Fix intel gpu stats (#4581)

* Fix return code for timeout call

* Rewrit intel gpu logic

* Fix test and output
This commit is contained in:
Nicolas Mowen 2022-12-03 09:28:01 -07:00 committed by GitHub
parent 94a2b6bb88
commit 007fa75294
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 30 additions and 18 deletions

View File

@ -35,11 +35,11 @@ class TestGpuStats(unittest.TestCase):
@patch("subprocess.run") @patch("subprocess.run")
def test_intel_gpu_stats(self, sp): def test_intel_gpu_stats(self, sp):
process = MagicMock() process = MagicMock()
process.returncode = 0 process.returncode = 124
process.stdout = self.intel_results process.stdout = self.intel_results
sp.return_value = process sp.return_value = process
intel_stats = get_intel_gpu_stats() intel_stats = get_intel_gpu_stats()
assert intel_stats == { assert intel_stats == {
"gpu": "10.73 %", "gpu": "1.34 %",
"mem": "- %", "mem": "- %",
} }

View File

@ -797,7 +797,7 @@ def get_intel_gpu_stats() -> dict[str, str]:
"""Get stats using intel_gpu_top.""" """Get stats using intel_gpu_top."""
intel_gpu_top_command = [ intel_gpu_top_command = [
"timeout", "timeout",
"0.1s", "0.5s",
"intel_gpu_top", "intel_gpu_top",
"-J", "-J",
"-o", "-o",
@ -812,27 +812,39 @@ def get_intel_gpu_stats() -> dict[str, str]:
capture_output=True, capture_output=True,
) )
if p.returncode != 0: # timeout has a non-zero returncode when timeout is reached
if p.returncode != 124:
logger.error(p.stderr) logger.error(p.stderr)
return None return None
else: else:
readings = json.loads(f"[{p.stdout}]") reading = "".join(p.stdout.split())
results: dict[str, str] = {} results: dict[str, str] = {}
for reading in readings: # render is used for qsv
if reading.get("engines", {}).get("Video/0", {}).get( render = []
"busy", 0 for result in re.findall('"Render/3D/0":{[a-z":\d.,%]+}', reading):
) or reading.get("engines", {}).get("Video/1", {}).get("busy", 0): packet = json.loads(result[14:])
gpu_usage = round( single = packet.get("busy", 0.0)
float(reading.get("engines", {}).get("Video/0", {}).get("busy", 0)) render.append(float(single))
+ float(
reading.get("engines", {}).get("Video/1", {}).get("busy", 0)
),
2,
)
results["gpu"] = f"{gpu_usage} %"
break
if render:
render_avg = sum(render) / len(render)
else:
render_avg = 1
# video is used for vaapi
video = []
for result in re.findall('"Video/\d":{[a-z":\d.,%]+}', reading):
packet = json.loads(result[10:])
single = packet.get("busy", 0.0)
video.append(float(single))
if video:
video_avg = sum(video) / len(video)
else:
video_avg = 1
results["gpu"] = f"{round((video_avg + render_avg) / 2, 2)} %"
results["mem"] = "- %" results["mem"] = "- %"
return results return results