in 2026b the last line of nrPolarEncode has he following
% Encode using indexing. This is equivalent to mod(u’*G,2)’
outd = mod(sum(G(logical(u),:)),2)’;
However this is not equivalent to mod(u’*G,2) which is what was in 2025b.
For e.g. if u = zeros(1,N) and u(1) = 1; Then G(logical(u),:) has dim 1xN. sum(G(logical(u),:)) reduces to a scalar and outd is reduced to a scalar. but mod(u’*g,2) in the original version retians the shape of 1,N. so this is causing a specific problem for certain test vectors. What am i missing?in 2026b the last line of nrPolarEncode has he following
% Encode using indexing. This is equivalent to mod(u’*G,2)’
outd = mod(sum(G(logical(u),:)),2)’;
However this is not equivalent to mod(u’*G,2) which is what was in 2025b.
For e.g. if u = zeros(1,N) and u(1) = 1; Then G(logical(u),:) has dim 1xN. sum(G(logical(u),:)) reduces to a scalar and outd is reduced to a scalar. but mod(u’*g,2) in the original version retians the shape of 1,N. so this is causing a specific problem for certain test vectors. What am i missing? in 2026b the last line of nrPolarEncode has he following
% Encode using indexing. This is equivalent to mod(u’*G,2)’
outd = mod(sum(G(logical(u),:)),2)’;
However this is not equivalent to mod(u’*G,2) which is what was in 2025b.
For e.g. if u = zeros(1,N) and u(1) = 1; Then G(logical(u),:) has dim 1xN. sum(G(logical(u),:)) reduces to a scalar and outd is reduced to a scalar. but mod(u’*g,2) in the original version retians the shape of 1,N. so this is causing a specific problem for certain test vectors. What am i missing? r2026 nr toolbox bug MATLAB Answers — New Questions