Code Execution JVM

Personal notes on JVM code execution. Mostly taken from "Optimizing Java" from Benjamin J. Evans, Chris Newland and James Gough

JVM executes interpreted bytecode, instead of machine code directly which can be considered slow. In order to be increase performance, Java environments solve this problem by providing dynamic compilation, or JIT (just-in-time) compilation.

JIT

JIT is a mechanism but which the JVM monitors witch methods are being executed in order to determine whether individual methods are eligible to directly executable code. Just in time compilation is a technique whereby programs are converted into highly optimized machine code at runtime. Hotspot and most other production-grade JVMs rely heavily on this approach.

The technique gathers information about the runtime and builds a profile that can be used to determine which parts of the program are used frequently and would benefit most from optimization.

Bytecode Interpretation

The JVM interpreter operates as a stack machine. Unlike physical CPU's, there are no registers that are used as immediate holding areas for computation. Instead, all values are put in a evaluation stack, and the stack machine instructions work by transforming the values(s) at the top of the stack.

Areas to hold data:

• Evaluation stack, which is local to a particular method

• Local variables to temporarily store results

• Heap - is shared between methods and between threads

JVM Bytecode

Each machine operation (opcode) is represented by 1 byte. Opcodes run from 0 to 255 (Java 10 only uses around 200).

Load and Store

Family name	Arguments	Description
load	(i1)	Loads value from local variable i1 onto the stack
store	(i1)	Stores top of stack into local variable i1
ldc	c1	Loads value from CP#c1
const		Loads simple constant value onto the stack
pop		Discards value on top of stack
dup		Duplicates value on top of stack
getfield	c1	Loads value from field CP#c1 in object on top of stack onto the stack
putfield	c1
getstatic	c1
putstatic	c1

Arithmetic

Family name	Description
add	Adds two values from top of stack
sub	Subtracts two values from top of stack
div	Divides two values from top of stack
mul	Multiplies two values from top of stack
(cast)	Casts value at top of stack to a different primitive type
neg	Negates value at top of stack
rem	Computes remainder (integer division) of top two values on stack

Flow control

Family name	Arguments	Description
if	(i1)	Branch to the location indicated by the argument
goto	i1	Unconditional branch to the supplied offset
tableswitch
lookupswitch

Method invocation

Opcode	Arguments	Description
invokevirtual	c1	Invokes the method found at CP#c1 via virtual dispatch
invokespecial	c1	Invokes the method found at CP#c1 via special dispatch
invokeinterface	c1, count, 0	Invokes the interface method found at CP#c1 using interface lookup
invokestatic	c1	Invokes the static method found at CP#c1
invokedynamic	c1, 0, 0	Dynamically looks up which method to invoke and executes it

Platform Opcodes

Opcode	Arguments	Description
new	c1	Allocates space for an object of type found at CP#c1
newarray	prim	Allocates space for a primitive array of type prim
anewarray	c1	Allocates space for an object array of type found at CP#c1
arraylength		Replaces array on top of stack with its length
monitorenter		Locks monitor of object on top of stack
monitorexit		Unlocks monitor of object on top of stack