[clang] [llvm] [Clang] Add __builtin_assume_dereferenceable to encode deref assumption. (PR #121789)

[Erich Keane via llvm-commits]

================
@@ -2761,6 +2761,41 @@ etc.).
 
 Query for this feature with ``__has_builtin(__builtin_assume_separate_storage)``.
 
+``__builtin_assume_dereferenceable``
+-------------------------------------
+
+``__builtin_assume_derefernceable`` is used to provide the optimizer with the
+knowledge that the pointer argument P is dereferenceable up to the specified
+number of bytes.
+
+**Syntax**:
+
+.. code-block:: c++
+
+    __builtin_assume_dereferenceable(const void *, size_t)
+
+**Example of Use**:
+
+.. code-block:: c++
+
+  int foo(int *x, int y) {
+      __builtin_assume_dereferenceable(x, 4);
+      int z = 0;
+      if (y == 1) {
+        // The optimizer may execute the load of x unconditionally.
----------------
erichkeane wrote:

A better description as to WHY it can do so unconditionally would be nice.  I originally assumed it was making some sort of assumption about the value of 'y' and was unclear why that would be the case.  

Here's a side note: Does this builtin ALSO imply the opposite?  A `__builtin_assume_dereferenceable` states that the 1st 4 bytes are dereferenceable (so only index 0?), but does that mean it is NOT `derferenceable` beyond that?  For example, do we NOW know that:

`x[3]` is UB?  I should hope not...

https://github.com/llvm/llvm-project/pull/121789