diff --git a/src/ggml-cuda/ggml-cuda.cu b/src/ggml-cuda/ggml-cuda.cu
index 060550c572..53a43241d6 100644
--- a/src/ggml-cuda/ggml-cuda.cu
+++ b/src/ggml-cuda/ggml-cuda.cu
@@ -1195,6 +1195,8 @@ static void ggml_cuda_op_mul_mat_cublas(
     const bool use_fp16 = (src0->type == GGML_TYPE_F16 || ggml_is_quantized(src0->type)) && ggml_is_contiguous(src0) && row_diff == src0->ne[1] && dst->op_params[0] == GGML_PREC_DEFAULT;
 
     if (src0->type == GGML_TYPE_BF16 && ggml_is_contiguous(src0) && row_diff == src0->ne[1]) {
+        // MIT licensed. Copyright (C) 2024 Iwan Kawrakow
+        // https://github.com/ikawrakow/ik_llama.cpp/pull/40
         ggml_cuda_pool_alloc<nv_bfloat16> src1_as_bf16(ctx.pool(id));
         if (src1->type != GGML_TYPE_BF16) {
             const to_bf16_cuda_t to_bf16_cuda = ggml_get_to_bf16_cuda(src1->type);