Re: [dpdk-dev] [PATCH v2 2/3] app/compress-perf: add performance measurement

["Jozwiak, TomaszX" <tomaszx.jozwiak@intel.com>]

> -----Original Message-----
> From: Verma, Shally [mailto:Shally.Verma@cavium.com]
> Sent: Tuesday, November 6, 2018 4:37 PM
> To: Jozwiak, TomaszX <tomaszx.jozwiak@intel.com>; dev@dpdk.org; Trahe,
> Fiona <fiona.trahe@intel.com>; akhil.goyal@nxp.com
> Subject: RE: [PATCH v2 2/3] app/compress-perf: add performance
> measurement
> 
> 
> 
> >-----Original Message-----
> >From: Jozwiak, TomaszX <tomaszx.jozwiak@intel.com>
> >Sent: 06 November 2018 14:19
> >To: Verma, Shally <Shally.Verma@cavium.com>; dev@dpdk.org; Trahe,
> Fiona
> ><fiona.trahe@intel.com>; akhil.goyal@nxp.com
> >Subject: RE: [PATCH v2 2/3] app/compress-perf: add performance
> >measurement
> >
> >External Email
> >
> >> -----Original Message-----
> >> From: Verma, Shally [mailto:Shally.Verma@cavium.com]
> >> Sent: Monday, November 5, 2018 9:57 AM
> >> To: Jozwiak, TomaszX <tomaszx.jozwiak@intel.com>; dev@dpdk.org;
> >> Trahe, Fiona <fiona.trahe@intel.com>; akhil.goyal@nxp.com
> >> Subject: RE: [PATCH v2 2/3] app/compress-perf: add performance
> >> measurement
> >>
> >>
> >>
> >> >-----Original Message-----
> >> >From: Tomasz Jozwiak <tomaszx.jozwiak@intel.com>
> >> >Sent: 02 November 2018 15:14
> >> >To: dev@dpdk.org; fiona.trahe@intel.com; tomaszx.jozwiak@intel.com;
> >> >Verma, Shally <Shally.Verma@cavium.com>; akhil.goyal@nxp.com
> >> >Subject: [PATCH v2 2/3] app/compress-perf: add performance
> >> measurement
> >> >
> >> >External Email
> >> >
> >> >Added performance measurement part into compression perf. test.
> >> >
> >> >Signed-off-by: De Lara Guarch, Pablo
> >> ><pablo.de.lara.guarch@intel.com>
> >> >Signed-off-by: Tomasz Jozwiak <tomaszx.jozwiak@intel.com>
> >> >---
> >> > app/test-compress-perf/comp_perf_options_parse.c |   8 +-
> >> > app/test-compress-perf/main.c                    | 886
> >> ++++++++++++++++++++++-
> >> > 2 files changed, 883 insertions(+), 11 deletions(-)
> >> >
> >> >diff --git a/app/test-compress-perf/comp_perf_options_parse.c
> >> >b/app/test-compress-perf/comp_perf_options_parse.c
> >> >index bef4d2f..e5da3ad 100644
> >> >--- a/app/test-compress-perf/comp_perf_options_parse.c
> >> >+++ b/app/test-compress-perf/comp_perf_options_parse.c
> >> >@@ -431,10 +431,6 @@ parse_huffman_enc(struct comp_test_data
> >> *test_data, const char *arg)
> >> >                {
> >> >                        "fixed",
> >> >                        RTE_COMP_HUFFMAN_FIXED
> >> >-               },
> >> >-               {
> >> >-                       "dynamic",
> >> >-                       RTE_COMP_HUFFMAN_DYNAMIC
> >> >                }
> >> >        };
> >> >
> >> >@@ -569,9 +565,9 @@ comp_perf_options_default(struct
> comp_test_data
> >> *test_data)
> >> >        test_data->seg_sz = 2048;
> >> >        test_data->burst_sz = 32;
> >> >        test_data->pool_sz = 8192;
> >> >-       test_data->max_sgl_segs = UINT16_MAX;
> >> >+       test_data->max_sgl_segs = 16;
> >> >        test_data->num_iter = 10000;
> >> >-       test_data->huffman_enc = RTE_COMP_HUFFMAN_DYNAMIC;
> >> >+       test_data->huffman_enc = RTE_COMP_HUFFMAN_FIXED;
> >> >        test_data->test_op = COMPRESS_DECOMPRESS;
> >> >        test_data->window_sz = -1;
> >> >        test_data->level.min = 1;
> >> >diff --git a/app/test-compress-perf/main.c
> >> >b/app/test-compress-perf/main.c index f52b98d..e3f4bf6 100644
> >> >--- a/app/test-compress-perf/main.c
> >> >+++ b/app/test-compress-perf/main.c
> >> >@@ -5,14 +5,728 @@
> >> > #include <rte_malloc.h>
> >> > #include <rte_eal.h>
> >> > #include <rte_log.h>
> >> >+#include <rte_cycles.h>
> >> > #include <rte_compressdev.h>
> >> >
> >> > #include "comp_perf_options.h"
> >> >
> >> >+#define NUM_MAX_XFORMS 16
> >> >+#define NUM_MAX_INFLIGHT_OPS 512
> >> >+#define EXPANSE_RATIO 1.05
> >> >+#define MIN_ISAL_SIZE 8
> >> Can we avoid ISAL specific naming ?
> >
> >TJ: yes true :) will be fixed in V3
> >
> >
> >
> >
> >> >+
> >> >+#define DIV_CEIL(a, b)  ((a) / (b) + ((a) % (b) != 0))
> >> >+
> >> >+/* Cleanup state machine */
> >> >+static enum cleanup_st {
> >> >+       ST_CLEAR = 0,
> >> >+       ST_TEST_DATA,
> >> >+       ST_COMPDEV,
> >> >+       ST_INPUT_DATA,
> >> >+       ST_MEMORY_ALLOC,
> >> >+       ST_PREPARE_BUF,
> >> >+       ST_DURING_TEST
> >> >+} cleanup = ST_CLEAR;
> >> >+
> >> >+static int
> >> >+param_range_check(uint16_t size, const struct rte_param_log2_range
> >> >+*range) {
> >> >+       unsigned int next_size;
> >> >+
> >> >+       /* Check lower/upper bounds */
> >> >+       if (size < range->min)
> >> >+               return -1;
> >> >+
> >> >+       if (size > range->max)
> >> >+               return -1;
> >> >+
> >> >+       /* If range is actually only one value, size is correct */
> >> >+       if (range->increment == 0)
> >> >+               return 0;
> >> >+
> >> >+       /* Check if value is one of the supported sizes */
> >> >+       for (next_size = range->min; next_size <= range->max;
> >> >+                       next_size += range->increment)
> >> >+               if (size == next_size)
> >> >+                       return 0;
> >> >+
> >> >+       return -1;
> >> >+}
> >> >+
> >> >+static int
> >> >+comp_perf_check_capabilities(struct comp_test_data *test_data) {
> >> >+       const struct rte_compressdev_capabilities *cap;
> >> >+
> >> >+       cap = rte_compressdev_capability_get(test_data->cdev_id,
> >> >+                                            RTE_COMP_ALGO_DEFLATE);
> >> >+
> >> >+       if (cap == NULL) {
> >> >+               RTE_LOG(ERR, USER1,
> >> >+                       "Compress device does not support DEFLATE\n");
> >> >+               return -1;
> >> >+       }
> >> >+
> >> >+       uint64_t comp_flags = cap->comp_feature_flags;
> >> >+
> >> >+       /* Huffman enconding */
> >> >+       if (test_data->huffman_enc == RTE_COMP_HUFFMAN_FIXED &&
> >> >+                       (comp_flags & RTE_COMP_FF_HUFFMAN_FIXED) == 0) {
> >> >+               RTE_LOG(ERR, USER1,
> >> >+                       "Compress device does not supported Fixed Huffman\n");
> >> >+               return -1;
> >> >+       }
> >> >+
> >> >+       if (test_data->huffman_enc == RTE_COMP_HUFFMAN_DYNAMIC
> &&
> >> >+                       (comp_flags & RTE_COMP_FF_HUFFMAN_DYNAMIC) == 0)
> {
> >> >+               RTE_LOG(ERR, USER1,
> >> >+                       "Compress device does not supported Dynamic
> Huffman\n");
> >> >+               return -1;
> >> >+       }
> >> >+
> >> >+       /* Window size */
> >> >+       if (test_data->window_sz != -1) {
> >> >+               if (param_range_check(test_data->window_sz, &cap-
> >> >window_size)
> >> >+                               < 0) {
> >> >+                       RTE_LOG(ERR, USER1,
> >> >+                               "Compress device does not support "
> >> >+                               "this window size\n");
> >> >+                       return -1;
> >> >+               }
> >> >+       } else
> >> >+               /* Set window size to PMD maximum if none was specified */
> >> >+               test_data->window_sz = cap->window_size.max;
> >> >+
> >> >+       /* Check if chained mbufs is supported */
> >> >+       if (test_data->max_sgl_segs > 1  &&
> >> >+                       (comp_flags & RTE_COMP_FF_OOP_SGL_IN_SGL_OUT) ==
> 0) {
> >> >+               RTE_LOG(INFO, USER1, "Compress device does not support "
> >> >+                               "chained mbufs. Max SGL segments set to 1\n");
> >> >+               test_data->max_sgl_segs = 1;
> >> >+       }
> >> >+
> >> >+       /* Level 0 support */
> >> >+       if (test_data->level.min == 0 &&
> >> >+                       (comp_flags &
> >> >+ RTE_COMP_FF_NONCOMPRESSED_BLOCKS) ==
> >> 0) {
> >> >+               RTE_LOG(ERR, USER1, "Compress device does not support "
> >> >+                               "level 0 (no compression)\n");
> >> >+               return -1;
> >> >+       }
> >> >+
> >> >+       return 0;
> >> >+}
> >> >+
> >> >+static int
> >> >+comp_perf_allocate_memory(struct comp_test_data *test_data) {
> >> >+       /* Number of segments for input and output
> >> >+        * (compression and decompression)
> >> >+        */
> >> >+       uint32_t total_segs = DIV_CEIL(test_data->input_data_sz,
> >> >+                       test_data->seg_sz);
> >> >+       test_data->comp_buf_pool =
> >> rte_pktmbuf_pool_create("comp_buf_pool",
> >> >+                               total_segs,
> >> >+                               0, 0, test_data->seg_sz + RTE_PKTMBUF_HEADROOM,
> >> >+                               rte_socket_id());
> >> >+       if (test_data->comp_buf_pool == NULL) {
> >> >+               RTE_LOG(ERR, USER1, "Mbuf mempool could not be
> created\n");
> >> >+               return -1;
> >> >+       }
> >> >+
> >> >+       cleanup = ST_MEMORY_ALLOC;
> >> >+       test_data->decomp_buf_pool =
> >> rte_pktmbuf_pool_create("decomp_buf_pool",
> >> >+                               total_segs,
> >> >+                               0, 0, test_data->seg_sz + RTE_PKTMBUF_HEADROOM,
> >> >+                               rte_socket_id());
> >> >+       if (test_data->decomp_buf_pool == NULL) {
> >> >+               RTE_LOG(ERR, USER1, "Mbuf mempool could not be
> created\n");
> >> >+               return -1;
> >> >+       }
> >> Unless am missing to see it, you need to free pre-allocated memories
> >> here before return call for all failed cases.
> >
> >TJ: There's only one 'freeing stack' at the end of main application
> >function to avoid double freeing resources (which was previously n V1).
> >We have state machine for that stuff (static enum cleanup_st) to know
> what should be free and what has been allocated already.
> >In case you mean the state machine is set just after first alloc in line 136:
> >
> >cleanup = ST_MEMORY_ALLOC;
> >
> >so we know what should be free at the end of application running in line
> 891:
> >
> >end:
> >        switch (cleanup) {
> >
> >        case ST_DURING_TEST:
> >        case ST_PREPARE_BUF:
> >                free_bufs(test_data);
> >                /* fallthrough */
> >        case ST_MEMORY_ALLOC:
> >                rte_free(test_data->decomp_bufs);
> Even if we are in this state but it doesn't guarantee all of the buffers in this
> state are allocated. So shouldn't every pointer be null-checked before?
> Thanks
> Shally



This check is inside free function already - not need to add double check. 

Thx, Tomek