sna: Add a fast path for the most common fallback for CPU-CPU blits

This path will mostly be upload for individual glyph uploads, for which
the malloc overhead is significant.

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

3 files changed, 46 insertions, 7 deletions

diff --git a/src/sna/sna_blt.c b/src/sna/sna_blt.c
index beef9a65..b4ef083f 100644
--- a/src/sna/sna_blt.c
+++ b/src/sna/sna_blt.c
@@ -1953,13 +1953,6 @@ prepare_blt_put(struct sna *sna,
 	return true;
 }
 
-#define alphaless(format) PICT_FORMAT(PICT_FORMAT_BPP(format),		\
-				      PICT_FORMAT_TYPE(format),		\
-				      0,				\
-				      PICT_FORMAT_R(format),		\
-				      PICT_FORMAT_G(format),		\
-				      PICT_FORMAT_B(format))
-
 static bool
 is_clear(PixmapPtr pixmap)
 {
diff --git a/src/sna/sna_composite.c b/src/sna/sna_composite.c
index ea4a1003..36e2bdee 100644
--- a/src/sna/sna_composite.c
+++ b/src/sna/sna_composite.c
@@ -470,6 +470,7 @@ sna_composite_fb(CARD8 op,
 	int src_xoff, src_yoff;
 	int msk_xoff, msk_yoff;
 	int dst_xoff, dst_yoff;
+	int16_t tx, ty;
 	unsigned flags;
 
 	DBG(("%s -- op=%d, fallback dst=(%d, %d)+(%d, %d), size=(%d, %d): region=((%d,%d), (%d, %d))\n",
@@ -519,6 +520,44 @@ sna_composite_fb(CARD8 op,
 	if (mask)
 		validate_source(mask);
 
+	if (mask == NULL &&
+	    src->pDrawable &&
+	    src->filter != PictFilterConvolution &&
+	    (op == PictOpSrc || (op == PictOpOver && !PICT_FORMAT_A(src->format))) &&
+	    (dst->format == src->format || dst->format == alphaless(src->format)) &&
+	    sna_transform_is_integer_translation(src->transform, &tx, &ty) &&
+	    region->extents.x1 + src_x + tx >= 0 &&
+	    region->extents.y1 + src_y + ty >= 0 &&
+	    region->extents.x2 + src_x + tx <= src->pDrawable->width &&
+	    region->extents.x2 + src_y + ty <= src->pDrawable->height) {
+		PixmapPtr dst_pixmap = get_drawable_pixmap(dst->pDrawable);
+		PixmapPtr src_pixmap = get_drawable_pixmap(src->pDrawable);
+		int nbox = RegionNumRects(region);
+		BoxPtr box = RegionRects(region);
+
+		src_x += tx; src_y += ty;
+
+		get_drawable_deltas(src->pDrawable, src_pixmap, &tx, &ty);
+		src_x += tx; src_y += ty;
+
+		get_drawable_deltas(dst->pDrawable, dst_pixmap, &tx, &ty);
+		dst_x += tx; dst_y += ty;
+
+		do {
+			memcpy_blt(src_pixmap->devPrivate.ptr,
+				   dst_pixmap->devPrivate.ptr,
+				   dst_pixmap->drawable.bitsPerPixel,
+				   src_pixmap->devKind,
+				   dst_pixmap->devKind,
+				   box->x1 + src_x, box->y1 + src_y,
+				   box->x1 + dst_x, box->y1 + dst_y,
+				   box->x2 - box->x1, box->y2 - box->y1);
+			box++;
+		} while (--nbox);
+
+		return;
+	}
+
 	src_image = image_from_pict(src, FALSE, &src_xoff, &src_yoff);
 	mask_image = image_from_pict(mask, FALSE, &msk_xoff, &msk_yoff);
 	dest_image = image_from_pict(dst, TRUE, &dst_xoff, &dst_yoff);
diff --git a/src/sna/sna_render.h b/src/sna/sna_render.h
index aace869e..26363945 100644
--- a/src/sna/sna_render.h
+++ b/src/sna/sna_render.h
@@ -775,4 +775,11 @@ static inline bool sna_vertex_wait__locked(struct sna_render *r)
 	return was_active;
 }
 
+#define alphaless(format) PICT_FORMAT(PICT_FORMAT_BPP(format),		\
+				      PICT_FORMAT_TYPE(format),		\
+				      0,				\
+				      PICT_FORMAT_R(format),		\
+				      PICT_FORMAT_G(format),		\
+				      PICT_FORMAT_B(format))
+
 #endif /* SNA_RENDER_H */