And in order to avoid a double dereference in the dispatcher, directly store
the normalEntry in the map.
The index to the block map becomes ((((DR<<1) | IR) << 30) | (address >> 2)).
This has been chosen since the msr bits change less often than the address,
thus we keep nearby entries together.
Also do not call the C dispatcher in case the assembly dispatcher didn't
find a block, since it wouldn't find a block either due to the 1:1 mapping,
except when falling back to the non shm segment lookup table.
By using a shm memory segment for the fast_block_map that is sparsely
allocated (i.e. on write by the OS) instead of a statically allocated
array we can make the block lookup faster by:
* Having a bigger space available for lookup that doesn't take up
too much memory, because the OS will only allocate the needed
pages when written to.
* Decrease the time spent to lookup a block in the assembly dispatcher
due to less comparisions and shorter code (for example the pc check
has been entirely dropped since only the msrBits need to be validated).
When the JIT block cache is full the shm segment will also be released
and reallocated to avoid allocating too much memory. It will also be
reset when the instruction cache is flushed by the PPC code to avoid
having stale entries.
Also fallback to the original method in case the memory segment couldn't
be allocated.
Frajo Haider