From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga05.intel.com (mga05.intel.com [192.55.52.43]) by dpdk.org (Postfix) with ESMTP id 6771A1B56C for ; Fri, 23 Nov 2018 14:07:03 +0100 (CET) X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from fmsmga001.fm.intel.com ([10.253.24.23]) by fmsmga105.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 23 Nov 2018 05:07:03 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.56,269,1539673200"; d="scan'208";a="110565191" Received: from tjozwiax-mobl.ger.corp.intel.com (HELO localhost.localdomain) ([10.103.104.46]) by fmsmga001.fm.intel.com with ESMTP; 23 Nov 2018 05:07:01 -0800 From: Tomasz Jozwiak To: dev@dpdk.org, fiona.trahe@intel.com, tomaszx.jozwiak@intel.com, Shally.Verma@cavium.com, akhil.goyal@nxp.com Date: Fri, 23 Nov 2018 14:06:49 +0100 Message-Id: <1542978409-28668-6-git-send-email-tomaszx.jozwiak@intel.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1542978409-28668-1-git-send-email-tomaszx.jozwiak@intel.com> References: <1541151842-8746-1-git-send-email-tomaszx.jozwiak@intel.com> <1542978409-28668-1-git-send-email-tomaszx.jozwiak@intel.com> Subject: [dpdk-dev] [PATCH v3 5/5] app/compress-perf: code refactoring X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 23 Nov 2018 13:07:04 -0000 Code refactoring to separate validation from benchmarking part. Signed-off-by: Tomasz Jozwiak --- app/test-compress-perf/Makefile | 2 + app/test-compress-perf/comp_perf_options.h | 12 + app/test-compress-perf/comp_perf_test_benchmark.c | 291 +++++++++++++++++ app/test-compress-perf/comp_perf_test_benchmark.h | 13 + app/test-compress-perf/comp_perf_test_verify.c | 337 ++++++++++++++++++++ app/test-compress-perf/comp_perf_test_verify.h | 13 + app/test-compress-perf/main.c | 366 +--------------------- app/test-compress-perf/meson.build | 4 +- 8 files changed, 685 insertions(+), 353 deletions(-) create mode 100644 app/test-compress-perf/comp_perf_test_benchmark.c create mode 100644 app/test-compress-perf/comp_perf_test_benchmark.h create mode 100644 app/test-compress-perf/comp_perf_test_verify.c create mode 100644 app/test-compress-perf/comp_perf_test_verify.h diff --git a/app/test-compress-perf/Makefile b/app/test-compress-perf/Makefile index 8aa7a22..d20e17e 100644 --- a/app/test-compress-perf/Makefile +++ b/app/test-compress-perf/Makefile @@ -12,5 +12,7 @@ CFLAGS += -O3 # all source are stored in SRCS-y SRCS-y := main.c SRCS-y += comp_perf_options_parse.c +SRCS-y += comp_perf_test_verify.c +SRCS-y += comp_perf_test_benchmark.c include $(RTE_SDK)/mk/rte.app.mk diff --git a/app/test-compress-perf/comp_perf_options.h b/app/test-compress-perf/comp_perf_options.h index 7516ea0..ca96a3c 100644 --- a/app/test-compress-perf/comp_perf_options.h +++ b/app/test-compress-perf/comp_perf_options.h @@ -2,6 +2,9 @@ * Copyright(c) 2018 Intel Corporation */ +#ifndef _COMP_PERF_OPS_ +#define _COMP_PERF_OPS_ + #define MAX_DRIVER_NAME 64 #define MAX_INPUT_FILE_NAME 64 #define MAX_LIST 32 @@ -46,6 +49,13 @@ struct comp_test_data { /* Store TSC duration for all levels (including level 0) */ uint64_t comp_tsc_duration[RTE_COMP_LEVEL_MAX + 1]; uint64_t decomp_tsc_duration[RTE_COMP_LEVEL_MAX + 1]; + size_t comp_data_sz; + size_t decomp_data_sz; + double ratio; + double comp_gbps; + double decomp_gbps; + double comp_tsc_byte; + double decomp_tsc_byte; }; int @@ -57,3 +67,5 @@ comp_perf_options_default(struct comp_test_data *test_data); int comp_perf_options_check(struct comp_test_data *test_data); + +#endif diff --git a/app/test-compress-perf/comp_perf_test_benchmark.c b/app/test-compress-perf/comp_perf_test_benchmark.c new file mode 100644 index 0000000..ae5f204 --- /dev/null +++ b/app/test-compress-perf/comp_perf_test_benchmark.c @@ -0,0 +1,291 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2018 Intel Corporation + */ + +#include +#include +#include +#include +#include + +#include "comp_perf_test_benchmark.h" + +static int +main_loop(struct comp_test_data *test_data, uint8_t level, + enum rte_comp_xform_type type) +{ + uint8_t dev_id = test_data->cdev_id; + uint32_t i, iter, num_iter; + struct rte_comp_op **ops, **deq_ops; + void *priv_xform = NULL; + struct rte_comp_xform xform; + struct rte_mbuf **input_bufs, **output_bufs; + int res = 0; + int allocated = 0; + + if (test_data == NULL || !test_data->burst_sz) { + RTE_LOG(ERR, USER1, + "Unknown burst size\n"); + return -1; + } + + ops = rte_zmalloc_socket(NULL, + 2 * test_data->total_bufs * sizeof(struct rte_comp_op *), + 0, rte_socket_id()); + + if (ops == NULL) { + RTE_LOG(ERR, USER1, + "Can't allocate memory for ops strucures\n"); + return -1; + } + + deq_ops = &ops[test_data->total_bufs]; + + if (type == RTE_COMP_COMPRESS) { + xform = (struct rte_comp_xform) { + .type = RTE_COMP_COMPRESS, + .compress = { + .algo = RTE_COMP_ALGO_DEFLATE, + .deflate.huffman = test_data->huffman_enc, + .level = level, + .window_size = test_data->window_sz, + .chksum = RTE_COMP_CHECKSUM_NONE, + .hash_algo = RTE_COMP_HASH_ALGO_NONE + } + }; + input_bufs = test_data->decomp_bufs; + output_bufs = test_data->comp_bufs; + } else { + xform = (struct rte_comp_xform) { + .type = RTE_COMP_DECOMPRESS, + .decompress = { + .algo = RTE_COMP_ALGO_DEFLATE, + .chksum = RTE_COMP_CHECKSUM_NONE, + .window_size = test_data->window_sz, + .hash_algo = RTE_COMP_HASH_ALGO_NONE + } + }; + input_bufs = test_data->comp_bufs; + output_bufs = test_data->decomp_bufs; + } + + /* Create private xform */ + if (rte_compressdev_private_xform_create(dev_id, &xform, + &priv_xform) < 0) { + RTE_LOG(ERR, USER1, "Private xform could not be created\n"); + res = -1; + goto end; + } + + uint64_t tsc_start, tsc_end, tsc_duration; + + tsc_start = tsc_end = tsc_duration = 0; + tsc_start = rte_rdtsc(); + num_iter = test_data->num_iter; + + for (iter = 0; iter < num_iter; iter++) { + uint32_t total_ops = test_data->total_bufs; + uint32_t remaining_ops = test_data->total_bufs; + uint32_t total_deq_ops = 0; + uint32_t total_enq_ops = 0; + uint16_t ops_unused = 0; + uint16_t num_enq = 0; + uint16_t num_deq = 0; + + while (remaining_ops > 0) { + uint16_t num_ops = RTE_MIN(remaining_ops, + test_data->burst_sz); + uint16_t ops_needed = num_ops - ops_unused; + + /* + * Move the unused operations from the previous + * enqueue_burst call to the front, to maintain order + */ + if ((ops_unused > 0) && (num_enq > 0)) { + size_t nb_b_to_mov = + ops_unused * sizeof(struct rte_comp_op *); + + memmove(ops, &ops[num_enq], nb_b_to_mov); + } + + /* Allocate compression operations */ + if (ops_needed && !rte_comp_op_bulk_alloc( + test_data->op_pool, + &ops[ops_unused], + ops_needed)) { + RTE_LOG(ERR, USER1, + "Could not allocate enough operations\n"); + res = -1; + goto end; + } + allocated += ops_needed; + + for (i = 0; i < ops_needed; i++) { + /* + * Calculate next buffer to attach to operation + */ + uint32_t buf_id = total_enq_ops + i + + ops_unused; + uint16_t op_id = ops_unused + i; + /* Reset all data in output buffers */ + struct rte_mbuf *m = output_bufs[buf_id]; + + m->pkt_len = test_data->seg_sz * m->nb_segs; + while (m) { + m->data_len = m->buf_len - m->data_off; + m = m->next; + } + ops[op_id]->m_src = input_bufs[buf_id]; + ops[op_id]->m_dst = output_bufs[buf_id]; + ops[op_id]->src.offset = 0; + ops[op_id]->src.length = + rte_pktmbuf_pkt_len(input_bufs[buf_id]); + ops[op_id]->dst.offset = 0; + ops[op_id]->flush_flag = RTE_COMP_FLUSH_FINAL; + ops[op_id]->input_chksum = buf_id; + ops[op_id]->private_xform = priv_xform; + } + + num_enq = rte_compressdev_enqueue_burst(dev_id, 0, ops, + num_ops); + if (num_enq == 0) { + struct rte_compressdev_stats stats; + + rte_compressdev_stats_get(dev_id, &stats); + if (stats.enqueue_err_count) { + res = -1; + goto end; + } + } + + ops_unused = num_ops - num_enq; + remaining_ops -= num_enq; + total_enq_ops += num_enq; + + num_deq = rte_compressdev_dequeue_burst(dev_id, 0, + deq_ops, + test_data->burst_sz); + total_deq_ops += num_deq; + + if (iter == num_iter - 1) { + for (i = 0; i < num_deq; i++) { + struct rte_comp_op *op = deq_ops[i]; + struct rte_mbuf *m = op->m_dst; + + m->pkt_len = op->produced; + uint32_t remaining_data = op->produced; + uint16_t data_to_append; + + while (remaining_data > 0) { + data_to_append = + RTE_MIN(remaining_data, + test_data->seg_sz); + m->data_len = data_to_append; + remaining_data -= + data_to_append; + m = m->next; + } + } + } + rte_mempool_put_bulk(test_data->op_pool, + (void **)deq_ops, num_deq); + allocated -= num_deq; + } + + /* Dequeue the last operations */ + while (total_deq_ops < total_ops) { + num_deq = rte_compressdev_dequeue_burst(dev_id, 0, + deq_ops, test_data->burst_sz); + if (num_deq == 0) { + struct rte_compressdev_stats stats; + + rte_compressdev_stats_get(dev_id, &stats); + if (stats.dequeue_err_count) { + res = -1; + goto end; + } + } + + total_deq_ops += num_deq; + + if (iter == num_iter - 1) { + for (i = 0; i < num_deq; i++) { + struct rte_comp_op *op = deq_ops[i]; + struct rte_mbuf *m = op->m_dst; + + m->pkt_len = op->produced; + uint32_t remaining_data = op->produced; + uint16_t data_to_append; + + while (remaining_data > 0) { + data_to_append = + RTE_MIN(remaining_data, + test_data->seg_sz); + m->data_len = data_to_append; + remaining_data -= + data_to_append; + m = m->next; + } + } + } + rte_mempool_put_bulk(test_data->op_pool, + (void **)deq_ops, num_deq); + allocated -= num_deq; + } + } + + tsc_end = rte_rdtsc(); + tsc_duration = tsc_end - tsc_start; + + if (type == RTE_COMP_COMPRESS) + test_data->comp_tsc_duration[level] = + tsc_duration / num_iter; + else + test_data->decomp_tsc_duration[level] = + tsc_duration / num_iter; + +end: + rte_mempool_put_bulk(test_data->op_pool, (void **)ops, allocated); + rte_compressdev_private_xform_free(dev_id, priv_xform); + rte_free(ops); + return res; +} + +int +cperf_benchmark(struct comp_test_data *test_data, uint8_t level) { + int i, ret = EXIT_SUCCESS; + + /* + * Run the tests twice, discarding the first performance + * results, before the cache is warmed up + */ + for (i = 0; i < 2; i++) { + if (main_loop(test_data, level, RTE_COMP_COMPRESS) < 0) { + ret = EXIT_FAILURE; + goto end; + } + } + + for (i = 0; i < 2; i++) { + if (main_loop(test_data, level, RTE_COMP_DECOMPRESS) < 0) { + ret = EXIT_FAILURE; + goto end; + } + } + + test_data->comp_tsc_byte = + (double)(test_data->comp_tsc_duration[level]) / + test_data->input_data_sz; + + test_data->decomp_tsc_byte = + (double)(test_data->decomp_tsc_duration[level]) / + test_data->input_data_sz; + + test_data->comp_gbps = rte_get_tsc_hz() / test_data->comp_tsc_byte * 8 / + 1000000000; + + test_data->decomp_gbps = rte_get_tsc_hz() / test_data->decomp_tsc_byte + * 8 / 1000000000; +end: + return ret; +} diff --git a/app/test-compress-perf/comp_perf_test_benchmark.h b/app/test-compress-perf/comp_perf_test_benchmark.h new file mode 100644 index 0000000..b193445 --- /dev/null +++ b/app/test-compress-perf/comp_perf_test_benchmark.h @@ -0,0 +1,13 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2018 Intel Corporation + */ + +#ifndef _COMP_PERF_TEST_BENCHMARK_ +#define _COMP_PERF_TEST_BENCHMARK_ + +#include "comp_perf_options.h" + +int +cperf_benchmark(struct comp_test_data *test_data, uint8_t level); + +#endif diff --git a/app/test-compress-perf/comp_perf_test_verify.c b/app/test-compress-perf/comp_perf_test_verify.c new file mode 100644 index 0000000..2ce2c31 --- /dev/null +++ b/app/test-compress-perf/comp_perf_test_verify.c @@ -0,0 +1,337 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2018 Intel Corporation + */ + +#include +#include +#include +#include + +#include "comp_perf_test_verify.h" + +static int +main_loop(struct comp_test_data *test_data, uint8_t level, + enum rte_comp_xform_type type, + uint8_t *output_data_ptr, + size_t *output_data_sz) +{ + uint8_t dev_id = test_data->cdev_id; + uint32_t i, iter, num_iter; + struct rte_comp_op **ops, **deq_ops; + void *priv_xform = NULL; + struct rte_comp_xform xform; + size_t output_size = 0; + struct rte_mbuf **input_bufs, **output_bufs; + int res = 0; + int allocated = 0; + + if (test_data == NULL || !test_data->burst_sz) { + RTE_LOG(ERR, USER1, + "Unknown burst size\n"); + return -1; + } + + ops = rte_zmalloc_socket(NULL, + 2 * test_data->total_bufs * sizeof(struct rte_comp_op *), + 0, rte_socket_id()); + + if (ops == NULL) { + RTE_LOG(ERR, USER1, + "Can't allocate memory for ops strucures\n"); + return -1; + } + + deq_ops = &ops[test_data->total_bufs]; + + if (type == RTE_COMP_COMPRESS) { + xform = (struct rte_comp_xform) { + .type = RTE_COMP_COMPRESS, + .compress = { + .algo = RTE_COMP_ALGO_DEFLATE, + .deflate.huffman = test_data->huffman_enc, + .level = level, + .window_size = test_data->window_sz, + .chksum = RTE_COMP_CHECKSUM_NONE, + .hash_algo = RTE_COMP_HASH_ALGO_NONE + } + }; + input_bufs = test_data->decomp_bufs; + output_bufs = test_data->comp_bufs; + } else { + xform = (struct rte_comp_xform) { + .type = RTE_COMP_DECOMPRESS, + .decompress = { + .algo = RTE_COMP_ALGO_DEFLATE, + .chksum = RTE_COMP_CHECKSUM_NONE, + .window_size = test_data->window_sz, + .hash_algo = RTE_COMP_HASH_ALGO_NONE + } + }; + input_bufs = test_data->comp_bufs; + output_bufs = test_data->decomp_bufs; + } + + /* Create private xform */ + if (rte_compressdev_private_xform_create(dev_id, &xform, + &priv_xform) < 0) { + RTE_LOG(ERR, USER1, "Private xform could not be created\n"); + res = -1; + goto end; + } + + num_iter = 1; + + for (iter = 0; iter < num_iter; iter++) { + uint32_t total_ops = test_data->total_bufs; + uint32_t remaining_ops = test_data->total_bufs; + uint32_t total_deq_ops = 0; + uint32_t total_enq_ops = 0; + uint16_t ops_unused = 0; + uint16_t num_enq = 0; + uint16_t num_deq = 0; + + output_size = 0; + + while (remaining_ops > 0) { + uint16_t num_ops = RTE_MIN(remaining_ops, + test_data->burst_sz); + uint16_t ops_needed = num_ops - ops_unused; + + /* + * Move the unused operations from the previous + * enqueue_burst call to the front, to maintain order + */ + if ((ops_unused > 0) && (num_enq > 0)) { + size_t nb_b_to_mov = + ops_unused * sizeof(struct rte_comp_op *); + + memmove(ops, &ops[num_enq], nb_b_to_mov); + } + + /* Allocate compression operations */ + if (ops_needed && !rte_comp_op_bulk_alloc( + test_data->op_pool, + &ops[ops_unused], + ops_needed)) { + RTE_LOG(ERR, USER1, + "Could not allocate enough operations\n"); + res = -1; + goto end; + } + allocated += ops_needed; + + for (i = 0; i < ops_needed; i++) { + /* + * Calculate next buffer to attach to operation + */ + uint32_t buf_id = total_enq_ops + i + + ops_unused; + uint16_t op_id = ops_unused + i; + /* Reset all data in output buffers */ + struct rte_mbuf *m = output_bufs[buf_id]; + + m->pkt_len = test_data->seg_sz * m->nb_segs; + while (m) { + m->data_len = m->buf_len - m->data_off; + m = m->next; + } + ops[op_id]->m_src = input_bufs[buf_id]; + ops[op_id]->m_dst = output_bufs[buf_id]; + ops[op_id]->src.offset = 0; + ops[op_id]->src.length = + rte_pktmbuf_pkt_len(input_bufs[buf_id]); + ops[op_id]->dst.offset = 0; + ops[op_id]->flush_flag = RTE_COMP_FLUSH_FINAL; + ops[op_id]->input_chksum = buf_id; + ops[op_id]->private_xform = priv_xform; + } + + num_enq = rte_compressdev_enqueue_burst(dev_id, 0, ops, + num_ops); + if (num_enq == 0) { + struct rte_compressdev_stats stats; + + rte_compressdev_stats_get(dev_id, &stats); + if (stats.enqueue_err_count) { + res = -1; + goto end; + } + } + + ops_unused = num_ops - num_enq; + remaining_ops -= num_enq; + total_enq_ops += num_enq; + + num_deq = rte_compressdev_dequeue_burst(dev_id, 0, + deq_ops, + test_data->burst_sz); + total_deq_ops += num_deq; + + for (i = 0; i < num_deq; i++) { + struct rte_comp_op *op = deq_ops[i]; + const void *read_data_addr = + rte_pktmbuf_read(op->m_dst, 0, + op->produced, output_data_ptr); + if (read_data_addr == NULL) { + RTE_LOG(ERR, USER1, + "Could not copy buffer in destination\n"); + res = -1; + goto end; + } + + if (read_data_addr != output_data_ptr) + rte_memcpy(output_data_ptr, + rte_pktmbuf_mtod(op->m_dst, + uint8_t *), + op->produced); + output_data_ptr += op->produced; + output_size += op->produced; + + } + + + if (iter == num_iter - 1) { + for (i = 0; i < num_deq; i++) { + struct rte_comp_op *op = deq_ops[i]; + struct rte_mbuf *m = op->m_dst; + + m->pkt_len = op->produced; + uint32_t remaining_data = op->produced; + uint16_t data_to_append; + + while (remaining_data > 0) { + data_to_append = + RTE_MIN(remaining_data, + test_data->seg_sz); + m->data_len = data_to_append; + remaining_data -= + data_to_append; + m = m->next; + } + } + } + rte_mempool_put_bulk(test_data->op_pool, + (void **)deq_ops, num_deq); + allocated -= num_deq; + } + + /* Dequeue the last operations */ + while (total_deq_ops < total_ops) { + num_deq = rte_compressdev_dequeue_burst(dev_id, 0, + deq_ops, test_data->burst_sz); + if (num_deq == 0) { + struct rte_compressdev_stats stats; + + rte_compressdev_stats_get(dev_id, &stats); + if (stats.dequeue_err_count) { + res = -1; + goto end; + } + } + + total_deq_ops += num_deq; + + for (i = 0; i < num_deq; i++) { + struct rte_comp_op *op = deq_ops[i]; + const void *read_data_addr = + rte_pktmbuf_read(op->m_dst, 0, + op->produced, output_data_ptr); + if (read_data_addr == NULL) { + RTE_LOG(ERR, USER1, + "Could not copy buffer in destination\n"); + res = -1; + goto end; + } + + if (read_data_addr != output_data_ptr) + rte_memcpy(output_data_ptr, + rte_pktmbuf_mtod( + op->m_dst, uint8_t *), + op->produced); + output_data_ptr += op->produced; + output_size += op->produced; + + } + + if (iter == num_iter - 1) { + for (i = 0; i < num_deq; i++) { + struct rte_comp_op *op = deq_ops[i]; + struct rte_mbuf *m = op->m_dst; + + m->pkt_len = op->produced; + uint32_t remaining_data = op->produced; + uint16_t data_to_append; + + while (remaining_data > 0) { + data_to_append = + RTE_MIN(remaining_data, + test_data->seg_sz); + m->data_len = data_to_append; + remaining_data -= + data_to_append; + m = m->next; + } + } + } + rte_mempool_put_bulk(test_data->op_pool, + (void **)deq_ops, num_deq); + allocated -= num_deq; + } + } + + if (output_data_sz) + *output_data_sz = output_size; +end: + rte_mempool_put_bulk(test_data->op_pool, (void **)ops, allocated); + rte_compressdev_private_xform_free(dev_id, priv_xform); + rte_free(ops); + return res; +} + + + +int +cperf_verification(struct comp_test_data *test_data, uint8_t level) { + int ret = EXIT_SUCCESS; + + test_data->ratio = 0; + + if (main_loop(test_data, level, RTE_COMP_COMPRESS, + test_data->compressed_data, + &test_data->comp_data_sz) < 0) { + ret = EXIT_FAILURE; + goto end; + } + + if (main_loop(test_data, level, RTE_COMP_DECOMPRESS, + test_data->decompressed_data, + &test_data->decomp_data_sz) < 0) { + ret = EXIT_FAILURE; + goto end; + } + + if (test_data->decomp_data_sz != test_data->input_data_sz) { + RTE_LOG(ERR, USER1, + "Decompressed data length not equal to input data length\n"); + RTE_LOG(ERR, USER1, + "Decompressed size = %zu, expected = %zu\n", + test_data->decomp_data_sz, test_data->input_data_sz); + ret = EXIT_FAILURE; + goto end; + } else { + if (memcmp(test_data->decompressed_data, + test_data->input_data, + test_data->input_data_sz) != 0) { + RTE_LOG(ERR, USER1, + "Decompressed data is not the same as file data\n"); + ret = EXIT_FAILURE; + goto end; + } + } + + test_data->ratio = (double) test_data->comp_data_sz / + test_data->input_data_sz * 100; + +end: + return ret; +} diff --git a/app/test-compress-perf/comp_perf_test_verify.h b/app/test-compress-perf/comp_perf_test_verify.h new file mode 100644 index 0000000..67c6b49 --- /dev/null +++ b/app/test-compress-perf/comp_perf_test_verify.h @@ -0,0 +1,13 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2018 Intel Corporation + */ + +#ifndef _COMP_PERF_TEST_VERIFY_ +#define _COMP_PERF_TEST_VERIFY_ + +#include "comp_perf_options.h" + +int +cperf_verification(struct comp_test_data *test_data, uint8_t level); + +#endif diff --git a/app/test-compress-perf/main.c b/app/test-compress-perf/main.c index 5950c96..4de913e 100644 --- a/app/test-compress-perf/main.c +++ b/app/test-compress-perf/main.c @@ -5,10 +5,11 @@ #include #include #include -#include #include #include "comp_perf_options.h" +#include "comp_perf_test_verify.h" +#include "comp_perf_test_benchmark.h" #define NUM_MAX_XFORMS 16 #define NUM_MAX_INFLIGHT_OPS 512 @@ -442,285 +443,7 @@ free_bufs(struct comp_test_data *test_data) } } -static int -main_loop(struct comp_test_data *test_data, uint8_t level, - enum rte_comp_xform_type type, - uint8_t *output_data_ptr, - size_t *output_data_sz, - unsigned int benchmarking) -{ - uint8_t dev_id = test_data->cdev_id; - uint32_t i, iter, num_iter; - struct rte_comp_op **ops, **deq_ops; - void *priv_xform = NULL; - struct rte_comp_xform xform; - size_t output_size = 0; - struct rte_mbuf **input_bufs, **output_bufs; - int res = 0; - int allocated = 0; - - if (test_data == NULL || !test_data->burst_sz) { - RTE_LOG(ERR, USER1, - "Unknown burst size\n"); - return -1; - } - - ops = rte_zmalloc_socket(NULL, - 2 * test_data->total_bufs * sizeof(struct rte_comp_op *), - 0, rte_socket_id()); - - if (ops == NULL) { - RTE_LOG(ERR, USER1, - "Can't allocate memory for ops strucures\n"); - return -1; - } - - deq_ops = &ops[test_data->total_bufs]; - - if (type == RTE_COMP_COMPRESS) { - xform = (struct rte_comp_xform) { - .type = RTE_COMP_COMPRESS, - .compress = { - .algo = RTE_COMP_ALGO_DEFLATE, - .deflate.huffman = test_data->huffman_enc, - .level = level, - .window_size = test_data->window_sz, - .chksum = RTE_COMP_CHECKSUM_NONE, - .hash_algo = RTE_COMP_HASH_ALGO_NONE - } - }; - input_bufs = test_data->decomp_bufs; - output_bufs = test_data->comp_bufs; - } else { - xform = (struct rte_comp_xform) { - .type = RTE_COMP_DECOMPRESS, - .decompress = { - .algo = RTE_COMP_ALGO_DEFLATE, - .chksum = RTE_COMP_CHECKSUM_NONE, - .window_size = test_data->window_sz, - .hash_algo = RTE_COMP_HASH_ALGO_NONE - } - }; - input_bufs = test_data->comp_bufs; - output_bufs = test_data->decomp_bufs; - } - - /* Create private xform */ - if (rte_compressdev_private_xform_create(dev_id, &xform, - &priv_xform) < 0) { - RTE_LOG(ERR, USER1, "Private xform could not be created\n"); - res = -1; - goto end; - } - uint64_t tsc_start, tsc_end, tsc_duration; - - tsc_start = tsc_end = tsc_duration = 0; - if (benchmarking) { - tsc_start = rte_rdtsc(); - num_iter = test_data->num_iter; - } else - num_iter = 1; - - for (iter = 0; iter < num_iter; iter++) { - uint32_t total_ops = test_data->total_bufs; - uint32_t remaining_ops = test_data->total_bufs; - uint32_t total_deq_ops = 0; - uint32_t total_enq_ops = 0; - uint16_t ops_unused = 0; - uint16_t num_enq = 0; - uint16_t num_deq = 0; - - output_size = 0; - - while (remaining_ops > 0) { - uint16_t num_ops = RTE_MIN(remaining_ops, - test_data->burst_sz); - uint16_t ops_needed = num_ops - ops_unused; - - /* - * Move the unused operations from the previous - * enqueue_burst call to the front, to maintain order - */ - if ((ops_unused > 0) && (num_enq > 0)) { - size_t nb_b_to_mov = - ops_unused * sizeof(struct rte_comp_op *); - - memmove(ops, &ops[num_enq], nb_b_to_mov); - } - - /* Allocate compression operations */ - if (ops_needed && !rte_comp_op_bulk_alloc( - test_data->op_pool, - &ops[ops_unused], - ops_needed)) { - RTE_LOG(ERR, USER1, - "Could not allocate enough operations\n"); - res = -1; - goto end; - } - allocated += ops_needed; - - for (i = 0; i < ops_needed; i++) { - /* - * Calculate next buffer to attach to operation - */ - uint32_t buf_id = total_enq_ops + i + - ops_unused; - uint16_t op_id = ops_unused + i; - /* Reset all data in output buffers */ - struct rte_mbuf *m = output_bufs[buf_id]; - - m->pkt_len = test_data->seg_sz * m->nb_segs; - while (m) { - m->data_len = m->buf_len - m->data_off; - m = m->next; - } - ops[op_id]->m_src = input_bufs[buf_id]; - ops[op_id]->m_dst = output_bufs[buf_id]; - ops[op_id]->src.offset = 0; - ops[op_id]->src.length = - rte_pktmbuf_pkt_len(input_bufs[buf_id]); - ops[op_id]->dst.offset = 0; - ops[op_id]->flush_flag = RTE_COMP_FLUSH_FINAL; - ops[op_id]->input_chksum = buf_id; - ops[op_id]->private_xform = priv_xform; - } - - num_enq = rte_compressdev_enqueue_burst(dev_id, 0, ops, - num_ops); - ops_unused = num_ops - num_enq; - remaining_ops -= num_enq; - total_enq_ops += num_enq; - - num_deq = rte_compressdev_dequeue_burst(dev_id, 0, - deq_ops, - test_data->burst_sz); - total_deq_ops += num_deq; - if (benchmarking == 0) { - for (i = 0; i < num_deq; i++) { - struct rte_comp_op *op = deq_ops[i]; - const void *read_data_addr = - rte_pktmbuf_read(op->m_dst, 0, - op->produced, output_data_ptr); - if (read_data_addr == NULL) { - RTE_LOG(ERR, USER1, - "Could not copy buffer in destination\n"); - res = -1; - goto end; - } - - if (read_data_addr != output_data_ptr) - rte_memcpy(output_data_ptr, - rte_pktmbuf_mtod( - op->m_dst, uint8_t *), - op->produced); - output_data_ptr += op->produced; - output_size += op->produced; - - } - } - - if (iter == num_iter - 1) { - for (i = 0; i < num_deq; i++) { - struct rte_comp_op *op = deq_ops[i]; - struct rte_mbuf *m = op->m_dst; - - m->pkt_len = op->produced; - uint32_t remaining_data = op->produced; - uint16_t data_to_append; - - while (remaining_data > 0) { - data_to_append = - RTE_MIN(remaining_data, - test_data->seg_sz); - m->data_len = data_to_append; - remaining_data -= - data_to_append; - m = m->next; - } - } - } - rte_mempool_put_bulk(test_data->op_pool, - (void **)deq_ops, num_deq); - allocated -= num_deq; - } - - /* Dequeue the last operations */ - while (total_deq_ops < total_ops) { - num_deq = rte_compressdev_dequeue_burst(dev_id, 0, - deq_ops, test_data->burst_sz); - total_deq_ops += num_deq; - if (benchmarking == 0) { - for (i = 0; i < num_deq; i++) { - struct rte_comp_op *op = deq_ops[i]; - const void *read_data_addr = - rte_pktmbuf_read(op->m_dst, 0, - op->produced, output_data_ptr); - if (read_data_addr == NULL) { - RTE_LOG(ERR, USER1, - "Could not copy buffer in destination\n"); - res = -1; - goto end; - } - - if (read_data_addr != output_data_ptr) - rte_memcpy(output_data_ptr, - rte_pktmbuf_mtod( - op->m_dst, uint8_t *), - op->produced); - output_data_ptr += op->produced; - output_size += op->produced; - - } - } - - if (iter == num_iter - 1) { - for (i = 0; i < num_deq; i++) { - struct rte_comp_op *op = deq_ops[i]; - struct rte_mbuf *m = op->m_dst; - - m->pkt_len = op->produced; - uint32_t remaining_data = op->produced; - uint16_t data_to_append; - - while (remaining_data > 0) { - data_to_append = - RTE_MIN(remaining_data, - test_data->seg_sz); - m->data_len = data_to_append; - remaining_data -= - data_to_append; - m = m->next; - } - } - } - rte_mempool_put_bulk(test_data->op_pool, - (void **)deq_ops, num_deq); - allocated -= num_deq; - } - } - - if (benchmarking) { - tsc_end = rte_rdtsc(); - tsc_duration = tsc_end - tsc_start; - - if (type == RTE_COMP_COMPRESS) - test_data->comp_tsc_duration[level] = - tsc_duration / num_iter; - else - test_data->decomp_tsc_duration[level] = - tsc_duration / num_iter; - } - - if (benchmarking == 0 && output_data_sz) - *output_data_sz = output_size; -end: - rte_mempool_put_bulk(test_data->op_pool, (void **)ops, allocated); - rte_compressdev_private_xform_free(dev_id, priv_xform); - rte_free(ops); - return res; -} int main(int argc, char **argv) @@ -743,6 +466,7 @@ main(int argc, char **argv) rte_exit(EXIT_FAILURE, "Cannot reserve memory in socket %d\n", rte_socket_id()); + ret = EXIT_SUCCESS; cleanup = ST_TEST_DATA; comp_perf_options_default(test_data); @@ -785,9 +509,6 @@ main(int argc, char **argv) else level = test_data->level.list[0]; - size_t comp_data_sz; - size_t decomp_data_sz; - printf("Burst size = %u\n", test_data->burst_sz); printf("File size = %zu\n", test_data->input_data_sz); @@ -798,84 +519,27 @@ main(int argc, char **argv) cleanup = ST_DURING_TEST; while (level <= test_data->level.max) { + /* * Run a first iteration, to verify compression and * get the compression ratio for the level */ - if (main_loop(test_data, level, RTE_COMP_COMPRESS, - test_data->compressed_data, - &comp_data_sz, 0) < 0) { - ret = EXIT_FAILURE; - goto end; - } - - if (main_loop(test_data, level, RTE_COMP_DECOMPRESS, - test_data->decompressed_data, - &decomp_data_sz, 0) < 0) { - ret = EXIT_FAILURE; - goto end; - } - - if (decomp_data_sz != test_data->input_data_sz) { - RTE_LOG(ERR, USER1, - "Decompressed data length not equal to input data length\n"); - RTE_LOG(ERR, USER1, - "Decompressed size = %zu, expected = %zu\n", - decomp_data_sz, test_data->input_data_sz); - ret = EXIT_FAILURE; - goto end; - } else { - if (memcmp(test_data->decompressed_data, - test_data->input_data, - test_data->input_data_sz) != 0) { - RTE_LOG(ERR, USER1, - "Decompressed data is not the same as file data\n"); - ret = EXIT_FAILURE; - goto end; - } - } - - double ratio = (double) comp_data_sz / - test_data->input_data_sz * 100; + if (cperf_verification(test_data, level) != EXIT_SUCCESS) + break; /* - * Run the tests twice, discarding the first performance - * results, before the cache is warmed up + * Run benchmarking test */ - for (i = 0; i < 2; i++) { - if (main_loop(test_data, level, RTE_COMP_COMPRESS, - NULL, NULL, 1) < 0) { - ret = EXIT_FAILURE; - goto end; - } - } - - for (i = 0; i < 2; i++) { - if (main_loop(test_data, level, RTE_COMP_DECOMPRESS, - NULL, NULL, 1) < 0) { - ret = EXIT_FAILURE; - goto end; - } - } - - uint64_t comp_tsc_duration = - test_data->comp_tsc_duration[level]; - double comp_tsc_byte = (double)comp_tsc_duration / - test_data->input_data_sz; - double comp_gbps = rte_get_tsc_hz() / comp_tsc_byte * 8 / - 1000000000; - uint64_t decomp_tsc_duration = - test_data->decomp_tsc_duration[level]; - double decomp_tsc_byte = (double)decomp_tsc_duration / - test_data->input_data_sz; - double decomp_gbps = rte_get_tsc_hz() / decomp_tsc_byte * 8 / - 1000000000; + if (cperf_benchmark(test_data, level) != EXIT_SUCCESS) + break; printf("%6u%12zu%17.2f%19"PRIu64"%21.2f" "%15.2f%21"PRIu64"%23.2f%16.2f\n", - level, comp_data_sz, ratio, comp_tsc_duration, - comp_tsc_byte, comp_gbps, decomp_tsc_duration, - decomp_tsc_byte, decomp_gbps); + level, test_data->comp_data_sz, test_data->ratio, + test_data->comp_tsc_duration[level], + test_data->comp_tsc_byte, test_data->comp_gbps, + test_data->decomp_tsc_duration[level], + test_data->decomp_tsc_byte, test_data->decomp_gbps); if (test_data->level.inc != 0) level += test_data->level.inc; @@ -886,8 +550,6 @@ main(int argc, char **argv) } } - ret = EXIT_SUCCESS; - end: switch (cleanup) { diff --git a/app/test-compress-perf/meson.build b/app/test-compress-perf/meson.build index ba6d64d..ec73e5e 100644 --- a/app/test-compress-perf/meson.build +++ b/app/test-compress-perf/meson.build @@ -3,5 +3,7 @@ allow_experimental_apis = true sources = files('comp_perf_options_parse.c', - 'main.c') + 'main.c', + 'comp_perf_test_verify.c', + 'comp_perf_test_benchmark.c') deps = ['compressdev'] -- 2.7.4