[CUDA] Implement DynamicSlice/DynamicSliceUpdate (#2533)

* Move DynamicSlice to gpu/primitives

* Implement compute_dynamic_offset in CUDA
This commit is contained in:
Cheng
2025-08-26 07:31:39 +09:00
committed by GitHub
parent 2ca75bb529
commit 4822c3dbe9
12 changed files with 226 additions and 134 deletions

View File

@@ -24,8 +24,6 @@ namespace mlx::core {
}
NO_GPU(BlockMaskedMM)
NO_GPU(DynamicSlice)
NO_GPU(DynamicSliceUpdate)
NO_GPU(FFT)
NO_GPU(GatherMM)
NO_GPU(GatherQMM)