reanimator/bee
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity

Fix NVIDIA persistence mode and add benchmark results table

Browse Source
This commit is contained in:
Mikhail Chusavitin
2026-04-06 10:47:07 +03:00
parent 6e94216f3b
commit fc5c100a29
6 changed files with 296 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
audit/internal/platform/nvidia_stress.go
View File

@@ -16,12 +16,12 @@ func (s *System) RunNvidiaStressPack(ctx context.Context, baseDir string, opts N
return "", err
}
return runAcceptancePackCtx(ctx, baseDir, nvidiaStressArchivePrefix(opts.Loader), []satJob{
{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
{name: "02-nvidia-smi-list.log", cmd: []string{"nvidia-smi", "-L"}},
return runAcceptancePackCtx(ctx, baseDir, nvidiaStressArchivePrefix(opts.Loader), withNvidiaPersistenceMode(
satJob{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
satJob{name: "02-nvidia-smi-list.log", cmd: []string{"nvidia-smi", "-L"}},
job,
{name: "04-nvidia-smi-after.log", cmd: []string{"nvidia-smi", "--query-gpu=index,name,temperature.gpu,power.draw,utilization.gpu,memory.used,memory.total", "--format=csv,noheader,nounits"}},
}, logFunc)
satJob{name: "04-nvidia-smi-after.log", cmd: []string{"nvidia-smi", "--query-gpu=index,name,temperature.gpu,power.draw,utilization.gpu,memory.used,memory.total", "--format=csv,noheader,nounits"}},
), logFunc)
}
func nvidiaStressArchivePrefix(loader string) string {

96
audit/internal/platform/sat.go
View File

@@ -278,13 +278,13 @@ func (s *System) RunNCCLTests(ctx context.Context, baseDir string, logFunc func(
if gpuCount < 1 {
gpuCount = 1
}
return runAcceptancePackCtx(ctx, baseDir, "nccl-tests", []satJob{
{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
{name: "02-all-reduce-perf.log", cmd: []string{
return runAcceptancePackCtx(ctx, baseDir, "nccl-tests", withNvidiaPersistenceMode(
satJob{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
satJob{name: "02-all-reduce-perf.log", cmd: []string{
"all_reduce_perf", "-b", "512M", "-e", "4G", "-f", "2",
"-g", strconv.Itoa(gpuCount), "--iters", "20",
}},
}, logFunc)
), logFunc)
}
func (s *System) RunNvidiaOfficialComputePack(ctx context.Context, baseDir string, durationSec int, gpuIndices []int, logFunc func(string)) (string, error) {
@@ -296,18 +296,18 @@ func (s *System) RunNvidiaOfficialComputePack(ctx context.Context, baseDir strin
if err != nil {
return "", err
}
return runAcceptancePackCtx(ctx, baseDir, "gpu-nvidia-compute", []satJob{
{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
{name: "02-dcgmi-version.log", cmd: []string{"dcgmi", "-v"}},
{
return runAcceptancePackCtx(ctx, baseDir, "gpu-nvidia-compute", withNvidiaPersistenceMode(
satJob{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
satJob{name: "02-dcgmi-version.log", cmd: []string{"dcgmi", "-v"}},
satJob{
name: "03-dcgmproftester.log",
cmd: profCmd,
env: nvidiaVisibleDevicesEnv(selected),
collectGPU: true,
gpuIndices: selected,
},
{name: "04-nvidia-smi-after.log", cmd: []string{"nvidia-smi", "--query-gpu=index,name,temperature.gpu,power.draw,utilization.gpu,memory.used,memory.total", "--format=csv,noheader,nounits"}},
}, logFunc)
satJob{name: "04-nvidia-smi-after.log", cmd: []string{"nvidia-smi", "--query-gpu=index,name,temperature.gpu,power.draw,utilization.gpu,memory.used,memory.total", "--format=csv,noheader,nounits"}},
), logFunc)
}
func (s *System) RunNvidiaTargetedPowerPack(ctx context.Context, baseDir string, durationSec int, gpuIndices []int, logFunc func(string)) (string, error) {
@@ -315,16 +315,16 @@ func (s *System) RunNvidiaTargetedPowerPack(ctx context.Context, baseDir string,
if err != nil {
return "", err
}
return runAcceptancePackCtx(ctx, baseDir, "gpu-nvidia-targeted-power", []satJob{
{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
{
return runAcceptancePackCtx(ctx, baseDir, "gpu-nvidia-targeted-power", withNvidiaPersistenceMode(
satJob{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
satJob{
name: "02-dcgmi-targeted-power.log",
cmd: nvidiaDCGMNamedDiagCommand("targeted_power", normalizeNvidiaBurnDuration(durationSec), selected),
collectGPU: true,
gpuIndices: selected,
},
{name: "03-nvidia-smi-after.log", cmd: []string{"nvidia-smi", "--query-gpu=index,name,temperature.gpu,power.draw,utilization.gpu,memory.used,memory.total", "--format=csv,noheader,nounits"}},
}, logFunc)
satJob{name: "03-nvidia-smi-after.log", cmd: []string{"nvidia-smi", "--query-gpu=index,name,temperature.gpu,power.draw,utilization.gpu,memory.used,memory.total", "--format=csv,noheader,nounits"}},
), logFunc)
}
func (s *System) RunNvidiaPulseTestPack(ctx context.Context, baseDir string, durationSec int, gpuIndices []int, logFunc func(string)) (string, error) {
@@ -332,16 +332,16 @@ func (s *System) RunNvidiaPulseTestPack(ctx context.Context, baseDir string, dur
if err != nil {
return "", err
}
return runAcceptancePackCtx(ctx, baseDir, "gpu-nvidia-pulse", []satJob{
{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
{
return runAcceptancePackCtx(ctx, baseDir, "gpu-nvidia-pulse", withNvidiaPersistenceMode(
satJob{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
satJob{
name: "02-dcgmi-pulse-test.log",
cmd: nvidiaDCGMNamedDiagCommand("pulse_test", normalizeNvidiaBurnDuration(durationSec), selected),
collectGPU: true,
gpuIndices: selected,
},
{name: "03-nvidia-smi-after.log", cmd: []string{"nvidia-smi", "--query-gpu=index,name,temperature.gpu,power.draw,utilization.gpu,memory.used,memory.total", "--format=csv,noheader,nounits"}},
}, logFunc)
satJob{name: "03-nvidia-smi-after.log", cmd: []string{"nvidia-smi", "--query-gpu=index,name,temperature.gpu,power.draw,utilization.gpu,memory.used,memory.total", "--format=csv,noheader,nounits"}},
), logFunc)
}
func (s *System) RunNvidiaBandwidthPack(ctx context.Context, baseDir string, gpuIndices []int, logFunc func(string)) (string, error) {
@@ -349,16 +349,16 @@ func (s *System) RunNvidiaBandwidthPack(ctx context.Context, baseDir string, gpu
if err != nil {
return "", err
}
return runAcceptancePackCtx(ctx, baseDir, "gpu-nvidia-bandwidth", []satJob{
{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
{
return runAcceptancePackCtx(ctx, baseDir, "gpu-nvidia-bandwidth", withNvidiaPersistenceMode(
satJob{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
satJob{
name: "02-dcgmi-nvbandwidth.log",
cmd: nvidiaDCGMNamedDiagCommand("nvbandwidth", 0, selected),
collectGPU: true,
gpuIndices: selected,
},
{name: "03-nvidia-smi-after.log", cmd: []string{"nvidia-smi", "--query-gpu=index,name,temperature.gpu,power.draw,utilization.gpu,memory.used,memory.total", "--format=csv,noheader,nounits"}},
}, logFunc)
satJob{name: "03-nvidia-smi-after.log", cmd: []string{"nvidia-smi", "--query-gpu=index,name,temperature.gpu,power.draw,utilization.gpu,memory.used,memory.total", "--format=csv,noheader,nounits"}},
), logFunc)
}
func (s *System) RunNvidiaAcceptancePack(baseDir string, logFunc func(string)) (string, error) {
@@ -389,16 +389,16 @@ func (s *System) RunNvidiaTargetedStressValidatePack(ctx context.Context, baseDi
logFunc(fmt.Sprintf("pre-flight: killed stale worker pid=%d name=%s", p.PID, p.Name))
}
}
return runAcceptancePackCtx(ctx, baseDir, "gpu-nvidia-targeted-stress", []satJob{
{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
{
return runAcceptancePackCtx(ctx, baseDir, "gpu-nvidia-targeted-stress", withNvidiaPersistenceMode(
satJob{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
satJob{
name: "02-dcgmi-targeted-stress.log",
cmd: nvidiaDCGMNamedDiagCommand("targeted_stress", normalizeNvidiaBurnDuration(durationSec), selected),
collectGPU: true,
gpuIndices: selected,
},
{name: "03-nvidia-smi-after.log", cmd: []string{"nvidia-smi", "--query-gpu=index,name,temperature.gpu,power.draw,utilization.gpu,memory.used,memory.total", "--format=csv,noheader,nounits"}},
}, logFunc)
satJob{name: "03-nvidia-smi-after.log", cmd: []string{"nvidia-smi", "--query-gpu=index,name,temperature.gpu,power.draw,utilization.gpu,memory.used,memory.total", "--format=csv,noheader,nounits"}},
), logFunc)
}
func resolveDCGMGPUIndices(gpuIndices []int) ([]int, error) {
@@ -568,14 +568,24 @@ type satStats struct {
Unsupported int
}
func withNvidiaPersistenceMode(jobs ...satJob) []satJob {
out := make([]satJob, 0, len(jobs)+1)
out = append(out, satJob{
name: "00-nvidia-smi-persistence-mode.log",
cmd: []string{"nvidia-smi", "-pm", "1"},
})
out = append(out, jobs...)
return out
}
func nvidiaSATJobs() []satJob {
return []satJob{
{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
{name: "02-dmidecode-baseboard.log", cmd: []string{"dmidecode", "-t", "baseboard"}},
{name: "03-dmidecode-system.log", cmd: []string{"dmidecode", "-t", "system"}},
{name: "04-nvidia-bug-report.log", cmd: []string{"nvidia-bug-report.sh", "--output-file", "{{run_dir}}/nvidia-bug-report.log"}},
{name: "05-bee-gpu-burn.log", cmd: []string{"bee-gpu-burn", "--seconds", "5", "--size-mb", "64"}},
}
return withNvidiaPersistenceMode(
satJob{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
satJob{name: "02-dmidecode-baseboard.log", cmd: []string{"dmidecode", "-t", "baseboard"}},
satJob{name: "03-dmidecode-system.log", cmd: []string{"dmidecode", "-t", "system"}},
satJob{name: "04-nvidia-bug-report.log", cmd: []string{"nvidia-bug-report.sh", "--output-file", "{{run_dir}}/nvidia-bug-report.log"}},
satJob{name: "05-bee-gpu-burn.log", cmd: []string{"bee-gpu-burn", "--seconds", "5", "--size-mb", "64"}},
)
}
func nvidiaDCGMJobs(diagLevel int, gpuIndices []int) []satJob {
@@ -590,12 +600,12 @@ func nvidiaDCGMJobs(diagLevel int, gpuIndices []int) []satJob {
}
diagArgs = append(diagArgs, "-i", strings.Join(ids, ","))
}
return []satJob{
{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
{name: "02-dmidecode-baseboard.log", cmd: []string{"dmidecode", "-t", "baseboard"}},
{name: "03-dmidecode-system.log", cmd: []string{"dmidecode", "-t", "system"}},
{name: "04-dcgmi-diag.log", cmd: diagArgs},
}
return withNvidiaPersistenceMode(
satJob{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
satJob{name: "02-dmidecode-baseboard.log", cmd: []string{"dmidecode", "-t", "baseboard"}},
satJob{name: "03-dmidecode-system.log", cmd: []string{"dmidecode", "-t", "system"}},
satJob{name: "04-dcgmi-diag.log", cmd: diagArgs},
)
}
func nvidiaDCGMNamedDiagCommand(name string, durationSec int, gpuIndices []int) []string {

29
audit/internal/platform/sat_test.go
View File

@@ -28,13 +28,19 @@ func TestRunNvidiaAcceptancePackIncludesGPUStress(t *testing.T) {
jobs := nvidiaSATJobs()
if len(jobs) != 5 {
t.Fatalf("jobs=%d want 5", len(jobs))
if len(jobs) != 6 {
t.Fatalf("jobs=%d want 6", len(jobs))
}
if got := jobs[4].cmd[0]; got != "bee-gpu-burn" {
if got := jobs[0].cmd[0]; got != "nvidia-smi" {
t.Fatalf("preflight command=%q want nvidia-smi", got)
}
if got := strings.Join(jobs[0].cmd, " "); got != "nvidia-smi -pm 1" {
t.Fatalf("preflight=%q want %q", got, "nvidia-smi -pm 1")
}
if got := jobs[5].cmd[0]; got != "bee-gpu-burn" {
t.Fatalf("gpu stress command=%q want bee-gpu-burn", got)
}
if got := jobs[3].cmd[1]; got != "--output-file" {
if got := jobs[4].cmd[1]; got != "--output-file" {
t.Fatalf("bug report flag=%q want --output-file", got)
}
}
@@ -82,7 +88,7 @@ func TestAMDStressJobsIncludeBandwidthAndGST(t *testing.T) {
func TestNvidiaSATJobsUseBuiltinBurnDefaults(t *testing.T) {
jobs := nvidiaSATJobs()
got := jobs[4].cmd
got := jobs[5].cmd
want := []string{"bee-gpu-burn", "--seconds", "5", "--size-mb", "64"}
if len(got) != len(want) {
t.Fatalf("cmd len=%d want %d", len(got), len(want))
@@ -94,6 +100,19 @@ func TestNvidiaSATJobsUseBuiltinBurnDefaults(t *testing.T) {
}
}
func TestNvidiaDCGMJobsEnablePersistenceModeBeforeDiag(t *testing.T) {
jobs := nvidiaDCGMJobs(3, []int{2, 0})
if len(jobs) != 5 {
t.Fatalf("jobs=%d want 5", len(jobs))
}
if got := strings.Join(jobs[0].cmd, " "); got != "nvidia-smi -pm 1" {
t.Fatalf("preflight=%q want %q", got, "nvidia-smi -pm 1")
}
if got := strings.Join(jobs[4].cmd, " "); got != "dcgmi diag -r 3 -i 2,0" {
t.Fatalf("diag=%q want %q", got, "dcgmi diag -r 3 -i 2,0")
}
}
func TestBuildNvidiaStressJobUsesSelectedLoaderAndDevices(t *testing.T) {
t.Parallel()