================
@@ -7532,6 +7532,25 @@ A managed variable can be accessed in both device and
host code.
}];
}
+def CUDAClusterDimsAttrDoc : Documentation {
+ let Category = DocCatDecl;
+ let Content = [{
+In CUDA/HIP programming, the ``__cluster_dims__`` attribute can be applied to
a kernel function
+to set the dimensions of a thread block cluster. This allows to group multiple
thread blocks into
+a larger unit called a "cluster". `__cluster_dims__` defines the cluster size
as ``(X, Y, Z)``,
+where each value is the number of thread blocks in that dimension.
+ }];
+}
+
+def CUDANoClusterAttrDoc : Documentation {
+ let Category = DocCatDecl;
+ let Content = [{
+In CUDA/HIP programming, the ``__no_cluster__`` attribute can be applied to a
kernel function to
+indicate that the thread block cluster feature will not be enabled at both
compile time and kernel
----------------
shiltian wrote:
Again, like I said, I don’t think we should do this. This isn't a CUDA/HIP
programming guide, and LLVM docs are definitely not the right place for it.
Turning LLVM docs into a CUDA/HIP programming guide isn't something we can just
decide here. I'd expect it to go through an NFC first, and if the community
accepts it, then we can put together a proper document.
As for "some ideas of what the thing does so they can research it more," I
think the current wording already works well. We mention the "thread block
cluster feature" for CUDA and HIP, and that's basically all the keywords
someone needs to search and find the CUDA docs for details.
https://github.com/llvm/llvm-project/pull/156686
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
