| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
| |
The bugs we had noticed previously are believed to be fixed. MSVC will
do per-function updating of AVX content where necessary and the ICC
issue is no longer relevant.
Change-Id: I2bbf422288924c198645fffd16a9235f2d73cc19
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Use the same approach we use for iOS, which is to set multiple
CMAKE_OSX_ARCHITECTURES values and let the clang front end
deal with lipo-ing the final libraries.
For now, Qt can be configured to build universal macOS libraries by
passing 2 architectures to CMake, either via:
-DCMAKE_OSX_ARCHITECTURES="x86_64;arm64"
or
-DCMAKE_OSX_ARCHITECTURES="arm64;x86_64"
Currently we recommend specifying the intel x86_64 arch as the first
one, to get an intel slice configuration that is comparable to a
non-universal intel build.
Specifying the arm64 slice first could pessimize optimizations and
reduce the feature set for the intel slice due to the limitation
that we run configure tests only once.
The first specified architecture is the one used to do all the
configure tests.
It 'mostly' defines the common feature set of both architecture
slices, with the excepion of some special handling for sse2 and
neon instructions.
In the future we might want to run at least the Qt architecture config
test for all specified architectures, so that we can extract all the
supported sub-arches and instruction sets in a reliable way.
For now, we use the same sse2 hack as for iOS simulator_and_device
builds, otherwise QtGui fails to link due to missing
qt_memfill32_sse2 and other symbols.
The hack is somewhat augmented to ensure that reconfiguration
still succeeds (same issue happened with iOS). Previously the sse2
feature condition was broken due to force setting the feature
to be ON. Now the condition also checks for a special
QT_FORCE_FEATURE_sse2 variable which we set internally.
Note that we shouldn't build for arm64e, because the binaries
get killed when running on AS with the following message:
kernel: exec_mach_imgact: not running binary built against
preview arm64e ABI.
Aslo, by default, we disable the arm64 slice for qt sql plugins,
mostly because the CI provisioned sql libraries that we depend on only
contain x86_64 slices, and trying to build the sql plugins for both
slices will fail with linker errors.
This behavior can be disabled for all targets marked by
qt_internal_force_macos_intel_arch, by setting the
QT_FORCE_MACOS_ALL_ARCHES CMake option to ON.
To disble it per-target one can set
QT_FORCE_MACOS_ALL_ARCHES_${target} to ON.
Task-number: QTBUG-85447
Change-Id: Iccb5dfcc1a21a8a8292bd3817df0ea46c3445f75
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
| |
The first replacement had missed objective-C++ code some places ourside
the src dir.
In C-files Q_DECL_NOTHROW is replaced with Q_DECL_NOEXCEPT as we still
need to turn it off when compiled in C mode, but can get rid of the old
NOTHROW moniker.
Change-Id: I6370f57066679c5120d0265a69e7e378e09d4759
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|
|
|
|
|
|
| |
Needed by gcc 4.8
Change-Id: I2daa5728761599255cf3912d37e7b9dd60ccb60c
Reviewed-by: Eirik Aavitsland <eirik.aavitsland@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Conflicts:
.qmake.conf
sc/corelib/io/qfsfileengine_p.h
src/corelib/io/qstorageinfo_unix.cpp
src/platformsupport/eglconvenience/qeglpbuffer_p.h
src/platformsupport/input/libinput/qlibinputkeyboard.cpp
src/platformsupport/input/libinput/qlibinputpointer.cpp
src/plugins/platforms/cocoa/qcocoamenu.mm
src/plugins/platforms/ios/qiosscreen.h
src/plugins/platforms/ios/qioswindow.h
src/plugins/platforms/ios/quiview.mm
src/printsupport/dialogs/qpagesetupdialog_unix_p.h
src/printsupport/dialogs/qprintpreviewdialog.cpp
src/printsupport/widgets/qcupsjobwidget_p.h
src/widgets/widgets/qmenu.cpp
tests/auto/corelib/tools/qdatetime/tst_qdatetime.cpp
tests/auto/widgets/itemviews/qtreeview/tst_qtreeview.cpp
Change-Id: Iecb4883122efe97ef0ed850271e6c51bab568e9c
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This is going to be the most common scenario for GCC: all recent
versions allow compiling F16C code in the same source, thus generating
better code.
MSVC is excluded unless AVX is already turned on by the user in the
mkspec file, because it fails to use the VEX-prefixed instructions for
everything else, printing a warning that it knows it should have done
so.
ICC is excluded because it's known to generate invalid code when using
the F16C intrinsics unless F16C is turned explicitly on.
Clang is excluded because it runs into an internal error compiling this
code unless F16C is turned explicitly on.
Change-Id: I57a1bd6e0c194530b732fffd14f58de6a46f0de1
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
| |
Also we can't use constructors for multiple reasons now that it is C.
Change-Id: I27f3e011cc1f67f5aa134eaf3ab934456cead902
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|
Both ARM and x86 can convert fp16 much faster in bulk than one at a
time. This also enables hardware accelerated conversion on x86, when
F16C isn't unconditionally available at compile time.
This code is implemented in C to ensure that there's no leakage of
inline symbols from the .obj file that was compiled by Visual Studio
with AVX support. Unfortunately, simd.prf uses $(CXX) instead of $(CC)
for all its sources, which means the file gets interpreted as C++ by
g++, clang++ and icpc. Those compilers at least don't leak any symbols.
Done-with: Thiago Macieira <thiago.macieira@intel.com>
Change-Id: I9d26d99e83392861fb09564e0e8e8d76cd8483b3
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|