Monitor: show GPU temperature and fan speed

Extends the existing per-tick nvidia-smi query with temperature.gpu and
fan.speed (no extra call); N/A values (cards without fan telemetry) are
simply omitted from the GPU line.
This commit is contained in:
mram
2026-09-22 16:56:01 +02:00
parent 2040b30c94
commit e58ff33912
5 changed files with 104 additions and 34 deletions
+25
View File
@@ -116,6 +116,31 @@ func TestParseGPUEngineInstance(t *testing.T) {
}
}
func TestParseGPUStats(t *testing.T) {
st, err := parseGPUStats("4300, 16384, 55, 42\n")
if err != nil {
t.Fatal(err)
}
if st.UsedMB != 4300 || st.TotalMB != 16384 || st.TempC != 55 || st.FanPct != 42 {
t.Errorf("got %+v", st)
}
// Cards that don't expose temperature/fan report N/A.
st, err = parseGPUStats("1024, 16384, N/A, N/A")
if err != nil {
t.Fatal(err)
}
if st.TempC != -1 || st.FanPct != -1 {
t.Errorf("got %+v, want -1 for N/A fields", st)
}
for _, bad := range []string{"", "1, 2", "x, 16384, 55, 42", "1024, x, 55, 42", "1, 2, 3, 4, 5"} {
if _, err := parseGPUStats(bad); err == nil {
t.Errorf("%q parsed, want failure", bad)
}
}
}
func TestProcessesLive(t *testing.T) {
if runtime.GOOS != "windows" && runtime.GOOS != "linux" {
t.Skip("no process listing on this platform")