[Mesa-dev] [PATCH 12/12] intel/nir: Enable nir_opt_find_array_copies

[Jason Ekstrand]

We have to be a bit careful with this one because we want it to run in
the optimization loop but only in the first brw_nir_optimize call.
Later calls assume that we've lowered away copy_deref instructions and
we don't want to introduce any more.
---
 src/intel/compiler/brw_nir.c | 16 +++++++++-------
 src/intel/compiler/brw_nir.h |  3 ++-
 2 files changed, 11 insertions(+), 8 deletions(-)

diff --git a/src/intel/compiler/brw_nir.c b/src/intel/compiler/brw_nir.c
index 506051d6419..ce701f854af 100644
--- a/src/intel/compiler/brw_nir.c
+++ b/src/intel/compiler/brw_nir.c
@@ -533,7 +533,7 @@ brw_nir_no_indirect_mask(const struct brw_compiler *compiler,
 
 nir_shader *
 brw_nir_optimize(nir_shader *nir, const struct brw_compiler *compiler,
-                 bool is_scalar)
+                 bool is_scalar, bool allow_copies)
 {
    nir_variable_mode indirect_mask =
       brw_nir_no_indirect_mask(compiler, nir->info.stage);
@@ -544,6 +544,8 @@ brw_nir_optimize(nir_shader *nir, const struct brw_compiler *compiler,
       OPT(nir_split_array_vars, nir_var_local);
       OPT(nir_narrow_vec_vars, nir_var_local);
       OPT(nir_lower_vars_to_ssa);
+      if (allow_copies)
+         OPT(nir_opt_find_array_copies);
       OPT(nir_opt_copy_prop_vars);
 
       if (is_scalar) {
@@ -664,7 +666,7 @@ brw_preprocess_nir(const struct brw_compiler *compiler, nir_shader *nir)
                         nir_lower_isign64 |
                         nir_lower_divmod64);
 
-   nir = brw_nir_optimize(nir, compiler, is_scalar);
+   nir = brw_nir_optimize(nir, compiler, is_scalar, true);
 
    /* This needs to be run after the first optimization pass but before we
     * lower indirect derefs away
@@ -701,7 +703,7 @@ brw_preprocess_nir(const struct brw_compiler *compiler, nir_shader *nir)
    nir_lower_indirect_derefs(nir, indirect_mask);
 
    /* Get rid of split copies */
-   nir = brw_nir_optimize(nir, compiler, is_scalar);
+   nir = brw_nir_optimize(nir, compiler, is_scalar, false);
 
    OPT(nir_remove_dead_variables, nir_var_local);
 
@@ -730,11 +732,11 @@ brw_nir_link_shaders(const struct brw_compiler *compiler,
 
       const bool p_is_scalar =
          compiler->scalar_stage[(*producer)->info.stage];
-      *producer = brw_nir_optimize(*producer, compiler, p_is_scalar);
+      *producer = brw_nir_optimize(*producer, compiler, p_is_scalar, false);
 
       const bool c_is_scalar =
          compiler->scalar_stage[(*producer)->info.stage];
-      *consumer = brw_nir_optimize(*consumer, compiler, c_is_scalar);
+      *consumer = brw_nir_optimize(*consumer, compiler, c_is_scalar, false);
    }
 }
 
@@ -761,7 +763,7 @@ brw_postprocess_nir(nir_shader *nir, const struct brw_compiler *compiler,
       OPT(nir_opt_algebraic_before_ffma);
    } while (progress);
 
-   nir = brw_nir_optimize(nir, compiler, is_scalar);
+   nir = brw_nir_optimize(nir, compiler, is_scalar, false);
 
    if (devinfo->gen >= 6) {
       /* Try and fuse multiply-adds */
@@ -857,7 +859,7 @@ brw_nir_apply_sampler_key(nir_shader *nir,
 
    if (nir_lower_tex(nir, &tex_options)) {
       nir_validate_shader(nir);
-      nir = brw_nir_optimize(nir, compiler, is_scalar);
+      nir = brw_nir_optimize(nir, compiler, is_scalar, false);
    }
 
    return nir;
diff --git a/src/intel/compiler/brw_nir.h b/src/intel/compiler/brw_nir.h
index 7d82edafe46..0193457c577 100644
--- a/src/intel/compiler/brw_nir.h
+++ b/src/intel/compiler/brw_nir.h
@@ -155,7 +155,8 @@ bool brw_nir_opt_peephole_ffma(nir_shader *shader);
 
 nir_shader *brw_nir_optimize(nir_shader *nir,
                              const struct brw_compiler *compiler,
-                             bool is_scalar);
+                             bool is_scalar,
+                             bool allow_copies);
 
 #define BRW_NIR_FRAG_OUTPUT_INDEX_SHIFT 0
 #define BRW_NIR_FRAG_OUTPUT_INDEX_MASK INTEL_MASK(0, 0)
-- 
2.17.1