Talk about how compilers actually work in Java language.

1. COMPILER IN
JAVA
complier java use and effect on
performance
San Jose State University
Computer Science Department
2014 Fall
Kaya Ota

2. AGENDA
 First of all CPU
 Java attempt to do
 Kinds of compilers
 The difference of each complier
 The reasons why Java has multiple compliers

3. FIRST OF ALL
 Any computer (CPU) can execute only assembly or binary code
 As well known, all programs must be translated into binary code.
CPU
Assembly or
binary
Java, python, c,
English,
Japanese, etc

4. WHAT JAVA ATTEMPT TO….
 Compiled Language
 Benefit : once source code
is translated to binary, it does
not have to re-translate. So,
it will be faster than an
interpreted language.
 Downside : compiled
languages are tightly coupled
with CPU. So, programs do
not have a portability.
 Interpreted Language
 Benefit : if a computer has a
appropriate interpreter, then
programs can run on the
machine. (it is portable.)
 Downside : an interpreter
translate each line one by
one, so it need to translate
again and again. Therefore, it
is slower than complied
language.
Java tries to have portability AND speed
of compiled language.

5. COMPILE ONCE & RUN
EVERYWHERE
 To be independent from CPU and OS, java make the process of
compilation into two steps.
 1st step : Source Code  intermediate code (= .class)
 So, only on JVM, an intermediate code is executable
 2nd step : intermediate code  binary
 This is still dependent on CPU and OS.

6. PROGRAM AT COMPILER TIME AND AT
EXECUTION TIME
SRC
.java
Java compiler
(javac [.java
file])
Java
byte
code
.class
Class loader
Byte code verifier
Byte code
source
through
new or file
system
JVM JIT
OS
hardware
1st step compiler
static compiling
2nd step compiler
dynamic
compiling

7. JAVA HAS MULTIPLE KINDS OF
COMPLIER
 Different applications have different needs.
 Long run enterprise application allows more optimizations.
 Smaller client-side application needs fast execution with less resource consumption.
 Client Compiler
 Server Compiler
 Tired Compilation
The primary difference between client compiler and server complier is their
aggressiveness in compiling code

8. DIFFERENCE BETWEEN
CLIENT AND SERVER
footPrint
(memory
usage)
Running time Warming up
time
Client
Server
Time interval between
download and
activation of an
application once

9. CLIENT
 The primary objective of the client compiler is “fast start up.”
 The client complier begins compiling sooner than server complier
does.
 During the beginning of execution, the client compiler will be
faster
 Because the client compiler compiles more code correspondingly
Usually client has a fewer
memory, so client compiler is
optimized in memory usage as
shown in a previous table.

10. SERVER
 Server side applications are expected to run in a long period of
time.
 Instead of using more resources,( such as CPU cycles, larger
code cache, etc) these applications delivers more advanced
optimizations for later in running cycle.

11. TIERED COMPILATION
 Tiered compilation combines the best parts of client side
compilation and server side compilation.
 Tiered compilation gather data during low-impact compiler activity.
 To apply more advanced optimizations later, the compiler can use
the data.

12. DEFAULT SETTING OF COMPILERS
Windows Mac

13. CHANGE YOUR COMPILER
 Windows
 Installer x86 for 32bit machine
 Java option: -client or –server (default client )
 Installer x64 for 64bit machine
 Java option: -hotspot or –server (default server)
 Mac 64 bit ---only for –server

14. HOT SPOT COMPLICATION
 Hot spots of application : the more the section of the code is
executed. the hotter that section is said to be.
 Typically, some subset of codes are executed frequently.
 And, a performance of an application depends on the frequently
executed codes.
 So, java starts running an application with interpreter like and
determine which chunks of codes are invoked frequently
 Then, java decides to compile those chunks of codes(no longer
translate them again!)
 Benefit : by avoiding infrequently invoked codes, java complier can
focus on the performance-critical-parts, without necessarily
increasing the overall compilation time

15. EXAMPLE OF HOT SPOT
 Priority Queue
 insert(int newItem)
 delete(int position)
 increaseKey(int posi, int delta)
 decreaseKey(int posi, int delta)
 shiftUp(int position)
 shiftDown(int position)
Those functions call shiftUp
or shiftDown somewhere in
their procedure .
Therefore, shiftUp and
shiftDown are said to be
hotter than other
functions.

16. REFERENCE
 JVM performance optimization
 http://www.javaworld.com/article/2078635/enterprise-middleware/jvm-performance-
optimization-part-2-compilers.html
 Java Performance : The Definitive Guide by Scott Oaks
 Why Java Is Better Then Other Programming Language And Byte
Code Concept
 http://codemink.com/why-java-is-better-byte-code-concept/