Dear Greg, Nathan, Gunnar,
I am finding the real-valued double-prec randutv quite useful (now I have the matlab interface) as a drop-in replacement for SVD in rank-deficient linear systems that need a quasi-minimum-norm LSQ solve, giving a 2x to 4x speedup. Are there any plans to make a complex version? (I don't want to pay 8x cost by faking it in a real-valued 2N sized matrix..). Likewise, single-prec? This seems like something that could be farmed out.
Thanks & best wishes, Alex
Dear Greg, Nathan, Gunnar,
I am finding the real-valued double-prec randutv quite useful (now I have the matlab interface) as a drop-in replacement for SVD in rank-deficient linear systems that need a quasi-minimum-norm LSQ solve, giving a 2x to 4x speedup. Are there any plans to make a complex version? (I don't want to pay 8x cost by faking it in a real-valued 2N sized matrix..). Likewise, single-prec? This seems like something that could be farmed out.
Thanks & best wishes, Alex