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

@@ -30,22 +30,23 @@ const (
// taskNames maps target → human-readable name for validate (SAT) runs.
var taskNames = map[string]string{
"nvidia": "NVIDIA SAT",
"nvidia-stress": "NVIDIA GPU Stress",
"memory": "Memory SAT",
"storage": "Storage SAT",
"cpu": "CPU SAT",
"amd": "AMD GPU SAT",
"amd-mem": "AMD GPU MEM Integrity",
"amd-bandwidth": "AMD GPU MEM Bandwidth",
"amd-stress": "AMD GPU Burn-in",
"memory-stress": "Memory Burn-in",
"sat-stress": "SAT Stress (stressapptest)",
"platform-stress": "Platform Thermal Cycling",
"audit": "Audit",
"support-bundle": "Support Bundle",
"install": "Install to Disk",
"install-to-ram": "Install to RAM",
"nvidia": "NVIDIA SAT",
"nvidia-benchmark": "NVIDIA Benchmark",
"nvidia-stress": "NVIDIA GPU Stress",
"memory": "Memory SAT",
"storage": "Storage SAT",
"cpu": "CPU SAT",
"amd": "AMD GPU SAT",
"amd-mem": "AMD GPU MEM Integrity",
"amd-bandwidth": "AMD GPU MEM Bandwidth",
"amd-stress": "AMD GPU Burn-in",
"memory-stress": "Memory Burn-in",
"sat-stress": "SAT Stress (stressapptest)",
"platform-stress": "Platform Thermal Cycling",
"audit": "Audit",
"support-bundle": "Support Bundle",
"install": "Install to Disk",
"install-to-ram": "Install to RAM",
}
// burnNames maps target → human-readable name when a burn profile is set.
@@ -108,8 +109,11 @@ type taskParams struct {
DiagLevel int `json:"diag_level,omitempty"`
GPUIndices []int `json:"gpu_indices,omitempty"`
ExcludeGPUIndices []int `json:"exclude_gpu_indices,omitempty"`
SizeMB int `json:"size_mb,omitempty"`
Loader string `json:"loader,omitempty"`
BurnProfile string `json:"burn_profile,omitempty"`
BenchmarkProfile string `json:"benchmark_profile,omitempty"`
RunNCCL bool `json:"run_nccl,omitempty"`
DisplayName string `json:"display_name,omitempty"`
Device string `json:"device,omitempty"` // for install
PlatformComponents []string `json:"platform_components,omitempty"`
@@ -547,6 +551,18 @@ func (q *taskQueue) runTask(t *Task, j *jobState, ctx context.Context) {
} else {
archive, err = a.RunNvidiaAcceptancePack("", j.append)
}
case "nvidia-benchmark":
if a == nil {
err = fmt.Errorf("app not configured")
break
}
archive, err = a.RunNvidiaBenchmarkCtx(ctx, "", platform.NvidiaBenchmarkOptions{
Profile: t.params.BenchmarkProfile,
SizeMB: t.params.SizeMB,
GPUIndices: t.params.GPUIndices,
ExcludeGPUIndices: t.params.ExcludeGPUIndices,
RunNCCL: t.params.RunNCCL,
}, j.append)
case "nvidia-stress":
if a == nil {
err = fmt.Errorf("app not configured")