Lerp vs fastLerp

Torsten · 2015-12-03T13:56:48

We all know what lerps does and how it works - but there are two ways to implement lerps: lerp = a * (1 - t) + b * t (2 multiplys, one substraction, one addition) fastLerp = a + (b - a) * t (one multiply, one substraction, one addition) In practice, i have never seen any differences between the both. If i write unit tests for - both have the same result for the same combinations: // lerp assert.strictEqual(math.lerp(0, 100, -1), -100, "lerp for 0 / 100 / -1.0"); assert.strictEqual(math.lerp(0, 100, -0.75), -75, "lerp for 0 / 100 / -0.75"); assert.strictEqual(math.lerp(0, 100, -0.5), -50, "lerp for 0 / 100 / -0.50"); assert.strictEqual(math.lerp(0, 100, -0.25), -25, "lerp for 0 / 100 / -0.25"); assert.strictEqual(math.lerp(0, 100, 0), 0, "lerp for 0 / 100 / 0.0"); assert.strictEqual(math.lerp(0, 100, 0.25), 25, "lerp for 0 / 100 / 0.25"); assert.strictEqual(math.lerp(0, 100, 0.5), 50, "lerp for 0 / 100 / 0.5"); assert.strictEqual(math.lerp(0, 100, 0.75), 75, "lerp for 0 / 100 / 0.75"); assert.strictEqual(math.lerp(0, 100, 1), 100, "lerp for 0 / 100 / 1.0"); assert.strictEqual(math.lerp(100, 0, -1), 200, "lerp for 100 / 0 / -1.0"); assert.strictEqual(math.lerp(100, 0, -0.75), 175, "lerp for 100 / 0 / -0.75"); assert.strictEqual(math.lerp(100, 0, -0.5), 150, "lerp for 100 / 0 / -0.5"); assert.strictEqual(math.lerp(100, 0, -0.25), 125, "lerp for 100 / 0 / -0.25"); assert.strictEqual(math.lerp(100, 0, 0), 100, "lerp for 100 / 0 / 0.0"); assert.strictEqual(math.lerp(100, 0, 0.25), 75, "lerp for 100 / 0 / 0.25"); assert.strictEqual(math.lerp(100, 0, 0.5), 50, "lerp for 100 / 0 / 0.5"); assert.strictEqual(math.lerp(100, 0, 0.75), 25, "lerp for 100 / 0 / 0.75"); assert.strictEqual(math.lerp(100, 0, 1), 0, "lerp for 100 / 0 / 1.0"); // lerpFast assert.strictEqual(math.lerpFast(0, 100, -1), -100, "lerpFast for 0 / 100 / -1.0"); assert.strictEqual(math.lerpFast(0, 100, -0.75), -75, "lerpFast for 0 / 100 / -0.75"); assert.strictEqual(math.lerpFast(0, 100, -0.5), -50, "lerpFast for 0 / 100 / -0.50"); assert.strictEqual(math.lerpFast(0, 100, -0.25), -25, "lerpFast for 0 / 100 / -0.25"); assert.strictEqual(math.lerpFast(0, 100, 0), 0, "lerpFast for 0 / 100 / 0.0"); assert.strictEqual(math.lerpFast(0, 100, 0.25), 25, "lerpFast for 0 / 100 / 0.25"); assert.strictEqual(math.lerpFast(0, 100, 0.5), 50, "lerpFast for 0 / 100 / 0.5"); assert.strictEqual(math.lerpFast(0, 100, 0.75), 75, "lerpFast for 0 / 100 / 0.75"); assert.strictEqual(math.lerpFast(0, 100, 1), 100, "lerpFast for 0 / 100 / 1.0"); assert.strictEqual(math.lerpFast(100, 0, -1), 200, "lerpFast for 100 / 0 / -1.0"); assert.strictEqual(math.lerpFast(100, 0, -0.75), 175, "lerpFast for 100 / 0 / -0.75"); assert.strictEqual(math.lerpFast(100, 0, -0.5), 150, "lerpFast for 100 / 0 / -0.5"); assert.strictEqual(math.lerpFast(100, 0, -0.25), 125, "lerpFast for 100 / 0 / -0.25"); assert.strictEqual(math.lerpFast(100, 0, 0), 100, "lerpFast for 100 / 0 / 0.0"); assert.strictEqual(math.lerpFast(100, 0, 0.25), 75, "lerpFast for 100 / 0 / 0.25"); assert.strictEqual(math.lerpFast(100, 0, 0.5), 50, "lerpFast for 100 / 0 / 0.5"); assert.strictEqual(math.lerpFast(100, 0, 0.75), 25, "lerpFast for 100 / 0 / 0.75"); assert.strictEqual(math.lerpFast(100, 0, 1), 0, "lerpFast for 100 / 0 / 1.0"); Which one is better? Thanks in advance

Math and Physics Programming

Started by Finalspace November 27, 2015 09:09 AM

10 comments, last by alvaro 9 years, 2 months ago

PeterStock

823

November 30, 2015 10:09 AM

This is a useful reference for IEEE floating point:

http://www.appinf.com/download/FPIssues.pdf

Finalspace

Author

1,222

December 03, 2015 12:51 PM

I ported your sample to javascript and it does not produce the same result - there are no change at all.

http://jsfiddle.net/0u3hp7s5/

My youtube channel: https://www.youtube.com/user/F1nalspace

alvaro

21,610

December 03, 2015 01:56 PM

I ported your sample to javascript and it does not produce the same result - there are no change at all.

http://jsfiddle.net/0u3hp7s5/

That link is not working for me, but in JavaScript every number is a double, so you have considerably more precision.

I always write a*(1-t)+b*t because I want the expression to really evaluate to a at t=0 and to b at t=1. Worry about speed only if your profiler tells you this is a bottleneck. [Guess: it probably isn't.]

Lerp vs fastLerp

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

Lerp vs fastLerp

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

Reticulating splines