diff --git a/RandLAPACK/misc/rl_linops.hh b/RandLAPACK/misc/rl_linops.hh
index 43f21fe3..db22cd6e 100644
--- a/RandLAPACK/misc/rl_linops.hh
+++ b/RandLAPACK/misc/rl_linops.hh
@@ -115,16 +115,20 @@ struct RegExplicitSymLinOp : public SymmetricLinearOperator<T> {
     using scalar_t = T;
 
     RegExplicitSymLinOp(
-        int64_t m, const T* A_buff, int64_t lda, vector<T> &arg_regs
+        int64_t m, const T* A_buff, int64_t lda, T* arg_regs, int64_t arg_num_regs
     ) : SymmetricLinearOperator<T>(m), A_buff(A_buff), lda(lda) {
         randblas_require(lda >= m);
         _eval_includes_reg = false;
-        num_regs = arg_regs.size();
+        num_regs = arg_num_regs;
         num_regs = std::max(num_regs, (int64_t) 1);
         regs = new T[num_regs]{};
-        std::copy(arg_regs.begin(), arg_regs.end(), regs);
+        std::copy(arg_regs, arg_regs, regs);
     }
 
+    RegExplicitSymLinOp(
+        int64_t m, const T* A_buff, int64_t lda, vector<T> &arg_regs
+    ) : RegExplicitSymLinOp<T>(m, A_buff, lda, arg_regs.data(), static_cast<int64_t>(arg_regs.size())) {}
+
     ~RegExplicitSymLinOp() {
         if (regs != nullptr) delete [] regs;
     }
@@ -244,11 +248,7 @@ struct SpectralPrecond {
         num_regs = arg_num_regs;
     }
 
-    void prep(vector<T> &eigvecs, vector<T> &eigvals, vector<T> &mus, int64_t arg_num_rhs) {
-        // assume eigvals are positive numbers sorted in decreasing order.
-        int64_t arg_num_regs = mus.size();
-        int64_t arg_dim_pre  = eigvals.size();
-        reset_owned_buffers(arg_dim_pre, arg_num_rhs, arg_num_regs);
+    void set_D_from_eigs_and_regs(T* eigvals, T* mus) {
         for (int64_t r = 0; r < num_regs; ++r) {
             T  mu_r = mus[r];
             T* D_r  = D + r*dim_pre;
@@ -257,6 +257,15 @@ struct SpectralPrecond {
                 D_r[i] = (numerator / (eigvals[i] + mu_r)) - 1.0;
             }
         }
+        return;
+    }
+
+    void prep(vector<T> &eigvecs, vector<T> &eigvals, vector<T> &mus, int64_t arg_num_rhs) {
+        // assume eigvals are positive numbers sorted in decreasing order.
+        int64_t arg_num_regs = mus.size();
+        int64_t arg_dim_pre  = eigvals.size();
+        reset_owned_buffers(arg_dim_pre, arg_num_rhs, arg_num_regs);
+        set_D_from_eigs_and_regs(eigvals.data(), mus.data());
         std::copy(eigvecs.begin(), eigvecs.end(), V);
         return;
     }