Jump to content

GPU deformSop - H15


Guest tar

Recommended Posts

Out of interest has anyone made this env work; on OsX it crashes out. H1.0.417, OsX 10.10.5

 

HOUDINI_FAST_SKIN_BLENDSHAPES

    Enable experimental feature to use the GPU to do deformation

    in SOP_Deform and SOP_BlendShapes.

Link to comment
Share on other sites

Pretty sure this approach was abandoned.

 

I'm curious, as to why?

 

Either way, i enabled it last night and found it difficult to find improvement when using OpenCL. ( Not sure what shared OGL Sharing is though ). This was only after spending a few minutes testing it. In some instances it was faster/smoother, others it was equaled when unchecked.    

 

A few versions ago I thought i would not be able to use the blendshape sop in general, because it can get a little sluggish the higher the polycount (understandably). But with each version of Houdini, it seems to be handling geometry faster and faster.  

 

I made a simple VEX based blendshape sop a while back and it ran noticeably faster than the standard sop. It lead me to believe the sop wasn't multi threaded? Which would have been a shame as it had decent blend options and the update channel names from inputs was a nice feature. 

 

But it was hard to tell the difference. Between my simplified vex version, vop, blendshape sop with OpenCL enabled or the Blendshape sop at default... they all hover very much in the same frame rate at times. This was with a 140k poly head with 2 shapes + 1 corrective handled by chops.

 

One thing that kept bothering me though is you cant apply a material beforehand (it wont show with ogl settings things like adjusted spec roughness is nice when viewing). You can obviously apply a material sop after and set the display flag on that but it runs noticeably slower. When i plop a material sop before my vex/vop versions i can view it all on one node and it runs like butter. Anyone know a work around?

 

Oh and the only crash i got was when i activated blend normals with OpenCL enabled. 

Link to comment
Share on other sites

I'd have to ask to be sure, but it was probably abandoned for the reasons you've mentioned - it's just not that much of an improvement to warrant the extra complexity.

 

It's also possible that the Blend Shape SOP wasn't the bottleneck in playback.

 

The material SOP should go before any time-dependent SOPs, otherwise you incur the penalty of reapplying the material to all 140k polygons every frame. I'm not sure why this wouldn't work beforehand as long as the material attribute is still there, unless the blendshape is causing a different GL shader to be used when GL sharing is active (GL and CL share the buffers, rather than reading the CL buffer back to memory and then pushing it back to GL).

Link to comment
Share on other sites

I disabled the fast_skin_blendshapes and now all is well, the shader displays properly.

 

In testing it a lil bit more, I do think it is a bit faster but i have an old system and an old card, im curious how well it performs. Also i can you ask about blendshape and sequence blend sop being multi-threaded, there is a big difference in my opinion between vop/vex sop solution and these sops.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...