/**
* 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).
*/
-static void
-brw_compute_vue_map(struct brw_context *brw, struct brw_vs_compile *c,
+void
+brw_compute_vue_map(struct brw_context *brw, struct brw_vue_map *vue_map,
GLbitfield64 slots_valid)
{
- const struct intel_context *intel = &brw->intel;
- struct brw_vue_map *vue_map = &c->prog_data.vue_map;
-
- /* Prior to Gen6, don't assign a slot for VARYING_SLOT_CLIP_VERTEX, since
- * it is unsupported.
- */
- if (intel->gen < 6)
- slots_valid &= ~VARYING_BIT_CLIP_VERTEX;
-
vue_map->slots_valid = slots_valid;
int i;
+ /* gl_Layer doesn't get its own varying slot--it's stored in the virst VUE
+ * slot (VARYING_SLOT_PSIZ).
+ */
+ slots_valid &= ~VARYING_BIT_LAYER;
+
/* Make sure that the values we store in vue_map->varying_to_slot and
* vue_map->slot_to_varying won't overflow the signed chars that are used
* to store them. Note that since vue_map->slot_to_varying sometimes holds
/* VUE header: format depends on chip generation and whether clipping is
* enabled.
*/
- switch (intel->gen) {
+ switch (brw->gen) {
case 4:
+ case 5:
/* There are 8 dwords in VUE header pre-Ironlake:
* dword 0-3 is indices, point width, clip flags.
* dword 4-7 is ndc position
* dword 8-11 is the first vertex data.
- */
- assign_vue_slot(vue_map, VARYING_SLOT_PSIZ);
- assign_vue_slot(vue_map, BRW_VARYING_SLOT_NDC);
- assign_vue_slot(vue_map, VARYING_SLOT_POS);
- break;
- case 5:
- /* There are 20 DWs (D0-D19) in VUE header on Ironlake:
- * dword 0-3 of the header is indices, point width, clip flags.
- * dword 4-7 is the ndc position
- * dword 8-11 of the vertex header is the 4D space position
- * dword 12-19 of the vertex header is the user clip distance.
- * dword 20-23 is a pad so that the vertex element data is aligned
- * dword 24-27 is the first vertex data we fill.
*
- * Note: future pipeline stages expect 4D space position to be
- * contiguous with the other varyings, so we make dword 24-27 a
- * duplicate copy of the 4D space position.
+ * On Ironlake the VUE header is nominally 20 dwords, but the hardware
+ * will accept the same header layout as Gen4 [and should be a bit faster]
*/
assign_vue_slot(vue_map, VARYING_SLOT_PSIZ);
assign_vue_slot(vue_map, BRW_VARYING_SLOT_NDC);
- assign_vue_slot(vue_map, BRW_VARYING_SLOT_POS_DUPLICATE);
- assign_vue_slot(vue_map, VARYING_SLOT_CLIP_DIST0);
- assign_vue_slot(vue_map, VARYING_SLOT_CLIP_DIST1);
- assign_vue_slot(vue_map, BRW_VARYING_SLOT_PAD);
assign_vue_slot(vue_map, VARYING_SLOT_POS);
break;
case 6:
*/
assign_vue_slot(vue_map, VARYING_SLOT_PSIZ);
assign_vue_slot(vue_map, VARYING_SLOT_POS);
- if (c->key.userclip_active) {
+ if (slots_valid & BITFIELD64_BIT(VARYING_SLOT_CLIP_DIST0))
assign_vue_slot(vue_map, VARYING_SLOT_CLIP_DIST0);
+ if (slots_valid & BITFIELD64_BIT(VARYING_SLOT_CLIP_DIST1))
assign_vue_slot(vue_map, VARYING_SLOT_CLIP_DIST1);
- }
+
/* front and back colors need to be consecutive so that we can use
* ATTRIBUTE_SWIZZLE_INPUTATTR_FACING to swizzle them when doing
* two-sided color.
}
}
+
bool
-brw_vs_prog_data_compare(const void *in_a, const void *in_b,
- int aux_size, const void *in_key)
+brw_vs_prog_data_compare(const void *in_a, const void *in_b)
{
const struct brw_vs_prog_data *a = in_a;
const struct brw_vs_prog_data *b = in_b;
- /* Compare all the struct up to the pointers. */
- if (memcmp(a, b, offsetof(struct brw_vs_prog_data, param)))
- return false;
-
- if (memcmp(a->param, b->param, a->nr_params * sizeof(void *)))
+ /* Compare the base vec4 structure. */
+ if (!brw_vec4_prog_data_compare(&a->base, &b->base))
return false;
- if (memcmp(a->pull_param, b->pull_param, a->nr_pull_params * sizeof(void *)))
+ /* Compare the rest of the struct. */
+ const unsigned offset = sizeof(struct brw_vec4_prog_data);
+ if (memcmp(((char *) a) + offset, ((char *) b) + offset,
+ sizeof(struct brw_vs_prog_data) - offset)) {
return false;
+ }
return true;
}
struct brw_vertex_program *vp,
struct brw_vs_prog_key *key)
{
- struct intel_context *intel = &brw->intel;
GLuint program_size;
const GLuint *program;
struct brw_vs_compile c;
+ struct brw_vs_prog_data prog_data;
void *mem_ctx;
int i;
struct gl_shader *vs = NULL;
memset(&c, 0, sizeof(c));
memcpy(&c.key, key, sizeof(*key));
+ memset(&prog_data, 0, sizeof(prog_data));
mem_ctx = ralloc_context(NULL);
} else {
param_count = vp->program.Base.Parameters->NumParameters * 4;
}
- /* We also upload clip plane data as uniforms */
- param_count += MAX_CLIP_PLANES * 4;
+ /* vec4_visitor::setup_uniform_clipplane_values() also uploads user clip
+ * planes as uniforms.
+ */
+ param_count += c.key.base.nr_userclip_plane_consts * 4;
- c.prog_data.param = rzalloc_array(NULL, const float *, param_count);
- c.prog_data.pull_param = rzalloc_array(NULL, const float *, param_count);
+ prog_data.base.param = rzalloc_array(NULL, const float *, param_count);
+ prog_data.base.pull_param = rzalloc_array(NULL, const float *, param_count);
GLbitfield64 outputs_written = vp->program.Base.OutputsWritten;
- c.prog_data.inputs_read = vp->program.Base.InputsRead;
+ prog_data.inputs_read = vp->program.Base.InputsRead;
if (c.key.copy_edgeflag) {
outputs_written |= BITFIELD64_BIT(VARYING_SLOT_EDGE);
- c.prog_data.inputs_read |= VERT_BIT_EDGEFLAG;
+ prog_data.inputs_read |= VERT_BIT_EDGEFLAG;
}
- if (intel->gen < 6) {
+ if (brw->gen < 6) {
/* Put dummy slots into the VUE for the SF to put the replaced
* point sprite coords in. We shouldn't need these dummy slots,
* which take up precious URB space, but it would mean that the SF
if (c.key.point_coord_replace & (1 << i))
outputs_written |= BITFIELD64_BIT(VARYING_SLOT_TEX0 + i);
}
+
+ /* if back colors are written, allocate slots for front colors too */
+ if (outputs_written & BITFIELD64_BIT(VARYING_SLOT_BFC0))
+ outputs_written |= BITFIELD64_BIT(VARYING_SLOT_COL0);
+ if (outputs_written & BITFIELD64_BIT(VARYING_SLOT_BFC1))
+ outputs_written |= BITFIELD64_BIT(VARYING_SLOT_COL1);
+ }
+
+ /* In order for legacy clipping to work, we need to populate the clip
+ * distance varying slots whenever clipping is enabled, even if the vertex
+ * shader doesn't write to gl_ClipDistance.
+ */
+ if (c.key.base.userclip_active) {
+ outputs_written |= BITFIELD64_BIT(VARYING_SLOT_CLIP_DIST0);
+ outputs_written |= BITFIELD64_BIT(VARYING_SLOT_CLIP_DIST1);
}
- brw_compute_vue_map(brw, &c, outputs_written);
+ brw_compute_vue_map(brw, &prog_data.base.vue_map, outputs_written);
if (0) {
_mesa_fprint_program_opt(stdout, &c.vp->program.Base, PROG_PRINT_DEBUG,
/* Emit GEN4 code.
*/
- program = brw_vs_emit(brw, prog, &c, mem_ctx, &program_size);
+ program = brw_vs_emit(brw, prog, &c, &prog_data, mem_ctx, &program_size);
if (program == NULL) {
ralloc_free(mem_ctx);
return false;
}
- 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) {
+ if (c.base.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*REG_SIZE);
+ prog_data.base.total_scratch
+ = brw_get_scratch_size(c.base.last_scratch*REG_SIZE);
- brw_get_scratch_bo(intel, &brw->vs.scratch_bo,
- c.prog_data.total_scratch * brw->max_vs_threads);
+ brw_get_scratch_bo(brw, &brw->vs.base.scratch_bo,
+ prog_data.base.total_scratch * brw->max_vs_threads);
}
brw_upload_cache(&brw->cache, BRW_VS_PROG,
&c.key, sizeof(c.key),
program, program_size,
- &c.prog_data, sizeof(c.prog_data),
- &brw->vs.prog_offset, &brw->vs.prog_data);
+ &prog_data, sizeof(prog_data),
+ &brw->vs.base.prog_offset, &brw->vs.prog_data);
ralloc_free(mem_ctx);
return true;
}
static bool
-key_debug(struct intel_context *intel, const char *name, int a, int b)
+key_debug(struct brw_context *brw, const char *name, int a, int b)
{
if (a != b) {
perf_debug(" %s %d->%d\n", name, a, b);
struct gl_shader_program *prog,
const struct brw_vs_prog_key *key)
{
- struct intel_context *intel = &brw->intel;
struct brw_cache_item *c = NULL;
const struct brw_vs_prog_key *old_key = NULL;
bool found = false;
if (c->cache_id == BRW_VS_PROG) {
old_key = c->key;
- if (old_key->program_string_id == key->program_string_id)
+ if (old_key->base.program_string_id == key->base.program_string_id)
break;
}
}
}
for (unsigned int i = 0; i < VERT_ATTRIB_MAX; i++) {
- found |= key_debug(intel, "Vertex attrib w/a flags",
+ found |= key_debug(brw, "Vertex attrib w/a flags",
old_key->gl_attrib_wa_flags[i],
key->gl_attrib_wa_flags[i]);
}
- found |= key_debug(intel, "user clip flags",
- old_key->userclip_active, key->userclip_active);
+ found |= key_debug(brw, "user clip flags",
+ old_key->base.userclip_active, key->base.userclip_active);
- found |= key_debug(intel, "user clipping planes as push constants",
- old_key->nr_userclip_plane_consts,
- key->nr_userclip_plane_consts);
+ found |= key_debug(brw, "user clipping planes as push constants",
+ old_key->base.nr_userclip_plane_consts,
+ key->base.nr_userclip_plane_consts);
- found |= key_debug(intel, "clip distance enable",
- old_key->uses_clip_distance, key->uses_clip_distance);
- found |= key_debug(intel, "clip plane enable bitfield",
- old_key->userclip_planes_enabled_gen_4_5,
- key->userclip_planes_enabled_gen_4_5);
- found |= key_debug(intel, "copy edgeflag",
+ found |= key_debug(brw, "clip distance enable",
+ old_key->base.uses_clip_distance, key->base.uses_clip_distance);
+ found |= key_debug(brw, "copy edgeflag",
old_key->copy_edgeflag, key->copy_edgeflag);
- found |= key_debug(intel, "PointCoord replace",
+ found |= key_debug(brw, "PointCoord replace",
old_key->point_coord_replace, key->point_coord_replace);
- found |= key_debug(intel, "vertex color clamping",
- old_key->clamp_vertex_color, key->clamp_vertex_color);
+ found |= key_debug(brw, "vertex color clamping",
+ old_key->base.clamp_vertex_color, key->base.clamp_vertex_color);
- found |= brw_debug_recompile_sampler_key(intel, &old_key->tex, &key->tex);
+ found |= brw_debug_recompile_sampler_key(brw, &old_key->base.tex,
+ &key->base.tex);
if (!found) {
perf_debug(" Something else\n");
}
}
+
+void
+brw_setup_vec4_key_clip_info(struct brw_context *brw,
+ struct brw_vec4_prog_key *key,
+ bool program_uses_clip_distance)
+{
+ struct gl_context *ctx = &brw->ctx;
+
+ key->userclip_active = (ctx->Transform.ClipPlanesEnabled != 0);
+ key->uses_clip_distance = program_uses_clip_distance;
+ if (key->userclip_active && !key->uses_clip_distance) {
+ key->nr_userclip_plane_consts
+ = _mesa_logbase2(ctx->Transform.ClipPlanesEnabled) + 1;
+ }
+}
+
+
static void brw_upload_vs_prog(struct brw_context *brw)
{
- struct intel_context *intel = &brw->intel;
- struct gl_context *ctx = &intel->ctx;
+ struct gl_context *ctx = &brw->ctx;
struct brw_vs_prog_key key;
/* BRW_NEW_VERTEX_PROGRAM */
struct brw_vertex_program *vp =
/* Just upload the program verbatim for now. Always send it all
* the inputs it asks for, whether they are varying or not.
*/
- key.program_string_id = vp->id;
- 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;
- }
- }
+ key.base.program_string_id = vp->id;
+ brw_setup_vec4_key_clip_info(brw, &key.base, vp->program.UsesClipDistance);
/* _NEW_POLYGON */
- if (intel->gen < 6) {
+ if (brw->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;
+ key.base.clamp_vertex_color = ctx->Light._ClampVertexColor;
/* _NEW_POINT */
- if (intel->gen < 6 && ctx->Point.PointSprite) {
+ if (brw->gen < 6 && ctx->Point.PointSprite) {
for (i = 0; i < 8; i++) {
if (ctx->Point.CoordReplace[i])
key.point_coord_replace |= (1 << i);
}
/* _NEW_TEXTURE */
- brw_populate_sampler_prog_key_data(ctx, prog, &key.tex);
+ brw_populate_sampler_prog_key_data(ctx, prog, brw->vs.base.sampler_count,
+ &key.base.tex);
/* BRW_NEW_VERTICES */
- if (intel->gen < 8 && !intel->is_haswell) {
+ if (brw->gen < 8 && !brw->is_haswell) {
/* Prior to Haswell, the hardware can't natively support GL_FIXED or
* 2_10_10_10_REV vertex formats. Set appropriate workaround flags.
*/
if (!brw_search_cache(&brw->cache, BRW_VS_PROG,
&key, sizeof(key),
- &brw->vs.prog_offset, &brw->vs.prog_data)) {
+ &brw->vs.base.prog_offset, &brw->vs.prog_data)) {
bool success = do_vs_prog(brw, ctx->Shader.CurrentVertexProgram,
vp, &key);
-
+ (void) success;
assert(success);
}
- if (memcmp(&brw->vs.prog_data->vue_map, &brw->vue_map_geom_out,
+ brw->vs.base.prog_data = &brw->vs.prog_data->base.base;
+
+ if (memcmp(&brw->vs.prog_data->base.vue_map, &brw->vue_map_geom_out,
sizeof(brw->vue_map_geom_out)) != 0) {
- brw->vue_map_geom_out = brw->vs.prog_data->vue_map;
- brw->state.dirty.brw |= BRW_NEW_VUE_MAP_GEOM_OUT;
+ brw->vue_map_vs = brw->vs.prog_data->base.vue_map;
+ brw->state.dirty.brw |= BRW_NEW_VUE_MAP_VS;
+ if (brw->gen < 7) {
+ /* No geometry shader support, so the VS VUE map is the VUE map for
+ * the output of the "geometry" portion of the pipeline.
+ */
+ brw->vue_map_geom_out = brw->vue_map_vs;
+ brw->state.dirty.brw |= BRW_NEW_VUE_MAP_GEOM_OUT;
+ }
}
}
{
struct brw_context *brw = brw_context(ctx);
struct brw_vs_prog_key key;
- uint32_t old_prog_offset = brw->vs.prog_offset;
+ uint32_t old_prog_offset = brw->vs.base.prog_offset;
struct brw_vs_prog_data *old_prog_data = brw->vs.prog_data;
bool success;
memset(&key, 0, sizeof(key));
- key.program_string_id = bvp->id;
- key.clamp_vertex_color = true;
+ key.base.program_string_id = bvp->id;
+ key.base.clamp_vertex_color = ctx->API == API_OPENGL_COMPAT;
- for (int i = 0; i < MAX_SAMPLERS; i++) {
+ unsigned sampler_count = _mesa_fls(vp->Base.SamplersUsed);
+ for (unsigned i = 0; i < sampler_count; i++) {
if (vp->Base.ShadowSamplers & (1 << i)) {
/* Assume DEPTH_TEXTURE_MODE is the default: X, X, X, 1 */
- key.tex.swizzles[i] =
+ key.base.tex.swizzles[i] =
MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_ONE);
} else {
/* Color sampler: assume no swizzling. */
- key.tex.swizzles[i] = SWIZZLE_XYZW;
+ key.base.tex.swizzles[i] = SWIZZLE_XYZW;
}
}
success = do_vs_prog(brw, prog, bvp, &key);
- brw->vs.prog_offset = old_prog_offset;
+ brw->vs.base.prog_offset = old_prog_offset;
brw->vs.prog_data = old_prog_data;
return success;
}
+
void
brw_vs_prog_data_free(const void *in_prog_data)
{
const struct brw_vs_prog_data *prog_data = in_prog_data;
- ralloc_free((void *)prog_data->param);
- ralloc_free((void *)prog_data->pull_param);
+ brw_vec4_prog_data_free(&prog_data->base);
}
projects / mesa.git / blobdiff