| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
| |
- Renamed LICENSE.LGPL to LICENSE.LGPLv21
- Added LICENSE.LGPLv3
- Removed LICENSE.GPL
Change-Id: Iec3406e3eb3f133be549092015cefe33d259a3f2
Reviewed-by: Iikka Eklund <iikka.eklund@digia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In the list of default implementations, the raster ops added in
ae0ddb8c729c105a5b4f32a4f6765af8fe546333 were all offset by one
composition mode because of a duplicate entry in the array. The
effect would be, e.g. that using the NotDestination operator would
resolve to the Set operator instead.
Most users will probably not have experienced this since any of
the asm-based functions will be preferred.
[ChangeLog][Painting] Fixed some very rare cases of mismatched
raster modes in QPainter.
Change-Id: Ia242b54c78acbe1c89d9b4ecd10936564ec134b2
Task-number: QTBUG-41413
Reviewed-by: Andy Shaw <andy.shaw@digia.com>
|
|
|
|
|
|
|
|
| |
This code hasn't been tested for at least 4 years. It's not maintained
and probably doesn't work.
Change-Id: I4b9a5179e34111b400914f91caa6b741b69771bb
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Adds basic support for 10-bit per color channel formats to QImage
and the XCB plugin. This will make it possible to paint to and from
these formats, but only at 8-bit per color channel accuracy.
This also fixes Qt5 applications on X11 with native 30bit depth.
[ChangeLog][QtGui][QImage] Added support for 10-bit per color channel image formats.
Task-number: QTBUG-25998
Change-Id: I93ccd3c74bfbb0bd94b352476e5fe58a94119e1f
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
|
|\
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
mkspecs/qnx-x86-qcc/qplatformdefs.h
src/corelib/global/qglobal.h
src/network/socket/qnativesocketengine_winrt.cpp
src/plugins/platforms/android/androidjniaccessibility.cpp
src/plugins/platforms/windows/qwindowswindow.cpp
Manually adjusted:
mkspecs/qnx-armle-v7-qcc/qplatformdefs.h
to include 9ce697f2d54be6d94381c72af28dda79cbc027d4
Thanks goes to Sergio for the qnx mkspecs adjustments.
Change-Id: I53b1fd6bc5bc884e5ee2c2b84975f58171a1cb8e
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
qt_alphamapblit_rgba8888 is only used on little-endian systems, where
qAlpha() returns the correct value.
qdrawhelper.cpp:6256:13: error: 'void qt_alphamapblit_rgba8888(QRasterBuffer*, int, int, quint32, const uchar*, int, int, int, const QClipData*)' defined but not used [-Werror=unused-function]
Change-Id: Ibba6dd6914138f7ae5d53a8e354597f5fff65433
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@digia.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
A vectorized codepath for rotating transforms were added in 5.3.0, but
was only properly tested for SSE2. The NEON version remains unfinished.
Since it was never working, this patch reverts the NEON version.
Task-number: QTBUG-39445
Change-Id: Ifbce0e03781d217ad976c6b18ac88381055cba66
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
|
|/
|
|
|
|
|
|
|
|
|
|
|
| |
Add detection of MIPS DSPr2 at run-time in qsimd.cpp. This makes it
possible to have generic Qt builds for MIPS that can enable the fast
code paths for processors with the DSP ASE at run-time. Also, this
makes it possible to manually disable them by setting the environment
variable "QT_NO_CPU_FEATURE=dspr2". Last, but not least, functions
requiring DSPr2 are not enabled when running in CPUs with version-1
DSP.
Change-Id: Ia5a01d84119553c22ab83386c74a6cb8ba5fee53
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch reuses the code and pattern used to vectorize scaling, to
also vectorize rotating transforms in fetchTransformedBilinearARGB32PM.
This provides significant improvements in QtWebKit on benchmarks using
rotating transforms.
Change-Id: If250e0f5dae1ff0f954301f96cc8970a99ae3e9d
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The calculation of the width/height required for the
scaling algorithm was prone to floating point rounding
issues, where the lower value got rounded down, the higher
one rounded up. This could lead to a situation where we
iterated over one more line/pixel in the line than we have
in the source image.
Correct this by passing the dimension of the source image into
the function and bounds checking the values before iterating.
Task-number: QTBUG-35927
Change-Id: If44b2235a479224660d508a0504fec40d724763a
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
|
|\
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
src/gui/painting/qdrawhelper.cpp
src/gui/painting/qdrawhelper_avx.cpp
src/gui/painting/qdrawhelper_x86_p.h
Change-Id: I09352d05119f6fd000319074a2705b38361b2c90
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
drawCachedGlyphs draws with the wrong color on RGBA8888. The issue
is that the draw routines bitmapblit_quint32 and alphamapblit_quint32
while safe to use on rgba formats, needs to have the input color
converted.
This patch adds small wrapper functions for bitmapblit and alphamapblit
that converts the formats. Two tests are extended to ensure we have
test coverage.
Change-Id: I5f99f3795eba46a69d4df5b167e6099024e9a060
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This patch adds a new config option to qmake to enable full optimization
where it makes sense. This currently is supported on all gcc like
compilers by exchanging -O2 for -O3.
In qtbase it is used to enable full optimizations on qtcore and qtgui
and in a later patch can be used to replace similar existing logic in
QtWebKit's WTF and JavaScriptCore modules.
This fixes a performance regression from gcc 4.7 to 4.8 in the software
renderer.
An aliasing error in qregion.cpp which was exposed by more aggresive
optimization has been solved as well.
Change-Id: Ic2c6c41b79cb3846212b40e7bcc11ff492beb27f
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The drawing code currently only optimizes the bilinear interpolation
under specific conditions that allows the optimizations used there.
The patch adds a SSE2 version of the fallback 4 pixel interpolation.
Change-Id: I4e8a2ba6cb44647105a9b24e38b3ab755a435050
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
|
|\|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
src/gui/image/qimage.cpp
src/gui/text/qtextengine.cpp
src/plugins/platforms/linuxfb/qlinuxfbscreen.cpp
src/printsupport/kernel/qprintengine_win.cpp
Change-Id: I09ce991a57f39bc7b1ad6978d0e0d858df0cd444
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The RGBA8888 formats was incorrectly using the qt_gradient_quint32
which is argb specific. This caused vertical gradients but only
vertical gradients to be drawn incorrectly.
This changes the RGBA8888 formats formats to use the generic gradient
method and renames qt_gradient_quint32 to qt_gradient_argb32 to
indicate its limitation.
Change-Id: Ia1cd48ca7f4f78b64f31d6263e81cd8ac3b0954e
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This patch optimizes the unpremultiply method further by using a
lookup table to avoid any divisions at all.
The opportunity is taken to export both premultiply and unpremultiply
since they are commonly used methods relevant to the exported QRgb type
that can be both premultiplied and unpremultipled ARGB.
[ChangeLog][QtGui][QColor] Exported highly optimized methods for
premultiply and unpremultiply of QRgb values.
Change-Id: I658bcf57b0bc73c34c1765b64617d43b63ae820b
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When drawing to and from the less common formats most of the cpu time
is spend in conversion. The conversion method is rather slow due to
using variable shifts and masks that the compiler does not have a chance
to optimize.
This patch changes the conversion methods to being templates fed by
constexpr methods. This allows the compiler to fully optimize the methods
yielding 2x->5x speedups.
The reliance on constexpr however means the optimized methods are only
used under C++11.
Change-Id: I2ec77c4c1c03f12ee463a694a2b59db0f0b52db1
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
px16 and py16 might be <0 in the first iteration of the loop. In order
to avoid accessing an invalid index the values have to be corrected
before px and py are accessed.
Change-Id: I790faeacc2b77ae8af52bf5e69a2d801c9fe973d
Reviewed-by: Andrew Knight <andrew.knight@digia.com>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@digia.com>
|
|\|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
src/corelib/global/qglobal.h
src/corelib/tools/qstring.cpp
src/gui/image/image.pri
src/gui/image/qimage.cpp
src/plugins/platforms/cocoa/qcocoawindow.h
src/plugins/platforms/cocoa/qcocoawindow.mm
src/plugins/platforms/eglfs/qeglfshooks_stub.cpp
tests/auto/corelib/io/qstandardpaths/tst_qstandardpaths.cpp
Change-Id: I3b9ba029c8f2263b011f204fdf68c3231c6d4ce5
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
storePixel<QPixelLayout::BPP32> would have been called by
storePixels<QPixelLayout::BPP32> but that function has an overload that
uses memcpy instead. So it's really unused.
Found by Clang 3.4.
Change-Id: I661804e273486d3469221fdd362a24cd77cfb8a7
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Changed MIPS DSP portion of the mkspecs/features/simd.prf file in order
to fix the corrupted build system for MIPS platforms.
List of the additionally optimized functions
from file src/gui/painting/qdrawhelper.cpp:
- qt_blend_rgb16_on_rgb16
- qt_fetchUntransformed_888
- qt_fetchUntransformed_444
- qt_fetchUntransformed_argb8565
from file src/gui/image/qimage.cpp:
- convert_ARGB_to_ARGB_PM_inplace
from file src/corelib/qstring.cpp:
- ucstrncmp
- toLatin1_helper
- fromLatin1_helper
Change-Id: I5c47a69784917eee29a8dbd2718828a390b27c93
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Currently the non-standard opaque formats all perform inverse premultiply
on semi transparent pixels when saving. This is only performed on RGB16
and RGB32 when explicitly converting to the formats, but not when drawing.
This patch changes the conversion to be consistent across the board so
less common formats are treated the same as the most common ones. This
also makes drawing and converting to the these formats much faster.
Task-number: QTBUG-36143
Change-Id: I877ddb1c1dbb4d2dd9b4b9192525e50375a1ae60
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Profiling indicates that this function is one of two hot spots causing
a noticeable latency when changing KDE virtual desktops.
Instead of computing two modulos per pixel in the inner loop, it is
possible to compute the modulos outside the loop and compute a modulo
sum in the inner loop for a reasonable speedup.
Change-Id: Ic4217b7686e031d7673b3e10aa977dae263096dc
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
On iOS we don't build the NEON drawhelpers as they are implemented using
GAS syntax assembly, so prevent drawhelpers.cpp from trying to use them.
Follow-up to 01c59ac857927043e49d4725062faea715ed49a3.
Change-Id: I3e3dc9b0e1d780db3184413d2ff3010ac8dcd37b
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Our inverse premultiply is rather unoptimized. It's major weakness is
using three divide operation which are the slowest integer operations.
This patch reduces the three divisions to just one but using an
approximation that is accurate for all uchar values.
The patch also adds a general short-cut for alpha==255.
Together these improvements makes it 2 to 16 times faster depending on
how many 0 and 255 alpha values are encountered.
Change-Id: I96d7098a1bc320793b0d0526637acd1fdb5a43eb
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
These two functions used to cause a runtime detection of the CPU
features in order to improve their performance. Since the last two
commits, there's no runtime detection of either SSE2 or Neon support,
so there's no point in attempting runtime detection.
Task-number: QTBUG-30440
Change-Id: I54fe92787c983003c2cc867ee636daec30063033
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Now the only way to enable Neon support is to change the mkspec.
[ChangeLog][Important Behavior Changes] Qt no longer checks for
support for the Neon FPU on ARM platforms at runtime. Code optimized
for Neon must be enabled unconditionally at compile time by ensuring
the compiler supports Neon. You may need to edit your mkspec for that.
Task-number: QTBUG-30440
Change-Id: I4df9b2bf3cd022f8ed70f02f16878cb2cb3fe6fb
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
If the compiler supports SSE2, we'll use our SSE2-optimised code
unconditionally. Runtime detection is left for SSSE3 code.
The SSE2 codebase is big and thus a timebomb if an inline function
gets leaked out and run without runtime check. In reality, it's
extremely unlikely people running CPUs without SSE2 support are
running Qt 5 at this moment (they're either too old or too new,
e.g. Intel Quark).
The SSSE3 codebase is a lot more manageable.
Task-number: QTBUG-30440
Change-Id: I3e586e4434e820365d5316b650ee3061d0acf767
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This feature was never completely implemented. Sure, it's nice to
build the SSE2 + SSSE3 code with the VEX prefix, which results in
better code. But the leap isn't that big anyway.
This is the first step to removing the runtime detection for the
drawhelpers. They create timebombs when we use inline functions.
Task-number: QTBUG-30440
Change-Id: Ic53b2cf5261106a1c940d4a36eb6111b7d998be1
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
|
|/
|
|
|
|
|
|
|
|
|
| |
After profiling drawing with RGBA8888 images most of the time appears
to be spend in converting to and from ARGB32PM.
This patch adds four small converters that are 3-4 times faster than the
generic converter.
Change-Id: I3c7498756f440ca3ea9c1417b26dd8e1953b9d06
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
|
|
|
|
|
|
|
|
|
| |
The fill color was not correctly converted before being filled into
RGBA8888 images. This patch adds a function with convertion and
adds tests for it to tst_qpainter.
Change-Id: If8b0e6db38b2794a60301842e25f377eb7216796
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
|
|
|
|
|
|
|
|
|
| |
Support the byte-ordered RGBA format which is used by OpenGL, and many
endian neutral byte formats.
Task-number: QTBUG-32201
Change-Id: I77cffb4c30c69545fa96ded2f537b2ebd9351acb
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@digia.com>
|
|\
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
src/plugins/platforms/cocoa/qcocoafiledialoghelper.mm
src/plugins/platforms/qnx/qqnxrasterbackingstore.cpp
tools/configure/configureapp.cpp
Change-Id: I3092bd3276af14304b7ab3ae1e1cc05d11cdede0
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
It is deprecated and clang is starting to warn about it.
Patch mostly generated by clang itself, with some careful grep
and sed for the platform-specific parts.
Change-Id: I8058e6db0f1b41b33a9e8f17a712739159982450
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|\|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
src/corelib/io/qdatastream.cpp
src/corelib/io/qdatastream.h
src/corelib/json/qjsonwriter.cpp
src/plugins/platforms/cocoa/qcocoawindow.mm
src/plugins/platforms/xcb/qxcbkeyboard.cpp
Change-Id: I46fef1455f5a9f2ce1ec394a3c65881093c51b62
|
| |
| |
| |
| |
| | |
Change-Id: Ic87be3b6c7afbde869bd011a3a585b0fe7dbf15d
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
|
|\|
| |
| |
| | |
Change-Id: Icff019d74ae04c628a80f66aa478e4db40fae464
|
| |\
| | |
| | |
| | | |
Change-Id: Icedabda08961326a0a447ec71f1b0f0f5df075eb
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
O3 leads to gcc bailing out with OOM on e.g. MinGW with gcc
4.7.3 onwards.
Task-number: QTBUG-29099
Change-Id: I3ce49794fa4857e756d2994454d4144cfb44ce58
Reviewed-by: Jonathan Liu <net147@gmail.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|/ /
| |
| |
| |
| |
| |
| |
| |
| |
| | |
We used to disable NEON completely, as the iOS toolchain does not
handle the GAS syntax of the pixman draw-helpers. But we can limit
the disabling to just the draw-helpers, which means we get NEON
optimization of eg. QImage and QString.
Change-Id: If350b06ce521cca8b24468be5a168ff21e9e7124
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@digia.com>
|
|\|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
src/corelib/io/qsavefile_p.h
src/corelib/tools/qregularexpression.cpp
src/gui/util/qvalidator.cpp
src/gui/util/qvalidator.h
Change-Id: I58fdf0358bd86e2fad5d9ad0556f3d3f1f535825
|
| |
| |
| |
| |
| | |
Change-Id: Ic804938fc352291d011800d21e549c10acac66fb
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
|
|/
|
|
|
|
|
|
| |
There is no reason those function shouldn't be static. This produces
no change in code, except that the functions may be better optimised.
Change-Id: Ia60b81d4e29a36190a453e8c7c8102c8f547e1a9
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
|
|
|
|
|
|
|
|
|
| |
Although not widely used, the raster operations from Qt 3 were useful
and some of them were already implemented, this brings the rest of them
back for those who need them.
Change-Id: Id538611eaaba9be3d39bf2dd33b6c532f5d4aba8
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
These functions have begun showing in the tst_symbols unit test as
"symbol does not start with q". Since they're never called directly,
they're never inlined. Since they were inline, the compiler was probably
deciding to not export them.
Something changed and it could be anything (new compiler version, new
options, etc.). So mark them static.
Change-Id: I838dfc94edd7f09c202743bff0daf9d20c10c3a6
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
|
|
|
|
|
|
|
|
| |
Change copyrights and license headers from Nokia to Digia
Change-Id: If1cc974286d29fd01ec6c19dd4719a67f4c3f00e
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Reviewed-by: Sergio Ahumada <sergio.ahumada@digia.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously QImage::fill() for Format_RGB888 expected a BGR value
instead of the RGB order defined by QRgb, making it counter intuitive to
use related to the 32-bit formats.
Fixed the QPixelLayout data for RGB888 and changed the byte order of
quint24 based on what the optimized image conversion routines expect.
Change-Id: I72926debbc6f5b5cb10b8aa0b2a2a916a04db946
Reviewed-by: Kim M. Kalland <kim.kalland@nokia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
List of optimized routines:
- comp_func_DestinationOver
- comp_func_SourceIn
- comp_func_DestinationIn
- comp_func_DestinationOut
- comp_func_SourceAtop
- comp_func_DestinationAtop
- comp_func_XOR
- comp_func_SourceOut
- comp_func_solid_SourceOver
- comp_func_solid_DestinationOver
- comp_func_solid_SourceIn
- comp_func_solid_DestinationIn
- comp_func_solid_SourceAtop
- comp_func_solid_DestinationAtop
- copm_func_solid_XOR
- comp_func_solid_SourceOut
Previously optimized routines qt_blend_argb32_on_argb32_mips_dsp and
comp_func_Source_mips_dsp are redesigned and rewritten.
Overall improvement by running tst_bench_blendbench benchmark app
from tests/benchmarks/gui/image/blendbench/ is 27%.
Change-Id: I6ab09b17cac10f4aded59787074ab4c89e72ccac
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
First, check that the option in question hasn't been already enabled
by the compiler, via compiler switches. If it has been, then we don't
need to verify anything, and we can assume that it's safe to use such
instructions. For example, on an x86-64 build, qCpuHasFeature(SSE2) is
always a constant true.
If the compile-time check fails, then we proceed to try and detect the
processor features at runtime. But instead of insisting on a call to
qDetectCPUFeatures, allow the code using the detection to read from a
variable and simply test it for values. Only if the variable isn't
initialised should it make a function call. The Q_ASSUME allows this
code to be very efficient even with multiple uses of qCpuHasFeature.
Change the uninitialised value from -1 to 0 so that simpler
instructions can be used to check for non-initialisation.
The qDetectCPUFeatures function is renamed to qDetectCpuFeatures to
match the Qt coding style and also to catch uses this code that need
to be adapted.
Change-Id: I24ca5a6ad21075e2e249e1a4f8f5057b8f68ce7c
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|