main: threaded simulation

The actual computation is embarrassingly parallelised - the only part
that requires a mutex is the picking of any two programs due to the
fact that both programs may be affected following a compute block.
This is why we do a uniqueness check in thread_pick.

src/main.c

@@ -94,19 +94,20 @@ void simulation_draw(struct Simulation *sim)
     DrawRectangle(x * CELL_WIDTH, y * CELL_WIDTH, CELL_WIDTH, CELL_WIDTH,
                   color);
 
-    if (i == sim->p1)
-    {
-      DrawRectangleLines(x * CELL_WIDTH, y * CELL_WIDTH, CELL_WIDTH, CELL_WIDTH,
-                         BLUE);
-    }
-    if (i == sim->p1)
-    {
-      DrawRectangleLines(x * CELL_WIDTH, y * CELL_WIDTH, CELL_WIDTH, CELL_WIDTH,
-                         RED);
-    }
-
     sv = sv_chop_left(sv, 64);
   }
+
+  for (u64 i = 0; i < THREAD_POOL; ++i)
+  {
+    u64 p1 = sim->states[i].p1;
+    u64 p2 = sim->states[i].p2;
+    DrawRectangleLines((p1 / GRID_WIDTH) * CELL_WIDTH,
+                       (p1 % GRID_WIDTH) * CELL_WIDTH, CELL_WIDTH, CELL_WIDTH,
+                       BLUE);
+    DrawRectangleLines((p2 / GRID_WIDTH) * CELL_WIDTH,
+                       (p2 % GRID_WIDTH) * CELL_WIDTH, CELL_WIDTH, CELL_WIDTH,
+                       RED);
+  }
 }
 
 int main(void)
@@ -114,16 +115,15 @@ int main(void)
   srand(time(NULL));
 
   struct Simulation sim = {0};
-  simulation_init(&sim);
+  simulation_start(&sim);
 
-  bool paused = false;
   InitWindow(WIDTH, HEIGHT, "CompLife");
   SetTargetFPS(60);
   for (size_t ticks = 0; !WindowShouldClose(); ++ticks)
   {
     if (IsKeyPressed(KEY_SPACE))
     {
-      paused = !paused;
+      sim.paused = !sim.paused;
     }
     if (IsMouseButtonPressed(MOUSE_BUTTON_LEFT))
     {
@@ -143,15 +143,12 @@ int main(void)
       }
       printf("\n");
     }
-    if (!paused)
-    {
-      simulation_update(&sim);
-    }
     BeginDrawing();
     ClearBackground(BLACK);
     simulation_draw(&sim);
     EndDrawing();
   }
+  simulation_stop(&sim);
   CloseWindow();
   return 0;
 }

src/simulation.c

@@ -5,38 +5,103 @@
  */
 
 #include <stdlib.h>
+#include <string.h>
 
 #include "simulation.h"
 
-void simulation_init(struct Simulation *sim)
+bool any_threads_using(const struct ThreadState *const states, u64 n)
+{
+  for (u64 i = 0; i < THREAD_POOL; ++i)
+  {
+    if (states[i].p1 == n || states[i].p2 == n)
+    {
+      return true;
+    }
+  }
+  return false;
+}
+
+void thread_pick(struct ThreadState *state)
+{
+  struct Simulation *sim = state->sim;
+  while (true)
+  {
+    u64 p1 = rand() % (SIMULATION_SIZE / SIZEOF_PROGRAM);
+    if (any_threads_using(sim->states, p1))
+    {
+      continue;
+    }
+    u64 p2 = rand() % (SIMULATION_SIZE / SIZEOF_PROGRAM);
+    while (p1 * 8 <= ((p2 * 8) + SIZEOF_PROGRAM) &&
+           p2 * 8 <= ((p1 * 8) + SIZEOF_PROGRAM))
+    {
+      p2 = rand() % (SIMULATION_SIZE / SIZEOF_PROGRAM);
+    }
+    if (any_threads_using(sim->states, p2))
+    {
+      continue;
+    }
+    state->p1 = p1;
+    state->p2 = p2;
+    break;
+  }
+}
+
+const struct timespec THREAD_DEFAULT_SLEEP = {.tv_sec = 1};
+
+int thread_update(void *rawptr)
+{
+  struct ThreadState *state     = rawptr;
+  struct Simulation *simulation = state->sim;
+
+  while (!simulation->stopped)
+  {
+    while (simulation->paused)
+    {
+      thrd_sleep(&THREAD_DEFAULT_SLEEP, NULL);
+    }
+
+    // We only need to lock when picking the programs
+    mtx_lock(&simulation->mutex);
+    thread_pick(state);
+    mtx_unlock(&simulation->mutex);
+
+    sv_t a = SV(simulation->buffer + (state->p1 * SIZEOF_PROGRAM), 64);
+    sv_t b = SV(simulation->buffer + (state->p2 * SIZEOF_PROGRAM), 64);
+
+    struct ProgramConcat prog_concat = {0};
+    program_concat(&prog_concat, a, b);
+    program_execute(&prog_concat);
+    program_split(&prog_concat);
+  }
+  return 0;
+}
+
+void simulation_start(struct Simulation *sim)
 {
   for (u64 i = 0; i < SIMULATION_SIZE / sizeof(u16); ++i)
   {
     ((u16 *)(sim->buffer))[i] = rand() % UINT16_MAX;
   }
-}
 
-void simulation_pick(struct Simulation *sim)
-{
-  sim->p1 = rand() % (SIMULATION_SIZE / SIZEOF_PROGRAM);
-  sim->p2 = rand() % (SIMULATION_SIZE / SIZEOF_PROGRAM);
-  while (sim->p1 * 8 <= ((sim->p2 * 8) + SIZEOF_PROGRAM) &&
-         sim->p2 * 8 <= ((sim->p1 * 8) + SIZEOF_PROGRAM))
+  sim->stopped = false;
+  sim->paused  = true;
+  mtx_init(&sim->mutex, mtx_plain);
+  for (u64 i = 0; i < THREAD_POOL; ++i)
   {
-    sim->p2 = rand() % (SIMULATION_SIZE / SIZEOF_PROGRAM);
+    memset(&sim->states[i], 0, sizeof(sim->states[i]));
+    sim->states[i].sim = sim;
+    thrd_create(&sim->threads[i], thread_update, &sim->states[i]);
   }
 }
 
-void simulation_update(struct Simulation *sim)
+void simulation_stop(struct Simulation *sim)
 {
-  simulation_pick(sim);
-  sv_t a = SV(sim->buffer + (sim->p1 * SIZEOF_PROGRAM), 64);
-  sv_t b = SV(sim->buffer + (sim->p2 * SIZEOF_PROGRAM), 64);
-
-  struct ProgramConcat prog_concat = {0};
-  program_concat(&prog_concat, a, b);
-  program_execute(&prog_concat);
-  program_split(&prog_concat);
+  sim->stopped = true;
+  for (u64 i = 0; i < THREAD_POOL; ++i)
+  {
+    thrd_join(sim->threads[i], NULL);
+  }
 }
 
 /* Copyright (C) 2026 Aryadev Chavali

src/simulation.h

@@ -9,21 +9,35 @@
 #define SIMULATION_H
 
 #include "program_iter.h"
+#include <stdatomic.h>
+#include <threads.h>
 
 #define NUM_PROGRAMS_POW_2 10
 #define NUM_PROGRAMS       (1LU << NUM_PROGRAMS_POW_2)
 #define SIMULATION_SIZE    (SIZEOF_PROGRAM * NUM_PROGRAMS)
 
+#define THREAD_POOL 8
+
+struct ThreadState
+{
+  u64 p1, p2;
+  void *sim;
+};
+
 // Simulation is simply a massive tape.  We pick two 64 byte portions of this
 // buffer for our update procedure, which is based on program_iter.
 struct Simulation
 {
   char buffer[SIMULATION_SIZE];
-  u64 p1, p2;
+
+  bool paused, stopped;
+  mtx_t mutex;
+  thrd_t threads[THREAD_POOL];
+  struct ThreadState states[THREAD_POOL];
 };
 
-void simulation_init(struct Simulation *sim);
-void simulation_update(struct Simulation *sim);
+void simulation_start(struct Simulation *sim);
+void simulation_stop(struct Simulation *sim);
 
 #endif