You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
If the data type long int is 32-bit, the denominator for calculating $P_W$ in LRS estimate overflows, i.e. the binomial coefficient $\binom{L - W + 1}{2}$ is beyond the upper limit of 32-bit taking into account that $L$ is order of 10^6 and actual values of $W$.
As a result, H_original becomes infinity and H_bitstring becimes 0.13 for truerand_8bit.bin.
The text was updated successfully, but these errors were encountered:
Yes, this tool only works on platforms on which "long int" is at least a 64 bit quantity. There are several places where this assumption is made, this is only one of them.
I should mention that the error-free rollover is clearly a bug, but the resolution to this bug would be to trigger an assert on such environments, so it wouldn't resolve in a particularly useful way for such architectures.
Thank you for its investigation. The temporary solution will be to use the proposed code in Windows 10 branch.
Taking into account that the denominator can not be expressed by 32-bit, the numerator (vector S in code) should also be expressed by 64-bit for consistency.
If the data type long int is 32-bit, the denominator for calculating$P_W$ in LRS estimate overflows, i.e. the binomial coefficient $\binom{L - W + 1}{2}$ is beyond the upper limit of 32-bit taking into account that $L$ is order of 10^6 and actual values of $W$ .
As a result, H_original becomes infinity and H_bitstring becimes 0.13 for truerand_8bit.bin.
The text was updated successfully, but these errors were encountered: