aco: use io semantics to get an intrinsic's slot

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6689>

diff --git a/src/amd/compiler/aco_instruction_selection.cpp b/src/amd/compiler/aco_instruction_selection.cpp
index 0f87c94..101bcf3 100644 (file)
--- a/src/amd/compiler/aco_instruction_selection.cpp
+++ b/src/amd/compiler/aco_instruction_selection.cpp
@@ -4096,14 +4096,13 @@ std::pair<Temp, unsigned> get_tcs_per_patch_output_vmem_offset(isel_context *ctx
    return offs;
 }
 
-bool tcs_driver_location_matches_api_mask(isel_context *ctx, nir_intrinsic_instr *instr, bool per_vertex, uint64_t mask, bool *indirect)
+bool tcs_compare_intrin_with_mask(isel_context *ctx, nir_intrinsic_instr *instr, bool per_vertex, uint64_t mask, bool *indirect)
 {
    assert(per_vertex || ctx->shader->info.stage == MESA_SHADER_TESS_CTRL);
 
    if (mask == 0)
       return false;
 
-   unsigned drv_loc = nir_intrinsic_base(instr);
    nir_src *off_src = nir_get_io_offset_src(instr);
 
    if (!nir_src_is_const(*off_src)) {
@@ -4112,9 +4111,10 @@ bool tcs_driver_location_matches_api_mask(isel_context *ctx, nir_intrinsic_instr
    }
 
    *indirect = false;
-   uint64_t slot = per_vertex
-                   ? ctx->output_drv_loc_to_var_slot[ctx->shader->info.stage][drv_loc / 4]
-                   : (ctx->output_tcs_patch_drv_loc_to_var_slot[drv_loc / 4] - VARYING_SLOT_PATCH0);
+   uint64_t slot = nir_intrinsic_io_semantics(instr).location;
+   if (!per_vertex)
+      slot -= VARYING_SLOT_PATCH0;
+
    return (((uint64_t) 1) << slot) & mask;
 }
 
@@ -4179,7 +4179,7 @@ void visit_store_ls_or_es_output(isel_context *ctx, nir_intrinsic_instr *instr)
    if (ctx->tcs_in_out_eq && store_output_to_temps(ctx, instr)) {
       /* When the TCS only reads this output directly and for the same vertices as its invocation id, it is unnecessary to store the VS output to LDS. */
       bool indirect_write;
-      bool temp_only_input = tcs_driver_location_matches_api_mask(ctx, instr, true, ctx->tcs_temp_only_inputs, &indirect_write);
+      bool temp_only_input = tcs_compare_intrin_with_mask(ctx, instr, true, ctx->tcs_temp_only_inputs, &indirect_write);
       if (temp_only_input && !indirect_write)
          return;
    }
@@ -4230,7 +4230,7 @@ bool tcs_output_is_read_by_tes(isel_context *ctx, nir_intrinsic_instr *instr, bo
                    : ctx->program->info->tcs.tes_patch_inputs_read;
 
    bool indirect_write = false;
-   bool output_read_by_tes = tcs_driver_location_matches_api_mask(ctx, instr, per_vertex, mask, &indirect_write);
+   bool output_read_by_tes = tcs_compare_intrin_with_mask(ctx, instr, per_vertex, mask, &indirect_write);
    return indirect_write || output_read_by_tes;
 }
 
@@ -4241,7 +4241,7 @@ bool tcs_output_is_read_by_tcs(isel_context *ctx, nir_intrinsic_instr *instr, bo
                    : ctx->shader->info.patch_outputs_read;
 
    bool indirect_write = false;
-   bool output_read = tcs_driver_location_matches_api_mask(ctx, instr, per_vertex, mask, &indirect_write);
+   bool output_read = tcs_compare_intrin_with_mask(ctx, instr, per_vertex, mask, &indirect_write);
    return indirect_write || output_read;
 }
 

diff --git a/src/amd/compiler/aco_instruction_selection.h b/src/amd/compiler/aco_instruction_selection.h
index a9249a8..8b0daf1 100644 (file)
--- a/src/amd/compiler/aco_instruction_selection.h
+++ b/src/amd/compiler/aco_instruction_selection.h
@@ -113,8 +113,6 @@ struct isel_context {
    /* I/O information */
    shader_io_state inputs;
    shader_io_state outputs;
-   uint8_t output_drv_loc_to_var_slot[MESA_SHADER_COMPUTE][VARYING_SLOT_MAX];
-   uint8_t output_tcs_patch_drv_loc_to_var_slot[VARYING_SLOT_MAX];
 };
 
 inline Temp get_arg(isel_context *ctx, struct ac_arg arg)

diff --git a/src/amd/compiler/aco_instruction_selection_setup.cpp b/src/amd/compiler/aco_instruction_selection_setup.cpp
index 6dfe28e..c0fa52c 100644 (file)
--- a/src/amd/compiler/aco_instruction_selection_setup.cpp
+++ b/src/amd/compiler/aco_instruction_selection_setup.cpp
@@ -478,9 +478,6 @@ setup_vs_variables(isel_context *ctx, nir_shader *nir)
    {
       if (ctx->stage == vertex_vs || ctx->stage == ngg_vertex_gs)
          variable->data.driver_location = variable->data.location * 4;
-
-      assert(variable->data.location >= 0 && variable->data.location <= UINT8_MAX);
-      ctx->output_drv_loc_to_var_slot[MESA_SHADER_VERTEX][variable->data.driver_location / 4] = variable->data.location;
    }
 
    if (ctx->stage == vertex_vs || ctx->stage == ngg_vertex_gs) {
@@ -566,19 +563,6 @@ setup_tcs_info(isel_context *ctx, nir_shader *nir, nir_shader *vs)
 }
 
 void
-setup_tcs_variables(isel_context *ctx, nir_shader *nir)
-{
-   nir_foreach_shader_out_variable(variable, nir) {
-      assert(variable->data.location >= 0 && variable->data.location <= UINT8_MAX);
-
-      if (variable->data.patch)
-         ctx->output_tcs_patch_drv_loc_to_var_slot[variable->data.driver_location / 4] = variable->data.location;
-      else
-         ctx->output_drv_loc_to_var_slot[MESA_SHADER_TESS_CTRL][variable->data.driver_location / 4] = variable->data.location;
-   }
-}
-
-void
 setup_tes_variables(isel_context *ctx, nir_shader *nir)
 {
    ctx->tcs_num_patches = ctx->args->options->key.tes.num_patches;
@@ -622,7 +606,6 @@ setup_variables(isel_context *ctx, nir_shader *nir)
       break;
    }
    case MESA_SHADER_TESS_CTRL: {
-      setup_tcs_variables(ctx, nir);
       break;
    }
    case MESA_SHADER_TESS_EVAL: {