Internal workhorse called by corr_slide(). Computes Pearson correlation
matrices directly on the input matrix without allocating per-window
submatrices or calling R's cor().
Usage
corr_slide_cpp(timeseries, window, step)
Arguments
- timeseries
NumericMatrix [N x Tmax].
- window
Integer window size in timepoints.
- step
Integer step between window onsets.
Value
List with:
- corr_mats
NumericVector with dim [N, N, n_windows].
- idx
IntegerVector of 1-indexed window onset positions.
Details
Loop ordering is chosen for cache efficiency on the column-major R matrix:
the innermost loop over t (timepoints) within a window iterates over
consecutive memory locations in each column of timeseries, so the
cross-product accumulation step is sequential in memory.
