XL Fortran for Blue Gene/Q

Optimization technologies for Blue Gene/Q systems

• Support the Quad Processing eXtension (QPX) instruction set to help simplify and improve manipulation of vector operations in applications.
• Take advantage of thread-level speculative execution (SE) to dynamically detect thread conflicts and roll back conflicting threads for re-execution.
• Use transactional memory to help simplify parallel programming processes by grouping read and write operations and running them as a single operation.
• Provide support for 64-bit application development to help simplify the development of large applications.
• Include built-in functions to retrieve statistical information, write statistics into log files and switch modes to thread-level speculative execution at run time.

High-performance libraries and debugging tools

• Use enhanced Mathematical Acceleration Subsystem (MASS) and MASS Vector (MASSV) libraries.
• Take advantage of a shared memory parallelization (SMP) runtime library that supports both explicit and automatic parallel processing.
• Generate debugging information for use by a symbolic debugger.
• Provide compatibility with object code or libraries compiled with previous versions of the XL Fortran compiler and runtime environment.

Compiler options and reporting capabilities

• Offer compiler suboptions for fine-grained control over optimizations and transformations.
• Deliver diagnostic reports that identify areas to improve performance by showing optimizations the compiler performed and missed.
• Provide diagnostic reports in XML or HTML format for expanded viewing options.
• Generate information about data reorganizations and the location of data prefetch instructions resulting from compiler operations.
• Take advantage of internal improvements, such as alias analysis, for faster data gathering and data sharing.

Conformance with international programming standards

• Supports Fortran 2003, Fortran 95, Fortran 90 and FORTRAN 77 standards and provides initial support for the Fortran 2008 standard.
• Works with the Open Multiprocessing (OpenMP) 3.1 application programming interface (API) specification to support the development of parallel applications.
• Includes language interoperability standards to support code portability between multiple operating systems and hardware platforms.
• Offers support for Linux operating systems.
• Uses built-in functions for atomic memory access that correspond to the GNU Compiler Collection (GCC).