[aarch64-port-dev ] RFR: Fix out by one in writing array barriers

[Edward Nevill]

On Thu, 2013-12-12 at 13:30 +0000, Andrew Haley wrote:

> The first store is done to [base + [end >> card_shift]].  I think this
> is wrong, because [base + [end >> card_shift]] is outside the range.
> We need to make in an inclusive pointer.  So, we need to subtract 1
> from end, not add it:
> 
>            __ lsr(start, start, CardTableModRefBS::card_shift);
>            __ sub(end, end, BytesPerHeapOop); // end - 1 to make inclusive
>            __ lsr(end, end, CardTableModRefBS::card_shift);
>            __ sub(end, end, start); // number of bytes to copy
> 
>           const Register count = end; // 'end' register contains bytes count now
> 	  __ mov(scratch, (address)ct->byte_map_base);
>           __ add(start, start, scratch);
> 	  __ BIND(L_loop);
> 	  __ strb(zr, Address(start, count));
>           __ subs(count, count, 1);
>           __ br(Assembler::HS, L_loop);

But the end pointer is already inclusive, looking at the 3 calls to gen_write_ref_array_post_barrier in stubGenerator_aarch64.cpp

1)

   1268       __ sub(count, count, 1); // make an inclusive end pointer
   1269       __ lea(count, Address(d, count, Address::uxtw(exact_log2(size))));
   1270       gen_write_ref_array_post_barrier(d, count, rscratch1);

2)

   1320       __ sub(count, count, 1); // make an inclusive end pointer
   1321       __ lea(count, Address(d, count, Address::uxtw(exact_log2(size))));
   1322       gen_write_ref_array_post_barrier(d, count, rscratch1);

3)

   1699     __ add(to, to, -heapOopSize);         // make an inclusive end point        er
   1700     gen_write_ref_array_post_barrier(start_to, to, rscratch1);

Regards,
Ed.