DarthAffe/stable-diffusion.cpp
1
0
Fork 0
mirror of https://github.com/leejet/stable-diffusion.cpp.git synced 2026-06-09 15:56:39 +00:00
Code Issues Packages Projects Releases Wiki Activity

perf: allocate CPU-offloaded params from runtime device pinned host buffer (#1601)

Browse Source
This commit is contained in:
fszontagh 2026-06-06 10:22:18 +02:00 committed by GitHub
parent 1f9ee88e09
commit 064001b524
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 12 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
src/ggml_extend.hpp
View File

@ -3017,7 +3017,18 @@ public:
LOG_DEBUG("%s skipping params allocation (no tensors)", get_desc().c_str());
return true;
}
params_buffer = ggml_backend_alloc_ctx_tensors(params_ctx, params_backend);
// Pinned host buffer when CPU-offloaded for DMA-direct H2D.
ggml_backend_buffer_type_t params_buft = nullptr;
if (params_backend != runtime_backend) {
ggml_backend_dev_t runtime_dev = ggml_backend_get_device(runtime_backend);
if (runtime_dev != nullptr) {
params_buft = ggml_backend_dev_host_buffer_type(runtime_dev);
}
}
if (params_buft == nullptr) {
params_buft = ggml_backend_get_default_buffer_type(params_backend);
}
params_buffer = ggml_backend_alloc_ctx_tensors_from_buft(params_ctx, params_buft);
if (params_buffer == nullptr) {
LOG_ERROR("%s alloc params backend buffer failed, num_tensors = %i",
get_desc().c_str(),