summaryrefslogtreecommitdiff
path: root/vcl/inc/skia/gdiimpl.hxx
diff options
context:
space:
mode:
authorLuboš Luňák <l.lunak@collabora.com>2022-01-06 18:33:04 +0100
committerLuboš Luňák <l.lunak@collabora.com>2022-01-12 13:27:37 +0100
commit9cb34aa3c98ac6254b7e70e26e92f3e2ba881fab (patch)
treef439a449d8b59a22def72548b62e0aa75e50f830 /vcl/inc/skia/gdiimpl.hxx
parentd39acd432df374ffdf4142c23d1c74df8821160c (diff)
reduce explicit Skia flushing
The docs say that Skia handles this itself, and with Vulkan excessive flushing may show up in profiling in some cases, as it's apparently a non-trivial operation in GPU mode. Remove even the two workarounds, I cannot reproduce the problems anymore, so let's assume it's been fixed in Skia. But still keep the flushing after a certain number of operations, as too many pending operations still may overload Skia (or Vulkan?), besides tdf#136369 I can also reproduce it while loading bsc#1183308 which does a large number of tiny VirtualDevice's and does GetBitmap() on them. Also change the counter to be global, as we use just one Skia drawing context. Change-Id: I48b96c2a59f8e1eeef3da154dbe373a37857c4be Reviewed-on: https://gerrit.libreoffice.org/c/core/+/128293 Tested-by: Jenkins Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
Diffstat (limited to 'vcl/inc/skia/gdiimpl.hxx')
-rw-r--r--vcl/inc/skia/gdiimpl.hxx2
1 files changed, 1 insertions, 1 deletions
diff --git a/vcl/inc/skia/gdiimpl.hxx b/vcl/inc/skia/gdiimpl.hxx
index b5144a249207..b2814b2a6463 100644
--- a/vcl/inc/skia/gdiimpl.hxx
+++ b/vcl/inc/skia/gdiimpl.hxx
@@ -368,7 +368,7 @@ protected:
double transparency;
};
LastPolyPolygonInfo mLastPolyPolygonInfo;
- int mPendingOperationsToFlush;
+ inline static int pendingOperationsToFlush = 0;
int mScaling; // The scale factor for HiDPI screens.
};