On Wednesday, November 16, 2016 8:37:50 PM PST Jason Ekstrand wrote: > On Mon, Nov 14, 2016 at 5:41 PM, Kenneth Graunke <[email protected]> > wrote: > > > Signed-off-by: Kenneth Graunke <[email protected]> > > --- > > src/compiler/Makefile.sources | 1 + > > src/compiler/nir/nir.h | 1 + > > .../nir/nir_lower_clip_cull_distance_arrays.c | 187 > > +++++++++++++++++++++ > > 3 files changed, 189 insertions(+) > > create mode 100644 src/compiler/nir/nir_lower_clip_cull_distance_arrays.c > > > > diff --git a/src/compiler/Makefile.sources b/src/compiler/Makefile.sources > > index 08d93e0..6c331b0 100644 > > --- a/src/compiler/Makefile.sources > > +++ b/src/compiler/Makefile.sources > > @@ -202,6 +202,7 @@ NIR_FILES = \ > > nir/nir_lower_bitmap.c \ > > nir/nir_lower_clamp_color_outputs.c \ > > nir/nir_lower_clip.c \ > > + nir/nir_lower_clip_cull_distance_arrays.c \ > > nir/nir_lower_double_ops.c \ > > nir/nir_lower_double_packing.c \ > > nir/nir_lower_drawpixels.c \ > > diff --git a/src/compiler/nir/nir.h b/src/compiler/nir/nir.h > > index 0b78242..f3aadb9 100644 > > --- a/src/compiler/nir/nir.h > > +++ b/src/compiler/nir/nir.h > > @@ -2441,6 +2441,7 @@ bool nir_lower_idiv(nir_shader *shader); > > > > void nir_lower_clip_vs(nir_shader *shader, unsigned ucp_enables); > > void nir_lower_clip_fs(nir_shader *shader, unsigned ucp_enables); > > +void nir_lower_clip_cull_distance_arrays(nir_shader *nir); > > > > void nir_lower_two_sided_color(nir_shader *shader); > > > > diff --git a/src/compiler/nir/nir_lower_clip_cull_distance_arrays.c > > b/src/compiler/nir/nir_lower_clip_cull_distance_arrays.c > > new file mode 100644 > > index 0000000..9f8a7de > > --- /dev/null > > +++ b/src/compiler/nir/nir_lower_clip_cull_distance_arrays.c > > @@ -0,0 +1,187 @@ > > +/* > > + * Copyright © 2015 Intel Corporation > > + * > > + * Permission is hereby granted, free of charge, to any person obtaining a > > + * copy of this software and associated documentation files (the > > "Software"), > > + * to deal in the Software without restriction, including without > > limitation > > + * the rights to use, copy, modify, merge, publish, distribute, > > sublicense, > > + * and/or sell copies of the Software, and to permit persons to whom the > > + * Software is furnished to do so, subject to the following conditions: > > + * > > + * The above copyright notice and this permission notice (including the > > next > > + * paragraph) shall be included in all copies or substantial portions of > > the > > + * Software. > > + * > > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, > > EXPRESS OR > > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF > > MERCHANTABILITY, > > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT > > SHALL > > + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR > > OTHER > > + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING > > + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER > > DEALINGS > > + * IN THE SOFTWARE. > > + */ > > + > > +#include "nir.h" > > +#include "nir_builder.h" > > + > > +/** > > + * @file > > + * > > + * This pass combines separate clip and cull distance arrays into a > > + * single array that contains both. Clip distances come first, then > > + * cull distances. It also populates nir_shader_info with the size > > + * of the original arrays so the driver knows which are which. > > + */ > > + > > +/** > > + * Get the length of the clip/cull distance array, looking past > > + * any interface block arrays. > > + */ > > +static unsigned > > +get_unwrapped_array_length(nir_variable *var) > > +{ > > + if (!var) > > + return 0; > > + > > + /* Unwrap GS input and TCS input/output interfaces. We want the > > + * underlying clip/cull distance array length, not the per-vertex > > + * array length. > > + */ > > > > Probably not needed, but we could put some stage-based asserts in here.
Yeah...we should probably put competent helpers for "is this
per-vertex IO?" and "get me the actual underlying type please"
in nir.h where everybody can use them.
> > + const struct glsl_type *type = var->type;
> > + if (glsl_type_is_array_of_arrays(type))
> > + type = glsl_get_array_element(type);
> > +
> > + return glsl_get_length(type);
> > +}
> > +
> > +/**
> > + * Get the type of the combined array (including interface block nesting).
> > + */
> > +static const struct glsl_type *
> > +get_new_array_type(const struct glsl_type *old_type, unsigned length)
> > +{
> > + const struct glsl_type *type = glsl_array_type(glsl_float_type(),
> > length);
> > +
> > + if (glsl_type_is_array_of_arrays(old_type))
> > + type = glsl_array_type(type, glsl_get_length(type));
> > +
> > + return type;
> > +}
> > +
> > +/**
> > + * Rewrite any clip/cull distances to refer to the new combined array.
> > + */
> > +static void
> > +rewrite_references(nir_instr *instr,
> > + nir_variable *combined,
> > + unsigned cull_offset)
> > +{
> > + if (instr->type != nir_instr_type_intrinsic)
> > + return;
> > +
> > + nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
> > +
> > + /* copy_var needs to be lowered to load/store before calling this pass
> > */
> > + assert(intrin->intrinsic != nir_intrinsic_copy_var);
> >
>
> Wouldn't be that hard to make it support it, but sure.
>
>
> > +
> > + if (intrin->intrinsic != nir_intrinsic_load_var &&
> > + intrin->intrinsic != nir_intrinsic_store_var)
> > + return;
> > +
> > + nir_deref_var *var_ref = intrin->variables[0];
> > + if (var_ref->var->data.mode != combined->data.mode)
> > + return;
> > +
> > + if (var_ref->var->data.location != VARYING_SLOT_CLIP_DIST0 &&
> > + var_ref->var->data.location != VARYING_SLOT_CULL_DIST0)
> > + return;
> > +
> > + /* Update types along the deref chain */
> > + const struct glsl_type *type = combined->type;
> > + nir_deref *deref = &var_ref->deref;
> > + while (deref) {
> > + deref->type = type;
> > + deref = deref->child;
> > + type = glsl_get_array_element(type);
> > + }
> > +
> > + /* For cull distances, add an offset to the array index */
> > + if (var_ref->var->data.location == VARYING_SLOT_CULL_DIST0) {
> > + nir_deref *tail = nir_deref_tail(&intrin->variables[0]->deref);
> > + nir_deref_array *array_ref = nir_deref_as_array(tail);
> > +
> > + array_ref->base_offset += cull_offset;
> > + }
> > +
> > + /* Point the deref at the combined array */
> > + var_ref->var = combined;
> > +
> > + /* There's no need to update writemasks; it's a scalar array. */
> > +}
> > +
> > +static void
> > +combine_clip_cull(nir_shader *nir,
> > + struct exec_list *vars,
> > + bool store_info)
> > +{
> > + nir_variable *cull = NULL;
> > + nir_variable *clip = NULL;
> > +
> > + nir_foreach_variable(var, vars) {
> > + if (var->data.location == VARYING_SLOT_CLIP_DIST0)
> > + clip = var;
> > +
> > + if (var->data.location == VARYING_SLOT_CULL_DIST0)
> > + cull = var;
> > + }
> > +
> > + const unsigned clip_array_size = get_unwrapped_array_length(clip);
> > + const unsigned cull_array_size = get_unwrapped_array_length(cull);
> > +
> > + if (store_info) {
> > + nir->info->clip_distance_array_size = clip_array_size;
> > + nir->info->cull_distance_array_size = cull_array_size;
> >
>
> Why is this optional? If the sizes here and in the info ever differ,
> you're going to have a problem.
We call this function twice - once for inputs, and once for outputs.
The info values should come from the shader outputs. Though, looking
below......
> > + }
> > +
> > + if (clip)
> > + clip->data.compact = true;
> > +
> > + if (cull)
> > + cull->data.compact = true;
> > +
> > + if (cull_array_size > 0) {
> > + if (clip_array_size == 0) {
> > + /* No clip distances, just change the cull distance location */
> > + cull->data.location = VARYING_SLOT_CLIP_DIST0;
> > + } else {
> > + /* Turn the ClipDistance array into a combined one */
> > + clip->type = get_new_array_type(clip->type,
> > + clip_array_size +
> > cull_array_size);
> > +
> > + /* Rewrite CullDistance to reference the combined array */
> > + nir_foreach_function(function, nir) {
> > + if (function->impl) {
> > + nir_foreach_block(block, function->impl) {
> > + nir_foreach_instr(instr, block) {
> > + rewrite_references(instr, clip, clip_array_size);
> > + }
> > + }
> > + }
> > + }
> > +
> > + /* Delete the old CullDistance variable */
> > + exec_node_remove(&cull->node);
> > + ralloc_free(cull);
> > + }
> > + }
> > +}
> > +
> > +void
> > +nir_lower_clip_cull_distance_arrays(nir_shader *nir)
> > +{
> > + if (nir->stage <= MESA_SHADER_GEOMETRY)
> > + combine_clip_cull(nir, &nir->outputs, &nir->info);
> > +
> > + if (nir->stage > MESA_SHADER_VERTEX)
> > + combine_clip_cull(nir, &nir->inputs, NULL);
> > +}
.....apparently I meant to pass "true" and "false" here.
(I'd originally passed the shader_info pointer, rather than a flag,
but once I started passing nir_shader as well, I changed it. Oops.)
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ mesa-dev mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/mesa-dev
