[llvm] [docs] Reword the alignment implications for atomic instructions. (PR #75871)

[Jonas Paulsson via llvm-commits]

https://github.com/JonPsson1 created https://github.com/llvm/llvm-project/pull/75871

Atomic instructions (load / store/ atomicrwm / cmpxchg) are not undefined behavior if they lack natural alignment. They will typically (with AtomicExpand pass enabled) be converted into libcalls.


Does this sound right? (suggestions welcome)



>From de127ec691cbf6221bedb70e51c6da0a2d0b89d2 Mon Sep 17 00:00:00 2001
From: Jonas Paulsson <paulson1 at linux.ibm.com>
Date: Mon, 18 Dec 2023 16:54:51 -0600
Subject: [PATCH] Suggested rewordering for atomic alignments.

---
 llvm/docs/LangRef.rst | 21 +++++++++++++--------
 1 file changed, 13 insertions(+), 8 deletions(-)

diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index 7f4a316a21acee..326b2c29ff38c6 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -10515,9 +10515,10 @@ Atomic loads produce :ref:`defined <memmodel>` results when they may see
 multiple atomic stores. The type of the pointee must be an integer, pointer, or
 floating-point type whose bit width is a power of two greater than or equal to
 eight and less than or equal to a target-specific size limit.  ``align`` must be
-explicitly specified on atomic loads, and the load has undefined behavior if the
-alignment is not set to a value which is at least the size in bytes of the
-pointee. ``!nontemporal`` does not have any defined semantics for atomic loads.
+explicitly specified on atomic loads and must typically be a power of two greater
+or equal to the size of the `<value>` type, or the AtomicExpand pass will convert
+it to a libcall. ``!nontemporal`` does not have any defined semantics for atomic
+loads.
 
 The optional constant ``align`` argument specifies the alignment of the
 operation (that is, the alignment of the memory address). It is the
@@ -10655,7 +10656,11 @@ Atomic loads produce :ref:`defined <memmodel>` results when they may see
 multiple atomic stores. The type of the pointee must be an integer, pointer, or
 floating-point type whose bit width is a power of two greater than or equal to
 eight and less than or equal to a target-specific size limit.  ``align`` must be
-explicitly specified on atomic stores, and the store has undefined behavior if
+explicitly specified on atomic stores and must typically be a power of two
+greater or equal to the size of the `<value>` type, or the AtomicExpand pass will
+convert it to a libcall.
+
+and the store has undefined behavior if
 the alignment is not set to a value which is at least the size in bytes of the
 pointee. ``!nontemporal`` does not have any defined semantics for atomic stores.
 
@@ -10807,8 +10812,8 @@ must be at least ``monotonic``, the failure ordering cannot be either
 A ``cmpxchg`` instruction can also take an optional
 ":ref:`syncscope <syncscope>`" argument.
 
-The alignment must be a power of two greater or equal to the size of the
-`<value>` type.
+Typically, the alignment must be a power of two greater or equal to the size
+of the `<value>` type, or the AtomicExpand pass will convert it to a libcall.
 
 The alignment is only optional when parsing textual IR; for in-memory IR, it is
 always present. If unspecified, the alignment is assumed to be equal to the
@@ -10910,8 +10915,8 @@ the ``atomicrmw`` is marked as ``volatile``, then the optimizer is not
 allowed to modify the number or order of execution of this
 ``atomicrmw`` with other :ref:`volatile operations <volatile>`.
 
-The alignment must be a power of two greater or equal to the size of the
-`<value>` type.
+Typically, the alignment must be a power of two greater or equal to the size
+of the `<value>` type, or the AtomicExpand pass will convert it to a libcall.
 
 The alignment is only optional when parsing textual IR; for in-memory IR, it is
 always present. If unspecified, the alignment is assumed to be equal to the