llama.cpp/ggml/src/ggml-cuda/mmv.cuh

#include "common.cuh"

// maximum number of src0 rows with which to use mul_mat_vec over cuBLAS if FP16 tensor cores are available
#define MMV_MAX_ROWS 512

void ggml_cuda_mul_mat_vec(ggml_backend_cuda_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);

void ggml_cuda_op_mul_mat_vec(
    ggml_backend_cuda_context & ctx,
    const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst, const char * src0_dd_i, const float * src1_ddf_i,
    const char * src1_ddq_i, float * dst_dd_i, const int64_t row_low, const int64_t row_high, const int64_t src1_ncols,
    const int64_t src1_padded_row_size, cudaStream_t stream);