Re: NeuralRack v0.4.1 bug fix release

Am 16.08.26 um 13:09 schrieb Jeremy Jongepier:
> On 8/3/26 5:24 PM, Hermann Meyer wrote:
>> NeuralRack is a Neural Model and Impulse Response File loader for 
>> Linux/ Windows available as Stand alone application, and in the Clap, 
>> LV2, vst2 and vst3 plugin format.
>> It supports [*.nam files](https://www.tone3000.com/search?tags=103) 
>> and, or [*.json or .aidax 
>> files](https://www.tone3000.com/search?tags=23562) by using the 
>> [NeuralAudio](https://github.com/mikeoliphant/NeuralAudio) engine.
>>
>> For Impulse Response File Convolution it use [FFTConvolver](https:// 
>> github.com/HiFi-LoFi/FFTConvolver)
>>
>> Resampling is done by [Libzita-resampler](https:// 
>> kokkinizita.linuxaudio.org/linuxaudio/zita-resampler/resampler.html)
>>
>> New in this release:
>> - fix clap plugin crash in Bitwig
>>
>> Neuralrack allow to load up to two model files and run them serial.
>> The input/output could be controlled separate for each model.
>> For tone sharping a 6 band EQ could be enabled.
>> Additional it allow to load up a separate Impulse Response file for 
>> each output channel (stereo).
>> Neuralrack provide a buffered Mode which introduce a one frame 
>> latency when enabled.
>> It could move one Neural Model, or the complete processing into a 
>> background thread. That will reduce the CPU load when needed.
>> The resulting latency will be reported to the host so that it could 
>> be compensated.
>>
>> Project Page:
>>
>> https://github.com/brummer10/NeuralRack
>>
>> Release Page:
>>
>> https://github.com/brummer10/NeuralRack/releases/tag/v0.4.1
>
> Hello Hermann,
>
> What was the difference between the v3 and non-v3 releases again? 
> Something with CPU architecture?
>
> Jeremy


Hi Jeremy

Yes, it's about CPU micro-architecture optimization. v3 here means 
x86-64-v3 level support. This brings a huge performance boost when 
working with neural models.

https://en.wikipedia.org/wiki/X86-64#Microarchitecture_levels

regards

hermann
//