#include <crates/ulib/includes.hpp>
#include "lib.hpp"

extern "C"
void sta_levelize_calc_outs_cpu(
	usize num_pins,
	const usize *__restrict__ arcs_st,
	const STAGraphArc *__restrict__ arcs,
	usize *__restrict__ num_outs)
{
	// parallel
	for(usize i = 0; i < num_pins; ++i) {
		usize l = arcs_st[i], r = arcs_st[i + 1];
		for(usize j = l; j < r; ++j) {
			// atomic
			++num_outs[arcs[j].prev];
		}
	}
}

extern "C"
void sta_levelize_get_first_layer_cpu(
	usize num_pins,
	const usize *__restrict__ arcs_st,
	const usize *__restrict__ num_outs,
	usize *__restrict__ levels_nd,
	usize *__restrict__ tmp)
{
	// parallel
	for(usize i = 0; i < num_pins; ++i) {
		if(num_outs[i] != 0) continue;
		if(arcs_st[i] == arcs_st[i + 1]) {
			++tmp[1];  // num_dead_pins
		}
		else {
			// atomic
			usize p = tmp[0]++;  // next_level
			levels_nd[p] = i;
		}
	}
}

extern "C"
void sta_levelize_one_level_cpu(
	const usize *__restrict__ arcs_st,
	const STAGraphArc *__restrict__ arcs,
	usize *__restrict__ num_outs,
	usize l0,
	usize l1,
	usize *__restrict__ levels_nd,
	usize *__restrict__ tmp)
{
	// parallel
	for(usize p = l0; p < l1; ++p) {
		usize i = levels_nd[p];
		usize l = arcs_st[i], r = arcs_st[i + 1];
		for(usize j = l; j < r; ++j) {
			usize prev = arcs[j].prev;
			// atomic
			usize ro = --num_outs[prev];
			if(ro == 0) {
				// atomic
				usize p = tmp[0]++;
				levels_nd[p] = prev;
			}
		}
	}
}