postgresql performance: bigint vs varchar

-- postgresql 13 default
-- ubuntu, core i5, ssd

create table t1 (i bigint, c varchar(20), t text);
create table t2 (i bigint, c varchar(20), t text);

insert into t1 (i,c,t) 
select random() * 1000000, '', '12345678912345678912345678900'
from generate_series(0,1000000);

insert into t2 (i,c,t) 
select random() * 1000000, '', '12345678912345678912345678900'
from generate_series(0,1000000);

update t1 set c=i::varchar;
update t2 set c=i::varchar;

create index t1_i on t1(i);
create index t1_c on t1(c);

create index t2_i on t2(i);
create index t2_c on t2(c);

analyze t1;
analyze t2;


select count(*) from (
  select distinct i from t1
) as d;

-- 164 ms


select count(*) from (
  select distinct c from t1
) as d;

-- 186 ms


select count(*) from (
  select distinct t1.c from t1 join t2 on t1.c = t2.c  
) as d;

-- 825 ms


select count(*) from (
  select distinct t1.i from t1 join t2 on t1.i = t2.i  
) as d;

-- 526 ms


select count(*) from (
  select t1.c from t1 join t2 on t1.c = t2.c  
) as d;

-- 450 ms


select count(*) from (
  select t1.i from t1 join t2 on t1.i = t2.i  
) as d;

-- 420 ms