshl, sinhl, and tanhl implementations from upstream.
Change-Id: I92f9cd53d12efb2e0f4f8a9590b9fe42398233f6
2014-09-12 16:09:40 -07:00
Elliott Hughes
460ad7454a Sync libm with upstream.
Change-Id: I8ac8ee52122ee19a2e423c3211092023cb4896eb
2014-09-12 14:00:02 -07:00
Chih-Hung Hsieh
356a6249d8 Disable Clang, due to x86 LDBL_MANT_DIG errors.
Clang -m32 has incorrect sizeof(long double) = 12.
With -m64, LDBL_MANT_DIG is incorrect 64.

BUG: 17163651
Change-Id: I4b157aa1a8572c3a10aece6070e119a292ab8e83
2014-09-04 12:35:19 -07:00
Elliott Hughes
e0c56efddf Use __GNUC_PREREQ rather than __GNUC_PREREQ__ to match glibc.
Bug: 16874785
Change-Id: I8512f8be3fd149d8720c5c3b4657bedd5ce2b1d1
2014-08-29 15:54:11 -07:00
Elliott Hughes
f4c948a9e9 Move the meat of <features.h> into <sys/cdefs.h>.
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
2014-08-19 11:18:20 -07:00
Elliott Hughes
5f5cc45cf0 Fix <features.h> (_BSD_SOURCE and _GNU_SOURCE).
<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
2014-08-18 16:04:03 -07:00
Elliott Hughes
1e83245bae Revert "Switch libm to building with clang."
This reverts commit aadc4b2ff0.

The patch broke x86.

Change-Id: I0cda903732a685838bbd1877f94e1593a2a32acc
2014-07-11 01:24:15 +00:00
Elliott Hughes
aadc4b2ff0 Switch libm to building with clang.
Bug: 16211965
Change-Id: I81ec9706a4f2b5e8a840508a0f711cd25a5826ca
2014-07-10 16:58:07 -07:00
Dan Albert
71f9b34fbe Revert "Revert "Prevent symbols from libgcc from being reexported.""
This reverts commit 4e4e1a5d78.

Change-Id: I8e78c3a59598e052e02965dd37b3fba8f567eed9
2014-06-24 04:31:08 +00:00
Dan Albert
4e4e1a5d78 Revert "Prevent symbols from libgcc from being reexported."
This reverts commit 65e80cf622.

Change-Id: I5f94b123f73db8d0b03cd71a5cffa0662776a017
2014-06-23 20:12:05 +00:00
Dan Albert
65e80cf622 Prevent symbols from libgcc from being reexported.
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
2014-06-21 00:09:31 -07:00
Elliott Hughes
b6ca747402 Merge "Add a flag control that PRIVATE libm can be chosen" 2014-06-17 20:56:23 +00:00
Elliott Hughes
4779a11a3d Merge "Move mips fenv implementation details into fenv.c." 2014-06-09 22:27:30 +00:00
Elliott Hughes
89d61243f2 Move mips fenv implementation details into fenv.c.
Change-Id: I2415e4808e40c2981d016c01969ba14ea22bf82e
2014-06-09 15:25:15 -07:00
Elliott Hughes
43bf81e544 Move x86 fenv implementation details into fenv.c.
Change-Id: I6cb8c730483c325dc3cb75c2b2fbdd2d8455a54c
2014-06-09 15:20:59 -07:00
Elliott Hughes
9812a0233c Clean up the ARM fenv.h, moving implementation details into fenv.c.
Change-Id: I919d35db27ec16a6f54a7a5c3c070f33f03d9bb9
2014-06-09 13:57:57 -07:00
Serban Constantinescu
a147a1da5c AArch64: libm: Fix ARM64 fenv_t and refactor ARM64 libm implementation.
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>
2014-06-09 13:52:28 -07:00
Ben Cheng
ff00dc8565 Workaround the GCC "(long)fn -> lfn" optimization bug.
Pass "-fno-builtin-rint -fno-builtin-rintl -fno-builtin-rintf" to avoid
generating self recursions in lrint, lrintl, and lrintf.

BUG: 14225968
Change-Id: Iffc4d8f0692580d636a529f4c8aa9a7a10d7acd8
2014-05-29 16:32:00 -07:00
Elliott Hughes
de9ac71053 Use -fvisibility=hidden to build libm.
Bug: 11156955
Change-Id: I77b7772a44313a1eb3970f5d2c8e2af63690efa4
2014-05-19 16:58:52 -07:00
Elliott Hughes
b4f2f28054 Declare sincos/sincosf/sincosl in <math.h>.
Change-Id: If7b98f98c7ab4d3d77d310be082bff64512d51a8
2014-05-08 21:19:12 -07:00
Elliott Hughes
9a5a3e8e74 Fix <math.h> to quieten most of our warnings.
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
2014-05-05 21:19:47 -07:00