Comparing

OSQP PARSDMM Current GroptSDMM
\(\tilde{z}\) \(s\) \(s\)
\(\alpha\tilde{z} + (1-\alpha)z\) \(\bar{x} = \gamma s + (1-\gamma)y\) \(\gamma s + (1-\gamma)z\)
\(z\) \(y\) \(z\)
\(y\) \(-v\) \(y\)
\(\alpha\) \(\gamma\) \(\gamma,\gamma_x\)

OSQP: Single \(\alpha\) value used for overrelaxation of x and primal updates
PARSDMM: Constraint specific \(\gamma\) for primal update, and nothing on x


Both \(\rho\) are the same thing, and both are constriant specific (even though OSQP doesnt show it in algorithm)