Optimization settings for x86 and armeabi-v7a versions
-
My application written with qt 5.11.1 on Android runs twice as fast on my tablet's atom cpu processor z3745 when the code is compiled in x86 than when it is compiled in armeabi-v7a...
In the application's pro file, I use the same optimization parameters in both cases:
QMAKE_CXXFLAGS_RELEASE +="-fsigned-char -funsigned-bitfields -O3 -DNDEBUG"
QMAKE_CFLAGS_RELEASE += "-fsigned-char -funsigned-bitfields -O3 -DNDEBUG"
QMAKE_LFLAGS_RELEASE += -O1How to explain such a speed difference between the x86 and armeabi-v7a versions ? nd what are the good optimization parameters for the armeabi-v7a version?
Thank you!
-
My application written with qt 5.11.1 on Android runs twice as fast on my tablet's atom cpu processor z3745 when the code is compiled in x86 than when it is compiled in armeabi-v7a...
In the application's pro file, I use the same optimization parameters in both cases:
QMAKE_CXXFLAGS_RELEASE +="-fsigned-char -funsigned-bitfields -O3 -DNDEBUG"
QMAKE_CFLAGS_RELEASE += "-fsigned-char -funsigned-bitfields -O3 -DNDEBUG"
QMAKE_LFLAGS_RELEASE += -O1How to explain such a speed difference between the x86 and armeabi-v7a versions ? nd what are the good optimization parameters for the armeabi-v7a version?
Thank you!
Hi @Francky033,
How to explain such a speed difference between the x86 and armeabi-v7a versions ?
Without data, nobody can explain. You need to find out the slow parts in your app and then you can start to optimize it.
QtCreator supports profiling. This should give you an idea about the slowest parts of your program.
Regards
-
Hi,
Out of curiosity, are you running your application on processors with similar specs ?
-
I have an ASUS tablet that works very well with this processor. The display is particularly fast. The tablet is 4 years old...
-
My application written with qt 5.11.1 on Android runs twice as fast on my tablet's atom cpu processor z3745 when the code is compiled in x86 than when it is compiled in armeabi-v7a...
In the application's pro file, I use the same optimization parameters in both cases:
QMAKE_CXXFLAGS_RELEASE +="-fsigned-char -funsigned-bitfields -O3 -DNDEBUG"
QMAKE_CFLAGS_RELEASE += "-fsigned-char -funsigned-bitfields -O3 -DNDEBUG"
QMAKE_LFLAGS_RELEASE += -O1How to explain such a speed difference between the x86 and armeabi-v7a versions ? nd what are the good optimization parameters for the armeabi-v7a version?
Thank you!
@Francky033 said in Optimization settings for x86 and armeabi-v7a versions:
My application written with qt 5.11.1 on Android runs twice as fast on my tablet's atom cpu processor z3745 when the code is compiled in x86 than when it is compiled in armeabi-v7a...
Just to be sure, because I may be missing something: are you saying that you can run your application compiled for X86 and then compiled for armeabi-v7a in the exact same tablet?
-
@Francky033 said in Optimization settings for x86 and armeabi-v7a versions:
My application written with qt 5.11.1 on Android runs twice as fast on my tablet's atom cpu processor z3745 when the code is compiled in x86 than when it is compiled in armeabi-v7a...
Just to be sure, because I may be missing something: are you saying that you can run your application compiled for X86 and then compiled for armeabi-v7a in the exact same tablet?
@Pablo-J.-Rogina yes... I'm running the two compilations on the same tablet(the CPU is compatible with the armv7 and the x86 binairies).. and I can see a speed difference on the exact same tablet... The x86 compiled version is faster...
-
@Pablo-J.-Rogina yes... I'm running the two compilations on the same tablet(the CPU is compatible with the armv7 and the x86 binairies).. and I can see a speed difference on the exact same tablet... The x86 compiled version is faster...
@Francky033 said in Optimization settings for x86 and armeabi-v7a versions:
same tablet(the CPU is compatible with the armv7 and the x86 binairies
Ok, I didn't know about that capability for the processor.
In that case and without any further data, I assume the Atom CPU is faster with x86 binaries because it runs native code for its architecture while when running ARM binaries some sort of emulation layer should be employed and thus slowing things down.
-
It's a possibility.... Indeed...
-
@Pablo-J.-Rogina yes... I'm running the two compilations on the same tablet(the CPU is compatible with the armv7 and the x86 binairies).. and I can see a speed difference on the exact same tablet... The x86 compiled version is faster...
@Francky033 said in Optimization settings for x86 and armeabi-v7a versions:
the CPU is compatible with the armv7 and the x86 binairies
Never heard about such CPU. Most probably it executes ARM binaries using an emulation and that's why ARM binaries are slower.
