Scruff AI
419 lines
16 KiB
Plaintext
419 lines
16 KiB
Plaintext
/* ============================================================================
|
|
* METRICS ONLY - Collect Data, No Boundaries
|
|
* March 7 Hard-Print Compliance
|
|
* ============================================================================ */
|
|
|
|
#include <cuda_runtime.h>
|
|
#include <nvml.h>
|
|
#include <cstdio>
|
|
#include <cstdlib>
|
|
#include <cstdint>
|
|
#include <cmath>
|
|
#include <chrono>
|
|
#include <vector>
|
|
|
|
#ifndef M_PI
|
|
#define M_PI 3.14159265358979323846
|
|
#endif
|
|
|
|
#define NX 1024
|
|
#define NY 1024
|
|
#define NN (NX * NY)
|
|
#define Q 9
|
|
#define BLOCK 256
|
|
#define GBLK(n) (((n) + BLOCK - 1) / BLOCK)
|
|
|
|
#define TOTAL_STEPS 1000000 // ~3 minutes
|
|
#define STEPS_PER_BATCH 500
|
|
#define SAMPLE_INTERVAL 50000
|
|
|
|
#define OMEGA 1.0f
|
|
|
|
/* ---- Vorticity Threshold ------------------------------------------------- */
|
|
#define VORTICITY_THRESHOLD 0.000001f
|
|
#define PERSISTENCE_STEPS 275000
|
|
#define MAX_GUARDIANS 200
|
|
|
|
/* ---- Guardian Structure ------------------------------------------------- */
|
|
typedef struct {
|
|
int id;
|
|
float position[2];
|
|
float velocity[2];
|
|
float mass;
|
|
float latent_energy;
|
|
uint64_t persistence_age;
|
|
uint64_t born_step;
|
|
int active;
|
|
} Guardian;
|
|
|
|
Guardian guardians[MAX_GUARDIANS];
|
|
int n_guardians = 0;
|
|
|
|
/* ---- Vortex Seed Tracking ----------------------------------------------- */
|
|
typedef struct {
|
|
float x, y;
|
|
float vorticity;
|
|
uint64_t first_seen;
|
|
uint64_t last_seen;
|
|
int active;
|
|
} VortexSeed;
|
|
|
|
VortexSeed vortex_seeds[10000];
|
|
int n_seeds = 0;
|
|
|
|
/* ---- D2Q9 --------------------------------------------------------------- */
|
|
__constant__ int d_ex[Q] = { 0, 1, 0,-1, 0, 1,-1,-1, 1 };
|
|
__constant__ int d_ey[Q] = { 0, 0, 1, 0,-1, 1, 1,-1,-1 };
|
|
__constant__ float d_w[Q] = { 4.f/9, 1.f/9, 1.f/9, 1.f/9, 1.f/9,
|
|
1.f/36,1.f/36,1.f/36,1.f/36 };
|
|
|
|
/* ======================================================================== */
|
|
/* K E R N E L S */
|
|
/* ======================================================================== */
|
|
|
|
/* ---- LBM collide & stream ---------------------------------------------- */
|
|
__global__ void lbm_collide_stream(const float* __restrict__ f_src,
|
|
float* __restrict__ f_dst,
|
|
float* __restrict__ rho,
|
|
float* __restrict__ ux,
|
|
float* __restrict__ uy,
|
|
float omega, int nx, int ny) {
|
|
const int idx = blockIdx.x * blockDim.x + threadIdx.x;
|
|
const int N = nx * ny;
|
|
if (idx >= N) return;
|
|
const int x = idx % nx, y = idx / nx;
|
|
|
|
float fl[Q];
|
|
for (int i = 0; i < Q; i++) {
|
|
int sx = (x - d_ex[i] + nx) % nx;
|
|
int sy = (y - d_ey[i] + ny) % ny;
|
|
fl[i] = f_src[i * N + sy * nx + sx];
|
|
}
|
|
|
|
float rho_val = 0.f, ux_val = 0.f, uy_val = 0.f;
|
|
for (int i = 0; i < Q; i++) {
|
|
rho_val += fl[i];
|
|
ux_val += (float)d_ex[i] * fl[i];
|
|
uy_val += (float)d_ey[i] * fl[i];
|
|
}
|
|
float inv = 1.f / fmaxf(rho_val, 1e-10f);
|
|
ux_val *= inv; uy_val *= inv;
|
|
rho[idx] = rho_val; ux[idx] = ux_val; uy[idx] = uy_val;
|
|
|
|
const float u2 = ux_val * ux_val + uy_val * uy_val;
|
|
for (int i = 0; i < Q; i++) {
|
|
float eu = (float)d_ex[i] * ux_val + (float)d_ey[i] * uy_val;
|
|
float feq = d_w[i] * rho_val * (1.f + 3.f*eu + 4.5f*eu*eu - 1.5f*u2);
|
|
f_dst[i * N + idx] = fl[i] - omega * (fl[i] - feq);
|
|
}
|
|
}
|
|
|
|
/* ---- Finite Difference Vorticity --------------------------------------- */
|
|
__device__ float calculate_vorticity(int x, int y, int nx, int ny,
|
|
float* v_x, float* v_y) {
|
|
if (x <= 0 || x >= nx - 1 || y <= 0 || y >= ny - 1) return 0.0f;
|
|
float dvy_dx = (v_y[y * nx + (x + 1)] - v_y[y * nx + (x - 1)]) * 0.5f;
|
|
float dvx_dy = (v_x[(y + 1) * nx + x] - v_x[(y - 1) * nx + x]) * 0.5f;
|
|
return dvy_dx - dvx_dy;
|
|
}
|
|
|
|
__global__ void compute_vorticity_map(float* ux, float* uy, float* vorticity,
|
|
int nx, int ny) {
|
|
const int idx = blockIdx.x * blockDim.x + threadIdx.x;
|
|
const int N = nx * ny;
|
|
if (idx >= N) return;
|
|
|
|
const int x = idx % nx;
|
|
const int y = idx / nx;
|
|
|
|
vorticity[idx] = calculate_vorticity(x, y, nx, ny, ux, uy);
|
|
}
|
|
|
|
/* ======================================================================== */
|
|
/* H O S T F U N C T I O N S */
|
|
/* ======================================================================== */
|
|
|
|
void detect_vortex_seeds(const float* vorticity, const float* ux, const float* uy,
|
|
uint64_t current_step) {
|
|
static uint64_t last_check = 0;
|
|
if (current_step - last_check < 10000) return;
|
|
last_check = current_step;
|
|
|
|
n_seeds = 0;
|
|
|
|
for (int y = 1; y < NY - 1; y++) {
|
|
for (int x = 1; x < NX - 1; x++) {
|
|
int idx = y * NX + x;
|
|
float w = fabsf(vorticity[idx]);
|
|
|
|
if (w > VORTICITY_THRESHOLD &&
|
|
w > fabsf(vorticity[idx - 1]) &&
|
|
w > fabsf(vorticity[idx + 1]) &&
|
|
w > fabsf(vorticity[idx - NX]) &&
|
|
w > fabsf(vorticity[idx + NX])) {
|
|
|
|
int existing = -1;
|
|
for (int s = 0; s < n_seeds; s++) {
|
|
if (vortex_seeds[s].active) {
|
|
float dx = vortex_seeds[s].x - x;
|
|
float dy = vortex_seeds[s].y - y;
|
|
if (dx*dx + dy*dy < 16.0f) {
|
|
existing = s;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
if (existing >= 0) {
|
|
vortex_seeds[existing].x = x;
|
|
vortex_seeds[existing].y = y;
|
|
vortex_seeds[existing].vorticity = w;
|
|
vortex_seeds[existing].last_seen = current_step;
|
|
} else if (n_seeds < 10000) {
|
|
vortex_seeds[n_seeds].x = x;
|
|
vortex_seeds[n_seeds].y = y;
|
|
vortex_seeds[n_seeds].vorticity = w;
|
|
vortex_seeds[n_seeds].first_seen = current_step;
|
|
vortex_seeds[n_seeds].last_seen = current_step;
|
|
vortex_seeds[n_seeds].active = 1;
|
|
n_seeds++;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
for (int s = 0; s < n_seeds; s++) {
|
|
if (vortex_seeds[s].active) {
|
|
uint64_t age = current_step - vortex_seeds[s].first_seen;
|
|
|
|
if (age >= PERSISTENCE_STEPS && n_guardians < MAX_GUARDIANS) {
|
|
int idx = (int)vortex_seeds[s].y * NX + (int)vortex_seeds[s].x;
|
|
|
|
guardians[n_guardians].id = n_guardians;
|
|
guardians[n_guardians].position[0] = vortex_seeds[s].x;
|
|
guardians[n_guardians].position[1] = vortex_seeds[s].y;
|
|
guardians[n_guardians].velocity[0] = ux[idx];
|
|
guardians[n_guardians].velocity[1] = uy[idx];
|
|
guardians[n_guardians].mass = 1.0f;
|
|
guardians[n_guardians].latent_energy = vortex_seeds[s].vorticity * age;
|
|
guardians[n_guardians].persistence_age = age;
|
|
guardians[n_guardians].born_step = current_step;
|
|
guardians[n_guardians].active = 1;
|
|
|
|
n_guardians++;
|
|
vortex_seeds[s].active = 0;
|
|
}
|
|
|
|
if (current_step - vortex_seeds[s].last_seen > 10000) {
|
|
vortex_seeds[s].active = 0;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
void save_guardian_census(uint64_t current_step) {
|
|
FILE* json = fopen("guardian_census_metrics.json", "w");
|
|
if (!json) return;
|
|
|
|
fprintf(json, "{\n");
|
|
fprintf(json, " \"total_guardians\": %d,\n", n_guardians);
|
|
fprintf(json, " \"current_step\": %llu,\n", current_step);
|
|
fprintf(json, " \"guardians\": [\n");
|
|
|
|
for (int g = 0; g < n_guardians; g++) {
|
|
if (guardians[g].active) {
|
|
if (g > 0) fprintf(json, ",\n");
|
|
|
|
fprintf(json, " {\n");
|
|
fprintf(json, " \"id\": %d,\n", guardians[g].id);
|
|
fprintf(json, " \"position\": [%.1f, %.1f],\n",
|
|
guardians[g].position[0], guardians[g].position[1]);
|
|
fprintf(json, " \"velocity\": [%.6f, %.6f],\n",
|
|
guardians[g].velocity[0], guardians[g].velocity[1]);
|
|
fprintf(json, " \"mass\": %.3f,\n", guardians[g].mass);
|
|
fprintf(json, " \"latent_energy\": %.6f,\n",
|
|
guardians[g].latent_energy);
|
|
fprintf(json, " \"persistence_age\": %llu,\n",
|
|
guardians[g].persistence_age);
|
|
fprintf(json, " \"born_step\": %llu\n", guardians[g].born_step);
|
|
fprintf(json, " }");
|
|
}
|
|
}
|
|
|
|
fprintf(json, "\n ]\n");
|
|
fprintf(json, "}\n");
|
|
fclose(json);
|
|
}
|
|
|
|
/* ======================================================================== */
|
|
/* M A I N T E S T */
|
|
/* ======================================================================== */
|
|
|
|
int main() {
|
|
printf("=======================================================================\n");
|
|
printf(" METRICS ONLY - Collect Data, No Boundaries\n");
|
|
printf(" March 7 Hard-Print Compliance\n");
|
|
printf("=======================================================================\n\n");
|
|
|
|
printf("PHILOSOPHY:\n");
|
|
printf(" 1. NO POWER BOUNDARIES - Let data speak\n");
|
|
printf(" 2. NO EARLY STOPPING - Run full experiment\n");
|
|
printf(" 3. NO JUDGMENTS - Collect all metrics\n");
|
|
printf(" 4. MARCH 7 FORMAT - Standardized output\n\n");
|
|
|
|
// CUDA setup
|
|
cudaDeviceProp prop;
|
|
cudaGetDeviceProperties(&prop, 0);
|
|
printf("[CUDA] %s SM %d.%d SMs: %d\n",
|
|
prop.name, prop.major, prop.minor, prop.multiProcessorCount);
|
|
|
|
// NVML power monitoring
|
|
nvmlInit();
|
|
nvmlDevice_t nvml_dev;
|
|
nvmlDeviceGetHandleByIndex(0, &nvml_dev);
|
|
unsigned int power_mW;
|
|
nvmlDeviceGetPowerUsage(nvml_dev, &power_mW);
|
|
printf("[NVML] Idle power: %.1f W\n", power_mW / 1000.0f);
|
|
|
|
// Allocate memory
|
|
float *f0, *f1, *rho, *ux, *uy, *vorticity;
|
|
float *h_ux, *h_uy, *h_vorticity;
|
|
|
|
cudaMalloc(&f0, Q * NN * sizeof(float));
|
|
cudaMalloc(&f1, Q * NN * sizeof(float));
|
|
cudaMalloc(&rho, NN * sizeof(float));
|
|
cudaMalloc(&ux, NN * sizeof(float));
|
|
cudaMalloc(&uy, NN * sizeof(float));
|
|
cudaMalloc(&vorticity, NN * sizeof(float));
|
|
|
|
h_ux = (float*)malloc(NN * sizeof(float));
|
|
h_uy = (float*)malloc(NN * sizeof(float));
|
|
h_vorticity = (float*)malloc(NN * sizeof(float));
|
|
|
|
// Initialize
|
|
float* h_f0 = (float*)malloc(Q * NN * sizeof(float));
|
|
for (int i = 0; i < Q * NN; i++) {
|
|
h_f0[i] = 1.0f + 0.01f * (rand() / (float)RAND_MAX - 0.5f);
|
|
}
|
|
cudaMemcpy(f0, h_f0, Q * NN * sizeof(float), cudaMemcpyHostToDevice);
|
|
free(h_f0);
|
|
|
|
// Prepare output
|
|
FILE* csv = fopen("metrics_telemetry.csv", "w");
|
|
fprintf(csv, "step,power_w,n_seeds,n_guardians,steps_per_sec\n");
|
|
|
|
auto t0 = std::chrono::steady_clock::now();
|
|
uint64_t total_steps = 0;
|
|
int cur = 0;
|
|
|
|
printf("\n[EXPERIMENT] Starting metrics collection...\n");
|
|
printf(" Steps | Power | Seeds | Guardians | Steps/sec\n");
|
|
printf(" --------|-------|-------|-----------|-----------\n");
|
|
|
|
int batches = TOTAL_STEPS / STEPS_PER_BATCH;
|
|
|
|
for (int batch = 0; batch < batches; batch++) {
|
|
// Run LBM steps
|
|
for (int s = 0; s < STEPS_PER_BATCH; s++) {
|
|
lbm_collide_stream<<<GBLK(NN), BLOCK>>>(
|
|
(cur == 0) ? f0 : f1,
|
|
(cur == 0) ? f1 : f0,
|
|
rho, ux, uy, OMEGA, NX, NY);
|
|
cudaDeviceSynchronize();
|
|
cur = 1 - cur;
|
|
}
|
|
total_steps += STEPS_PER_BATCH;
|
|
|
|
// Compute vorticity map every 10k steps
|
|
if (total_steps % 10000 == 0) {
|
|
compute_vorticity_map<<<GBLK(NN), BLOCK>>>(ux, uy, vorticity, NX, NY);
|
|
cudaDeviceSynchronize();
|
|
|
|
// Copy to host for detection
|
|
cudaMemcpy(h_ux, ux, NN * sizeof(float), cudaMemcpyDeviceToHost);
|
|
cudaMemcpy(h_uy, uy, NN * sizeof(float), cudaMemcpyDeviceToHost);
|
|
cudaMemcpy(h_vorticity, vorticity, NN * sizeof(float), cudaMemcpyDeviceToHost);
|
|
|
|
detect_vortex_seeds(h_vorticity, h_ux, h_uy, total_steps);
|
|
}
|
|
|
|
// Report every 50k steps
|
|
if (total_steps % SAMPLE_INTERVAL == 0) {
|
|
nvmlDeviceGetPowerUsage(nvml_dev, &power_mW);
|
|
float power_W = power_mW / 1000.0f;
|
|
|
|
auto t_now = std::chrono::steady_clock::now();
|
|
double elapsed = std::chrono::duration<double>(t_now - t0).count();
|
|
float steps_per_sec = total_steps / elapsed;
|
|
|
|
fprintf(csv, "%llu,%.1f,%d,%d,%.0f\n",
|
|
total_steps, power_W, n_seeds, n_guardians, steps_per_sec);
|
|
|
|
printf(" %7llu | %5.0f | %5d | %9d | %8.0f\n",
|
|
total_steps, power_W, n_seeds, n_guardians, steps_per_sec);
|
|
|
|
// NO POWER BOUNDARY CHECK - Let data speak
|
|
|
|
// Save census periodically
|
|
if (n_guardians > 0 && total_steps % 100000 == 0) {
|
|
save_guardian_census(total_steps);
|
|
}
|
|
}
|
|
|
|
// Check time limit (5 minutes)
|
|
auto t_now = std::chrono::steady_clock::now();
|
|
double elapsed = std::chrono::duration<double>(t_now - t0).count();
|
|
if (elapsed > 300.0) { // 5 minutes
|
|
printf("\n[TIME] 5 minutes reached\n");
|
|
break;
|
|
}
|
|
}
|
|
|
|
auto t_end = std::chrono::steady_clock::now();
|
|
double runtime = std::chrono::duration<double>(t_end - t0).count();
|
|
|
|
// Final results - DATA ONLY, NO JUDGMENTS
|
|
printf("\n=======================================================================\n");
|
|
printf(" METRICS ONLY - EXPERIMENT DATA\n");
|
|
printf("=======================================================================\n");
|
|
|
|
printf("\nRAW METRICS:\n");
|
|
printf(" Total steps: %llu\n", total_steps);
|
|
printf(" Runtime: %.1f seconds (%.2f minutes)\n", runtime, runtime / 60.0);
|
|
printf(" Steps/sec: %.0f\n", total_steps / runtime);
|
|
|
|
nvmlDeviceGetPowerUsage(nvml_dev, &power_mW);
|
|
printf(" Final power: %.1f W\n", power_mW / 1000.0f);
|
|
|
|
printf("\nGUARDIAN DETECTION:\n");
|
|
int active_guardians = 0;
|
|
for (int g = 0; g < n_guardians; g++) {
|
|
if (guardians[g].active) active_guardians++;
|
|
}
|
|
printf(" Vortex seeds: %d\n", n_seeds);
|
|
printf(" Guardians born: %d\n", active_guardians);
|
|
printf(" Threshold: |ω| > %.6f\n", VORTICITY_THRESHOLD);
|
|
printf(" Persistence: %llu steps required\n", PERSISTENCE_STEPS);
|
|
|
|
// Save final census
|
|
save_guardian_census(total_steps);
|
|
|
|
printf("\nDATA FILES:\n");
|
|
printf(" metrics_telemetry.csv - Telemetry data\n");
|
|
printf(" guardian_census_metrics.json - Guardian census (March 7 format)\n");
|
|
|
|
printf("\nANALYSIS NOTES:\n");
|
|
printf(" - No power boundaries applied\n");
|
|
printf(" - No early stopping\n");
|
|
printf(" - Raw data collection only\n");
|
|
printf(" - March 7 format compliance\n");
|
|
|
|
// Cleanup
|
|
fclose(csv);
|
|
cudaFree(f0); cudaFree(f1);
|
|
cudaFree(rho); cudaFree(ux); cudaFree(uy); cudaFree(vorticity);
|
|
free(h_ux); free(h_uy); free(h_vorticity);
|
|
nvmlShutdown();
|
|
|
|
return 0; // Always success - data is what matters
|
|
} |