For many years the main objective of this research was to find solutions to enable the construction of so-called super computers, or machines capable of supporting scientists in solving mathematical problems very sophisticated. The progress made in recent years in the manufacturing processes of integrated circuits, on the one hand I agree on the same chip to achieve an extraordinary number of devices, the other has forced designers to rethink radically the structure of the CPU. This is why, unlike to what happened a few years ago, the frequency of operation is no longer an appropriate metric for measuring processor performance.
Already in 1996 Professor Kunle Olukotun, Stanford University, has shown that making four simple processors on the same silicon chip was possible to achieve superior performance compared to those of a traditional single processor, having the same size, and based on the solutions more advanced techniques at that time. The work of Prof. Olukotun really represented the beginning of the multi-core processors. The Cell processor used in the PlayStation 3 is one of the first multi-core products that has had commercial importance and, as is known, today many of the PC desktop and laptop processors available on the market have more than one core.
Because the major processor manufacturers can continue developing these solutions, selling over the coming years with hundreds of CPU cores, it is necessary to solve some technical problems which the scientific community is dealing with. The most important of these is the availability of a software technology that can "parallelize" the applications and make them suitable to be performed by multi-core CPU, to its full potential. The researchers are currently working on the development of compilers capable of split applications runnable processes independently on different CPU cores. E 'reasonable to expect that in a few years our mount a PC with at least 100 processor cores and is capable of performing, with outstanding performance, and video games realistic scientific applications. Here you can use a compiler that hides the complexity of the hardware and allowing the same programming approach used today for traditional x86 machines.
0 comments:
Post a Comment