Re: [hatari-devel] Falcon tests (audio regression)

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

On 1.6.2022 23.18, Nicolas Pomarède wrote:
> Le 01/06/2022 à 21:08, Laurent Sallafranque a écrit :
> > I've encountered something bad after the first demo I tested : 4musiK 
> > by ... me ;)
> >
> > The sound is really bad.

It's really, really bad also in Flextracker.  It sound more like noise 
generator than a music / module player.

This means that I cannot check Falcon audio programs for the release 
before this regression gets fixed. :-/


	- Eero

> > I've bissected hatari and it seems to be related to :
> >
> > 650979aa06c0ba9e4ef6f7978ada0dce1c687594 is the first bad commit
> > commit 650979aa06c0ba9e4ef6f7978ada0dce1c687594
> > Author: Nicolas Pomarede <npomarede@xxxxxxxxxxxx>
> > Date:   Wed Mar 9 15:02:44 2022 +0100
> >
> >      Use Cycles_GetClockCounterImmediate() instead of 
> > CyclesGlobalClockCounter to handle MFP updates
> >      This will take into account the current cycles already processed 
> > when MFP is updated
> >      during the processing of an instruction in Cycle Exact mode
> >
> >   src/cycInt.c          | 29 +++++++++++++++--------------
> >   src/cycles.c          | 17 +++++++++++++++++
> >   src/includes/cycles.h |  1 +
> >   src/m68000.c          |  3 ++-
> >   src/mfp.c             |  9 +++++++++
> >   5 files changed, 44 insertions(+), 15 deletions(-)
> >
> >
> > Any idea ?
>
> thanks for bissecting, I will have a look, maybe some clocks ratios are 
> not correct after those changes I made.