FS4 Request: Pack Stereo

[wlangfor@uoguelph.ca]

It's nice to be able to choose between stereo and 4 channel packing.

An official primitive would be a good idea. The asm versions have s few bugs depending on usage and application.

[aronb]

I think it should be 6 channels for 5.1, or even (better) 8 channels (7.1)

That way, we can record musical parts from Flowstone (or executable file), to be assembled later in a DAW of our / users choice.

I use typically a minimum of 6 channels, with the 2 main channels prerecorded.

Just my 2 cents

Aron

[wlangfor@uoguelph.ca]

You make a good point, both would be better. Or 2,4,6 and 8 as you say.

I think it should be 6 channels for 5.1, or even (better) 8 channels (7.1)

That way, we can record musical parts from Flowstone (or executable file), to be assembled later in a DAW of our / users choice.

I use typically a minimum of 6 channels, with the 2 main channels prerecorded.

Just my 2 cents

Aron

[adamszabo]

The asm versions have s few bugs depending on usage and application.

What bugs do the asm packing have? Can you post an example?

[wlangfor@uoguelph.ca]

The bug is when using as parent depack and pack (stereo fast pack) it interferes with stereo width correlation.
It completely supersedes it in some cases. I've used it though with success even in eq coef's, but I wouldn't use it in the tense of treating the entire signal, only in small batches.

I suppose I could rip apart the fast pack ams though and reverse engineer that to only output two channels. But the out uses a mono4 from packed signal as a shortcut. So, I guess technically I'd only have to make one.

[adamszabo]

I still dont understand, post a simple test project so we can see how it interferes with stereo width

[MichaelBenjamin]

.

[wlangfor@uoguelph.ca]

the 4pack thing in sm is SSE2, at 4*32bit width.
so it is a given by Hardware, since these operations are resolved at chip level.
the only trickery i remember that could confuse the channels is shuffling.
for example there was some filter that calculates 2*32bit stereo on lane 1&2, then shuffles the result to the lanes 3&4 and feedsback this into the filter again, to have a double as steep stereo filter with a one sample delay at cost of one sse2 operation for two passes.

maybe look if there is a shufps command in the asm and check if it is in original state, maybe you edited it per accident.

edit: or aswell simple loading, because i had a error once on manual loading 4pack, since the sse2 intrinsics are mm_load_ps(3,2,1,0), and not (0,1,2,3), and i spent some time finding that out.

Thanks MB

I still dont understand, post a simple test project so we can see how it interferes with stereo width

will do in a few days, waiting for internet access, browsing through a firewall for the time being.