vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-09 19:54:34 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

reserve 200 mib for internal gpu usage

Browse source
This commit is contained in:
Lizonghang 2025-01-27 22:14:12 +04:00
parent 1ca9e7974b
commit 2934cf3e8e
2 changed files with 5 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

3
common/common.cpp
View file

@ -1201,8 +1201,7 @@ static bool assign_layers_to_device(
}
if (dev_gpu[m]) {
float reserved_mem = 0.1f; // reserved shared memory to avoid potential OOM, set to 100 MiB by default
vec_z_gpu[m] = (double)((dev.gpu_props.memory_free - reserved_mem) * GIGABYTE - c_gpu[m]) / (double)(n_layer * b_prime);
vec_z_gpu[m] = (double)(dev.gpu_props.memory_free * GIGABYTE - c_gpu[m]) / (double)(n_layer * b_prime);
if (dev.gpu_support.metal && m == 0 && cparams.keep_out_in_metal) {
vec_z_gpu[m] -= (double)bo / (double)(n_layer * b_prime);
}