Skip to content

fix: reuse graph cut plans across CFG passes #1943

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
leejet merged 1 commit into from
Sep 6, 2026
Merged

fix: reuse graph cut plans across CFG passes #1943

Show file tree
Hide file tree
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
31 changes: 22 additions & 9 deletions src/core/ggml_graph_cut.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -537,10 +537,12 @@ namespace sd::ggml_graph_cut {
return signature;
}

bool plan_matches_graph(ggml_cgraph* gf, const Plan& plan) {
static bool plan_matches_graph(ggml_cgraph* gf,
const Plan& plan,
const std::vector<uint64_t>& layout) {
GGML_ASSERT(gf != nullptr);
if (plan.leaf_names.size() != static_cast<size_t>(gf->n_leafs) ||
plan.layout != graph_layout(gf, false)) {
plan.layout != layout) {
return false;
}
for (int i = 0; i < gf->n_leafs; ++i) {
Expand All @@ -558,6 +560,11 @@ namespace sd::ggml_graph_cut {
return cut_markers == plan.cut_markers;
}

bool plan_matches_graph(ggml_cgraph* gf, const Plan& plan) {
GGML_ASSERT(gf != nullptr);
return plan_matches_graph(gf, plan, graph_layout(gf, false));
}

ggml_tensor* output_tensor(ggml_cgraph* gf, const Segment& segment, size_t output_index) {
GGML_ASSERT(gf != nullptr);
if (output_index >= segment.output_node_indices.size()) {
Expand Down Expand Up @@ -938,20 +945,26 @@ namespace sd::ggml_graph_cut {
GGML_ASSERT(gf != nullptr);
GGML_ASSERT(cache != nullptr);

if (cache->graph_cut_plan.available &&
plan_matches_graph(gf, cache->graph_cut_plan)) {
return cache->graph_cut_plan;
const auto layout = graph_layout(gf, false);
auto& plans = cache->graph_cut_plans;
for (auto it = plans.begin(); it != plans.end(); ++it) {
if (it->available && plan_matches_graph(gf, *it, layout)) {
plans.splice(plans.begin(), plans, it);
return plans.front();
}
}

int64_t t_plan_begin = ggml_time_ms();
Plan plan = build_plan(backend, gf, params_tensor_set, log_desc);
cache->graph_cut_plan = plan;
int64_t t_plan_begin = ggml_time_ms();
plans.push_front(build_plan(backend, gf, params_tensor_set, log_desc));
if (plans.size() > PlanCache::MAX_PLANS) {
plans.pop_back();
}
if (log_desc != nullptr) {
LOG_INFO("%s build cached graph cut plan done (taking %lld ms)",
log_desc,
ggml_time_ms() - t_plan_begin);
}
return plan;
return plans.front();
}

} // namespace sd::ggml_graph_cut
4 changes: 3 additions & 1 deletion src/core/ggml_graph_cut.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@

#include <array>
#include <cstdint>
#include <list>
#include <string>
#include <unordered_map>
#include <unordered_set>
Expand Down Expand Up @@ -48,7 +49,8 @@ namespace sd::ggml_graph_cut {
};

struct PlanCache {
Plan graph_cut_plan;
static constexpr size_t MAX_PLANS = 4;
std::list<Plan> graph_cut_plans;
};

static constexpr const char* GGML_RUNNER_CUT_PREFIX = "ggml_runner_cut:";
Expand Down
Loading