[llvm] 6d28dff - [AMDGPU] Update GFX10 memory model to account for MALL
Carl Ritson via llvm-commits
llvm-commits at lists.llvm.org
Wed Nov 17 16:29:55 PST 2021
Author: Carl Ritson
Date: 2021-11-18T09:29:30+09:00
New Revision: 6d28dffb6bf4c97848290b9aee3c19025470e54a
URL: https://github.com/llvm/llvm-project/commit/6d28dffb6bf4c97848290b9aee3c19025470e54a
DIFF: https://github.com/llvm/llvm-project/commit/6d28dffb6bf4c97848290b9aee3c19025470e54a.diff
LOG: [AMDGPU] Update GFX10 memory model to account for MALL
Document memory attached last level (MALL) cache added in GFX10.3.
Reviewed By: t-tye
Differential Revision: https://reviews.llvm.org/D114076
Added:
Modified:
llvm/docs/AMDGPUUsage.rst
Removed:
################################################################################
diff --git a/llvm/docs/AMDGPUUsage.rst b/llvm/docs/AMDGPUUsage.rst
index 3fa80e56f288d..8984ab1a202f3 100644
--- a/llvm/docs/AMDGPUUsage.rst
+++ b/llvm/docs/AMDGPUUsage.rst
@@ -373,7 +373,7 @@ Every processor supports every OS ABI (see :ref:`amdgpu-os`) with the following
- Ryzen 3 Pro 4350G
- Ryzen 3 Pro 4350GE
- **GCN GFX10 (RDNA 1)** [AMD-GCN-GFX10-RDNA1]_
+ **GCN GFX10.1 (RDNA 1)** [AMD-GCN-GFX10-RDNA1]_
-----------------------------------------------------------------------------------------------------------------------
``gfx1010`` ``amdgcn`` dGPU - cumode - Absolute - *rocm-amdhsa* - Radeon RX 5700
- wavefrontsize64 flat - *pal-amdhsa* - Radeon RX 5700 XT
@@ -393,7 +393,7 @@ Every processor supports every OS ABI (see :ref:`amdgpu-os`) with the following
Add product
names.
- **GCN GFX10 (RDNA 2)** [AMD-GCN-GFX10-RDNA2]_
+ **GCN GFX10.3 (RDNA 2)** [AMD-GCN-GFX10-RDNA2]_
-----------------------------------------------------------------------------------------------------------------------
``gfx1030`` ``amdgcn`` dGPU - cumode - Absolute - *rocm-amdhsa* - Radeon RX 6800
- wavefrontsize64 flat - *pal-amdhsa* - Radeon RX 6800 XT
@@ -8571,6 +8571,9 @@ For GFX10:
requirements of acquire, release and sequential consistency.
* The L2 cache can be kept coherent with other agents on some targets, or ranges
of virtual addresses can be set up to bypass it to ensure system coherence.
+* On GFX10.3 a memory attached last level (MALL) cache exists for GPU memory.
+ The MALL cache is fully coherent with GPU memory and has no impact on system
+ coherence. All agents (GPU and CPU) access GPU memory through the MALL cache.
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
More information about the llvm-commits
mailing list