youngkingdom/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[Doc] Make C++ code more plausible (#2156)

Browse Source 
Co-authored-by: Haicheng Wu <haichengw@nvidia.com>
This commit is contained in:
Ronan Keryell
2025-04-10 11:35:46 -07:00
committed by GitHub
parent 19cc2a5feb
commit dd76dec4ef
1 changed files with 7 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
media/docs/cpp/blackwell_cluster_launch_control.md
View File

@ -43,7 +43,7 @@ __device__ non_persistent_kernel(...) {
setup_common_data_structures();
dim3 workCoordinates = blockIdx;
coordinate_specific_compute(workCoordinates);
}
}
```
#### Static Persistent Kernel
``` c++
@ -51,9 +51,10 @@ __device__ non_persistent_kernel(...) {
__device__ static_persistent_kernel(...) {
setup_common_data_structures(...);
dim3 workCoordinates = blockIdx;
bool isValidId;
do {
coordinate_specific_compute(workCoordinates);
isValidId, workCoordinates = staticTileScheduler.fetch_next_work();
std::tie(isValidId, workCoordinates) = staticTileScheduler.fetch_next_work();
} while (isValidId);
}
```
@ -65,9 +66,11 @@ __device__ static_persistent_kernel(...) {
__device__ clc_dynamic_persistent_kernel(...) {
setup_common_data_structures(...);
dim3 workCoordinates = blockIdx;
dim3 newClcID;
bool isValidId;
do {
coordinate_specific_compute(workCoordinates);
isValidId, newClcID = clcTileScheduler.fetch_next_work();
std::tie(isValidId, newClcID) = clcTileScheduler.fetch_next_work();
workCoordinates = newClcID;
} while (isValidId);
}
@ -76,7 +79,7 @@ __device__ clc_dynamic_persistent_kernel(...) {
### Cluster Launch Control Pipeline Class
Please refer to the `PipelineCLCFetchAsync` pipeline class defined in [Cluster launch control pipeline class](https://github.com/NVIDIA/cutlass/tree/main/include/cutlass/pipeline/sm100_pipeline.hpp). Cluster launch control queries can be pipelined and mananged by an asynchronous pipeline with producer-consumer relationship (See
Please refer to the `PipelineCLCFetchAsync` pipeline class defined in [Cluster launch control pipeline class](https://github.com/NVIDIA/cutlass/tree/main/include/cutlass/pipeline/sm100_pipeline.hpp). Cluster launch control queries can be pipelined and managed by an asynchronous pipeline with producer-consumer relationship (See
[pipeline](pipeline.md) document). The producer is the scheduler warp of the 0th CTA in the cluster and the consumers are all warps that need `ClcID`s.
To setup a CLC pipeline correctly, we need to make sure the params are set to the right values: