/**
* Compute the VUE map for vertex shader program.
+ *
+ * Note that consumers of this map using cache keys must include
+ * prog_data->userclip and prog_data->outputs_written in their key
+ * (generated by CACHE_NEW_VS_PROG).
*/
-void
-brw_compute_vue_map(struct brw_vue_map *vue_map,
- const struct intel_context *intel, int nr_userclip,
- GLbitfield64 outputs_written)
+static void
+brw_compute_vue_map(struct brw_vs_compile *c)
{
+ struct brw_context *brw = c->func.brw;
+ const struct intel_context *intel = &brw->intel;
+ struct brw_vue_map *vue_map = &c->prog_data.vue_map;
+ GLbitfield64 outputs_written = c->prog_data.outputs_written;
int i;
vue_map->num_slots = 0;
assign_vue_slot(vue_map, VERT_RESULT_PSIZ);
assign_vue_slot(vue_map, BRW_VERT_RESULT_NDC);
assign_vue_slot(vue_map, BRW_VERT_RESULT_HPOS_DUPLICATE);
- assign_vue_slot(vue_map, BRW_VERT_RESULT_CLIP0);
- assign_vue_slot(vue_map, BRW_VERT_RESULT_CLIP1);
+ assign_vue_slot(vue_map, VERT_RESULT_CLIP_DIST0);
+ assign_vue_slot(vue_map, VERT_RESULT_CLIP_DIST1);
assign_vue_slot(vue_map, BRW_VERT_RESULT_PAD);
assign_vue_slot(vue_map, VERT_RESULT_HPOS);
break;
*/
assign_vue_slot(vue_map, VERT_RESULT_PSIZ);
assign_vue_slot(vue_map, VERT_RESULT_HPOS);
- if (nr_userclip) {
- assign_vue_slot(vue_map, BRW_VERT_RESULT_CLIP0);
- assign_vue_slot(vue_map, BRW_VERT_RESULT_CLIP1);
+ if (c->key.userclip_active) {
+ assign_vue_slot(vue_map, VERT_RESULT_CLIP_DIST0);
+ assign_vue_slot(vue_map, VERT_RESULT_CLIP_DIST1);
}
/* front and back colors need to be consecutive so that we can use
* ATTRIBUTE_SWIZZLE_INPUTATTR_FACING to swizzle them when doing
/* The hardware doesn't care about the rest of the vertex outputs, so just
* assign them contiguously. Don't reassign outputs that already have a
* slot.
+ *
+ * Also, prior to Gen6, don't assign a slot for VERT_RESULT_CLIP_VERTEX,
+ * since it is unsupported. In Gen6 and above, VERT_RESULT_CLIP_VERTEX may
+ * be needed for transform feedback; since we don't want to have to
+ * recompute the VUE map (and everything that depends on it) when transform
+ * feedback is enabled or disabled, just go ahead and assign a slot for it.
*/
for (int i = 0; i < VERT_RESULT_MAX; ++i) {
+ if (intel->gen < 6 && i == VERT_RESULT_CLIP_VERTEX)
+ continue;
if ((outputs_written & BITFIELD64_BIT(i)) &&
vue_map->vert_result_to_slot[i] == -1) {
assign_vue_slot(vue_map, i);
}
}
+
+/**
+ * Decide which set of clip planes should be used when clipping via
+ * gl_Position or gl_ClipVertex.
+ */
+gl_clip_plane *brw_select_clip_planes(struct gl_context *ctx)
+{
+ if (ctx->Shader.CurrentVertexProgram) {
+ /* There is currently a GLSL vertex shader, so clip according to GLSL
+ * rules, which means compare gl_ClipVertex (or gl_Position, if
+ * gl_ClipVertex wasn't assigned) against the eye-coordinate clip planes
+ * that were stored in EyeUserPlane at the time the clip planes were
+ * specified.
+ */
+ return ctx->Transform.EyeUserPlane;
+ } else {
+ /* Either we are using fixed function or an ARB vertex program. In
+ * either case the clip planes are going to be compared against
+ * gl_Position (which is in clip coordinates) so we have to clip using
+ * _ClipUserPlane, which was transformed into clip coordinates by Mesa
+ * core.
+ */
+ return ctx->Transform._ClipUserPlane;
+ }
+}
+
+
static bool
do_vs_prog(struct brw_context *brw,
struct gl_shader_program *prog,
if (c.key.copy_edgeflag) {
c.prog_data.outputs_written |= BITFIELD64_BIT(VERT_RESULT_EDGE);
- c.prog_data.inputs_read |= 1<<VERT_ATTRIB_EDGEFLAG;
+ c.prog_data.inputs_read |= VERT_BIT_EDGEFLAG;
}
/* Put dummy slots into the VUE for the SF to put the replaced
c.prog_data.outputs_written |= BITFIELD64_BIT(VERT_RESULT_TEX0 + i);
}
+ brw_compute_vue_map(&c);
+
if (0) {
_mesa_fprint_program_opt(stdout, &c.vp->program.Base, PROG_PRINT_DEBUG,
- GL_TRUE);
+ true);
}
/* Emit GEN4 code.
*/
- if (brw->new_vs_backend && prog) {
+ if (prog) {
if (!brw_vs_emit(prog, &c)) {
ralloc_free(mem_ctx);
return false;
brw_old_vs_emit(&c);
}
+ if (c.prog_data.nr_pull_params)
+ c.prog_data.num_surfaces = 1;
+ if (c.vp->program.Base.SamplersUsed)
+ c.prog_data.num_surfaces = SURF_INDEX_VS_TEXTURE(BRW_MAX_TEX_UNIT);
+ if (prog &&
+ prog->_LinkedShaders[MESA_SHADER_VERTEX]->NumUniformBlocks) {
+ c.prog_data.num_surfaces =
+ SURF_INDEX_VS_UBO(prog->_LinkedShaders[MESA_SHADER_VERTEX]->NumUniformBlocks);
+ }
+
/* Scratch space is used for register spilling */
if (c.last_scratch) {
+ perf_debug("Vertex shader triggered register spilling. "
+ "Try reducing the number of live vec4 values to "
+ "improve performance.\n");
+
c.prog_data.total_scratch = brw_get_scratch_size(c.last_scratch);
brw_get_scratch_bo(intel, &brw->vs.scratch_bo,
- c.prog_data.total_scratch * brw->vs_max_threads);
+ c.prog_data.total_scratch * brw->max_vs_threads);
}
/* get the program
return true;
}
+static bool
+key_debug(const char *name, int a, int b)
+{
+ if (a != b) {
+ perf_debug(" %s %d->%d\n", name, a, b);
+ return true;
+ }
+ return false;
+}
+
+void
+brw_vs_debug_recompile(struct brw_context *brw,
+ struct gl_shader_program *prog,
+ const struct brw_vs_prog_key *key)
+{
+ struct brw_cache_item *c = NULL;
+ const struct brw_vs_prog_key *old_key = NULL;
+ bool found = false;
+
+ perf_debug("Recompiling vertex shader for program %d\n", prog->Name);
+
+ for (unsigned int i = 0; i < brw->cache.size; i++) {
+ for (c = brw->cache.items[i]; c; c = c->next) {
+ if (c->cache_id == BRW_VS_PROG) {
+ old_key = c->key;
+
+ if (old_key->program_string_id == key->program_string_id)
+ break;
+ }
+ }
+ if (c)
+ break;
+ }
+
+ if (!c) {
+ perf_debug(" Didn't find previous compile in the shader cache for "
+ "debug\n");
+ return;
+ }
+
+ for (unsigned int i = 0; i < VERT_ATTRIB_MAX; i++) {
+ found |= key_debug("GL_FIXED rescaling",
+ old_key->gl_fixed_input_size[i],
+ key->gl_fixed_input_size[i]);
+ }
+
+ found |= key_debug("user clip flags",
+ old_key->userclip_active, key->userclip_active);
+
+ found |= key_debug("user clipping planes as push constants",
+ old_key->nr_userclip_plane_consts,
+ key->nr_userclip_plane_consts);
+
+ found |= key_debug("clip distance enable",
+ old_key->uses_clip_distance, key->uses_clip_distance);
+ found |= key_debug("clip plane enable bitfield",
+ old_key->userclip_planes_enabled_gen_4_5,
+ key->userclip_planes_enabled_gen_4_5);
+ found |= key_debug("copy edgeflag",
+ old_key->copy_edgeflag, key->copy_edgeflag);
+ found |= key_debug("PointCoord replace",
+ old_key->point_coord_replace, key->point_coord_replace);
+ found |= key_debug("vertex color clamping",
+ old_key->clamp_vertex_color, key->clamp_vertex_color);
+
+ found |= brw_debug_recompile_sampler_key(&old_key->tex, &key->tex);
+
+ if (!found) {
+ perf_debug(" Something else\n");
+ }
+}
static void brw_upload_vs_prog(struct brw_context *brw)
{
- struct gl_context *ctx = &brw->intel.ctx;
+ struct intel_context *intel = &brw->intel;
+ struct gl_context *ctx = &intel->ctx;
struct brw_vs_prog_key key;
+ /* BRW_NEW_VERTEX_PROGRAM */
struct brw_vertex_program *vp =
(struct brw_vertex_program *)brw->vertex_program;
+ struct gl_program *prog = (struct gl_program *) brw->vertex_program;
int i;
memset(&key, 0, sizeof(key));
* the inputs it asks for, whether they are varying or not.
*/
key.program_string_id = vp->id;
- key.nr_userclip = brw_count_bits(ctx->Transform.ClipPlanesEnabled);
- key.copy_edgeflag = (ctx->Polygon.FrontMode != GL_FILL ||
- ctx->Polygon.BackMode != GL_FILL);
+ key.userclip_active = (ctx->Transform.ClipPlanesEnabled != 0);
+ key.uses_clip_distance = vp->program.UsesClipDistance;
+ if (key.userclip_active && !key.uses_clip_distance) {
+ if (intel->gen < 6) {
+ key.nr_userclip_plane_consts
+ = _mesa_bitcount_64(ctx->Transform.ClipPlanesEnabled);
+ key.userclip_planes_enabled_gen_4_5
+ = ctx->Transform.ClipPlanesEnabled;
+ } else {
+ key.nr_userclip_plane_consts
+ = _mesa_logbase2(ctx->Transform.ClipPlanesEnabled) + 1;
+ }
+ }
+
+ /* _NEW_POLYGON */
+ if (intel->gen < 6) {
+ key.copy_edgeflag = (ctx->Polygon.FrontMode != GL_FILL ||
+ ctx->Polygon.BackMode != GL_FILL);
+ }
/* _NEW_LIGHT | _NEW_BUFFERS */
key.clamp_vertex_color = ctx->Light._ClampVertexColor;
}
}
+ /* _NEW_TEXTURE */
+ brw_populate_sampler_prog_key_data(ctx, prog, &key.tex);
+
/* BRW_NEW_VERTICES */
for (i = 0; i < VERT_ATTRIB_MAX; i++) {
- if (vp->program.Base.InputsRead & (1 << i) &&
+ if (vp->program.Base.InputsRead & BITFIELD64_BIT(i) &&
brw->vb.inputs[i].glarray->Type == GL_FIXED) {
key.gl_fixed_input_size[i] = brw->vb.inputs[i].glarray->Size;
}
const struct brw_tracked_state brw_vs_prog = {
.dirty = {
.mesa = (_NEW_TRANSFORM | _NEW_POLYGON | _NEW_POINT | _NEW_LIGHT |
+ _NEW_TEXTURE |
_NEW_BUFFERS),
.brw = (BRW_NEW_VERTEX_PROGRAM |
BRW_NEW_VERTICES),
.cache = 0
},
- .prepare = brw_upload_vs_prog
+ .emit = brw_upload_vs_prog
};
bool
{
struct brw_context *brw = brw_context(ctx);
struct brw_vs_prog_key key;
- struct gl_vertex_program *vp = prog->VertexProgram;
- struct brw_vertex_program *bvp = brw_vertex_program(vp);
uint32_t old_prog_offset = brw->vs.prog_offset;
struct brw_vs_prog_data *old_prog_data = brw->vs.prog_data;
bool success;
- if (!vp)
+ if (!prog->_LinkedShaders[MESA_SHADER_VERTEX])
return true;
+ struct gl_vertex_program *vp = (struct gl_vertex_program *)
+ prog->_LinkedShaders[MESA_SHADER_VERTEX]->Program;
+ struct brw_vertex_program *bvp = brw_vertex_program(vp);
+
memset(&key, 0, sizeof(key));
key.program_string_id = bvp->id;