[PATCH] D121242: [AMDGPU] gfx940 memory model

[Scott Linder via Phabricator via llvm-commits]

scott.linder added inline comments.


================
Comment at: llvm/docs/AMDGPUUsage.rst:8673
 
+.. _amdgpu-amdhsa-memory-model-gfx940:
+
----------------
rampitec wrote:
> foad wrote:
> > I don't know how to review 2500 lines of text. How is it different from GFX90A?
> The main change is the replacement of glc/scc/slc with sc0/sc1/nt bits. While looks functionally similar it is conceptually different. Instead of locking individual caches we now requesting coherency at a specific memory scope.
One approach is to isolate the "Memory Model GFX90A" and "Memory Model GFX940" sections into their own files, and then use your preferred diff-tool with more granular highlighting (i.e. word-based rather than line-based) to compare the two models.

At that point the diff is actually very concise.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D121242/new/

https://reviews.llvm.org/D121242