function ip = spinterpm(d,z,y,levelseq,purgedata) % SPINTERPM Multi-linear interpolation (maximum-norm-based grid) % IP = SPINTERPM(N,D,Z,Y) Computes the interpolated % values at [Y1, ..., YN] over the sparse grid at level N. Y % may be a double array for vectorized processing, with each row % representing one point. The sparse grid data must be given as % an array Z containing the node weights (computed with % SPVALS). Note that the sparse grid is always normalized to % the unit cube [0,1]^D, i.e., if the weights have been computed % for a different domain, the values Y have to be rescaled % accordingly. (Internal function) % Author : Andreas Klimke, Universitaet Stuttgart % Version: 1.2 % Date : February 3, 2006 % Change log: % V1.0 : September 24, 2003 % Initial revision % V1.1 : January 24, 2006 % Changed data types to operate on uint arrays % V1.2 : February 3, 2006 % Added droptol processing. % ------------------------------------------------------------ % Sparse Grid Interpolation Toolbox % Copyright (c) 2006 W. Andreas Klimke, Universitaet Stuttgart % Copyright (c) 2007-2008 W. A. Klimke. All Rights Reserved. % See LICENSE.txt for license. % email: klimkeas@ians.uni-stuttgart.de % web : http://www.ians.uni-stuttgart.de/spinterp % ------------------------------------------------------------ ninterp = uint32(size(y,1)); ip = zeros(ninterp,1); % Get the number of levels nlevels = uint32(size(levelseq,1)); if ~isempty(purgedata), purge = true; else purge = false; end % index contains the index of the resulting array containing all % subdomains of the level. index = uint32(1); index2 = zeros(d,1,'uint32'); repvec = ones(d,1,'uint32'); level = ones(d,1,'uint8'); repeat = zeros(d,1); for kl = 1:nlevels npoints = uint32(1); lval = uint8(0); nlevelzero = 0; for k = 1:d lval = levelseq(kl,k); level(k) = lval; if lval == 0 repvec(k) = 3; nlevelzero = nlevelzero + 1; else repvec(k) = 2^uint32(lval); end repeat(k) = 0; npoints = npoints * repvec(k); if k > 1 repvec(k) = repvec(k) * repvec(k-1); end end % Skip subgrids with all surpluses below droptol. if purge if purgedata(kl) == 0 index = index + npoints; continue; end end nrepeats = 2^nlevelzero-1; yt = 0; k = uint32(1); repstep = 0; while k <= ninterp temp = 1; l = uint16(1); while l <= d lval = level(l); yt = y(k,l); % Compute the scaling factor and the array position of % the weight if lval == 0 if repeat(l) == 0 index2(l) = 1; temp = temp * (1 - 2 * abs(yt-0.5)); else if yt == 1 index2(l) = 2; else xp = floor(yt * 2) * 2; if xp == 0 temp = temp * 2 * (0.5 - yt); else temp = temp * 2 * (yt - 0.5); end index2(l) = xp; end end elseif yt == 1 temp = 0; break; else scale = 2^double(lval); xp = floor(yt * scale); temp = temp * ... (1 - 2 * scale * abs( yt - (xp+0.5)/scale)); index2(l) = xp; end l = l + 1; if temp == 0 break; end end % If the scaling factor is not Zero, add the computed value if temp > 0 index3 = index + index2(1); for l = 2:d index3 = index3 + repvec(l-1)*index2(l); end ip(k) = ip(k) + temp*z(index3); end if repstep == nrepeats k = k + 1; if nrepeats > 0 for l = 1:d repeat(l) = 0; end repstep = 0; end else for l = 1:d if level(l) == 0 repeat(l) = repeat(l) + 1; if repeat(l) > 1 repeat(l) = 0; else break; end end end repstep = repstep + 1; end end index = index + npoints; end