1
Здравствуйте кто может помочь обойти проверку наличие активации в приложении ?
Size: a a a
2020 October 21
2020 October 23
MG
would it be faster to measure text in JNI than an java?
MG
in*
DG
Matthew Good
would it be faster to measure text in JNI than an java?
is it possible to do so?
1. if you use monospace font and no text wrapping then there's nothing to do actually just integer division
2. if you use monospace but need text wrapping then for transferring string and converting it into char array I guess you'll spent more time than just loop through in java
3. if you need non monospace font - there's nothing to do in cpp, use built-in functions from android
smth like paint.getTextBounds or pain.measureText
1. if you use monospace font and no text wrapping then there's nothing to do actually just integer division
2. if you use monospace but need text wrapping then for transferring string and converting it into char array I guess you'll spent more time than just loop through in java
3. if you need non monospace font - there's nothing to do in cpp, use built-in functions from android
smth like paint.getTextBounds or pain.measureText
MG
is it possible to do so?
1. if you use monospace font and no text wrapping then there's nothing to do actually just integer division
2. if you use monospace but need text wrapping then for transferring string and converting it into char array I guess you'll spent more time than just loop through in java
3. if you need non monospace font - there's nothing to do in cpp, use built-in functions from android
smth like paint.getTextBounds or pain.measureText
1. if you use monospace font and no text wrapping then there's nothing to do actually just integer division
2. if you use monospace but need text wrapping then for transferring string and converting it into char array I guess you'll spent more time than just loop through in java
3. if you need non monospace font - there's nothing to do in cpp, use built-in functions from android
smth like paint.getTextBounds or pain.measureText
hm ok
MG
welp heres an interesting fact:
i out-perform android's
android's line breaking is the last thing I expected to be a performance hit from its text layout system
but instead it is like 96% of the performance hit
With most of it being in
like holy crap 😱
i out-perform android's
TextView by a factor of 10 o.oandroid's line breaking is the last thing I expected to be a performance hit from its text layout system
but instead it is like 96% of the performance hit
With most of it being in
android.text.StaticLayout.generatemeanwhile i take ~220ms to do line breaking
android.graphics.text.LineBreaker.computeLineBreaks
android.graphics.text.LineBreaker.nComputeLineBreaks
2.29s
like holy crap 😱
DG
you're doing for monospace right?
MG
no
MG
im just using the default typeface
MG
that is initalized in the paint
MG
even with the same text size for both (
setTextSize(30.0f);)MG
TextView still takes 2.3 seconds
MG
the text paint i cant really change in TextView
mRes = getResources();
mTextBookPaint = new TextPaint(Paint.ANTI_ALIAS_FLAG);
mTextBookPaint.density = mRes.getDisplayMetrics().density;
mBackgroundPaint = new Paint();
mBackgroundPaint.setColor(Color.WHITE);
mBackgroundPaint.setStyle(Paint.Style.FILL);
MG
even with
textBookView.setBreakStrategy(LineBreaker.BREAK_STRATEGY_SIMPLE); it still takes over 2 secondsMG
so yea i dont know wtf
nComputeLineBreaks is doing that takes it over 2 seconds to completeMG
also would it be faster to like... build my line info structure in CPP instead of Java?
MG
as i need to get this to at least 16ms
MG
to be fast enough fot real-time text layout and measurement (eg with no drawing lag as the user inputs text)
MG
as i managed to get it down from 630 ms average to around 220 ms to 330 ms average
by moving the allocations from the inside of the loop to the outside of the loop, measuring the entire line once, and then using the consumed count as an offset to the measured widths index
instead of allocating and measuring each loop
but i am not sure what else i can do
by moving the allocations from the inside of the loop to the outside of the loop, measuring the entire line once, and then using the consumed count as an offset to the measured widths index
instead of allocating and measuring each loop
but i am not sure what else i can do