The Kalray MPPA-256 processor is based on a recent low-energy manycore architecture. In this article, we investigate its performance in multiprecision arithmetic for number-theoretic applications. We have developed a library for modular arithmetic that takes full advantage of the particularities of this architecture. This is in turn used in an implementation of the ECM, an algorithm for integer factorization using el-liptic curves. For parameters corresponding to a crypt-analytic context, our implementation compares well to state-of-the-art implementations on GPU, while using much less energy.