May 17, 2018

Compute Bound Median Filter On GPU

On May 16th 2018 we presented at PerfUG.
Our presentation was about taking an existing C# code and run it on the GPU with Hybridizer.
We went through all optimization phases, from simple C# parallel for to a rolling buffer of registers with sorting networks.

Check out the slides:

March 15, 2018

Hybridizer Essentials For Linux – Beta

With version 1.1.6151.9352 comes a beta version of Hybridizer Essentials for Linux.

You can download the package from our github.


  • Tested on Ubuntu 16.04 – should work on other debian-based systems
  • Install a version of CUDA (starting at 8.0 – 9.0, 9.1 and 9.2 supported)
  • Install Mono (version or later)


  • Intall deb package #> sudo dpkg -i hybridizer-essentials-amd64-.deb
  • Package will install in /opt/altimesh/hybridizer-essentials

Request a license:

  • Run /opt/altimesh/hybridizer-essentials/bin/RequestLicense.exe and follow the instructions.
  • The license key for your machine will be sent to you by email to the address you are invited to provide.
  • Edit a .hybridizerrc configuration in your home directory and paste the license after the hostname and column. Here is an example:
    MyMachineHostName: yTfgNaoGHDZYpP7t0+ECTLeaUuNr<...>lPg==

January 31, 2018

Run C# code on GPU in a minute

We updated Hybridizer Essentials today.
We fixed numerous bugs, and added some new features:

  • Cublas, Cusparse and Nvblas are better supported in CUDA 9.0 and 9.1
  • We now support intrinsic includes
  • Visual studio integration should be faster, due to a large code refactoring and performance improvements
  • We added an Item template containing all the required boilerplate C# code. That should save you some precious time
  • Some users reported a freeze of Visual Studio 2017. This should be fixed
  • We had a time zone issue for PST users. Licenses now work immediately from their zone
  • cudaDeviceProperties are now correctly mapped in cuda 9.0 and 9.1
  • Generated CUDA projects now contain the -G option in debug
  • Binary is compiled for the machine’s GPU. That means it’s compiled with the best compute capability available, but can’t be run on another machine

Finally we updated Hybridizer Samples to integrate those changes.
Given those changes, it’s now possible to create a new C# project and run code on a GPU in less than a minute:

Stay tuned!

January 11, 2018

Hybridizer Essentials supports CUDA 9.1

cuda 9.1 on marketplace
Since the publication in parallelforall blog, we had 500 installations. We received both great and constructive feedbacks. The most important issue was the lack of CUDA 9.1 support. This is now done with version 1.1.5685.8524 of Hybridizer Essentials.

You can download it from Visual Studio Marketplace or from our release page on our github.

We also updated our SDK accordingly. We added four Visual Studio solutions, one for each Visual Studio version we support.

Be aware that version 15.5.3 of Visual Studio breaks CUDA support (even 9.1). Keep version 15.5.2 (or below) if you want CUDA with Visual Studio 2017.