Compiler Optimization

[Luc The Perverse]

Hello. This is more a question of curiousity than actually needing a
performance increase.

I was wondering if the Java compiler (I am using Sun's j2sdk 1.4.2_10) will
change Math.PI * 2.0 into a constant in bytecode, or should I make a final
double above my function call?

 

[Stefan Ram]

I was wondering if the Java compiler (I am using Sun's j2sdk 1.4.2_10) will
change Math.PI * 2.0 into a constant in bytecode, or should I make a final
double above my function call?

You can use

javap -c <classname>

to see the bytecode.

 

[NullBock]

Equations solely consisting of compile-time constants are typically
solved before being compiled to byte code. This is also true of
strings, so :

String s = "I want " + ((23 + 7) / 3) + " big macs";

would be stored in byte code as "I want 10 big macs".

Walter Gildersleeve
Freiburg, Germany

______________________________________________________
http://linkfrog.net
URL Shortening
Free and easy, small and green.

 

[Luc The Perverse]

Equations solely consisting of compile-time constants are typically
solved before being compiled to byte code. This is also true of
strings, so :

String s = "I want " + ((23 + 7) / 3) + " big macs";

would be stored in byte code as "I want 10 big macs".

Oh that is even smarter than I thought

Alright cool thanks.

 

[Thomas Hawtin]

Hello. This is more a question of curiousity than actually needing a
performance increase.

I was wondering if the Java compiler (I am using Sun's j2sdk 1.4.2_10) will
change Math.PI * 2.0 into a constant in bytecode, or should I make a final
double above my function call?

Should the value of PI change, then you will need to recompile your code.

Looking at the bytecode is not a particularly good way of determining
performance. The real optimisations happen at runtime, once your code
has been run 10,000 times. So when performance counts, benchmark your
actual code in a close to real life situation.

Tom Hawtin