Hair even distribution is not even #62891

New Issue

Eneko Castresana Vara · 2019-03-24T10:12:45+01:00

2019-03-24 10:12:45 +01:00

System Information
Operating system: Ubuntu 18.04.2 LTS
Graphics card: GeForce GTX 950M

Blender Version
Broken: 2.80, 9105e1d084, 2019-03-20 22:53

Short description of error
Hair even distribution is not even.
Apologies in case. I'm unsure if this is a duplicate of https://developer.blender.org/T50536

Exact steps for others to reproduce the error
Please open attached .blend and switch even distribution.

Note how with even distribution disabled, each face emits 8 hair objects, which roughly matches the total of 50 particles set: 8x6 faces= 48 objects.

Now enable even distribution. The top and bottom faces emit 16 objects each, whereas the side faces emit 4 objects each: 16x2 + 4x4 = 48 objects.

EDIT: tl;dr: With even distribution on, it is expected (among other things, perhaps, such as similar distance between particles) the number of emitted particles should be proportional to face area. Currently it isn't proportional, and bigger faces are getting so many more particles than they should and smaller faces fewer! (Sorry about the next nonsense paragraph, I was in a rush).

Note that the emitter object has scaling applied and its side faces' areas are half of its top/bottom faces' areas. Therefore proper even distribution should work like this: Top and bottom faces have an area of 1 unit each. Side faces have an area of 0.5 units each. For every two side faces there should be the same number of particles than for any top/bottom face. Target is 50 total particles. 50/4 (every two side faces have the same area as 1 top/bottom face. So we're dividing by 4 area units) = 12.5 whose floor is 12. 12x4= 48. Therefore the top and bottom faces should emit 12 particles each, and the side faces should emit 6 particles each: 12x2 + 6x4 = 48.

hair-even-distribution.blend

**System Information** Operating system: Ubuntu 18.04.2 LTS Graphics card: GeForce GTX 950M **Blender Version** Broken: 2.80, 9105e1d08467, 2019-03-20 22:53 **Short description of error** Hair even distribution is not even. Apologies in case. I'm unsure if this is a duplicate of https://developer.blender.org/T50536 **Exact steps for others to reproduce the error** Please open attached .blend and switch even distribution. Note how with even distribution disabled, each face emits 8 hair objects, which roughly matches the total of 50 particles set: 8x6 faces= 48 objects. Now enable even distribution. The top and bottom faces emit 16 objects each, whereas the side faces emit 4 objects each: 16x2 + 4x4 = 48 objects. **EDIT: tl;dr:** With even distribution on, it is expected (among other things, perhaps, such as similar distance between particles) the number of emitted particles should be proportional to face area. Currently it isn't proportional, and bigger faces are getting so many more particles than they should and smaller faces fewer! (Sorry about the next nonsense paragraph, I was in a rush). Note that the emitter object has **scaling applied** and its side faces' areas are **half** of its top/bottom faces' areas. Therefore proper even distribution should work like this: Top and bottom faces have an area of 1 unit each. Side faces have an area of 0.5 units each. For every two side faces there should be the same number of particles than for any top/bottom face. Target is 50 total particles. 50/4 (every two side faces have the same area as 1 top/bottom face. So we're dividing by 4 area units) = 12.5 whose floor is 12. 12x4= 48. Therefore the top and bottom faces should emit 12 particles each, and the side faces should emit 6 particles each: 12x2 + 6x4 = 48. [hair-even-distribution.blend](https://archive.blender.org/developer/F6875241/hair-even-distribution.blend)

Eneko Castresana Vara commented

2019-03-24 10:12:45 +01:00

Added subscriber: @ecv

Jacques Lucke commented

2019-03-25 13:00:48 +01:00

Added subscribers: @Sergey, @brecht, @JacquesLucke

Jacques Lucke commented

2019-03-25 13:00:49 +01:00

The issue is that in psys_thread_context_init_distribute all the vertex positions are transformed with BKE_mesh_orco_verts_transform.
In the example file, this multiplies the z coordinate of every vertex with 0.5.
This is because mesh->size is (1, 1, 0.5).
This seems to be used for texture scaling or so, but I don't understand why the vertices are transformed like that before the area is calculated.

diff --git a/source/blender/blenkernel/intern/particle_distribute.c b/source/blender/blenkernel/intern/particle_distribute.c
index dd0a4a24d7b..8d0e25dcc07 100644
--- a/source/blender/blenkernel/intern/particle_distribute.c
+++ b/source/blender/blenkernel/intern/particle_distribute.c
@@ -969,12 +969,8 @@ static int psys_thread_context_init_distribute(ParticleThreadContext *ctx, Parti
 				copy_v3_v3(co1, orcodata[mf->v1]);
 				copy_v3_v3(co2, orcodata[mf->v2]);
 				copy_v3_v3(co3, orcodata[mf->v3]);
-				BKE_mesh_orco_verts_transform(ob->data, &co1, 1, 1);
-				BKE_mesh_orco_verts_transform(ob->data, &co2, 1, 1);
-				BKE_mesh_orco_verts_transform(ob->data, &co3, 1, 1);
 				if (mf->v4) {
 					copy_v3_v3(co4, orcodata[mf->v4]);
-					BKE_mesh_orco_verts_transform(ob->data, &co4, 1, 1);
 				}
 			}
 			else {

@Sergey or @brecht, can you tell me, why this is done?

The issue is that in `psys_thread_context_init_distribute` all the vertex positions are transformed with `BKE_mesh_orco_verts_transform`. In the example file, this multiplies the z coordinate of every vertex with `0.5`. This is because `mesh->size` is `(1, 1, 0.5)`. This seems to be used for texture scaling or so, but I don't understand why the vertices are transformed like that before the area is calculated. ``` diff --git a/source/blender/blenkernel/intern/particle_distribute.c b/source/blender/blenkernel/intern/particle_distribute.c index dd0a4a24d7b..8d0e25dcc07 100644 --- a/source/blender/blenkernel/intern/particle_distribute.c +++ b/source/blender/blenkernel/intern/particle_distribute.c @@ -969,12 +969,8 @@ static int psys_thread_context_init_distribute(ParticleThreadContext *ctx, Parti copy_v3_v3(co1, orcodata[mf->v1]); copy_v3_v3(co2, orcodata[mf->v2]); copy_v3_v3(co3, orcodata[mf->v3]); - BKE_mesh_orco_verts_transform(ob->data, &co1, 1, 1); - BKE_mesh_orco_verts_transform(ob->data, &co2, 1, 1); - BKE_mesh_orco_verts_transform(ob->data, &co3, 1, 1); if (mf->v4) { copy_v3_v3(co4, orcodata[mf->v4]); - BKE_mesh_orco_verts_transform(ob->data, &co4, 1, 1); } } else { ``` @Sergey or @brecht, can you tell me, why this is done?

blender-admin commented

2019-03-25 13:55:03 +01:00

This issue was referenced by 489c015e70

This issue was referenced by 489c015e706cf3bcfd246a2a081664d290ec7096

Brecht Van Lommel commented

2019-03-25 14:17:36 +01:00

Undeformed / orco texture coordinates are supposed to be stored normalized to a 0..1 range. So particle distribution then transforms them back.

There's a case here where they are stored incorrectly, I'll fix that.

Undeformed / orco texture coordinates are supposed to be stored normalized to a 0..1 range. So particle distribution then transforms them back. There's a case here where they are stored incorrectly, I'll fix that.

Jacques Lucke commented

2019-03-25 14:24:12 +01:00

It's funny that CD_ORCO coordinates (I guess that means "original coordinates"?) are actually deformed.

Maybe a comment next to the definition of CD_ORCO could help here (or a better name).

It's funny that `CD_ORCO` coordinates (I guess that means "original coordinates"?) are actually deformed. Maybe a comment next to the definition of `CD_ORCO` could help here (or a better name).

Brecht Van Lommel commented

2019-03-25 14:29:13 +01:00

Yes, it's old Blender Internal terminology for original coordinations. We should rename that once.

Brecht Van Lommel commented

2019-03-25 14:30:54 +01:00

Changed status from 'Open' to: 'Resolved'

Brecht Van Lommel closed this issue

2019-03-25 14:30:54 +01:00

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Hair even distribution is not even #62891