Files
llama-swap/internal/perf/d3dkmt_windows_test.go
T
George 0ab214d1c8 perf: add vendor-agnostic GPU monitoring for Windows (experimental) (#779)
Add GPU monitoring support for AMD and Intel GPUs on Windows using
D3DKMT (DirectX) and PDH performance counters.

- Add PDH-based GPU utilization via \GPU Engine(*)\Utilization
Percentage counter, summing all engine types per adapter (3D, Compute,
Copy, Video).
- Add D3DKMT bindings for adapter enumeration, memory segments, and
adapter perf data.
- Use PDH as primary utilization source (works on all vendors), with
D3DKMT RunningTime as fallback for systems without PDH counters.
- Prefer nvidia-smi when available, fall back to D3DKMT + PDH for
AMD/Intel.
- Backend priority: nvidia-smi -> D3DKMT + PDH -> ErrNoGpuTool.

Verified on AMD 7900XTX GPU with llama.cpp Vulkan & ROCm backend: GPU
utilization correctly shows ~99% during inference, ~0-2% when idle.

---

LLM disclosure: GLM 5.1 & Kimi K2.6 have been used extensively during
exploration and coding to the point that the LLM's wrote over 3/4 of the
code, and I have done additional verification myself.
As such, it should be considered experimental.
Additional verification is needed.

I have tested it on my 7900XTX system with Windows 11, and it works
correctly, but as I only have this one rig, I cannot verify it
everywhere.
2026-06-16 21:49:09 -07:00

99 lines
2.6 KiB
Go

//go:build windows
package perf
import (
"testing"
"github.com/stretchr/testify/assert"
)
func TestD3dkmtNodeUtil_FullLoad(t *testing.T) {
prev := nodeRunningTimes{Global: 1000, System: 10000}
cur := nodeRunningTimes{Global: 5000, System: 14000}
got := d3dkmtNodeUtil(prev, cur, 100000)
assert.Equal(t, 100.0, got)
}
func TestD3dkmtNodeUtil_PartialUtil(t *testing.T) {
prev := nodeRunningTimes{Global: 1000, System: 10000}
cur := nodeRunningTimes{Global: 3000, System: 14000}
got := d3dkmtNodeUtil(prev, cur, 100000)
assert.Equal(t, 50.0, got)
}
func TestD3dkmtNodeUtil_Identical(t *testing.T) {
prev := nodeRunningTimes{Global: 10000, System: 10000}
cur := nodeRunningTimes{Global: 20000, System: 20000}
got := d3dkmtNodeUtil(prev, cur, 100000)
assert.Equal(t, 100.0, got)
}
func TestD3dkmtNodeUtil_CounterWrap(t *testing.T) {
prev := nodeRunningTimes{Global: 9000, System: 10000}
cur := nodeRunningTimes{Global: 1000, System: 10000}
got := d3dkmtNodeUtil(prev, cur, 100000)
assert.Equal(t, -1.0, got)
}
func TestD3dkmtNodeUtil_SystemWrap(t *testing.T) {
prev := nodeRunningTimes{Global: 1000, System: 9000}
cur := nodeRunningTimes{Global: 5000, System: 1000}
got := d3dkmtNodeUtil(prev, cur, 100000)
assert.Equal(t, -1.0, got)
}
func TestD3dkmtNodeUtil_ZeroDelta(t *testing.T) {
prev := nodeRunningTimes{Global: 1000, System: 10000}
cur := nodeRunningTimes{Global: 1000, System: 10000}
got := d3dkmtNodeUtil(prev, cur, 100000)
assert.Equal(t, 0.0, got)
}
func TestD3dkmtNodeUtil_ElapsedFallback(t *testing.T) {
prev := nodeRunningTimes{Global: 1000, System: 10000}
cur := nodeRunningTimes{Global: 6000, System: 10000}
got := d3dkmtNodeUtil(prev, cur, 50000)
assert.InDelta(t, 10.0, got, 0.01)
}
func TestD3dkmtFanPct_Normal(t *testing.T) {
assert.Equal(t, 50.0, d3dkmtFanPct(1500, 3000))
}
func TestD3dkmtFanPct_MaxFan(t *testing.T) {
assert.Equal(t, 100.0, d3dkmtFanPct(3000, 3000))
}
func TestD3dkmtFanPct_OverMaxClamped(t *testing.T) {
assert.Equal(t, 100.0, d3dkmtFanPct(4000, 3000))
}
func TestD3dkmtFanPct_ZeroMaxFan(t *testing.T) {
assert.Equal(t, 0.0, d3dkmtFanPct(1500, 0))
}
func TestD3dkmtFanPct_ZeroFanRPM(t *testing.T) {
assert.Equal(t, 0.0, d3dkmtFanPct(0, 3000))
}
func TestD3dkmtFanPct_BothZero(t *testing.T) {
assert.Equal(t, 0.0, d3dkmtFanPct(0, 0))
}
func TestD3dkmtPowerW(t *testing.T) {
assert.Equal(t, 250.0, d3dkmtPowerW(2500))
}
func TestD3dkmtPowerW_Zero(t *testing.T) {
assert.Equal(t, 0.0, d3dkmtPowerW(0))
}
func TestD3dkmtTempC(t *testing.T) {
assert.Equal(t, 65, d3dkmtTempC(650))
}
func TestD3dkmtTempC_Zero(t *testing.T) {
assert.Equal(t, 0, d3dkmtTempC(0))
}