[tim-janik/beast] Resample accuracy fails (#139)

From: Tim Janik <notifications github com>
To: tim-janik/beast <beast noreply github com>
Cc: Subscribed <subscribed noreply github com>
Subject: [tim-janik/beast] Resample accuracy fails (#139)
Date: Wed, 12 Feb 2020 16:29:31 -0800

Since we enabled -ffast-math, resampler accuracy tests have been failing.
Just build todays Beats master and run the checks:

out/tests/suite1 testresampler_check_precision_sub24
  RUN…     testresampler_check_precision_sub24
suite1: tests/testresampler.cc:810: testresampler_check_precision_sub24: assertion failed: run_accuracy (RES_SUBSAMPLE, false, 24, 90, 9000, 983, 126)

Maybe the error margins need to be extended, now that math precision can potentially be reduced, or maybe some float instructions really need adaptions. In particular -fassociative-math and -freciprocal-math may be the culprits here. See also: https://mail.gnome.org/archives/beast/2020-January/msg00009.html

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or unsubscribe.

Follow-Ups:
- Re: [tim-janik/beast] Resampler accuracy tests fail (#139)
  - From: Tim Janik

[Date Prev][Date Next] [Thread Prev][Thread Next] [Thread Index] [Date Index] [Author Index]