sna/gen4+: Add box emitters for the generic vertex paths

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

1 files changed, 111 insertions, 6 deletions

diff --git a/src/sna/gen4_vertex.c b/src/sna/gen4_vertex.c
index 0360b862..aaecb78b 100644
--- a/src/sna/gen4_vertex.c
+++ b/src/sna/gen4_vertex.c
@@ -320,7 +320,7 @@ emit_texcoord(struct sna *sna,
 	      int16_t x, int16_t y)
 {
 	if (channel->is_solid) {
-		OUT_VERTEX_F(x);
+		OUT_VERTEX_F(0.5);
 		return;
 	}
 
@@ -377,6 +377,63 @@ emit_primitive(struct sna *sna,
 		    r->dst.x,  r->dst.y);
 }
 
+sse2 inline static float *
+vemit_texcoord(float *v,
+	      const struct sna_composite_channel *channel,
+	      int16_t x, int16_t y)
+{
+	if (channel->is_solid) {
+		*v++ = 0.5;
+	} else {
+		x += channel->offset[0];
+		y += channel->offset[1];
+
+		if (channel->is_affine) {
+			float s, t;
+
+			sna_get_transformed_coordinates(x, y,
+							channel->transform,
+							&s, &t);
+			*v++ = s * channel->scale[0];
+			*v++ = t * channel->scale[1];
+		} else {
+			float s, t, w;
+
+			sna_get_transformed_coordinates_3d(x, y,
+							   channel->transform,
+							   &s, &t, &w);
+			*v++ = s * channel->scale[0];
+			*v++ = t * channel->scale[1];
+			*v++ = w;
+		}
+	}
+
+	return v;
+}
+
+sse2 force_inline static float *
+vemit_vertex(float *v,
+	     const struct sna_composite_op *op,
+	     int16_t x, int16_t y)
+{
+	*v++ = pack_2s(x, y);
+	return vemit_texcoord(v, &op->src, x, y);
+}
+
+sse2 fastcall static void
+emit_boxes(const struct sna_composite_op *op,
+	   const BoxRec *box, int nbox,
+	   float *v)
+{
+	do {
+		v = vemit_vertex(v, op, box->x2, box->y2);
+		v = vemit_vertex(v, op, box->x1, box->y2);
+		v = vemit_vertex(v, op, box->x1, box->y1);
+
+		box++;
+	} while (--nbox);
+}
+
 sse2 force_inline static void
 emit_vertex_mask(struct sna *sna,
 		 const struct sna_composite_op *op,
@@ -408,6 +465,32 @@ emit_primitive_mask(struct sna *sna,
 			 r->dst.x,  r->dst.y);
 }
 
+sse2 force_inline static float *
+vemit_vertex_mask(float *v,
+		  const struct sna_composite_op *op,
+		  int16_t x, int16_t y)
+{
+	*v++ = pack_2s(x, y);
+	v = vemit_texcoord(v, &op->src, x, y);
+	v = vemit_texcoord(v, &op->mask, x, y);
+	return v;
+}
+
+sse2 fastcall static void
+emit_boxes_mask(const struct sna_composite_op *op,
+		const BoxRec *box, int nbox,
+		float *v)
+{
+	do {
+		v = vemit_vertex_mask(v, op, box->x2, box->y2);
+		v = vemit_vertex_mask(v, op, box->x1, box->y2);
+		v = vemit_vertex_mask(v, op, box->x1, box->y1);
+
+		box++;
+	} while (--nbox);
+}
+
+
 sse2 fastcall static void
 emit_primitive_solid(struct sna *sna,
 		     const struct sna_composite_op *op,
@@ -1837,6 +1920,7 @@ unsigned gen4_choose_composite_emitter(struct sna *sna, struct sna_composite_op
 			}
 		} else {
 			tmp->prim_emit = emit_primitive_mask;
+			tmp->emit_boxes = emit_boxes_mask;
 			tmp->floats_per_vertex = 1;
 			vb = 0;
 			if (tmp->mask.is_solid) {
@@ -1943,6 +2027,7 @@ unsigned gen4_choose_composite_emitter(struct sna *sna, struct sna_composite_op
 			DBG(("%s: projective src, no mask\n", __FUNCTION__));
 			assert(!tmp->src.is_solid);
 			tmp->prim_emit = emit_primitive;
+			tmp->emit_boxes = emit_boxes;
 			tmp->floats_per_vertex = 4;
 			vb = 3;
 		}
@@ -1962,10 +2047,10 @@ emit_span_vertex(struct sna *sna,
 }
 
 sse2 fastcall static void
-emit_composite_spans_primitive(struct sna *sna,
-			       const struct sna_composite_spans_op *op,
-			       const BoxRec *box,
-			       float opacity)
+emit_span_primitive(struct sna *sna,
+		    const struct sna_composite_spans_op *op,
+		    const BoxRec *box,
+		    float opacity)
 {
 	emit_span_vertex(sna, op, box->x2, box->y2);
 	OUT_VERTEX_F(opacity);
@@ -1978,6 +2063,25 @@ emit_composite_spans_primitive(struct sna *sna,
 }
 
 sse2 fastcall static void
+emit_span_boxes(const struct sna_composite_spans_op *op,
+		const struct sna_opacity_box *b, int nbox,
+		float *v)
+{
+	do {
+		v = vemit_vertex(v, &op->base, b->box.x2, b->box.y2);
+		*v++ = b->alpha;
+
+		v = vemit_vertex(v, &op->base, b->box.x1, b->box.y2);
+		*v++ = b->alpha;
+
+		v = vemit_vertex(v, &op->base, b->box.x1, b->box.y1);
+		*v++ = b->alpha;
+
+		b++;
+	} while (--nbox);
+}
+
+sse2 fastcall static void
 emit_span_solid(struct sna *sna,
 		const struct sna_composite_spans_op *op,
 		const BoxRec *box,
@@ -3085,7 +3189,8 @@ unsigned gen4_choose_spans_emitter(struct sna *sna,
 		vb = 1 << 2 | 2;
 	} else {
 		DBG(("%s: projective transform\n", __FUNCTION__));
-		tmp->prim_emit = emit_composite_spans_primitive;
+		tmp->prim_emit = emit_span_primitive;
+		tmp->emit_boxes = emit_span_boxes;
 		tmp->base.floats_per_vertex = 5;
 		vb = 1 << 2 | 3;
 	}