[Mesa-dev] [PATCH v2 1/8] nir: evaluate if condition uses inside the if branches

[Timothy Arceri]

On 23/07/18 18:02, Timothy Arceri wrote:
> Since we know what side of the branch we ended up on we can just
> replace the use with a constant.
> 
> All the spill changes in shader-db are from Dolphin uber shaders,
> despite some small regressions the change is clearly positive.
> 
> shader-db results IVB:
> 
> total instructions in shared programs: 9999201 -> 9993483 (-0.06%)
> instructions in affected programs: 163235 -> 157517 (-3.50%)
> helped: 132
> HURT: 2
> 
> total cycles in shared programs: 231670754 -> 219476091 (-5.26%)
> cycles in affected programs: 143424120 -> 131229457 (-8.50%)
> helped: 115
> HURT: 24
> 
> total spills in shared programs: 4383 -> 4370 (-0.30%)
> spills in affected programs: 1656 -> 1643 (-0.79%)
> helped: 9
> HURT: 18
> 
> total fills in shared programs: 4610 -> 4581 (-0.63%)
> fills in affected programs: 374 -> 345 (-7.75%)
> helped: 6
> HURT: 0
> ---
>   src/compiler/nir/nir_opt_if.c | 124 ++++++++++++++++++++++++++++++++++
>   1 file changed, 124 insertions(+)
> 
> diff --git a/src/compiler/nir/nir_opt_if.c b/src/compiler/nir/nir_opt_if.c
> index b3d0bf1decb..b3d5046a76e 100644
> --- a/src/compiler/nir/nir_opt_if.c
> +++ b/src/compiler/nir/nir_opt_if.c
> @@ -369,6 +369,87 @@ opt_if_loop_terminator(nir_if *nif)
>      return true;
>   }
>   
> +static void
> +replace_if_condition_use_with_const(nir_src *use, unsigned nir_boolean,
> +                                    void *mem_ctx, bool if_condition)
> +{
> +   /* Create const */
> +   nir_load_const_instr *load = nir_load_const_instr_create(mem_ctx, 1, 32);
> +   load->value.u32[0] = nir_boolean;
> +
> +   if (if_condition) {
> +      nir_instr_insert_before_cf(&use->parent_if->cf_node,  &load->instr);
> +   } else if (use->parent_instr->type == nir_instr_type_phi) {
> +      nir_phi_instr *cond_phi = nir_instr_as_phi(use->parent_instr);
> +
> +      bool UNUSED found = false;
> +      nir_foreach_phi_src(phi_src, cond_phi) {
> +         if (phi_src->src.ssa == use->ssa) {
> +            nir_instr_insert_before_block(phi_src->pred, &load->instr);

I've just noticed this needs to be nir_instr_insert_after_block() 
otherwise we can end up inserting the const before so other phi which 
trips up validation.

I've fixed this locally.