DarthAffe/stable-diffusion.cpp
1
0 Fork 0
mirror of https://github.com/leejet/stable-diffusion.cpp.git synced 2026-11-11 03:24:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

perf: optimize H3 inference and segmented weight prefetch (#2240)

Browse Source
This commit is contained in:
leejet authored and GitHub committed 2026-10-30 00:38:14 +08:00
1 parent e7b7ec6d6f
commit 37bc88b5e9
17 files changed
+367 -210
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

src/core/ggml_runner.h
+4 -4
View File
@@ -141,12 +141,12 @@ private:
size_t logged_segment_count_ = 0;
ggml_status last_compute_status_ = GGML_STATUS_SUCCESS;
sd::ComputeWorkspace::Measurement measure(ggml_cgraph* graph);
sd::ComputeWorkspace::Measurement measure(ggml_cgraph* graph, const std::unordered_set<const ggml_tensor*>* external_outputs = nullptr);
std::vector<DeviceMemoryRequest> memory_requests(const std::vector<sd::BackendBufferSize>& sizes,
size_t pending_cache_bytes) const;
bool fits(const std::vector<DeviceMemoryRequest>& requests,
const std::vector<ggml_tensor*>& params) const;
bool execute_segment(ggml_cgraph* graph, int n_threads);
bool execute_segment(ggml_cgraph* graph, int n_threads, const std::function<void()>& on_before_compute);
std::optional<sd::Tensor<float>> execute_graph(ggml_cgraph* graph, int n_threads, bool no_return, const std::function<bool()>& read_outputs);
protected:
@@ -162,8 +162,8 @@ protected:
sd::ComputeWorkspace workspace_;
ggml_context* compute_ctx = nullptr;
std::unique_ptr<std::max_align_t[]> compute_ctx_buffer_;
bool runner_started_ = false;
bool graph_active_ = false;
bool runner_started_ = false;
bool graph_active_ = false;
size_t max_graph_vram_bytes = 0;
bool graph_cut_layer_split_enabled = false;