Verification is currently the slowest part of the LRSW-DAA algorithm. As mentioned in Chen, et al (doi: 10.1007/978-3-642-12510-2_16) this can be sped up using batch verification.
AMCL supports this (it's why the Miller Loop and the exponentiation are split in their implementation). In fact, the test of the pairing in their code includes a batch multiplication.
Chen, et all claim this resulted in a 40% speedup of the verification.