Functions branch: Add more Float Math ops
ClosedPublic
Actions

Authored by artem ivanov (ixd) on Aug 7 2019, 7:41 PM.

Details

Reviewers

Jacques Lucke (JacquesLucke)

Summary

Added:

Subtract
Divide
Power
Log
Square Root
Absolute
Modulo
Cosine
Tangent
Arcsine
Arccosine
Arctangent
Arctan2
Fract
Ceil
Floor
Round
Snap

Notes:
None in EnumProperty is a separator.
May corrupt .blend files that used Float Math node with Multiply, Min, Max, Sin (Enum values are changed)

Diff Detail

Repository: rB Blender

Event Timeline

artem ivanov (ixd) created this revision.Aug 7 2019, 7:41 PM

artem ivanov (ixd) retitled this revision from Functions branch: Add more Float Math nodes to Functions branch: Add more Float Math ops.

We need to create tasks so we don’t end up working on the same thing

Yeah right 😅

I’m working on the vector math node

Jacques Lucke (JacquesLucke) requested changes to this revision.Aug 7 2019, 8:00 PM

Jacques Lucke (JacquesLucke) added inline comments.

source/blender/functions/functions/scalar_math.cpp
142	what is `intpt`? Please use a more descriptive name.
166	I think in this case you can also just return zero. Don't try to be clever.
187	how are negative numbers handled?
383	I think we should assert `isfinite` instead of having a condition. Same above.

This revision now requires changes to proceed.Aug 7 2019, 8:00 PM

artem ivanov (ixd) added inline comments.Aug 7 2019, 9:30 PM

source/blender/functions/functions/scalar_math.cpp
187	`sasqrtf` takes care of it. source/blender/blenlib/intern/math_base_inline.c:159 MINLINE float sasqrtf(float fac) { if (UNLIKELY(fac <= 0.0f)) { return 0.0f; } else { return sqrtf(fac); } }

artem ivanov (ixd) added inline comments.Aug 8 2019, 6:55 PM

source/blender/functions/functions/scalar_math.cpp
383	`isfinite` is pretty cheap function (something like `fabsf(value) <= FLT_MAX`). Inf can happen when user is tweaking nodes (e.g. pow -> multiply -> pow chain can produce such values fairly easy for 32-bit floats). So IMHO no need to crash in such a case.