Skip to content

vulkan: allow graph_optimize for prompt processing workloads #17475

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Nov 26, 2025
Merged

vulkan: allow graph_optimize for prompt processing workloads #17475

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 0 additions & 18 deletions ggml/src/ggml-vulkan/ggml-vulkan.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12990,24 +12990,6 @@ static void ggml_vk_graph_optimize(ggml_backend_t backend, struct ggml_cgraph *
return false;
};

// This function tries to reorder the graph to allow nodes to run in parallel.
// This helps with small batches, but for large batches its a slowdown, probably
// due to cache contention. So only reorder if the majority of nodes have few rows.
int num_small_nodes = 0;
int num_counted_nodes = 0;
for (int i = 0; i < graph->n_nodes; ++i) {
if (!is_empty(graph->nodes[i]) &&
graph->nodes[i]->op != GGML_OP_SET_ROWS) {
if (ggml_nrows(graph->nodes[i]) <= 8) {
num_small_nodes++;
}
num_counted_nodes++;
}
}
if (num_small_nodes < num_counted_nodes / 2) {
return;
}

std::vector<ggml_tensor *> new_order;
std::vector<bool> used(graph->n_nodes, false);
std::set<ggml_tensor *> used_node_set;
Expand Down
Loading