b10481: CUDA: MMVQ nwarps=8 for bs=1 for dense models on DGX Spark (#26843)

<table> <tr><td> <a href="https://www.reddit.com/r/LocalLLaMA/comments/1x8bqqs/cuda_faster_mmvq_for_lowbit_types_by_am17an_pull/"> <img alt="CUDA: faster MMVQ for low-bit types by

<ul> <li>cuda, hip: GDN chunked kernel</li> </ul> <p>Ported-from: <a class="issue-link js-issue-link"

<ul> <li>cuda: cap Stream-K blocks at the number of k-blocks</li> </ul> <p>Stream-K normally launches one CUDA block per SM. For small matrices,<br /> this can exceed the total

<table> <tr><td> <a href="https://www.reddit.com/r/LocalLLaMA/comments/1svdjfa/cuda_reduce_mmq_streamk_overhead_by/"> <img alt="CUDA: reduce MMQ stream-k overhead by

<ul> <li> <p>CUDA: refactor swizzling code</p> </li> <li> <p>fix templates/loop bounds</p> </li> </ul>

<table> <tr><td> <a href="https://www.reddit.com/r/LocalLLaMA/comments/1x5skta/cuda_hip_gdn_chunked_kernel_by_am17an_pull/"> <img alt="cuda, hip: GDN chunked kernel by am17an ·

<ul> <li>cuda: stage the lightning indexer queries in head passes for MUSA</li> </ul> <p>MUSA archs 21 and 22 cap static shared memory at 28 KB, and the tile<br /> kernel staged