Snorm range max

[lasalvavida]

Adds the ability to vary the range size for SNORM and oct-encoding values for #100.

@pjcozzi, can you look at this when you get a chance?

You can use this to get oct24P and oct32P.

[pjcozzi]

For CesiumGS/3d-tiles#100

[pjcozzi]

result parameters should come last so we need to introduce a new function (and we can implement this function using that new one; no need to deprecate this).

[pjcozzi]

Same comment.

[pjcozzi]

@bagnell can you take a quick look at the math and the tests?

[lasalvavida]

@pjcozzi Updated

[bagnell]

The code looks good to me. The only comment I have is do we want to use an arbitrary range? Shouldn't it take the number of bits?

[pjcozzi]

Did you remember to commit CHANGES.md? I don't see it.

[lasalvavida]

Did you remember to commit CHANGES.md? I don't see it.

You're right, I missed it. It has been added.

[lasalvavida]

The code looks good to me. The only comment I have is do we want to use an arbitrary range? Shouldn't it take the number of bits?

I wanted to avoid calling Math.pow; I would be fine with it either way though.

[bagnell]

I'm fine with it either way as well, but if we keep it this way, it should be mentioned in the doc.

[lasalvavida]

Could you be more specific? Is there a part of the doc that is unclear?

[bagnell]

rangeMax isn't just the maximum value. It is supposed to be a power of 2 and determines how many bits the result is stored in. Doesn't the doc say an oct encoded normal is stored in 2 bytes? If not, it should. There's other functions in this class that assume that's true.

[lasalvavida]

The behavior of octEncode and octDecode is unchanged - it is a specific case of octEncodeInRange and octDecodeInRange where rangeMax=255, and the doc for those functions still says 2 bytes.

octEncodeInRange and octDecodeInRange take an arbitrary range currently. Should they throw an exception if rangeMax isn't a power of two? I'm fine with doing that, I'm just not sure if it adds any value.

[bagnell]

octEncodeInRange and octDecodeInRange take an arbitrary range currently. Should they throw an exception if rangeMax isn't a power of two?

Yes, it can be an arbitrary range but it determines how many bits the number can be stored in. No need to throw an exception. Add something like used to store in log2(rangemax) bits to the doc.

[lasalvavida]

Updated

[lasalvavida]

Can I get an update on this? I'm going to need this to finish i3dm.

[bagnell]

Looks good to me. @pjcozzi do you want to take another look?

[pjcozzi]

Thanks @lasalvavida.

@lilleyse can you merge master into 3d-tiles to make @lasalvavida next merge clean?

[lilleyse]

@lilleyse can you merge master into 3d-tiles to make @lasalvavida next merge clean?

Ok

[lilleyse]

Ready now.