The key to performance optimization is low-level processorspecific
code. The most prominent are the SIMD instruction set
enhancements like MMX™ technology, Intel® Streaming SIMD Extensions,
and Intel SSE4. High-level languages do not usually have standard
constructs for the fine-grained data parallelism enabled by using SIMD
instructions, requiring the use of specialized tools or programming directly in
assembly language. Performance libraries that utilize SIMD instructions and
target key application domains have been developed to address this
discrepancy between low-level performance tuning and high-level application
development. The Intel® Integrated Performance Primitives (Intel® IPP)
library is one such solution. The exercises in this article cover the use of the
Intel IPP library in embedded applications. Most of the questions can be
answered by reviewing the material found in the corresponding chapter of
the companion reference book Break Away with Intel® Atom™ Processors: A
Guide to Architecture Migration.
