maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[NFC][AMDGPU] Minor editorial improvements to AMDGPUUsage.rst 

Differential Revision: https://reviews.llvm.org/D90661
This commit is contained in:
Tony 2020-11-03 02:14:45 +00:00
parent e1af54296c
commit 45bcbe46d7
1 changed files with 4 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
llvm/docs/AMDGPUUsage.rst
View File

@ -4328,7 +4328,7 @@ For GFX6-GFX9:
Scalar memory operations are only used to access memory that is proven to not
change during the execution of the kernel dispatch. This includes constant
address space and global address space for program scope const variables.
address space and global address space for program scope ``const`` variables.
Therefore, the kernel machine code does not have to maintain the scalar cache to
ensure it is coherent with the vector caches. The scalar and vector caches are
invalidated between kernel dispatches by CP since constant address space data
@ -4402,8 +4402,7 @@ in table :ref:`amdgpu-amdhsa-memory-model-code-sequences-gfx6-gfx9-table`.
------------------------------------------------------------------------------------
load atomic unordered *any* *any* *Same as non-atomic*.
store atomic unordered *any* *any* *Same as non-atomic*.
atomicrmw unordered *any* *any* *Same as monotonic
atomic*.
atomicrmw unordered *any* *any* *Same as monotonic atomic*.
**Monotonic Atomic**
------------------------------------------------------------------------------------
load atomic monotonic - singlethread - global 1. buffer/global/ds/flat_load
@ -5581,7 +5580,7 @@ For GFX10:
Scalar memory operations are only used to access memory that is proven to not
change during the execution of the kernel dispatch. This includes constant
address space and global address space for program scope const variables.
address space and global address space for program scope ``const`` variables.
Therefore, the kernel machine code does not have to maintain the scalar cache to
ensure it is coherent with the vector caches. The scalar and vector caches are
invalidated between kernel dispatches by CP since constant address space data
@ -5675,8 +5674,7 @@ table :ref:`amdgpu-amdhsa-memory-model-code-sequences-gfx10-table`.
------------------------------------------------------------------------------------
load atomic unordered *any* *any* *Same as non-atomic*.
store atomic unordered *any* *any* *Same as non-atomic*.
atomicrmw unordered *any* *any* *Same as monotonic
atomic*.
atomicrmw unordered *any* *any* *Same as monotonic atomic*.
**Monotonic Atomic**
------------------------------------------------------------------------------------
load atomic monotonic - singlethread - global 1. buffer/global/flat_load