| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Dear all,There is an interesting survey regarding open hardware and its application in scientific computing:
https://docs.google.com/spreadsheet/viewform?formkey=dFc4cjlfRm1PRWdjS0xheGZqRWcxcWc6MQ Furthermore, CERN is a major supporter of open hardware: http://joinup.ec.europa.eu/news/eu-cern-launches-open-hardware-initiativeFrom this point of view, I believe we can provide adequate support for interval computations, via an open hardware framework. Of course, in the near future, we will have to rely only on FPGA implementations, but I think that it represents a good starting point before any major chip manufacturer will include interval arithmetic support (however, I believe that interval extensions to open-source cores (such as the OpenRisc) or hardware accelerators for interval computations on FPGAs will be sufficient for scientific purposes).
Best regards, Alexandru Amaricai On 3/15/2013 10:46 PM, Kreinovich, Vladik wrote:
We had an interesting talk on computer architecture today, and the presenter said that, to her big surprise, a lot of architecture design details are motivated by the National Labs. Companies like Intel and IBM regularly ask researchers from the national labs about the features these researchers want to see -- since the labs, via Federal funding, are among the main buyers of state-of-the-art high-performance computers. Sometimes, if a minor computer architecture feature is not needed much for a general customer but is useful for national labs, a part of a big grant given to a national lab for solving an important problem goes to the company as a subcontractor to subsidize the company's development of this feature. From this viewpoint, if we can get a good national lab support for interval computations, this will boost our chances that relevant features will be implemented in hardware.