this is not yet the official announcement, but version 1.15.4 of mpg123
is in preparation (being uploaded to sf.net). It contains some
notorious bug fixes that I want out of the way before bringing out the
glorious 1.16.0 (Taihei: should I merge the AVX branch to trunk?). This
note has the main purpose to stir up the binary builders so that they
have a chance to have things ready when the official announcement is
The audio output fixup will be part of the next release (looking at
you, OS/2 folks).
> Taihei: should I merge the AVX branch to trunk?
It is OK to merge the branch, but we have to consider for which processors we should enable dct36 asm optimizations (including existing 3dnow code for K6/K7) before the 1.16.0 release.
And testing on windows 64bit environment is still required.
Am Tue, 21 May 2013 00:01:20 +0900
schrieb Taihei Momma <tmkk@...>:
> but we have to consider for which processors we should enable dct36 asm optimizations (including existing 3dnow code for K6/K7) before the 1.16.0 release.
Yes, I am aware of the K6 issue. I have the K6-III+ still standing on my desk. I thought about simple check for CPU family below K7 to enable 3DNow dct36. I need to get time for this, but I'll then also run it on an older K7 (T-bird, or some Duron) to compare. I might even have a K6-2 CPU lying around; but I'm not sure if I'll invest the time to replace the K6-III with it just for this test. A basic question could be: disable if AMD K7 detected or enable if AMD K6 detected? I'm thinking about Winchips, Cyrix, VIA C3.
I'm leaning towards disabling dct36 specifically if an AMD CPU with newer family is detected, considering that the FPU of those other CPUs is also not that stellar (or am I wrong for VIA?). Doesn't make sense to waste too much time on this decision.
The main question for me is how far we are going with static decoder configurations. Since we are messing with function pointers in any case for configuring decoders for the differing MPEG variants, we might just as well drop non-OPT_MULTI builds. Then one could focus on supporting a personal selection of decoders to include in a generic multi build.
> And testing on windows 64bit environment is still required.
Hear, hear! Windows folks, your turn!
I merged things to trunk now; also preparing a current http://mpg123.org/snapshot (20130520173221) for convenience.