一个用group bound优化timing path的例子

有一条timing path的其中一段如下所示：

...
buf_SMC_2/A (THHBUFFXT)     0.013 *   21.471 f    (1126.96,3018.00)  
buf_SMC_2/Y (THHBUFFXT)     0.263     21.735 f    (1130.33,3018.60)  
buf_SMC_N_2 (net)           0.000     21.735 f    # dont_touch,long wire
buf_SMC_1/A (THHBUFFXT)     9.619 *   31.353 f    (1279.84,-4059.60) 
buf_SMC_1/Y (THHBUFFXT)     1.145     32.499 f    (1283.21,-4060.20) 
buf_SMC_N_1 (net)           0.000     32.499 f    # dont_touch
buf_SMC/A (THHBUFFXT)       0.006 *   32.504 f    (1364.60,-4071.60) 
buf_SMC/Y (THHBUFFXT)       0.121     32.625 f    (1367.97,-4071.00) 
...

一个buffer放置在design的顶部，然后纵向“长途跋涉”7000nm多，穿过一个thin channel来到design的底部，造成了9ns多的delay。而且由于net被标注为dont_touch，不能进行优化。

好的布局(placement)结果应该是把3个buffer放在一起。

这固然是因为工具没有足够“聪明”，我们也可以通过更好的constraints来引导工具做得更好。

create_bounds -name buf_SMC -effort medium  [get_cell {buf_SMC_2 buf_SMC_1 buf_SMC}]