I am attaching a proof-of-concept patch that changes the bytecode 
generated from {1:1,2:2,3:3,4:4} from 

BUILD_MAP                4  
LOAD_CONST               2 (1)  
LOAD_CONST               2 (1)  
STORE_MAP             
LOAD_CONST               1 (2)  
LOAD_CONST               1 (2)  
STORE_MAP             
LOAD_CONST               4 (3)  
LOAD_CONST               4 (3)  
STORE_MAP             
LOAD_CONST               3 (4)  
LOAD_CONST               3 (4)  
STORE_MAP             

to

LOAD_CONST               1 (4)  
LOAD_CONST               1 (4)  
LOAD_CONST               2 (3)   
LOAD_CONST               2 (3)                                                    
LOAD_CONST               3 (2)                                        
LOAD_CONST               3 (2)                                         
LOAD_CONST               4 (1)  
LOAD_CONST               4 (1)  
BUILD_MAP                4  
 
and changes BUILD_MAP to behave similarly to BUILD_(SET|LIST|TUPLE) and 
consume 2*n items from the stack.

The advantage is more compact and faster bytecode and uniform treatment 
of BUILD_* instructions.