diff --git a/src/main.c b/src/main.c
index 403f75e..05c817c 100644
--- a/src/main.c
+++ b/src/main.c
@@ -106,8 +106,8 @@ int main(int argc, char **argv)
 		return -1;
 	}
 
-	RPROF_START("Read files");
 	size_t json_size = get_file_size(f);
+	RPROF_START_BYTES("Read JSON file", json_size);
 	char *json_data = malloc(json_size);
 	size_t bytes_read = fread(json_data, 1, json_size, f);
 	if (bytes_read != json_size) {
@@ -115,12 +115,14 @@ int main(int argc, char **argv)
 		return -1;
 	}
 	fclose(f);
+	RPROF_STOP();
 
 	f64 *reference_harvensines = NULL;
 	size_t reference_harvensines_count = 0;
 	f64 reference_harvensine_sum = 0;
 	if (argc >= 3)
 	{
+		RPROF_START("Read answer file");
 		char *answers_filename = argv[2];
 		FILE *f = fopen(answers_filename, "r");
 		if (f == NULL) {
@@ -137,12 +139,12 @@ int main(int argc, char **argv)
 		fread(&reference_harvensine_sum, sizeof(f64), 1, f);
 
 		fclose(f);
+		RPROF_STOP();
 	}
-	RPROF_STOP();
 
 	// Step 1. Read json file
 
-	RPROF_START("Parse JSON");
+	RPROF_START_BYTES("Parse JSON", json_size);
 	struct json_value *parsed = NULL;
 	{
 		parsed = malloc(sizeof(struct json_value));
@@ -164,7 +166,7 @@ int main(int argc, char **argv)
 
 	// Step 3. Calculate harvensine distances
 
-	RPROF_START("Compute harvensines");
+	RPROF_START_BYTES("Compute harvensines", sizeof(struct point_pair)*pairs->count);
 	f64 *harvensines = malloc(pairs->count*sizeof(f64));
 	for (int i = 0; i < pairs->count; i++) {
 		struct point_pair *p = &pairs->pairs[i];
@@ -172,19 +174,24 @@ int main(int argc, char **argv)
 	}
 	RPROF_STOP();
 
-	RPROF_START("Sum harvensines");
+	RPROF_START_BYTES("Sum harvensines", sizeof(f64)*pairs->count);
 	f64 harvensine_sum = 0;
 	for (int i = 0; i < pairs->count; i++) {
 		harvensine_sum += harvensines[i];
 	}
 	RPROF_STOP();
 
-	RPROF_START("Free memory");
+	RPROF_START_BYTES("Free struct memory", sizeof(f64)*pairs->count + sizeof(struct point_pair)*pairs->count);
 	free(reference_harvensines);
 	free(harvensines);
+	RPROF_STOP();
+	RPROF_START("Free json memory");
 	free_json_value(parsed);
 	free_point_pairs(pairs);
 	RPROF_STOP();
+	RPROF_START_BYTES("Free json file", json_size);
+	free(json_data);
+	RPROF_STOP();
 
 	rprof_end();
 
diff --git a/src/rprof.h b/src/rprof.h
index 5234621..8bb5e02 100644
--- a/src/rprof.h
+++ b/src/rprof.h
@@ -40,6 +40,7 @@ typedef struct {
 	uint32_t calls;
 	uint64_t inclusive_duration;
 	uint64_t exclusive_duration;
+	uint64_t bytes_processed;
 } rprof_slot;
 
 typedef struct {
@@ -63,7 +64,7 @@ static rprof g_rprof = { 0 };
 
 void rprof_init();
 void rprof_end();
-void rprof_start(size_t slot_idx, char *label);
+void rprof_start(size_t slot_idx, char *label, uint64_t bytes_processed);
 void rprof_stop();
 
 int rprof_cmp_by_calls(const rprof_slot **A, const rprof_slot **B);
@@ -71,7 +72,8 @@ int rprof_cmp_by_exclusive_duration(const rprof_slot **A, const rprof_slot **B);
 int rprof_cmp_by_inclusive_duration(const rprof_slot **A, const rprof_slot **B);
 void rprof_output(prof_sort_cmp_cb sort_cb);
 
-#define RPROF_START(label) rprof_start(__COUNTER__, label)
+#define RPROF_START(label) rprof_start(__COUNTER__, label, 0)
+#define RPROF_START_BYTES(label, bytes) rprof_start(__COUNTER__, label, bytes)
 #define RPROF_STOP() rprof_stop()
 
 #ifdef RPROF_IMPLEMENTATION
@@ -187,7 +189,7 @@ int rprof_cmp_by_inclusive_duration(const rprof_slot **A, const rprof_slot **B)
 		#define ARRAY_LEN(x) (sizeof(x)/sizeof(x[0]))
 	#endif
 
-	void rprof_start(size_t slot_idx, char *label)
+	void rprof_start(size_t slot_idx, char *label, uint64_t bytes_processed)
 	{
 		assert(g_rprof.started);
 		assert(!g_rprof.finished);
@@ -198,6 +200,7 @@ int rprof_cmp_by_inclusive_duration(const rprof_slot **A, const rprof_slot **B)
 		rprof_slot *slot = &g_rprof.slots[slot_idx];
 		slot->label = label;
 		slot->calls++;
+		slot->bytes_processed += bytes_processed;
 
 		g_rprof.duration_stack[g_rprof.stack_size] = slot->inclusive_duration;
 		g_rprof.slot_stack[g_rprof.stack_size]     = slot_idx;
@@ -230,6 +233,7 @@ int rprof_cmp_by_inclusive_duration(const rprof_slot **A, const rprof_slot **B)
 
 		uint64_t total_time = g_rprof.end_time - g_rprof.init_time;
 		uint64_t cpu_hz = rprof_get_cpu_timer_hz(100);
+		float total_time_secs = (float)total_time / cpu_hz;
 
 		rprof_slot *slots[RPROF_MAX_SLOTS+1] = { 0 };
 		uint32_t slot_count = 0;
@@ -260,7 +264,7 @@ int rprof_cmp_by_inclusive_duration(const rprof_slot **A, const rprof_slot **B)
 			qsort(slots, slot_count, sizeof(rprof_slot*), (void*)sort_cb);
 		}
 
-		printf("\nTotal time taken: %.3fms (%lu)\n", (float)total_time*1000/cpu_hz, total_time);
+		printf("\nTotal time taken: %.3fms (%lu) (CPU: ~%.3fGHz)\n", total_time_secs*1000, total_time, (float)cpu_hz/1000000000);
 
 		uint32_t duration_max_width = 0;
 		uint32_t percent_max_width = 0;
@@ -281,12 +285,22 @@ int rprof_cmp_by_inclusive_duration(const rprof_slot **A, const rprof_slot **B)
 		}
 
 		char line_format[128];
-		snprintf(line_format, ARRAY_LEN(line_format), "  %%%ds - %%%dlu %%-%ds [%%d]\n", label_width, duration_max_width, percent_max_width);
+		snprintf(line_format, ARRAY_LEN(line_format), "  %%%ds - %%%dlu %%-%ds [%%d]", label_width, duration_max_width, percent_max_width);
 
 		for (int i = 0; i < slot_count; i++) {
 			rprof_slot *slot = slots[i];
 
 			printf(line_format, slot->label, slot->inclusive_duration, percent_column[i], slot->calls);
+			if (slot->bytes_processed > 0) {
+				float time_spent = (float)slot->inclusive_duration / cpu_hz;
+				float megabytes = (float)slot->bytes_processed / (1024 * 1024);
+				if (megabytes > 10) {
+					printf(" at %.3fgb/s", (megabytes / 1024) / time_spent);
+				} else {
+					printf(" at %.3fmb/s", megabytes / time_spent);
+				}
+			}
+			printf("\n");
 		}
 	}