Brian Whitman

looking into this regomodo-- one thing, your FP code for "face to face" and a few others are empty. do you know this? ``` {"metadata":{"artist":"Daft Punk", "release":"Discovery", "title":"Superheroes", "genre":"House", "bitrate":256,"sample_rate":44100,...

also, i don't like this: ``` "codegen_time":10.096123, "decode_time":19.019990 ``` what kind of machine are you running on? there's no way a full file can take that long unless something else...

thanks. so i think there are two separate issues here: one is the failure to match of known DB tracks for your query and the ohter is the long codegen...

But in practice codegenning? (mobile) On Jul 11, 2011, at 1:47 PM, [email protected] wrote: > MurmurHashNeutral2 is twice as slow as the little-endian only version. We should only use it...

The issue was my check for how many ops were alive to compute the amplitude but during note offs, the ops were set to 0.. but shouldn't they still be...

(also, super weird that local WASM doesn't crash on nans but remote does -- a security thing?)

```(cd .libs && rm -f libloris.la && ln -s ../libloris.la libloris.la) test -z "/usr/local/lib" || ../config/install-sh -c -d "/usr/local/lib" /bin/sh ../libtool --mode=install /usr/bin/install -c libloris.la '/usr/local/lib' /usr/bin/install -c .libs/libloris.13.0.0.dylib /usr/local/lib/libloris.13.0.0.dylib...

@dpwe also noticed the loris package fails with python subversions with more than one character (e.g. python3.11 is read as python3.1).

@davay42 @octetta check out the latest AudioWorklet example, in the `docs` folder here and at https://shorepine.github.com/amy -- works great for me on Chrome & Safari. Not yet on iOS, but...

One big thing people are running into is the `amp_coefs`, vs `amp`, `freq_coefs` vs `freq`, etc. In python we do `amp=X` or `amp=X,Y,Z` and it doesn't translate into C events.