[doc][GlobalISel]Improving generic opcodes for memory operations

[pooja2299]

Summary

Adding examples for each of the generic opcodes involved in memory operations. Yet to add examples for some of the opcodes.

[pooja2299]

llvm/docs/GlobalISel/GenericOpcode.rst

684–685

I am not aware that with what reference previous author has written this example, so needed guidance before changing it .

709

Should I add the description of below example instead of this line?

[paquette]

llvm/docs/GlobalISel/GenericOpcode.rst

656

I don't think you need to change this.

662

I think for this example, you want to use G_ZEXTLOAD.

You'll also want a memoperand.

E.g.

%load:_(s64) = G_LOAD %ptr(p0) :: (load (s32)) ; High 32 bits are undefined.
%load:_(s64) = G_SEXTLOAD %ptr(p0) :: (load (s32)) ; High 32 bits are sign-extended.
%load:_(s64) = G_ZEXTLOAD %ptr(p0) :: (load (s32)) ; High 32 bits are zero-extended.

672

Can we change GEP to G_PTR_ADD?

676

I think I'd drop this part and only have G_INDEXED_LOAD examples. Or, I'd make it clear that these are being combined to a G_INDEXED_LOAD.

I'd also add an example for the post-indexed G_INDEXED_LOAD.

709

I think it's fine as is.

713

I don't think we need to show the original G_STORE + G_PTR_ADD in the example. Only the G_INDEXED_STORE.

[pooja2299]

Edited the redundant examples

[pooja2299]

replaced GEP with G_PTR_ADD

[gandhi21299]

ping

[pooja2299]

In D108319#3073130, @gandhi21299 wrote:

ping

Hello. Was busy with my placements and final year project. Should I commit this patch or does it need any further changes?

[gandhi21299]

LGTM, appreciate your efforts!

[jansvoboda11]

FYI, it seems like a branch containing this patch was accidentally pushed to GitHub: https://github.com/llvm/llvm-project/tree/arcpatch-D108319

[pooja2299]

In D108319#3077600, @jansvoboda11 wrote:

FYI, it seems like a branch containing this patch was accidentally pushed to GitHub: https://github.com/llvm/llvm-project/tree/arcpatch-D108319

Apologies for that. I will revert the pull request.

llvm/docs/GlobalISel/GenericOpcode.rst

 in the top bits is unspecified.
 
 Memory Operations
 -----------------
 
 G_LOAD, G_SEXTLOAD, G_ZEXTLOAD
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
-Generic load. Expects a MachineMemOperand in addition to explicit
+Generic load expects a MachineMemOperand in addition to explicit
 operands. If the result size is larger than the memory size, the
 high bits are undefined, sign-extended, or zero-extended respectively.
 
+.. code-block:: none
+
+  %dst(s8) = G_LOAD %ptr
+  %dst(s32) = G_ZEXT %dst(s8)
+
 Only G_LOAD is valid if the result is a vector type. If the result is larger
 than the memory size, the high elements are undefined (i.e. this is not a
 per-element, vector anyextload)
 
 G_INDEXED_LOAD
 ^^^^^^^^^^^^^^
 
-Generic indexed load. Combines a GEP with a load. $newaddr is set to $base + $offset.
+Generic indexed load opcode combines a GEP with a load. 
+
+.. code-block:: none
+
+  $addr = G_PTR_ADD $base, $offset
+  [...]
+  $val = G_LOAD $addr
+  [...]
+
+  $val, $addr = G_INDEXED_LOAD $base, $offset, 1 (IsPre)   
+
+$addr is set to $base + $offset.
 If $am is 0 (post-indexed), then the value is loaded from $base; if $am is 1 (pre-indexed)
-then the value is loaded from $newaddr.
+then the value is loaded from $addr. 
 
 G_INDEXED_SEXTLOAD
 ^^^^^^^^^^^^^^^^^^
 
 Same as G_INDEXED_LOAD except that the load performed is sign-extending, as with G_SEXTLOAD.
 
 G_INDEXED_ZEXTLOAD
 ^^^^^^^^^^^^^^^^^^
 
 Same as G_INDEXED_LOAD except that the load performed is zero-extending, as with G_ZEXTLOAD.
 
 G_STORE
 ^^^^^^^
 
 Generic store. Expects a MachineMemOperand in addition to explicit
 operands. If the stored value size is greater than the memory size,
 the high bits are implicitly truncated. If this is a vector store, the
 high elements are discarded (i.e. this does not function as a per-lane
 vector, truncating store)
 
 G_INDEXED_STORE
 ^^^^^^^^^^^^^^^
 
 Combines a store with a GEP. See description of G_INDEXED_LOAD for indexing behaviour.
 
+.. code:: none
+
+  G_STORE $val, $base
+  [...]
+  $addr = G_PTR_ADD $base, $offset
+  [...]
+
+  $addr = G_INDEXED_STORE $val, $base, $offset 
+
 G_ATOMIC_CMPXCHG_WITH_SUCCESS
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
 Generic atomic cmpxchg with internal success check. Expects a
 MachineMemOperand in addition to explicit operands.
 
 G_ATOMIC_CMPXCHG
 ^^^^^^^^^^^^^^^^