[Mesa-dev] [PATCH] register_allocate: don't consider trivially colorable registers for spilling.

[Matt Turner]

On Mon, Oct 1, 2012 at 3:43 PM, Paul Berry <stereotype441 at gmail.com> wrote:
> Previously, we considered all registers as candidates for spilling.
> This was counterproductive--for any registers that have already been
> removed from the interference graph, there is no benefit to spilling
> them, since they don't contribute to register pressure.
>
> This patch ensures that we will only try to spill registers that are
> still in the interference graph after register allocation has failed.
>
> This is consistent with the recommendations of the paper "Retargetable
> Graph-Coloring Register Allocation for Irregular Architectures", on
> which our register allocator is based.
> ---
>
> Note: I haven't found a shader whose performance is significantly
> improved by this patch (it seems that usually, our heuristics happen
> to select registers that are still in the interference graph), but it
> still seems justified on theoretical grounds.
>
>  src/mesa/program/register_allocate.c | 7 +++++++
>  1 file changed, 7 insertions(+)
>
> diff --git a/src/mesa/program/register_allocate.c b/src/mesa/program/register_allocate.c
> index 97d4e33..88793db 100644
> --- a/src/mesa/program/register_allocate.c
> +++ b/src/mesa/program/register_allocate.c
> @@ -555,6 +555,13 @@ ra_get_best_spill_node(struct ra_graph *g)
>        if (cost <= 0.0)
>          continue;
>
> +      /* Only consider registers for spilling if they are still in the
> +       * interference graph (those on the stack have already been proven to be
> +       * allocatable without spilling).
> +       */
> +      if (g->nodes[n].in_stack)
> +         continue;
> +
>        benefit = ra_get_spill_benefit(g, n);
>
>        if (benefit / cost > best_benefit) {
> --
> 1.7.12.1

This is indeed what the paper says, and makes sense to me.

Reviewed-by: Matt Turner <mattst88 at gmail.com>