Using IBM XL Compilers to Optimize Applications on Linux on POWER8

Bookmark and Share

IBM has broadened its Linux on POWER offering to include support for Linux little Endian distributions running on IBM POWER8 systems. One great way to boost the performance of your POWER8 systems is to compile your applications with the IBM XL compilers (C/C++ and Fortran). These compilers have delivered industry leading performance for over 20 years on AIX, and are now available for IBM POWER little Endian. In combination with IBM POWER8 system processor announcement, the XL compilers delivered leadership scores in the SPEC CPU2006 benchmarks. 
The XL C/C++ and XL Fortran compilers offer a wide range of optimization levels, from –O2 through –O5, with varying tradeoffs between compile time and runtime performance. The –O2 option provides basic optimization and delivers performance on C/C++ applications similar to the gcc –O2 option in most cases.  –O3 or 
O3 qipa is recommended for most commercial code, while O3 or O3 qhot delivers good performance results for most technical computing code. With these recommended options, the IBM XL compilers typically perform 10-30% better than GCC. For example, some open source software (OSS) applications can benefit from the IBM XL compilers' optimization and Power8 exploitation to run pretty faster on Linux on Power8 systems, including scripting languages, data bases and analytics. 
The XL compilers fully exploit features of the Power8 processor including the vector registers and instructions (allowing a single instruction to act on multiple data elements at the same time), cache prefetch instructions (to minimize cache miss delays when accessing data from RAM), and simultaneous multi-threading (SMT) which allows a single POWER8 core to execute up to 8 threads at once. 
For more information, go here.