leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 27 Packages Projects Releases Wiki Activity

Fix typos in the text (#2417)

Browse Source
This commit is contained in:
Oleksandr Pavlyk
2025-07-16 21:51:12 -04:00
committed by GitHub
parent a1aaf2300a
commit 9892624b66
1 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
media/docs/cpp/dependent_kernel_launch.md
+4 -4
View File
@@ -5,11 +5,11 @@ overlap their execution, named
[Programmatic Dependent Launch (PDL)](https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#programmatic-dependent-launch-and-synchronization). [Programmatic Dependent Launch (PDL)](https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#programmatic-dependent-launch-and-synchronization).
This allows kernels with conflict in global memory to programmatically and safely overlap portions This allows kernels with conflict in global memory to programmatically and safely overlap portions
of their execution. Primary kernel can signal it is about to finish execution, and the next kernel is expected to of their execution. Primary kernel can signal it is about to finish execution, and the next kernel is expected to
programatically wait on the previous kernel to finish flushing its memory. programmatically wait on the previous kernel to finish flushing its memory.
We enable PDL by setting a flag through the extended CUDA launch APIs. All CUTLASS kernels with PDL support We enable PDL by setting a flag through the extended CUDA launch APIs. All CUTLASS kernels with PDL support
will wait on the prior kernel to flush its output to memory and signal the next kernel to start. This means will wait on the prior kernel to flush its output to memory and signal the next kernel to start. This means
they can safely be dropped in with any other set of kernels using PDL as long as they also adhear to waiting on they can safely be dropped in with any other set of kernels using PDL as long as they also adhere to waiting on
the prior to flush its memory as well. the prior to flush its memory as well.
For more information, we refer you to the [PDL section in the CUDA Programming Guide](https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#programmatic-dependent-launch-and-synchronization). For more information, we refer you to the [PDL section in the CUDA Programming Guide](https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#programmatic-dependent-launch-and-synchronization).
@@ -36,10 +36,10 @@ gemm.run(
## Model-Aware Optimizations with PDL ## Model-Aware Optimizations with PDL
In [example 63](https://github.com/NVIDIA/cutlass/tree/main/examples/63_hopper_gemm_with_weight_prefetch/README.md), we use PDL to explicitly optimize for In [example 63](https://github.com/NVIDIA/cutlass/tree/main/examples/63_hopper_gemm_with_weight_prefetch/README.md), we use PDL to explicitly optimize for
performance of kernels where we know that one of the input matricies (our weights) will not be produced by a prior performance of kernels where we know that one of the input matrices (our weights) will not be produced by a prior
kernel. In that case, we only need to wait on the prior kernels memory flush in order to load the other input matrix kernel. In that case, we only need to wait on the prior kernels memory flush in order to load the other input matrix
(our activations). During our prologue, we can prefetch our weights to improve performance for memory bandwidth-bound (our activations). During our prologue, we can prefetch our weights to improve performance for memory bandwidth-bound
problem sizes. For more informations we refer the reader to [the example](https://github.com/NVIDIA/cutlass/tree/main/examples/63_hopper_gemm_with_weight_prefetch/README.md). problem sizes. For more information, we refer the reader to [the example](https://github.com/NVIDIA/cutlass/tree/main/examples/63_hopper_gemm_with_weight_prefetch/README.md).
## Copyright ## Copyright