sysinfo and compute Commands
sysinfo - display hardware profile
Syntax
sysinfo
No parameters. Always performs a fresh hardware detection (force=True). The previously saved compute_mode is re-applied after detection.
Full output - CPU + RAM only (no GPU)
exf> sysinfo
┌─────────────────────────────┐
│ CPU │
├──────────────┬──────────────┤
│ Property │ Value │
├──────────────┼──────────────┤
│ Model │ Intel Core i7-11800H @ 2.30GHz │
│ Architecture │ x86_64 │
│ Cores │ 8 │
│ Threads │ 16 │
│ Frequency │ 2300 MHz │
└──────────────┴──────────────┘
┌─────────────────────────────┐
│ Memory (RAM) │
├──────────────┬──────────────┤
│ Property │ Value │
├──────────────┼──────────────┤
│ Total │ 32,768 MB │
│ Available │ 18,432 MB │
└──────────────┴──────────────┘
[WARN] No GPU detected on this system
Compute mode: auto -> cpu | Best backend: cpu
Full output - with NVIDIA GPU
exf> sysinfo
┌─────────────────────────────────────────────────┐
│ CPU │
├──────────────────┬──────────────────────────────┤
│ Property │ Value │
├──────────────────┼──────────────────────────────┤
│ Model │ AMD Ryzen 9 7950X @ 4.50GHz │
│ Architecture │ x86_64 │
│ Cores │ 16 │
│ Threads │ 32 │
│ Frequency │ 4500 MHz │
└──────────────────┴──────────────────────────────┘
┌───────────────────────────┐
│ Memory (RAM) │
├───────────────┬───────────┤
│ Property │ Value │
├───────────────┼───────────┤
│ Total │ 65,536 MB │
│ Available │ 52,480 MB │
└───────────────┴───────────┘
┌──────────────────────────────────────────────────────────────────────────┐
│ GPU Devices │
├───┬───────────────────────────┬────────┬──────────┬────────┬─────┬───────┤
│ # │ Name │ Vendor │ VRAM │Backend │Driver│Compute│
├───┼───────────────────────────┼────────┼──────────┼────────┼─────┼───────┤
│ 0 │ NVIDIA GeForce RTX 4090 │ nvidia │ 24,576 MB│ cuda │555.85│ 8.9 │
└───┴───────────────────────────┴────────┴──────────┴────────┴─────┴───────┘
Compute mode: auto -> hybrid | Best backend: cuda
Full output - with AMD GPU (ROCm)
exf> sysinfo
... (CPU and RAM tables as above) ...
┌──────────────────────────────────────────────────────────────────────────┐
│ GPU Devices │
├───┬───────────────────────────┬────────┬──────────┬────────┬──────┬──────┤
│ # │ Name │ Vendor │ VRAM │Backend │Driver│Compute│
├───┼───────────────────────────┼────────┼──────────┼────────┼──────┼──────┤
│ 0 │ AMD Radeon RX 7900 XTX │ amd │ 24,576 MB│ rocm │ 6.0.2│gfx1100│
└───┴───────────────────────────┴────────┴──────────┴────────┴──────┴──────┘
Compute mode: auto -> hybrid | Best backend: rocm
Full output - multi-GPU system
exf> sysinfo
... (CPU and RAM tables) ...
┌──────────────────────────────────────────────────────────────────────────────┐
│ GPU Devices │
├───┬─────────────────────────────┬────────┬──────────┬────────┬──────┬────────┤
│ # │ Name │ Vendor │ VRAM │Backend │Driver│Compute │
├───┼─────────────────────────────┼────────┼──────────┼────────┼──────┼────────┤
│ 0 │ NVIDIA GeForce RTX 4070 │ nvidia │ 8,192 MB│ cuda │555.85│ 8.9 │
│ 1 │ NVIDIA GeForce RTX 3060 │ nvidia │ 12,288 MB│ cuda │555.85│ 8.6 │
└───┴─────────────────────────────┴────────┴──────────┴────────┴──────┴────────┘
Compute mode: auto -> hybrid | Best backend: cuda
What sysinfo detects
| Field | Source | Notes |
|---|---|---|
| CPU Model | /proc/cpuinfo (Linux), wmic cpu (Windows), sysctl (macOS) | |
| CPU Architecture | platform.machine() | e.g. x86_64, aarch64, armv7l |
| Cores | os.cpu_count() / psutil | Physical cores |
| Threads | os.cpu_count() | Logical threads (including HT) |
| Frequency (MHz) | /proc/cpuinfo cpu MHz, psutil | May show 0 on throttled systems |
| RAM Total | psutil / /proc/meminfo | In MB |
| RAM Available | psutil available | Excludes buffers/cache |
| GPU Name | nvidia-smi, rocm-smi, clinfo, wmic path Win32_VideoController | |
| GPU VRAM | nvidia-smi --query-gpu=memory.total | |
| GPU Backend | CUDA / ROCm / OpenCL auto-detected | |
| GPU Driver | nvidia-smi, amdgpu version | N/A if unavailable |
| Compute Capability | CUDA sm_XX, ROCm gfxXXXX | N/A for other backends |
Startup system line
The hardware one-liner shown in the startup banner is generated by HWProfile.one_liner():
[system] Intel Core i7-11800H (16T) | 32768MB RAM | No GPU detected | compute_mode: auto->cpu
[system] AMD Ryzen 9 7950X (32T) | 65536MB RAM | NVIDIA GeForce RTX 4090 24576MB | compute_mode: auto->hybrid
compute - set compute mode
Syntax
compute <cpu|gpu|hybrid|auto>
Parameters
| Value | Meaning |
|---|---|
cpu | Force CPU-only for all operations. All GPU-accelerated modules fall back to CPU paths. |
gpu | Force GPU for all operations. Falls back to CPU silently if no GPU is present. |
hybrid | Use GPU for ML-accelerated modules (banner fingerprinting, credential ranking, fuzzing entropy estimation) and CPU for all standard exploits. |
auto | Automatically resolve at runtime: hybrid if GPU is present, cpu if not. (Default) |
The selected mode is persisted to the EmbedXPL config file (~/.exf_config.json) and restored on the next launch.
Tab completion
exf> compute <TAB>
auto cpu gpu hybrid
compute cpu
exf> compute cpu
[+] compute_mode => cpu
compute gpu (GPU present)
exf> compute gpu
[+] compute_mode => gpu
compute gpu (no GPU - automatic fallback)
exf> compute gpu
[WARN] No GPU detected -- falling back to compute_mode=cpu
The mode is saved as cpu even though gpu was requested.
compute hybrid
exf> compute hybrid
[+] compute_mode => hybrid
compute hybrid (no GPU - automatic fallback)
exf> compute hybrid
[WARN] No GPU detected -- falling back to compute_mode=cpu
compute auto
exf> compute auto
[+] compute_mode => auto
auto resolves to: cpu
exf> compute auto
[+] compute_mode => auto
auto resolves to: hybrid
Invalid mode
exf> compute quantum
[-] Invalid compute mode 'quantum'. Choose from: cpu, gpu, hybrid, auto
sysinfo after changing compute mode
sysinfo re-reads the saved config and shows the current effective mode:
exf> compute gpu
[+] compute_mode => gpu
exf> sysinfo
...
Compute mode: gpu | Best backend: cuda
Which modules use GPU compute?
| Module category | GPU use |
|---|---|
| Standard exploits | Never (CPU only) |
| Standard scanners | Never (CPU only) |
| Standard creds | Never (CPU only) |
core/ml/banner_fingerprint.py | Yes - device fingerprinting via transformer model |
core/ml/advisor.py | Yes - module priority ranking using ML scoring |
core/gpu/backend.py | Yes - GPU backend dispatcher for ML modules |
| Fuzzing entropy estimator | Yes (hybrid: GPU for scoring, CPU for delivery) |
If no GPU is present, all ML modules fall back to CPU paths automatically - they are slower but functionally equivalent.