Speedup of SVC prediction

[Dannynis]

Hey, im exploring your cool package and when using SVC with polynomial kernel i do see improvement during fitting phase.
But when benchmarking prediction time I see slow-down in results. Is this behavior expected ?
Thank you !

[Alexsandruss]

Can you specify data and SVC parameters, please? Without them, it is hard to say is prediction slow-down expected.

I run this example and prediction performance with sklearnex is better than stock:

import logging
logging.getLogger().setLevel(logging.INFO)

from timeit import default_timer as timer
from sklearn.datasets import make_classification

from sklearn.svm import SVC as SVCStock
from sklearnex.svm import SVC as SVCOptimized


random_state = 42
data_params = {
    'n_samples': 5000,
    'n_features': 1024,
    'n_informative': 512,
    'n_classes': 5,
    'random_state': random_state
}
x, y = make_classification(**data_params)

svm_params = {
    'kernel': 'poly'
}

n_runs = 10
svc = SVCStock(**svm_params).fit(x, y)
t0 = timer()
for i in range(n_runs):
    svc.predict(x)
t1 = timer()
svc = SVCOptimized(**svm_params).fit(x, y)
t2 = timer()
for i in range(n_runs):
    svc.predict(x)
t3 = timer()

print(f'Stock time[s]: {(t1 - t0) / n_runs}')
print(f'Optim time[s]: {(t3 - t2) / n_runs}')