4 articles found
Unlock high-performance LLM inference with the TokenSpeed-kernel. Learn how this open-source subsyst...
Discover how In-Kernel Broadcast Optimization (IKBO) reduces compute-intensive net latency by up to...
Optimize your Mamba-2 SSD modules with a fused Triton kernel for 1.50x-2.51x speedups on NVIDIA A100...
NVIDIA H100 and RTX4000 Tensor Cores truncate FP32 outputs to 13-bit mantissas during FP8 matmuls. S...