Files
llama-swap/internal/perf/monitor_windows.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

122 lines
2.7 KiB
Go

package perf
import (
"bufio"
"context"
"fmt"
"os/exec"
"strings"
"time"
"github.com/mostlygeek/llama-swap/internal/logmon"
"github.com/shirou/gopsutil/v4/cpu"
"github.com/shirou/gopsutil/v4/mem"
"github.com/shirou/gopsutil/v4/net"
)
func getGpuStats(ctx context.Context, every time.Duration, logger *logmon.Monitor) (chan []GpuStat, error) {
if ch, err := tryNvidiaSmiWindows(ctx, every, logger); err == nil {
logger.Info("using nvidia-smi for GPU monitoring")
return ch, nil
} else {
logger.Debugf("nvidia-smi: %s", err.Error())
}
if ch, err := tryD3DKMT(ctx, every, logger); err == nil {
logger.Info("using D3DKMT for GPU monitoring")
return ch, nil
} else {
logger.Debugf("D3DKMT: %s", err.Error())
}
return nil, ErrNoGpuTool
}
// tryNvidiaSmiWindows starts nvidia-smi in loop mode on Windows and returns
// a channel receiving GPU stat snapshots. Returns ErrNoGpuTool if nvidia-smi
// is not available.
func tryNvidiaSmiWindows(ctx context.Context, every time.Duration, logger *logmon.Monitor) (chan []GpuStat, error) {
if _, err := exec.LookPath("nvidia-smi"); err != nil {
return nil, ErrNoGpuTool
}
sec := int(every.Seconds())
if sec < 1 {
sec = 1
}
cmd := exec.CommandContext(ctx, "nvidia-smi",
"--query-gpu=index,name,uuid,temperature.gpu,utilization.gpu,memory.used,memory.total,fan.speed,power.draw",
"--format=csv,noheader,nounits",
"--loop", fmt.Sprintf("%d", sec),
)
stdout, err := cmd.StdoutPipe()
if err != nil {
return nil, fmt.Errorf("nvidia-smi stdout pipe failed: %w", err)
}
if err := cmd.Start(); err != nil {
return nil, fmt.Errorf("nvidia-smi start failed: %w", err)
}
ch := make(chan []GpuStat, 1)
go func() {
defer close(ch)
scanner := bufio.NewScanner(stdout)
for scanner.Scan() {
line := strings.TrimSpace(scanner.Text())
if line == "" {
continue
}
stat := ParseNvidiaSmiLine(line)
if stat != nil {
select {
case ch <- []GpuStat{*stat}:
default:
}
}
}
cmd.Wait()
}()
return ch, nil
}
func readSysStats() (SysStat, error) {
cpuPcts, err := cpu.Percent(0, true)
if err != nil {
return SysStat{}, err
}
vmStat, err := mem.VirtualMemory()
if err != nil {
return SysStat{}, err
}
const toMB = 1024 * 1024
netIO := make([]NetIOStat, 0)
if ioCounters, err := net.IOCounters(true); err == nil {
for _, ioc := range ioCounters {
netIO = append(netIO, NetIOStat{
Name: ioc.Name,
BytesRecv: ioc.BytesRecv,
BytesSent: ioc.BytesSent,
})
}
}
return SysStat{
Timestamp: time.Now(),
CpuUtilPerCore: cpuPcts,
MemTotalMB: int(vmStat.Total / toMB),
MemUsedMB: int(vmStat.Used / toMB),
MemFreeMB: int(vmStat.Free / toMB),
NetIO: netIO,
}, nil
}