[Bug] CMake build adds sse flags for all source files, not only for SSE-optimized ones #154
Comments
|
let me try to fix that |
|
this has been merged to master |
|
@xnorpx thank you so much! Do you know when a new version with this fix will be released? |
|
That I don't know sorry would have to consume master for now. |
|
Closing as fixed. Thanks for pushing things upstream! |
|
Thanks for fixing! |
daleglass
added a commit
to daleglass/vircadia
that referenced
this issue
Jun 7, 2020
This will build the very latest Opus available (May 30 2020). The absolute minimum version of Opus is commit 927de8453c502586c03e25c169ec08f2a93ebc02, released on April 21, 2020. This incorporates a fix for xiph/opus#154
This was referenced Oct 14, 2022
sumitkandpal
added a commit
to airtimemedia/opus
that referenced
this issue
Jun 26, 2023
* Avoid processing LPC coeffs beyond the given order in NEON optimizations * Fix NEON optimizations buffer read overrun Thanks to Ray Essick * Don't update null data pointer after each multistream decoder The data pointer could be null in the case of DTX or packet loss. * OPUS_GET_IN_DTX handles Silk middle channel only Signed-off-by: Felicia Lim <flim@google.com> * CMake: add option to set BUILD_SHARED_LIBS variable From https://cmake.org/cmake/help/latest/variable/BUILD_SHARED_LIBS.html: "Global flag to cause add_library() to create shared libraries if on. If present and true, this will cause all libraries to be built shared unless the library was explicitly added as a static library. This variable is often added to projects as an option() so that each user of a project can decide if they want to build the project using shared or static libraries." Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * CMake: Add shared library to features Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * CMake: Add alias Opus::opus for opus library. Useful for CMake superbuild pattern. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * CMake: use PACKAGE_VERSION for the pkg-config file The old variable `OPUS_LIBRARY_VERSION`, does not line up with what's used in autoconf, which uses the PACKAGE_VERSION instead, which in turn lines up with what version the projects `opusfile` and `libsndfile` check for, for example. While at it, I also cleaned up the accidental double up in code. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * Revert "Attenuate SILK PLC gain only for unvoiced speech" This reverts commit 4f5557c. Signed-off-by: Felicia Lim <flim@google.com> * Fixes to the the activity flag that is passed to Silk so it represents the final activity flag used in the DTX decision This flag was modified after calling the Silk encoder function. This commit corrects that behavior by introducing those modifications before calling the Silk encoder. Slightly modified comments by Felicia Lim Signed-off-by: Felicia Lim <flim@google.com> * Revert "Fixes to the the activity flag that is passed to Silk so it represents the final activity flag used in the DTX decision" This reverts commit ea3b30f. * Reland "Fixes to the the activity flag that is passed to Silk so it represents the final activity flag used in the DTX decision" This flag was modified after calling the Silk encoder function. This commit corrects that behavior by introducing those modifications before calling the Silk encoder. Slightly modified comments by Felicia Lim Signed-off-by: Felicia Lim <flim@google.com> * Fix for an assertion when running the fixed point tests Signed-off-by: Felicia Lim <flim@google.com> * Fix signed integer overflows in silk_noise_shape_quantizer_del_dec * Fix another signed integer overflow in silk_noise_shape_quantizer_del_dec * Fix celt decoder assertion when using OPUS_CUSTOM When using OPUS_CUSTOM, `CELTDecoder->end` can be larger than 21. Assert against 25 instead in OPUS_CUSTOM builds. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * cmake - make package version parsing more robust Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * cmake - add CPack and default to TGZ package Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * cmake - Add support for detecting the presence of lrint and lrintf. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * cmake - Fix CMake install include directory Install include directory must be `include/opus`, not `include`. Old path is still here for compatibility. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * cmake - add math library for test programs demo and compare when building dynamic library Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * cmake - intinsics fixes, only compile SSE source with SSE flags on non-windows when target use runtime check (GH xiph#154). For windows we only use /arch flag when target is presumed to support SSE to avoid AVX function pollution (GH xiph#132). Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * cmake - Add option for testing to improve cli Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * cmake - Add OPUS prefix to feature info to show correct commandline option Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * cmake - Fix OPUS_STACK_PROTECTOR option Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * CMake Changes - Fix typo in OPUS_USE_NEON description. - Set OPUS_PRESUME_NEON for iOS platforms as all armv7 and higher iOS devices support NEON. - Fix detection of aarch64 for OPUS_CPU_ARM and adding sources from celt_sources_arm. (previously would miss armcpu.c and arm_celt_map.c) - Change "armv7-a" to "arm" in MATCHES checks against CMAKE_SYSTEM_PROCESSOR as systems like the RPi3 report as "armv7l". - Rename OPUS_MAY_SUPPORT_NEON to OPUS_MAY_HAVE_NEON as this name is used everywhere else in the CMake build system. Without this, runtime capability detection is broken on aarch64. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * cmake - Add variable length detection and alloca detection Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * cmake - add option for float_approx for IEEE 754 compatible targets Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * cmake - add option for fast math Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * Fix a typo in in opus_custom.h. Signed-off-by: Ralph Giles <giles@thaumas.net> * gitlab-ci: Add a build description. Describe builds for the gitlab continuous integration service runners. This does a trial build under both autotools and cmake, so we get some coverage for changes on that hosting platform. After the same script in the vorbis and ogg projects. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * Repository moved to gitlab.xiph.org * fix equivalent bitrate calculation for <20ms frame sizes Signed-off-by: Hector Martin <marcan@marcan.st> Signed-off-by: Mark Harris <mark.hsj@gmail.com> * trivial_example: open raw pcm files in binary mode. The simple codec round-trip example file in the doc directory opens an input and output pcm file. It was working fine on POSIX systems, but not on Windows, which treats text files differently. This is confusing in a example, so it's better to add an explicit binary flag to the fopen() calls. This does nothing on unix-like systems, but should make the example work for developers on Windows. Thanks to Wavesonics who reported this on irc. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * trivial_example: Check the return value of fread(). Silence a gcc warning by checking the return value of the fread() call instead of the feof() guard. This prevents an infinite loop in the case of a read error. Otherwise, when end-of-file is reached fread() will certainly return a smaller number of elements read than requested, so both cases are handled now. Add a comment to clarify that we're dropping a partial frame on purpose to keep the code simple. Also add more braces around conditional bodies for less error-prone style. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * Build trivial_example by default. Add doc/trivial_example.c to the autotools build so we get some minimal verification that the code compiles. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * Update Doxygen config file, header and footer to Doxygen 1.8.18 Signed-off-by: Ralph Giles <giles@thaumas.net> * Build and run test for cmake build in gitlab-ci Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * Disable message box when calling abort(). The message box is causing hangs in tests. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * Build time improvement, for MSVC use intrin0.h instead of intrin.h and remove usage of stdio.h in production code Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * Fix intrin0.h include guard. This lighter-weight intrinsics header is available starting with Microsoft Visual Studio 2017, so the previous change to allow this header failed when building with Visual Studio 2015. Restores the appveyor continuous integration build. * Add support to CMake build for building frameworks on Apple systems. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * cmake - add headers to project Signed-off-by: Mark Harris <mark.hsj@gmail.com> * cmake - only publish opus_custom.h if custom modes is enabled xiph#121 Signed-off-by: Mark Harris <mark.hsj@gmail.com> * cmake - move cmake related files to cmake folder to make root dir less busy Signed-off-by: Mark Harris <mark.hsj@gmail.com> * cmake - add include guards to cmake files Signed-off-by: Mark Harris <mark.hsj@gmail.com> * cmake - move all compiler feature detection to OpusConfig Signed-off-by: Mark Harris <mark.hsj@gmail.com> * cmake - add option to disable intrinsics Signed-off-by: Mark Harris <mark.hsj@gmail.com> * cmake/CFeatureCheck.cmake: fix feature tests failing when Opus is a submodule CMAKE_SOURCE_DIR corresponds to the top project's source directory. CMAKE_BINARY_DIR corresponds to the top project's binary directory. The usage of these variables doesn't cause any problems when Opus is built as a standalone project. This is not the case when Opus is added as submodule: the variables are set by the project that calls "add_subdirectory()". The fix consists in using PROJECT_SOURCE_DIR and PROJECT_BINARY_DIR, which always refer to the current project. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * cmake - refactor and cleanup options Signed-off-by: Mark Harris <mark.hsj@gmail.com> * opus draft: fix run failed under dash [[ ]], the compound command is not supported by all shell interpreter. [ ], the buildin command is more common. Signed-off-by: Ralph Giles <giles@thaumas.net> * fix doc/build_draft.sh run error Add newer source directories to the destdir file tree so we can include all sources referenced from opus_sources.mk. Signed-off-by: Ralph Giles <giles@thaumas.net> * Silence clang silk_encoder alignment warning * Prefer SSE and ASM implementation of float2int before lrintf for MSVC Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * Fix trailing whitspace in previous commit Signed-off-by: Ralph Giles <giles@thaumas.net> * Fix arm build with rtcd enabled. The autotools build doesn't set OPUS_HAVE_RTCD for arm targets, assuming all the supported intrinsics will work on the runtime cpu. The cmake build however defines this by default when the neon extension is available on the target. On Linux, the runtime cpu detection reads /proc/cpuinfo, so removing stdio.h from celt/os_support.h meant that the cmake build for arm targets failed. We don't currently have ci runs for that configuration, so this only became apparent through manual testing. Signed-off-by: Marcus Asteborg <maastebo@microsoft.com> Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * Add CI to check for trailing whitespaces in merge requests * Remove trailing whitespaces in vla.c that was missed previously * Fix and clean up opus_decode_fuzzer Use the fuzzed sub-length of the input data instead of the whole input. * configure: adjust x86 get cpu info inline assembly method for PIC case .. just like the way it is done in celt/x86/x86cpu.c. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * gitlab-ci: Run jobs on the gcc:9 image. Use a versioned gcc container image for more consistent test results. This is the same version we're using for other projects, but of course it will need to be bumped periodically. The current gcc release is 10.2. The oldest supported release is 8.4, so this is in the middle of the support window. Signed-off-by: Marvin Scholz <epirat07@gmail.com> * cmake - add warning flags for clang, gcc etc. * cmake - Fix typo bug in enable_float_api option * cmake - fix bugs around consuming Opus as a submodule in cmake and package version parsing * CMake: Make _FORTIFY_SOURCE optional * cmake - MINGW check for ssp lib and link if security features is enabled * opus_decoder_fuzzer: limit the number of decodes to avoid timeout * silk: Fix incorrect ifdef in debug.c Signed-off-by: Mark Harris <mark.hsj@gmail.com> * celt: Fix broken SSE pre-processor check due to typo This broke 5 years ago in 43120f0 Signed-off-by: Mark Harris <mark.hsj@gmail.com> * Replace WIN32 with _WIN32 everywhere _WIN32 is defined on all Windows platforms by every compiler that targets Windows. We do not need WIN32 at all. Signed-off-by: Mark Harris <mark.hsj@gmail.com> Resolves xiph#104 * repacketizer_demo: check for read errors to fix compiler warnings Actually check for read errors instead of just storing the return value in a variable that then never gets checked. Also fixes "conversion from 'size_t' to 'int', possible loss of data" compiler warnings on Windows with MSVC caused by storing the size_t returned by fread() into an int variable. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * Fix MSVC warning about trunction from double to float Specify the precision as float to avoid truncating from double. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * Add support for Meson build system Tested on: - Linux/x86* with gcc - Android armv7 arm64 x86 x86_64 with clang - Windows x86 x86_64 with Visual Studio 2017 - Windows x86 x86_64 with MinGW - macOS x86_64 with clang - iOS arm64 x86_64 with clang Co-authored by: Nirbheek Chauhan <nirbheek@centricular.com> https://gitlab.xiph.org/xiph/opus/-/merge_requests/13 * cmake - change logging of api version to debug Signed-off-by: Ralph Giles <giles@thaumas.net> * CMakeLists.txt: specify path to target file in add_test() directives This is required in case the output path for tests is changed by a project adding Opus as submodule. Signed-off-by: Ralph Giles <giles@thaumas.net> * CMakeLists.txt: specify working directory in add_test() directives This is required for Windows because it doesn't have RPATH, thus it fails to find Opus if it's not in the same directory as the executables or in PATH. Signed-off-by: Ralph Giles <giles@thaumas.net> * cmake - add option for assertions Signed-off-by: Ralph Giles <giles@thaumas.net> * cmake - add option for hardening Signed-off-by: Ralph Giles <giles@thaumas.net> * cmake - add option for fuzzing Signed-off-by: Ralph Giles <giles@thaumas.net> * cmake - add option for check asm Signed-off-by: Ralph Giles <giles@thaumas.net> * cmake - add option for fixed point debug Signed-off-by: Ralph Giles <giles@thaumas.net> * Meson: Fix doc build when opus is a subproject meson.source_root() and meson.build_root() have been deprecated in latest Meson release because they are a trap. They point to the root of parent project instead of root of current subproject. Meson 0.56.0 added meson.project_source/build_root() but it is just as easy to use meson.current_source/build_dir() in the root meson.build file and avoids bumping required meson version. Signed-off-by: Ralph Giles <giles@thaumas.net> * docs: fix simple typo, neareast -> nearest There is a small typo in celt/fixed_generic.h. Should read `nearest` rather than `neareast`. Signed-off-by: Ralph Giles <giles@thaumas.net> * ci: fix pipeline run for merge requests This way CI pipeline runs for branches and tags and makes it show up in merge requests where a branch is used as source branch. Makes all jobs show up in merge request CI indicator (not just the whitespace job) and only runs a single CI pipeline, without the additional detached pipeline. https://docs.gitlab.com/ce/ci/yaml/#workflowrules-templates Signed-off-by: Ralph Giles <giles@thaumas.net> * Fix float-approx negative left shift UB Reported by toto. * Sending refresh DTX packets every 400 ms independently of the encoded frame size. Signed-off-by: Felicia Lim <flim@google.com> * celt_lpc: avoid overflows when computing lpcs in fixed point The LPCs are computed in 32-bit, so increase the allowed range from +/-8 to +/-64 to avoid overflows caught during fuzzing. Before downshifting back down to the +/-8 range in the final 16-bit output, perform bandwidth extension to avoid any additional overflow issues. * Fix trailing whitespace. This was introduced in February, and fails the corresponding check in gitlab ci runs. Also indent the subsequent lines to match and correct typos. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * cmake - add support to run ctest on android #2347 Signed-off-by: Ralph Giles <giles@thaumas.net> * Relax comparison to 0 to avoid a floating point divide-by-zero error. * Revert relaxing comparison to 0 for fixed point only * meson: fix get-version script for git worktrees For git worktree directories .git is not a directory but a file that points to the real .git dir. The `update_version` script used by other builds works correctly with git worktrees. Signed-off-by: Ralph Giles <giles@thaumas.net> * Remove an unused parameter * Check channels/stream counts and mapping when creating the multistream encoder * Fix buffer overflow in xcorr_kernel_sse4_1 Before, an overflow can occur in the last loop if `len` is not a multiple of 4 as OP_CVTEPI16_EPI32_M64 tries to load 64 bits, but there are insufficient bits allocated in `x`. * Disable dangerous SSE 4.1 intrinsic optimizations These could result in 16-byte-aligned loads on unaligned data, causing a segfault. * Initialize non-zero test arrays. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * Prevent int32 overflow when applying HARM FIR filter in NSQ.c by using a saturating sum. This matches behavior in NSQ_del_dec.c. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * Update and re-enable SILK SSE4.1 optimisations * build test scripts * Cleanup testing directories to save space * Fix lrint/lrintf detection Prevents using lrint/lrintf when compiling with -std=c90 even though the functions are in libm. This was causing tests to fail, likely due to incorrect prototypes. * print rate used for testvectors * Correct redundancy handling with lost/DTX frames In xiph#253, the encoder generates a Hybrid frame with redundancy, to switch to CELT-only mode, and then activates DTX immediately afterwards. The decoder ran Hybrid PLC, which isn't right. Use CELT PLC instead if there was already a transition to CELT via redundancy at the end of the previous frame. Also do not use a stale CELT decoder to decode a second redundancy frame when the first redundancy frame for a transition from SILK-only mode was lost. Instead of mixing in old audio from the last time that CELT was used, ignore the second redundancy frame in this case. Alternatively the CELT decoder could be reset before decoding, but it would not be ready until after the 2.5 ms of audio that is needed. Reviewed by Jean-Marc Valin. * Improve background noise estimation for CELT DTX We now update the background noise estimate even in frames classified as transient. It shouldn't be a problem because we're using min statistics. Also, it avoids problems when update frames get missclassified as transient. In addition, we now use the duration of losses rather than the number of lost packets to make decisions. That should make PLC/DTX behaviour more consistent across frame sizes. * Fix 8101b33 to decode ignored redundancy Even if the redundancy is ignored, the final range from the decoder is needed for testing. Reviewed by Timothy B. Terriberry. * Fixes wrap-around in silk_inner_prod16_sse4_1() Thanks Tim * Silence some warnings for fixed-point debug builds Reviewed by Timothy B. Terriberry. * Fixes --disable-rtcd Make sure we don't try to use the rtcd table when rtcd is disabled. That code still needs a lot more cleaning up. * Fixes valgrind failure caused by silk_find_pred_coefs_*() The function copies NLSFs from the stack to the state which for order 10 means we were copying uninitialized values. That in turn breaks check-asm when comparing the state under valgrind. Reviewed by Timothy B. Terriberry. * Check the return value of __get_cpuid(). This function can fail if CPUID is not supported or the maximum supported value of EAX is less than the requested one. Check the return value and explicitly disable all SIMD if it does fail. This was happening before implicitly because of the initialization of info[] to zero, but being explicit about it makes it less likely someone will break this behavior because they did not realize what was going on. * Work around a valgrind false-positive in CPUID. Valgrind versions prior to 3.17.0 assume that an uninitialized value in ECX causes the whole output of CPUID to be uninitialized, even though ECX is only "read" by CPUID for certain values of EAX. Work around that by guaranteeing that ECX is initialized. * Adds fuzzing to CPU detection Makes ti possible to randomize (with --enable-fuzzing) the CPU flags so we can better test all the intrinsics implementations. * Estimate the inner product accuracy to fix check-asm Estimate the rounding error so that we can have a useful margin of error when checking the asm against the C code even when the float operations get reordered due to -ffast-math. * Adds OPUS_SET_INBAND_FEC(2) option Unlike OPUS_SET_INBAND_FEC(1), the encoder does not necessarily switch to SILK if we have music. * Add asan/ubsan support in random tests * Work around UBSan unaligned access errors. The underlying objects are all 8-bit integers. Verified that the generated assembly still just uses MOVD. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * Avoid left shifts of negative values in debug macros Reviewed by Mark Harris * Fix some 16-bit overflows (using 32-bit macros) Reviewed by Mark Harris * Fix fixed-point overflow in pitch downsampling Reviewed by Mark Harris * Avoid undefined behaviour within the debug macros Even when the macro itself would overflow. Reviewed by Mark Harris * Avoids incrementing uninitialized values The values were never used, but ubsan + valgrind would complain. Reviewed by Mark Harris * Silence GCC 11+ -Wmaybe-uninitialized warnings Reviewed by Timothy B. Terriberry. * Fix quoting and whitespace errors in build test Reviewed by Jean-Marc Valin. * Fix warning with --disable-rfc8251 * Replace assert with test_failed function in test This will fix -Wunused-but-set-variable on gcc 9.3 release build. Also remove unused assert.h. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * Remove unused variable in tests Signed-off-by: Mark Harris <mark.hsj@gmail.com> * Fix opus.h for doxygen when referencing alternative values Doxygen was not able to resolve the references because it looked for OPUS_APPLICATION_VOIP/@ref. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * doc: Use consistent alternative notation Signed-off-by: Mark Harris <mark.hsj@gmail.com> * cmake - Add OPUS_BUILD to test targets Signed-off-by: Mark Harris <mark.hsj@gmail.com> * Fix warnings when compiling with FUZZING enabled * Update x86 CPU detection configure check. Commit 6577534 switched from using __get_cpuid() to __get_cpuid_count(), but the corresponding configure check was not updated. Since __get_cpuid_count() was introduced much later, make sure we check for the function we actually use. Thanks to Mark Harris for the report. * meson: Fix reporting of cpu family if intrinsics not supported Signed-off-by: Doug Nazar <nazard@nazar.ca> * Fix uninitialized field on custom mode malloc fail * cmake - fix lrintf, lrint detection This commit addresses the issues of not finding lrintf and lrint. We switch to check_symbol_exists instead per cmake documentation. Also make sure to link math lib for detection for nix. For MSVC the issue for non x86 builds was that the standard was set to default which is 199409L. This resulted in not using lrintf even that it was found. To address this we set the C standard to C11 and it will only apply to newer versions of MSVC where the /std flag is supported. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * cmake - fix rtcd detection on x86 non windows Signed-off-by: Mark Harris <mark.hsj@gmail.com> * cmake - move warning C4244 to level 4 Opus compare is used to generate test vectors so no cosmetic changes is taken. Hence we move this warning to level 4 for opus compare. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * update doc on custom mode Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * Fix C90-related warnings * Make silk/x86 header indentation consistent. The indentation for nested #ifs was all over the place. * Don't compile x86 cpu detection without RTCD. Also #error if RTCD is enabled without a detection method, like Arm. A number of SILK functions also still used the lookup tables, even when RTCD was disabled. Fix those, too. * Only build platform RTCD sources when enabled. To avoid issues with empty compilation units. * Silence Clang 13+ null-pointer-subtraction warning * Silence MSVC C4244 warning When building with FLOAT_APPROX. Signed-off-by: Mark Harris <mark.hsj@gmail.com> * Fix NORM_ALIASING_HACK We need to move the history out of the way before we write to the shape array X, or else we get corruption of the audio. Signed-off-by: Jean-Marc Valin <jmvalin@amazon.com> * Relaxing checks for MULT16_32_QX() MULT16_32_QX() is now implemented using a signed-unsigned multiply, so the second argument can now have one extra bit compared to the old signed-signed implementation. Reviewed by Mark Harris * Using saturating round to fix some wrap-arounds Reviewed by Mark Harris * More ubsan fixes for the debug macros themselves Reviewed by Mark Harris * Ensuring we can see where crashes occur Reviewed by Mark Harris * Re-tuning the use of LTP scaling Making LTP scaling depend on the bitrate and whether FEC is on. The thresholds for scaling 1 and 2 are now independent. * More FEC tuning: lowering the LBRR bitrate a bit * Smooth out the LBRR rate estimate Reduces fluctuations in the non-FEC target bitrate. * Change pitch scaling behavior wrt nFramesPerPacket Not sure if it was the original intent, but we now reduce the loss percentage threshold for pitch scaling as 1/nFramesPerPacket since only the first frame will have pitch scaling anyway. As a side effect, this brings back the original behavior of disabling pitch scaling for 0% loss. * Fix typo in MacroDebug.h comment. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * opus.m4: fix -Wstrict-prototypes Signed-off-by: Sam James <sam@gentoo.org> Signed-off-by: Mark Harris <mark.hsj@gmail.com> * Make CELT FFT twiddle complex type aligned This makes kiss_twiddle_cpx 4-byte aligned (instead of 2-byte) for fixed-point builds. Tested with an armv6j+nofp development board, CELT encoding becomes 1.4x as fast, and decoding over 2x. Performance gain is mostly attributed to the proper alignment of the static const array mdct_twiddles960. Co-authored-by: David Gao <davidgao@google.com> Signed-off-by: Felicia Lim <flim@google.com> * Bump LT version Added OPUS_SET_INBAND_FEC(2) since previous version * Avoid "ISO C forbids an empty translation unit" Add dummy typedef to avoid the warning * oops, avoid using a reserved identifier * docs: replace fgrep with grep -F It's been deprecated for decades and in Debian system it's starting to print warnings. Just use grep -F instead. Signed-off-by: Ralph Giles <giles@thaumas.net> * meson: fix build on arm64 Would fail like: Checking if "compiler supports ARMv7/AArch64 NEON intrinsics" : links: NO Checking if "compiler supports ARMv7/AArch64 NEON intrinsics with -mfpu=neon" : links: YES Checking if "compiler supports AArch64 NEON intrinsics" : links: NO Checking if "compiler supports AArch64 NEON intrinsics with -mfpu=neon" : links: NO Message: Compiler does not support AArch64 NEON intrinsics ../silk/meson.build:28:45: ERROR: Unknown variable "have_arm_intrinsics_or_asm". since commit 0808841. * ci: add arm64 CI * ci: add ci-fairy linter to make sure commits are GPG signed * cmake instructions Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * CAR-2144 test some changes * Fix Gitlab CI Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> * CAR-2144 remove old endif * CAR-2144 remove run_config script * Feature/car 2144 fix opus cmake sumit (#6) * CAR-2144 set intall to OFF * CAR-2144 remove opus cpack --------- Signed-off-by: Felicia Lim <flim@google.com> Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> Signed-off-by: Ralph Giles <giles@thaumas.net> Signed-off-by: Mark Harris <mark.hsj@gmail.com> Signed-off-by: Hector Martin <marcan@marcan.st> Signed-off-by: Marcus Asteborg <maastebo@microsoft.com> Signed-off-by: Marvin Scholz <epirat07@gmail.com> Signed-off-by: Doug Nazar <nazard@nazar.ca> Signed-off-by: Jean-Marc Valin <jmvalin@amazon.com> Signed-off-by: Sam James <sam@gentoo.org> Co-authored-by: Felicia Lim <flim@google.com> Co-authored-by: Gustaf Ullberg <gustaf.ullberg@gmail.com> Co-authored-by: Davide Beatrici <davidebeatrici@gmail.com> Co-authored-by: Marcus Asteborg <maastebo@microsoft.com> Co-authored-by: Nathaniel R. Lewis <linux.robotdude@gmail.com> Co-authored-by: DeadSix27 <DeadSix27@users.noreply.github.com> Co-authored-by: Jesús de Vicente Peña <devicentepena@webrtc.org> Co-authored-by: Niclas Olmenius <niclas@voysys.se> Co-authored-by: evpobr <evpobr@gmail.com> Co-authored-by: willson-chen <willson.chenwx@gmail.com> Co-authored-by: Hua, Chunbo <chunbo.hua@intel.com> Co-authored-by: Ralph Giles <giles@thaumas.net> Co-authored-by: Mark Harris <mark.hsj@gmail.com> Co-authored-by: Hector Martin <marcan@marcan.st> Co-authored-by: Simon Jackson <si@sonocent.com> Co-authored-by: Davide Beatrici <git@davidebeatrici.dev> Co-authored-by: sezero <sezero@users.sourceforge.net> Co-authored-by: Zeno Endemann <zeno.endemann@googlemail.com> Co-authored-by: Nirbheek Chauhan <nirbheek@centricular.com> Co-authored-by: Tim-Philipp Müller <tim@centricular.com> Co-authored-by: Xavier Claessens <xavier.claessens@collabora.com> Co-authored-by: Tim Gates <tim.gates@iress.com> Co-authored-by: Jean-Marc Valin <jmvalin@jmvalin.ca> Co-authored-by: Tom Denton <tomdenton@google.com> Co-authored-by: Francis Quiers <fquiers@cisco.com> Co-authored-by: Jean-Marc Valin <jmvalin@amazon.com> Co-authored-by: Timothy B. Terriberry <tterribe@xiph.org> Co-authored-by: Alexander Traud <pabstraud@compuserve.com> Co-authored-by: Timothy B. Terriberry <territim@amazon.com> Co-authored-by: Doug Nazar <nazard@nazar.ca> Co-authored-by: Nathan E. Egge <eggenath@amazon.com> Co-authored-by: Sam James <sam@gentoo.org> Co-authored-by: Zheng Lv <lvzheng@google.com> Co-authored-by: David Gao <davidgao@google.com> Co-authored-by: Claudio Saavedra <csaavedra@igalia.com> Co-authored-by: Marcus Asteborg <xnorpx@outlook.com> Co-authored-by: sumitkandpal <sumit@airtime.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
For example:
opus/opus_functions.cmake
Lines 142 to 155 in f9d3d43
check_and_set_flagsadds the specified flag to compile options if supported:opus/opus_functions.cmake
Lines 91 to 97 in f9d3d43
On the other hand,
Makefile.amadds SSE options only for SSE-optimized source files:opus/Makefile.am
Lines 329 to 344 in ad8fe90
Can we fix that for CMake builds? This may lead to SSE 4.1 compiler optimizations in non-SSE-optimized files and therefore to invalid instruction errors on machines without SSE 4.1 support.
The text was updated successfully, but these errors were encountered: