Intel Advisor, an integral part of Intel Parallel Studio XE 2017, can help identify portions of code that could be good candidates for parallelization (both vectorization and threading). It can also help determine when it might not be appropriate to parallelize a section of code, depending on the platform, processor, and configuration it’s running on. Intel Advisor Roofline Analysis reveals the gap between an application’s performance and its expected performance.
Discovering where the performance bottlenecks are and knowing what to do about it can be a mysterious and complex art, needing some very sophisticated performance analysis tools for success. That’s where Intel® VTune™ Amplifier XE 2017, part of Intel Parallel Studio XE, comes in.
When used in a TBB environment, Intel has demonstrated a many-fold performance improvement over the same parallelized code using Intel MKL in an OpenMP environment. Intel TBB-enabled Intel MKL is ideal when there is heavy threading in the Intel TBB application. Intel TBB-enabled Intel MKL shows solid performance improvements through better interoperability with other parts of the workload.
With the release of Intel Parallel Studio XE 2017, the focus is on making applications perform better on Intel architecture-based clusters. Intel MPI Library 2017, a fully integrated component of Intel Parallel Studio XE 2017, implements the high-performance MPI-3.1 specification on multiple fabrics. It enables programmers to quickly deliver the best parallel performance, even if you change or upgrade to new interconnects, without requiring changes to the software or operating environment.
Intel DAAL is a high-performance library specifically optimized for big data analysis on the latest Intel platforms, including Intel Xeon®, and Intel Xeon Phi™. It provides the algorithmic building blocks for all stages in data analysis in offline, batch, streaming, and distributed processing environments. It was designed for efficient use over all the popular data platforms and APIs in use today, including MPI, Hadoop, Spark, R, MATLAB, Python, C++, and Java.
“By implementing popular Python packages such as NumPy, SciPy, scikit-learn, to call the Intel Math Kernel Library (Intel MKL) and the Intel Data Analytics Acceleration Library (Intel DAAL), Python applications are automatically optimized to take advantage of the latest architectures. These libraries have also been optimized for multithreading through calls to the Intel Threading Building Blocks (Intel TBB) library. This means that existing Python applications will perform significantly better merely by switching to the Intel distribution.”
“Many of the libraries developed in the 70s and 80s for core linear algebra and scientific math computation, such as BLAS, LAPACK, FFT, are still in use today with C, C++, Fortran, and even Python programs. With MKL, Intel has engineered a ready-to-use, royalty-free library that implements these numerical algorithms optimized specifically to take advantage of the latest features of Intel chip architectures. Even the best compiler can’t compete with the level of performance possible from a hand-optimized library. Any application that already relies on the BLAS or LAPACK functionality will achieve better performance on Intel and compatible architectures just by downloading and re-linking with Intel MKL.”
While HPC developers worry about squeezing out the ultimate performance while running an application on dedicated cores, Intel TBB tackles a problem that HPC users never worry about: How can you make parallelism work well when you share the cores that you run upon?” This is more of a concern if you’re running that application on a many-core laptop or workstation than a dedicated supercomputer because who knows what will also be running on those shared cores. Intel Threaded Building Blocks reduce the delays from other applications by utilizing a revolutionary task-stealing scheduler. This is the real magic of TBB.
“Managing the work on each node can be referred to as Domain parallelism. During the run of the application, the work assigned to each node can be generally isolated from other nodes. The node can work on its own and needs little communication with other nodes to perform the work. The tools that are needed for this are MPI for the developer, but can take advantage of frameworks such as Hadoop and Spark (for big data analytics). Managing the work for each core or thread will need one level down of control. This type of work will typically invoke a large number of independent tasks that must then share data between the tasks.”