> On Jul 8, 2025, at 17:46, Siddhesh Poyarekar <siddh...@gotplt.org> wrote:
>
> On 2025-07-08 17:17, Qing Zhao wrote:
>> Are the above the correct and efficient updates to the .ACCESS_WITH_SIZE to
>> resolve both PR121000 and the issue
>> we have with counted_by for pointers?
>
> I don't know about PR121000, but for counted_by with pointers, I think the
> REF_TO_OBJ (and the result_type) would also have to be a->fam and not
> &a->fam, i.e. don't generate an INDIRECT_REF to the .ACCESS_WITH_SIZE.
Yes. That’s right. I already changed this in my local workspace. And worked
well.
With the current .ACCESS_WITH_SIZE, in order to get the correct TYPE_SIZE_UNIT
of the element TYPE for the pointer array, we have to distinguish whether the
TYPE passed by the 6th argument is for FAM or for pointer, therefore, I have to
add one more argument to the .ACCESS_WITH_SIZE:
+ the 7th argument of the call is 1 when for FAM, 0 for pointers.
With the new .ACCESS_WITH_SIZE I proposed in the previous email as:
ACCESS_WITH_SIZE (REF_TO_OBJ, REF_TO_SIZE, CLASS_OF_SIZE,
TYPE_OF_SIZE, ACCESS_MODE)
which returns the REF_TO_OBJ same as the 1st argument;
1st argument REF_TO_OBJ: The reference to the object;
2nd argument REF_TO_SIZE: The reference to the size of the object,
3rd argument CLASS_OF_SIZE + TYPE_OF_SIZE: An integer constant with a TYPE:
The integer constant value of this argument represents:
0: means that the size referenced by the REF_TO_SIZE is the
number of bytes.
1: means that the size referenced by the REF_TO_SIZE is the
number of the elements of the object type;
The TYPE is the same as the TYPE of the object referenced by REF_TO_SIZE.
4th argument ACCESS_MODE:
-1: Unknown access semantics
0: none
1: read_only
2: write_only
3: read_write
5th argument: The TYPE_SIZE_UNIT of the element TYPE of the FAM or the pointer
array. “
Since we pass the TYPE_SIZE_UNIT of the element TYPE directly to the call to
.ACCESS_WITH_SIZE, no need to distinguish whether the TYPE is for FAM or
pointer anymore.
Hope this is clear.
thanks.
Qing
>
> Thanks,
> Sid
