ZI optimization?

armlink will initialize the RW area either by

• copying an image of the RW area to the execution address at runtime (as a special case, if the load and execute addresses are equal then no copying needs to happen)
• compressing the image of the RW area at link time and decompressing it to the execution address (which might overlap the load address) at runtime  (see http://infocenter.arm.com/help/topic/com.arm.doc.dui0474i/CHDDIBFI.html)

If a bit of ZI is in the RW area it will be treated as any other bit of RW but will have the initial value of zero.

Large ZI objects are not moved to ZI and are effectively run-length compressed.

armlink will initialize the RW area either by

• copying an image of the RW area to the execution address at runtime (as a special case, if the load and execute addresses are equal then no copying needs to happen)
• compressing the image of the RW area at link time and decompressing it to the execution address (which might overlap the load address) at runtime  (see http://infocenter.arm.com/help/topic/com.arm.doc.dui0474i/CHDDIBFI.html)

If a bit of ZI is in the RW area it will be treated as any other bit of RW but will have the initial value of zero.

Large ZI objects are not moved to ZI and are effectively run-length compressed.

Hi Scott,

Thanks, all makes sense now! I'd not joined the dots regarding the RW compression and the repeated 0's for the ZI area small objects.

Cheers,

Niall.

Hi Niall, do you need any more precision on this?

If not, would you please be able to mark the question answered?

Thanks!