feat: multithreading

demo/binary.c

@@ -7,11 +7,11 @@
 #define POLESITTER_IMPLEMENTATION
 #include "../src/polesitter.h"
 
-#define PARTICLE_COUNT     5000
+#define PARTICLE_COUNT     10000
 #define BLACKHOLE_MASS     1000.0F
 #define ARENA_SIZE         (1024ULL * 1024ULL * 256ULL)
 #define MAX_EXPECTED_SPEED 20
-#define THETA              3.4641F // sqrt of 12
+#define THETA              2.0F // sqrt of 12
 
 float    vx[PARTICLE_COUNT]           = {0};
 float    vy[PARTICLE_COUNT]           = {0};
@@ -51,7 +51,7 @@ void* init_polesitter(void) {
         return NULL;
     }
 
-    ps_config_t cfg = {buffer, ARENA_SIZE, THETA};
+    ps_config_t cfg = {buffer, ARENA_SIZE, PARTICLE_COUNT, THETA, 1};
     ps_init(&ctx, &cfg);
 
     arrs.x    = px;
@@ -155,7 +155,7 @@ void update(void) {
 
     UpdateCamera(&camera, CAMERA_ORBITAL);
 
-    ps_arena_clear(&ctx->arena);
+    ps_impl_arena_clear(&ctx->arena);
 
     for (int i = 0; i < PARTICLE_COUNT; ++i) {
         id[i] = i;
@@ -253,7 +253,7 @@ void draw(void) {
                               (unsigned char)(180 - (f * 100)),
                               (unsigned char)(255 - (f * 255)), 255};
         }
-        DrawPoint3D((Vector3){px[i], py[i], pz[i]}, p_color);
+        DrawSphereEx((Vector3){px[i], py[i], pz[i]}, 0.1F, 3, 3, p_color);
     }
 
     EndBlendMode();

demo/unary.c

@@ -4,14 +4,15 @@
 #include <stdint.h>
 #include <stdlib.h>
 
+#define PS_MULTITHREADING
 #define POLESITTER_IMPLEMENTATION
 #include "../src/polesitter.h"
 
-#define PARTICLE_COUNT     5000
+#define PARTICLE_COUNT     40000
 #define BLACKHOLE_MASS     1000.0F
 #define ARENA_SIZE         (1024ULL * 1024ULL * 256ULL)
 #define MAX_EXPECTED_SPEED 20
-#define THETA              3.4641F // sqrt of 12
+#define THETA              2.00F // sqrt of 12
 
 float    vx[PARTICLE_COUNT]           = {0};
 float    vy[PARTICLE_COUNT]           = {0};
@@ -51,7 +52,7 @@ void* init_polesitter(void) {
         return NULL;
     }
 
-    ps_config_t cfg = {buffer, ARENA_SIZE, THETA};
+    ps_config_t cfg = {buffer, ARENA_SIZE, PARTICLE_COUNT, THETA, 4};
     ps_init(&ctx, &cfg);
 
     arrs.x    = px;
@@ -116,7 +117,7 @@ int main(void) {
 void update(void) {
     UpdateCamera(&camera, CAMERA_ORBITAL);
 
-    ps_arena_clear(&ctx->arena);
+    ps_impl_arena_clear(&ctx->arena);
 
     for (int i = 0; i < PARTICLE_COUNT; ++i) {
         id[i] = i;

docs/plot.py

@@ -4,7 +4,8 @@
 COL_TEXT = '#C9D1D9'
 COL_BORDER = '#8B949E'
 COL_NAIVE = '#D29922'
-COL_FMM = '#1F6FEB'
+COL_FMM_ST = '#1F6FEB'
+COL_FMM_MT = '#8957E5'
 COL_BG = '#21262D'
 
 plt.rcParams.update({
@@ -26,12 +27,14 @@
 data = pd.read_csv('results.csv')
 N = data['N']
 t_direct = data['Direct_Time']
-t_fmm = data['FMM_Time']
+t_fmm_st = data['FMM_Time_ST']
+t_fmm_mt = data['FMM_Time_MT']
 
 fig, ax = plt.subplots(figsize=(8, 5))
 
 ax.plot(N, t_direct, marker='o', linewidth=2.5, color=COL_NAIVE, label='naive $\mathcal{O}(N^2)$')
-ax.plot(N, t_fmm, marker='o', linewidth=2.5, color=COL_FMM, label='FMM $\mathcal{O}(N)$')
+ax.plot(N, t_fmm_st, marker='o', linewidth=2.5, color=COL_FMM_ST, label='FMM $\mathcal{O}(N)$ 1-core')
+ax.plot(N, t_fmm_mt, marker='o', linewidth=2.5, color=COL_FMM_MT, label='FMM $\mathcal{O}(N)$ 8-core')
 
 ax.set_title('perf scaling: direct evaluation vs FMM', fontsize=14, pad=15, fontweight='bold')
 ax.set_xlabel('# of particles (N)', fontsize=12)
@@ -48,16 +51,14 @@
 
 fig_log, ax_log = plt.subplots(figsize=(8, 5))
 
-ax_log.plot(N, t_direct, marker='o', linewidth=2.5, color=COL_NAIVE,
-label='naive $\mathcal{O}(N^2)$')
-ax_log.plot(N, t_fmm, marker='o', linewidth=2.5, color=COL_FMM,
-label='FMM $\mathcal{O}(N)$')
+ax_log.plot(N, t_direct, marker='o', linewidth=2.5, color=COL_NAIVE,  label='naive $\mathcal{O}(N^2)$')
+ax_log.plot(N, t_fmm_st, marker='o', linewidth=2.5, color=COL_FMM_ST, label='FMM $\mathcal{O}(N)$ 1-core')
+ax_log.plot(N, t_fmm_mt, marker='o', linewidth=2.5, color=COL_FMM_MT, label='FMM $\mathcal{O}(N)$ 8-core')
 
 ax_log.set_xscale('log')
 ax_log.set_yscale('log')
 
-ax_log.set_title('algorithmic scaling (log-log)', fontsize=14, pad=15,
-fontweight='bold')
+ax_log.set_title('algorithmic scaling (log-log)', fontsize=14, pad=15, fontweight='bold')
 ax_log.set_xlabel('# of particles (N)', fontsize=12)
 ax_log.set_ylabel('exec time (secs)', fontsize=12)