GSK provide extensive disclosures about how they test goodwill for impairment. However, it appears that these tests are done at the segment level which means a total of three tests for the whole group.
GSK uses fair value less costs to sell rather than the value in use methodology to determine the recoverable amount of their CGUs. IAS 36 defines recoverable amount as the higher of the two measures but if one, such as the fair value approach, produces no impairment then there is no requirement to calculate both.
Fair value less costs to sell often works out to be higher than value in use because the latter is restricted by not being able to take into account the value of future investment.
GSK aggregates individual cash generating units to produce the 3 combined CGUs that are tested. We assume this is because goodwill cannot be allocated to the individual CGUs in a non-arbitrary manner.