Age | Commit message (Collapse) | Author | |
---|---|---|---|
2023-12-20 | Merge LACE/NoLACE under OSCE frameworkopus-ng-lace-integration5 | Jan Buethe | |
2023-12-15 | use opus_(re)alloc and opus_free for dnn and DRED related functionsopus-ng-fix-alloc | Michael Klingbeil | |
2023-11-30 | don't redefine _mm_loadu_si32 on MSVC | Michael Klingbeil | |
2023-11-29 | Trying to fix/update meson build | Jean-Marc Valin | |
Still don't quite know what I'm doing | |||
2023-11-28 | Oops, fix the fixed-point build | Jean-Marc Valin | |
2023-11-28 | Fixes for ARMv7/AArch32 | Jean-Marc Valin | |
1) Enable asm/intrinsics even for floating-point 2) Make sure ARMv8 asimd enables EDSP/MEDIA/Neon 3) Add dotp architecture to rtcd table since AArch *can* have dotp | |||
2023-11-21 | Add rtcd for silk_inner_product_FLP() | Jean-Marc Valin | |
2023-11-21 | Start enabling AVX2 silk_inner_product_FLP() | Jean-Marc Valin | |
Not yet with rtcd | |||
2023-11-21 | Avoids AVX2 optimizations being disabled | Jean-Marc Valin | |
2023-11-21 | Misc fixes on previous patch | Jean-Marc Valin | |
Fixes warnings, undefined behaviour, and check-asm failure | |||
2023-11-21 | Optimize NSQ_del_dec() for AVX2 | Victor Ding | |
The optimization is bit-exact with C function. This optimization speeds up SILK encoder (floating point) as following: AMD Zen: Complexity 0-5 : 0% Complexity 6-7 : 3 - 7% Complexity 8-10: 8 - 15% Intel Skylake: Complexity 0-5 : 0% Complexity 6-7 : 14 - 18% Complexity 8-10: 17 - 22% Adapted by Jean-Marc Valin | |||
2023-11-21 | AVX2 version of silk_inner_product_FLP() | Jean-Marc Valin | |
Not hooked up | |||
2023-11-18 | Speed up silk_warped_autocorrelation_FLP() | Jean-Marc Valin | |
Reducing the dependency chain between tmp1 and tmp2 at the cost of an extra multiply. | |||
2023-11-16 | Remove C99 comment | Jean-Marc Valin | |
2023-11-16 | Adding RTCD for compute_activation()exp_dnn_rtcd6 | Jean-Marc Valin | |
2023-11-16 | Adding RTCD for DNN code | Jean-Marc Valin | |
Starting with compute_linear() | |||
2023-11-15 | Using sparse GRUs in DRED decoder | Jean-Marc Valin | |
Saves ~270 kB of weights in the decoder | |||
2023-11-08 | DRED: quantize scale and dead zone to 8 bits | Jean-Marc Valin | |
2023-11-07 | DRED code cleanup | Jean-Marc Valin | |
Removing some indirections | |||
2023-11-07 | Split stats in two and remove useless dimensions | Jean-Marc Valin | |
2023-11-06 | DRED: quantize r and p0 parameters with 8 bits | Jean-Marc Valin | |
Only code non-degenerate symbols, which makes the encoder faster | |||
2023-11-06 | Vectorize DRED quantization | Jean-Marc Valin | |
2023-11-04 | Match silenced overflow checks in the sse4.1 version | Felicia Lim | |
Update silk/x86/NSQ_del_dec_sse4_1.c to match the remaining silk/NSQ_del_dec.c changes made in https://gitlab.xiph.org/xiph/opus/-/commit/c913dc38 Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> | |||
2023-11-03 | Silence some overflow checks | Felicia Lim | |
Co-authored-by: James Zern <jzern@google.com> | |||
2023-10-30 | Don't try to use models that aren't loaded | Jean-Marc Valin | |
2023-10-30 | Don't use reserved identifiers for include guards | Jean-Marc Valin | |
2023-10-30 | Update blob loading code | Jean-Marc Valin | |
2023-10-21 | Cleanup | Jean-Marc Valin | |
2023-10-19 | Support OPUS_SET_COMPLEXITY() on decoder side | Jean-Marc Valin | |
Controls whether deep PLC is enabled | |||
2023-10-18 | Don't call the libm tanh() | Jean-Marc Valin | |
2023-10-18 | Making the build possible without the models | Jean-Marc Valin | |
No dependency on the data files if no DNN code enabled | |||
2023-10-15 | Rename ENABLE_NEURAL_FED to ENABLE_DRED | Jean-Marc Valin | |
Signed-off-by: Jean-Marc Valin <jmvalin@amazon.com> | |||
2023-10-15 | Rename NEURAL_PLC to ENABLE_DEEP_PLC | Jean-Marc Valin | |
Signed-off-by: Jean-Marc Valin <jmvalin@amazon.com> | |||
2023-10-08 | Fix warning from casting between 1D and 2D arrays | Jean-Marc Valin | |
2023-10-08 | Silencing alignment warnings on x86 intrinsics | Jean-Marc Valin | |
Those intrinsics don't actually require alignment so we're OK | |||
2023-10-06 | New model with wider range of bitrates | Jean-Marc Valin | |
Using a max lambda of 0.04 | |||
2023-10-06 | Switching to neural pitch estimator | Jean-Marc Valin | |
Remove old pitch estimator and retrain all models | |||
2023-10-02 | update model | Jean-Marc Valin | |
2023-09-28 | Fix stats indexing for state | Jean-Marc Valin | |
2023-09-26 | RDOVAE model update | Jean-Marc Valin | |
2023-09-24 | Infinite loops are bad | Jean-Marc Valin | |
2023-09-21 | Handle the case where the initial state didn't fit | Jean-Marc Valin | |
2023-09-21 | Quantizing initial state with rdovae too | Jean-Marc Valin | |
More efficient than PVQ | |||
2023-08-10 | Changing DRED exp. ID so we can reserve 127 | Jean-Marc Valin | |
There was a suggestion to use 127 for extending the extension ID space | |||
2023-08-10 | Making it easier to remove DRED experimental ID | Jean-Marc Valin | |
When ready, change DRED_EXTENSION_ID to the final ID, remove DRED_EXPERIMENTAL_VERSION completely, and change DRED_EXPERIMENTAL_BYTES to zero (eventually remove it). | |||
2023-07-28 | Make RDOVAE encoder use LinearLayer directlyopus-ng-linear4 | Jean-Marc Valin | |
2023-07-13 | Make "VBR with cap" less aggressive | Jean-Marc Valin | |
The bits we don't use won't be wasted, so it's less important to get exactly the optimal number of bits below the cap. | |||
2023-07-13 | Some general SILK CBR tuning | Jean-Marc Valin | |
The gain*2 when overshooting was too aggressive and the undershoot case wasn't aggressive enough. This now seems to work reasonably well. | |||
2023-07-12 | Fix DRED/neural PLC for SILK stereo | Jean-Marc Valin | |
Don't attempt to run the neural PLC on the side channel since we only have one state. | |||
2023-07-03 | Properly compute and use the DRED offset field | Jean-Marc Valin | |
Also, don't code DRED that's redundant with the main packet |