DPDK patches and discussions
 help / color / mirror / Atom feed
From: "Dewar, Alan" <ad759e@intl.att.com>
To: "'Dumitrescu, Cristian'" <cristian.dumitrescu@intel.com>,
	"'alangordondewar@gmail.com'" <alangordondewar@gmail.com>
Cc: "'dev@dpdk.org'" <dev@dpdk.org>, "'Alan Dewar'" <alan.dewar@att.com>
Subject: Re: [dpdk-dev] [PATCH v2] test: new sched WRR unit-test
Date: Wed, 3 Jan 2018 14:02:40 +0000	[thread overview]
Message-ID: <3F9268EEC0E43747A5FFFC6B48EF0321FBE62D@gbcdcmbx03.intl.att.com> (raw)
In-Reply-To: <3EB4FA525960D640B5BDFFD6A3D891267BAF7EDA@IRSMSX108.ger.corp.intel.com>

Hi Cristian,

In response to your comment at the very end, "Please keep the unit test in the same patch set with the code", any revised change will include the new unit-tests.

Regards
Alan

> -----Original Message-----
> From: Dumitrescu, Cristian [mailto:cristian.dumitrescu@intel.com] 
> Sent: Tuesday, January 02, 2018 4:15 PM
> To: alangordondewar@gmail.com
> Cc: dev@dpdk.org; Alan Dewar <alan.dewar@att.com>
> Subject: RE: [PATCH v2] test: new sched WRR unit-test
>
>
>
> > -----Original Message-----
> > From: alangordondewar@gmail.com [mailto:alangordondewar@gmail.com]
> > Sent: Thursday, November 30, 2017 9:05 AM
> > To: Dumitrescu, Cristian <cristian.dumitrescu@intel.com>
> > Cc: dev@dpdk.org; Alan Dewar <alan.dewar@att.com>
> > Subject: [PATCH v2] test: new sched WRR unit-test
> > 
> > From: Alan Dewar <alan.dewar@att.com>
> > 
> > New unit-test for the librte_sched WRR weighting code.
> > 
> > With the standard 17.11 code, the first three sub-tests pass, but the 
> > last three fail due to bugs in the WRR weighting code.
> > 
> > With v1 of the "sched: fix overflow errors in WRR weighting code"
> > patch the first five sub-tests pass, and the last sub-test fails badly.
> > 
> > With v2 of the "sched: fix overflow errors in WRR weighting code"
> > patch the first five sub-tests pass, and the last sub-test is a very 
> > near miss (i.e. measured packets counts are one away from the expected 
> > counts).
> > 
> > Signed-off-by: Alan Dewar <alan.dewar@att.com>
> > ---
> > v2 - add new 255-254-253-1 weightings sub-test
> > 
> >  test/test/Makefile         |   1 +
> >  test/test/test_sched_wrr.c | 491
> > +++++++++++++++++++++++++++++++++++++++++++++
> >  2 files changed, 492 insertions(+)
> >  create mode 100644 test/test/test_sched_wrr.c
> > 
> > diff --git a/test/test/Makefile b/test/test/Makefile index 
> > bb54c98..0ab0ed3 100644
> > --- a/test/test/Makefile
> > +++ b/test/test/Makefile
> > @@ -173,6 +173,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_NET) += test_crc.c  ifeq 
> > ($(CONFIG_RTE_LIBRTE_SCHED),y)  SRCS-y += test_red.c  SRCS-y += 
> > test_sched.c
> > +SRCS-y += test_sched_wrr.c
> >  endif
> > 
> >  SRCS-$(CONFIG_RTE_LIBRTE_METER) += test_meter.c diff --git 
> > a/test/test/test_sched_wrr.c b/test/test/test_sched_wrr.c new file 
> > mode 100644 index 0000000..df5a231
> > --- /dev/null
> > +++ b/test/test/test_sched_wrr.c
> > @@ -0,0 +1,491 @@
> > +/*-
> > + *   BSD LICENSE
> > + *
> > + *   Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
> > + *   Copyright(c) 2017 ATT Intellectual Property. 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 <stdlib.h>
> > +#include <stdio.h>
> > +#include <string.h>
> > +#include <stdint.h>
> > +#include <unistd.h>
> > +
> > +#include "test.h"
> > +
> > +#include <rte_cycles.h>
> > +#include <rte_ether.h>
> > +#include <rte_ip.h>
> > +#include <rte_byteorder.h>
> > +#include <rte_sched.h>
> > +
> > +
> > +#define SUBPORT         0
> > +#define PIPE            1
> > +#define TC              0
> > +#define QUEUE           3
> > +
> > +static struct rte_sched_subport_params subport_param[] = {
> > +	{
> > +		.tb_rate = 1250000000,
> > +		.tb_size = 1000000,
> > +
> > +		.tc_rate = {1250000000, 1250000000, 1250000000,
> > 1250000000},
> > +		.tc_period = 10,
> > +	},
> > +};
> > +
> > +static struct rte_sched_pipe_params pipe_profile[] = {
> > +	{ /* Profile #0 */
> > +		.tb_rate = 3051750,
> > +		.tb_size = 1000000,
> > +
> > +		.tc_rate = {3051750, 3051750, 3051750, 3051750},
> > +		.tc_period = 160,
> > +
> > +		.wrr_weights = {1, 1, 1, 1,
> > +				1, 1, 1, 1,
> > +				1, 1, 1, 1,
> > +				1, 1, 1, 1},
> > +	},
> > +};
> > +
> > +static struct rte_sched_port_params port_param = {
> > +	.socket = 0, /* computed */
> > +	.rate = 0, /* computed */
> > +	.mtu = 1522,
> > +	.frame_overhead = RTE_SCHED_FRAME_OVERHEAD_DEFAULT,
> > +	.n_subports_per_port = 1,
> > +	.n_pipes_per_subport = 1024,
> > +	.qsize = {32, 32, 32, 32},
> > +	.pipe_profiles = pipe_profile,
> > +	.n_pipe_profiles = 1,
> > +};
> > +
> > +#define NB_MBUF          (4 * 32)
> > +#define MBUF_DATA_SZ     (2048 + RTE_PKTMBUF_HEADROOM)
> > +#define MEMPOOL_CACHE_SZ 0
> > +#define SOCKET           0
> > +
> > +
> > +static struct rte_mempool *
> > +create_mempool(uint32_t total_packets) {
> > +	struct rte_mempool *mp;
> > +
> > +	mp = rte_mempool_lookup("test_sched wrr");
> > +	if (!mp)
> > +		mp = rte_pktmbuf_pool_create("test_sched wrr",
> > total_packets,
> > +			MEMPOOL_CACHE_SZ, 0, MBUF_DATA_SZ,
> > SOCKET);
> > +
> > +	return mp;
> > +}
> > +
> > +static void
> > +delete_mempool(struct rte_mempool *mp) {
> > +	rte_mempool_free(mp);
> > +}
> > +
> > +static void
> > +prepare_pkt(struct rte_mbuf *mbuf, uint32_t tc, uint32_t queue) {
> > +	struct ether_hdr *eth_hdr;
> > +	struct vlan_hdr *vlan1, *vlan2;
> > +	struct ipv4_hdr *ip_hdr;
> > +
> > +	/* Simulate a classifier */
> > +	eth_hdr = rte_pktmbuf_mtod(mbuf, struct ether_hdr *);
> > +	vlan1 = (struct vlan_hdr *)(&eth_hdr->ether_type);
> > +	vlan2 = (struct vlan_hdr *)((uintptr_t)&eth_hdr->ether_type +
> > +				    sizeof(struct vlan_hdr));
> > +	eth_hdr = (struct ether_hdr *)((uintptr_t)&eth_hdr->ether_type +
> > +				       2 * sizeof(struct vlan_hdr));
> > +	ip_hdr = (struct ipv4_hdr *)((uintptr_t)eth_hdr +
> > +				     sizeof(eth_hdr->ether_type));
> > +
> > +	vlan1->vlan_tci = rte_cpu_to_be_16(SUBPORT);
> > +	vlan2->vlan_tci = rte_cpu_to_be_16(PIPE);
> > +	eth_hdr->ether_type =  rte_cpu_to_be_16(ETHER_TYPE_IPv4);
> > +	ip_hdr->dst_addr = IPv4(0, 0, TC, QUEUE);
> > +
> > +
> > +	rte_sched_port_pkt_write(mbuf, SUBPORT, PIPE, tc, queue,
> > +				 e_RTE_METER_YELLOW);
> > +
> > +	/* 64 byte packet */
> > +	mbuf->pkt_len  = 60;
> > +	mbuf->data_len = 60;
> > +}
> > +
> > +/*
> > + * This function carries out the core of the enqueue/dequeue testing.
> > + * This is where we should detect failures if the WRR code is broken.
> > + */
> > +static int
> > +test_sched_wrr_enqueue_dequeue(const char *subtest_name,
> > +			       struct rte_mempool *mp,
> > +			       struct rte_sched_port *port,
> > +			       struct rte_mbuf **in_mbufs,
> > +			       struct rte_mbuf **out_mbufs,
> > +			       int32_t enqueue_packets,
> > +			       int32_t dequeue_packets,
> > +			       uint32_t *expected_counts)
> > +{
> > +	uint32_t wrr_counts[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS] = {
> > 0 };
> > +	uint32_t pipe;
> > +	int32_t i;
> > +	int err;
> > +
> > +	/*
> > +	 * Create the packets to be enqueued, spread them evenly across
> > +	 * each of the four WRR queues of the test-TC
> > +	 */
> > +	for (i = 0; i < enqueue_packets; i++) {
> > +		in_mbufs[i] = rte_pktmbuf_alloc(mp);
> > +		TEST_ASSERT_NOT_NULL(in_mbufs[i],
> > +				     "%s: Packet allocation failed on packet "
> > +				     "%d\n", subtest_name, i);
> > +		prepare_pkt(in_mbufs[i], TC, (i & 0x3));
> > +	}
> > +
> > +	/*
> > +	 * Queue all the enqueue packets, none should be dropped as the
> > +	 * four queues should be long enough.
> > +	 */
> > +	err = rte_sched_port_enqueue(port, in_mbufs, enqueue_packets);
> > +	TEST_ASSERT_EQUAL(err, enqueue_packets,
> > +			  "%s: Enqueue failed, err: %d != %d\n",
> > +			  subtest_name, err, enqueue_packets);
> > +
> > +	/*
> > +	 * Dequeue the required number of packets.
> > +	 */
> > +	err = rte_sched_port_dequeue(port, out_mbufs, dequeue_packets);
> > +	TEST_ASSERT_EQUAL(err, dequeue_packets,
> > +			  "%s: Dequeue failed, err: %d != %d\n",
> > +			  subtest_name, err, dequeue_packets);
> > +
> > +	/*
> > +	 * Check each packet and count which WRR queue it came from.
> > +	 */
> > +	for (i = 0; i < dequeue_packets; i++) {
> > +		enum rte_meter_color color;
> > +		uint32_t subport, traffic_class, queue;
> > +
> > +		color = rte_sched_port_pkt_read_color(out_mbufs[i]);
> > +		TEST_ASSERT_EQUAL(color, e_RTE_METER_YELLOW,
> > +				  "%s: Wrong color\n", subtest_name);
> > +
> > +		rte_sched_port_pkt_read_tree_path(out_mbufs[i],
> > +				&subport, &pipe, &traffic_class, &queue);
> > +
> > +		TEST_ASSERT_EQUAL(subport, SUBPORT, "%s: Wrong
> > subport\n",
> > +				  subtest_name);
> > +		TEST_ASSERT_EQUAL(pipe, PIPE, "%s: Wrong pipe\n",
> > subtest_name);
> > +		TEST_ASSERT_EQUAL(traffic_class, TC,
> > +				  "%s: Wrong traffic-class\n",
> > subtest_name);
> > +		wrr_counts[queue]++;
> > +		rte_pktmbuf_free(out_mbufs[i]);
> > +	}
> > +
> > +	/*
> > +	 * Check the number of packets dequeued from each WRR queue
> > +	 * against the expected counts.
> > +	 */
> > +	err = 0;
> > +	for (i = 0; i < RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS; i++) {
> > +		printf("%s - WRR queue %d, dequeued: %u, expected:
> > %u\n",
> > +		       subtest_name, i, wrr_counts[i], expected_counts[i]);
> > +		if (wrr_counts[i] != expected_counts[i])
> > +			err = -1;
> > +	}
> > +	return err;
> > +}
> > +
> > +/*
> > + * This function does the test config set-up and tear-down.
> > + * If we see failures in here it is probably due to the test configuration.
> > + */
> > +static int
> > +test_sched_wrr_test(const char *subtest_name, uint16_t *tc_qlengths,
> > +		    uint8_t *wrr_weights, int32_t enqueue_packets,
> > +		    int32_t dequeue_packets, uint32_t *expected_counts) {
> > +	struct rte_mbuf **in_mbufs;
> > +	struct rte_mbuf **out_mbufs;
> > +	struct rte_mempool *mp = NULL;
> > +	struct rte_sched_port *port = NULL;
> > +	int32_t total_expected = 0;
> > +	uint32_t pipe;
> > +	int32_t i;
> > +	int err;
> > +
> > +	/*
> > +	 * Some inbound argument checking
> > +	 */
> > +	TEST_ASSERT_EQUAL((tc_qlengths[TC] * 4), enqueue_packets,
> > +			  "%s: Queue length/Enqueue packet mismatch - "
> > +			  "%u vs %d\n",
> > +			  subtest_name, (tc_qlengths[TC] * 4),
> > enqueue_packets);
> > +	TEST_ASSERT((dequeue_packets <= enqueue_packets),
> > +		    "%s: Dequeue packets %d > Enqueue packets %d\n",
> > +		    subtest_name, dequeue_packets, enqueue_packets);
> > +
> > +	for (i = 0; i < RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS; i++)
> > +		total_expected += expected_counts[i];
> > +
> > +	TEST_ASSERT((dequeue_packets == total_expected),
> > +		    "%s: Dequeue packets %d != Total expected %u\n",
> > +		    subtest_name, dequeue_packets, total_expected);
> > +
> > +	/*
> > +	 * Create the mempool and allocate arrays to hold the rte_mbuf
> > pointers
> > +	 */
> > +	mp = create_mempool(enqueue_packets);
> > +	TEST_ASSERT_NOT_NULL(mp, "%s: Error creating mempool\n",
> > subtest_name);
> > +
> > +	in_mbufs = malloc(sizeof(struct rte_mbuf *) * enqueue_packets);
> > +	TEST_ASSERT_NOT_NULL(in_mbufs, "%s: Error creating in_mbuf
> > array\n",
> > +			     subtest_name);
> > +
> > +	out_mbufs = malloc(sizeof(struct rte_mbuf *) * dequeue_packets);
> > +	TEST_ASSERT_NOT_NULL(out_mbufs, "%s: Error creating out_mbuf
> > array\n",
> > +			     subtest_name);
> > +
> > +	/*
> > +	 * Set up the port and pipe profiles with the TC's queue lengths
> > +	 * and the WRR queue weightings
> > +	 */
> > +	port_param.socket = 0;
> > +	port_param.rate = (uint64_t) 10000 * 1000 * 1000 / 8;
> > +	for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++)
> > +		port_param.qsize[i] = tc_qlengths[i];
> > +
> > +	for (i = 0; i < RTE_SCHED_QUEUES_PER_PIPE; i++)
> > +		pipe_profile[0].wrr_weights[i] = wrr_weights[i];
> > +
> > +	port = rte_sched_port_config(&port_param);
> > +	TEST_ASSERT_NOT_NULL(port, "%s: Error config sched port\n",
> > +			     subtest_name);
> > +
> > +	err = rte_sched_subport_config(port, SUBPORT, subport_param);
> > +	TEST_ASSERT_SUCCESS(err, "%s: Error config sched, err=%d\n",
> > +			    subtest_name, err);
> > +
> > +	for (pipe = 0; pipe < port_param.n_pipes_per_subport; pipe++) {
> > +		err = rte_sched_pipe_config(port, SUBPORT, pipe, 0);
> > +		TEST_ASSERT_SUCCESS(err,
> > +				    "%s: Error config sched pipe %u,
> > err=%d\n",
> > +				    subtest_name, pipe, err);
> > +	}
> > +
> > +	/*
> > +	 * Enqueue and dequeue packets checking that each WRR queue
> > dequeued
> > +	 * the correct number of packets.
> > +	 */
> > +	err = test_sched_wrr_enqueue_dequeue(subtest_name, mp, port,
> > in_mbufs,
> > +					     out_mbufs, enqueue_packets,
> > +					     dequeue_packets,
> > +					     expected_counts);
> > +
> > +	/*
> > +	 * Free up allocated resources
> > +	 */
> > +	free(in_mbufs);
> > +	free(out_mbufs);
> > +	rte_sched_port_free(port);
> > +	delete_mempool(mp);
> > +
> > +	return err;
> > +}
> > +
> > +static int
> > +test_sched_wrr_even_weights(void)
> > +{
> > +	/*
> > +	 * Even weighting - each WRR queue should dequeue the same
> > number of
> > +	 * packets
> > +	 */
> > +	uint16_t tc_qlengths[] = { 64, 64, 64, 64 };
> > +	uint8_t wrr_weights[] = { 1, 1, 1, 1,     // TC-0
> > +				  1, 1, 1, 1,     // TC-1
> > +				  1, 1, 1, 1,     // TC-2
> > +				  1, 1, 1, 1 };   // TC-3
> > +	int32_t enqueue_packets = 64 * 4;
> > +	int32_t dequeue_packets = 32;
> > +	uint32_t expected_counts[] = { 8, 8, 8, 8 };
> > +
> > +	return test_sched_wrr_test("wrr-even-weights", tc_qlengths,
> > wrr_weights,
> > +				   enqueue_packets, dequeue_packets,
> > +				   expected_counts);
> > +}
> > +
> > +static int
> > +test_sched_wrr_8_4_2_1_weights(void)
> > +{
> > +	/*
> > +	 * Uneven weightings but with a low LCM of 8 * 4 * 2 * 1 = 64
> > +	 */
> > +	uint16_t tc_qlengths[] = { 64, 64, 64, 64 };
> > +	uint8_t wrr_weights[] = { 8, 4, 2, 1,     // TC-0
> > +				  1, 1, 1, 1,
> > +				  1, 1, 1, 1,
> > +				  1, 1, 1, 1 };
> > +	int32_t enqueue_packets = 64 * 4;
> > +	int32_t dequeue_packets = 15;
> > +	uint32_t expected_counts[] = { 8, 4, 2, 1 };
> > +
> > +	return test_sched_wrr_test("wrr-8-4-2-1-weights", tc_qlengths,
> > +				   wrr_weights, enqueue_packets,
> > +				   dequeue_packets, expected_counts); }
> > +
> > +static int
> > +test_sched_wrr_1_2_3_4_weights(void)
> > +{
> > +	/*
> > +	 * Uneven weightings with a low LCM, but weightings in reverse
> > order
> > +	 * from previous test
> > +	 */
> > +	uint16_t tc_qlengths[] = { 64, 64, 64, 64 };
> > +	uint8_t wrr_weights[] = { 1, 2, 3, 4,
> > +				  1, 1, 1, 1,
> > +				  1, 1, 1, 1,
> > +				  1, 1, 1, 1 };
> > +	int32_t enqueue_packets = 64 * 4;
> > +	int32_t dequeue_packets = 40;
> > +	uint32_t expected_counts[] = { 4, 8, 12, 16 };
> > +
> > +	return test_sched_wrr_test("wrr-1-2-3-4-weights", tc_qlengths,
> > +				   wrr_weights, enqueue_packets,
> > +				   dequeue_packets, expected_counts); }
> > +
> > +static int
> > +test_sched_wrr_11_7_5_3_weights(void)
> > +{
> > +	/*
> > +	 * This test generates a LCM of 11 * 7 * 5 * 3 = 1155.
> > +	 * 1155 / 3 = 385 which is more than can be fitted in a uint8_t
> > +	 */
> > +	uint16_t tc_qlengths[] = { 64, 64, 64, 64 };
> > +	uint8_t wrr_weights[] = { 11, 7, 5, 3,
> > +				  1, 1, 1, 1,
> > +				  1, 1, 1, 1,
> > +				  1, 1, 1, 1 };
> > +	int32_t enqueue_packets = 64 * 4;
> > +	int32_t dequeue_packets = 26;
> > +	uint32_t expected_counts[] = { 11, 7, 5, 3 };
> > +
> > +	return test_sched_wrr_test("wrr-11-7-5-3-weights", tc_qlengths,
> > +				   wrr_weights, enqueue_packets,
> > +				   dequeue_packets, expected_counts); }
> > +
> > +static int
> > +test_sched_wrr_100_to_97_weights(void)
> > +{
> > +	/*
> > +	 * The ratios between the weightings is small how well will the
> > +	 * pseudo floating point wrr_cost perform.
> > +	 */
> > +	uint16_t tc_qlengths[] = { 128, 1, 1, 1 };
> > +	uint8_t wrr_weights[] = { 100, 99, 98, 97,
> > +				  1, 1, 1, 1,
> > +				  1, 1, 1, 1,
> > +				  1, 1, 1, 1 };
> > +	int32_t enqueue_packets = 128 * 4;
> > +	int32_t dequeue_packets = 394;
> > +	uint32_t expected_counts[] = { 100, 99, 98, 97 };
> > +
> > +	return test_sched_wrr_test("wrr-100-to-97-weights", tc_qlengths,
> > +				   wrr_weights, enqueue_packets,
> > +				   dequeue_packets, expected_counts); }
> > +
> > +static int
> > +test_sched_wrr_255_254_253_1_weights(void)
> > +{
> > +	/*
> > +	 * This test generates a LCM of 255 * 254 * 253 * 1 = 16386810
> > +	 * 16386810 / 255 =    64262 =   0xFB06
> > +	 * 16386810 / 1   = 16386810 = 0xFA0AFF
> > +	 * 0xFA0AFF needs to be shifted 16 bits left to fit into a uint8_t
> > +	 * but shifting 0xFB06 16 bits left results in a zero wrr_cost.
> > +	 *
> > +	 * This test can fail in two different ways.
> > +	 *
> > +	 * Very badly when the dequeued packet counts aren't even close to
> > the
> > +	 * expected counts due a bug in a recent WRR patch.
> > +	 *
> > +	 * A very near miss, when the dequeued packet counts are just one
> > or
> > +	 * two packets away from the expected counts.  This happens
> > because
> > +	 * the weights of 255, 254, 254 and 1 get converted into wrr_costs
> > +	 * of 1, 1, 1, 250 respectively.  I think that this happens due to
> > +	 * rounding errors in the pseudo floating point code used by WRR.
> > +	 */
> > +	uint16_t tc_qlengths[] = { 1024, 1, 1, 1 };
> > +	uint8_t wrr_weights[] = { 254, 1, 253, 255,
> > +				  1, 1, 1, 1,
> > +				  1, 1, 1, 1,
> > +				  1, 1, 1, 1 };
> > +	int32_t enqueue_packets = 1024 * 4;
> > +	int32_t dequeue_packets = 255 + 254 + 253 + 1;
> > +	uint32_t expected_counts[] = { 254, 1, 253, 255 };
> > +
> > +	return test_sched_wrr_test("wrr-255-254-253-1-weights",
> > tc_qlengths,
> > +				   wrr_weights, enqueue_packets,
> > +				   dequeue_packets, expected_counts); }
> > +
> > +/**
> > + * WRR test main entrance for library sched  */ static int
> > +test_sched_wrr(void)
> > +{
> > +	TEST_ASSERT_SUCCESS(test_sched_wrr_even_weights(),
> > +			    "even-weight test failed\n");
> > +	TEST_ASSERT_SUCCESS(test_sched_wrr_8_4_2_1_weights(),
> > +			    "8-4-2-1-weight test failed\n");
> > +	TEST_ASSERT_SUCCESS(test_sched_wrr_1_2_3_4_weights(),
> > +			    "1-2-3-4-weight test failed\n");
> > +	TEST_ASSERT_SUCCESS(test_sched_wrr_11_7_5_3_weights(),
> > +			    "11-7-5-3-weight test failed\n");
> > +	TEST_ASSERT_SUCCESS(test_sched_wrr_100_to_97_weights(),
> > +			    "100-to-97-weight test failed\n");
> > +	TEST_ASSERT_SUCCESS(test_sched_wrr_255_254_253_1_weights(),
> > +			    "255-254-253-1-weight test failed\n");
> > +	return 0;
> > +}
> > +
> > +REGISTER_TEST_COMMAND(sched_wrr_test, test_sched_wrr);
> > --
> > 2.1.4
>
> Please keep the unit test in the same patch set with the code.
>

      reply	other threads:[~2018-01-03 14:03 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-11-30  9:05 alangordondewar
2017-11-30 10:48 ` Luca Boccassi
2018-01-02 16:15 ` Dumitrescu, Cristian
2018-01-03 14:02   ` Dewar, Alan [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=3F9268EEC0E43747A5FFFC6B48EF0321FBE62D@gbcdcmbx03.intl.att.com \
    --to=ad759e@intl.att.com \
    --cc=alan.dewar@att.com \
    --cc=alangordondewar@gmail.com \
    --cc=cristian.dumitrescu@intel.com \
    --cc=dev@dpdk.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).