Implement SkPicture::bytesUsed() for SkRecord backend

src/core/SkTileGrid.cpp

Index: src/core/SkTileGrid.cpp
diff --git a/src/core/SkTileGrid.cpp b/src/core/SkTileGrid.cpp
index e285cccd0bff506a67614b4a3305a4fe1ffb1170..e023e3539b04d2b0fa72468cabcf44fd494af529 100644
--- a/src/core/SkTileGrid.cpp
+++ b/src/core/SkTileGrid.cpp
@@ -175,3 +175,12 @@ void SkTileGrid::search(const SkRect& originalQuery, SkTDArray<unsigned>* result
     }
 }
 
+size_t SkTileGrid::bytesUsed() const {
+    size_t byteCount = sizeof(SkTileGrid);
+
+    for (int i = 0; i < fXTiles * fYTiles; i++) {

[chrishtr, 2014/11/17 21:34:21]

Will this be slow? Overall, what asymptotic runtime should we expect for the
public bytesUsed() method?

[tomhudson, 2014/11/17 22:12:15]

If the tile grid just spans a screen on the Nexus 9, it'd be 8x6, so ~50 adds
and multiplies, but there should be a buffer above and below the screen. If
that's 2kpx each way we're looking at 24x6 -> 150 adds and multiplies?

If it's expensive in practice, I suspect it'll be much more reasonable to cache
than to try to track as we go, but either ought to be feasible.

+        byteCount = fTiles[i].reserved() * sizeof(unsigned);

[mtklein, 2014/11/17 23:15:33]

I think you want += here?

While you're at it, can't hurt to move the multiply out of the loop.

+    }
+
+    return byteCount;
+}