* the VUE that are not needed by the fragment shader. It is measured in
* 256-bit increments.
*/
-uint32_t
-get_attr_override(struct brw_vue_map *vue_map, int urb_entry_read_offset,
- int fs_attr, bool two_side_color)
+static uint32_t
+get_attr_override(const struct brw_vue_map *vue_map, int urb_entry_read_offset,
+ int fs_attr, bool two_side_color, uint32_t *max_source_attr)
{
- int attr_override, slot;
- int vs_attr = _mesa_frag_attrib_to_vert_result(fs_attr);
- if (vs_attr < 0 || vs_attr == VERT_RESULT_HPOS) {
- /* These attributes will be overwritten by the fragment shader's
- * interpolation code (see emit_interp() in brw_wm_fp.c), so just let
- * them reference the first available attribute.
- */
- return 0;
- }
-
/* Find the VUE slot for this attribute. */
- slot = vue_map->vert_result_to_slot[vs_attr];
+ int slot = vue_map->varying_to_slot[fs_attr];
/* If there was only a back color written but not front, use back
* as the color instead of undefined
*/
- if (slot == -1 && vs_attr == VERT_RESULT_COL0)
- slot = vue_map->vert_result_to_slot[VERT_RESULT_BFC0];
- if (slot == -1 && vs_attr == VERT_RESULT_COL1)
- slot = vue_map->vert_result_to_slot[VERT_RESULT_BFC1];
+ if (slot == -1 && fs_attr == VARYING_SLOT_COL0)
+ slot = vue_map->varying_to_slot[VARYING_SLOT_BFC0];
+ if (slot == -1 && fs_attr == VARYING_SLOT_COL1)
+ slot = vue_map->varying_to_slot[VARYING_SLOT_BFC1];
if (slot == -1) {
/* This attribute does not exist in the VUE--that means that the vertex
- * shader did not write to it. Behavior is undefined in this case, so
+ * shader did not write to it. This means that either:
+ *
+ * (a) This attribute is a texture coordinate, and it is going to be
+ * replaced with point coordinates (as a consequence of a call to
+ * glTexEnvi(GL_POINT_SPRITE, GL_COORD_REPLACE, GL_TRUE)), so the
+ * hardware will ignore whatever attribute override we supply.
+ *
+ * (b) This attribute is read by the fragment shader but not written by
+ * the vertex shader, so its value is undefined. Therefore the
+ * attribute override we supply doesn't matter.
+ *
+ * In either case the attribute override we supply doesn't matter, so
* just reference the first available attribute.
*/
return 0;
* Each increment of urb_entry_read_offset represents a 256-bit value, so
* it counts for two 128-bit VUE slots.
*/
- attr_override = slot - 2 * urb_entry_read_offset;
- assert (attr_override >= 0 && attr_override < 32);
+ int source_attr = slot - 2 * urb_entry_read_offset;
+ assert(source_attr >= 0 && source_attr < 32);
/* If we are doing two-sided color, and the VUE slot following this one
* represents a back-facing color, then we need to instruct the SF unit to
* do back-facing swizzling.
*/
- if (two_side_color) {
- if (vue_map->slot_to_vert_result[slot] == VERT_RESULT_COL0 &&
- vue_map->slot_to_vert_result[slot+1] == VERT_RESULT_BFC0)
- attr_override |= (ATTRIBUTE_SWIZZLE_INPUTATTR_FACING << ATTRIBUTE_SWIZZLE_SHIFT);
- else if (vue_map->slot_to_vert_result[slot] == VERT_RESULT_COL1 &&
- vue_map->slot_to_vert_result[slot+1] == VERT_RESULT_BFC1)
- attr_override |= (ATTRIBUTE_SWIZZLE_INPUTATTR_FACING << ATTRIBUTE_SWIZZLE_SHIFT);
+ bool swizzling = two_side_color &&
+ ((vue_map->slot_to_varying[slot] == VARYING_SLOT_COL0 &&
+ vue_map->slot_to_varying[slot+1] == VARYING_SLOT_BFC0) ||
+ (vue_map->slot_to_varying[slot] == VARYING_SLOT_COL1 &&
+ vue_map->slot_to_varying[slot+1] == VARYING_SLOT_BFC1));
+
+ /* Update max_source_attr. If swizzling, the SF will read this slot + 1. */
+ if (*max_source_attr < source_attr + swizzling)
+ *max_source_attr = source_attr + swizzling;
+
+ if (swizzling) {
+ return source_attr |
+ (ATTRIBUTE_SWIZZLE_INPUTATTR_FACING << ATTRIBUTE_SWIZZLE_SHIFT);
}
- return attr_override;
+ return source_attr;
}
+
+/**
+ * Create the mapping from the FS inputs we produce to the previous pipeline
+ * stage (GS or VS) outputs they source from.
+ */
+void
+calculate_attr_overrides(const struct brw_context *brw,
+ uint16_t *attr_overrides,
+ uint32_t *point_sprite_enables,
+ uint32_t *flat_enables,
+ uint32_t *urb_entry_read_length)
+{
+ const int urb_entry_read_offset = BRW_SF_URB_ENTRY_READ_OFFSET;
+ uint32_t max_source_attr = 0;
+
+ /* _NEW_LIGHT */
+ bool shade_model_flat = brw->ctx.Light.ShadeModel == GL_FLAT;
+
+ /* Initialize all the attr_overrides to 0. In the loop below we'll modify
+ * just the ones that correspond to inputs used by the fs.
+ */
+ memset(attr_overrides, 0, 16*sizeof(*attr_overrides));
+
+ for (int attr = 0; attr < VARYING_SLOT_MAX; attr++) {
+ enum glsl_interp_qualifier interp_qualifier =
+ brw->fragment_program->InterpQualifier[attr];
+ bool is_gl_Color = attr == VARYING_SLOT_COL0 || attr == VARYING_SLOT_COL1;
+ /* CACHE_NEW_WM_PROG */
+ int input_index = brw->wm.prog_data->urb_setup[attr];
+
+ if (input_index < 0)
+ continue;
+
+ /* _NEW_POINT */
+ if (brw->ctx.Point.PointSprite &&
+ (attr >= VARYING_SLOT_TEX0 && attr <= VARYING_SLOT_TEX7) &&
+ brw->ctx.Point.CoordReplace[attr - VARYING_SLOT_TEX0]) {
+ *point_sprite_enables |= (1 << input_index);
+ }
+
+ if (attr == VARYING_SLOT_PNTC)
+ *point_sprite_enables |= (1 << input_index);
+
+ /* flat shading */
+ if (interp_qualifier == INTERP_QUALIFIER_FLAT ||
+ (shade_model_flat && is_gl_Color &&
+ interp_qualifier == INTERP_QUALIFIER_NONE))
+ *flat_enables |= (1 << input_index);
+
+ /* BRW_NEW_VUE_MAP_GEOM_OUT | _NEW_LIGHT | _NEW_PROGRAM */
+ uint16_t attr_override =
+ get_attr_override(&brw->vue_map_geom_out,
+ urb_entry_read_offset, attr,
+ brw->ctx.VertexProgram._TwoSideEnabled,
+ &max_source_attr);
+
+ /* The hardware can only do the overrides on 16 overrides at a
+ * time, and the other up to 16 have to be lined up so that the
+ * input index = the output index. We'll need to do some
+ * tweaking to make sure that's the case.
+ */
+ if (input_index < 16)
+ attr_overrides[input_index] = attr_override;
+ else
+ assert(attr_override == input_index);
+ }
+
+ /* From the Sandy Bridge PRM, Volume 2, Part 1, documentation for
+ * 3DSTATE_SF DWord 1 bits 15:11, "Vertex URB Entry Read Length":
+ *
+ * "This field should be set to the minimum length required to read the
+ * maximum source attribute. The maximum source attribute is indicated
+ * by the maximum value of the enabled Attribute # Source Attribute if
+ * Attribute Swizzle Enable is set, Number of Output Attributes-1 if
+ * enable is not set.
+ * read_length = ceiling((max_source_attr + 1) / 2)
+ *
+ * [errata] Corruption/Hang possible if length programmed larger than
+ * recommended"
+ *
+ * Similar text exists for Ivy Bridge.
+ */
+ *urb_entry_read_length = ALIGN(max_source_attr + 1, 2) / 2;
+}
+
+
static void
upload_sf_state(struct brw_context *brw)
{
- struct intel_context *intel = &brw->intel;
- struct gl_context *ctx = &intel->ctx;
- uint32_t urb_entry_read_length;
- /* BRW_NEW_FRAGMENT_PROGRAM */
- uint32_t num_outputs = _mesa_bitcount_64(brw->fragment_program->Base.InputsRead);
- /* _NEW_LIGHT */
- bool shade_model_flat = ctx->Light.ShadeModel == GL_FLAT;
+ struct gl_context *ctx = &brw->ctx;
+ /* CACHE_NEW_WM_PROG */
+ uint32_t num_outputs = brw->wm.prog_data->num_varying_inputs;
uint32_t dw1, dw2, dw3, dw4, dw16, dw17;
int i;
/* _NEW_BUFFER */
- bool render_to_fbo = _mesa_is_user_fbo(brw->intel.ctx.DrawBuffer);
- bool multisampled_fbo = false;
- if (ctx->DrawBuffer->_ColorDrawBuffers[0])
- multisampled_fbo = ctx->DrawBuffer->_ColorDrawBuffers[0]->NumSamples > 0;
+ bool render_to_fbo = _mesa_is_user_fbo(ctx->DrawBuffer);
+ bool multisampled_fbo = ctx->DrawBuffer->Visual.samples > 1;
- int attr = 0, input_index = 0;
- int urb_entry_read_offset = 1;
+ const int urb_entry_read_offset = BRW_SF_URB_ENTRY_READ_OFFSET;
float point_size;
- uint16_t attr_overrides[FRAG_ATTRIB_MAX];
+ uint16_t attr_overrides[16];
uint32_t point_sprite_origin;
- /* CACHE_NEW_VS_PROG */
- urb_entry_read_length = ((brw->vs.prog_data->vue_map.num_slots + 1) / 2 -
- urb_entry_read_offset);
- if (urb_entry_read_length == 0) {
- /* Setting the URB entry read length to 0 causes undefined behavior, so
- * if we have no URB data to read, set it to 1.
- */
- urb_entry_read_length = 1;
- }
-
- dw1 =
- GEN6_SF_SWIZZLE_ENABLE |
- num_outputs << GEN6_SF_NUM_OUTPUTS_SHIFT |
- urb_entry_read_length << GEN6_SF_URB_ENTRY_READ_LENGTH_SHIFT |
- urb_entry_read_offset << GEN6_SF_URB_ENTRY_READ_OFFSET_SHIFT;
+ dw1 = GEN6_SF_SWIZZLE_ENABLE | num_outputs << GEN6_SF_NUM_OUTPUTS_SHIFT;
dw2 = GEN6_SF_STATISTICS_ENABLE |
GEN6_SF_VIEWPORT_TRANSFORM_ENABLE;
(1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT);
}
- /* Create the mapping from the FS inputs we produce to the VS outputs
- * they source from.
+ /* BRW_NEW_VUE_MAP_GEOM_OUT | _NEW_POINT | _NEW_LIGHT | _NEW_PROGRAM |
+ * CACHE_NEW_WM_PROG
*/
- for (; attr < FRAG_ATTRIB_MAX; attr++) {
- enum glsl_interp_qualifier interp_qualifier =
- brw->fragment_program->InterpQualifier[attr];
- bool is_gl_Color = attr == FRAG_ATTRIB_COL0 || attr == FRAG_ATTRIB_COL1;
-
- if (!(brw->fragment_program->Base.InputsRead & BITFIELD64_BIT(attr)))
- continue;
-
- /* _NEW_POINT */
- if (ctx->Point.PointSprite &&
- (attr >= FRAG_ATTRIB_TEX0 && attr <= FRAG_ATTRIB_TEX7) &&
- ctx->Point.CoordReplace[attr - FRAG_ATTRIB_TEX0]) {
- dw16 |= (1 << input_index);
- }
-
- if (attr == FRAG_ATTRIB_PNTC)
- dw16 |= (1 << input_index);
-
- /* flat shading */
- if (interp_qualifier == INTERP_QUALIFIER_FLAT ||
- (shade_model_flat && is_gl_Color &&
- interp_qualifier == INTERP_QUALIFIER_NONE))
- dw17 |= (1 << input_index);
-
- /* The hardware can only do the overrides on 16 overrides at a
- * time, and the other up to 16 have to be lined up so that the
- * input index = the output index. We'll need to do some
- * tweaking to make sure that's the case.
- */
- assert(input_index < 16 || attr == input_index);
-
- /* CACHE_NEW_VS_PROG | _NEW_LIGHT | _NEW_PROGRAM */
- attr_overrides[input_index++] =
- get_attr_override(&brw->vs.prog_data->vue_map,
- urb_entry_read_offset, attr,
- ctx->VertexProgram._TwoSideEnabled);
- }
-
- for (; input_index < FRAG_ATTRIB_MAX; input_index++)
- attr_overrides[input_index] = 0;
+ uint32_t urb_entry_read_length;
+ calculate_attr_overrides(brw, attr_overrides, &dw16, &dw17,
+ &urb_entry_read_length);
+ dw1 |= (urb_entry_read_length << GEN6_SF_URB_ENTRY_READ_LENGTH_SHIFT |
+ urb_entry_read_offset << GEN6_SF_URB_ENTRY_READ_OFFSET_SHIFT);
BEGIN_BATCH(20);
OUT_BATCH(_3DSTATE_SF << 16 | (20 - 2));
_NEW_POINT |
_NEW_MULTISAMPLE),
.brw = (BRW_NEW_CONTEXT |
- BRW_NEW_FRAGMENT_PROGRAM),
- .cache = CACHE_NEW_VS_PROG
+ BRW_NEW_FRAGMENT_PROGRAM |
+ BRW_NEW_VUE_MAP_GEOM_OUT),
+ .cache = CACHE_NEW_WM_PROG
},
.emit = upload_sf_state,
};