以下是一些测试以供比较：
面具中的几个Truthy值：

import numpy as np

test = np.arange(100000)
mask = test < 100

%timeit np.multiply(test, 3, out=test, where=mask)
# 6.98 µs ± 90.2 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)

%timeit test[mask] *= 3
# 21.3 µs ± 159 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)

%timeit test[mask] = test[mask] * 3
# 21.9 µs ± 214 ns per loop (mean ± std. dev. of 7 runs, 10,000 loops each)

有很多真理的价值观：

test = np.arange(100000)
mask = test < 90000

%timeit np.multiply(test, 3, out=test, where=mask)
# 60.9 µs ± 554 ns per loop (mean ± std. dev. of 7 runs, 10,000 loops each)

%timeit test[mask] *= 3
# 100 µs ± 474 ns per loop (mean ± std. dev. of 7 runs, 10,000 loops each)

%timeit test[mask] = test[mask] * 3
# 120 µs ± 1.09 µs per loop (mean ± std. dev. of 7 runs, 10,000 loops each)

随机掩码：

test = np.arange(100000)
mask = np.random.choice([True, False], size=100000)

%timeit np.multiply(test, 3, out=test, where=mask)
# 411 µs ± 4.25 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)

%timeit test[mask] *= 3
# 712 µs ± 7.93 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)

%timeit test[mask] = test[mask] * 3
# 714 µs ± 5.72 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)

为了完整性，不带掩码：

%timeit np.multiply(test, 3, out=test)
# 22.4 µs ± 126 ns per loop (mean ± std. dev. of 7 runs, 10,000 loops each)

所以我推荐使用np.multiply和out=arr以及where=mask。