This systemtap probe measures memory cost per require. Be aware that require sum up, so subrequires form the sum of the initial require.

require_mem.stp

global before;
global after;

probe process("ruby").mark("require__entry")
{
    module = kernel_string($arg1)
    mem = proc_mem_size_pid(pid())
    before [module] = mem;
}

probe process("ruby").mark("require__return")
{
    module = kernel_string($arg1)
    mem = proc_mem_size_pid(pid())
    after [module] = mem;
}

probe end {
    foreach ([mod-] in after) {
        printf("%s\t%s\n", pages_to_string(after[mod] - before[mod]), mod);
    }
    delete before;
    delete after;
}