Skip to content

Reduce allocation size in test Gather_oveflow_check from 4GiB to >2GiB #27483

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

Merged
adrianlizarraga merged 4 commits into from
Feb 28, 2026
Merged

Reduce allocation size in test Gather_oveflow_check from 4GiB to >2GiB #27483

Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
fe9a780
Reduce allocation size in test Gather_oveflow_check from 4GiB to >2GiB
adrianlizarraga Feb 27, 2026
3e066a8
Skip test on 32-bit platforms due to limitation on std::vector::max_s…
adrianlizarraga Feb 27, 2026
ec8e2ca
Increase Android vram to 7GiB. Address review comments
adrianlizarraga Feb 27, 2026
608804b
Decrease Android emulator memory to 5GiB. That should work
adrianlizarraga Feb 27, 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
24 changes: 9 additions & 15 deletions onnxruntime/test/providers/cpu/tensor/gather_op_test.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -342,16 +342,10 @@ TEST(GatherOpTest, Gather_axis1_indices2d_string) {
}

TEST(GatherOpTest, Gather_overflow_check) {
// Skip on 32-bit platforms where size_t overflow would truncate the large expected
// output shape and where allocating the full reference tensor is infeasible.
#if SIZE_MAX <= UINT32_MAX
GTEST_SKIP() << "Gather_overflow_check skipped on 32-bit platforms.";
#endif

// The test uses dimensions (65537, 2) and indices of length 65537, which produce an output
// shape of (65537, 65537).
// The test uses dimensions (46341, 2) and indices of length 46341, which produce an output
// shape of (46341, 46341).
//
// 65537 x 65537 = 4,295,098,369 which is greater than the maximum value of a 32-bit integer (2,147,483,647).
// 46341 x 46341 = 2,147,488,281 which is just greater than the maximum value of a 32-bit integer (2,147,483,647).
//
// This test is to verify CPU implementation of the Gather operator doesn't overflow when calculating
// the output shape and generating the output tensor.
Expand All @@ -360,15 +354,15 @@ TEST(GatherOpTest, Gather_overflow_check) {
test.AddAttribute<int64_t>("axis", 1LL);

// Inputs
const std::vector<int64_t> data_dims{65537, 2};
const std::vector<int64_t> indices_dims{65537};
const std::vector<int64_t> data_dims{46341, 2};
const std::vector<int64_t> indices_dims{46341};
std::vector<uint8_t> data_values(static_cast<size_t>(data_dims[0] * data_dims[1]), 1);
std::vector<int64_t> indices_values(static_cast<size_t>(indices_dims[0]), 1);
std::vector<uint8_t> expected_output_values(static_cast<size_t>(65537) * static_cast<size_t>(65537), 1);
std::vector<uint8_t> expected_output_values(static_cast<size_t>(46341) * static_cast<size_t>(46341), 1);

test.AddInput<uint8_t>("data", {65537, 2}, data_values);
test.AddInput<int64_t>("indices", {65537}, indices_values);
test.AddOutput<uint8_t>("output", {65537, 65537}, expected_output_values);
test.AddInput<uint8_t>("data", {46341, 2}, data_values);
test.AddInput<int64_t>("indices", {46341}, indices_values);
test.AddOutput<uint8_t>("output", {46341, 46341}, expected_output_values);
Comment thread
adrianlizarraga marked this conversation as resolved.
Outdated

std::vector<std::unique_ptr<IExecutionProvider>> execution_providers;
execution_providers.emplace_back(DefaultCpuExecutionProvider());
Expand Down
Loading