From 64a538ee99d9f04c5c44aa15c9be2c021a50f0f4 Mon Sep 17 00:00:00 2001 From: Daniel Lustig Date: Wed, 9 Feb 2022 09:22:18 -0500 Subject: [PATCH] Two small wording changes 1) Clarify that atomics in any GPU's memory (not just in peer GPU memory) are also atomic if only accessed by GPU threads 2) Remove an extra period --- docs/extended_api/thread_scopes.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/extended_api/thread_scopes.md b/docs/extended_api/thread_scopes.md index f249d8608e..67215090e5 100644 --- a/docs/extended_api/thread_scopes.md +++ b/docs/extended_api/thread_scopes.md @@ -84,7 +84,7 @@ An atomic operation is atomic at the scope it specifies if: `1`, or - it affects an object in CPU memory and [`hostNativeAtomicSupported`] is `1`, or -- it affects an object in GPU peer memory and only GPU threads access it. +- it affects an object in GPU memory and only GPU threads access it. Refer to the [CUDA programming guide] for more information on unified memory, CPU memory, and GPU peer memory. @@ -112,7 +112,7 @@ Modify [thread.latch.class paragraph 2] of ISO/IEC IS 14882 (the C++ Standard) as follows: > 2. Concurrent invocations of the member functions of `latch`, other than its > destructor, do not introduce data races -> ***as if they were atomic operations.***. +> ***as if they were atomic operations***. Modify [thread.sema.cnt paragraph 3] of ISO/IEC IS 14882 (the C++ Standard) as follows: