upstream-archive/stable-diffusion.cpp
1
0
Fork 0
mirror of https://github.com/leejet/stable-diffusion.cpp.git synced 2026-09-26 07:57:43 -05:00
Code Issues Packages Projects Releases Wiki Activity

perf: reduce CPU overhead in graph execution and sampling (#1997)

Browse Source
This commit is contained in:
leejet
2026-09-19 18:32:58 +08:00
committed by GitHub
parent d32b4e893b
commit 275ab58e01
8 changed files with 195 additions and 123 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/pipeline/diffusion_engine.cpp
+7 -3
View File
@@ -2158,6 +2158,10 @@ sd::Tensor<float> StableDiffusionGGML::sample(const std::shared_ptr<DiffusionMod
RunnerEndOnExit sample_control_runner_end{!control_image.empty() && control_net != nullptr ? control_net.get() : nullptr};
const bool apply_denoise_mask = !denoise_mask.empty() &&
std::any_of(denoise_mask.values().begin(), denoise_mask.values().end(),
[](float value) { return value != 1.f; });
std::vector<int> skip_layers(guidance.slg.layers, guidance.slg.layers + guidance.slg.layer_count);
float cfg_scale = guidance.txt_cfg;
float img_cfg_scale = guidance.img_cfg;
@@ -2287,13 +2291,13 @@ sd::Tensor<float> StableDiffusionGGML::sample(const std::shared_ptr<DiffusionMod
hunyuan_timestep_r_tensor = sd::Tensor<float>::from_vector({sigmas[step + 1]});
}
sd::Tensor<float> noised_input = x * c_in;
if (!denoise_mask.empty() && (version == VERSION_WAN2_2_TI2V || sd_version_is_ltxav(version) || sd_version_is_lingbot_video(version))) {
if (apply_denoise_mask && (version == VERSION_WAN2_2_TI2V || sd_version_is_ltxav(version) || sd_version_is_lingbot_video(version))) {
noised_input = noised_input * denoise_mask + sampling_init_latent * (1.0f - denoise_mask);
}
if (cache_runtime.spectrum_enabled && cache_runtime.spectrum.should_predict()) {
cache_runtime.spectrum.predict(&denoised);
if (!denoise_mask.empty()) {
if (apply_denoise_mask) {
denoised = denoised * denoise_mask + sampling_init_latent * (1.0f - denoise_mask);
}
if (preview_needed && sd_should_preview_denoised()) {
@@ -2516,7 +2520,7 @@ sd::Tensor<float> StableDiffusionGGML::sample(const std::shared_ptr<DiffusionMod
if (cache_runtime.spectrum_enabled) {
cache_runtime.spectrum.update(denoised);
}
if (!denoise_mask.empty()) {
if (apply_denoise_mask) {
denoised = denoised * denoise_mask + sampling_init_latent * (1.0f - denoise_mask);
}
if (preview_needed && sd_should_preview_denoised()) {