use Inline C => <<'END_OF_C_CODE';

void p_process_2(SV *svv) {
    char prev = 0;
    long count = 0;
    long i=0;
    long len = SvCUR(svv);
    AV *array;
    AV *list;
    char *s = SvPV(svv, len);
    
    Inline_Stack_Vars;
    Inline_Stack_Reset;
    
    list = newAV();
    av_extend(list, 500);

    prev = *s;

    for (i=0; i<len; i++) {
        if (prev == *s) { 
            count++;
        } else {
            array = newAV(); 
            av_push(array,newSVpvn(&prev,1));
            av_push(array,newSViv(i-count));
            av_push(array,newSViv(count));
            av_push(list,newRV_inc(array));
            prev = *s;
            count=1;
        }
              
        s++;
    }
    Inline_Stack_Push(newRV_inc(list));
    Inline_Stack_Done;
}

END_OF_C_CODE
[download]

Featuring no problems with null chars (binary clean), array pre-extension and the fastest performance so far :)

Its kinda like golf with benchmarks.