Skip to content

[QNN EP] Add dynamic option to set HTP performance mode #26135

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
merged 4 commits into from
Sep 26, 2025
Merged

[QNN EP] Add dynamic option to set HTP performance mode #26135

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
5844185
[QNN EP] Add a EP dynamic option to set Htp performance mode
quic-tirupath Sep 23, 2025
9d6a073
[QNN EP] Add Unit tests for EP dynamic option htp_performance_mode
quic-tirupath Sep 23, 2025
0286bd6
[QNN EP] Fix a bug in htp_performance_mode setting
quic-tirupath Sep 23, 2025
c5f0715
[QNN EP] Change option name to make it QNN specific
quic-ashwshan Sep 25, 2025
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: 7 additions & 0 deletions include/onnxruntime/core/session/onnxruntime_session_options_config_keys.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -408,3 +408,10 @@ static const char* const kOrtSessionOptionsDisableModelCompile = "session.disabl
// Note: UNSUPPORTED models always fail regardless of this setting.
static const char* const kOrtSessionOptionsFailOnSuboptimalCompiledModel =
"session.fail_on_suboptimal_compiled_model";

// THIS OPTION IS NOT A REGULAR SESSION OPTION SINCE IT CAN BE MODIFIED AT ANY TIME
// Meant to be used with SetEpDynamicOptions
// options for HTP performance mode: "burst", "balanced", "default", "high_performance",
// "high_power_saver", "low_balanced", "extreme_power_saver", "low_power_saver", "power_saver",
// "sustained_high_performance". Default to "default".
static const char* const kOrtEpDynamicOptionsHtpPerformanceMode = "ep.dynamic.htp_performance_mode";
11 changes: 11 additions & 0 deletions onnxruntime/core/providers/qnn/qnn_execution_provider.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1576,6 +1576,17 @@ Status QNNExecutionProvider::SetEpDynamicOptions(gsl::span<const char* const> ke
LOGS_DEFAULT(ERROR) << "Invalid EP Workload Type: " << value;
return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT, "Invalid EP Workload Type.");
}
} else if (key == kOrtEpDynamicOptionsHtpPerformanceMode) {
auto backend_type = qnn_backend_manager_->GetQnnBackendType();
if (qnn::QnnBackendType::HTP != backend_type && qnn::QnnBackendType::DSP != backend_type) {
return Status::OK();
}
qnn::HtpPerformanceMode htp_performance_mode = qnn::HtpPerformanceMode::kHtpDefault;
ParseHtpPerformanceMode(value, htp_performance_mode);
if (GetPerThreadContext().IsHtpPowerConfigIdValid()) {
ORT_RETURN_IF_ERROR(qnn_backend_manager_->SetHtpPowerConfig(GetPerThreadContext().GetHtpPowerConfigId(),
htp_performance_mode));
}
} else {
LOGS_DEFAULT(ERROR) << "EP Dynamic Option \"" << key << "\" is not currently supported.";
return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT, "Unsupported EP Dynamic Option");
Expand Down
15 changes: 15 additions & 0 deletions onnxruntime/test/providers/qnn/qnn_ep_context_test.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2076,6 +2076,21 @@ TEST_F(QnnHTPBackendTests, QnnEpDynamicOptions) {
} catch (const std::exception& e) {
EXPECT_STREQ("Unsupported EP Dynamic Option", e.what());
}

const char* const htp_perf_mode_type[] = {"ep.dynamic.htp_performance_mode"};
const char* const eps_type[] = {"extreme_power_saver"};
const char* const shp_type[] = {"sustained_high_performance"};
session.SetEpDynamicOptions(htp_perf_mode_type, shp_type, 1);
ort_output = session.Run(Ort::RunOptions{}, input_names_c.data(), ort_inputs.data(), ort_inputs.size(),
output_names_c.data(), 1);

session.SetEpDynamicOptions(htp_perf_mode_type, eps_type, 1);
ort_output = session.Run(Ort::RunOptions{}, input_names_c.data(), ort_inputs.data(), ort_inputs.size(),
output_names_c.data(), 1);

session.SetEpDynamicOptions(htp_perf_mode_type, shp_type, 1);
ort_output = session.Run(Ort::RunOptions{}, input_names_c.data(), ort_inputs.data(), ort_inputs.size(),
output_names_c.data(), 1);
}

// Implementation of OrtOutStreamWriteFunc that writes the compiled model to a file.
Expand Down
Loading