v3.9 update (#2203)

* v3.9 update

* voidD

---------

Co-authored-by: yuzhai <yuzhai@nvidia.com>

examples/65_distributed_gemm/65_distributed_gemm.cu

@@ -834,10 +834,10 @@ int main(int argc, char const **args) {
   CUDA_CHECK(cudaGetDevice(&current_device_id));
   CUDA_CHECK(cudaGetDeviceProperties(&props, current_device_id));
   cudaError_t error = cudaGetDeviceProperties(&props, 0);
-  if (props.major < 9) {
+  if (props.major != 9 || props.minor != 0) {
     std::cerr
-      << "This example requires a GPU of NVIDIA's Hopper Architecture or "
-      << "later (compute capability 90 or greater)." << std::endl;
+      << "This example requires a GPU of NVIDIA's Hopper Architecture "
+      << "(compute capability 90)." << std::endl;
     return 0;
   }
 

examples/65_distributed_gemm/README.md

@@ -63,6 +63,10 @@ procedure is the same, simply modify the following line in the example:
 using TP = _8;
 ```
 
+## References
+* [Distributed GEMM Blog](https://blog.shi-labs.com/distributed-gemm-88be6a481e2b)
+* [Distributed GEMM Talk on CUDA Mode](https://www.youtube.com/watch?v=NHRTCQBZokg)
+
 ## Copyright
 
 Copyright (c) 2017 - 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.

examples/65_distributed_gemm/REQUIREMENTS.md

@@ -17,6 +17,8 @@ Like all other CUTLASS examples, the NVIDIA driver, runtime, and CUDA Toolkit ar
 This example specifically requires CUDA Toolkit 12.6 or newer, due to some of the necessary
 CUDA graph APIs.
 
+The minimum CUDA driver version for running this example is [560.28.03](https://docs.nvidia.com/cuda/archive/12.6.0/cuda-toolkit-release-notes/index.html#id5).
+
 ### Hardware / driver settings
 
 This example requires Hopper GPUs with NVLink network.