PIO Performance Issue

[GuillaumeLeclerc]

Hello,

I was trying to use the PIOs to read 8 bit serial data and speed seem terrible, any idea what is going on:

import time
import rp2
from time import ticks_ms, ticks_diff
from rp2 import PIO
import micropython
from machine import Pin

@rp2.asm_pio(autopull=True, sideset_init=PIO.OUT_LOW)
def blink():
    set(x, 1)
    wrap_target()
    in_(x, 8).side(1)
    push(iffull, block).side(0)
    wrap()

# Instantiate a state machine with the blink program, at 2000Hz, with set bound to Pin(25) (LED on the Pico board)
sm = rp2.StateMachine(0, blink, freq=16_000_000, set_base=Pin(25), in_base=Pin(0))
sm.active(1)

@micropython.viper
def read(sm, buffer):
    for i in range(128):
        sm.get(buffer)
# Run the state machine for 3 seconds.  The LED should blink.
buffer = memoryview(bytearray(1024))

start = ticks_ms()
read(sm, buffer)
end = ticks_ms()
print(ticks_diff(end, start) / 1000)

Frequency of the PIO is 16MHz, it reads 8 bits every other cycle so 8 * 16 /2 = 64Mb/s = 8MB/s
I try to fill a 1024bytes 128 times so I expect it to be give or take 125ms. When I run the code it takes 1376 so roughly 10x what I expect.

What is going on here ?