| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
- Replace Nokia contact email address with Qt Project website.
- Remove "All rights reserved" line from license headers.
As in the past, to avoid rewriting various autotests that contain
line-number information, an extra blank line has been inserted at the
end of the license text to ensure that this commit does not change the
total number of lines in the license header.
Change-Id: Ie7ba62011752fcb149b99b26317c54f2a0cfa931
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
|
| |
|
|
|
|
|
|
|
| |
Simple search and replace. This commit doesn't touch 3rd-party files,
nor translations (where the change is not so simple and will be handled
in a separate commit).
Change-Id: I4e48513b8078a44a8cd272326685b25338890148
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
When drawing such pixmaps (used by both the openvg and opengl graphics
systems) onto another pixmap or to a QImage, the performance was
sub-optimal due to missing and accidentally disabled support
specific to QVolatileImage. This is now fixed and drawing pixmaps into
a QImage is also made optimal by using the QS60PaintEngine for
QImage too. This will cause a 5-7x (or even up to 12x on certain
hardware and platform) increase in offscreen pixmap drawing
performance.
Task-number: QTBUG-19880
Reviewed-by: Jani Hautakangas
|
| |
|
|
|
|
|
| |
Updated version of LGPL and FDL licenseheaders.
Apply release phase licenseheaders for all source files.
Reviewed-by: Trust Me
|
| |
|
|
|
|
|
|
|
| |
Forcing VG (and GL) pixmapdata to support toNativeType() with type
VolatileImage is unnecessary. It is cleaner to try doing a
fromNativeType() with bitmap+mask and fall back if that results in
a null pixmapdata.
Reviewed-by: Jani Hautakangas
|
| |
|
|
| |
Reviewed-by: Jani Hautakangas
|
| |
|
|
|
|
|
|
|
|
|
| |
Inverting the pixels for bitmaps of mode EGray2 is done in-place,
which is wrong if the bitmap handle was duplicated. Instead, we
need to make a copy. This also means that we cannot treat EGray2
a conversion-less format in openvg, but such images are likely
to be used as masks only, so optimizing just for mere drawing
in case of this format is not really necessary.
Reviewed-by: Jani Hautakangas
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
From now on the image data coming from CFbsBitmap will not be forced
to RGB32 or ARGB32_Premultiplied in openvg. This allows copy-less
creation and drawing of pixmaps not just from bitmaps with display mode
EColor16MAP and EColor16MU, but also a few other modes, like EColor64K.
Painting into such pixmaps or drawing them via the raster engine is
potentially slower in such cases, which is now reflected in the
fromSymbianCFbsBitmap documentation.
Note that this patch has no effect on extended bitmaps (e.g. skin
graphics), those must always be rasterized first regardless of the
display mode.
Task-number: QTBUG-18027
Reviewed-by: Jani Hautakangas
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
QNativeImageHandleProvider is a thin interface consisting of get()
and release() functions. Pixmaps constructed with such a provider
will call these functions to acquire and release a native handle,
e.g. a CFbsBitmap or RSgImage pointer in case of Symbian. The behavior
is largely similar to constructing pixmaps via fromSymbianCFbsBitmap
or fromSymbianRSgImage, with the exception of pixmap hibernation:
release() (and subsequently get()) is guaranteed to be called also
in case of hibernation, allowing more fine-grained tracking of
the usage and lifetime of image data.
Task-number: QT-4632
Reviewed-by: Jani Hautakangas
|
| |
|
|
|
|
|
|
|
|
|
| |
Enable readback of pixel data for pixmaps that are constructed directly
from a VGImage. These do not have any backing data in main memory
(i.e. 'source' is null), however certain operations, like toImage(),
fill(), or painting into the pixmap do not work without it. With this
patch the data is read back via vgGetImageSubData when needed.
Task-number: QT-4669
Reviewed-by: Jani Hautakangas
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Similarly to QS60PixmapData, that is used on raster, QVGPixmapData
is also backed by a CFbsBitmap from now on (at least when it makes
sense to do so). This allows copy-less bitmap handle duplication in
from- and toSymbianCFbsBitmap() in case of certain image formats,
reduces local heap usage and improves performance with the s60 style
due to a reduced number of pixel data copies.
Task-number: QT-2505
Reviewed-by: Jason Barron
Reviewed-by: Jani Hautakangas
|
| |
|
|
|
|
|
|
| |
The retrieved VGImage handle was accidentally assigned to a local
variable instead of the member with the same name so it always
ended up creating an empty pixmap.
Reviewed-by: Jani Hautakangas
|
| |
|
|
| |
Reviewed-by: Trust Me
|
| |
|
|
|
|
|
| |
Functions were accidentally thought to be in private scope.
Renaming broke BC.
Reviewed-by: Jason Barron
|
| |
|
|
|
|
|
|
|
|
| |
Use vgWritePixels, if possible, in case where VGImage
for QPixmap can't be allocated due to low GPU memory
situation. This patch also renames some VG paint engine
internal functions to clarify their purpose.
Task-number: QT-3589
Reviewed-by: Jason Barron
|
| |
|
|
|
|
|
|
|
|
| |
Symbian bitmap formats may have different scanline length
when compared to QImage formats. We need to define scanline
length for intermediate QImage when converting from CFbsBitmap
to QPixmap.
Task-number: QTBUG-14218
Reviewed-by: Jason Barron
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
In the cases where an error occured while converting a QPixmap to a
VGImage this function would return without deleting the RSgImage
pointer that it created. Fix is to use a QScopedPointer instead. Also
don't use q_check_ptr() since this isn't a CBase derived class.
In case you are wondering why I didn't use a custom deleter here so
that Close() was also called, we need to make sure that Close() is
called on the RSgImage instance before calling Close() on the driver.
Reviewed-by: mread
|
| |
|
|
|
|
|
|
|
|
|
| |
Calling the various Khronos getError() functions here is inefficient
since it involves a round trip to the GPU and gains nothing since the
null handle can be used to test for error conditions at no cost. Also
no need to open the RSgDriver because we don't actually need to call
any functions on RSgImage, just convert it.
Reviewed-by: Alessandro Portale
Reviewed-by: Jani Hautakangas
|
| |
|
|
|
|
|
|
|
|
|
|
| |
After the latest round of API reviews, the Symbian glyph cache has now
been moved from GDI to FBS and this required some changes in Qt's
implementation.
Also incorporate an optimization where we first iterate over the glyph
vector to eliminate glyphs that are already cached in Qt's glyph cache.
This way we only open the glyph iterator on glyphs that we need.
Reviewed-by: Alessandro Portale
|
| |
|
|
|
|
|
| |
Recent changes in Qt's EGL layer required some changes in the code used
to convert RSgImage to/from VGImage.
Reviewed-by: Gunnar Sletta
|
| |
|
|
|
|
|
| |
If the hardware glyph cache API is not represent and in use, then the
glyphs will not be inverted.
Reviewed-by: TrustMe
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The OpenVG paint engine traditionally takes glyph images from the
alphaMapForGlyph() function which returns the glyph image in the
upright projection. When it constructs a VGImage from this image, it
passes a positive data stride which will read the top scanline of the
source image into the bottom scanline of the VGImage due to the VG
coordinate system. It then uses the path transform where the 'sy' value
of the matrix is set to -1 and this re-inverts everything when drawing.
With the Symbian based glyph cache, the VGImage is constructed from a
RSgImage which is a hardware resource and compensates for the
coordinate system used by VG and GL at the time it is created. In
the case of the hardware glyph cache, the glyph image is read into the
RSgImage using a negative data stride so it does not need to be
inverted when drawn. To allow for this, introduce a flag which
indicates that the 'sy' entry of the matrix should be flipped such that
the glyph is drawn normally.
Also in this patch is a change to the glyph origin which now uses the
bottom of the glyph metric bounding rect instead of the top due to
orientation of the glyph inside the VGImage.
Reviewed-by: Alessandro Portale
|
| |
|
|
|
|
|
|
|
|
|
| |
This introduces a specialized OpenVG font cache for the Symbian
platform. By using RGlyphDataIterator we can create a VGImage to store
inside a VGFont without having to upload our own glyph image. This
works by utilizing RSgImage which can be used to share handles to
graphics memory across processes thus allowing glyph images to be
shared by multiple processes and reducing graphics memory usage.
Reviewed-by: Alessandro Portale
|
| |
|
|
|
|
|
|
|
|
| |
This code will now be used by both the fromNativeType() function and
the upcoming glyph cache implementation. We also change the #ifdef's
location here slightly because even if we do not have support for
RSgImage, we still want to be able to support the CFbsBitmap conversion
functions for the OpenVG graphics system.
Reviewed-by: Alessandro Portale
|
|
|
Factor out the Symbian specific bits of QVGPixmapData into a separate
file and introduce our subclassed implementation of the VG font cache.
Also make our font cache a friend of the S60 font engine because the
cache needs to access the CFont member to create the glyph image.
Reviewed-by: Alessandro Portale
|
