1: \begin{abstract}

2: We develop and implement a parallel flatPERM algorithm \cite{G97,PK04}

3: with mutually interacting parallel flatPERM sequences and use it to 

4: sample self-avoiding walks in 2 and 3 dimensions.  Our data show that 

5: the parallel implementation accelerates the convergence of the 

6: flatPERM algorithm.  Moreover, increasing the number of interacting flatPERM

7: sequences (rather than running longer simulations) improves the rate of

8: convergence.  This suggests that a more efficient implementation 

9: of flatPERM will be a massively parallel implementation, rather than long 

10: simulations of one, or a few parallel sequences.  We also use the algorithm 

11: to estimate the growth constant of the self-avoiding walk in two and in 

12: three dimensions using simulations over 12 parallel sequences.  Our 

13: best results are

14: \[

15: \mu_d = \cases{

16: 2.6381585(1), & \hbox{if $d=2$}; \cr

17: 4.684039(1), & \hbox{if $d=3$}.

18: } 

19: \]

20: \end{abstract}

21: