•(NEW) MPI Debugging and Profiling included in all Linux products. PGI Workstation
　 and PGI Server for Linux now include the ability to debug and profile up to four local
　 MPI processes. Both products also include a precompiled MPICH library.

•(NEW) Debugging and Profiling of Microsoft MPI programs (MSMPI) running on
　 Microsoft Compute Cluster Server (CCS).

•(NEW) Expanded Products and Platforms
　　　　- PGI Workstation for MacOS 10.4.9 and 10.5.
　　　　- PGI Server cross-platform (Linux, MacOS & Windows/SUA)
　　　　- PGI Visual Fortran Server network floating license

•(NEW) Expanded Linux Distributions Support
　　　　- SuSE 10.2 and SLES 10
　　　　- Fedora Core 7
　　　　- Red Hat Enterprise Linux 5
　(NEW) Unlimited OpenMP thread counts in all PGI products

•Quad-Core AMD Opteron micro-architecture optimizations
　　　　- Optimize code selection to take full advantage of 128-bit FPU (reduce merge
　　　　　dependencies)
　　　　- Aggressive use of 2MB huge pages to exploit increased number of TLB entries
　　　　- Optimized heap allocation to improve use of 2-way set associative L1 cache

•Over 5% performance increase over latest update of release 7.0

•Memory hierarchy and memory allocation optimizations including huge pages support

•Enhanced auto-parallelization of vector loops across multiple cores using memory
　bandwidth-based profitability analysis

•PGCC and PGC++ inlining enhancements and optimizations
　　　　- inlining of file static functions across files
　　　　- PFO-driven inlining of indirect calls through IPA

•Improved IPA compilation speed

•Enhanced profile-feedback optimizations
　　　　- Feedback aggregation over multiple runs
　　　　- Function inlining
　　　　- Register allocation

•Enhance code placement optimizations
　　　　- Improved handling of large functions
　　　　- Optimize placement of successor blocks with similar probabilities
　　　　- Improved placement in absence of dynamic profile feedback

•Support for dynamic linking on Windows SFU

•Support for static linking on Windows

•Vectorization optimizations
　　　　- Extended SSE vectorization for loops with indirect addressing
　　　　- Additional SSE3 extensions
　　　　- Optimized register selection
　　　　- Cache alignment tuning for multi-core

•Enhanced gcc/g++ compatibility
　　　　- source correlation and handling of include files
　　　　- support for namespaces, overloading and reference variables

•PGDBG debugger enhancements
　　　　- Improved stack traces
　　　　- interoperability with Microsoft Visual C++
　　　　- fast disassemply
　　　　- performance improvements
　　　　- improved interoperability with gcc/g++

•PGPROF MPI profiling of aggregate communication routines

•PGI CDK supports MVAPICH running over InfiniBand*

•Updated PGI User's Guide and PGI Visual Fortran Users Guide

•Pre-validation of de facto standard support libraries including NetCDF,
　F90 OpenGL, ATLAS, ScaLAPACK, MPI-CH and LAM MPI

•Pre-validation of POP, MM5, CAM, MOM4, WRF2, AMBER, CHARMM, GAMESS,
　 MOLPRO, MCNP5 and TBMD