[llvm] [NVPTX] Add Intrinsics for discard.* (PR #128404)

[via llvm-commits]

================
@@ -671,6 +671,43 @@ level on which the priority is to be applied. The only supported value for the s
 For more information, refer to the PTX ISA
 `<https://docs.nvidia.com/cuda/parallel-thread-execution/#data-movement-and-conversion-instructions-applypriority>`_.
 
+``llvm.nvvm.discard.*``'
+^^^^^^^^^^^^^^^^^^^^^^^^
+
+Syntax:
+"""""""
+
+.. code-block:: llvm
+
+  declare void  @llvm.nvvm.discard.global.L2(ptr addrspace(1) %global_ptr, i64 immarg)
+  declare void  @llvm.nvvm.discard.L2(ptr %ptr, i64 immarg)
+
+Overview:
+"""""""""
+
+The '``@llvm.nvvm.discard.*``' semantically behaves like a weak write of an *unstable indeterminate value*: 
+reads of memory locations with *unstable indeterminate values* may return different 
+bit patterns each time until the memory is overwritten.
+This operation *hints* to the implementation that data in the specified cache ``.level`` 
+can be destructively discarded without writing it back to memory. The operand ``size`` is an 
+integer constant that specifies the length in bytes of the address range ``[a, a + size)`` to write 
+*unstable indeterminate values* into. The only supported value for the ``size`` operand is ``128``. 
+If no state space is specified then `generic-addressing` is used. If the specified address does 
+not fall within the address window of ``.global`` state space then the behavior is undefined.
+
+LLVM does not define anywhere what an *unstable indeterminate values* is, and the closest concept 
+LLVM has breaks the example below:
+
+.. code-block:: text
+  
+  discard.global.L2 [ptr], 128;
+  ld.weak.u32 r0, [ptr];
+  ld.weak.u32 r1, [ptr];
+  // The values in r0 and r1 may differ!
----------------
gonzalobg wrote:

@Artem-B is something like this clear enough?

https://github.com/llvm/llvm-project/pull/128404