alphard/stable-diffusion.cpp
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

perf: make ggml_conv_2d faster

Browse Source
This commit is contained in:
leejet 2023-08-26 16:58:15 +08:00
parent 008d80a0b1
commit d765b95ed1
2 changed files with 13 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
ggml

@ -1 +1 @@
Subproject commit d73671e5f1fdb359ea52c6744f19ebce5149ae91
Subproject commit 3a0b87bde946d2d5d4896f1e700272e54d07968a

12
stable-diffusion.cpp
View File

@ -3256,6 +3256,10 @@ class StableDiffusionGGML {
}
}
}
#ifdef GGML_PERF
ggml_graph_print(&diffusion_graph);
#endif
int64_t t1 = ggml_time_ms();
LOG_INFO("step %d sampling completed, taking %.2fs", i + 1, (t1 - t0) * 1.0f / 1000);
LOG_DEBUG("diffusion graph use %.2fMB runtime memory: static %.2fMB, dynamic %.2fMB",
@ -3345,6 +3349,10 @@ class StableDiffusionGGML {
int64_t t0 = ggml_time_ms();
ggml_graph_compute_with_ctx(ctx, &vae_graph, n_threads);
int64_t t1 = ggml_time_ms();
#ifdef GGML_PERF
ggml_graph_print(&vae_graph);
#endif
LOG_DEBUG("computing vae graph completed, taking %.2fs", (t1 - t0) * 1.0f / 1000);
result = ggml_dup_tensor(res_ctx, moments);
@ -3470,6 +3478,10 @@ class StableDiffusionGGML {
int64_t t0 = ggml_time_ms();
ggml_graph_compute_with_ctx(ctx, &vae_graph, n_threads);
int64_t t1 = ggml_time_ms();
#ifdef GGML_PERF
ggml_graph_print(&vae_graph);
#endif
LOG_DEBUG("computing vae graph completed, taking %.2fs", (t1 - t0) * 1.0f / 1000);
result_img = ggml_dup_tensor(res_ctx, img);