Risto Honkanen

Constant coefficients linear prediction for lossless compression of ultraspectral sounder data using a graphics processing unit

The amount of data generated by ultraspectral sounders is so large that considerable savings in data storage and transmission bandwidth can be achieved using data compression. Due to this large amount of data, the data compression time is of utmost importance. Increasing the programmability of the commodity Graphics Processing Units (GPUs) offer potential for considerable increases in computation speeds in applications that are data parallel. In our experiments, we implemented a spectral image data compression method called Linear Prediction with Constant Coefficients (LP-CC) using NVIDIAs CUDA parallel computing architecture. LP-CC compression method represents a current state-of-the-art technique in lossless compression of ultraspectral sounder data. The method showed an average compression ratio of 3.39 when applied to publicly available NASA AIRS data. We achieved a speed-up of 86 compared to a single threaded CPU version. Thus, the commodity GPU was able to significantly decrease the computational time of a compression algorithm based on a constant coefficient linear prediction.

Hot-potato routing algorithms for sparse optical torus

In this paper we present an optical network architecture and deflection (or hot potato) routing algorithms supporting efficient communication between n processor nodes in a shared memory parallel computer. The sparse optical torus network consists of an n/spl times/n torus, where processor nodes are situated diagonally, and routing nodes are optical deflection nodes of two inputs and two outputs. A design of optical deflection node is presented. Several routing algorithms, based on the greedy routing algorithm, are developed. By experiments and partial theoretical analyses these algorithms run efficiently on sparse optical torus.

Address-free all-to-all routing in sparse torus

In this work we present a simple network design for all-to-all routing and study deflection routing on it. We present a time-scheduled routing algorithm where packets are routed address-free. We show that a total exchange relation, where every processor has a packet to route to every other processor, can be routed with routing cost of 1/2 + o(1) time units per packet. The network consists of an n-sided d-dimensional torus, where the nd-1 processor (or input/output) nodes are sparsely but regularly situated among nd - nd-1 deflection routing nodes, having d input and d output links. The finite-state routing nodes change their states by a fixed, preprogrammed pattern.

Work-optimal routing in wavelength-division multiplexed three-dimensional dense optical tori

In this paper, we present an all-optical network architecture and an all-optical router for it. The dense 3-dimensional optical torus network (WDOT) consists of an n x n x n torus, each node having a processor. The number of processors of the network is P = n3 and the number of optical links is L = 3n3. Routing is based on the scheduled transmission of packets and wavelength-division multiplexing. The routing protocol ensures that no electro-optical conversion is needed at the intermediate nodes and all the packets injected into the routing machinery reach their targets without collisions. A work-optimal routing of h-relations is achieved with a reasonable size of h ∈ Θ(P log P).

Nearly-All-Optical Routing in Sparse Optical Tori

In this paper we present an all-optical network architecture and a nearly-all-optical router for it. The sparse optical torus network consists of an ntimesn torus, where processors are deployed diagonally. Addresses of packets are encoded and recognized by using fiber Bragg grating arrays. The optical address recognition ensures that only a few logical gates are needed to implement routing decisions at the routing nodes. A work-optimal routing of h-relation is achieved with a reasonable size of hisinOmega(n log n)

Work-optimal two-phase routing in a sparse optical torus

In this paper we present an all-optical network architecture and a two-phase routing protocol for it. The layered sparse optical torus (LSOT) of degree d consists of n x n nodes at the layer 1, where n = d2. Processors are deployed diagonally at each ith diagonal. The overall number of processors is then P = d3. Additionally a LSOT consists of d2 d x d subnetworks at the layer 2. Routing is based on the scheduled transmission of packets and wavelength-division multiplexing. The routing protocol ensures that no electro-optical conversion is needed at the intermediate nodes and all the packets injected into the routing machinery reach their targets without collisions. A work optimal routing of h-relation is achieved with the size of h ε θ(P).

Balanced WDM and TDM Routing in Coloured Sparse Optical Tori

In this paper, we present an all-optical network architecture and a routing protocol for it. The coloured sparse optical torus network (CSOT ) consists of an n × n torus for which n = b^2 for some b in {2, 3, 4, . . .}. Processors of the network are deployed at nodes for which (i + j) mod b = 0, where i and j are row and column indices of a node andb is the block size and the number of wavelengths used. The number of processors is P = b3. Routing is based on scheduled transmission of packets and wavelength-division multiplexing. The routing protocol ensures that no electro-optical conversionis needed at the intermediate nodes and all the packets injected into the routing machinery reach their targets without collisions. A work-optimal routing of h-relations is achieved fora reasonable size of h in (P log P).

Lambda-Systolic Routing in a Wavelength-Division Multiplexed All-Optical Butterfly

In this paper we present an all-optical network architecture and a routing protocol for it. A 2r-dimensional coloured optical butterfly (COBF) network consists of P =

Lambda-Systolic Routing in a Dense Optical Torus.

2^{2r}

Comparison of nearest neighbour and neural network based classifications of patient's activity

processors, N =