Search
Search Funnelback University
- Refined by:
- Date: 2000
1 -
9 of
9
search results for news |u:www.tcm.phy.cam.ac.uk
Fully-matching results
-
4. New parallel implementation
www.tcm.phy.cam.ac.uk/~pdh1001/papers/paper4/node4.html7 Jan 2000: Another advantage of the new method is that the Hamiltonian matrix is now effectively blocked. ... operations. The new method distributes the data such that each block of the matrix can be applied in. -
8. Conclusions
www.tcm.phy.cam.ac.uk/~pdh1001/papers/paper4/node8.html7 Jan 2000: 8. Conclusions. We have presented a new method for performing FFTs on parallel computers which scales to a larger number of nodes than the traditional method due to the reduced latency ... The new method automatically satsifies the demand of load -
Bibliography
www.tcm.phy.cam.ac.uk/~pdh1001/papers/paper4/node10.html7 Jan 2000: Turnbull, eds., Solid State Physics Vol. 24 (Academic Press, New York, 1970) 1-36. -
3. Traditional parallel implementation
www.tcm.phy.cam.ac.uk/~pdh1001/papers/paper4/node3.html7 Jan 2000: Next: 4. New parallel implementation Up: Parallel fast Fourier transforms Previous: 2. ... Figure 1: Distribution of data for traditional implementation. Next: 4. New parallel implementation Up: Parallel fast Fourier transforms Previous: 2. -
2. Formulation of the problem
www.tcm.phy.cam.ac.uk/~pdh1001/papers/paper5/node2.html11 Jul 2000: 8). When Eq. (7) is solved, a new output electron density. ... is obtained and a new input electron density for the next iteration can be constructed by a linear (or more sophisticated [18]) mixing scheme e.g. -
7. Load balancing
www.tcm.phy.cam.ac.uk/~pdh1001/papers/paper4/node7.html7 Jan 2000: We note here that the new method automatically satisifies the demand of load balancing. ... In fact, the cutoff sphere on the full FFT grid is transformed into a cutoff sphere on the FFT grid on each node in the new method. -
6. Results
www.tcm.phy.cam.ac.uk/~pdh1001/papers/paper4/node6.html7 Jan 2000: 5, we would not expect this machine to give particularly favourable results for the new method because its latency cost is small. ... On the machine used here, the new method is out-performed by the old method for the reasons given above, and would only -
5. Cost comparison
www.tcm.phy.cam.ac.uk/~pdh1001/papers/paper4/node5.html7 Jan 2000: Next: 6. Results Up: Parallel fast Fourier transforms Previous: 4. New parallel implementation.. ... Next: 6. Results Up: Parallel fast Fourier transforms Previous: 4. New parallel implementation. -
1. Introduction
www.tcm.phy.cam.ac.uk/~pdh1001/papers/paper4/node1.html7 Jan 2000: In this paper, we present a new method for performing FFTs on parallel computers which minimises the latency cost and thus offers the prospect of scaling plane-wave electronic structure calculations ... 2 we outline those aspects of the derivation of the
Search history
Recently clicked results
Recently clicked results
Your click history is empty.
Recent searches
Recent searches
Your search history is empty.