Nvidia on Tuesday will unveil upgrades to Parallel Nsight, a toolkit for building GPU-accelerated applications from within Microsoft's Visual Studio IDE, and Cuda Toolkit, for leveraging massively parallel processing capabilities of GPUs (graphics processing units). Nvidia Parallel Nsight 1.5 features support for the Visual Studio 2010 software development platform released in April.
"For 6 million Windows developers already using Windows and Visual Studio, this new release of Parallel Nsight provides some significant features [for building GPU-accelerated, parallel applications]," said Will Ramey, senior product manager for GPU computing at Nvidia. These types of applications are prominent in industries such as oil and gas exploration, as well as in areas such as graphics, he said.
"With Parallel Nsight, we're bringing GPU, many-core debugging into Visual Studio 2010," said Stephen Jones, product line manager for developer tools at Nvidia. Multicore debugging is specific to GPUs, according to Nvidia. Single-system debugging in version 1.5, meanwhile, allows a developer to debug on one GPU while using another to display the desktop. Developers do not need to use multiple machines.
Also featured in version 1.5 is system analysis functionality, in which Visual Studio users can visualise events occurring on the CPU and GPU across a correlated timeline. Events will be correlated on cores. "[This capability] allows the developer to see exactly what's going on in their application across these different processing units" and optimize performance characteristics, said Jones. Tesla Compute Cluster (TCC) debugging is featured in version 1.5 as well.
Parallel Nsight 1.5 supports the other product being introduced Tuesday, Cuda Toolkit 3.2. The upgrade to Cuda Toolkit offers new math libraries, performance improvements, and advanced cluster management.
Cuda Toolkit 3.2, which is free, will be available as a release candidate in mid-September. The Professional edition of Parallel Nsight is available now as a release candidate. A date has not been set for general availability, but the product will be priced at $349.
A free Standard edition, with less functionality than the Professional variant, will be available on September 22. Standard Edition lacks capabilities such as Tesla Compute Cluster and data breakpoints support.