Author Topic: NVIDIA CUDA Toolkit 4.1 RC1 Available For Registered Developer  (Read 5079 times)

0 Members and 1 Guest are viewing this topic.


  • Global Moderator
  • Hero Member
  • *****
  • Posts: 4222
    • View Profile
NVIDIA CUDA Toolkit 4.1 RC1 Available For Registered Developer
« on: November 09, 2011, 10:20:20 PM »
CUDA 4.1 Release Candidate Now Available
by Ian Buck, posted Nov 09 2011 at 04:07AM

With the posting of CUDA 4.1 Release Candidate 1.0 to our registered Developer site there are 3 new technology pieces you should know about in 4.1.

New Compiler: I’m excited to say that we have transitioned to a new foundation for our compiler technology, LLVM.  The LLVM open source compiler has taken the compiler world by storm with adopters ranging from academia to commercial users.   We’ve spent man-years working on tuning and optimizing this new compiler for you. Early feedback we’re hearing is that many applications are seeing a nice increase in performance as well.

Image Processing: Developers focused on image processing are going to want to dive into our 1,000+ new image processing functions added to our NPP (NVIDIA Performance Primitives) Library.  We consider these a core foundation for anyone engaged in GPU acceleration of computer vision, image processing, geospatial intelligence, augmented reality and computational photography.  Find out more about the image processing functions on our NPP library page

Dev Tools: Profiling and Tuning tools were dramatically improved with a new version of our Visual Profiler with automated performance analysis and expert guidance. We built this product with an eye for customers who are tuning and optimizing heterogeneous code bases with CUDA C and CUDA C++. The new expert system will analyze your program and provide step by step improvement suggestions and links you directly to the optimization documentation.

See the full new feature list on the CUDA 4.1 home page.


  • NewsPosters
  • Jr. Member
  • *
  • Posts: 92
    • View Profile
Re: NVIDIA CUDA Toolkit 4.1 RC1 Available For Registered Developer
« Reply #1 on: November 12, 2011, 10:54:53 AM »
Cool I will download this and try it out.

Glad to see they are now using LLVM.

Both AMD and Intel use LLVM for OpenCL so we should get consistency between how the compilers interpret OpenCL code.