Skip to content

KFI-203 Improve thread safety of packing in convolve_kleidiai.cpp #26575

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
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

KFI-203 Improve thread safety of packing in convolve_kleidiai.cpp #26575

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
8 changes: 4 additions & 4 deletions onnxruntime/core/mlas/lib/kleidiai/convolve_kleidiai.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -332,8 +332,8 @@ static std::shared_ptr<std::byte[]> RhsPackWeightsBiasSme(const size_t co, const
const float* weights, const float* bias,
MLAS_THREADPOOL* ThreadPool)
{
//cache of prepacked kai rhs weights and biases
static std::unordered_map<RhsCacheKey, std::shared_ptr<std::byte[]>> rhs_cache;
// Cache of prepacked kai rhs weights and biases. thread_local to prevent interference from parallel sessions.
thread_local std::unordered_map<RhsCacheKey, std::shared_ptr<std::byte[]>> rhs_cache;

RhsCacheKey key = { co, ci, kh, kw, dilationh, dilationw, HashWeights(weights) };

Expand Down Expand Up @@ -474,8 +474,8 @@ static std::unique_ptr<std::byte[]> LhsPackImageDataSme(const size_t ci, const s

auto nhwc = NChwToNhwc(1, ci, ih, iw, in, 1, 1, false, ThreadPool);

//cache of computed lhs ptr offsets
static std::unordered_map<LhsCacheKey, std::shared_ptr<const void*[]>> lhs_ptrs_cache;
// Cache of computed lhs ptr offsets. thread_local to prevent interference from parallel sessions.
thread_local std::unordered_map<LhsCacheKey, std::shared_ptr<const void*[]>> lhs_ptrs_cache;

std::shared_ptr<const void*[]> lhs_ptrs;
if (auto found = lhs_ptrs_cache.find(key); found != lhs_ptrs_cache.end()) {
Expand Down
Loading