3 ms·

Any guesses as to why the Java implementation is relatively slow? JNI boundary crossings for floating point arithmetic, or lack of use of the math coprocessor?

by jbooth 17y ago

smanek 17y ago
I don't know, but it's much slower than I expected. It's the latest Java installable by apt on Debian (all these tests are run in Debian running under VMWare). $ java -version java version "1.6.0_0" OpenJDK Runtime Environment (build 1.6.0_0-b11) OpenJDK 64-Bit Server VM (build 1.6.0_0-b11, mixed mode)
- jbooth 17y ago
  Dohhhhh don't use OpenJDK :) Get the sun version, it's probably faster. Although it may not be much faster, found this unfixed bug report: http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=5005861 http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=5005861
  - jbooth 17y ago
    In more depth : the sun JDK probably does a lot better job of JIT compilation, it's the difference between running interpreted bytecode and compiled native code. On the Hadoop project, checksumming was a major CPU constraint, because you have to assume every file everywhere is possibly on faulty hardware, and checksum everything. They found that it was actually faster to implement CRC32 in pure Java and let the JIT compilation do it's magic than it was to make native calls to zlibc. But that may not apply to floating point if the Java implementation is bad enough.
- igouy 17y ago
  Java sin and cos might not be using the machine hardware. Plain vanilla use of Math.sin and Math.cos = 9.15s Constrain to the range +45 to -45 degrees = 4.88s As demonstrated 20 months ago http://shootout.alioth.debian.org/gp4/program.php?test=partialsums&lang=java&id=4 http://shootout.alioth.debian.org/gp4/program.php?test=parti...
ArnoldLayne 17y ago
It's an issue with the trig functions in Java. See this Factor vs Java comparison: http://factor-language.blogspot.com/2009/08/performance-comparison-between-factor.html http://factor-language.blogspot.com/2009/08/performance-comp...

Any guesses as to why the Java implementation is relatively slow? JNI boundary crossings for floating point arithmetic, or lack of use of the math coprocessor?

by jbooth 17y ago

Any guesses as to why the Java implementation is relatively slow? JNI boundary crossings for floating point arithmetic, or lack of use of the math coprocessor?

smanek 17y ago

I don't know, but it's much slower than I expected. It's the latest Java installable by apt on Debian (all these tests are run in Debian running under VMWare). $ java -version java version "1.6.0_0" OpenJDK Runtime Environment (build 1.6.0_0-b11) OpenJDK 64-Bit Server VM (build 1.6.0_0-b11, mixed mode)

jbooth 17y ago
Dohhhhh don't use OpenJDK :) Get the sun version, it's probably faster. Although it may not be much faster, found this unfixed bug report: http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=5005861 http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=5005861
- jbooth 17y ago
  In more depth : the sun JDK probably does a lot better job of JIT compilation, it's the difference between running interpreted bytecode and compiled native code. On the Hadoop project, checksumming was a major CPU constraint, because you have to assume every file everywhere is possibly on faulty hardware, and checksum everything. They found that it was actually faster to implement CRC32 in pure Java and let the JIT compilation do it's magic than it was to make native calls to zlibc. But that may not apply to floating point if the Java implementation is bad enough.
igouy 17y ago
Java sin and cos might not be using the machine hardware. Plain vanilla use of Math.sin and Math.cos = 9.15s Constrain to the range +45 to -45 degrees = 4.88s As demonstrated 20 months ago http://shootout.alioth.debian.org/gp4/program.php?test=partialsums&lang=java&id=4 http://shootout.alioth.debian.org/gp4/program.php?test=parti...

ArnoldLayne 17y ago

It's an issue with the trig functions in Java. See this Factor vs Java comparison: http://factor-language.blogspot.com/2009/08/performance-comparison-between-factor.html http://factor-language.blogspot.com/2009/08/performance-comp...