1: \begin{abstract}

2: 

3: We carry out a high-precision simulation of the

4: two-dimensional $SU(3)$ principal chiral model

5: at correlation lengths $\xi$ up to $\sim\! 4 \times 10^5$,

6: using a multi-grid Monte Carlo (MGMC) algorithm

7: and approximately one year of Cray C-90 CPU time.

8: We extrapolate

9: the finite-volume Monte Carlo data to infinite volume

10: using finite-size-scaling theory,

11: and we discuss carefully the systematic and statistical errors

12: in this extrapolation.

13: We then compare the extrapolated data to the renormalization-group predictions.

14: The deviation from asymptotic scaling,

15: which is $\approx 12\%$ at $\xi \sim 25$,

16: decreases to $\approx 2\%$  at 

17: $\xi \sim 4 \times 10^5$.

18: We also analyze the dynamic critical behavior of the MGMC 

19: algorithm using lattices up to $256 \times 256$,

20: finding the dynamic critical exponent

21: $z_{int,{\cal M}^2} \approx 0.45 \pm 0.02$

22: (subjective 68\% confidence interval).

23: Thus, for this asymptotically free model,

24: critical slowing-down is greatly reduced compared to local algorithms,

25: but not completely eliminated.

26: 

27: \end{abstract}

28: