error_demo1.py3 1.0e16*((1.0+1.2e-16)-1.0)
only about 16 digits accuracy, power of 10 not exact
ans= 2.220446049250313
 
force store, break optimization, no difference here
ans= 2.220446049250313
 
expression with double ans= 0.0
 
error_demo1.py3 (1.0+1.0e-16)+1.0  too small
ans= 0.0
 
force store, break optimization, no difference here
ans= 0.0