Hello, I have patched and compiled linuxsampler for Raspberry Pi 4 4Gb with arm64 PiOS (Debian) running a 6.1 Realtime patched Kernel. In every respect this system is running very well with only one issue. Whenever it exceeds 128 (or thereabouts) Voices the audio starts to include noise like tipping rice grains into a metal bowl. The noise persists and gets worse as the voice count increases until the voice count drops below ~120 the noise ceases and linuxsampler continues normally. No voice stealing is occurring during this
Subscribing to all the various statistics produces no errors during thes sequence of events. I have a single SFZ instrument only, 'SET VOICES 256' and 'SET STREAMS 300' plus two filter chains each with one ladspa filter. CPU load and memory usage is very low (vmstat) even when exceeding the critical voice count so is unlikely to be the issue. Overclocking makes no improvement. My Audio output is: CREATE AUDIO_OUTPUT_DEVICE ALSA ACTIVE='true' CARD='UR22mkII' CHANNELS='2' FRAGMENTS='2' FRAGMENTSIZE='128' SAMPLERATE='48000' Changing 'fragments' to higher values only adds latency again no change to the behaviour. I am aware of the compile options: enable-refill-streams, enable-stream-min-refill, enable-stream-max-refill and enable-stream-size. Perhaps someone can clarify what is happening, I'd rather not recompile 'in the dark'. Yes I am reviewing the code but early days. BTW I hit this voice limit playing the Salamander Piano on minimally complex passages, note that that piano has note off and key off samples triggered in addition to the regular note on samples. It plays beautifully BTW until that ~128 or more voice count. Hopefully someone can give me some direction. TIA Doug
_______________________________________________ Linuxsampler-devel mailing list Linuxsampler-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxsampler-devel
