I found the following behaviour very unintuitive :
R128=RealField(128)
pi_r128=R128.pi()
print pi_r128*2
print pi_r128*2.
print RR(pi)*2
The corresponding output is :
6.2831853071795864769252867665590057684
6.28318530717959
6.28318530717959
The first line was the expected behavior.
The second line was not expected because floating number 2. has the default precision contrary to pi_r128 whose is precision is wider. As the 3rd line shows, the 128 bit precision is completely ignored from pi_r128*2 calculation.
So I was wondering how conversion between RealField numbers was done.