GCC can apparently execute some benchmarks in 0.00 ns/op, probably by means
of eliminating the benchmark loop. This caused problems for our R script
because it takes the log of the ratio, and of course log(0) = -Inf. This
change simply modifies the script to ignore zeros in its input.
GCC can apparently execute some benchmarks in 0.00 ns/op, probably by means of eliminating the benchmark loop. This caused problems for our R script because it takes the log of the ratio, and of course log(0) = -Inf. This change simply modifies the script to ignore zeros in its input.