running on ubuntu20.04, with 16 cores, 32G memory.
tested using the same tracefile(about 700 files, 300k+ lines, 20k+ functions, 100k+ branches), genhtml command:

time genhtml --output-directory $OUTPUT --branch-coverage --prefix $PREFIX coverage.info 

with ubuntu20.04 default lcov 1.14:

real 0m23.972s user 0m22.009s sys 0m1.823s 

with lcov 2.0:

real 1m14.681s user 1m12.299s sys 0m1.970s 

with lcov 2.0(--parallel):

real 2m36.937s user 21m53.142s sys 14m31.093s 

It seems that adding --parallel makes things worse, which consumes more CPU and memory, but have no advantage.(I thought parallelism should be faster?)