performance question

i am using hyperscan in BLOCK mode, compiled about 4k patterns in one database by using hs_compile_multi.

about 400 patterns are real regex, and about 3600 patterns are pure literal like abc

to improve the scan performance, will it be helpful if i split 4k pattern into two databases?
that is, for 3600 patterns( pure literal ), use hs_compile_lit_multi to compile, and for 400 patterns(real regex), use hs_compile_multi.

and hs_scan 2 times. is it helpful? thanks.