WIP: checkpoint current tree

This commit is contained in:
2026-04-05 12:05:00 +03:00
parent a14ec8631c
commit 20abff7f90
15 changed files with 539 additions and 59 deletions

View File

@@ -232,6 +232,54 @@ func (h *handler) handleAPISATRun(target string) http.HandlerFunc {
}
}
func (h *handler) handleAPIBenchmarkNvidiaRun(w http.ResponseWriter, r *http.Request) {
if h.opts.App == nil {
writeError(w, http.StatusServiceUnavailable, "app not configured")
return
}
var body struct {
Profile string `json:"profile"`
SizeMB int `json:"size_mb"`
GPUIndices []int `json:"gpu_indices"`
ExcludeGPUIndices []int `json:"exclude_gpu_indices"`
RunNCCL *bool `json:"run_nccl"`
DisplayName string `json:"display_name"`
}
if r.Body != nil {
if err := json.NewDecoder(r.Body).Decode(&body); err != nil && !errors.Is(err, io.EOF) {
writeError(w, http.StatusBadRequest, "invalid request body")
return
}
}
runNCCL := true
if body.RunNCCL != nil {
runNCCL = *body.RunNCCL
}
t := &Task{
ID: newJobID("benchmark-nvidia"),
Name: taskDisplayName("nvidia-benchmark", "", ""),
Target: "nvidia-benchmark",
Priority: 15,
Status: TaskPending,
CreatedAt: time.Now(),
params: taskParams{
GPUIndices: body.GPUIndices,
ExcludeGPUIndices: body.ExcludeGPUIndices,
SizeMB: body.SizeMB,
BenchmarkProfile: body.Profile,
RunNCCL: runNCCL,
DisplayName: body.DisplayName,
},
}
if strings.TrimSpace(body.DisplayName) != "" {
t.Name = body.DisplayName
}
globalQueue.enqueue(t)
writeJSON(w, map[string]string{"task_id": t.ID, "job_id": t.ID})
}
func (h *handler) handleAPISATStream(w http.ResponseWriter, r *http.Request) {
id := r.URL.Query().Get("job_id")
if id == "" {
@@ -491,6 +539,22 @@ func (h *handler) handleAPIExportUSBBundle(w http.ResponseWriter, r *http.Reques
// ── GPU presence ──────────────────────────────────────────────────────────────
func (h *handler) handleAPIGNVIDIAGPUs(w http.ResponseWriter, _ *http.Request) {
if h.opts.App == nil {
writeError(w, http.StatusServiceUnavailable, "app not configured")
return
}
gpus, err := h.opts.App.ListNvidiaGPUs()
if err != nil {
writeError(w, http.StatusInternalServerError, err.Error())
return
}
if gpus == nil {
gpus = []platform.NvidiaGPU{}
}
writeJSON(w, gpus)
}
func (h *handler) handleAPIGPUPresence(w http.ResponseWriter, r *http.Request) {
if h.opts.App == nil {
writeError(w, http.StatusServiceUnavailable, "app not configured")
@@ -516,8 +580,10 @@ func (h *handler) handleAPIGPUTools(w http.ResponseWriter, _ *http.Request) {
_, amdErr := os.Stat("/dev/kfd")
nvidiaUp := nvidiaErr == nil
amdUp := amdErr == nil
_, dcgmErr := exec.LookPath("dcgmi")
writeJSON(w, []toolEntry{
{ID: "bee-gpu-burn", Available: nvidiaUp, Vendor: "nvidia"},
{ID: "dcgm", Available: nvidiaUp && dcgmErr == nil, Vendor: "nvidia"},
{ID: "john", Available: nvidiaUp, Vendor: "nvidia"},
{ID: "nccl", Available: nvidiaUp, Vendor: "nvidia"},
{ID: "rvs", Available: amdUp, Vendor: "amd"},