[Mesa-dev] [PATCH 4/4] nir: Make nir_opt_remove_phis see through moves.
Iago Toral
itoral at igalia.com
Fri Jul 29 09:42:36 UTC 2016
On Fri, 2016-07-29 at 01:29 -0700, Kenneth Graunke wrote:
> I found a shader in Tales of Maj'Eyal that contains:
>
> if ssa_21 {
> block block_1:
> /* preds: block_0 */
> ...instructions that prevent the select peephole...
> vec1 32 ssa_23 = imov ssa_4
> vec1 32 ssa_24 = imov ssa_4.y
> vec1 32 ssa_25 = imov ssa_4.z
> /* succs: block_3 */
> } else {
> block block_2:
> /* preds: block_0 */
> vec1 32 ssa_26 = imov ssa_4
> vec1 32 ssa_27 = imov ssa_4.y
> vec1 32 ssa_28 = imov ssa_4.z
> /* succs: block_3 */
> }
> block block_3:
> /* preds: block_1 block_2 */
> vec1 32 ssa_29 = phi block_1: ssa_23, block_2: ssa_26
> vec1 32 ssa_30 = phi block_1: ssa_24, block_2: ssa_27
> vec1 32 ssa_31 = phi block_1: ssa_25, block_2: ssa_28
>
> Here, copy propagation will bail because phis cannot perform
> swizzles,
> and CSE won't do anything because there is no dominance relationship
> between the imovs. By making nir_opt_remove_phis handle identical
> moves,
> we can eliminate the phis and rewrite everything to use ssa_4
> directly,
> so all the moves become dead and get eliminated.
>
> I don't think we need to check "exact" - just the alu sources.
> Presumably phi sources should match in their exactness.
>
> On Broadwell:
>
> total instructions in shared programs: 11639872 -> 11638535 (-0.01%)
> instructions in affected programs: 134222 -> 132885 (-1.00%)
> helped: 338
> HURT: 0
>
> Signed-off-by: Kenneth Graunke <kenneth at whitecape.org>
> ---
> src/compiler/nir/nir_opt_remove_phis.c | 24 +++++++++++++++++++++++-
> 1 file changed, 23 insertions(+), 1 deletion(-)
>
> diff --git a/src/compiler/nir/nir_opt_remove_phis.c
> b/src/compiler/nir/nir_opt_remove_phis.c
> index f1e7bb0..ebcc856 100644
> --- a/src/compiler/nir/nir_opt_remove_phis.c
> +++ b/src/compiler/nir/nir_opt_remove_phis.c
> @@ -27,6 +27,26 @@
>
> #include "nir.h"
>
> +static nir_alu_instr *
> +get_parent_mov(nir_ssa_def *ssa)
> +{
> + if (ssa->parent_instr->type != nir_instr_type_alu)
> + return false;
You should return NULL instead of false here.
> + nir_alu_instr *alu = nir_instr_as_alu(ssa->parent_instr);
> + return (alu->op == nir_op_imov || alu->op == nir_op_fmov) ? alu :
> NULL;
> +}
> +
> +static bool
> +matching_mov(nir_alu_instr *mov1, nir_ssa_def *ssa)
> +{
> + if (!mov1)
> + return false;
> +
> + nir_alu_instr *mov2 = get_parent_mov(ssa);
> + return mov2 && nir_alu_srcs_equal(mov1, mov2, 0, 0);
> +}
> +
> /*
> * This is a pass for removing phi nodes that look like:
> * a = phi(b, b, b, ...)
> @@ -54,6 +74,7 @@ remove_phis_block(nir_block *block)
> nir_phi_instr *phi = nir_instr_as_phi(instr);
>
> nir_ssa_def *def = NULL;
> + nir_alu_instr *mov = NULL;
> bool srcs_same = true;
>
> nir_foreach_phi_src(src, phi) {
> @@ -75,8 +96,9 @@ remove_phis_block(nir_block *block)
>
> if (def == NULL) {
> def = src->src.ssa;
> + mov = get_parent_mov(def);
> } else {
> - if (src->src.ssa != def) {
> + if (src->src.ssa != def && !matching_mov(mov, src-
> >src.ssa)) {
> srcs_same = false;
> break;
> }
More information about the mesa-dev
mailing list