If CDS archived heap objects can be loaded at their "requested addresses", we can avoid relocation of their oop fields.

The calculation of the requested addresses is too complicated with -XX:-UseCompressedOops. We first pick a point close to the top of the *dumptime* heap:

https://github.com/openjdk/jdk/blob/0c86c31bccd676e1cfbd35898ee16e89d5752688/src/hotspot/share/cds/archiveHeapWriter.cpp#L342

Then, we find out where this point would be if the *runtime* heap starts at 0x10000000:

https://github.com/openjdk/jdk/blob/0c86c31bccd676e1cfbd35898ee16e89d5752688/src/hotspot/share/cds/heapShared.cpp#L1695-L1720

The problem is for -XX:-UseCompressedOops, the word "requested address" has two different meaning (between the two blocks of code referenced above). This difference is not explained in the comments here:

https://github.com/openjdk/jdk/blob/0c86c31bccd676e1cfbd35898ee16e89d5752688/src/hotspot/share/cds/archiveHeapWriter.hpp#L73

==============================
With -XX:-UseCompressedOops, the OS will randomly pick a range for the runtime heap, so we will always need to relocate anyway. We can greatly simplify the dumping code by writing the first archived heap object at 0x10000000.