Re: [Jikesrvm-researchers] OBJECT_ADDR_POSITION

[Perry C. <per...@us...>]

Hezi,

With respect to your statement   "int objaddr = rawaddr + size -
OBJECT_ADDR_POSITION;",
 first note that

      (1) static final int OBJECT_ADDR_POSITION = SCALAR_HEADER_SIZE +
OBJECT_HEADER_OFFSET;
      (2) the "size" in the original expression is the total size of the
object (including header).

With that, the original statement can be made more unstandable by
considering the equivalent expressions and diagram below.
Also, you should look at the VM_ObjectLayouts.java file for more related
information.

      objaddr = rawaddr + total_size - scalar_header_size -
object_header_offset
                objaddr = (rawaddr + datafield_size) - object_header_offset
      objaddr + object_header_offset = rawaddr + datafield_size

+------+------+------+------+------+------+ - -  + - -  +
|fldN-1| fldx | fld1 | fld0 | tib  |status|
+------+------+------+------+------+------+ - -  + - -  +
^rawaddr                                         ^objaddr
<--------------datafield_size------------><----object_hdr_offset---->


Perry


Hezi Azatchi <he...@cs...>@www-124.southbury.usf.ibm.com on
02/17/2002 12:34:46 PM

Please respond to jik...@ww...

Sent by:    jik...@ww...


To:    <jik...@ww...>
cc:
Subject:    [Jikesrvm-researchers] OBJECT_ADDR_POSITION




My question refers to the Allocator of the markAndSweep
collector/Concurrent collector.

The object address is calculated from the raw address by the following
simple equation:

int objaddr = rawaddr + size - OBJECT_ADDR_POSITION;

My problem: The object which is allocated is associated with
VM_SizeControl of `size` whereas the object allocation can take more than
`size` (because the OBJECT_ADDR_POSITION is negative number, for instance).
Is it ok?
Why?

What is the duty of the OBJECT_ADDR_POSITION, is it used just to cause the
offsets of the status fields of array/scalar to be the same.

Thanks,
--Hezi.

_______________________________________________
Jikesrvm-researchers mailing list
Jik...@ww...
http://www-124.ibm.com/developerworks/oss/mailman/listinfo/jikesrvm-researchers