[Mesa-dev] [PATCH 2/3] nir/dominance: Add a constant-time mechanism for comparing blocks

[Connor Abbott]

On Fri, Feb 6, 2015 at 5:12 PM, Jason Ekstrand <jason at jlekstrand.net> wrote:
> This is mostly thanks to Connor.  The idea is to do a depth-first search
> that computes pre and post indices for all the blocks.  We can then figure
> out if one block dominates another in constant time by two simple
> comparison operations.
> ---
>  src/glsl/nir/nir.h           |  9 +++++++++
>  src/glsl/nir/nir_dominance.c | 27 +++++++++++++++++++++++++++
>  2 files changed, 36 insertions(+)
>
> diff --git a/src/glsl/nir/nir.h b/src/glsl/nir/nir.h
> index 886dcd2..5d84343 100644
> --- a/src/glsl/nir/nir.h
> +++ b/src/glsl/nir/nir.h
> @@ -1135,6 +1135,14 @@ typedef struct nir_block {
>     /* Set of nir_block's on the dominance frontier of this block */
>     struct set *dom_frontier;
>
> +   /*
> +    * These two indices have the property that dom_{pre,post}_index for each
> +    * child of this block in the dominance tree will always be between
> +    * dom_pre_index and dom_post_index for this block, which makes testing if
> +    * a given block is dominated by another block an O(1) operation.
> +    */
> +   unsigned dom_pre_index, dom_post_index;
> +
>     /* live in and out for this block; used for liveness analysis */
>     BITSET_WORD *live_in;
>     BITSET_WORD *live_out;
> @@ -1501,6 +1509,7 @@ void nir_calc_dominance_impl(nir_function_impl *impl);
>  void nir_calc_dominance(nir_shader *shader);
>
>  nir_block *nir_dominance_lca(nir_block *b1, nir_block *b2);
> +bool nir_block_dominates(nir_block *parent, nir_block *child);
>
>  void nir_dump_dom_tree_impl(nir_function_impl *impl, FILE *fp);
>  void nir_dump_dom_tree(nir_shader *shader, FILE *fp);
> diff --git a/src/glsl/nir/nir_dominance.c b/src/glsl/nir/nir_dominance.c
> index 1022692..76508f5 100644
> --- a/src/glsl/nir/nir_dominance.c
> +++ b/src/glsl/nir/nir_dominance.c
> @@ -177,6 +177,17 @@ calc_dom_children(nir_function_impl* impl)
>     nir_foreach_block(impl, block_add_child, NULL);
>  }
>
> +static void
> +calc_dfs_indicies(nir_block *block, unsigned *index)
> +{
> +   block->dom_pre_index = (*index)++;
> +
> +   for (unsigned i = 0; i < block->num_dom_children; i++)
> +      calc_dfs_indicies(block->dom_children[i], index);
> +
> +   block->dom_post_index = (*index)++;
> +}
> +
>  void
>  nir_calc_dominance_impl(nir_function_impl *impl)
>  {
> @@ -201,6 +212,9 @@ nir_calc_dominance_impl(nir_function_impl *impl)
>     impl->start_block->imm_dom = NULL;
>
>     calc_dom_children(impl);
> +
> +   unsigned dfs_index = 0;
> +   calc_dfs_indicies(impl->start_block, &dfs_index);
>  }
>
>  void
> @@ -224,6 +238,19 @@ nir_dominance_lca(nir_block *b1, nir_block *b2)
>     return intersect(b1, b2);
>  }
>
> +bool
> +nir_block_dominates(nir_block *parent, nir_block *child)
> +{
> +   assert(nir_cf_node_get_function(&parent->cf_node) ==
> +          nir_cf_node_get_function(&child->cf_node));
> +
> +   assert(nir_cf_node_get_function(&parent->cf_node)->valid_metadata &
> +          nir_metadata_dominance);
> +
> +   return child->dom_pre_index >= parent->dom_pre_index &&
> +          child->dom_pre_index <= parent->dom_post_index;

It doesn't matter too much, but I think doing child->dom_post_index <=
parent->dom_post_index will make it less likely for people to think
it's a bug (I did at first). Other than that,

Reviewed-by: Connor Abbott <cwabbott0 at gmail.com>

> +}
> +
>  static bool
>  dump_block_dom(nir_block *block, void *state)
>  {
> --
> 2.2.2
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/mesa-dev