Skip to content

[GPU] coverity fix: avoid overflow in MoE weight offsets #36911

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
sungeunk wants to merge 2 commits into
base: master
Choose a base branch
from
+4 −3
Open

[GPU] coverity fix: avoid overflow in MoE weight offsets #36911

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8534f60
fix(gpu): avoid overflow in MoE weight offsets
sungeunk Jul 16, 2026
5d6f077
fix clangformat issue
sungeunk Jul 16, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 4 additions & 3 deletions src/plugins/intel_gpu/src/graph/impls/ocl_v2/moe/moe_3gemm_swiglu_opt.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1105,12 +1105,13 @@ class moe_3gemm_swiglu_opt_impl : public PrimitiveImplOCL {
}
}
// weight shape: [ic, oc], type: u4/i8
int64_t wei_offset = j * get_bytes_count(dnnl_weights[i].ic * dnnl_weights[i].oc, moe_fusion_wei_addr.weight[i]->get_layout());
int64_t wei_offset =
j * get_bytes_count(static_cast<int64_t>(dnnl_weights[i].ic) * dnnl_weights[i].oc, moe_fusion_wei_addr.weight[i]->get_layout());
dnnl_weights[i].weight =
convert2dnnl(moe_fusion_wei_addr.weight[i], {dnnl_weights[i].ic, dnnl_weights[i].oc}, dnnl::memory::format_tag::ba, wei_offset);

// scale shape: [ic / ic_group_size, oc], type: f16
int64_t scale_offset = j * get_bytes_count(dnnl_weights[i].ic * dnnl_weights[i].oc / dnnl_weights[i].ic_group_size,
int64_t scale_offset = j * get_bytes_count(static_cast<int64_t>(dnnl_weights[i].ic) * dnnl_weights[i].oc / dnnl_weights[i].ic_group_size,
moe_fusion_wei_addr.scale[i]->get_layout());
dnnl_weights[i].scale = convert2dnnl(moe_fusion_wei_addr.scale[i],
{dnnl_weights[i].ic / dnnl_weights[i].ic_group_size, dnnl_weights[i].oc},
Expand All @@ -1120,7 +1121,7 @@ class moe_3gemm_swiglu_opt_impl : public PrimitiveImplOCL {
// zp shape: [ic / ic_group_size, oc], type: u4/i8
// Skip ZP memory allocation for symmetric quantization (has_zp=false) to save memory
if (cur_moe->_config.has_zp) {
int64_t zp_offset = j * get_bytes_count(dnnl_weights[i].ic * dnnl_weights[i].oc / dnnl_weights[i].ic_group_size,
int64_t zp_offset = j * get_bytes_count(static_cast<int64_t>(dnnl_weights[i].ic) * dnnl_weights[i].oc / dnnl_weights[i].ic_group_size,
moe_fusion_wei_addr.zp[i]->get_layout());
dnnl_weights[i].zp = convert2dnnl(moe_fusion_wei_addr.zp[i],
{dnnl_weights[i].ic / dnnl_weights[i].ic_group_size, dnnl_weights[i].oc},
Expand Down
Loading