Our subject is the class U of all positive solutions of a semilinear equation L u = ψ ( u ) in E where L is a second order elliptic differential operator, E is a domain in R d and ψ belongs to a convex class Ψ of C 1 functions which contains functions ψ ( u ) = u α with α > 1 . A special role is played by a class U 0 of solutions which we call σ-moderate. A solution u is moderate if there exists h ⩾ u such that L h = 0 in E. We say that u ∈ U is σ-moderate if u is the limit of an increasing sequence of moderate solutions. In [E.B. Dynkin, S.E. Kuznetsov, Fine topology and fine trace on the boundary associated with a class of quasilinear differential equations, Comm. Pure Appl. Math. 51 (1998) 897–936] all σ-moderate solutions were classified by using their fine boundary traces. 2 2 It is known (see [B. Mselati, Classification and probabilistic representation of the positive solutions of a semilinear elliptic equation, Mem. Amer. Math. Soc. 168 (798) (2004); E.B. Dynkin, Superdiffusions and Positive Solutions of Nonlinear Partial Differential Equations, Amer. Math. Soc., Providence, RI, 2004]) that U 0 = U in the case of the equation Δ u = u α with 1 < α ⩽ 2 in a bounded smooth domain. Therefore in this case the description of all solutions of L u = ψ ( u ) follows from the results on σ-moderate solutions. In [M. Marcus, L. Véron, The precise boundary trace of positive solutions of the equation Δ u = u q in the supercritical case, in: Perspectives in Nonlinear Partial Differential Equations, in honor of Haim Brezis, Contemp. Math., Amer. Math. Soc., Providence, RI, 2007, arxiv.org/math/0610102] Marcus and Véron introduced a different characteristic (called the precise trace) for solutions of the equation Δ u = u α with α > 1 in a bounded C 2 domain. In the present paper we develop a general scheme covering both approaches and we prove the equivalence, in a certain sense, of the fine and precise traces. An implication of this equivalence is a Wiener type criterion for vanishing of the Poisson kernel of the equation L u ( x ) = a ( x ) u ( x ) .