test/performance: port to NAC3

This commit is contained in:
Sébastien Bourdeauducq 2024-08-30 13:21:27 +08:00
parent a7b0450593
commit 016691cee3
1 changed files with 72 additions and 107 deletions

View File

@ -1,44 +1,45 @@
import os
import time
import unittest
from typing import Literal
import numpy
from numpy import int32, float64
from numpy import int32, float64, ndarray
from artiq.experiment import *
from artiq.test.hardware_testbench import ExperimentCase
from artiq.coredevice.core import Core
bool_list_large = [True] * (1 << 20)
bool_list_small = [True] * (1 << 10)
# large: 1MB payload
# small: 1KB payload
bytes_large = b"\x00" * (1 << 20)
bytes_small = b"\x00" * (1 << 10)
list_large = [123] * (1 << 18)
list_small = [123] * (1 << 8)
array_large = numpy.array(list_large, int32)
array_small = numpy.array(list_small, int32)
byte_list_large = [True] * (1 << 20)
byte_list_small = [True] * (1 << 10)
received_bytes = 0
time_start = 0
time_end = 0
@nac3
class _Transfer(EnvExperiment):
core: KernelInvariant[Core]
count: KernelInvariant[int32]
h2d: Kernel[list[float]]
d2h: Kernel[list[float]]
def build(self):
self.setattr_device("core")
self.count = 10
self.h2d = [0.0] * self.count
self.d2h = [0.0] * self.count
@rpc
def get_bytes(self, large: bool) -> bytes:
if large:
return bytes_large
else:
return bytes_small
@rpc
def get_list(self, large: bool) -> list[int32]:
if large:
@ -47,14 +48,14 @@ class _Transfer(EnvExperiment):
return list_small
@rpc
def get_byte_list(self, large: bool) -> list[bool]:
def get_bool_list(self, large: bool) -> list[bool]:
if large:
return byte_list_large
return bool_list_large
else:
return byte_list_small
return bool_list_small
@rpc
def get_array(self, large: bool) -> numpy.ndarray: # NAC3TODO: [int32]
def get_array(self, large: bool) -> ndarray[int32, Literal[1]]:
if large:
return array_large
else:
@ -65,11 +66,19 @@ class _Transfer(EnvExperiment):
return string_list
@rpc
def sink(self, data):
def sink_bool_list(self, data: list[bool]):
pass
@rpc(flags={"async"})
def sink_async(self, data):
@rpc
def sink_list(self, data: list[int32]):
pass
@rpc
def sink_array(self, data: ndarray[int32, Literal[1]]):
pass
@rpc # NAC3TODO (flags={"async"})
def sink_async(self, data: list[int32]):
global received_bytes, time_start, time_end
if received_bytes == 0:
time_start = time.time()
@ -81,73 +90,48 @@ class _Transfer(EnvExperiment):
def get_async_throughput(self) -> float:
return 128.0 / (time_end - time_start)
@kernel
def test_bytes(self, large):
def inner():
def test_bool_list(self, large: bool):
for i in range(self.count):
t0 = self.core.get_rtio_counter_mu()
data = self.get_bytes(large)
data = self.get_bool_list(large)
t1 = self.core.get_rtio_counter_mu()
self.sink(data)
self.sink_bool_list(data)
t2 = self.core.get_rtio_counter_mu()
self.h2d[i] = self.core.mu_to_seconds(t1 - t0)
self.d2h[i] = self.core.mu_to_seconds(t2 - t1)
for i in range(self.count):
inner()
return (self.h2d, self.d2h)
@kernel
def test_byte_list(self, large):
def inner():
t0 = self.core.get_rtio_counter_mu()
data = self.get_byte_list(large)
t1 = self.core.get_rtio_counter_mu()
self.sink(data)
t2 = self.core.get_rtio_counter_mu()
self.h2d[i] = self.core.mu_to_seconds(t1 - t0)
self.d2h[i] = self.core.mu_to_seconds(t2 - t1)
def test_list(self, large: bool):
for i in range(self.count):
inner()
return (self.h2d, self.d2h)
@kernel
def test_list(self, large):
def inner():
t0 = self.core.get_rtio_counter_mu()
data = self.get_list(large)
t1 = self.core.get_rtio_counter_mu()
self.sink(data)
self.sink_list(data)
t2 = self.core.get_rtio_counter_mu()
self.h2d[i] = self.core.mu_to_seconds(t1 - t0)
self.d2h[i] = self.core.mu_to_seconds(t2 - t1)
for i in range(self.count):
inner()
return (self.h2d, self.d2h)
@kernel
def test_array(self, large):
def inner():
def test_array(self, large: bool):
for i in range(self.count):
t0 = self.core.get_rtio_counter_mu()
data = self.get_array(large)
t1 = self.core.get_rtio_counter_mu()
self.sink(data)
self.sink_array(data)
t2 = self.core.get_rtio_counter_mu()
self.h2d[i] = self.core.mu_to_seconds(t1 - t0)
self.d2h[i] = self.core.mu_to_seconds(t2 - t1)
for i in range(self.count):
inner()
return (self.h2d, self.d2h)
@kernel
def test_async(self):
data = self.get_bytes(True)
def test_async(self) -> float:
data = self.get_list(True)
for _ in range(128):
self.sink_async(data)
return self.get_async_throughput()
class TransferTest(ExperimentCase):
@classmethod
def setUpClass(self):
@ -169,59 +153,35 @@ class TransferTest(ExperimentCase):
print("| {} | {:>12.2f} | {:>12.2f} |".format(
pad(v[0]), v[1], v[2]))
def test_bytes_large(self):
def test_bool_list_large(self):
exp = self.create(_Transfer)
results = exp.test_bytes(True)
host_to_device = (1 << 20) / numpy.array(results[0], float64)
device_to_host = (1 << 20) / numpy.array(results[1], float64)
exp.test_bool_list(True)
host_to_device = (1 << 20) / numpy.array(exp.h2d, float64)
device_to_host = (1 << 20) / numpy.array(exp.d2h, float64)
host_to_device /= 1024*1024
device_to_host /= 1024*1024
self.results.append(["Bytes (1MB) H2D", host_to_device.mean(),
self.results.append(["Bool List (1MB) H2D", host_to_device.mean(),
host_to_device.std()])
self.results.append(["Bytes (1MB) D2H", device_to_host.mean(),
self.results.append(["Bool List (1MB) D2H", device_to_host.mean(),
device_to_host.std()])
def test_bytes_small(self):
def test_bool_list_small(self):
exp = self.create(_Transfer)
results = exp.test_bytes(False)
host_to_device = (1 << 10) / numpy.array(results[0], float64)
device_to_host = (1 << 10) / numpy.array(results[1], float64)
exp.test_bool_list(False)
host_to_device = (1 << 10) / numpy.array(exp.h2d, float64)
device_to_host = (1 << 10) / numpy.array(exp.d2h, float64)
host_to_device /= 1024*1024
device_to_host /= 1024*1024
self.results.append(["Bytes (1KB) H2D", host_to_device.mean(),
self.results.append(["Bool List (1KB) H2D", host_to_device.mean(),
host_to_device.std()])
self.results.append(["Bytes (1KB) D2H", device_to_host.mean(),
device_to_host.std()])
def test_byte_list_large(self):
exp = self.create(_Transfer)
results = exp.test_byte_list(True)
host_to_device = (1 << 20) / numpy.array(results[0], float64)
device_to_host = (1 << 20) / numpy.array(results[1], float64)
host_to_device /= 1024*1024
device_to_host /= 1024*1024
self.results.append(["Bytes List (1MB) H2D", host_to_device.mean(),
host_to_device.std()])
self.results.append(["Bytes List (1MB) D2H", device_to_host.mean(),
device_to_host.std()])
def test_byte_list_small(self):
exp = self.create(_Transfer)
results = exp.test_byte_list(False)
host_to_device = (1 << 10) / numpy.array(results[0], float64)
device_to_host = (1 << 10) / numpy.array(results[1], float64)
host_to_device /= 1024*1024
device_to_host /= 1024*1024
self.results.append(["Bytes List (1KB) H2D", host_to_device.mean(),
host_to_device.std()])
self.results.append(["Bytes List (1KB) D2H", device_to_host.mean(),
self.results.append(["Bool List (1KB) D2H", device_to_host.mean(),
device_to_host.std()])
def test_list_large(self):
exp = self.create(_Transfer)
results = exp.test_list(True)
host_to_device = (1 << 20) / numpy.array(results[0], float64)
device_to_host = (1 << 20) / numpy.array(results[1], float64)
exp.test_list(True)
host_to_device = (1 << 20) / numpy.array(exp.h2d, float64)
device_to_host = (1 << 20) / numpy.array(exp.d2h, float64)
host_to_device /= 1024*1024
device_to_host /= 1024*1024
self.results.append(["I32 List (1MB) H2D", host_to_device.mean(),
@ -231,9 +191,9 @@ class TransferTest(ExperimentCase):
def test_list_small(self):
exp = self.create(_Transfer)
results = exp.test_list(False)
host_to_device = (1 << 10) / numpy.array(results[0], float64)
device_to_host = (1 << 10) / numpy.array(results[1], float64)
exp.test_list(False)
host_to_device = (1 << 10) / numpy.array(exp.h2d, float64)
device_to_host = (1 << 10) / numpy.array(exp.d2h, float64)
host_to_device /= 1024*1024
device_to_host /= 1024*1024
self.results.append(["I32 List (1KB) H2D", host_to_device.mean(),
@ -243,9 +203,9 @@ class TransferTest(ExperimentCase):
def test_array_large(self):
exp = self.create(_Transfer)
results = exp.test_array(True)
host_to_device = (1 << 20) / numpy.array(results[0], float64)
device_to_host = (1 << 20) / numpy.array(results[1], float64)
exp.test_array(True)
host_to_device = (1 << 20) / numpy.array(exp.h2d, float64)
device_to_host = (1 << 20) / numpy.array(exp.d2h, float64)
host_to_device /= 1024*1024
device_to_host /= 1024*1024
self.results.append(["I32 Array (1MB) H2D", host_to_device.mean(),
@ -255,9 +215,9 @@ class TransferTest(ExperimentCase):
def test_array_small(self):
exp = self.create(_Transfer)
results = exp.test_array(False)
host_to_device = (1 << 10) / numpy.array(results[0], float64)
device_to_host = (1 << 10) / numpy.array(results[1], float64)
exp.test_array(False)
host_to_device = (1 << 10) / numpy.array(exp.h2d, float64)
device_to_host = (1 << 10) / numpy.array(exp.d2h, float64)
host_to_device /= 1024*1024
device_to_host /= 1024*1024
self.results.append(["I32 Array (1KB) H2D", host_to_device.mean(),
@ -265,12 +225,17 @@ class TransferTest(ExperimentCase):
self.results.append(["I32 Array (1KB) D2H", device_to_host.mean(),
device_to_host.std()])
@unittest.skip("NAC3TODO https://git.m-labs.hk/M-Labs/nac3/issues/182")
def test_async_throughput(self):
exp = self.create(_Transfer)
results = exp.test_async()
print("Async throughput: {:>6.2f}MiB/s".format(results))
@nac3
class _KernelOverhead(EnvExperiment):
core: KernelInvariant[Core]
def build(self):
self.setattr_device("core")