Unverified50% confidenceFactExact time
At the scheduling level, GC-MoE uses a three-node CUDA Graph to launch the dispatch kernel and the grouped-GEMM kernel as concurrent root nodes (on Hopper the graph has four nodes because of the fused shared experts).
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/4/2026
First Seen
Valid until: 2/2/2027
Sources
Related Entities
Related Claims
UnverifiedThe performance of CUDA Kernels is extremely dependent on alignment; slight mismatches in matrix dimensions, thread block sizes, or memory access patterns will cause throughput to drop significantly, and because Anifer only serves fixed model shapes and GPU architectures, each Kernel can be written in exactly the optimal form.68% similarUnverifiedMatrix multiplication on GPUs is computed in parallel across thousands of CUDA cores via the cuBLAS library.65% similarUnverifiedThe same convolution layer may have multiple kernel implementations, such as direct convolution, Winograd, and implicit GEMM based on Tensor Cores.64% similarUnverifiedGC-MoE is an open-source FP8/FP4 MoE inference operator for NVIDIA Hopper and Blackwell GPUs, which uses CUDA Green Contexts to partition the device into two partitions: a small communication partition and a large compute partition.62% similarUnverifiedMeta's KernelBench and related research from institutions such as DeepMind have attempted to use LLMs to automatically generate and evaluate GPU kernels61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1133729API
curl https://kongchang.com/api/v1/knowledge/claims/1133729MCP
get_claim(id=1133729)