Possible irregularity around 0,0,0 for Noise3Dgrad

[TheWand3rer]

First of all, many thanks for this noise library.
I am using your Noise3D gradient function (converted to HLSL though) for a terrain generation system. I have noticed that whenever the function is called with input (0,0,0) it glitches. See here:

As you can see one vertex is completely wrong. If I add a bias of 0.1 on any of the coordinates it goes away (i.e. passing 0, 0.1, 0). I am using the result of the noise function as the height of the vertex.

As far as I can tell the HLSL version of the function is identical, save obviously for the vecn / floatn changes. I realise debugging this must be super hard, but maybe it does this on your end too.

Here's just one octave of the noise, with the other terrain chunks removed:

[stegu]

Weird. Although integer positions should typically be avoided for any gradient noise (because the function is zero at those points), this should have caught my eye before. My considerably more recent function "psrdnoise" could be a better choice. Also on Github, in stegu/psrdnoise. But I should still fix this bug. Thanks for reporting it! Den sön 19 jan. 2025 19:43TheWand3rer ***@***.***> skrev:

…

First of all, many thanks for this noise library. I am using your Noise3D gradient function (converted to HLSL though) for a terrain generation system. I have noticed that whenever the function is called with input (0,0,0) it glitches. See here: Screenshot.2025-01-19.192629.png (view on web) <https://github.com/user-attachments/assets/69d17ddd-078b-4187-bc65-4c59e2cfb2d1> As you can see one vertex is completely wrong. If I add a bias of 0.1 on any of the coordinates it goes away (i.e. passing 0, 0.1, 0). I am using the result of the noise function as the height of the vertex. image.png (view on web) <https://github.com/user-attachments/assets/0442cc5e-ea5c-4551-949c-7c7dde3e57ea> As far as I can tell the HLSL version of the function is identical, save obviously for the vecn / floatn changes. I realise debugging this must be super hard, but maybe it does this on your end too. Here's just one octave of the noise, with the other terrain chunks removed: image.png (view on web) <https://github.com/user-attachments/assets/9aa4a666-6d15-47c7-a483-04209031539e> — Reply to this email directly, view it on GitHub <#13>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAFGK2VQICUUQK56S7LOOOD2LPW6BAVCNFSM6AAAAABVO6VK7CVHI2DSMVQWIX3LMV43ASLTON2WKOZSG44TONZXGE4TGMQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>