Optimize a C Function

You're a firmware engineer at Arm. This function runs on every sample batch coming off a sensor peripheral, byte-swapping and packing each 16-bit reading before handing the buffer to the application layer. It works, but it is eating far more of the CPU budget than a function this simple should.

typedef struct {
    uint16_t reading;
    uint8_t sensor_id;
} SensorSample;

void process_samples(SensorSample *samples, int count, uint8_t *output) {
    for (int i = 0; i < get_sample_count(samples); i++) {
        SensorSample *temp = malloc(sizeof(SensorSample));
        temp->reading = samples[i].reading;
        temp->sensor_id = samples[i].sensor_id;

        uint16_t swapped = 0;
        for (int bit = 0; bit < 16; bit++) {
            if (bit < 8) {
                swapped = swapped | (((temp->reading >> bit) & 1) << (bit + 8));
            } else {
                swapped = swapped | (((temp->reading >> bit) & 1) << (bit - 8));
            }
        }

        output[i * 3] = (uint8_t)(swapped >> 8);
        output[i * 3 + 1] = (uint8_t)(swapped & 0xFF);
        output[i * 3 + 2] = temp->sensor_id;

        free(temp);
    }
}

How can the following function in C be optimized? Explain what you would change and why, for a function that runs on every sample batch on a resource-constrained device.