Intel today released version 2016 of Parallel Studio XE. It now includes the new Intel Data Analytics Acceleration Library (DAAL), as well as full support for C++11.
James Reinders, chief evangelist at Intel Software, said that the new edition can plug developers into existing workflows that might include Big Data processing systems like Hadoop.
(Related: Intel impresses at its developer forum)
“In Big Data, when you’re doing analytics, you’re looking for correlations, so computation is king,” he said. “The Intel Math Kernel Library (MKL) has enormous number-crunching capabilities, but with Big Data, you need to move data around. Where do you get it? How do you stream it? We had to be able to attach different data sources. Being able to fit into Hadoop and that sort of an environment is what DAAL does: It takes MKL and puts things around it to make that possible.”
Many of the changes and additions in Intel Parallel Studio XE 2016 are aimed at improving performance of the code these tools are used against. “We have a new advisor capability,” said Reinders. “The parallel advisor that has given threading advice in the past, runs code, analyzes it, and builds a model of how it runs.”
For this release, however, he said that Parallel Studio XE 2016 now supports advising users on vectorization. “Using the vector instructions, how do you get your code to take advantage of these vector instructions? With Xeon 5, single precision code could potentially do 16 threads at a time. The core of your algorithm could be staring at a 16x advantage with vectorization. Sixteen times or even something less than that is hard to ignore.”
This is the first major release of Intel Parallel Studio XE to include expanded support for C++14. Also included in this release are new MPI snapshotting capabilities. Reinders said that Parallel Studio’s more heavyweight profiling and tracing tool was too heavy for large-scale applications. Where previously applications across two to 3,000 nodes were supported, he said that up to 32,000 nodes can be snapshotted with the new tool, without incurring a significant slowdown.