| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Am 01.05.2015 um 10:48 schrieb David Lester:
On 30 Apr 2015, at 21:42, Vincent Lefevre<vincent@xxxxxxxxxx> wrote: On 2015-04-30 17:49:51 +0200, Ulrich Kulisch wrote:It computes the exact dot product totally on chip without any memory involvement.You need memory on chip for the long accumulator. At least one for each core.Just so. My suggestion to Ulrich is the following approach (which we use in SpiNNaker/Human Brain Project): Use a minimal processor, attach a small amount of instruction memory (16-32K), and data memory (32-64K) in a Harvard configuration (separate instruction/data paths). That’s your processing node.
Yes, that is coming very close to what I claimed. With a small amount of register memory you can compute all dot products exactly and at extreme speed just by avoiding the memory bottle-neck for intermediate results.. One needs about 600 bit for IEEE 754 short. We needed about 1000 bit for the /370 format long in the IBM product ACRITH-XSC and about 2000 bit should suffice for 754 double.
The exact dot product is a key operation for verified computing with high accuracy.
Thanks for the attachments. Are the instructions available? With best regards Ulrich Kulisch -- Karlsruher Institut für Technologie (KIT) Institut für Angewandte und Numerische Mathematik D-76128 Karlsruhe, Germany Prof. Ulrich Kulisch KIT Distinguished Senior Fellow Telefon: +49 721 608-42680 Fax: +49 721 608-46679 E-Mail:ulrich.kulisch@xxxxxxx www.kit.edu www.math.kit.edu/ianm2/~kulisch/ KIT - Universität des Landes Baden-Württemberg und nationales Großforschungszentrum in der Helmholtz-Gesellschaft