Reducing the cost of garbage collecting small arrays

[Johan Tibell]

On Thu, Jun 23, 2011 at 3:27 PM, Jan-Willem Maessen
<jmaessen at alum.mit.edu> wrote:
> Yes, I'd still expect that; internal node churn with fat nodes
> exhausts heap more quickly than usual.  If large nodes become the
> norm, cranking up GC nursery size might be in order.
>
> It's great to see that fat nodes finally work well in GHC, though.
> When I first tried this in the 90's it was so bad (due to the extra
> level of indirection & write barrier problems) I gave up in
> frustration.

I just tested reducing the fanout to 8, which actually makes things slower.

Johan