sound: Remove FEEDEQ_CLAMP()

[christos]

Summary

There is PCM_CLAMP_S32() already.

No functional change intended.

Sponsored by: The FreeBSD Foundation
MFC after: 1 week

[dev_submerge.ch]

Hm, PCM_CLAMP_S32 does something different when the SND_PCM_64 option is not set?

[christos]

In D48033#1095717, @dev_submerge.ch wrote:

Hm, PCM_CLAMP_S32 does something different when the SND_PCM_64 option is not set?

It does, but I am wondering whether the non-SND_PCM_64 is useful. I am not exactly sure why it does what it does.

[dev_submerge.ch]

In D48033#1096266, @christos wrote:

In D48033#1095717, @dev_submerge.ch wrote:

Hm, PCM_CLAMP_S32 does something different when the SND_PCM_64 option is not set?

It does, but I am wondering whether the non-SND_PCM_64 is useful. I am not exactly sure why it does what it does.

From sys/conf/NOTES:

# SND_PCM_64                   (Only applicable for i386/32bit arch)
#                              Process 32bit samples through 64bit
#                              integer/arithmetic. Slight increase of dynamic
#                              range at a cost of possible slowdown.

I think the default on 32bit platforms is to process 32bit samples at only 24bit precision, to prevent overflow while using 32bit integers for performance reasons. Option SND_PCM_64 turns on proper 64bit integer processing, which can handle 32bit samples easily. See also PCM_FXSHIFT and the comment you deleted in D47932.

What is really questionable from a readability POV is typedef int64_t intpcm32_t;, IMHO. We have to be extra careful refactoring this stuff.