Fluid Sim: Conditional inaccurate frame mapping #39583

New Issue

Willi · 2014-04-04T01:30:59+02:00

Willi commented

2014-04-04 01:30:59 +02:00

System Information
Win7/64, AMD HD 6850

Blender Version
Broken: 2.70 official, 2.69, 2.67, 2.66.1, 2.65 (r53189)
Worked: 2.63, 2.64a (r51232)

Short description of error
If an inflow object is enabled by a key frame, the simulation may start one frame too late depending on the position of the key frame and the end time of the domain object.

Exact steps for others to reproduce the error

Open the appended blend file (including prepared UI): The scene contains one domain object and one inflow object. The inflow object is enabled at frame 100.
Bake the domain object.
Navigate to frame 100: The domain object is replaced by the fluid mesh. It's the frame at which the inflow object is enabled. This is expected behavior.
In the dope sheet window, move the key frame from 100 to 160.

Bake the domain object again and navigate to frame 160: The domain object is not replaced by the fluid mesh. Instead it's replaced at frame 161.

Problem: Depending on at which frame the inflow object is enabled, and also depending on the "end time" of the domain object, the domain object is replaced by the fluid mesh at that frame or one frame later. The result should be more consistent and reliable.

FluidSim_Offset.blend

**System Information** Win7/64, AMD HD 6850 **Blender Version** Broken: 2.70 official, 2.69, 2.67, 2.66.1, 2.65 (r53189) Worked: 2.63, 2.64a (r51232) **Short description of error** If an inflow object is enabled by a key frame, the simulation may start one frame too late depending on the position of the key frame and the end time of the domain object. **Exact steps for others to reproduce the error** - Open the appended blend file (including prepared UI): The scene contains one domain object and one inflow object. The inflow object is enabled at frame 100. - Bake the domain object. - Navigate to frame 100: The domain object is replaced by the fluid mesh. It's the frame at which the inflow object is enabled. This is expected behavior. - In the dope sheet window, move the key frame from 100 to 160. # Bake the domain object again and navigate to frame 160: The domain object is *not* replaced by the fluid mesh. Instead it's replaced at frame 161. Problem: Depending on at which frame the inflow object is enabled, and also depending on the "end time" of the domain object, the domain object is replaced by the fluid mesh at that frame or one frame later. The result should be more consistent and reliable. [FluidSim_Offset.blend](https://archive.blender.org/developer/F83885/FluidSim_Offset.blend)

Willi commented

2014-04-04 01:30:59 +02:00

Changed status to: 'Open'

Willi commented

2014-04-04 01:30:59 +02:00

Added subscriber: @willi-2

Willi commented

2014-04-04 01:43:08 +02:00

Sorry, little correction: Instead of "end time of the domain object" I actually meant "end frame of the scene", but I think the end time could also have an influence.

Daniel Genrich was assigned by Thomas Dinges

2014-04-07 09:07:54 +02:00

Thomas Dinges commented

2014-04-07 09:07:54 +02:00

Added subscriber: @ThomasDinges

Thomas Dinges commented

2014-04-07 09:07:54 +02:00

Daniel, can you please take a look?

Sergey Sharybin commented

2014-09-17 14:05:53 +02:00

Added subscriber: @Sergey

Sergey Sharybin commented

2014-09-17 14:05:53 +02:00

Tried to solve this, and it's rather tricky one. Calculating doubles on blender side, passing them as floats to elbeem and casting them to doubles again is gonna to give really nasty precision loss. it's not so easy to bring back to consistency without some major API reconsiderations..

Also, assuming there's ongoing project about better fluid sim integration, not sure we need to spend time on this one. I've made sure this report is listed from the TODO http://wiki.blender.org/index.php/Dev:2.5/Source/Development/Todo/Tools#Fluid_Simulation

Thanks for the report, but closing it now.

Tried to solve this, and it's rather tricky one. Calculating doubles on blender side, passing them as floats to elbeem and casting them to doubles again is gonna to give really nasty precision loss. it's not so easy to bring back to consistency without some major API reconsiderations.. Also, assuming there's ongoing project about better fluid sim integration, not sure we need to spend time on this one. I've made sure this report is listed from the TODO http://wiki.blender.org/index.php/Dev:2.5/Source/Development/Todo/Tools#Fluid_Simulation Thanks for the report, but closing it now.

Sergey Sharybin commented

2014-09-17 14:06:54 +02:00

Changed status from 'Open' to: 'Archived'

Sergey Sharybin closed this issue

2014-09-17 14:06:54 +02:00

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

Fluid Sim: Conditional inaccurate frame mapping #39583

Bake the domain object again and navigate to frame 160: The domain object is not replaced by the fluid mesh. Instead it's replaced at frame 161.