How to use the pyperf._utils.percentile function in pyperf

def test_calibrate_warmups(self, nwarmup, unit):
        half = nwarmup + (len(self.warmups) - nwarmup) // 2
        sample1 = [value for loops, value in self.warmups[nwarmup:half]]
        sample2 = [value for loops, value in self.warmups[half:]]
        first_value = sample1[0]

        # test if the first value is an outlier
        values = sample1[1:] + sample2
        q1 = percentile(values, 0.25)
        q3 = percentile(values, 0.75)
        iqr = q3 - q1
        outlier_max = (q3 + 1.5 * iqr)
        # only check maximum, not minimum
        outlier = not(first_value <= outlier_max)

        mean1 = statistics.mean(sample1)
        mean2 = statistics.mean(sample2)
        mean_diff = (mean1 - mean2) / float(mean2)

        s1_q1 = percentile(sample1, 0.25)
        s2_q1 = percentile(sample2, 0.25)
        s1_q3 = percentile(sample1, 0.75)
        s2_q3 = percentile(sample2, 0.75)
        q1_diff = (s1_q1 - s2_q1) / float(s2_q1)
        q3_diff = (s1_q3 - s2_q3) / float(s2_q3)

values = sample1[1:] + sample2
        q1 = percentile(values, 0.25)
        q3 = percentile(values, 0.75)
        iqr = q3 - q1
        outlier_max = (q3 + 1.5 * iqr)
        # only check maximum, not minimum
        outlier = not(first_value <= outlier_max)

        mean1 = statistics.mean(sample1)
        mean2 = statistics.mean(sample2)
        mean_diff = (mean1 - mean2) / float(mean2)

        s1_q1 = percentile(sample1, 0.25)
        s2_q1 = percentile(sample2, 0.25)
        s1_q3 = percentile(sample1, 0.75)
        s2_q3 = percentile(sample2, 0.75)
        q1_diff = (s1_q1 - s2_q1) / float(s2_q1)
        q3_diff = (s1_q3 - s2_q3) / float(s2_q3)

        mad1 = median_abs_dev(sample1)
        mad2 = median_abs_dev(sample2)
        # FIXME: handle division by zero
        mad_diff = (mad1 - mad2) / float(mad2)

        if self.args.verbose:
            stdev1 = statistics.stdev(sample1)
            stdev2 = statistics.stdev(sample2)
            stdev_diff = (stdev1 - stdev2) / float(stdev2)

            sample1_str = format_values(unit, (s1_q1, mean1, s1_q3, stdev1, mad1))
            sample2_str = format_values(unit, (s2_q1, mean2, s2_q3, stdev2, mad2))
            print("Calibration: warmups=%s" % format_number(nwarmup))

first_value = sample1[0]

        # test if the first value is an outlier
        values = sample1[1:] + sample2
        q1 = percentile(values, 0.25)
        q3 = percentile(values, 0.75)
        iqr = q3 - q1
        outlier_max = (q3 + 1.5 * iqr)
        # only check maximum, not minimum
        outlier = not(first_value <= outlier_max)

        mean1 = statistics.mean(sample1)
        mean2 = statistics.mean(sample2)
        mean_diff = (mean1 - mean2) / float(mean2)

        s1_q1 = percentile(sample1, 0.25)
        s2_q1 = percentile(sample2, 0.25)
        s1_q3 = percentile(sample1, 0.75)
        s2_q3 = percentile(sample2, 0.75)
        q1_diff = (s1_q1 - s2_q1) / float(s2_q1)
        q3_diff = (s1_q3 - s2_q3) / float(s2_q3)

        mad1 = median_abs_dev(sample1)
        mad2 = median_abs_dev(sample2)
        # FIXME: handle division by zero
        mad_diff = (mad1 - mad2) / float(mad2)

        if self.args.verbose:
            stdev1 = statistics.stdev(sample1)
            stdev2 = statistics.stdev(sample2)
            stdev_diff = (stdev1 - stdev2) / float(stdev2)

# test if the first value is an outlier
        values = sample1[1:] + sample2
        q1 = percentile(values, 0.25)
        q3 = percentile(values, 0.75)
        iqr = q3 - q1
        outlier_max = (q3 + 1.5 * iqr)
        # only check maximum, not minimum
        outlier = not(first_value <= outlier_max)

        mean1 = statistics.mean(sample1)
        mean2 = statistics.mean(sample2)
        mean_diff = (mean1 - mean2) / float(mean2)

        s1_q1 = percentile(sample1, 0.25)
        s2_q1 = percentile(sample2, 0.25)
        s1_q3 = percentile(sample1, 0.75)
        s2_q3 = percentile(sample2, 0.75)
        q1_diff = (s1_q1 - s2_q1) / float(s2_q1)
        q3_diff = (s1_q3 - s2_q3) / float(s2_q3)

        mad1 = median_abs_dev(sample1)
        mad2 = median_abs_dev(sample2)
        # FIXME: handle division by zero
        mad_diff = (mad1 - mad2) / float(mad2)

        if self.args.verbose:
            stdev1 = statistics.stdev(sample1)
            stdev2 = statistics.stdev(sample2)
            stdev_diff = (stdev1 - stdev2) / float(stdev2)

            sample1_str = format_values(unit, (s1_q1, mean1, s1_q3, stdev1, mad1))
            sample2_str = format_values(unit, (s2_q1, mean2, s2_q3, stdev2, mad2))

# test if the first value is an outlier
        values = sample1[1:] + sample2
        q1 = percentile(values, 0.25)
        q3 = percentile(values, 0.75)
        iqr = q3 - q1
        outlier_max = (q3 + 1.5 * iqr)
        # only check maximum, not minimum
        outlier = not(first_value <= outlier_max)

        mean1 = statistics.mean(sample1)
        mean2 = statistics.mean(sample2)
        mean_diff = (mean1 - mean2) / float(mean2)

        s1_q1 = percentile(sample1, 0.25)
        s2_q1 = percentile(sample2, 0.25)
        s1_q3 = percentile(sample1, 0.75)
        s2_q3 = percentile(sample2, 0.75)
        q1_diff = (s1_q1 - s2_q1) / float(s2_q1)
        q3_diff = (s1_q3 - s2_q3) / float(s2_q3)

        mad1 = median_abs_dev(sample1)
        mad2 = median_abs_dev(sample2)
        # FIXME: handle division by zero
        mad_diff = (mad1 - mad2) / float(mad2)

        if self.args.verbose:
            stdev1 = statistics.stdev(sample1)
            stdev2 = statistics.stdev(sample2)
            stdev_diff = (stdev1 - stdev2) / float(stdev2)

            sample1_str = format_values(unit, (s1_q1, mean1, s1_q3, stdev1, mad1))