Best practices

Asynchronicity

While the SimAI client doesn’t use async/await mechanics, it is somewhat asynchronous in nature. While uploading geometries is a blocking method, running a prediction or a postprocessing returns the created object immediately, before the result is computed on the servers or available locally. This behavior makes it possible to request that multiple computations be run on the SimAI platform without waiting for any of the data to be available.

To wait for an object to be fully available, you can call the wait() method on the object. For example, you can call the Prediction.wait() method on a prediction. Or, you can call the global SimAIClient.wait() method to wait for all requests to complete.

Alternatively, you can try to access the object’s data, in which case the SimAI client automatically waits for the data to be ready if needed. Because of this behavior, when running a large number of computations, you should send all requests before accessing any of the data.

This example requests the predictions and postprocessings sequentially, which requires waiting for the data to be available and used before requesting the next one.

import ansys.simai.core

simai = ansys.simai.core.from_config()
speeds = [5.9, 5.10, 5.11]

for geom in simai.geometries.list():
    for vx in speeds:
        # Run prediction
        pred = geom.run_prediction(Vx=vx)
        # Request global coefficients postprocessing
        # Because you are accessing the data, you must wait for the computation to finish
        coeffs = pred.post.global_coefficients().data
        # do something with the data
        print(coeffs)

This more efficient example requests all the predictions and postprocessings right away and then processes the data once they are all available.

import ansys.simai.core

simai = ansys.simai.core.from_config()
speeds = [5.9, 5.10, 5.11]
predictions = []

for geom in simai.geometries.list():
    for vx in speeds:
        # Run prediction
        pred = geom.run_prediction(Vx=vx)
        # Request global coefficients postprocessing
        # Because you are not accessing the data, you are not blocked
        pred.post.global_coefficients()
        predictions.append(pred)

simai.wait()  # Wait for all objects requested locally to be complete
for pred in predictions:
    # do something with the data
    print(pred.post.global_coefficients().data)