path: root/lib/librte_sched/rte_sched.c

/*-
 *   BSD LICENSE
 *
 *   Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
 *   All rights reserved.
 *
 *   Redistribution and use in source and binary forms, with or without
 *   modification, are permitted provided that the following conditions
 *   are met:
 *
 *     * Redistributions of source code must retain the above copyright
 *       notice, this list of conditions and the following disclaimer.
 *     * Redistributions in binary form must reproduce the above copyright
 *       notice, this list of conditions and the following disclaimer in
 *       the documentation and/or other materials provided with the
 *       distribution.
 *     * Neither the name of Intel Corporation nor the names of its
 *       contributors may be used to endorse or promote products derived
 *       from this software without specific prior written permission.
 *
 *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */

#include <stdio.h>
#include <string.h>

#include <rte_common.h>
#include <rte_log.h>
#include <rte_memory.h>
#include <rte_malloc.h>
#include <rte_cycles.h>
#include <rte_prefetch.h>
#include <rte_branch_prediction.h>
#include <rte_mbuf.h>
#include <rte_bitmap.h>

#include "rte_sched.h"
#include "rte_sched_common.h"
#include "rte_approx.h"
#include "rte_reciprocal.h"

#ifdef __INTEL_COMPILER
#pragma warning(disable:2259) /* conversion may lose significant bits */
#endif

#ifdef RTE_SCHED_VECTOR
#include <rte_vect.h>

#ifdef RTE_ARCH_X86
#define SCHED_VECTOR_SSE4
#elif defined(RTE_MACHINE_CPUFLAG_NEON)
#define SCHED_VECTOR_NEON
#endif

#endif

#define RTE_SCHED_TB_RATE_CONFIG_ERR          (1e-7)
#define RTE_SCHED_WRR_SHIFT                   3
#define RTE_SCHED_GRINDER_PCACHE_SIZE         (64 / RTE_SCHED_QUEUES_PER_PIPE)
#define RTE_SCHED_PIPE_INVALID                UINT32_MAX
#define RTE_SCHED_BMP_POS_INVALID             UINT32_MAX

/* Scaling for cycles_per_byte calculation
 * Chosen so that minimum rate is 480 bit/sec
 */
#define RTE_SCHED_TIME_SHIFT		      8

struct rte_sched_subport {
	/* Token bucket (TB) */
	uint64_t tb_time; /* time of last update */
	uint32_t tb_period;
	uint32_t tb_credits_per_period;
	uint32_t tb_size;
	uint32_t tb_credits;

	/* Traffic classes (TCs) */
	uint64_t tc_time; /* time of next update */
	uint32_t tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
	uint32_t tc_credits[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
	uint32_t tc_period;

	/* TC oversubscription */
	uint32_t tc_ov_wm;
	uint32_t tc_ov_wm_min;
	uint32_t tc_ov_wm_max;
	uint8_t tc_ov_period_id;
	uint8_t tc_ov;
	uint32_t tc_ov_n;
	double tc_ov_rate;

	/* Statistics */
	struct rte_sched_subport_stats stats;
};

struct rte_sched_pipe_profile {
	/* Token bucket (TB) */
	uint32_t tb_period;
	uint32_t tb_credits_per_period;
	uint32_t tb_size;

	/* Pipe traffic classes */
	uint32_t tc_period;
	uint32_t tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
	uint8_t tc_ov_weight;

	/* Pipe queues */
	uint8_t  wrr_cost[RTE_SCHED_QUEUES_PER_PIPE];
};

struct rte_sched_pipe {
	/* Token bucket (TB) */
	uint64_t tb_time; /* time of last update */
	uint32_t tb_credits;

	/* Pipe profile and flags */
	uint32_t profile;

	/* Traffic classes (TCs) */
	uint64_t tc_time; /* time of next update */
	uint32_t tc_credits[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];

	/* Weighted Round Robin (WRR) */
	uint8_t wrr_tokens[RTE_SCHED_QUEUES_PER_PIPE];

	/* TC oversubscription */
	uint32_t tc_ov_credits;
	uint8_t tc_ov_period_id;
	uint8_t reserved[3];
} __rte_cache_aligned;

struct rte_sched_queue {
	uint16_t qw;
	uint16_t qr;
};

struct rte_sched_queue_extra {
	struct rte_sched_queue_stats stats;
#ifdef RTE_SCHED_RED
	struct rte_red red;
#endif
};

enum grinder_state {
	e_GRINDER_PREFETCH_PIPE = 0,
	e_GRINDER_PREFETCH_TC_QUEUE_ARRAYS,
	e_GRINDER_PREFETCH_MBUF,
	e_GRINDER_READ_MBUF
};

/*
 * Path through the scheduler hierarchy used by the scheduler enqueue
 * operation to identify the destination queue for the current
 * packet. Stored in the field pkt.hash.sched of struct rte_mbuf of
 * each packet, typically written by the classification stage and read
 * by scheduler enqueue.
 */
struct rte_sched_port_hierarchy {
	uint16_t queue:2;                /**< Queue ID (0 .. 3) */
	uint16_t traffic_class:2;        /**< Traffic class ID (0 .. 3)*/
	uint32_t color:2;                /**< Color */
	uint16_t unused:10;
	uint16_t subport;                /**< Subport ID */
	uint32_t pipe;		         /**< Pipe ID */
};

struct rte_sched_grinder {
	/* Pipe cache */
	uint16_t pcache_qmask[RTE_SCHED_GRINDER_PCACHE_SIZE];
	uint32_t pcache_qindex[RTE_SCHED_GRINDER_PCACHE_SIZE];
	uint32_t pcache_w;
	uint32_t pcache_r;

	/* Current pipe */
	enum grinder_state state;
	uint32_t productive;
	uint32_t pindex;
	struct rte_sched_subport *subport;
	struct rte_sched_pipe *pipe;
	struct rte_sched_pipe_profile *pipe_params;

	/* TC cache */
	uint8_t tccache_qmask[4];
	uint32_t tccache_qindex[4];
	uint32_t tccache_w;
	uint32_t tccache_r;

	/* Current TC */
	uint32_t tc_index;
	struct rte_sched_queue *queue[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
	struct rte_mbuf **qbase[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
	uint32_t qindex[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
	uint16_t qsize;
	uint32_t qmask;
	uint32_t qpos;
	struct rte_mbuf *pkt;

	/* WRR */
	uint16_t wrr_tokens[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS];
	uint16_t wrr_mask[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS];
	uint8_t wrr_cost[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS];
};

struct rte_sched_port {
	/* User parameters */
	uint32_t n_subports_per_port;
	uint32_t n_pipes_per_subport;
	uint32_t rate;
	uint32_t mtu;
	uint32_t frame_overhead;
	uint16_t qsize[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
	uint32_t n_pipe_profiles;
	uint32_t pipe_tc3_rate_max;
#ifdef RTE_SCHED_RED
	struct rte_red_config red_config[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE][e_RTE_METER_COLORS];
#endif

	/* Timing */
	uint64_t time_cpu_cycles;     /* Current CPU time measured in CPU cyles */
	uint64_t time_cpu_bytes;      /* Current CPU time measured in bytes */
	uint64_t time;                /* Current NIC TX time measured in bytes */
	struct rte_reciprocal inv_cycles_per_byte; /* CPU cycles per byte */

	/* Scheduling loop detection */
	uint32_t pipe_loop;
	uint32_t pipe_exhaustion;

	/* Bitmap */
	struct rte_bitmap *bmp;
	uint32_t grinder_base_bmp_pos[RTE_SCHED_PORT_N_GRINDERS] __rte_aligned_16;

	/* Grinders */
	struct rte_sched_grinder grinder[RTE_SCHED_PORT_N_GRINDERS];
	uint32_t busy_grinders;
	struct rte_mbuf **pkts_out;
	uint32_t n_pkts_out;

	/* Queue base calculation */
	uint32_t qsize_add[RTE_SCHED_QUEUES_PER_PIPE];
	uint32_t qsize_sum;

	/* Large data structures */
	struct rte_sched_subport *subport;
	struct rte_sched_pipe *pipe;
	struct rte_sched_queue *queue;
	struct rte_sched_queue_extra *queue_extra;
	struct rte_sched_pipe_profile *pipe_profiles;
	uint8_t *bmp_array;
	struct rte_mbuf **queue_array;
	uint8_t memory[0] __rte_cache_aligned;
} __rte_cache_aligned;

enum rte_sched_port_array {
	e_RTE_SCHED_PORT_ARRAY_SUBPORT = 0,
	e_RTE_SCHED_PORT_ARRAY_PIPE,
	e_RTE_SCHED_PORT_ARRAY_QUEUE,
	e_RTE_SCHED_PORT_ARRAY_QUEUE_EXTRA,
	e_RTE_SCHED_PORT_ARRAY_PIPE_PROFILES,
	e_RTE_SCHED_PORT_ARRAY_BMP_ARRAY,
	e_RTE_SCHED_PORT_ARRAY_QUEUE_ARRAY,
	e_RTE_SCHED_PORT_ARRAY_TOTAL,
};

#ifdef RTE_SCHED_COLLECT_STATS

static inline uint32_t
rte_sched_port_queues_per_subport(struct rte_sched_port *port)
{
	return RTE_SCHED_QUEUES_PER_PIPE * port->n_pipes_per_subport;
}

#endif

static inline uint32_t
rte_sched_port_queues_per_port(struct rte_sched_port *port)
{
	return RTE_SCHED_QUEUES_PER_PIPE * port->n_pipes_per_subport * port->n_subports_per_port;
}

static inline struct rte_mbuf **
rte_sched_port_qbase(struct rte_sched_port *port, uint32_t qindex)
{
	uint32_t pindex = qindex >> 4;
	uint32_t qpos = qindex & 0xF;

	return (port->queue_array + pindex *
		port->qsize_sum + port->qsize_add[qpos]);
}

static inline uint16_t
rte_sched_port_qsize(struct rte_sched_port *port, uint32_t qindex)
{
	uint32_t tc = (qindex >> 2) & 0x3;

	return port->qsize[tc];
}

static int
rte_sched_port_check_params(struct rte_sched_port_params *params)
{
	uint32_t i, j;

	if (params == NULL)
		return -1;

	/* socket */
	if ((params->socket < 0) || (params->socket >= RTE_MAX_NUMA_NODES))
		return -3;

	/* rate */
	if (params->rate == 0)
		return -4;

	/* mtu */
	if (params->mtu == 0)
		return -5;

	/* n_subports_per_port: non-zero, limited to 16 bits, power of 2 */
	if (params->n_subports_per_port == 0 ||
	    params->n_subports_per_port > 1u << 16 ||
	    !rte_is_power_of_2(params->n_subports_per_port))
		return -6;

	/* n_pipes_per_subport: non-zero, power of 2 */
	if (params->n_pipes_per_subport == 0 ||
	    !rte_is_power_of_2(params->n_pipes_per_subport))
		return -7;

	/* qsize: non-zero, power of 2,
	 * no bigger than 32K (due to 16-bit read/write pointers)
	 */
	for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) {
		uint16_t qsize = params->qsize[i];

		if (qsize == 0 || !rte_is_power_of_2(qsize))
			return -8;
	}

	/* pipe_profiles and n_pipe_profiles */
	if (params->pipe_profiles == NULL ||
	    params->n_pipe_profiles == 0 ||
	    params->n_pipe_profiles > RTE_SCHED_PIPE_PROFILES_PER_PORT)
		return -9;

	for (i = 0; i < params->n_pipe_profiles; i++) {
		struct rte_sched_pipe_params *p = params->pipe_profiles + i;

		/* TB rate: non-zero, not greater than port rate */
		if (p->tb_rate == 0 || p->tb_rate > params->rate)
			return -10;

		/* TB size: non-zero */
		if (p->tb_size == 0)
			return -11;

		/* TC rate: non-zero, less than pipe rate */
		for (j = 0; j < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; j++) {
			if (p->tc_rate[j] == 0 || p->tc_rate[j] > p->tb_rate)
				return -12;
		}

		/* TC period: non-zero */
		if (p->tc_period == 0)
			return -13;

#ifdef RTE_SCHED_SUBPORT_TC_OV
		/* TC3 oversubscription weight: non-zero */
		if (p->tc_ov_weight == 0)
			return -14;
#endif

		/* Queue WRR weights: non-zero */
		for (j = 0; j < RTE_SCHED_QUEUES_PER_PIPE; j++) {
			if (p->wrr_weights[j] == 0)
				return -15;
		}
	}

	return 0;
}

static uint32_t
rte_sched_port_get_array_base(struct rte_sched_port_params *params, enum rte_sched_port_array array)
{
	uint32_t n_subports_per_port = params->n_subports_per_port;
	uint32_t n_pipes_per_subport = params->n_pipes_per_subport;
	uint32_t n_pipes_per_port = n_pipes_per_subport * n_subports_per_port;
	uint32_t n_queues_per_port = RTE_SCHED_QUEUES_PER_PIPE * n_pipes_per_subport * n_subports_per_port;

	uint32_t size_subport = n_subports_per_port * sizeof(struct rte_sched_subport);
	uint32_t size_pipe = n_pipes_per_port * sizeof(struct rte_sched_pipe);
	uint32_t size_queue = n_queues_per_port * sizeof(struct rte_sched_queue);
	uint32_t size_queue_extra
		= n_queues_per_port * sizeof(struct rte_sched_queue_extra);
	uint32_t size_pipe_profiles
		= RTE_SCHED_PIPE_PROFILES_PER_PORT * sizeof(struct rte_sched_pipe_profile);
	uint32_t size_bmp_array = rte_bitmap_get_memory_footprint(n_queues_per_port);
	uint32_t size_per_pipe_queue_array, size_queue_array;

	uint32_t base, i;

	size_per_pipe_queue_array = 0;
	for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) {
		size_per_pipe_queue_array += RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS
			* params->qsize[i] * sizeof(struct rte_mbuf *);
	}
	size_queue_array = n_pipes_per_port * size_per_pipe_queue_array;

	base = 0;

	if (array == e_RTE_SCHED_PORT_ARRAY_SUBPORT)
		return base;
	base += RTE_CACHE_LINE_ROUNDUP(size_subport);

	if (array == e_RTE_SCHED_PORT_ARRAY_PIPE)
		return base;
	base += RTE_CACHE_LINE_ROUNDUP(size_pipe);

	if (array == e_RTE_SCHED_PORT_ARRAY_QUEUE)
		return base;
	base += RTE_CACHE_LINE_ROUNDUP(size_queue);

	if (array == e_RTE_SCHED_PORT_ARRAY_QUEUE_EXTRA)
		return base;
	base += RTE_CACHE_LINE_ROUNDUP(size_queue_extra);

	if (array == e_RTE_SCHED_PORT_ARRAY_PIPE_PROFILES)
		return base;
	base += RTE_CACHE_LINE_ROUNDUP(size_pipe_profiles);

	if (array == e_RTE_SCHED_PORT_ARRAY_BMP_ARRAY)
		return base;
	base += RTE_CACHE_LINE_ROUNDUP(size_bmp_array);

	if (array == e_RTE_SCHED_PORT_ARRAY_QUEUE_ARRAY)
		return base;
	base += RTE_CACHE_LINE_ROUNDUP(size_queue_array);

	return base;
}

uint32_t
rte_sched_port_get_memory_footprint(struct rte_sched_port_params *params)
{
	uint32_t size0, size1;
	int status;

	status = rte_sched_port_check_params(params);
	if (status != 0) {
		RTE_LOG(NOTICE, SCHED,
			"Port scheduler params check failed (%d)\n", status);

		return 0;
	}

	size0 = sizeof(struct rte_sched_port);
	size1 = rte_sched_port_get_array_base(params, e_RTE_SCHED_PORT_ARRAY_TOTAL);

	return size0 + size1;
}

static void
rte_sched_port_config_qsize(struct rte_sched_port *port)
{
	/* TC 0 */
	port->qsize_add[0] = 0;
	port->qsize_add[1] = port->qsize_add[0] + port->qsize[0];
	port->qsize_add[2] = port->qsize_add[1] + port->qsize[0];
	port->qsize_add[3] = port->qsize_add[2] + port->qsize[0];

	/* TC 1 */
	port->qsize_add[4] = port->qsize_add[3] + port->qsize[0];
	port->qsize_add[5] = port->qsize_add[4] + port->qsize[1];
	port->qsize_add[6] = port->qsize_add[5] + port->qsize[1];
	port->qsize_add[7] = port->qsize_add[6] + port->qsize[1];

	/* TC 2 */
	port->qsize_add[8] = port->qsize_add[7] + port->qsize[1];
	port->qsize_add[9] = port->qsize_add[8] + port->qsize[2];
	port->qsize_add[10] = port->qsize_add[9] + port->qsize[2];
	port->qsize_add[11] = port->qsize_add[10] + port->qsize[2];

	/* TC 3 */
	port->qsize_add[12] = port->qsize_add[11] + port->qsize[2];
	port->qsize_add[13] = port->qsize_add[12] + port->qsize[3];
	port->qsize_add[14] = port->qsize_add[13] + port->qsize[3];
	port->qsize_add[15] = port->qsize_add[14] + port->qsize[3];

	port->qsize_sum = port->qsize_add[15] + port->qsize[3];
}

static void
rte_sched_port_log_pipe_profile(struct rte_sched_port *port, uint32_t i)
{
	struct rte_sched_pipe_profile *p = port->pipe_profiles + i;

	RTE_LOG(DEBUG, SCHED, "Low level config for pipe profile %u:\n"
		"    Token bucket: period = %u, credits per period = %u, size = %u\n"
		"    Traffic classes: period = %u, credits per period = [%u, %u, %u, %u]\n"
		"    Traffic class 3 oversubscription: weight = %hhu\n"
		"    WRR cost: [%hhu, %hhu, %hhu, %hhu], [%hhu, %hhu, %hhu, %hhu], [%hhu, %hhu, %hhu, %hhu], [%hhu, %hhu, %hhu, %hhu]\n",
		i,

		/* Token bucket */
		p->tb_period,
		p->tb_credits_per_period,
		p->tb_size,

		/* Traffic classes */
		p->tc_period,
		p->tc_credits_per_period[0],
		p->tc_credits_per_period[1],
		p->tc_credits_per_period[2],
		p->tc_credits_per_period[3],

		/* Traffic class 3 oversubscription */
		p->tc_ov_weight,

		/* WRR */
		p->wrr_cost[ 0], p->wrr_cost[ 1], p->wrr_cost[ 2], p->wrr_cost[ 3],
		p->wrr_cost[ 4], p->wrr_cost[ 5], p->wrr_cost[ 6], p->wrr_cost[ 7],
		p->wrr_cost[ 8], p->wrr_cost[ 9], p->wrr_cost[10], p->wrr_cost[11],
		p->wrr_cost[12], p->wrr_cost[13], p->wrr_cost[14], p->wrr_cost[15]);
}

static inline uint64_t
rte_sched_time_ms_to_bytes(uint32_t time_ms, uint32_t rate)
{
	uint64_t time = time_ms;

	time = (time * rate) / 1000;

	return time;
}

static void
rte_sched_port_config_pipe_profile_table(struct rte_sched_port *port, struct rte_sched_port_params *params)
{
	uint32_t i, j;

	for (i = 0; i < port->n_pipe_profiles; i++) {
		struct rte_sched_pipe_params *src = params->pipe_profiles + i;
		struct rte_sched_pipe_profile *dst = port->pipe_profiles + i;

		/* Token Bucket */
		if (src->tb_rate == params->rate) {
			dst->tb_credits_per_period = 1;
			dst->tb_period = 1;
		} else {
			double tb_rate = (double) src->tb_rate
				/ (double) params->rate;
			double d = RTE_SCHED_TB_RATE_CONFIG_ERR;

			rte_approx(tb_rate, d,
				   &dst->tb_credits_per_period, &dst->tb_period);
		}
		dst->tb_size = src->tb_size;

		/* Traffic Classes */
		dst->tc_period = rte_sched_time_ms_to_bytes(src->tc_period,
							    params->rate);

		for (j = 0; j < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; j++)
			dst->tc_credits_per_period[j]
				= rte_sched_time_ms_to_bytes(src->tc_period,
							     src->tc_rate[j]);

#ifdef RTE_SCHED_SUBPORT_TC_OV
		dst->tc_ov_weight = src->tc_ov_weight;
#endif

		/* WRR */
		for (j = 0; j < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; j++) {
			uint32_t wrr_cost[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS];
			uint32_t lcd, lcd1, lcd2;
			uint32_t qindex;

			qindex = j * RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS;

			wrr_cost[0] = src->wrr_weights[qindex];
			wrr_cost[1] = src->wrr_weights[qindex + 1];
			wrr_cost[2] = src->wrr_weights[qindex + 2];
			wrr_cost[3] = src->wrr_weights[qindex + 3];

			lcd1 = rte_get_lcd(wrr_cost[0], wrr_cost[1]);
			lcd2 = rte_get_lcd(wrr_cost[2], wrr_cost[3]);
			lcd = rte_get_lcd(lcd1, lcd2);

			wrr_cost[0] = lcd / wrr_cost[0];
			wrr_cost[1] = lcd / wrr_cost[1];
			wrr_cost[2] = lcd / wrr_cost[2];
			wrr_cost[3] = lcd / wrr_cost[3];

			dst->wrr_cost[qindex] = (uint8_t) wrr_cost[0];
			dst->wrr_cost[qindex + 1] = (uint8_t) wrr_cost[1];
			dst->wrr_cost[qindex + 2] = (uint8_t) wrr_cost[2];
			dst->wrr_cost[qindex + 3] = (uint8_t) wrr_cost[3];
		}

		rte_sched_port_log_pipe_profile(port, i);
	}

	port->pipe_tc3_rate_max = 0;
	for (i = 0; i < port->n_pipe_profiles; i++) {
		struct rte_sched_pipe_params *src = params->pipe_profiles + i;
		uint32_t pipe_tc3_rate = src->tc_rate[3];

		if (port->pipe_tc3_rate_max < pipe_tc3_rate)
			port->pipe_tc3_rate_max = pipe_tc3_rate;
	}
}

struct rte_sched_port *
rte_sched_port_config(struct rte_sched_port_params *params)
{
	struct rte_sched_port *port = NULL;
	uint32_t mem_size, bmp_mem_size, n_queues_per_port, i, cycles_per_byte;

	/* Check user parameters. Determine the amount of memory to allocate */
	mem_size = rte_sched_port_get_memory_footprint(params);
	if (mem_size == 0)
		return NULL;

	/* Allocate memory to store the data structures */
	port = rte_zmalloc("qos_params", mem_size, RTE_CACHE_LINE_SIZE);
	if (port == NULL)
		return NULL;

	/* compile time checks */
	RTE_BUILD_BUG_ON(RTE_SCHED_PORT_N_GRINDERS == 0);
	RTE_BUILD_BUG_ON(RTE_SCHED_PORT_N_GRINDERS & (RTE_SCHED_PORT_N_GRINDERS - 1));

	/* User parameters */
	port->n_subports_per_port = params->n_subports_per_port;
	port->n_pipes_per_subport = params->n_pipes_per_subport;
	port->rate = params->rate;
	port->mtu = params->mtu + params->frame_overhead;
	port->frame_overhead = params->frame_overhead;
	memcpy(port->qsize, params->qsize, sizeof(params->qsize));
	port->n_pipe_profiles = params->n_pipe_profiles;

#ifdef RTE_SCHED_RED
	for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) {
		uint32_t j;

		for (j = 0; j < e_RTE_METER_COLORS; j++) {
			/* if min/max are both zero, then RED is disabled */
			if ((params->red_params[i][j].min_th |
			     params->red_params[i][j].max_th) == 0) {
				continue;
			}

			if (rte_red_config_init(&port->red_config[i][j],
				params->red_params[i][j].wq_log2,
				params->red_params[i][j].min_th,
				params->red_params[i][j].max_th,
				params->red_params[i][j].maxp_inv) != 0) {
				return NULL;
			}
		}
	}
#endif

	/* Timing */
	port->time_cpu_cycles = rte_get_tsc_cycles();
	port->time_cpu_bytes = 0;
	port->time = 0;

	cycles_per_byte = (rte_get_tsc_hz() << RTE_SCHED_TIME_SHIFT)
		/ params->rate;
	port->inv_cycles_per_byte = rte_reciprocal_value(cycles_per_byte);

	/* Scheduling loop detection */
	port->pipe_loop = RTE_SCHED_PIPE_INVALID;
	port->pipe_exhaustion = 0;

	/* Grinders */
	port->busy_grinders = 0;
	port->pkts_out = NULL;
	port->n_pkts_out = 0;

	/* Queue base calculation */
	rte_sched_port_config_qsize(port);

	/* Large data structures */
	port->subport = (struct rte_sched_subport *)
		(port->memory + rte_sched_port_get_array_base(params,
							      e_RTE_SCHED_PORT_ARRAY_SUBPORT));
	port->pipe = (struct rte_sched_pipe *)
		(port->memory + rte_sched_port_get_array_base(params,
							      e_RTE_SCHED_PORT_ARRAY_PIPE));
	port->queue = (struct rte_sched_queue *)
		(port->memory + rte_sched_port_get_array_base(params,
							      e_RTE_SCHED_PORT_ARRAY_QUEUE));
	port->queue_extra = (struct rte_sched_queue_extra *)
		(port->memory + rte_sched_port_get_array_base(params,
							      e_RTE_SCHED_PORT_ARRAY_QUEUE_EXTRA));
	port->pipe_profiles = (struct rte_sched_pipe_profile *)
		(port->memory + rte_sched_port_get_array_base(params,
							      e_RTE_SCHED_PORT_ARRAY_PIPE_PROFILES));
	port->bmp_array =  port->memory
		+ rte_sched_port_get_array_base(params, e_RTE_SCHED_PORT_ARRAY_BMP_ARRAY);
	port->queue_array = (struct rte_mbuf **)
		(port->memory + rte_sched_port_get_array_base(params,
							      e_RTE_SCHED_PORT_ARRAY_QUEUE_ARRAY));

	/* Pipe profile table */
	rte_sched_port_config_pipe_profile_table(port, params);

	/* Bitmap */
	n_queues_per_port = rte_sched_port_queues_per_port(port);
	bmp_mem_size = rte_bitmap_get_memory_footprint(n_queues_per_port);
	port->bmp = rte_bitmap_init(n_queues_per_port, port->bmp_array,
				    bmp_mem_size);
	if (port->bmp == NULL) {
		RTE_LOG(ERR, SCHED, "Bitmap init error\n");
		return NULL;
	}

	for (i = 0; i < RTE_SCHED_PORT_N_GRINDERS; i++)
		port->grinder_base_bmp_pos[i] = RTE_SCHED_PIPE_INVALID;


	return port;
}

void
rte_sched_port_free(struct rte_sched_port *port)
{
	uint32_t qindex;
	uint32_t n_queues_per_port;

	/* Check user parameters */
	if (port == NULL)
		return;

	n_queues_per_port = rte_sched_port_queues_per_port(port);

	/* Free enqueued mbufs */
	for (qindex = 0; qindex < n_queues_per_port; qindex++) {
		struct rte_mbuf **mbufs = rte_sched_port_qbase(port, qindex);
		uint16_t qsize = rte_sched_port_qsize(port, qindex);
		struct rte_sched_queue *queue = port->queue + qindex;
		uint16_t qr = queue->qr & (qsize - 1);
		uint16_t qw = queue->qw & (qsize - 1);

		for (; qr != qw; qr = (qr + 1) & (qsize - 1))
			rte_pktmbuf_free(mbufs[qr]);
	}

	rte_bitmap_free(port->bmp);
	rte_free(port);
}

static void
rte_sched_port_log_subport_config(struct rte_sched_port *port, uint32_t i)
{
	struct rte_sched_subport *s = port->subport + i;

	RTE_LOG(DEBUG, SCHED, "Low level config for subport %u:\n"
		"    Token bucket: period = %u, credits per period = %u, size = %u\n"
		"    Traffic classes: period = %u, credits per period = [%u, %u, %u, %u]\n"
		"    Traffic class 3 oversubscription: wm min = %u, wm max = %u\n",
		i,

		/* Token bucket */
		s->tb_period,
		s->tb_credits_per_period,
		s->tb_size,

		/* Traffic classes */
		s->tc_period,
		s->tc_credits_per_period[0],
		s->tc_credits_per_period[1],
		s->tc_credits_per_period[2],
		s->tc_credits_per_period[3],

		/* Traffic class 3 oversubscription */
		s->tc_ov_wm_min,
		s->tc_ov_wm_max);
}

int
rte_sched_subport_config(struct rte_sched_port *port,
	uint32_t subport_id,
	struct rte_sched_subport_params *params)
{
	struct rte_sched_subport *s;
	uint32_t i;

	/* Check user parameters */
	if (port == NULL ||
	    subport_id >= port->n_subports_per_port ||
	    params == NULL)
		return -1;

	if (params->tb_rate == 0 || params->tb_rate > port->rate)
		return -2;

	if (params->tb_size == 0)
		return -3;

	for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) {
		if (params->tc_rate[i] == 0 ||
		    params->tc_rate[i] > params->tb_rate)
			return -4;
	}

	if (params->tc_period == 0)
		return -5;

	s = port->subport + subport_id;

	/* Token Bucket (TB) */
	if (params->tb_rate == port->rate) {
		s->tb_credits_per_period = 1;
		s->tb_period = 1;
	} else {
		double tb_rate = ((double) params->tb_rate) / ((double) port->rate);
		double d = RTE_SCHED_TB_RATE_CONFIG_ERR;

		rte_approx(tb_rate, d, &s->tb_credits_per_period, &s->tb_period);
	}

	s->tb_size = params->tb_size;
	s->tb_time = port->time;
	s->tb_credits = s->tb_size / 2;

	/* Traffic Classes (TCs) */
	s->tc_period = rte_sched_time_ms_to_bytes(params->tc_period, port->rate);
	for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) {
		s->tc_credits_per_period[i]
			= rte_sched_time_ms_to_bytes(params->tc_period,
						     params->tc_rate[i]);
	}
	s->tc_time = port->time + s->tc_period;
	for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++)
		s->tc_credits[i] = s->tc_credits_per_period[i];

#ifdef RTE_SCHED_SUBPORT_TC_OV
	/* TC oversubscription */
	s->tc_ov_wm_min = port->mtu;
	s->tc_ov_wm_max = rte_sched_time_ms_to_bytes(params->tc_period,
						     port->pipe_tc3_rate_max);
	s->tc_ov_wm = s->tc_ov_wm_max;
	s->tc_ov_period_id = 0;
	s->tc_ov = 0;
	s->tc_ov_n = 0;
	s->tc_ov_rate = 0;
#endif

	rte_sched_port_log_subport_config(port, subport_id);

	return 0;
}

int
rte_sched_pipe_config(struct rte_sched_port *port,
	uint32_t subport_id,
	uint32_t pipe_id,
	int32_t pipe_profile)
{
	struct rte_sched_subport *s;
	struct rte_sched_pipe *p;
	struct rte_sched_pipe_profile *params;
	uint32_t deactivate, profile, i;

	/* Check user parameters */
	profile = (uint32_t) pipe_profile;
	deactivate = (pipe_profile < 0);

	if (port == NULL ||
	    subport_id >= port->n_subports_per_port ||
	    pipe_id >= port->n_pipes_per_subport ||
	    (!deactivate && profile >= port->n_pipe_profiles))
		return -1;


	/* Check that subport configuration is valid */
	s = port->subport + subport_id;
	if (s->tb_period == 0)
		return -2;

	p = port->pipe + (subport_id * port->n_pipes_per_subport + pipe_id);

	/* Handle the case when pipe already has a valid configuration */
	if (p->tb_time) {
		params = port->pipe_profiles + p->profile;

#ifdef RTE_SCHED_SUBPORT_TC_OV
		double subport_tc3_rate = (double) s->tc_credits_per_period[3]
			/ (double) s->tc_period;
		double pipe_tc3_rate = (double) params->tc_credits_per_period[3]
			/ (double) params->tc_period;
		uint32_t tc3_ov = s->tc_ov;

		/* Unplug pipe from its subport */
		s->tc_ov_n -= params->tc_ov_weight;
		s->tc_ov_rate -= pipe_tc3_rate;
		s->tc_ov = s->tc_ov_rate > subport_tc3_rate;

		if (s->tc_ov != tc3_ov) {
			RTE_LOG(DEBUG, SCHED,
				"Subport %u TC3 oversubscription is OFF (%.4lf >= %.4lf)\n",
				subport_id, subport_tc3_rate, s->tc_ov_rate);
		}
#endif

		/* Reset the pipe */
		memset(p, 0, sizeof(struct rte_sched_pipe));
	}

	if (deactivate)
		return 0;

	/* Apply the new pipe configuration */
	p->profile = profile;
	params = port->pipe_profiles + p->profile;

	/* Token Bucket (TB) */
	p->tb_time = port->time;
	p->tb_credits = params->tb_size / 2;

	/* Traffic Classes (TCs) */
	p->tc_time = port->time + params->tc_period;
	for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++)
		p->tc_credits[i] = params->tc_credits_per_period[i];

#ifdef RTE_SCHED_SUBPORT_TC_OV
	{
		/* Subport TC3 oversubscription */
		double subport_tc3_rate = (double) s->tc_credits_per_period[3]
			/ (double) s->tc_period;
		double pipe_tc3_rate = (double) params->tc_credits_per_period[3]
			/ (double) params->tc_period;
		uint32_t tc3_ov = s->tc_ov;

		s->tc_ov_n += params->tc_ov_weight;
		s->tc_ov_rate += pipe_tc3_rate;
		s->tc_ov = s->tc_ov_rate > subport_tc3_rate;

		if (s->tc_ov != tc3_ov) {
			RTE_LOG(DEBUG, SCHED,
				"Subport %u TC3 oversubscription is ON (%.4lf < %.4lf)\n",
				subport_id, subport_tc3_rate, s->tc_ov_rate);
		}
		p->tc_ov_period_id = s->tc_ov_period_id;
		p->tc_ov_credits = s->tc_ov_wm;
	}
#endif

	return 0;
}

void
rte_sched_port_pkt_write(struct rte_mbuf *pkt,
			 uint32_t subport, uint32_t pipe, uint32_t traffic_class,
			 uint32_t queue, enum rte_meter_color color)
{
	struct rte_sched_port_hierarchy *sched
		= (struct rte_sched_port_hierarchy *) &pkt->hash.sched;

	RTE_BUILD_BUG_ON(sizeof(*sched) > sizeof(pkt->hash.sched));

	sched->color = (uint32_t) color;
	sched->subport = subport;
	sched->pipe = pipe;
	sched->traffic_class = traffic_class;
	sched->queue = queue;
}

void
rte_sched_port_pkt_read_tree_path(const struct rte_mbuf *pkt,
				  uint32_t *subport, uint32_t *pipe,
				  uint32_t *traffic_class, uint32_t *queue)
{
	const struct rte_sched_port_hierarchy *sched
		= (const struct rte_sched_port_hierarchy *) &pkt->hash.sched;

	*subport = sched->subport;
	*pipe = sched->pipe;
	*traffic_class = sched->traffic_class;
	*queue = sched->queue;
}

enum rte_meter_color
rte_sched_port_pkt_read_color(const struct rte_mbuf *pkt)
{
	const struct rte_sched_port_hierarchy *sched
		= (const struct rte_sched_port_hierarchy *) &pkt->hash.sched;

	return (enum rte_meter_color) sched->color;
}

int
rte_sched_subport_read_stats(struct rte_sched_port *port,
			     uint32_t subport_id,
			     struct rte_sched_subport_stats *stats,
			     uint32_t *tc_ov)
{
	struct rte_sched_subport *s;

	/* Check user parameters */
	if (port == NULL || subport_id >= port->n_subports_per_port ||
	    stats == NULL || tc_ov == NULL)
		return -1;

	s = port->subport + subport_id;

	/* Copy subport stats and clear */
	memcpy(stats, &s->stats, sizeof(struct rte_sched_subport_stats));
	memset(&s->stats, 0, sizeof(struct rte_sched_subport_stats));

	/* Subport TC ovesubscription status */
	*tc_ov = s->tc_ov;

	return 0;
}

int
rte_sched_queue_read_stats(struct rte_sched_port *port,
	uint32_t queue_id,
	struct rte_sched_queue_stats *stats,
	uint16_t *qlen)
{
	struct rte_sched_queue *q;
	struct rte_sched_queue_extra *qe;

	/* Check user parameters */
	if ((port == NULL) ||
	    (queue_id >= rte_sched_port_queues_per_port(port)) ||
		(stats == NULL) ||
		(qlen == NULL)) {
		return -1;
	}
	q = port->queue + queue_id;
	qe = port->queue_extra + queue_id;

	/* Copy queue stats and clear */
	memcpy(stats, &qe->stats, sizeof(struct rte_sched_queue_stats));
	memset(&qe->stats, 0, sizeof(struct rte_sched_queue_stats));

	/* Queue length */
	*qlen = q->qw - q->qr;

	return 0;
}

static inline uint32_t
rte_sched_port_qindex(struct rte_sched_port *port, uint32_t subport, uint32_t pipe, uint32_t traffic_class, uint32_t queue)
{
	uint32_t result;

	result = subport * port->n_pipes_per_subport + pipe;
	result = result * RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE + traffic_class;
	result = result * RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS + queue;

	return result;
}

#ifdef RTE_SCHED_DEBUG

static inline int
rte_sched_port_queue_is_empty(struct rte_sched_port *port, uint32_t qindex)
{
	struct rte_sched_queue *queue = port->queue + qindex;

	return queue->qr == queue->qw;
}

#endif /* RTE_SCHED_DEBUG */

#ifdef RTE_SCHED_COLLECT_STATS

static inline void
rte_sched_port_update_subport_stats(struct rte_sched_port *port, uint32_t qindex, struct rte_mbuf *pkt)
{
	struct rte_sched_subport *s = port->subport + (qindex / rte_sched_port_queues_per_subport(port));
	uint32_t tc_index = (qindex >> 2) & 0x3;
	uint32_t pkt_len = pkt->pkt_len;

	s->stats.n_pkts_tc[tc_index] += 1;
	s->stats.n_bytes_tc[tc_index] += pkt_len;
}

#ifdef RTE_SCHED_RED
static inline void
rte_sched_port_update_subport_stats_on_drop(struct rte_sched_port *port,
						uint32_t qindex,
						struct rte_mbuf *pkt, uint32_t red)
#else
static inline void
rte_sched_port_update_subport_stats_on_drop(struct rte_sched_port *port,
						uint32_t qindex,
						struct rte_mbuf *pkt, __rte_unused uint32_t red)
#endif
{
	struct rte_sched_subport *s = port->subport + (qindex / rte_sched_port_queues_per_subport(port));
	uint32_t tc_index = (qindex >> 2) & 0x3;
	uint32_t pkt_len = pkt->pkt_len;

	s->stats.n_pkts_tc_dropped[tc_index] += 1;
	s->stats.n_bytes_tc_dropped[tc_index] += pkt_len;
#ifdef RTE_SCHED_RED
	s->stats.n_pkts_red_dropped[tc_index] += red;
#endif
}

static inline void
rte_sched_port_update_queue_stats(struct rte_sched_port *port, uint32_t qindex, struct rte_mbuf *pkt)
{
	struct rte_sched_queue_extra *qe = port->queue_extra + qindex;
	uint32_t pkt_len = pkt->pkt_len;

	qe->stats.n_pkts += 1;
	qe->stats.n_bytes += pkt_len;
}

#ifdef RTE_SCHED_RED
static inline void
rte_sched_port_update_queue_stats_on_drop(struct rte_sched_port *port,
						uint32_t qindex,
						struct rte_mbuf *pkt, uint32_t red)
#else
static inline void
rte_sched_port_update_queue_stats_on_drop(struct rte_sched_port *port,
						uint32_t qindex,
						struct rte_mbuf *pkt, __rte_unused uint32_t red)
#endif
{
	struct rte_sched_queue_extra *qe = port->queue_extra + qindex;
	uint32_t pkt_len = pkt->pkt_len;

	qe->stats.n_pkts_dropped += 1;
	qe->stats.n_bytes_dropped += pkt_len;
#ifdef RTE_SCHED_RED
	qe->stats.n_pkts_red_dropped += red;
#endif
}

#endif /* RTE_SCHED_COLLECT_STATS */

#ifdef RTE_SCHED_RED

static inline int
rte_sched_port_red_drop(struct rte_sched_port *port, struct rte_mbuf *pkt, uint32_t qindex, uint16_t qlen)
{
	struct rte_sched_queue_extra *qe;
	struct rte_red_config *red_cfg;
	struct rte_red *red;
	uint32_t tc_index;
	enum rte_meter_color color;

	tc_index = (qindex >> 2) & 0x3;
	color = rte_sched_port_pkt_read_color(pkt);
	red_cfg = &port->red_config[tc_index][color];

	if ((red_cfg->min_th | red_cfg->max_th) == 0)
		return 0;

	qe = port->queue_extra + qindex;
	red = &qe->red;

	return rte_red_enqueue(red_cfg, red, qlen, port->time);
}

static inline void
rte_sched_port_set_queue_empty_timestamp(struct rte_sched_port *port, uint32_t qindex)
{
	struct rte_sched_queue_extra *qe = port->queue_extra + qindex;
	struct rte_red *red = &qe->red;

	rte_red_mark_queue_empty(red, port->time);
}

#else

#define rte_sched_port_red_drop(port, pkt, qindex, qlen)             0

#define rte_sched_port_set_queue_empty_timestamp(port, qindex)

#endif /* RTE_SCHED_RED */

#ifdef RTE_SCHED_DEBUG

static inline void
debug_check_queue_slab(struct rte_sched_port *port, uint32_t bmp_pos,
		       uint64_t bmp_slab)
{
	uint64_t mask;
	uint32_t i, panic;

	if (bmp_slab == 0)
		rte_panic("Empty slab at position %u\n", bmp_pos);

	panic = 0;
	for (i = 0, mask = 1; i < 64; i++, mask <<= 1) {
		if (mask & bmp_slab) {
			if (rte_sched_port_queue_is_empty(port, bmp_pos + i)) {
				printf("Queue %u (slab offset %u) is empty\n", bmp_pos + i, i);
				panic = 1;
			}
		}
	}

	if (panic)
		rte_panic("Empty queues in slab 0x%" PRIx64 "starting at position %u\n",
			bmp_slab, bmp_pos);
}

#endif /* RTE_SCHED_DEBUG */

static inline uint32_t
rte_sched_port_enqueue_qptrs_prefetch0(struct rte_sched_port *port,
				       struct rte_mbuf *pkt)
{
	struct rte_sched_queue *q;
#ifdef RTE_SCHED_COLLECT_STATS
	struct rte_sched_queue_extra *qe;
#endif
	uint32_t subport, pipe, traffic_class, queue, qindex;

	rte_sched_port_pkt_read_tree_path(pkt, &subport, &pipe, &traffic_class, &queue);

	qindex = rte_sched_port_qindex(port, subport, pipe, traffic_class, queue);
	q = port->queue + qindex;
	rte_prefetch0(q);
#ifdef RTE_SCHED_COLLECT_STATS
	qe = port->queue_extra + qindex;
	rte_prefetch0(qe);
#endif

	return qindex;
}

static inline void
rte_sched_port_enqueue_qwa_prefetch0(struct rte_sched_port *port,
				     uint32_t qindex, struct rte_mbuf **qbase)
{
	struct rte_sched_queue *q;
	struct rte_mbuf **q_qw;
	uint16_t qsize;

	q = port->queue + qindex;
	qsize = rte_sched_port_qsize(port, qindex);
	q_qw = qbase + (q->qw & (qsize - 1));

	rte_prefetch0(q_qw);
	rte_bitmap_prefetch0(port->bmp, qindex);
}

static inline int
rte_sched_port_enqueue_qwa(struct rte_sched_port *port, uint32_t qindex,
			   struct rte_mbuf **qbase, struct rte_mbuf *pkt)
{
	struct rte_sched_queue *q;
	uint16_t qsize;
	uint16_t qlen;

	q = port->queue + qindex;
	qsize = rte_sched_port_qsize(port, qindex);
	qlen = q->qw - q->qr;

	/* Drop the packet (and update drop stats) when queue is full */
	if (unlikely(rte_sched_port_red_drop(port, pkt, qindex, qlen) ||
		     (qlen >= qsize))) {
		rte_pktmbuf_free(pkt);
#ifdef RTE_SCHED_COLLECT_STATS
		rte_sched_port_update_subport_stats_on_drop(port, qindex, pkt,
							    qlen < qsize);
		rte_sched_port_update_queue_stats_on_drop(port, qindex, pkt,
							  qlen < qsize);
#endif
		return 0;
	}

	/* Enqueue packet */
	qbase[q->qw & (qsize - 1)] = pkt;
	q->qw++;

	/* Activate queue in the port bitmap */
	rte_bitmap_set(port->bmp, qindex);

	/* Statistics */
#ifdef RTE_SCHED_COLLECT_STATS
	rte_sched_port_update_subport_stats(port, qindex, pkt);
	rte_sched_port_update_queue_stats(port, qindex, pkt);
#endif

	return 1;
}


/*
 * The enqueue function implements a 4-level pipeline with each stage
 * processing two different packets. The purpose of using a pipeline
 * is to hide the latency of prefetching the data structures. The
 * naming convention is presented in the diagram below:
 *
 *   p00  _______   p10  _______   p20  _______   p30  _______
 * ----->|       |----->|       |----->|       |----->|       |----->
 *       |   0   |      |   1   |      |   2   |      |   3   |
 * ----->|_______|----->|_______|----->|_______|----->|_______|----->
 *   p01            p11            p21            p31
 *
 */
int
rte_sched_port_enqueue(struct rte_sched_port *port, struct rte_mbuf **pkts,
		       uint32_t n_pkts)
{
	struct rte_mbuf *pkt00, *pkt01, *pkt10, *pkt11, *pkt20, *pkt21,
		*pkt30, *pkt31, *pkt_last;
	struct rte_mbuf **q00_base, **q01_base, **q10_base, **q11_base,
		**q20_base, **q21_base, **q30_base, **q31_base, **q_last_base;
	uint32_t q00, q01, q10, q11, q20, q21, q30, q31, q_last;
	uint32_t r00, r01, r10, r11, r20, r21, r30, r31, r_last;
	uint32_t result, i;

	result = 0;

	/*
	 * Less then 6 input packets available, which is not enough to
	 * feed the pipeline
	 */
	if (unlikely(n_pkts < 6)) {
		struct rte_mbuf **q_base[5];
		uint32_t q[5];

		/* Prefetch the mbuf structure of each packet */
		for (i = 0; i < n_pkts; i++)
			rte_prefetch0(pkts[i]);

		/* Prefetch the queue structure for each queue */
		for (i = 0; i < n_pkts; i++)
			q[i] = rte_sched_port_enqueue_qptrs_prefetch0(port,
								      pkts[i]);

		/* Prefetch the write pointer location of each queue */
		for (i = 0; i < n_pkts; i++) {
			q_base[i] = rte_sched_port_qbase(port, q[i]);
			rte_sched_port_enqueue_qwa_prefetch0(port, q[i],
							     q_base[i]);
		}

		/* Write each packet to its queue */
		for (i = 0; i < n_pkts; i++)
			result += rte_sched_port_enqueue_qwa(port, q[i],
							     q_base[i], pkts[i]);

		return result;
	}

	/* Feed the first 3 stages of the pipeline (6 packets needed) */
	pkt20 = pkts[0];
	pkt21 = pkts[1];
	rte_prefetch0(pkt20);
	rte_prefetch0(pkt21);

	pkt10 = pkts[2];
	pkt11 = pkts[3];
	rte_prefetch0(pkt10);
	rte_prefetch0(pkt11);

	q20 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt20);
	q21 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt21);

	pkt00 = pkts[4];
	pkt01 = pkts[5];
	rte_prefetch0(pkt00);
	rte_prefetch0(pkt01);

	q10 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt10);
	q11 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt11);

	q20_base = rte_sched_port_qbase(port, q20);
	q21_base = rte_sched_port_qbase(port, q21);
	rte_sched_port_enqueue_qwa_prefetch0(port, q20, q20_base);
	rte_sched_port_enqueue_qwa_prefetch0(port, q21, q21_base);

	/* Run the pipeline */
	for (i = 6; i < (n_pkts & (~1)); i += 2) {
		/* Propagate stage inputs */
		pkt30 = pkt20;
		pkt31 = pkt21;
		pkt20 = pkt10;
		pkt21 = pkt11;
		pkt10 = pkt00;
		pkt11 = pkt01;
		q30 = q20;
		q31 = q21;
		q20 = q10;
		q21 = q11;
		q30_base = q20_base;
		q31_base = q21_base;

		/* Stage 0: Get packets in */
		pkt00 = pkts[i];
		pkt01 = pkts[i + 1];
		rte_prefetch0(pkt00);
		rte_prefetch0(pkt01);

		/* Stage 1: Prefetch queue structure storing queue pointers */
		q10 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt10);
		q11 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt11);

		/* Stage 2: Prefetch queue write location */
		q20_base = rte_sched_port_qbase(port, q20);
		q21_base = rte_sched_port_qbase(port, q21);
		rte_sched_port_enqueue_qwa_prefetch0(port, q20, q20_base);
		rte_sched_port_enqueue_qwa_prefetch0(port, q21, q21_base);

		/* Stage 3: Write packet to queue and activate queue */
		r30 = rte_sched_port_enqueue_qwa(port, q30, q30_base, pkt30);
		r31 = rte_sched_port_enqueue_qwa(port, q31, q31_base, pkt31);
		result += r30 + r31;
	}

	/*
	 * Drain the pipeline (exactly 6 packets).
	 * Handle the last packet in the case
	 * of an odd number of input packets.
	 */
	pkt_last = pkts[n_pkts - 1];
	rte_prefetch0(pkt_last);

	q00 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt00);
	q01 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt01);

	q10_base = rte_sched_port_qbase(port, q10);
	q11_base = rte_sched_port_qbase(port, q11);
	rte_sched_port_enqueue_qwa_prefetch0(port, q10, q10_base);
	rte_sched_port_enqueue_qwa_prefetch0(port, q11, q11_base);

	r20 = rte_sched_port_enqueue_qwa(port, q20, q20_base, pkt20);
	r21 = rte_sched_port_enqueue_qwa(port, q21, q21_base, pkt21);
	result += r20 + r21;

	q_last = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt_last);

	q00_base = rte_sched_port_qbase(port, q00);
	q01_base = rte_sched_port_qbase(port, q01);
	rte_sched_port_enqueue_qwa_prefetch0(port, q00, q00_base);
	rte_sched_port_enqueue_qwa_prefetch0(port, q01, q01_base);

	r10 = rte_sched_port_enqueue_qwa(port, q10, q10_base, pkt10);
	r11 = rte_sched_port_enqueue_qwa(port, q11, q11_base, pkt11);
	result += r10 + r11;

	q_last_base = rte_sched_port_qbase(port, q_last);
	rte_sched_port_enqueue_qwa_prefetch0(port, q_last, q_last_base);

	r00 = rte_sched_port_enqueue_qwa(port, q00, q00_base, pkt00);
	r01 = rte_sched_port_enqueue_qwa(port, q01, q01_base, pkt01);
	result += r00 + r01;

	if (n_pkts & 1) {
		r_last = rte_sched_port_enqueue_qwa(port, q_last, q_last_base, pkt_last);
		result += r_last;
	}

	return result;
}

#ifndef RTE_SCHED_SUBPORT_TC_OV

static inline void
grinder_credits_update(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	struct rte_sched_subport *subport = grinder->subport;
	struct rte_sched_pipe *pipe = grinder->pipe;
	struct rte_sched_pipe_profile *params = grinder->pipe_params;
	uint64_t n_periods;

	/* Subport TB */
	n_periods = (port->time - subport->tb_time) / subport->tb_period;
	subport->tb_credits += n_periods * subport->tb_credits_per_period;
	subport->tb_credits = rte_sched_min_val_2_u32(subport->tb_credits, subport->tb_size);
	subport->tb_time += n_periods * subport->tb_period;

	/* Pipe TB */
	n_periods = (port->time - pipe->tb_time) / params->tb_period;
	pipe->tb_credits += n_periods * params->tb_credits_per_period;
	pipe->tb_credits = rte_sched_min_val_2_u32(pipe->tb_credits, params->tb_size);
	pipe->tb_time += n_periods * params->tb_period;

	/* Subport TCs */
	if (unlikely(port->time >= subport->tc_time)) {
		subport->tc_credits[0] = subport->tc_credits_per_period[0];
		subport->tc_credits[1] = subport->tc_credits_per_period[1];
		subport->tc_credits[2] = subport->tc_credits_per_period[2];
		subport->tc_credits[3] = subport->tc_credits_per_period[3];
		subport->tc_time = port->time + subport->tc_period;
	}

	/* Pipe TCs */
	if (unlikely(port->time >= pipe->tc_time)) {
		pipe->tc_credits[0] = params->tc_credits_per_period[0];
		pipe->tc_credits[1] = params->tc_credits_per_period[1];
		pipe->tc_credits[2] = params->tc_credits_per_period[2];
		pipe->tc_credits[3] = params->tc_credits_per_period[3];
		pipe->tc_time = port->time + params->tc_period;
	}
}

#else

static inline uint32_t
grinder_tc_ov_credits_update(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	struct rte_sched_subport *subport = grinder->subport;
	uint32_t tc_ov_consumption[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
	uint32_t tc_ov_consumption_max;
	uint32_t tc_ov_wm = subport->tc_ov_wm;

	if (subport->tc_ov == 0)
		return subport->tc_ov_wm_max;

	tc_ov_consumption[0] = subport->tc_credits_per_period[0] - subport->tc_credits[0];
	tc_ov_consumption[1] = subport->tc_credits_per_period[1] - subport->tc_credits[1];
	tc_ov_consumption[2] = subport->tc_credits_per_period[2] - subport->tc_credits[2];
	tc_ov_consumption[3] = subport->tc_credits_per_period[3] - subport->tc_credits[3];

	tc_ov_consumption_max = subport->tc_credits_per_period[3] -
		(tc_ov_consumption[0] + tc_ov_consumption[1] + tc_ov_consumption[2]);

	if (tc_ov_consumption[3] > (tc_ov_consumption_max - port->mtu)) {
		tc_ov_wm  -= tc_ov_wm >> 7;
		if (tc_ov_wm < subport->tc_ov_wm_min)
			tc_ov_wm = subport->tc_ov_wm_min;

		return tc_ov_wm;
	}

	tc_ov_wm += (tc_ov_wm >> 7) + 1;
	if (tc_ov_wm > subport->tc_ov_wm_max)
		tc_ov_wm = subport->tc_ov_wm_max;

	return tc_ov_wm;
}

static inline void
grinder_credits_update(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	struct rte_sched_subport *subport = grinder->subport;
	struct rte_sched_pipe *pipe = grinder->pipe;
	struct rte_sched_pipe_profile *params = grinder->pipe_params;
	uint64_t n_periods;

	/* Subport TB */
	n_periods = (port->time - subport->tb_time) / subport->tb_period;
	subport->tb_credits += n_periods * subport->tb_credits_per_period;
	subport->tb_credits = rte_sched_min_val_2_u32(subport->tb_credits, subport->tb_size);
	subport->tb_time += n_periods * subport->tb_period;

	/* Pipe TB */
	n_periods = (port->time - pipe->tb_time) / params->tb_period;
	pipe->tb_credits += n_periods * params->tb_credits_per_period;
	pipe->tb_credits = rte_sched_min_val_2_u32(pipe->tb_credits, params->tb_size);
	pipe->tb_time += n_periods * params->tb_period;

	/* Subport TCs */
	if (unlikely(port->time >= subport->tc_time)) {
		subport->tc_ov_wm = grinder_tc_ov_credits_update(port, pos);

		subport->tc_credits[0] = subport->tc_credits_per_period[0];
		subport->tc_credits[1] = subport->tc_credits_per_period[1];
		subport->tc_credits[2] = subport->tc_credits_per_period[2];
		subport->tc_credits[3] = subport->tc_credits_per_period[3];

		subport->tc_time = port->time + subport->tc_period;
		subport->tc_ov_period_id++;
	}

	/* Pipe TCs */
	if (unlikely(port->time >= pipe->tc_time)) {
		pipe->tc_credits[0] = params->tc_credits_per_period[0];
		pipe->tc_credits[1] = params->tc_credits_per_period[1];
		pipe->tc_credits[2] = params->tc_credits_per_period[2];
		pipe->tc_credits[3] = params->tc_credits_per_period[3];
		pipe->tc_time = port->time + params->tc_period;
	}

	/* Pipe TCs - Oversubscription */
	if (unlikely(pipe->tc_ov_period_id != subport->tc_ov_period_id)) {
		pipe->tc_ov_credits = subport->tc_ov_wm * params->tc_ov_weight;

		pipe->tc_ov_period_id = subport->tc_ov_period_id;
	}
}

#endif /* RTE_SCHED_TS_CREDITS_UPDATE, RTE_SCHED_SUBPORT_TC_OV */


#ifndef RTE_SCHED_SUBPORT_TC_OV

static inline int
grinder_credits_check(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	struct rte_sched_subport *subport = grinder->subport;
	struct rte_sched_pipe *pipe = grinder->pipe;
	struct rte_mbuf *pkt = grinder->pkt;
	uint32_t tc_index = grinder->tc_index;
	uint32_t pkt_len = pkt->pkt_len + port->frame_overhead;
	uint32_t subport_tb_credits = subport->tb_credits;
	uint32_t subport_tc_credits = subport->tc_credits[tc_index];
	uint32_t pipe_tb_credits = pipe->tb_credits;
	uint32_t pipe_tc_credits = pipe->tc_credits[tc_index];
	int enough_credits;

	/* Check queue credits */
	enough_credits = (pkt_len <= subport_tb_credits) &&
		(pkt_len <= subport_tc_credits) &&
		(pkt_len <= pipe_tb_credits) &&
		(pkt_len <= pipe_tc_credits);

	if (!enough_credits)
		return 0;

	/* Update port credits */
	subport->tb_credits -= pkt_len;
	subport->tc_credits[tc_index] -= pkt_len;
	pipe->tb_credits -= pkt_len;
	pipe->tc_credits[tc_index] -= pkt_len;

	return 1;
}

#else

static inline int
grinder_credits_check(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	struct rte_sched_subport *subport = grinder->subport;
	struct rte_sched_pipe *pipe = grinder->pipe;
	struct rte_mbuf *pkt = grinder->pkt;
	uint32_t tc_index = grinder->tc_index;
	uint32_t pkt_len = pkt->pkt_len + port->frame_overhead;
	uint32_t subport_tb_credits = subport->tb_credits;
	uint32_t subport_tc_credits = subport->tc_credits[tc_index];
	uint32_t pipe_tb_credits = pipe->tb_credits;
	uint32_t pipe_tc_credits = pipe->tc_credits[tc_index];
	uint32_t pipe_tc_ov_mask1[] = {UINT32_MAX, UINT32_MAX, UINT32_MAX, pipe->tc_ov_credits};
	uint32_t pipe_tc_ov_mask2[] = {0, 0, 0, UINT32_MAX};
	uint32_t pipe_tc_ov_credits = pipe_tc_ov_mask1[tc_index];
	int enough_credits;

	/* Check pipe and subport credits */
	enough_credits = (pkt_len <= subport_tb_credits) &&
		(pkt_len <= subport_tc_credits) &&
		(pkt_len <= pipe_tb_credits) &&
		(pkt_len <= pipe_tc_credits) &&
		(pkt_len <= pipe_tc_ov_credits);

	if (!enough_credits)
		return 0;

	/* Update pipe and subport credits */
	subport->tb_credits -= pkt_len;
	subport->tc_credits[tc_index] -= pkt_len;
	pipe->tb_credits -= pkt_len;
	pipe->tc_credits[tc_index] -= pkt_len;
	pipe->tc_ov_credits -= pipe_tc_ov_mask2[tc_index] & pkt_len;

	return 1;
}

#endif /* RTE_SCHED_SUBPORT_TC_OV */


static inline int
grinder_schedule(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	struct rte_sched_queue *queue = grinder->queue[grinder->qpos];
	struct rte_mbuf *pkt = grinder->pkt;
	uint32_t pkt_len = pkt->pkt_len + port->frame_overhead;

	if (!grinder_credits_check(port, pos))
		return 0;

	/* Advance port time */
	port->time += pkt_len;

	/* Send packet */
	port->pkts_out[port->n_pkts_out++] = pkt;
	queue->qr++;
	grinder->wrr_tokens[grinder->qpos] += pkt_len * grinder->wrr_cost[grinder->qpos];
	if (queue->qr == queue->qw) {
		uint32_t qindex = grinder->qindex[grinder->qpos];

		rte_bitmap_clear(port->bmp, qindex);
		grinder->qmask &= ~(1 << grinder->qpos);
		grinder->wrr_mask[grinder->qpos] = 0;
		rte_sched_port_set_queue_empty_timestamp(port, qindex);
	}

	/* Reset pipe loop detection */
	port->pipe_loop = RTE_SCHED_PIPE_INVALID;
	grinder->productive = 1;

	return 1;
}

#ifdef SCHED_VECTOR_SSE4

static inline int
grinder_pipe_exists(struct rte_sched_port *port, uint32_t base_pipe)
{
	__m128i index = _mm_set1_epi32(base_pipe);
	__m128i pipes = _mm_load_si128((__m128i *)port->grinder_base_bmp_pos);
	__m128i res = _mm_cmpeq_epi32(pipes, index);

	pipes = _mm_load_si128((__m128i *)(port->grinder_base_bmp_pos + 4));
	pipes = _mm_cmpeq_epi32(pipes, index);
	res = _mm_or_si128(res, pipes);

	if (_mm_testz_si128(res, res))
		return 0;

	return 1;
}

#elif defined(SCHED_VECTOR_NEON)

static inline int
grinder_pipe_exists(struct rte_sched_port *port, uint32_t base_pipe)
{
	uint32x4_t index, pipes;
	uint32_t *pos = (uint32_t *)port->grinder_base_bmp_pos;

	index = vmovq_n_u32(base_pipe);
	pipes = vld1q_u32(pos);
	if (!vminvq_u32(veorq_u32(pipes, index)))
		return 1;

	pipes = vld1q_u32(pos + 4);
	if (!vminvq_u32(veorq_u32(pipes, index)))
		return 1;

	return 0;
}

#else

static inline int
grinder_pipe_exists(struct rte_sched_port *port, uint32_t base_pipe)
{
	uint32_t i;

	for (i = 0; i < RTE_SCHED_PORT_N_GRINDERS; i++) {
		if (port->grinder_base_bmp_pos[i] == base_pipe)
			return 1;
	}

	return 0;
}

#endif /* RTE_SCHED_OPTIMIZATIONS */

static inline void
grinder_pcache_populate(struct rte_sched_port *port, uint32_t pos, uint32_t bmp_pos, uint64_t bmp_slab)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	uint16_t w[4];

	grinder->pcache_w = 0;
	grinder->pcache_r = 0;

	w[0] = (uint16_t) bmp_slab;
	w[1] = (uint16_t) (bmp_slab >> 16);
	w[2] = (uint16_t) (bmp_slab >> 32);
	w[3] = (uint16_t) (bmp_slab >> 48);

	grinder->pcache_qmask[grinder->pcache_w] = w[0];
	grinder->pcache_qindex[grinder->pcache_w] = bmp_pos;
	grinder->pcache_w += (w[0] != 0);

	grinder->pcache_qmask[grinder->pcache_w] = w[1];
	grinder->pcache_qindex[grinder->pcache_w] = bmp_pos + 16;
	grinder->pcache_w += (w[1] != 0);

	grinder->pcache_qmask[grinder->pcache_w] = w[2];
	grinder->pcache_qindex[grinder->pcache_w] = bmp_pos + 32;
	grinder->pcache_w += (w[2] != 0);

	grinder->pcache_qmask[grinder->pcache_w] = w[3];
	grinder->pcache_qindex[grinder->pcache_w] = bmp_pos + 48;
	grinder->pcache_w += (w[3] != 0);
}

static inline void
grinder_tccache_populate(struct rte_sched_port *port, uint32_t pos, uint32_t qindex, uint16_t qmask)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	uint8_t b[4];

	grinder->tccache_w = 0;
	grinder->tccache_r = 0;

	b[0] = (uint8_t) (qmask & 0xF);
	b[1] = (uint8_t) ((qmask >> 4) & 0xF);
	b[2] = (uint8_t) ((qmask >> 8) & 0xF);
	b[3] = (uint8_t) ((qmask >> 12) & 0xF);

	grinder->tccache_qmask[grinder->tccache_w] = b[0];
	grinder->tccache_qindex[grinder->tccache_w] = qindex;
	grinder->tccache_w += (b[0] != 0);

	grinder->tccache_qmask[grinder->tccache_w] = b[1];
	grinder->tccache_qindex[grinder->tccache_w] = qindex + 4;
	grinder->tccache_w += (b[1] != 0);

	grinder->tccache_qmask[grinder->tccache_w] = b[2];
	grinder->tccache_qindex[grinder->tccache_w] = qindex + 8;
	grinder->tccache_w += (b[2] != 0);

	grinder->tccache_qmask[grinder->tccache_w] = b[3];
	grinder->tccache_qindex[grinder->tccache_w] = qindex + 12;
	grinder->tccache_w += (b[3] != 0);
}

static inline int
grinder_next_tc(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	struct rte_mbuf **qbase;
	uint32_t qindex;
	uint16_t qsize;

	if (grinder->tccache_r == grinder->tccache_w)
		return 0;

	qindex = grinder->tccache_qindex[grinder->tccache_r];
	qbase = rte_sched_port_qbase(port, qindex);
	qsize = rte_sched_port_qsize(port, qindex);

	grinder->tc_index = (qindex >> 2) & 0x3;
	grinder->qmask = grinder->tccache_qmask[grinder->tccache_r];
	grinder->qsize = qsize;

	grinder->qindex[0] = qindex;
	grinder->qindex[1] = qindex + 1;
	grinder->qindex[2] = qindex + 2;
	grinder->qindex[3] = qindex + 3;

	grinder->queue[0] = port->queue + qindex;
	grinder->queue[1] = port->queue + qindex + 1;
	grinder->queue[2] = port->queue + qindex + 2;
	grinder->queue[3] = port->queue + qindex + 3;

	grinder->qbase[0] = qbase;
	grinder->qbase[1] = qbase + qsize;
	grinder->qbase[2] = qbase + 2 * qsize;
	grinder->qbase[3] = qbase + 3 * qsize;

	grinder->tccache_r++;
	return 1;
}

static inline int
grinder_next_pipe(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	uint32_t pipe_qindex;
	uint16_t pipe_qmask;

	if (grinder->pcache_r < grinder->pcache_w) {
		pipe_qmask = grinder->pcache_qmask[grinder->pcache_r];
		pipe_qindex = grinder->pcache_qindex[grinder->pcache_r];
		grinder->pcache_r++;
	} else {
		uint64_t bmp_slab = 0;
		uint32_t bmp_pos = 0;

		/* Get another non-empty pipe group */
		if (unlikely(rte_bitmap_scan(port->bmp, &bmp_pos, &bmp_slab) <= 0))
			return 0;

#ifdef RTE_SCHED_DEBUG
		debug_check_queue_slab(port, bmp_pos, bmp_slab);
#endif

		/* Return if pipe group already in one of the other grinders */
		port->grinder_base_bmp_pos[pos] = RTE_SCHED_BMP_POS_INVALID;
		if (unlikely(grinder_pipe_exists(port, bmp_pos)))
			return 0;

		port->grinder_base_bmp_pos[pos] = bmp_pos;

		/* Install new pipe group into grinder's pipe cache */
		grinder_pcache_populate(port, pos, bmp_pos, bmp_slab);

		pipe_qmask = grinder->pcache_qmask[0];
		pipe_qindex = grinder->pcache_qindex[0];
		grinder->pcache_r = 1;
	}

	/* Install new pipe in the grinder */
	grinder->pindex = pipe_qindex >> 4;
	grinder->subport = port->subport + (grinder->pindex / port->n_pipes_per_subport);
	grinder->pipe = port->pipe + grinder->pindex;
	grinder->pipe_params = NULL; /* to be set after the pipe structure is prefetched */
	grinder->productive = 0;

	grinder_tccache_populate(port, pos, pipe_qindex, pipe_qmask);
	grinder_next_tc(port, pos);

	/* Check for pipe exhaustion */
	if (grinder->pindex == port->pipe_loop) {
		port->pipe_exhaustion = 1;
		port->pipe_loop = RTE_SCHED_PIPE_INVALID;
	}

	return 1;
}


static inline void
grinder_wrr_load(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	struct rte_sched_pipe *pipe = grinder->pipe;
	struct rte_sched_pipe_profile *pipe_params = grinder->pipe_params;
	uint32_t tc_index = grinder->tc_index;
	uint32_t qmask = grinder->qmask;
	uint32_t qindex;

	qindex = tc_index * 4;

	grinder->wrr_tokens[0] = ((uint16_t) pipe->wrr_tokens[qindex]) << RTE_SCHED_WRR_SHIFT;
	grinder->wrr_tokens[1] = ((uint16_t) pipe->wrr_tokens[qindex + 1]) << RTE_SCHED_WRR_SHIFT;
	grinder->wrr_tokens[2] = ((uint16_t) pipe->wrr_tokens[qindex + 2]) << RTE_SCHED_WRR_SHIFT;
	grinder->wrr_tokens[3] = ((uint16_t) pipe->wrr_tokens[qindex + 3]) << RTE_SCHED_WRR_SHIFT;

	grinder->wrr_mask[0] = (qmask & 0x1) * 0xFFFF;
	grinder->wrr_mask[1] = ((qmask >> 1) & 0x1) * 0xFFFF;
	grinder->wrr_mask[2] = ((qmask >> 2) & 0x1) * 0xFFFF;
	grinder->wrr_mask[3] = ((qmask >> 3) & 0x1) * 0xFFFF;

	grinder->wrr_cost[0] = pipe_params->wrr_cost[qindex];
	grinder->wrr_cost[1] = pipe_params->wrr_cost[qindex + 1];
	grinder->wrr_cost[2] = pipe_params->wrr_cost[qindex + 2];
	grinder->wrr_cost[3] = pipe_params->wrr_cost[qindex + 3];
}

static inline void
grinder_wrr_store(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	struct rte_sched_pipe *pipe = grinder->pipe;
	uint32_t tc_index = grinder->tc_index;
	uint32_t qindex;

	qindex = tc_index * 4;

	pipe->wrr_tokens[qindex] = (grinder->wrr_tokens[0] & grinder->wrr_mask[0])
		>> RTE_SCHED_WRR_SHIFT;
	pipe->wrr_tokens[qindex + 1] = (grinder->wrr_tokens[1] & grinder->wrr_mask[1])
		>> RTE_SCHED_WRR_SHIFT;
	pipe->wrr_tokens[qindex + 2] = (grinder->wrr_tokens[2] & grinder->wrr_mask[2])
		>> RTE_SCHED_WRR_SHIFT;
	pipe->wrr_tokens[qindex + 3] = (grinder->wrr_tokens[3] & grinder->wrr_mask[3])
		>> RTE_SCHED_WRR_SHIFT;
}

static inline void
grinder_wrr(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	uint16_t wrr_tokens_min;

	grinder->wrr_tokens[0] |= ~grinder->wrr_mask[0];
	grinder->wrr_tokens[1] |= ~grinder->wrr_mask[1];
	grinder->wrr_tokens[2] |= ~grinder->wrr_mask[2];
	grinder->wrr_tokens[3] |= ~grinder->wrr_mask[3];

	grinder->qpos = rte_min_pos_4_u16(grinder->wrr_tokens);
	wrr_tokens_min = grinder->wrr_tokens[grinder->qpos];

	grinder->wrr_tokens[0] -= wrr_tokens_min;
	grinder->wrr_tokens[1] -= wrr_tokens_min;
	grinder->wrr_tokens[2] -= wrr_tokens_min;
	grinder->wrr_tokens[3] -= wrr_tokens_min;
}


#define grinder_evict(port, pos)

static inline void
grinder_prefetch_pipe(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;

	rte_prefetch0(grinder->pipe);
	rte_prefetch0(grinder->queue[0]);
}

static inline void
grinder_prefetch_tc_queue_arrays(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	uint16_t qsize, qr[4];

	qsize = grinder->qsize;
	qr[0] = grinder->queue[0]->qr & (qsize - 1);
	qr[1] = grinder->queue[1]->qr & (qsize - 1);
	qr[2] = grinder->queue[2]->qr & (qsize - 1);
	qr[3] = grinder->queue[3]->qr & (qsize - 1);

	rte_prefetch0(grinder->qbase[0] + qr[0]);
	rte_prefetch0(grinder->qbase[1] + qr[1]);

	grinder_wrr_load(port, pos);
	grinder_wrr(port, pos);

	rte_prefetch0(grinder->qbase[2] + qr[2]);
	rte_prefetch0(grinder->qbase[3] + qr[3]);
}

static inline void
grinder_prefetch_mbuf(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;
	uint32_t qpos = grinder->qpos;
	struct rte_mbuf **qbase = grinder->qbase[qpos];
	uint16_t qsize = grinder->qsize;
	uint16_t qr = grinder->queue[qpos]->qr & (qsize - 1);

	grinder->pkt = qbase[qr];
	rte_prefetch0(grinder->pkt);

	if (unlikely((qr & 0x7) == 7)) {
		uint16_t qr_next = (grinder->queue[qpos]->qr + 1) & (qsize - 1);

		rte_prefetch0(qbase + qr_next);
	}
}

static inline uint32_t
grinder_handle(struct rte_sched_port *port, uint32_t pos)
{
	struct rte_sched_grinder *grinder = port->grinder + pos;

	switch (grinder->state) {
	case e_GRINDER_PREFETCH_PIPE:
	{
		if (grinder_next_pipe(port, pos)) {
			grinder_prefetch_pipe(port, pos);
			port->busy_grinders++;

			grinder->state = e_GRINDER_PREFETCH_TC_QUEUE_ARRAYS;
			return 0;
		}

		return 0;
	}

	case e_GRINDER_PREFETCH_TC_QUEUE_ARRAYS:
	{
		struct rte_sched_pipe *pipe = grinder->pipe;

		grinder->pipe_params = port->pipe_profiles + pipe->profile;
		grinder_prefetch_tc_queue_arrays(port, pos);
		grinder_credits_update(port, pos);

		grinder->state = e_GRINDER_PREFETCH_MBUF;
		return 0;
	}

	case e_GRINDER_PREFETCH_MBUF:
	{
		grinder_prefetch_mbuf(port, pos);

		grinder->state = e_GRINDER_READ_MBUF;
		return 0;
	}

	case e_GRINDER_READ_MBUF:
	{
		uint32_t result = 0;

		result = grinder_schedule(port, pos);

		/* Look for next packet within the same TC */
		if (result && grinder->qmask) {
			grinder_wrr(port, pos);
			grinder_prefetch_mbuf(port, pos);

			return 1;
		}
		grinder_wrr_store(port, pos);

		/* Look for another active TC within same pipe */
		if (grinder_next_tc(port, pos)) {
			grinder_prefetch_tc_queue_arrays(port, pos);

			grinder->state = e_GRINDER_PREFETCH_MBUF;
			return result;
		}

		if (grinder->productive == 0 &&
		    port->pipe_loop == RTE_SCHED_PIPE_INVALID)
			port->pipe_loop = grinder->pindex;

		grinder_evict(port, pos);

		/* Look for another active pipe */
		if (grinder_next_pipe(port, pos)) {
			grinder_prefetch_pipe(port, pos);

			grinder->state = e_GRINDER_PREFETCH_TC_QUEUE_ARRAYS;
			return result;
		}

		/* No active pipe found */
		port->busy_grinders--;

		grinder->state = e_GRINDER_PREFETCH_PIPE;
		return result;
	}

	default:
		rte_panic("Algorithmic error (invalid state)\n");
		return 0;
	}
}

static inline void
rte_sched_port_time_resync(struct rte_sched_port *port)
{
	uint64_t cycles = rte_get_tsc_cycles();
	uint64_t cycles_diff = cycles - port->time_cpu_cycles;
	uint64_t bytes_diff;

	/* Compute elapsed time in bytes */
	bytes_diff = rte_reciprocal_divide(cycles_diff << RTE_SCHED_TIME_SHIFT,
					   port->inv_cycles_per_byte);

	/* Advance port time */
	port->time_cpu_cycles = cycles;
	port->time_cpu_bytes += bytes_diff;
	if (port->time < port->time_cpu_bytes)
		port->time = port->time_cpu_bytes;

	/* Reset pipe loop detection */
	port->pipe_loop = RTE_SCHED_PIPE_INVALID;
}

static inline int
rte_sched_port_exceptions(struct rte_sched_port *port, int second_pass)
{
	int exceptions;

	/* Check if any exception flag is set */
	exceptions = (second_pass && port->busy_grinders == 0) ||
		(port->pipe_exhaustion == 1);

	/* Clear exception flags */
	port->pipe_exhaustion = 0;

	return exceptions;
}

int
rte_sched_port_dequeue(struct rte_sched_port *port, struct rte_mbuf **pkts, uint32_t n_pkts)
{
	uint32_t i, count;

	port->pkts_out = pkts;
	port->n_pkts_out = 0;

	rte_sched_port_time_resync(port);

	/* Take each queue in the grinder one step further */
	for (i = 0, count = 0; ; i++)  {
		count += grinder_handle(port, i & (RTE_SCHED_PORT_N_GRINDERS - 1));
		if ((count == n_pkts) ||
		    rte_sched_port_exceptions(port, i >= RTE_SCHED_PORT_N_GRINDERS)) {
			break;
		}
	}

	return count;
}