Page History

Compressed oops in the Hotspot JVM

What's an oop

...

?

An "oop", or "ordinary object pointer" in HotSpot parlance is a managed pointer to an object. It  It is normally the same size as a native machine pointer.  A managed pointer is carefully tracked by the Java application and GC subsystem, so that storage for unused objects can be reclaimed.  This process can also involve relocating (copying) objects which are in use, so that storage can be compacted.

The term "oop" is traditional to certain VMs that derive from Smalltalk and Self, including the following:

• Self (an prototype-based relative of Smalltalk) https://github.com/russellallen/self/blob/master/vm/src/any/objects/oop.hh
• Strongtalk (a Smalltalk implementation) http://code.google.com/p/strongtalk/wiki/VMTypesForSmalltalkObjects
• Hotspot  http://hg.openjdk.java.net/hsx/hotspot-main/hotspot/file/0/src/share/vm/oops/oop.hpp
• V8 http://code.google.com/p/v8/source/browse/trunk/src/objects.h (mentions "smi" but not "oop")

(In some of these systems, the term "smi" refers to a special non-oop word, a pseudo-pointer, which encodes a small, 30-bit integer.  This term can also be found in the V8 implementation of Smalltalk.)

Why should they be compressed?

On an LP64 system, a machine word, and hence an oop, requires 64 bits, while on an ILP32 system, oops are only 32 bits.  But on an ILP32 system pointer, which means 64 bits on an LP64 system. On an ILP32 system, there is a maximum heap size of somewhat less than 4Gb or so, which is not enough for many applications. On an LP64 system, though, the heap of for any given run is almost twice as big as the corresponding IPL32 may have to be around 1.5 times as large as for the corresponding ILP32 system (assuming the run fits both modes). This is due to the expanded size of managed pointers. Memory is pretty cheap, but these days bandwidth and cache is in short supply, so doubling significantly increasing the size of the heap , just to get over the 4Gb limit , is painful.

(Additionally, on x86 chips, the ILP32 mode provides half the usable registers that the LP64 mode does. SPARC is not affected this way; RISC chips start out with lots of registers and just widen them for LP64 mode.)

...

• the klass field of every object
• every oop instance field
• every element of an oop array (objArray)
• in a constant pool... (what is the rule here??)
• (what else is compressed??)

The following oops in the heap are never compressed:

...

The Hotspot VM's data structures to manage Java classes are not compressed. These are generally found in the section of the Java heap known as the Permanent Generation (PermGen).

In the interpreter, oops are never compressed. These include JVM locals and stack elements, outgoing call arguments, and return values. The interpreter eagerly decodes oops loaded from the heap, and encodes them before storing them to the heap.

Likewise, method calling sequences, either interpreted or compiled, do not deal with use compressed oops.

In compiled code, oops are compressed or not according to the outcome of various optimizations. Optimized code may succeed in moving a compressed oop from one location in the managed heap to another without ever decoding it. Likewise, if the chip (i.e., x86) supports addressing modes which can be used for the decode operation, compressed oops might not be decoded even if they are used to address object fields or array elements.

...

! int R8; oop[] R9;  // R9 is 64 bits
! oop R10 = R9[R8];  // R10 is 32 bits
! load compressed ptr from wide base ptr:
movl R10, [R9 + R8<<2R8<<3 + 16]
! klassOop R11 = R10._klass;  // R11 is 32 bits
! void* const R12 = GetHeapBase();
! load compressed klass ptr from compressed base ptr:
movl R11, [R12 + R10<<3 + 8]

...

If UseCompressedOops is true, the klass is 32 bits. Non-arrays have a gap field immediately after the klass, while arrays store the length field immediately after the klass.

Zero based compressed oops

Compressed oops use an arbitrary address for the narrow oop base which is calculated as java heap base minus one (protected) page size for implicit NULL checks to work. This means a generic field reference is next:

<narrow-oop-base> + (<narrow-oop> << 3) + <field-offset>.

If the narrow oop base can be made to be zero (the java heap doesn't actually have to start at offset zero), then a generic field reference can be just next:

(<narrow-oop << 3) + <field-offset>

Theoretically it allows to save the heap base add (current Register Allocator does not allow to save register). Also with zero base the null check of compressed oop is not needed.
Current code for decoding compressed oops looks like this:

if (<narrow-oop> == NULL)
    <wide_oop> = NULL
else
    <wide_oop> = <narrow-oop-base> + (<narrow-oop> << 3)

With zero narrow oop base the code is much simpler. It needs only shift to decode/encode a compressed oop:

<wide_oop> = <narrow-oop> << 3

Also if java heap size < 4Gb and it can be moved into low virtual address space (below 4Gb) then compressed oops can be used without encoding/decoding.

Zero based implementation tries to allocated java heap using different strategies based on the heap size and a platform it runs on.
First, it tries to allocate java heap below 4Gb to use compressed oops without decoding if heap size < 4Gb.
If it fails or heap size > 4Gb it will try to allocate the heap below 32Gb to use zero based compressed oops.
If this also fails it will switch to regular compressed oops with narrow oop base.