It looks like you're new here. If you want to get involved, click one of these buttons!
This new Cuda 7.5 the dll is now listed as nvvm64_30_0.dll was looking into this as to see why GPU Compute was not working.
In the Doc it looks like the specific library that GPU compute is looking for is nvvm_20_0.dll
@Darksuit What version of Fabric are you using? What operating system are you on? We made a change that should have made this work on Windows 10 in Fabric 2.2. Pretty hard to say what is going on here without more specifics.
Not a problem, I have the latest 2.2.1 of Fabric (FabricEngine-2.2.1-Windows-x86_64) running on Windows 10. I pulled Cuda from Nvidia this afternoon. So this may be a recent Nvidia change though it looks like the nvvm file was built 8/15/15
the Path is set to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.5\nvvm\bin
The specific Error that Fabric is reporting states that it is not able to find nvvm64_20_0.dll which makes since as the file is labelled as nvvm64_30_0.dll
This is fresh install in a clean directory.
This was also done running as Standalone. promt to start up then runnig canvas.
It is finding the other DLL needed for CUDA and GPU compute. This is at least confirming that I have my ENV paths setup correctly. I will test this again on a separate machine that I have fabric running on.
Cuda 7.5 should load fine (even if nvvm64_20_0.dll is now nvvm64_30_0.dll).
Looking at your PATH seems you have only added
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.5\nvvm\bin in your path
While you also need to add C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.5\bin
C:\fabric\FabricEngine-2.2.1-Windows-x86_64>set PATH=%PATH%;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.5\bin;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.5\nvvm\bin
Let us know if it works for you
Technical Product Manager
Fabric Software Inc.
Work fine with CUDA v8.0.27 too.
I just rename new one nvvm64_31_0.dll to nvvm64_30_0.dll.
This is good to now Dmitry
Why fabric not support cuda GPU with computing power less than 3
I have nvidia quadro 6000 card & two tesla c 2075 cards " power capability 2"
Is there any way to make GPU work
We need computing capability of 3 because our implementation is using unified memory that came with the Kepler GPU architecture
I am sorry there is no current workaround on this.
©Copyright 2017 Fabric Software Inc. All rights reserved. | Privacy