BOOT

1234LONG

ENTRY64

TensorOS v0.1.0 "Neuron" booting...

[BOOT] Kernel loaded at 0x200000, size 94208 KB

[BOOT] SSE2 enabled --- SIMD tensor operations available

[BOOT] TSC calibrated: 4009 MHz

[PHASE 1] Hardware initialization

[EXCEPT] 32 CPU exception handlers installed (vectors 0-31)

[EXCEPT] Page fault, GPF, double fault: full register dump + stack trace

[WDOG] Timer ISR installed (1000 Hz tick counter)

[WDOG] Watchdog timer available (set/kick/disable)

[IDT] Loaded with 256 entries (32 exception + 224 IRQ)

[TSS] Initialized with IST stacks for NMI/#DF/#MC

[OK] IDT initialized

[CPU] Vendor: AuthenticAMD

[OK] 1 CPU(s) detected

[PCI] Enumerating devices...

[PCI] 0:0.0 vendor=8086 device=1237

[PCI] 0:1.0 vendor=8086 device=7000

[PCI] 0:2.0 vendor=1234 device=1111

[PCI] 0:3.0 vendor=1af4 device=1000

[PCI] 0:4.0 vendor=1af4 device=1001

[OK] PCI bus enumerated

[OK] 1 GPU(s) detected and initialized

GPU 0: Unknown GPU (0 MB VRAM, 0 CUDA cores)

[PIC] Remapped IRQs 0x20-0x2F

[TIMER] PIT configured at 1000 Hz

[OK] Interrupts configured (1000 Hz timer) - enabling after boot

[PHASE 2] Subsystem initialization

[MM] Multiboot mmap: 8191 MB usable, 9216 MB address space

[DBG] [MM] Initialized: 9216 MB total, 129 MB free

[DBG] [MM] Tensor heap: 4992 MB @ 0x100000000, Model cache: 2976 MB @ 0x5e00000

[OK] Tensor memory manager initialized

Tensor heap: 4992 MB, Model cache: 2976 MB

[DBG] [VMM] W^X: starting rodata

[DBG] [VMM] W^X: rodata done (26 pages), starting data

[DBG] [VMM] W^X: data done (17 pages), starting bss (0x2b9000 - 0x5c58488)

[DBG] [VMM] BSS lead: 0x2b9000-0x400000 (327 4K pages)

[DBG] [VMM] BSS huge: 0x400000-0x5c00000 (44 2M pages)

[DBG] [VMM] BSS trail: 0x5c00000-0x5c59000 (89 4K pages)

[DBG] [VMM] W^X: bss done (22944 pages), starting text

[VMM] W^X enforced: .text=142 RX, .rodata=26 R, .data=17 RW, .bss=22944 RW

[DBG] [SCHED] Initialized with 1 GPUs, 0 TPUs, policy=THROUGHPUT

[OK] Tensor scheduler initialized

[TFS] No disk found, using RAM-only mode

[OK] TensorFS mounted (RAM-only)

[OK] Tensor IPC initialized

[DBG] [GIT] Native git subsystem initialized

[OK] Native git subsystem initialized

[NET] Network init (stub)

[OK] Network stack initialized

[PHASE 3] AI runtime initialization

[DBG] [JIT] Pseudocode JIT compiler initialized

[OK] Pseudocode JIT compiler ready

[DBG] [JIT] x86_64 JIT compiler initialized

[OK] x86_64 native JIT backend initialized

[DBG] [ENGINE] Tensor execution engine initialized

[OK] Tensor execution engine ready (SSE2 SIMD)

[OK] CPU tensor math verified (matmul, relu, softmax, dot)

[OK] Model sandbox initialized

[DBG] [PKG] Model package manager initialized with 2 registries

[OK] Model package manager ready

[DBG] [VIRT] Capabilities: AMD-V NPT

[OK] Near-zero-cost virtualization layer ready

[PHASE 4] Launching userland

[MONITOR] Tensor monitor daemon initialized

[OK] Tensor monitor daemon started

[DEPLOY] Deployment service initialized

[DEPLOY] Deployment daemon initialized

[OK] Model deployment service started

[OK] Userland services initialized

[PHASE 5] CPU Feature Detection

[CPU] AVX state saving enabled (XCR0=0x7)

[CPU] Vendor: AuthenticAMD

[CPU] ISA: SSE SSE2 SSE3 SSSE3 SSE4.1 SSE4.2 AVX AVX2 FMA AES-NI POPCNT BMI1 BMI2

[CPU] GEMM dispatch: AVX2+FMA 8-wide (256-bit)

[PHASE 6] Production Self-Test Suite

=== TensorOS Self-Test Suite ===

[PASS] memset

[PASS] memcpy

[PASS] strlen

[PASS] strcmp

[PASS] strcpy

[PASS] exp

[PASS] sqrt

[PASS] tanh

[PASS] heap_alloc

[PASS] arena

[PASS] gemm_2x2

[PASS] softmax

[PASS] relu

[PASS] watchdog

[PASS] kstrlcpy

[PASS] sha256

------------------------------------------------------------

[TEST] ALL 16 TESTS PASSED

------------------------------------------------------------

[BRANIAC] Running self-test...

[BRANIAC] Network initialized: 4 layers [4->8->4->2]

[BRANIAC] JIT compiling 3 column kernels...

[BRANIAC] JIT: column 0 [4->8] compiled

[BRANIAC] JIT: column 1 [8->4] compiled

[BRANIAC] JIT: column 2 [4->2] compiled

[BRANIAC] JIT compilation complete (6 kernels)

[BRANIAC] Self-test: initial_loss=1.0972 final_loss=0.9055 PASS

[PHASE 7] SMP Multi-Core Bootstrap

=== SMP Multi-Core Demo ===

[SMP] LAPIC base: 0xfee00000, BSP APIC ID: 0

[SMP] ACPI MADT: 1 CPUs enumerated

[SMP] Detected 1 logical CPUs

[SMP] 1 CPUs, BSP APIC ID 0

CPU 0: APIC 0, state=IDLE

[SMP] Multi-core infrastructure ready

[SMP] Single-core mode (APs can be started with real hardware)

[SYSCALL] GDT expanded: user code=0x30, user data=0x28

[SYSCALL] SYSCALL/SYSRET configured (LSTAR=0x207ff0)

[SYSCALL] Kernel/user boundary active: ring-0 ↔ ring-3

[PHASE 8] Storage Driver

[VBLK] Found virtio-blk at PCI 0:4.0

[VBLK] I/O base: 0xc000

[VBLK] Capacity: 282835 sectors (138 MB)

[VBLK] Virtio-blk initialized successfully

[VBLK] PCI 0:4.0, I/O 0xc000, IRQ 11

[VBLK] Capacity: 138 MB (282835 sectors)

[VBLK] Stats: 0 reads (0 KB), 0 writes (0 KB)

[PHASE 9] Network Stack

[VNET] Found virtio-net at PCI 0:3.0

[VNET] I/O base: 0xc080

[VNET] IRQ: 11

[VNET] Device features: 0x79bf8064

[VNET] MAC: 52:54:00:12:34:56

[VNET] RX queue size: 256

[VNET] TX queue size: 256

[VNET] Virtio-net initialized successfully

[NET] Stack configured: 10.0.2.15/255.255.255.0 gw 10.0.2.2

[NET] Registered UDP handler on port 8080

==========================================================

TensorOS LLM API Server - OpenAI Compatible

==========================================================

Listening: http://10.0.2.15:8080

Model: (none)

Endpoints:

GET /v1/models List models

POST /v1/completions Text completion

POST /v1/chat/completions Chat completion

GET /health Health check

Quick start (from any device on the network):

# List models

curl http://10.0.2.15:8080/v1/models

# Chat completion

curl -X POST http://10.0.2.15:8080/v1/chat/completions \

-H 'Content-Type: application/json' \

-d '{"messages":[{"role":"user","content":"Hello"}]}'

# Python (OpenAI SDK)

from openai import OpenAI

client = OpenAI(

base_url="http://10.0.2.15:8080/v1",

api_key="tensoros")

r = client.chat.completions.create(

model="default",

messages=[{"role":"user","content":"Hello"}])

print(r.choices[0].message.content)

==========================================================

[TLS] TLS 1.3 initialized (ChaCha20-Poly1305 + X25519 + Ed25519)

[HTTPS] TLS 1.3 server on port 8443

[HTTPS] https://10.0.2.15:8443/v1/models

[OK] Network ready: 10.0.2.15 HTTP:8080 HTTPS:8443

[E1000] No Intel E1000 NIC found

[PHASE 10] AI Model Loader

Loading 138 MB model from disk...

[LLM] Loading model from disk: 138 MB

[LLM] ... 64 MB loaded

[LLM] ... 128 MB loaded

[LLM] Loaded 138 MB in 483 ms (292788 KB/s)

[LLM] Parsing GGUF...

[GGUF] Version 3, 272 tensors, 37 KV pairs

[LLM] Model: Smollm2 135M 8k Lc100K Mix1 Ep2 (llama)

[LLM] 30 layers, 576-dim, 49152 vocab, 9 heads (3 KV)

[LLM] Note: output.weight tied to token_embd

[LLM] Tensor mapping complete: 30 layers

[LLM] Vocab: 49152 tokens, BOS=1, EOS=2

[LLM] Allocating 360 MB scratch arena

[LLM] Model loaded successfully! Ready for inference.

[OK] Smollm2 135M 8k Lc100K Mix1 Ep2 (30 layers, 28311552 params)

============================================================

TensorOS v0.1.0 "Neuron" -- Boot Complete

============================================================

Arch: x86_64 (AVX2+FMA 256-bit)

CPUs: 1 RAM: 9216 MB

Storage: virtio-blk

LLM: Smollm2 135M 8k Lc100K Mix1 Ep2 (ready)

============================================================

[READY] TensorOS is operational. 1 CPUs, 9216 MB RAM

[LLM-DEMO] loaded=1

[LLM-DEMO] Running automatic inference benchmark...

--- Prompt: "What is an operating system?" ---

[JIT] AVX2 available — using C+AVX2 for dot/axpy/vadd

[JIT] 1 CPUs for parallel GEMV

[JIT] AVX2 GEMV 576x576 compiled (1090 bytes)

[JIT] AVX2 GEMV 192x576 compiled (1090 bytes)

[JIT] AVX2 GEMV 1536x576 compiled (1090 bytes)

[JIT] AVX2 GEMV 576x1536 compiled (1090 bytes)

[JIT] AVX2 GEMV 49152x576 compiled (1090 bytes)

[263 tok, 1544 ms/tok, prefill 20467 ms, 1 cpus]

Response (263 tokens):

1. A programming language.

2. A collection of pre-defined rules and algorithms.

3. A set of programs that use these rules and algorithms to perform a specific task.

4. A system that manages the resources of a computer system, such as memory, storage, and input/output devices.

5. A way to give instructions to other computers or software programs that use these rules and algorithms.

A typical operating system includes:

* A kernel: the core part of the operating system that manages the system's hardware and manages the flow of data between the CPU and other parts of the system.

* A bootloader: the first instruction in the boot process that loads the operating system from a disk or other device into the CPU's memory.

* A file system: a way to organize and manage files and directories in the system.

* A file system hierarchy: a way to organize and structure the files and directories in the system.

* A user interface: a way to provide a user with a way to interact with the system.

An operating system provides a framework for the system to run, manage, and interact with other software programs. It provides the basic building blocks for a computer system, allowing other programs to run and interact with it.

[PERF] 263 tokens in 427201 ms (0.6 tok/s)

--- Prompt: "Write a haiku about silicon." ---

[87 tok, 1482 ms/tok, prefill 20502 ms, 1 cpus]

Response (87 tokens):

The silicon in your house is a silent companion.

It dances in the light, a silent whisper.

Silicon's secrets, hidden deep within.

The word 'silicon' is a symbol of stability and strength.

It stands as a reminder of the precious metal.

It's a symbol of the beauty that lies within.

The haiku captures the essence of silicon's role in our world.

[PERF] 87 tokens in 150097 ms (0.6 tok/s)

--- Prompt: "Explain how neural networks learn in 2 sentences." ---