+
+static bool
+is_drawing_points(const struct brw_context *brw)
+{
+ /* Determine if the primitives *reaching the SF* are points */
+ /* _NEW_POLYGON */
+ if (brw->ctx.Polygon.FrontMode == GL_POINT ||
+ brw->ctx.Polygon.BackMode == GL_POINT) {
+ return true;
+ }
+
+ if (brw->geometry_program) {
+ /* BRW_NEW_GEOMETRY_PROGRAM */
+ return brw->geometry_program->OutputType == GL_POINTS;
+ } else {
+ /* BRW_NEW_PRIMITIVE */
+ return brw->primitive == _3DPRIM_POINTLIST;
+ }
+}
+
+
+/**
+ * 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,
+ uint32_t *urb_entry_read_offset)
+{
+ uint32_t max_source_attr = 0;
+
+ *point_sprite_enables = 0;
+ *flat_enables = 0;
+
+ *urb_entry_read_offset = BRW_SF_URB_ENTRY_READ_OFFSET;
+
+ /* BRW_NEW_FRAGMENT_PROGRAM
+ *
+ * If the fragment shader reads VARYING_SLOT_LAYER, then we need to pass in
+ * the full vertex header. Otherwise, we can program the SF to start
+ * reading at an offset of 1 (2 varying slots) to skip unnecessary data:
+ * - VARYING_SLOT_PSIZ and BRW_VARYING_SLOT_NDC on gen4-5
+ * - VARYING_SLOT_{PSIZ,LAYER} and VARYING_SLOT_POS on gen6+
+ */
+
+ bool fs_needs_vue_header = brw->fragment_program->Base.InputsRead &
+ (VARYING_BIT_LAYER | VARYING_BIT_VIEWPORT);
+
+ *urb_entry_read_offset = fs_needs_vue_header ? 0 : 1;
+
+ /* _NEW_LIGHT */
+ bool shade_model_flat = brw->ctx.Light.ShadeModel == GL_FLAT;
+
+ /* From the Ivybridge PRM, Vol 2 Part 1, 3DSTATE_SBE,
+ * description of dw10 Point Sprite Texture Coordinate Enable:
+ *
+ * "This field must be programmed to zero when non-point primitives
+ * are rendered."
+ *
+ * The SandyBridge PRM doesn't explicitly say that point sprite enables
+ * must be programmed to zero when rendering non-point primitives, but
+ * the IvyBridge PRM does, and if we don't, we get garbage.
+ *
+ * This is not required on Haswell, as the hardware ignores this state
+ * when drawing non-points -- although we do still need to be careful to
+ * correctly set the attr overrides.
+ */
+ /* BRW_NEW_PRIMITIVE | BRW_NEW_GEOMETRY_PROGRAM */
+ bool drawing_points = is_drawing_points(brw);
+
+ /* 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++) {
+ /* BRW_NEW_FRAGMENT_PROGRAM */
+ enum glsl_interp_qualifier interp_qualifier =
+ brw->fragment_program->InterpQualifier[attr];
+ bool is_gl_Color = attr == VARYING_SLOT_COL0 || attr == VARYING_SLOT_COL1;
+ /* BRW_NEW_FS_PROG_DATA */
+ int input_index = brw->wm.prog_data->urb_setup[attr];
+
+ if (input_index < 0)
+ continue;
+
+ /* _NEW_POINT */
+ bool point_sprite = false;
+ if (drawing_points) {
+ if (brw->ctx.Point.PointSprite &&
+ (attr >= VARYING_SLOT_TEX0 && attr <= VARYING_SLOT_TEX7) &&
+ brw->ctx.Point.CoordReplace[attr - VARYING_SLOT_TEX0]) {
+ point_sprite = true;
+ }
+
+ if (attr == VARYING_SLOT_PNTC)
+ point_sprite = true;
+
+ if (point_sprite)
+ *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 = point_sprite ? 0 :
+ 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;
+}
+
+