Run-time CPU detection (RTCD) is enabled by default if target platform support
it.
It can be disable at compile time with --disable-rtcd option.
Add RTCD support for ARM architecture.
Thanks to Timothy B. Terriberry for help and code review
Signed-off-by: Timothy B. Terriberry <tterribe@xiph.org>
Needed by commit 972a34ec2c.
Use autoreconf in autogen.sh instead of the handwritten version,
it's simpler, and also updates things that we weren't handling.
Drop the hand-written INSTALL file. Its information content was
~zero, and autotools wants to overwrite it with its own version,
so don't fight that, just .gitignore it.
Original patch by Aurélien Zanelli <aurelien.zanelli@parrot.com>:
http://lists.xiph.org/pipermail/opus/2013-May/002078.html
Revised version:
- Add autconf detection (ported from libtheora).
- Rename ARM5E to ARMv5E (an ARM5 is not the same thing as ARMv5!).
- Use actual macros so they can still be selectively overridden.
- Split out ARMv4 parts and add a few more ARMv4 macros.
- Label blocks to make them easy to find in generated assembly.
- Fix MULT16_32_Q15() so we can pass make check.
The MDCT test passes in values larger than 2**30 for b.
The new version should be just as fast (or faster, since it's
easier to merge the shift with following instructions), and
there's no appreciable impact on accuracy (FFT/MDCT SNR actually
goes up in most cases).
- Fix register constraints.
We were using early-clobber flags in a bunch of places that
didn't need them, and commutative-pair flags in a bunch of
places that weren't actually commutative.
This was Jean-Marc's fault (the original code came from Speex).
- Simplify silk_CLZ16().
- Port over iFFT C_MULC asm by Andree Buschmann
<AndreeBuschmann@t-online.de> from Rockbox.
- Speed up the C_MULC asm by using LDRD, allowing more flexible
addressing, re-ordering instructions to avoid some stalls,
allowing more flexible register allocation, and getting things
out of the inline asm block so the compiler can schedule them
better.
- Add C_MUL and C_MUL4 asm for the FFT to the encoder based, on the
new C_MULC.
In total, this patch gives a 22.3% speed-up on test_opus_encoder on
a 600 MHz Cortex A8 using gcc 4.2.1,
When restricted to ARMv4 optimizations, it gives a 9.6% speed-up
on the same processor/compiler.
On the conformance test vectors:
Average mono quality is 97.0583 %
Average stereo quality is 97.775 %
We shouldn't ever have any trailing newlines that need trimming here,
and the _s version wasn't added to m4sugar.m4 until autoconf 2.63b,
so this will let it work with 2.13 again.
This one meets or exceeds the following requirements:
- Version is checked/updated for every build action when in the git repo.
Does not require the user to re- ./configure to get the correct version.
- Version is not updated automatically when using exported tarball source.
Avoids accidentally getting a wrong version from some other git repo in
a parent directory of the source, and allows setting the correct version
for distro package exports.
- Automatic updating can be manually suppressed.
For developers doing lots of change/rebuild cycles they don't plan to
release, when they don't want a full rebuild triggered for every commit,
and again for every change made immediately after a commit.
The version will still always be updated if they do a `make dist`.
- Does not require any manual updating of versions in the mainline git
repo for each release aside from normal tagging. The version is
recorded in one file only, that is automatically generated and will
never need to be committed.
- Does not require gnu-make features for the autoconf builds.
It does not currently:
- Keep a checksum of every source file in tarball releases to mangle the
version if people modify the tarball source. Responsible people can
manually update the version easily though in such cases.
The version.mk file is now only used by the VC project files. Once they
are updated to use the package_version file too, then it can be deleted
from the repository.
This avoids a warning about obsolete AM_CONFIG_HEADERS() on more
recent autoconf. The new macro has been around at least since
autoconf 2.60. So this should be safe.
Library soname versioning is something we normally change
only right before each release. However, with 1.0.2 being
released from the 1.0.x branch with 3.0.3, it makes sense
to make the same update here. That way when we do a new
release from master we can just increment again and have
the version properly reflect ranges relative to the stable
branch.
Passing 'enable' as an argument to AM_MAINTAINER_MODE flips the
default to enabled, rather than disabled until automake 1.11.
This is a safer default for a developer-oriented library.
./configure --disable-maintainer-mode is still available for
packagers who what to preserve the upstream build distribution.
On earlier automake versions, the argument appears to have no
effect, so there is no behaviour change for developers using
older autotools.
Based on a patch for opusfile by ron@debian.org.
Doing it this way means it's only the weird corner case that actually
pays for being weird, and we don't litter the build dir with an extra
link that normal builds really don't need at all.
It also avoids the problem of platforms where LN_S isn't well defined.
Mostly this is for people building other things using an uninstalled
opus tree so that the opus-uninstalled.pc will return a working -I
with --cflags when PKG_CONFIG_PATH is set to point to the build dir.
The version of AC_OUTPUT that takes parameters has long been
deprecated now, and replaced by AC_CONFIG_* macros, so pass them
with AC_CONFIG_FILES, since we're now using AC_CONFIG_COMMANDS
to do the above.
The floating-point build calls a number of math library
functions, and linking with libm is technically necessary.
It wasn't obvious because most systems support shlib
dependencies and pulled it in that way, or supply the
referenced functions with intrinsics. Discovered the issue
trying to build libopusfile against the uninstalled static
libopus, which unlike opus-tools, doesn't itself need libm.
The -lm argument is only added to Libs for the floating-
point (default) build. It's not necessary for the fixed-
point build.
Also mark which build was used in the .pc file description.
This isn't used anywhere in the code, floating point is just the
default unless FIXED_POINT is defined.
In the speex codebase, arch.h has a check that both FIXED_POINT
and FLOATING_POINT aren't defined simultaneously, in part as a
check that the build system was constructed with thought about
these and other defines. However, we don't have such a check
and to me it seems unnecessary code.
On some systems (HPPA+HPUX+GCC) -fstatck-protector was causing failures not
at build or link time but at actual runtime. This is much less reasonable to
detect from autotools. It looks this this really can only safely be a white-
list, and the systems which would be whitelisted often already pick up the
setting from the OS build environment in any case. It isn't important for
OPUS, we were just using it as belt-and-suspenders security and because it
makes some failure types easier to troubleshoot.
I'm not sure how this worked before, the the previous version
string fiddling commit ended up not defining OPUS_VERSION in
the autoconf build, so opus_get_version_string() returned
'unknown'.
Previously we defined the release version string in configure.ac,
and overrode that with 'git describe --tags' if possible. This
made it difficult for non-autoconf builds to set their version
string correctly.
Instead we create, and check into version control, a file called
version.mk which defines OPUS_VERSION. The configure script reads
that file and uses it as a fallback if the git revision isn't available.
The expectation is that version.mk will be manually updated for
releases, just as the previous configure.ac version was. However,
since this is a simpler format, it is easier for alternate build
systems to use, reducing the number of places which must be updated.
Also removes the OPUS_MINOR_VERSION, etc. defines from config.h.
As best we recall, this was used to version the
library filename during rapid development. It's
no longer needed now that the bitstream is frozen
and isn't hooked up to anything in the build system.
The config.h version defines were declared in the
middle of the math feature testing. This commit moves
it to the top of configure.ac where the other version
number code resides.
The glibc 2.14 NEWS file says __malloc_hook will be removed
in the next release, so future-proof our use by checking
for this symbol at configure time and only compiling the
malloc failure tests if it is present.
Doxygen is a tool for generating programming documentation
based on comments in header and source files. This commit
adds the necessary configuration file and associated support
in the autotools build.
Right now it doesn't generate much documentation because our
public header files aren't marked up. Warnings are printed
for undocumented members and arguments.
With subdir-objects, both the top-level makefile, and
libcelt/Makefile, which runs celt-specific unit tests,
were sharing libcelt/.deps for generated dependencies.
Each thought it owned the directory, and so would remove
the files it created in that directory, followed by the
directory itself. The second makefile would then fail to
because the files it created weren't there to remove,
blocking a successful 'make distcheck' with in runs
'make distclean' as part of its verification tests.
We can work around the problem by generating the makefile
for the tests in the actual test directory, so it doesn't
share autotools scratch space with the top level makefile.
