nir: fix lowering arrays to elements for XFB outputs

If we have a transform feedback output like:

float[2] x2_out (VARYING_SLOT_VAR1.x, 0, 0)

which is lowered by nir_lower_io_arrays_to_elements to,

float x2_out (VARYING_SLOT_VAR1.x, 0, 0)
float x2_out@5 (VARYING_SLOT_VAR2.x, 0, 0)

We have to update the destination offset to avoid overwriting
the same value.

v2 (Jason Ekstrand):
 - Compute the correct offsets for arrays of vectors and/or doubles

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
This commit is contained in:
Samuel Pitoiset 2018-09-10 21:59:31 +02:00 committed by Jason Ekstrand
parent 9f4e0aa7c1
commit b2bbd978d0
1 changed files with 11 additions and 2 deletions

View File

@ -34,7 +34,8 @@
static unsigned
get_io_offset(nir_builder *b, nir_deref_instr *deref, nir_variable *var,
unsigned *element_index, nir_ssa_def **vertex_index)
unsigned *element_index, unsigned *xfb_offset,
nir_ssa_def **vertex_index)
{
nir_deref_path path;
nir_deref_path_init(&path, deref, NULL);
@ -51,6 +52,7 @@ get_io_offset(nir_builder *b, nir_deref_instr *deref, nir_variable *var,
}
unsigned offset = 0;
*xfb_offset = 0;
for (; *p; p++) {
if ((*p)->deref_type == nir_deref_type_array) {
/* must not be indirect dereference */
@ -59,6 +61,8 @@ get_io_offset(nir_builder *b, nir_deref_instr *deref, nir_variable *var,
unsigned size = glsl_count_attribute_slots((*p)->type, false);
offset += size * index;
xfb_offset += index * glsl_get_component_slots((*p)->type) * 4;
unsigned num_elements = glsl_type_is_array((*p)->type) ?
glsl_get_aoa_size((*p)->type) : 1;
@ -116,14 +120,19 @@ lower_array(nir_builder *b, nir_intrinsic_instr *intr, nir_variable *var,
nir_ssa_def *vertex_index = NULL;
unsigned elements_index = 0;
unsigned xfb_offset = 0;
unsigned io_offset = get_io_offset(b, nir_src_as_deref(intr->src[0]),
var, &elements_index, &vertex_index);
var, &elements_index, &xfb_offset,
&vertex_index);
nir_variable *element = elements[elements_index];
if (!element) {
element = nir_variable_clone(var, b->shader);
element->data.location = var->data.location + io_offset;
if (var->data.explicit_offset)
element->data.offset = var->data.offset + xfb_offset;
const struct glsl_type *type = glsl_without_array(element->type);
/* This pass also splits matrices so we need give them a new type. */