Profiling

ScopedTimer

wp.ScopedTimer objects can be used to gain some basic insight into the performance of Warp applications:

@wp.kernel
def inc_loop(a: wp.array(dtype=float), num_iters: int):
    i = wp.tid()
    for j in range(num_iters):
        a[i] += 1.0

n = 10_000_000
devices = wp.get_cuda_devices()

# pre-allocate host arrays for readback
host_arrays = [
    wp.empty(n, dtype=float, device="cpu", pinned=True) for _ in devices
]

# code for profiling
with wp.ScopedTimer("Demo"):
    for i, device in enumerate(devices):
        a = wp.zeros(n, dtype=float, device=device)
        wp.launch(inc_loop, dim=n, inputs=[a, 500], device=device)
        wp.launch(inc_loop, dim=n, inputs=[a, 1000], device=device)
        wp.launch(inc_loop, dim=n, inputs=[a, 1500], device=device)
        wp.copy(host_arrays[i], a)

The only required argument for the ScopedTimer constructor is a string label, which can be used to distinguish multiple timed code sections when reading the output. The snippet above will print a message like this:

Demo took 0.52 ms

By default, ScopedTimer measures the elapsed time on the CPU and does not introduce any CUDA synchronization. Since most CUDA operations are asynchronous, the result does not include the time spent executing kernels and memory transfers on the CUDA device. It’s still a useful measurement, because it shows how long it took to schedule the CUDA operations on the CPU.

To get the total amount of time including the device executions time, create the ScopedTimer with the synchronize=True flag. This is equivalent to calling wp.synchronize() before and after the timed section of code. Synchronizing at the beginning ensures that all prior CUDA work has completed prior to starting the timer. Synchronizing at the end ensures that all timed work finishes before stopping the timer. With the example above, the result might look like this:

Demo took 4.91 ms

The timing values will vary slightly from run to run and will depend on the system hardware and current load. The sample results presented here were obtained on a system with one RTX 4090 GPU, one RTX 3090 GPU, and an AMD Ryzen Threadripper Pro 5965WX CPU. For each GPU, the code allocates and initializes an array with 10 million floating point elements. It then launches the inc_loop kernel three times on the array. The kernel increments each array element a given number of times - 500, 1000, and 1500. Finally, the code copies the array contents to the CPU.

Profiling complex programs with many asynchronous and concurrent operations can be tricky. Profiling tools like NVIDIA Nsight Systems can present the results in a visual way and capture a plethora of timing information for deeper study. For profiling tools capable of visualizing NVTX ranges, ScopedTimer can be created with the use_nvtx=True argument. This will mark the CPU execution range on the timeline for easier visual inspection. The color can be customized using the color argument, as shown below:

with wp.ScopedTimer("Demo", use_nvtx=True, color="yellow"):
    ...

To use NVTX integration, you will need to install the NVIDIA NVTX Python package.

pip install nvtx

The package allows you to insert custom NVTX ranges into your code (nvtx.annotate) and customize the colors.

Here is what the demo code looks like in Nsight Systems (click to enlarge the image):

There are a few noteworthy observations we can make from this capture. Scheduling and launching the work on the CPU takes about half a millisecond, as shown in the NVTX / Start & End row. This time also includes the allocation of arrays on both CUDA devices. We can see that the execution on each device is asynchronous with respect to the host, since CUDA operations start running before the yellow Demo NVTX range finishes. We can also see that the operations on different CUDA devices execute concurrently, including kernels and memory transfers. The kernels run faster on the first CUDA device (RTX 4090) than the second device (RTX 3090). Memory transfers take about the same time on each device. Using pinned CPU arrays for the transfer destinations allows the transfers to run asynchronously without involving the CPU.

Check out the concurrency documentation for more information about asynchronous operations.

Note that synchronization was not enabled in this run, so the NVTX range only spans the CPU operations used to schedule the CUDA work. When synchronization is enabled, the timer will wait for all CUDA work to complete, so the NVTX range will span the synchronization of both devices:

with wp.ScopedTimer("Demo", use_nvtx=True, color="yellow", synchronize=True):
    ...

CUDA Activity Profiling

ScopedTimer supports timing individual CUDA activities like kernels and memory operations. This is done by measuring the time taken between CUDA events on the device. To get information about CUDA activities, pass the cuda_filter argument to the ScopedTimer constructor. The cuda_filter can be a bitwise combination of the following values:

CUDA profiling flags

wp.TIMING_KERNEL	Warp kernels (this includes all kernels written in Python as @wp.kernel)
wp.TIMING_KERNEL_BUILTIN	Builtin kernels (this includes kernels used by the Warp library under the hood)
wp.TIMING_MEMCPY	CUDA memory transfers (host-to-device, device-to-host, device-to-device, and peer-to-peer)
wp.TIMING_MEMSET	CUDA memset operations (e.g., zeroing out memory in wp.zeros())
wp.TIMING_GRAPH	CUDA graph launches
wp.TIMING_ALL	Combines all of the above for convenience.

When a non-zero cuda_filter is specified, Warp will inject CUDA events for timing purposes and report the results when the ScopeTimer finishes. This adds some overhead to the code, so should be used only during profiling.

CUDA event timing resolution is about 0.5 microseconds. The reported execution time of short operations will likely be longer than the operations actually took on the device. This is due to the timing resolution and the overhead of added instrumentation code. For more precise analysis of short operations, a tool like Nsight Systems can report more accurate data.

Enabling CUDA profiling with the demo code can be done like this:

with wp.ScopedTimer("Demo", cuda_filter=wp.TIMING_ALL):
    ...

This adds additional information to the output:

CUDA timeline:
----------------+---------+------------------------
Time            | Device  | Activity
----------------+---------+------------------------
    0.021504 ms | cuda:0  | memset
    0.163840 ms | cuda:0  | forward kernel inc_loop
    0.306176 ms | cuda:0  | forward kernel inc_loop
    0.451584 ms | cuda:0  | forward kernel inc_loop
    2.455520 ms | cuda:0  | memcpy DtoH
    0.051200 ms | cuda:1  | memset
    0.374784 ms | cuda:1  | forward kernel inc_loop
    0.707584 ms | cuda:1  | forward kernel inc_loop
    1.042432 ms | cuda:1  | forward kernel inc_loop
    2.136096 ms | cuda:1  | memcpy DtoH

CUDA activity summary:
----------------+---------+------------------------
Total time      | Count   | Activity
----------------+---------+------------------------
    0.072704 ms |       2 | memset
    3.046400 ms |       6 | forward kernel inc_loop
    4.591616 ms |       2 | memcpy DtoH

CUDA device summary:
----------------+---------+------------------------
Total time      | Count   | Device
----------------+---------+------------------------
    3.398624 ms |       5 | cuda:0
    4.312096 ms |       5 | cuda:1
Demo took 0.92 ms

The first section is the CUDA timeline, which lists all captured activities in issue order. We see a memset on device cuda:0, which corresponds to clearing the memory in wp.zeros(). This is followed by three launches of the inc_loop kernel on cuda:0 and a memory transfer from device to host issued by wp.copy(). The remaining entries repeat similar operations on device cuda:1.

The next section is the CUDA activity summary, which reports the cumulative time taken by each activity type. Here, the memsets, kernel launches, and memory transfer operations are grouped together. This is a good way to see where time is being spent overall. The memsets are quite fast. The inc_loop kernel launches took about three milliseconds of combined GPU time. The memory transfers took the longest, over four milliseconds.

The CUDA device summary shows the total time taken per device. We see that device cuda:0 took about 3.4 ms to complete the tasks and device cuda:1 took about 4.3 ms. This summary can be used to asses the workload distribution in multi-GPU applications.

The final line shows the time taken by the CPU, as with the default ScopedTimer options (without synchronization in this case).

Customizing the output

It is possible to customize how the activity timing results are reported. The function warp.timing_print() is used by default. To use a different reporting function, pass it as the report_func argument to ScopedTimer. The custom report function should take a list of warp.TimingResult objects as the first argument. Each result in the list corresponds to a single activity and the list represents the complete recorded timeline. By manually traversing the list, you can customize the formatting of the output, apply custom sorting rules, and aggregate the results as desired. The second argument is a string indent that should be printed at the beginning of each line. This is for compatibility with ScopedTimer indenting rules used with nested timers.

Here is an example of a custom reporting function, which aggregates the total time spend in forward and backward kernels:

def print_custom_report(results, indent=""):
    forward_time = 0
    backward_time = 0

    for r in results:
        # aggregate all forward kernels
        if r.name.startswith("forward kernel"):
            forward_time += r.elapsed
        # aggregate all backward kernels
        elif r.name.startswith("backward kernel"):
            backward_time += r.elapsed

    print(f"{indent}Forward kernels  : {forward_time:.6f} ms")
    print(f"{indent}Backward kernels : {backward_time:.6f} ms")

Let’s apply it to one of the Warp examples:

from warp.examples.optim.example_cloth_throw import Example

example = Example(None)
example.use_graph = False  # disable graphs so we get timings for individual kernels

with wp.ScopedTimer("Example", cuda_filter=wp.TIMING_KERNEL, report_func=print_custom_report):
    for iteration in range(5):
        example.step()

This produces a report like this:

Forward kernels  : 187.098367 ms
Backward kernels : 245.070177 ms

Using the activity timing functions directly

It is also possible to capture activity timings without using the ScopedTimer at all. Simply call warp.timing_begin() to start recording activity timings and warp.timing_end() to stop and get a list of recorded activities. You can use warp.timing_print() to print the default activity report or generate your own report from the list of results.

wp.timing_begin(cuda_filter=wp.TIMING_ALL)
...
results = wp.timing_end()

wp.timing_print(results)

Limitations

Currently, detailed activity timing is only available for CUDA devices, but support for CPU timing may be added in the future.

The activity profiling only records activities initiated using the Warp API. It does not capture CUDA activity initiated by other frameworks. A profiling tool like Nsight Systems can be used to examine whole program activities.

CUDA Events Timing

CUDA events can be used for timing purposes outside of the ScopedTimer. Here is an example:

with wp.ScopedDevice("cuda:0") as device:

    # ensure the module is loaded
    wp.load_module(device=device)

    # create events with enabled timing
    e1 = wp.Event(enable_timing=True)
    e2 = wp.Event(enable_timing=True)

    n = 10000000

    # start timing...
    wp.record_event(e1)

    a = wp.zeros(n, dtype=float)
    wp.launch(inc, dim=n, inputs=[a])

    # ...end timing
    wp.record_event(e2)

    # get elapsed time between the two events
    elapsed = wp.get_event_elapsed_time(e1, e2)
    print(elapsed)

The events must be created with the flag enable_timing=True. The first event is recorded at the start of the timed code and the second event is recorded at the end. The function warp.get_event_elapsed_time() is used to compute the time difference between the two events. We must ensure that both events have completed on the device before calling warp.get_event_elapsed_time(). By default, this function will synchronize on the second event using warp.synchronize_event(). If that is not desired, the user may pass the synchronize=False flag and must use some other means of ensuring that both events have completed prior to calling the function.

Note that timing very short operations may yield inflated results, due to the timing resolution of CUDA events and the overhead of the profiling code. In most cases, CUDA activity profiling with ScopedTimer will have less overhead and better precision. For the most accurate results, a profiling tool such as NVIDIA Nsight Systems should be used. The main benefit of using the manual event timing API is that it allows timing arbitrary sections of code rather than individual activities.

Profiling API Reference

class warp.ScopedTimer(

name,

active=True,

print=True,

detailed=False,

dict=None,

use_nvtx=False,

color='rapids',

synchronize=False,

cuda_filter=0,

report_func=None,

skip_tape=False,

)

indent = -1

enabled = True

__init__(

name,

active=True,

print=True,

detailed=False,

dict=None,

use_nvtx=False,

color='rapids',

synchronize=False,

cuda_filter=0,

report_func=None,

skip_tape=False,

)

Context manager object for a timer

Parameters:

• name (str) – Name of timer

• active (bool) – Enables this timer

• print (bool) – At context manager exit, print elapsed time to sys.stdout

• detailed (bool) – Collects additional profiling data using cProfile and calls print_stats() at context exit

• dict (dict) – A dictionary of lists to which the elapsed time will be appended using name as a key

• use_nvtx (bool) – If true, timing functionality is replaced by an NVTX range

• color (int or str) – ARGB value (e.g. 0x00FFFF) or color name (e.g. ‘cyan’) associated with the NVTX range

• synchronize (bool) – Synchronize the CPU thread with any outstanding CUDA work to return accurate GPU timings

• cuda_filter (int) – Filter flags for CUDA activity timing, e.g. warp.TIMING_KERNEL or warp.TIMING_ALL

• report_func (Callable) – A callback function to print the activity report (wp.timing_print() is used by default)

• skip_tape (bool) – If true, the timer will not be recorded in the tape

extra_msg

Can be set to a string that will be added to the printout at context exit.

Type:

str

elapsed

The duration of the with block used with this object

Type:

float

timing_results

The list of activity timing results, if collection was requested using cuda_filter

Type:

list[TimingResult]

class warp.TimingResult(device, name, filter, elapsed)

Timing result for a single activity.

Parameters:

raw_result (warp.utils.timing_result_t) – The result structure obtained from C++ (internal use only)

device

The device where the activity was recorded.

Type:

warp.Device

name

The activity name.

Type:

str

filter

The type of activity (e.g., warp.TIMING_KERNEL).

Type:

int

elapsed

The elapsed time in milliseconds.

Type:

float

__init__(device, name, filter, elapsed)

warp.timing_begin(cuda_filter=TIMING_ALL, synchronize=True)

Begin detailed activity timing.

Parameters:

• cuda_filter (int) – Filter flags for CUDA activity timing, e.g. warp.TIMING_KERNEL or warp.TIMING_ALL

• synchronize (bool) – Whether to synchronize all CUDA devices before timing starts

warp.timing_end(synchronize=True)

End detailed activity timing.

Parameters:

synchronize (bool) – Whether to synchronize all CUDA devices before timing ends

Returns:

A list of TimingResult objects for all recorded activities.

Return type:

list[TimingResult]

warp.timing_print(results, indent='')

Print timing results.

Parameters:

• results (list[TimingResult]) – List of TimingResult objects.

• indent (str) – Optional indentation for the output.