/**************************************************************************
*
- * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2007 VMware, Inc.
* All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* 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 NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS 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 "draw/draw_context.h"
#include "draw/draw_vbuf.h"
#include "draw/draw_vertex.h"
+#include "draw/draw_prim_assembler.h"
#include "draw/draw_pt.h"
#include "draw/draw_vs.h"
#include "draw/draw_gs.h"
unsigned vertex_data_offset;
unsigned vertex_size;
unsigned input_prim;
- unsigned output_prim;
unsigned opt;
};
-static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle,
- unsigned in_prim,
- unsigned out_prim,
- unsigned opt,
- unsigned *max_vertices )
+
+/** cast wrapper */
+static inline struct fetch_pipeline_middle_end *
+fetch_pipeline_middle_end(struct draw_pt_middle_end *middle)
{
- struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
+ return (struct fetch_pipeline_middle_end *) middle;
+}
+
+
+/**
+ * Prepare/validate middle part of the vertex pipeline.
+ * NOTE: if you change this function, also look at the LLVM
+ * function llvm_middle_end_prepare() for similar changes.
+ */
+static void
+fetch_pipeline_prepare(struct draw_pt_middle_end *middle,
+ unsigned prim,
+ unsigned opt,
+ unsigned *max_vertices)
+{
+ struct fetch_pipeline_middle_end *fpme = fetch_pipeline_middle_end(middle);
struct draw_context *draw = fpme->draw;
struct draw_vertex_shader *vs = draw->vs.vertex_shader;
+ struct draw_geometry_shader *gs = draw->gs.geometry_shader;
unsigned i;
unsigned instance_id_index = ~0;
-
- /* Add one to num_outputs because the pipeline occasionally tags on
- * an additional texcoord, eg for AA lines.
- */
- unsigned nr = MAX2( vs->info.num_inputs,
- vs->info.num_outputs + 1 );
+ const unsigned gs_out_prim = (gs ? gs->output_primitive :
+ u_assembled_prim(prim));
+ unsigned nr_vs_outputs = draw_total_vs_outputs(draw);
+ unsigned nr = MAX2(vs->info.num_inputs, nr_vs_outputs);
+ unsigned point_clip = draw->rasterizer->fill_front == PIPE_POLYGON_MODE_POINT ||
+ gs_out_prim == PIPE_PRIM_POINTS;
+
+ if (gs) {
+ nr = MAX2(nr, gs->info.num_outputs + 1);
+ }
/* Scan for instanceID system value.
*/
}
}
- fpme->input_prim = in_prim;
- fpme->output_prim = out_prim;
+ fpme->input_prim = prim;
fpme->opt = opt;
/* Always leave room for the vertex header whether we need it or
* not. It's hard to get rid of it in particular because of the
- * viewport code in draw_pt_post_vs.c.
+ * viewport code in draw_pt_post_vs.c.
*/
fpme->vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float);
-
-
- draw_pt_fetch_prepare( fpme->fetch,
+ draw_pt_fetch_prepare( fpme->fetch,
vs->info.num_inputs,
fpme->vertex_size,
instance_id_index );
- /* XXX: it's not really gl rasterization rules we care about here,
- * but gl vs dx9 clip spaces.
- */
draw_pt_post_vs_prepare( fpme->post_vs,
- (boolean)draw->bypass_clipping,
- (boolean)draw->identity_viewport,
- (boolean)draw->rasterizer->gl_rasterization_rules,
- (draw->vs.edgeflag_output ? true : false) );
+ draw->clip_xy,
+ draw->clip_z,
+ draw->clip_user,
+ point_clip ? draw->guard_band_points_xy :
+ draw->guard_band_xy,
+ draw->bypass_viewport,
+ draw->rasterizer->clip_halfz,
+ (draw->vs.edgeflag_output ? TRUE : FALSE) );
- draw_pt_so_emit_prepare( fpme->so_emit );
+ draw_pt_so_emit_prepare( fpme->so_emit, FALSE );
if (!(opt & PT_PIPELINE)) {
draw_pt_emit_prepare( fpme->emit,
- out_prim,
+ gs_out_prim,
max_vertices );
- *max_vertices = MAX2( *max_vertices,
- DRAW_PIPE_MAX_VERTICES );
+ *max_vertices = MAX2( *max_vertices, 4096 );
}
else {
- *max_vertices = DRAW_PIPE_MAX_VERTICES;
+ /* limit max fetches by limiting max_vertices */
+ *max_vertices = 4096;
}
- /* return even number */
- *max_vertices = *max_vertices & ~1;
-
/* No need to prepare the shader.
*/
vs->prepare(vs, draw);
+
+ /* Make sure that the vertex size didn't change at any point above */
+ assert(nr_vs_outputs == draw_total_vs_outputs(draw));
+}
+
+
+static void
+fetch_pipeline_bind_parameters(struct draw_pt_middle_end *middle)
+{
+ /* No-op since the vertex shader executor and drawing pipeline
+ * just grab the constants, viewport, etc. from the draw context state.
+ */
}
prim_info );
}
-static void emit(struct pt_emit *emit,
- const struct draw_vertex_info *vert_info,
- const struct draw_prim_info *prim_info)
+
+static void
+emit(struct pt_emit *emit,
+ const struct draw_vertex_info *vert_info,
+ const struct draw_prim_info *prim_info)
{
if (prim_info->linear) {
draw_pt_emit_linear(emit, vert_info, prim_info);
}
-static void draw_vertex_shader_run(struct draw_vertex_shader *vshader,
- const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
- const struct draw_vertex_info *input_verts,
- struct draw_vertex_info *output_verts )
+static void
+draw_vertex_shader_run(struct draw_vertex_shader *vshader,
+ const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
+ unsigned const_size[PIPE_MAX_CONSTANT_BUFFERS],
+ const struct draw_fetch_info *fetch_info,
+ const struct draw_vertex_info *input_verts,
+ struct draw_vertex_info *output_verts)
{
output_verts->vertex_size = input_verts->vertex_size;
output_verts->stride = input_verts->vertex_size;
output_verts->count = input_verts->count;
output_verts->verts =
(struct vertex_header *)MALLOC(output_verts->vertex_size *
- output_verts->count);
+ align(output_verts->count, 4));
vshader->run_linear(vshader,
(const float (*)[4])input_verts->verts->data,
( float (*)[4])output_verts->verts->data,
constants,
+ const_size,
input_verts->count,
input_verts->vertex_size,
- input_verts->vertex_size);
+ input_verts->vertex_size,
+ fetch_info->elts);
}
-static void fetch_pipeline_generic( struct draw_pt_middle_end *middle,
- const struct draw_fetch_info *fetch_info,
- const struct draw_prim_info *prim_info )
+
+static void
+fetch_pipeline_generic(struct draw_pt_middle_end *middle,
+ const struct draw_fetch_info *fetch_info,
+ const struct draw_prim_info *in_prim_info)
{
- struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
+ struct fetch_pipeline_middle_end *fpme = fetch_pipeline_middle_end(middle);
struct draw_context *draw = fpme->draw;
struct draw_vertex_shader *vshader = draw->vs.vertex_shader;
struct draw_geometry_shader *gshader = draw->gs.geometry_shader;
- struct draw_prim_info gs_prim_info;
+ struct draw_prim_info gs_prim_info[TGSI_MAX_VERTEX_STREAMS];
struct draw_vertex_info fetched_vert_info;
struct draw_vertex_info vs_vert_info;
- struct draw_vertex_info gs_vert_info;
+ struct draw_vertex_info gs_vert_info[TGSI_MAX_VERTEX_STREAMS];
struct draw_vertex_info *vert_info;
+ struct draw_prim_info ia_prim_info;
+ struct draw_vertex_info ia_vert_info;
+ const struct draw_prim_info *prim_info = in_prim_info;
+ boolean free_prim_info = FALSE;
unsigned opt = fpme->opt;
+ int num_vertex_streams = 1;
fetched_vert_info.count = fetch_info->count;
fetched_vert_info.vertex_size = fpme->vertex_size;
assert(0);
return;
}
+ if (draw->collect_statistics) {
+ draw->statistics.ia_vertices += prim_info->count;
+ draw->statistics.ia_primitives +=
+ u_decomposed_prims_for_vertices(prim_info->prim, fetch_info->count);
+ draw->statistics.vs_invocations += fetch_info->count;
+ }
/* Fetch into our vertex buffer.
*/
fetch( fpme->fetch, fetch_info, (char *)fetched_vert_info.verts );
- /* Finished with fetch:
- */
- fetch_info = NULL;
vert_info = &fetched_vert_info;
/* Run the shader, note that this overwrites the data[] parts of
* the pipeline verts.
+ * Need fetch info to get vertex id correct.
*/
if (fpme->opt & PT_SHADE) {
draw_vertex_shader_run(vshader,
draw->pt.user.vs_constants,
+ draw->pt.user.vs_constants_size,
+ fetch_info,
vert_info,
&vs_vert_info);
vert_info = &vs_vert_info;
}
+ /* Finished with fetch:
+ */
+ fetch_info = NULL;
+
if ((fpme->opt & PT_SHADE) && gshader) {
draw_geometry_shader_run(gshader,
draw->pt.user.gs_constants,
+ draw->pt.user.gs_constants_size,
vert_info,
prim_info,
- &gs_vert_info,
- &gs_prim_info);
+ &vshader->info,
+ gs_vert_info,
+ gs_prim_info);
FREE(vert_info->verts);
- vert_info = &gs_vert_info;
- prim_info = &gs_prim_info;
+ vert_info = &gs_vert_info[0];
+ prim_info = &gs_prim_info[0];
+ num_vertex_streams = gshader->num_vertex_streams;
+
+ /*
+ * pt emit can only handle ushort number of vertices (see
+ * render->allocate_vertices).
+ * vsplit guarantees there's never more than 4096, however GS can
+ * easily blow this up (by a factor of 256 (or even 1024) max).
+ */
+ if (vert_info->count > 65535) {
+ opt |= PT_PIPELINE;
+ }
+ } else {
+ if (draw_prim_assembler_is_required(draw, prim_info, vert_info)) {
+ draw_prim_assembler_run(draw, prim_info, vert_info,
+ &ia_prim_info, &ia_vert_info);
+
+ if (ia_vert_info.count) {
+ FREE(vert_info->verts);
+ vert_info = &ia_vert_info;
+ prim_info = &ia_prim_info;
+ free_prim_info = TRUE;
+ }
+ }
+ }
+ if (prim_info->count == 0) {
+ debug_printf("GS/IA didn't emit any vertices!\n");
+
+ FREE(vert_info->verts);
+ if (free_prim_info) {
+ FREE(prim_info->primitive_lengths);
+ }
+ return;
}
* XXX: Stream output surely needs to respect the prim_info->elt
* lists.
*/
- draw_pt_so_emit( fpme->so_emit,
- vert_info,
- prim_info );
-
- if (draw_pt_post_vs_run( fpme->post_vs,
- vert_info ))
- {
- opt |= PT_PIPELINE;
- }
+ draw_pt_so_emit( fpme->so_emit, num_vertex_streams, vert_info, prim_info );
+
+ draw_stats_clipper_primitives(draw, prim_info);
- /* Do we need to run the pipeline?
+ /*
+ * if there's no position, need to stop now, or the latter stages
+ * will try to access non-existent position output.
*/
- if (opt & PT_PIPELINE) {
- pipeline( fpme,
- vert_info,
- prim_info );
- }
- else {
- emit( fpme->emit,
- vert_info,
- prim_info );
+ if (draw_current_shader_position_output(draw) != -1) {
+
+ if (draw_pt_post_vs_run( fpme->post_vs, vert_info, prim_info ))
+ {
+ opt |= PT_PIPELINE;
+ }
+
+ /* Do we need to run the pipeline?
+ */
+ if (opt & PT_PIPELINE) {
+ pipeline( fpme, vert_info, prim_info );
+ }
+ else {
+ emit( fpme->emit, vert_info, prim_info );
+ }
}
FREE(vert_info->verts);
+ if (free_prim_info) {
+ FREE(prim_info->primitive_lengths);
+ }
}
-static void fetch_pipeline_run( struct draw_pt_middle_end *middle,
- const unsigned *fetch_elts,
- unsigned fetch_count,
- const ushort *draw_elts,
- unsigned draw_count )
+
+static inline unsigned
+prim_type(unsigned prim, unsigned flags)
{
- struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
+ if (flags & DRAW_LINE_LOOP_AS_STRIP)
+ return PIPE_PRIM_LINE_STRIP;
+ else
+ return prim;
+}
+
+
+static void
+fetch_pipeline_run(struct draw_pt_middle_end *middle,
+ const unsigned *fetch_elts,
+ unsigned fetch_count,
+ const ushort *draw_elts,
+ unsigned draw_count,
+ unsigned prim_flags)
+{
+ struct fetch_pipeline_middle_end *fpme = fetch_pipeline_middle_end(middle);
struct draw_fetch_info fetch_info;
struct draw_prim_info prim_info;
prim_info.start = 0;
prim_info.count = draw_count;
prim_info.elts = draw_elts;
- prim_info.prim = fpme->input_prim;
+ prim_info.prim = prim_type(fpme->input_prim, prim_flags);
+ prim_info.flags = prim_flags;
prim_info.primitive_count = 1;
prim_info.primitive_lengths = &draw_count;
}
-static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle,
- unsigned start,
- unsigned count)
+static void
+fetch_pipeline_linear_run(struct draw_pt_middle_end *middle,
+ unsigned start,
+ unsigned count,
+ unsigned prim_flags)
{
- struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
+ struct fetch_pipeline_middle_end *fpme = fetch_pipeline_middle_end(middle);
struct draw_fetch_info fetch_info;
struct draw_prim_info prim_info;
prim_info.start = 0;
prim_info.count = count;
prim_info.elts = NULL;
- prim_info.prim = fpme->input_prim;
+ prim_info.prim = prim_type(fpme->input_prim, prim_flags);
+ prim_info.flags = prim_flags;
prim_info.primitive_count = 1;
prim_info.primitive_lengths = &count;
-static boolean fetch_pipeline_linear_run_elts( struct draw_pt_middle_end *middle,
- unsigned start,
- unsigned count,
- const ushort *draw_elts,
- unsigned draw_count )
+static boolean
+fetch_pipeline_linear_run_elts(struct draw_pt_middle_end *middle,
+ unsigned start,
+ unsigned count,
+ const ushort *draw_elts,
+ unsigned draw_count,
+ unsigned prim_flags )
{
- struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
+ struct fetch_pipeline_middle_end *fpme = fetch_pipeline_middle_end(middle);
struct draw_fetch_info fetch_info;
struct draw_prim_info prim_info;
prim_info.start = 0;
prim_info.count = draw_count;
prim_info.elts = draw_elts;
- prim_info.prim = fpme->input_prim;
+ prim_info.prim = prim_type(fpme->input_prim, prim_flags);
+ prim_info.flags = prim_flags;
prim_info.primitive_count = 1;
prim_info.primitive_lengths = &draw_count;
}
-
-static void fetch_pipeline_finish( struct draw_pt_middle_end *middle )
+static void
+fetch_pipeline_finish( struct draw_pt_middle_end *middle )
{
/* nothing to do */
}
-static void fetch_pipeline_destroy( struct draw_pt_middle_end *middle )
+
+static void
+fetch_pipeline_destroy( struct draw_pt_middle_end *middle )
{
- struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
+ struct fetch_pipeline_middle_end *fpme = fetch_pipeline_middle_end(middle);
if (fpme->fetch)
draw_pt_fetch_destroy( fpme->fetch );
}
-struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit( struct draw_context *draw )
+struct draw_pt_middle_end *
+draw_pt_fetch_pipeline_or_emit(struct draw_context *draw)
{
- struct fetch_pipeline_middle_end *fpme = CALLOC_STRUCT( fetch_pipeline_middle_end );
+ struct fetch_pipeline_middle_end *fpme =
+ CALLOC_STRUCT( fetch_pipeline_middle_end );
if (!fpme)
goto fail;
fpme->base.prepare = fetch_pipeline_prepare;
+ fpme->base.bind_parameters = fetch_pipeline_bind_parameters;
fpme->base.run = fetch_pipeline_run;
fpme->base.run_linear = fetch_pipeline_linear_run;
fpme->base.run_linear_elts = fetch_pipeline_linear_run_elts;