LLM inference in C/C++
C++ ggml
CUDA: add a reserve to avoid spurious warning on older GCC builds (#29317)
Aman Gupta66fba63af1f4
diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cuindex 60d72046e3..27b83503d2 100644--- a/ggml/src/ggml-cuda/ggml-cuda.cu+++ b/ggml/src/ggml-cuda/ggml-cuda.cu@@ -3470,6 +3470,7 @@ static int ggml_cuda_try_fuse(ggml_backend_cuda_context * cuda_ctx, ggml_cgraphggml_cuda_topk_moe_args args;const bool can_fuse = ggml_cuda_topk_moe_fusion(cgraph, i, args);std::vector<ggml_op> ops;+ ops.reserve(13); // max ops; avoids gcc -Wstringop-overflow false positiveif (can_fuse) {const ggml_tensor * logits = node->src[0];@@ -4539,6 +4540,7 @@ static void ggml_backend_cuda_graph_optimize(ggml_backend_t backend, ggml_cgraphggml_cuda_topk_moe_args args;const bool can_fuse = ggml_cuda_topk_moe_fusion(cgraph, i, args);std::vector<ggml_op> ops;+ ops.reserve(13); // max ops; avoids gcc -Wstringop-overflow false positiveconst ggml_tensor * node = cgraph->nodes[i];