From 9f3917bf372aa19f85875dbe30ca12adc9b67b90 Mon Sep 17 00:00:00 2001 From: Kenneth Graunke Date: Wed, 9 Dec 2015 18:26:19 -0800 Subject: i965: Fix partial variable access for geometry shaders in SSO mode. Without varying packing, if a VS writes a compound variable, and the GS only reads part of it, the base location of the variable may not actually be in the VUE map. To cope with this, we do lowering in terms of varying slots, add any constant offsets to the base, and then do the VUE map remapping. This ensures we only look up VUE map entries for slots which actually exist. Signed-off-by: Kenneth Graunke Reviewed-by: Jason Ekstrand --- src/mesa/drivers/dri/i965/brw_nir.c | 38 +++++++++++++++++++++++++++++++++---- 1 file changed, 34 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i965/brw_nir.c b/src/mesa/drivers/dri/i965/brw_nir.c index c9a2767b393..88457e4f9cb 100644 --- a/src/mesa/drivers/dri/i965/brw_nir.c +++ b/src/mesa/drivers/dri/i965/brw_nir.c @@ -114,6 +114,27 @@ remap_vs_attrs(nir_block *block, void *closure) return true; } +static bool +remap_inputs_with_vue_map(nir_block *block, void *closure) +{ + const struct brw_vue_map *vue_map = closure; + + nir_foreach_instr(block, instr) { + if (instr->type != nir_instr_type_intrinsic) + continue; + + nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr); + + if (intrin->intrinsic == nir_intrinsic_load_input || + intrin->intrinsic == nir_intrinsic_load_per_vertex_input) { + int vue_slot = vue_map->varying_to_slot[intrin->const_index[0]]; + assert(vue_slot != -1); + intrin->const_index[0] = vue_slot; + } + } + return true; +} + static void brw_nir_lower_inputs(nir_shader *nir, const struct brw_device_info *devinfo, @@ -182,15 +203,24 @@ brw_nir_lower_inputs(nir_shader *nir, brw_compute_vue_map(devinfo, &input_vue_map, inputs_read, nir->info.separate_shader); - /* Start with the slot for the variable's base. */ foreach_list_typed(nir_variable, var, node, &nir->inputs) { - assert(input_vue_map.varying_to_slot[var->data.location] != -1); - var->data.driver_location = - input_vue_map.varying_to_slot[var->data.location]; + var->data.driver_location = var->data.location; } /* Inputs are stored in vec4 slots, so use type_size_vec4(). */ nir_lower_io(nir, nir_var_shader_in, type_size_vec4); + + /* This pass needs actual constants */ + nir_opt_constant_folding(nir); + + nir_foreach_overload(nir, overload) { + if (overload->impl) { + nir_builder_init(¶ms.b, overload->impl); + nir_foreach_block(overload->impl, add_const_offset_to_base, ¶ms); + nir_foreach_block(overload->impl, remap_inputs_with_vue_map, + &input_vue_map); + } + } } break; } -- cgit v1.2.3