feat(metrics): persist history in sqlite and add AMD memory validate tests

This commit is contained in:
2026-03-29 12:28:06 +03:00
parent 98f0cf0d52
commit e15bcc91c5
10 changed files with 539 additions and 29 deletions

View File

@@ -136,6 +136,54 @@ func (s *System) RunAMDAcceptancePack(ctx context.Context, baseDir string, logFu
}, logFunc)
}
// RunAMDMemIntegrityPack runs the official RVS MEM module as a validate-style memory integrity test.
func (s *System) RunAMDMemIntegrityPack(ctx context.Context, baseDir string, logFunc func(string)) (string, error) {
if err := ensureAMDRuntimeReady(); err != nil {
return "", err
}
cfgFile := "/tmp/bee-amd-mem.conf"
cfg := `actions:
- name: mem_integrity
device: all
module: mem
parallel: true
duration: 60000
copy_matrix: false
target_stress: 90
matrix_size: 8640
`
_ = os.WriteFile(cfgFile, []byte(cfg), 0644)
return runAcceptancePackCtx(ctx, baseDir, "gpu-amd-mem", []satJob{
{name: "01-rocm-smi.log", cmd: []string{"rocm-smi"}},
{name: "02-rvs-mem.log", cmd: []string{"rvs", "-c", cfgFile}},
{name: "03-rocm-smi-after.log", cmd: []string{"rocm-smi", "--showtemp", "--showpower", "--showmemuse", "--csv"}},
}, logFunc)
}
// RunAMDMemBandwidthPack runs AMD's memory/interconnect bandwidth-oriented tools.
func (s *System) RunAMDMemBandwidthPack(ctx context.Context, baseDir string, logFunc func(string)) (string, error) {
if err := ensureAMDRuntimeReady(); err != nil {
return "", err
}
cfgFile := "/tmp/bee-amd-babel.conf"
cfg := `actions:
- name: babel_mem_bw
device: all
module: babel
parallel: true
copy_matrix: true
target_stress: 90
matrix_size: 134217728
`
_ = os.WriteFile(cfgFile, []byte(cfg), 0644)
return runAcceptancePackCtx(ctx, baseDir, "gpu-amd-bandwidth", []satJob{
{name: "01-rocm-smi.log", cmd: []string{"rocm-smi"}},
{name: "02-rocm-bandwidth-test.log", cmd: []string{"rocm-bandwidth-test"}},
{name: "03-rvs-babel.log", cmd: []string{"rvs", "-c", cfgFile}},
{name: "04-rocm-smi-after.log", cmd: []string{"rocm-smi", "--showtemp", "--showpower", "--showmemuse", "--csv"}},
}, logFunc)
}
// RunAMDStressPack runs an AMD GPU burn-in pack.
// Missing tools are reported as UNSUPPORTED, consistent with the existing SAT pattern.
func (s *System) RunAMDStressPack(ctx context.Context, baseDir string, durationSec int, logFunc func(string)) (string, error) {
@@ -161,7 +209,7 @@ func amdStressRVSConfig(seconds int) string {
module: gst
parallel: true
duration: %d
copy_matrix: true
copy_matrix: false
target_stress: 90
matrix_size_a: 8640
matrix_size_b: 8640