Algorithm · Microsoft · Hard
Prompt During a live coding session, the interviewer gives you a CUDA operation to implement or optimize in an editor. You are expected to produce executable CUDA code quickly, including any necessary host-side launch logic. Typical variants You may be asked to implement a CUDA kernel and its host invocation for one of these operations: Elementwise vector addition or scaling, such as AXPY Matrix transposition A simplified LayerNorm or RMSNorm A simplified softmax A sum or…
Checking your access…