[sv-bc] Reading array with out-of-range or x/z index

Hi,

Steven Sharp wrote in Mantis 1067, then on 1364:
Section 5.2.2 states that for an array reference, if the index is out of
bounds or any bit in the address is x or z, the value of the reference
shall be x. But we have now allowed arrays of reals, which can't be x,
so there is no value specified for that. 

One possibility would be to have the reference be 0.0, since that is the
initialization value for reals the same way that x is the initialization
value for non-reals. Another possibility would be NaN, though that has
not been used in any other such situations (e.g. initialization).

That section is now 11.5.2. The sentence has not changed since then.
Steven later wrote,
I believe that the value to be used has been specified as 0.0 in the
1800 LRM.
but I did not find such a statement.

I found the following:
11.5.1 Vector bit-select and part-select addressing contains the
following statements:
If the bit-select is out of the address bounds or the bit-select is x or
z, then the value returned by the reference shall be x.

A part-select that addresses a range of bits that are completely out of
the address bounds of the vector, packed array, packed structure,
parameter or concatenation, or a part-select that is x or z shall yield
the value x when read and shall have no effect on the data stored when
written. Part-selects that are partially out of range shall, when read,
return x for the bits that are out of range and shall, when written,
only affect the bits that are in range.

11.5.2 Array and memory addressing contains the following:
If the index is out of the address bounds or if any bit in the address
is x or z, then the value of the reference shall be x.

7.4.6 Indexing and slicing of arrays contains:
If an index expression is out of the address bounds or if any bit in the
address is X or Z, then the index shall be invalid. The result of
reading from an array with an invalid index shall return the default
uninitialized value for the array element type. Writing to an array with
an invalid index shall perform no operation.

In 6.8 (Variable declarations), we have
Table 6-7 contains the default values for variables if no initializer is
specified.
Table 6-7-Default values
Type 				Default initial value
4-state integral 		'X
2-state integral 		'0
real, shortreal 			0.0
enumeration 			base type default initial value
string 				"" (empty string)
event 				new event
class 				null
chandle (Opaque handle) null

In 7.9.6 (Accessing invalid indices) in the subclause on associative
arrays, we have:
If an invalid index (i.e., 4-state expression has X's) is used during a
read operation or an attempt is made to read a nonexistent entry, then a
warning shall be issued; and the default initial value for the array
type shall be returned, as shown in Table 7-2. A user specified default
shall not generate a warning and returns the value specified in 7.10.11.
Table 7-2-Value read from a nonexistent associative array entry
Type of array 			Value read
4-state integral type 	'X
2-state integral type 	'0
enumeration 			base type default initial value
string 				""
class 				null
event 				null
If an invalid index is used during a write operation, the write is
ignored, and a warning is issued.


7.11.1 (Queue operators) says,
An invalid index value (i.e., a 4-state expression with X's or Z's, or a
value that lies outside 0...$) shall cause a read operation (e = Q[n])
to return the default initial value for the type of queue item (as
described in Table 7-2).

11.4.14.4 (Streaming dynamically sized data) says,
If the range expression evaluates to a range greater than the extent of
the array size, the entire array is streamed, and the remaining items
are generated using the default value (as described in Table 7-2) for
the given array.

I see several issues:
1. Table 6-7 differs from Table 7-2 in the case of events. Yet both are
referred to as 'default initial value'.

2. Table 6-7 refers only to variables. But (most) nets have 'z as their
default initial value, not 'x. That is ok for Table 6-7 because it is
not talking about net contexts. But elsewhere nets have to be taken into
account.

3. Table 7-2 omits real and chandle associative array element types. But
these can also be element types of associative arrays. Only the index
types of the such arrays are restricted. The other types need to be
added.

4. Should reading an out-of-range/non-existent array element return a
default initial value or something else? We see that Table 7-2 currently
differs from Table 6-7 for events. Presumably that is deliberate.
Assuming Table 7-2 is correct, it should not be referred to as a
'default initial value', even though that term has some meaning in the
context of associative arrays.

5. What should be returned if one reads an array element with an address
containing x/z? The same as the answer to the previous question?

6. What about net arrays? If reading a net array with an address which
is out-of-range or x/z should return X instead of Z, then it is again
not the default initial value. Reading an out-of-range net bit-select in
Verilog-XL returns X.

7. 11.5.1 specifies that writing a vector with an invalid part-select
has no effect. It should say the same for bit-selects.

8. The sentence quoted from 11.5.2 should perhaps be deleted and being
inconsistent with 7.4.6 and also redundant, replacing it instead with a
reference to 7.4.6. Ideally, the subclauses should be merged, but
lacking that, at least remove the contradictions. 

9. 7.4.6 says, "The result of reading from an array with an invalid
index shall return the default uninitialized value for the array element
type." That sentence probably needs to be updated.

10. References to 'default initial value' in 7.9.6 and 7.10.11 should be
changed.

11. If Table 7-2 is to be a general table specifying the value returned
when reading an array with an out-of-range or x/z index, it should be
taken out of the associative array subclause and moved to 7.4.6, and
7.4.6 should reference that table. 

12. 7.11.1 on queues should presumably not talk about a default initial
value and should reference the table in its new location.

13. What about 11.4.14.4? This refers to writing new elements.
Considering the types legal in this context, maybe there is no
difference between the two tables, but the terminology and the table
should be consistent with each other. That is, if what is now Table 7-2
is to be used, then they should not be called default values.

Thanks,
Shalom

Shalom Bresticker
Intel Jerusalem LAD DA
+972 2 589-6582
+972 54 721-1033

---------------------------------------------------------------------
Intel Israel (74) Limited

This e-mail and any attachments may contain confidential material for
the sole use of the intended recipient(s). Any review or distribution
by others is strictly prohibited. If you are not the intended
recipient, please contact the sender and delete all copies.

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.