Hyper-threading allows a single microprocessor to behave like two virtual chips and can run code faster, in a similar way to a dual-processor system.
To take advantage of this, the program code has to be split into two threads. This requires thought and adds to development time. The new compilers aim to aid this process by looking for sections of the program that could benefit from being run as parallel threads.
The compilers also support the OpenMP standard to allow high-level directives to be used to simplify the process of turning programmer source code into the machine language understood by the processor.
Hyper-threading has been a feature of Intel's server and high-end PC chips released over the past year (Itanium 2, Xeon, Pentium 4). Previous releases were criticised because the power of the chip was outstripping the performance of the software but, with threading, the company claims that software will run up to 40% faster than normal code.
The Version 7 compilers are available for C++ and Fortran for Windows and Linux operating systems. They have been designed to integrate with Microsoft Visual Studio and support many of the features of Compaq Visual Fortran, including command line compatibility.
The new compilers could benefit compute-intensive applications such as transaction-oriented programs, intensive financial, engineering and scientific operations and complex visual rendering systems for digital media, special effects and games.
Cern, the European organisation for nuclear research, has tested the C++ compiler on its software. Fons Rademakers, senior scientist at Cern, said the porting of 800,000 lines of C++ and 90,000 lines of C source code was compiled in an afternoon, a job that would previously have taken a week.