cublas - Robuta Search

https://docs.nvidia.com/cuda/archive/11.4.1/cublas/index.html cuBLAS :: CUDA Toolkit Documentation The API Reference guide for cuBLAS, the CUDA Basic Linear Algebra Subroutine library. cuda toolkit documentation https://deepwiki.com/chengzeyi/stable-fast/4.2-cublas-operations cuBLAS Operations | chengzeyi/stable-fast | DeepWiki This page documents the cuBLAS-accelerated matrix operations implemented in stable-fast. These operations leverage NVIDIA's cuBLAS library to achieve... operations stable fast deepwiki https://forums.developer.nvidia.com/t/cublas-batched-fp32-sgemm-dispatcher-picks-suboptimal-kernel-on-rtx-5090-sm-120/366212 cuBLAS batched FP32 SGEMM dispatcher picks suboptimal kernel on RTX 5090 (sm_120) - GPU-Accelerated... Apr 10, 2026 - While profiling FP32 SGEMM performance on the RTX 5090, I noticed that cublasSgemmStridedBatched dispatches the same cutlass_80_simt_sgemm_128x32_8x5 kernel... https://gist.github.com/csullivan/34d16d6db684dcc48cec9dee613751d4 Sharded decode, sharding rewrite done after FuseOpsByPattern (cublas/cutlass byoc) -- With debug... Sharded decode, sharding rewrite done after FuseOpsByPattern (cublas/cutlass byoc) -- With debug tracing calls - sharded_decode.py https://groups.google.com/g/caffe-users/c/1c6lKwFmBNs Check failed: status == CUBLAS_STATUS_SUCCESS (11 vs. 0) CUBLAS_STATUS_MAPPING_ERROR *** Check fail check failed status success https://dev.to/soytuber/rtx-5090-cublas-bug-neural-texture-compression-multi-gpu-vllm-inference-1cbj RTX 5090 cuBLAS Bug, Neural Texture Compression, Multi-GPU vLLM Inference - DEV Community RTX 5090 cuBLAS Bug, Neural Texture Compression, Multi-GPU vLLM Inference Today's... Tagged with gpu, nvidia, hardware.