Doubt on Group by PIPE

< Just a Question >

I have a large table around 47 million rows on which I am doing  a group by on set of columns.. due to nature of data and count(*) group by , I expect less reduce in rows...as final output....may be to half....

< explain says as Group by PIPE, as I have sorted the columns>

My doubt 

As per documentation...
a) does  group by pipe..never do spill to disk ? 
b) Do the output computed  only in memory ?

In this case , result dataset if large and may not fin to memory.