Intel has announced updates to a number of its compilers, libraries, and cluster tools. In general, the upgrades focus on parallel programming support and optimizations. The tools that are part of the upgrades and the new version numbers include:
- Intel Compilers 11.1 (Fortran and C/C++, for Windows, Linux, Mac OS X)
- Intel Integrated Performance Primitives (IPP) 6.1 (for Windows, Linux, Mac OS X)
- IntelMath Kernel Library (MKL) 10.2 (for Windows, Linux, Mac OS X)
- Intel Cluster Toolkit, Compiler Edition 3.2.1 (for Windows, Linux)
With the upgrades, support is included for the new AVX and AES instructions. AVX, short for "Advanced Vector Extensions," is a 256-bit instruction set extension to SSE and is designed for applications that are floating-point intensive. AVX can improve performance on existing and new applications that lend themselves to largely vectorizable data sets. Wider vector data sets can process up to twice the throughput of 128-bit data sets. The AES (short for "Advanced Encryption Standard") instructions set is the U.S. Government standard for symmetric encryption. It includes four instructions to facilitate high-performance AES encryption and decryption, and two instructions support the AES key expansion procedure. The Intel Software Development Emulator implements both AVX and AES and can be downloaded here.
Among other enhancements, Intel Compilers 11.1 provides support for Eclipse CDT 5.0 and SUSE Linux Enterprise Server 11 (SLES11). Moreover, the Intel Parallel Debugger Extension added to Intel C++ Compiler, Professional Edition for Windows. This lets you step through a parallel region in serial mode during debug session, without rebuilding. You can also find data sharing violations and breakpoint when the occur, and identify re-entrant functions and, breakpoint when one is encountered. This works with OpenMP applications compiled with /debug:parallel from Intel C++ Compiler.
Upgrades to the Integrated Performance Primitives (IPP) 6.1 support:
- Deferred Mode Image Processing (DMIP) Framework that delivers pipelined parallelism, improving performance of pipelined image operations, especially on larger images. The 6.1 version introduces task parallelism and, Intel claims, as much as 250% multicore performance scaling.
- PNG codec has been added to Unified Image Codec framework with 300% faster encoding than open source reference implementation.
- Improved Visual Studio Integration . Intellisense autocompletes function names and exposes parameter details for inclusion of Intel PP functions.
- Processor optimizations for future AVX 256-bit CPU architecture. A set of 65 optimized functions is included which average a 50% speedup over Core i7 processor performance.
- Improved data compression deflate/inflate APIs provide better zlib compatibility. Supports broader set of zlib usage scenarios so more applications can take advantage of IPP zlib acceleration.
- New cryptography functions (RSA_SSA1.5,RSA_PKCSv1.5) added to support HDCP 2.0 standard.
- Texture compression, advanced lighting, and 3D geometric super sampling functions added for improved image processing performance.
The Math Kernel Library (MKL) 10.2 now includes support for Intel Xeon 5500 optimizations, .NET/C#, AVX, improved FFTs (Scaling for scaling factors 1/N, 1/sqrt(N), DFTI_FORWARD_SIGN, Radices mix of 7, 11 and 13. and optimizer real data transforms in Cluster FFT), single-precision support in PARDISO ("Parallel Direct and Iterative Solvers"), and more.