| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| |
|
|
|
|
|
|
| |
This should improve accuracy of sinusoid extraction in case when
frequency in subband placed just before Nyquist frequency.
Note: we still unable to handle situation with low frequency in
subband which reuired much more samples for analysis.
|
| |
|
|
|
| |
This version doesn't use variable length arrays.
It it better from compiler compatibility perspective and slightly faster.
|
| |
|
|
|
|
|
|
|
|
| |
sin/cos computation with double precision inside libgha cause
significant cpu usage. Actualy we do not need such presition
due to discrete values in the bitstream.
New libgha version also does not perform multiple float to double
conversion in the hot loop (although it can be not good from cache
locality perspective). But it still gives speed up on modern CPUs.
|
| | |
|
| |
|
|
|
|
|
| |
* Multidimensional gha, rework residual check
* Possibility to look into the next frame during GHA
* Possibility to pass envelope into bitstream
* Tool to create oma file from tsv gha description (to test envelope processing)
|
| | |
|
|
|
Why we start with GHA:
* GHA is one of the most computation intencive part of the encoder.
Probably we need to implement multuthreading processing here
to achive good perfomance. If so it is better to make MT frendly design
at the begining of development.
* GHA is the most interesting part of AT3P.
* GHA is the one of the first processing step.
* We already have libgha (https://github.com/dcherednik/libgha)
library which should be tested and used in real application )))
|
