Skip to content

GitLab

Unverified Commit 98522149 authored by yizhang2077's avatar yizhang2077 Committed by GitHub
Browse files

mirror fix for custom allreduce (#3124)

parent 5d9d15e7
Hide whitespace changes
Inline Side-by-side
Showing with 1 addition and 1 deletion
sgl-kernel/src/sgl-kernel/csrc/trt_reduce_internal.cu
View file @ 98522149
...@@ -160,7 +160,7 @@ __inline__ __device__ void block_barrier(uint32_t** signals, uint32_t const flag ...@@ -160,7 +160,7 @@ __inline__ __device__ void block_barrier(uint32_t** signals, uint32_t const flag
} }
template <typename T, int RANKS_PER_NODE, bool COPY_INPUT = true> template <typename T, int RANKS_PER_NODE, bool COPY_INPUT = true>
static __global__ void oneShotAllReduceKernel(AllReduceParams params) { static __global__ void __launch_bounds__(512, 1) oneShotAllReduceKernel(AllReduceParams params) {
// Suppose that two GPUs participate in the AR exchange, and we start four blocks. // Suppose that two GPUs participate in the AR exchange, and we start four blocks.
// The message is partitioned into chunks as detailed below: // The message is partitioned into chunks as detailed below:
// message // message
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment