In the EOS smart contract, it support the float point computation, such as calculating the voting weight.
And I think that, for different operation systems and different compilers, the float point procession could cause an inconsistent result.
So for different block producers, the data stored in their ram or database will be not same, and to the end, maybe we can't make the +2/3 producers to attrive the final conformation.
Is that right?
Yes, different platforms have inconsistent floating point arithmetic implementations wich could cause problems.
That is why EOS uses a C library which avoids hardware floating point calculations and brings consistency to those who require floating points.
The down side of this decision is of course performance, since software implementations are always slower than hardware solutions.
Using such a library at the core of the EOS system eliminates the need for each dApp developer who needs floating point calculations to include such a library for themselves. This would be wasteful and will force each dApp developer to "invent the wheel" repeatedly. Possibly coming up with different solutions. This is already happening in other dApp platforms which have integer only.
