Skip to content

ProfilingSettings takes a nonnegative_int and a positive_int #1657

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
elliottslaughter wants to merge 2 commits into
base: master
Choose a base branch
from
Open

ProfilingSettings takes a nonnegative_int and a positive_int #1657

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e9058a7
ProfilingSettings takes a nonnegative_int and a positive_int.
elliottslaughter Jun 19, 2026
eb68dcc
Remove dead code.
elliottslaughter Jun 19, 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
4 changes: 2 additions & 2 deletions bin/run-model/src/run-model/main.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -99,7 +99,7 @@ int main(int argc, char **argv) {
/*optimizer=*/optimizer_attrs,
/*loss=*/std::nullopt,
/*input_tensors=*/input_tensors,
/*profiling_settings=*/ProfilingSettings{0, 0},
/*profiling_settings=*/ProfilingSettings{0_n, 1_p},
/*device_handle=*/device_handle,
/*device_type=*/DeviceType::GPU);

Expand All @@ -108,7 +108,7 @@ int main(int argc, char **argv) {
for (int i = 0; i < num_epochs; i++) {
perform_all_passes_for_pcg_instance(
/*instance=*/pcg_instance,
/*profiling_settings=*/ProfilingSettings{0, 1},
/*profiling_settings=*/ProfilingSettings{0_n, 1_p},
/*device_handle=*/device_handle);
}
});
Expand Down
29 changes: 13 additions & 16 deletions lib/kernels/include/kernels/profiling.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,8 +17,8 @@ std::optional<milliseconds_t> profiling_wrapper(F const &f,
Ts &&...ts) {
if (enable_profiling) {
ProfilingSettings settings = ProfilingSettings{
/*warmup_iters=*/0,
/*measure_iters=*/1,
/*warmup_iters=*/0_n,
/*measure_iters=*/1_p,
};
return profiling_wrapper<F, Ts...>(f, settings, std::forward<Ts>(ts)...);
} else {
Expand All @@ -33,10 +33,6 @@ std::optional<milliseconds_t>
ProfilingSettings const &settings,
DeviceType device_type,
Ts &&...ts) {
if (settings.measure_iters <= 0) {
return std::nullopt;
}

if (device_type == DeviceType::GPU) {
return gpu_profiling_wrapper(f, settings, std::forward<Ts>(ts)...);
} else {
Expand All @@ -49,25 +45,26 @@ template <typename F, typename... Ts>
milliseconds_t cpu_profiling_wrapper(F const &f,
ProfilingSettings const &settings,
Ts &&...ts) {
ASSERT(settings.measure_iters > 0);

device_stream_t stream = get_cpu_device_stream();

using TimePoint = std::chrono::time_point<std::chrono::steady_clock>;

std::optional<TimePoint> start = std::nullopt;
std::optional<TimePoint> end = std::nullopt;

for (int i = 0; i < settings.warmup_iters + settings.measure_iters; i++) {
if (i == settings.warmup_iters) {
for (int i = 0; i < settings.warmup_iters.int_from_nonnegative_int() +
settings.measure_iters.int_from_positive_int();
i++) {
if (i == settings.warmup_iters.int_from_nonnegative_int()) {
start = std::chrono::steady_clock::now();
}
f(stream, std::forward<Ts>(ts)...);
}
end = std::chrono::steady_clock::now();

std::chrono::duration<double, std::milli> avg_duration =
(end.value() - start.value()) / settings.measure_iters;
(end.value() - start.value()) /
settings.measure_iters.int_from_positive_int();

return milliseconds_t{
static_cast<float>(avg_duration.count()),
Expand All @@ -78,16 +75,16 @@ template <typename F, typename... Ts>
milliseconds_t gpu_profiling_wrapper(F const &f,
ProfilingSettings const &settings,
Ts &&...ts) {
ASSERT(settings.measure_iters > 0);

device_stream_t stream = get_gpu_device_stream();

ffEvent_t t_start, t_end;
checkCUDA(ffEventCreate(&t_start));
checkCUDA(ffEventCreate(&t_end));

for (int i = 0; i < settings.warmup_iters + settings.measure_iters; i++) {
if (i == settings.warmup_iters) {
for (int i = 0; i < settings.warmup_iters.int_from_nonnegative_int() +
settings.measure_iters.int_from_positive_int();
i++) {
if (i == settings.warmup_iters.int_from_nonnegative_int()) {
checkCUDA(ffEventRecord(t_start, stream.require_gpu()));
}
f(stream, std::forward<Ts>(ts)...);
Expand All @@ -100,7 +97,7 @@ milliseconds_t gpu_profiling_wrapper(F const &f,
checkCUDA(ffEventDestroy(t_start));
checkCUDA(ffEventDestroy(t_end));
return milliseconds_t{
elapsed / settings.measure_iters,
elapsed / settings.measure_iters.int_from_positive_int(),
};
}

Expand Down
10 changes: 8 additions & 2 deletions lib/kernels/include/kernels/profiling_settings.dtg.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,10 +10,16 @@ features = [
"fmt",
]

includes = [
"utils/nonnegative_int/nonnegative_int.h",
"utils/positive_int/positive_int.h",
]

[[fields]]
name = "warmup_iters"
type = "int"
type = "::FlexFlow::nonnegative_int"


[[fields]]
name = "measure_iters"
type = "int"
type = "::FlexFlow::positive_int"
12 changes: 6 additions & 6 deletions lib/local-execution/test/src/local-execution/computation_graph_instance.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -161,7 +161,7 @@ TEST_SUITE(FF_TEST_SUITE) {
},
/*input_tensors=*/input_tensors,
/*allocator=*/allocator,
/*profiling_settings=*/ProfilingSettings{0, 0},
/*profiling_settings=*/ProfilingSettings{0_n, 1_p},
/*device_handle=*/ff_handle,
/*global_device_id=*/global_device_id);

Expand All @@ -172,7 +172,7 @@ TEST_SUITE(FF_TEST_SUITE) {
for (int i = 0; i < num_epochs; i++) {
perform_all_passes_for_computation_graph_instance(
/*instance=*/computation_graph_instance,
/*profiling_settings=*/ProfilingSettings{0, 0},
/*profiling_settings=*/ProfilingSettings{0_n, 1_p},
/*ff_handle=*/ff_handle,
/*global_device_id=*/global_device_id);
loss_values.push_back(copy_tensor_accessor_r(
Expand Down Expand Up @@ -335,7 +335,7 @@ TEST_SUITE(FF_CUDA_TEST_SUITE) {
},
/*input_tensors=*/input_tensors,
/*allocator=*/allocator,
/*profiling_settings=*/ProfilingSettings{0, 0},
/*profiling_settings=*/ProfilingSettings{0_n, 1_p},
/*device_handle=*/ff_handle,
/*device_idx=*/device_idx);

Expand All @@ -348,7 +348,7 @@ TEST_SUITE(FF_CUDA_TEST_SUITE) {
for (int i = 0; i < num_epochs; i++) {
perform_all_passes_for_computation_graph_instance(
/*instance=*/computation_graph_instance,
/*profiling_settings=*/ProfilingSettings{0, 0},
/*profiling_settings=*/ProfilingSettings{0_n, 1_p},
/*ff_handle=*/ff_handle,
/*device_idx=*/device_idx);
loss_values.push_back(copy_tensor_accessor_r(
Expand Down Expand Up @@ -459,13 +459,13 @@ TEST_SUITE(FF_CUDA_TEST_SUITE) {
},
/*input_tensors=*/input_tensors,
/*allocator=*/allocator,
/*profiling_settings=*/ProfilingSettings{0, 1},
/*profiling_settings=*/ProfilingSettings{0_n, 1_p},
/*device_handle=*/ff_handle,
/*device_idx=*/device_idx);

perform_all_passes_for_computation_graph_instance(
/*instance=*/computation_graph_instance,
/*profiling_settings=*/ProfilingSettings{0, 0},
/*profiling_settings=*/ProfilingSettings{0_n, 1_p},
/*ff_handle=*/ff_handle,
/*device_idx=*/device_idx);
assert_unwrap(computation_graph_instance.get_loss_tensor_accessor());
Expand Down
8 changes: 4 additions & 4 deletions lib/local-execution/test/src/local-execution/cost_estimator/local_cost_estimator.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,8 +44,8 @@ TEST_SUITE(FF_TEST_SUITE) {
/*interconnect_specification=*/interconnect_specification,
/*allocator=*/allocator,
/*profiling_settings=*/
ProfilingSettings{/*warmup_iters=*/0,
/*measure_iters=*/1},
ProfilingSettings{/*warmup_iters=*/0_n,
/*measure_iters=*/1_p},
/*device_handle=*/ff_handle,
/*device_idx=*/device_idx);

Expand Down Expand Up @@ -121,8 +121,8 @@ TEST_SUITE(FF_CUDA_TEST_SUITE) {
/*interconnect_specification=*/interconnect_specification,
/*allocator=*/allocator,
/*profiling_settings=*/
ProfilingSettings{/*warmup_iters=*/0,
/*measure_iters=*/1},
ProfilingSettings{/*warmup_iters=*/0_n,
/*measure_iters=*/1_p},
/*device_handle=*/ff_handle,
/*device_idx=*/device_idx);

Expand Down
2 changes: 1 addition & 1 deletion lib/local-execution/test/src/local-execution/local_task_argument_accessor.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,7 +62,7 @@ TEST_SUITE(FF_TEST_SUITE) {
LocalTaskArgumentAccessor acc = LocalTaskArgumentAccessor{
/*allocator=*/allocator,
/*tensor_slots_backing=*/tensor_slots_backing,
/*profiling_settings=*/ProfilingSettings{0, 0},
/*profiling_settings=*/ProfilingSettings{0_n, 1_p},
/*ff_handle=*/cpu_make_device_handle_t(),
/*op_attrs=*/PCGOperatorAttrs{InputAttrs{input_tensor_shape}},
/*loss_attrs=*/std::nullopt,
Expand Down
8 changes: 4 additions & 4 deletions lib/realm-execution/test/src/realm-execution/test_e2e.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -253,7 +253,7 @@ TEST_SUITE(FF_TEST_SUITE) {
/*loss_mapping=*/cfg.loss_mapping,
},
/*input_tensors=*/input_tensors,
/*profiling_settings=*/ProfilingSettings{0, 0},
/*profiling_settings=*/ProfilingSettings{0_n, 1_p},
/*device_handle=*/device_handle,
/*device_type=*/DeviceType::CPU);

Expand All @@ -264,7 +264,7 @@ TEST_SUITE(FF_TEST_SUITE) {
for (int i = 0; i < num_epochs; i++) {
perform_all_passes_for_pcg_instance(
/*instance=*/pcg_instance,
/*profiling_settings=*/ProfilingSettings{0, 0},
/*profiling_settings=*/ProfilingSettings{0_n, 1_p},
/*device_handle=*/device_handle);
loss_values.push_back(copy_tensor_accessor_r(
dynamic_tensor_accessor_from_instance(
Expand Down Expand Up @@ -332,7 +332,7 @@ TEST_SUITE(FF_CUDA_TEST_SUITE) {
/*loss_mapping=*/cfg.loss_mapping,
},
/*input_tensors=*/input_tensors,
/*profiling_settings=*/ProfilingSettings{0, 0},
/*profiling_settings=*/ProfilingSettings{0_n, 1_p},
/*device_handle=*/device_handle,
/*device_type=*/DeviceType::GPU);

Expand All @@ -343,7 +343,7 @@ TEST_SUITE(FF_CUDA_TEST_SUITE) {
for (int i = 0; i < num_epochs; i++) {
perform_all_passes_for_pcg_instance(
/*instance=*/pcg_instance,
/*profiling_settings=*/ProfilingSettings{0, 0},
/*profiling_settings=*/ProfilingSettings{0_n, 1_p},
/*device_handle=*/device_handle);

loss_values.push_back(copy_tensor_accessor_r(
Expand Down
Loading