Gold Rush Around New Generation Compilers
The invasion of multi-core processors and the ubiquitous parallelism came somehow unexpectedly for the software industry (Intel had drastically changed its mind to multi-core projects in 2004). As a result, modern software development tools were completely unsuitable for parallel PC architecture. But almost all modern processors are multi-core, and the number of these cores grows exponentially in them: 4, 8, 16, etc. If the programs do not learn how to use parallelism effectively in the near future, then all the technical progress in the process of designing will go down the drain.
In fact, this is a new Grail for development companies: who will be the first to develop compilers for the efficient parallelization of programs? Here begins the real gold rush, writes the NY Times .
A month ago, Intel and Microsoft announced a research program in this direction. They finance the work of the group at the University of California at Berkeley. Researchers from the University of Illinois are fighting the same problem. Now it has become known that another - the third - the alliance of “gold miners” at Stanford (Pervasive Parallelism Lab) formed the corporations Sun Microsystems, AMD, Nvidia, IBM, HP and the same Intel.
Programmers from Stanford have been allocated a budget of $ 6 million for three years. Professor Kunle Olukotun, who was involved in the creation of the first multi-core processors, was appointed to lead.
')
For programs to effectively use multi-threading, only new compilers may not be enough. At Stanford, they will conduct various studies, including trying to create a fundamentally new programming language and introduce some kind of hardware innovations. Scientists even plan to use virtual reality and robots to test their theories.
In Berkeley, researchers have broken the spectrum of parallelism problems into seven classes of problems that will be solved by different methods.
In fact, this is a new Grail for development companies: who will be the first to develop compilers for the efficient parallelization of programs? Here begins the real gold rush, writes the NY Times .
A month ago, Intel and Microsoft announced a research program in this direction. They finance the work of the group at the University of California at Berkeley. Researchers from the University of Illinois are fighting the same problem. Now it has become known that another - the third - the alliance of “gold miners” at Stanford (Pervasive Parallelism Lab) formed the corporations Sun Microsystems, AMD, Nvidia, IBM, HP and the same Intel.
Programmers from Stanford have been allocated a budget of $ 6 million for three years. Professor Kunle Olukotun, who was involved in the creation of the first multi-core processors, was appointed to lead.
')
For programs to effectively use multi-threading, only new compilers may not be enough. At Stanford, they will conduct various studies, including trying to create a fundamentally new programming language and introduce some kind of hardware innovations. Scientists even plan to use virtual reality and robots to test their theories.
In Berkeley, researchers have broken the spectrum of parallelism problems into seven classes of problems that will be solved by different methods.
Source: https://habr.com/ru/post/25072/
All Articles