The BGR to Lab documentation mentions that
The equations above are given on the assumption that initial B, G, R values are normalized to the range [0..1]. The computed values of the L component are in the range [0..100], a and b component values are in the range [-128..127].
Is this intended to mean that the conversion functions cannot be used if the floating point data does not lie in that range? Either way, these calls on my system using the edition of IPP that shipped with composer_xe_2015.1.133 cause a segmentation fault if the floating point data does not lie in that range.
you are right - the input data MUST be normalized as it is mentioned in the manual. Internal algorithm uses table-based approach - so non-normalized data leads to hits beyond the table boundaries and access violation (segfault) as a result.