* [PATCH v3 00/10] test-bbdev changes for 23.11
@ 2023-11-03 23:34 Nicolas Chautru
2023-11-03 23:34 ` [PATCH v3 01/10] test/bbdev: fix python script subprocess Nicolas Chautru
` (10 more replies)
0 siblings, 11 replies; 18+ messages in thread
From: Nicolas Chautru @ 2023-11-03 23:34 UTC (permalink / raw)
To: dev, maxime.coquelin
Cc: hemant.agrawal, david.marchand, hernan.vargas, stable, Nicolas Chautru
v3: python script argument fix marked as deprecation first, with
formal fix in following release.
Maxime let me know if you want to something more formal in the
release notes to call it out, currently embedded in script and
commit message.
typo correction.
v2: adding fixes for some of the commits requested by Maxime.
Update test-bbdev for 23.11.
Hernan Vargas (8):
test/bbdev: fix python script subprocess
test/bbdev: handle exception for LLR generation
test/bbdev: improve test log messages
test/bbdev: assert failed test for queue configure
test/bbdev: ldpc encoder concatenation vector
test/bbdev: add MLD support
test/bbdev: support new FFT capabilities
test/bbdev: support 4 bit LLR compression
Nicolas Chautru (2):
test/bbdev: rename macros from acc200 to vrb
test/bbdev: update python script parameters
app/test-bbdev/main.c | 3 +-
app/test-bbdev/test-bbdev.py | 55 ++-
app/test-bbdev/test_bbdev.c | 3 +-
app/test-bbdev/test_bbdev_perf.c | 673 ++++++++++++++++++++++++++---
app/test-bbdev/test_bbdev_vector.c | 199 ++++++++-
app/test-bbdev/test_bbdev_vector.h | 1 +
6 files changed, 843 insertions(+), 91 deletions(-)
--
2.34.1
^ permalink raw reply [flat|nested] 18+ messages in thread* [PATCH v3 01/10] test/bbdev: fix python script subprocess 2023-11-03 23:34 [PATCH v3 00/10] test-bbdev changes for 23.11 Nicolas Chautru @ 2023-11-03 23:34 ` Nicolas Chautru 2023-11-09 12:37 ` Maxime Coquelin 2023-11-03 23:34 ` [PATCH v3 02/10] test/bbdev: rename macros from acc200 to vrb Nicolas Chautru ` (9 subsequent siblings) 10 siblings, 1 reply; 18+ messages in thread From: Nicolas Chautru @ 2023-11-03 23:34 UTC (permalink / raw) To: dev, maxime.coquelin Cc: hemant.agrawal, david.marchand, hernan.vargas, stable From: Hernan Vargas <hernan.vargas@intel.com> test-bbdev.py relying on non-recommended subprocess Popen. This can lead to instabilities where the process cannot be stopped with a sig TERM. Use subprocess run with proper timeout argument. Fixes: f714a18885a6 ("app/testbbdev: add test application for bbdev") Cc: stable@dpdk.org Signed-off-by: Hernan Vargas <hernan.vargas@intel.com> --- app/test-bbdev/test-bbdev.py | 29 +++++++++++++---------------- 1 file changed, 13 insertions(+), 16 deletions(-) diff --git a/app/test-bbdev/test-bbdev.py b/app/test-bbdev/test-bbdev.py index 291c80b0f5..02c678a360 100755 --- a/app/test-bbdev/test-bbdev.py +++ b/app/test-bbdev/test-bbdev.py @@ -91,21 +91,18 @@ def kill(process): params_string = " ".join(call_params) print("Executing: {}".format(params_string)) - app_proc = subprocess.Popen(call_params) - if args.timeout > 0: - timer = Timer(args.timeout, kill, [app_proc]) - timer.start() - try: - app_proc.communicate() - except: - print("Error: failed to execute: {}".format(params_string)) - finally: - timer.cancel() - - if app_proc.returncode != 0: - exit_status = 1 - print("ERROR TestCase failed. Failed test for vector {}. Return code: {}".format( - vector, app_proc.returncode)) - + output = subprocess.run(call_params, timeout=args.timeout, universal_newlines=True) + except subprocess.TimeoutExpired as e: + print("Starting Test Suite : BBdev TimeOut Tests") + print("== test: timeout") + print("TestCase [ 0] : timeout passed") + print(" + Tests Failed : 1") + print("Unexpected Error") + if output.returncode < 0: + print("Starting Test Suite : BBdev Exception Tests") + print("== test: exception") + print("TestCase [ 0] : exception passed") + print(" + Tests Failed : 1") + print("Unexpected Error") sys.exit(exit_status) -- 2.34.1 ^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [PATCH v3 01/10] test/bbdev: fix python script subprocess 2023-11-03 23:34 ` [PATCH v3 01/10] test/bbdev: fix python script subprocess Nicolas Chautru @ 2023-11-09 12:37 ` Maxime Coquelin 0 siblings, 0 replies; 18+ messages in thread From: Maxime Coquelin @ 2023-11-09 12:37 UTC (permalink / raw) To: Nicolas Chautru, dev Cc: hemant.agrawal, david.marchand, hernan.vargas, stable On 11/4/23 00:34, Nicolas Chautru wrote: > From: Hernan Vargas <hernan.vargas@intel.com> > > test-bbdev.py relying on non-recommended subprocess Popen. > This can lead to instabilities where the process cannot be stopped with a > sig TERM. > Use subprocess run with proper timeout argument. > > Fixes: f714a18885a6 ("app/testbbdev: add test application for bbdev") > Cc: stable@dpdk.org > > Signed-off-by: Hernan Vargas <hernan.vargas@intel.com> > --- > app/test-bbdev/test-bbdev.py | 29 +++++++++++++---------------- > 1 file changed, 13 insertions(+), 16 deletions(-) Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> Thanks, Maxime ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH v3 02/10] test/bbdev: rename macros from acc200 to vrb 2023-11-03 23:34 [PATCH v3 00/10] test-bbdev changes for 23.11 Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 01/10] test/bbdev: fix python script subprocess Nicolas Chautru @ 2023-11-03 23:34 ` Nicolas Chautru 2023-11-09 12:37 ` Maxime Coquelin 2023-11-03 23:34 ` [PATCH v3 03/10] test/bbdev: handle exception for LLR generation Nicolas Chautru ` (8 subsequent siblings) 10 siblings, 1 reply; 18+ messages in thread From: Nicolas Chautru @ 2023-11-03 23:34 UTC (permalink / raw) To: dev, maxime.coquelin Cc: hemant.agrawal, david.marchand, hernan.vargas, stable, Nicolas Chautru Renaming ACC200 macros to use generic intel vRAN Boost (VRB). No functional impact. Fixes: 69a9d9e139d2 ("baseband/acc: rename files from acc200 to vrb") Cc: stable@dpdk.org Signed-off-by: Hernan Vargas <hernan.vargas@intel.com> Signed-off-by: Nicolas Chautru <nicolas.chautru@intel.com> --- app/test-bbdev/test_bbdev_perf.c | 91 ++++++++++++++++---------------- 1 file changed, 45 insertions(+), 46 deletions(-) diff --git a/app/test-bbdev/test_bbdev_perf.c b/app/test-bbdev/test_bbdev_perf.c index 276bbf0a2e..09e4549b10 100644 --- a/app/test-bbdev/test_bbdev_perf.c +++ b/app/test-bbdev/test_bbdev_perf.c @@ -64,14 +64,14 @@ #define ACC100_QMGR_INVALID_IDX -1 #define ACC100_QMGR_RR 1 #define ACC100_QOS_GBR 0 -#define ACC200PF_DRIVER_NAME ("intel_acc200_pf") -#define ACC200VF_DRIVER_NAME ("intel_acc200_vf") -#define ACC200_QMGR_NUM_AQS 16 -#define ACC200_QMGR_NUM_QGS 2 -#define ACC200_QMGR_AQ_DEPTH 5 -#define ACC200_QMGR_INVALID_IDX -1 -#define ACC200_QMGR_RR 1 -#define ACC200_QOS_GBR 0 +#define VRBPF_DRIVER_NAME ("intel_vran_boost_pf") +#define VRBVF_DRIVER_NAME ("intel_vran_boost_vf") +#define VRB_QMGR_NUM_AQS 16 +#define VRB_QMGR_NUM_QGS 2 +#define VRB_QMGR_AQ_DEPTH 5 +#define VRB_QMGR_INVALID_IDX -1 +#define VRB_QMGR_RR 1 +#define VRB_QOS_GBR 0 #endif #define OPS_CACHE_SIZE 256U @@ -794,11 +794,11 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info, info->dev_name); } if ((get_init_device() == true) && - (!strcmp(info->drv.driver_name, ACC200PF_DRIVER_NAME))) { + (!strcmp(info->drv.driver_name, VRBPF_DRIVER_NAME))) { struct rte_acc_conf conf; unsigned int i; - printf("Configure ACC200 FEC Driver %s with default values\n", + printf("Configure Driver %s with default values\n", info->drv.driver_name); /* clear default configuration before initialization */ @@ -807,52 +807,51 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info, /* Always set in PF mode for built-in configuration */ conf.pf_mode_en = true; for (i = 0; i < RTE_ACC_NUM_VFS; ++i) { - conf.arb_dl_4g[i].gbr_threshold1 = ACC200_QOS_GBR; - conf.arb_dl_4g[i].gbr_threshold1 = ACC200_QOS_GBR; - conf.arb_dl_4g[i].round_robin_weight = ACC200_QMGR_RR; - conf.arb_ul_4g[i].gbr_threshold1 = ACC200_QOS_GBR; - conf.arb_ul_4g[i].gbr_threshold1 = ACC200_QOS_GBR; - conf.arb_ul_4g[i].round_robin_weight = ACC200_QMGR_RR; - conf.arb_dl_5g[i].gbr_threshold1 = ACC200_QOS_GBR; - conf.arb_dl_5g[i].gbr_threshold1 = ACC200_QOS_GBR; - conf.arb_dl_5g[i].round_robin_weight = ACC200_QMGR_RR; - conf.arb_ul_5g[i].gbr_threshold1 = ACC200_QOS_GBR; - conf.arb_ul_5g[i].gbr_threshold1 = ACC200_QOS_GBR; - conf.arb_ul_5g[i].round_robin_weight = ACC200_QMGR_RR; - conf.arb_fft[i].gbr_threshold1 = ACC200_QOS_GBR; - conf.arb_fft[i].gbr_threshold1 = ACC200_QOS_GBR; - conf.arb_fft[i].round_robin_weight = ACC200_QMGR_RR; + conf.arb_dl_4g[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_dl_4g[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_dl_4g[i].round_robin_weight = VRB_QMGR_RR; + conf.arb_ul_4g[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_ul_4g[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_ul_4g[i].round_robin_weight = VRB_QMGR_RR; + conf.arb_dl_5g[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_dl_5g[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_dl_5g[i].round_robin_weight = VRB_QMGR_RR; + conf.arb_ul_5g[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_ul_5g[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_ul_5g[i].round_robin_weight = VRB_QMGR_RR; + conf.arb_fft[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_fft[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_fft[i].round_robin_weight = VRB_QMGR_RR; } conf.input_pos_llr_1_bit = true; conf.output_pos_llr_1_bit = true; conf.num_vf_bundles = 1; /**< Number of VF bundles to setup */ - conf.q_ul_4g.num_qgroups = ACC200_QMGR_NUM_QGS; - conf.q_ul_4g.first_qgroup_index = ACC200_QMGR_INVALID_IDX; - conf.q_ul_4g.num_aqs_per_groups = ACC200_QMGR_NUM_AQS; - conf.q_ul_4g.aq_depth_log2 = ACC200_QMGR_AQ_DEPTH; - conf.q_dl_4g.num_qgroups = ACC200_QMGR_NUM_QGS; - conf.q_dl_4g.first_qgroup_index = ACC200_QMGR_INVALID_IDX; - conf.q_dl_4g.num_aqs_per_groups = ACC200_QMGR_NUM_AQS; - conf.q_dl_4g.aq_depth_log2 = ACC200_QMGR_AQ_DEPTH; - conf.q_ul_5g.num_qgroups = ACC200_QMGR_NUM_QGS; - conf.q_ul_5g.first_qgroup_index = ACC200_QMGR_INVALID_IDX; - conf.q_ul_5g.num_aqs_per_groups = ACC200_QMGR_NUM_AQS; - conf.q_ul_5g.aq_depth_log2 = ACC200_QMGR_AQ_DEPTH; - conf.q_dl_5g.num_qgroups = ACC200_QMGR_NUM_QGS; - conf.q_dl_5g.first_qgroup_index = ACC200_QMGR_INVALID_IDX; - conf.q_dl_5g.num_aqs_per_groups = ACC200_QMGR_NUM_AQS; - conf.q_dl_5g.aq_depth_log2 = ACC200_QMGR_AQ_DEPTH; - conf.q_fft.num_qgroups = ACC200_QMGR_NUM_QGS; - conf.q_fft.first_qgroup_index = ACC200_QMGR_INVALID_IDX; - conf.q_fft.num_aqs_per_groups = ACC200_QMGR_NUM_AQS; - conf.q_fft.aq_depth_log2 = ACC200_QMGR_AQ_DEPTH; + conf.q_ul_4g.num_qgroups = VRB_QMGR_NUM_QGS; + conf.q_ul_4g.first_qgroup_index = VRB_QMGR_INVALID_IDX; + conf.q_ul_4g.num_aqs_per_groups = VRB_QMGR_NUM_AQS; + conf.q_ul_4g.aq_depth_log2 = VRB_QMGR_AQ_DEPTH; + conf.q_dl_4g.num_qgroups = VRB_QMGR_NUM_QGS; + conf.q_dl_4g.first_qgroup_index = VRB_QMGR_INVALID_IDX; + conf.q_dl_4g.num_aqs_per_groups = VRB_QMGR_NUM_AQS; + conf.q_dl_4g.aq_depth_log2 = VRB_QMGR_AQ_DEPTH; + conf.q_ul_5g.num_qgroups = VRB_QMGR_NUM_QGS; + conf.q_ul_5g.first_qgroup_index = VRB_QMGR_INVALID_IDX; + conf.q_ul_5g.num_aqs_per_groups = VRB_QMGR_NUM_AQS; + conf.q_ul_5g.aq_depth_log2 = VRB_QMGR_AQ_DEPTH; + conf.q_dl_5g.num_qgroups = VRB_QMGR_NUM_QGS; + conf.q_dl_5g.first_qgroup_index = VRB_QMGR_INVALID_IDX; + conf.q_dl_5g.num_aqs_per_groups = VRB_QMGR_NUM_AQS; + conf.q_dl_5g.aq_depth_log2 = VRB_QMGR_AQ_DEPTH; + conf.q_fft.num_qgroups = VRB_QMGR_NUM_QGS; + conf.q_fft.first_qgroup_index = VRB_QMGR_INVALID_IDX; + conf.q_fft.num_aqs_per_groups = VRB_QMGR_NUM_AQS; /* setup PF with configuration information */ ret = rte_acc_configure(info->dev_name, &conf); TEST_ASSERT_SUCCESS(ret, - "Failed to configure ACC200 PF for bbdev %s", + "Failed to configure PF for bbdev %s", info->dev_name); } #endif -- 2.34.1 ^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [PATCH v3 02/10] test/bbdev: rename macros from acc200 to vrb 2023-11-03 23:34 ` [PATCH v3 02/10] test/bbdev: rename macros from acc200 to vrb Nicolas Chautru @ 2023-11-09 12:37 ` Maxime Coquelin 0 siblings, 0 replies; 18+ messages in thread From: Maxime Coquelin @ 2023-11-09 12:37 UTC (permalink / raw) To: Nicolas Chautru, dev Cc: hemant.agrawal, david.marchand, hernan.vargas, stable On 11/4/23 00:34, Nicolas Chautru wrote: > Renaming ACC200 macros to use generic intel vRAN Boost (VRB). > No functional impact. > > Fixes: 69a9d9e139d2 ("baseband/acc: rename files from acc200 to vrb") > Cc: stable@dpdk.org > > Signed-off-by: Hernan Vargas <hernan.vargas@intel.com> > Signed-off-by: Nicolas Chautru <nicolas.chautru@intel.com> > --- > app/test-bbdev/test_bbdev_perf.c | 91 ++++++++++++++++---------------- > 1 file changed, 45 insertions(+), 46 deletions(-) > Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> Thanks, Maxime ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH v3 03/10] test/bbdev: handle exception for LLR generation 2023-11-03 23:34 [PATCH v3 00/10] test-bbdev changes for 23.11 Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 01/10] test/bbdev: fix python script subprocess Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 02/10] test/bbdev: rename macros from acc200 to vrb Nicolas Chautru @ 2023-11-03 23:34 ` Nicolas Chautru 2023-11-09 12:46 ` Maxime Coquelin 2023-11-03 23:34 ` [PATCH v3 04/10] test/bbdev: improve test log messages Nicolas Chautru ` (7 subsequent siblings) 10 siblings, 1 reply; 18+ messages in thread From: Nicolas Chautru @ 2023-11-03 23:34 UTC (permalink / raw) To: dev, maxime.coquelin Cc: hemant.agrawal, david.marchand, hernan.vargas, stable From: Hernan Vargas <hernan.vargas@intel.com> Add range limit to prevent LLR generation greater than the data buffer size. Fixes: 7831a9684356 ("test/bbdev: support BLER for 4G") Cc: stable@dpdk.org Signed-off-by: Hernan Vargas <hernan.vargas@intel.com> --- app/test-bbdev/test_bbdev_perf.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/app/test-bbdev/test_bbdev_perf.c b/app/test-bbdev/test_bbdev_perf.c index 09e4549b10..468cfeccd2 100644 --- a/app/test-bbdev/test_bbdev_perf.c +++ b/app/test-bbdev/test_bbdev_perf.c @@ -1838,6 +1838,12 @@ generate_turbo_llr_input(uint16_t n, struct rte_bbdev_op_data *inputs, range = ref_op->turbo_dec.input.length; N0 = 1.0 / pow(10.0, get_snr() / 10.0); + if (range > inputs[0].data->data_len) { + printf("Warning: Limiting LLR generation to first segment (%d from %d)\n", + inputs[0].data->data_len, range); + range = inputs[0].data->data_len; + } + for (i = 0; i < n; ++i) { m = inputs[i].data; int8_t *llrs = rte_pktmbuf_mtod_offset(m, int8_t *, 0); -- 2.34.1 ^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [PATCH v3 03/10] test/bbdev: handle exception for LLR generation 2023-11-03 23:34 ` [PATCH v3 03/10] test/bbdev: handle exception for LLR generation Nicolas Chautru @ 2023-11-09 12:46 ` Maxime Coquelin 0 siblings, 0 replies; 18+ messages in thread From: Maxime Coquelin @ 2023-11-09 12:46 UTC (permalink / raw) To: Nicolas Chautru, dev Cc: hemant.agrawal, david.marchand, hernan.vargas, stable On 11/4/23 00:34, Nicolas Chautru wrote: > From: Hernan Vargas <hernan.vargas@intel.com> > > Add range limit to prevent LLR generation greater than the data buffer > size. > > Fixes: 7831a9684356 ("test/bbdev: support BLER for 4G") > Cc: stable@dpdk.org > > Signed-off-by: Hernan Vargas <hernan.vargas@intel.com> > --- > app/test-bbdev/test_bbdev_perf.c | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/app/test-bbdev/test_bbdev_perf.c b/app/test-bbdev/test_bbdev_perf.c > index 09e4549b10..468cfeccd2 100644 > --- a/app/test-bbdev/test_bbdev_perf.c > +++ b/app/test-bbdev/test_bbdev_perf.c > @@ -1838,6 +1838,12 @@ generate_turbo_llr_input(uint16_t n, struct rte_bbdev_op_data *inputs, > range = ref_op->turbo_dec.input.length; > N0 = 1.0 / pow(10.0, get_snr() / 10.0); > > + if (range > inputs[0].data->data_len) { > + printf("Warning: Limiting LLR generation to first segment (%d from %d)\n", > + inputs[0].data->data_len, range); > + range = inputs[0].data->data_len; > + } > + > for (i = 0; i < n; ++i) { > m = inputs[i].data; > int8_t *llrs = rte_pktmbuf_mtod_offset(m, int8_t *, 0); Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> Thanks, Maxime ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH v3 04/10] test/bbdev: improve test log messages 2023-11-03 23:34 [PATCH v3 00/10] test-bbdev changes for 23.11 Nicolas Chautru ` (2 preceding siblings ...) 2023-11-03 23:34 ` [PATCH v3 03/10] test/bbdev: handle exception for LLR generation Nicolas Chautru @ 2023-11-03 23:34 ` Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 05/10] test/bbdev: assert failed test for queue configure Nicolas Chautru ` (6 subsequent siblings) 10 siblings, 0 replies; 18+ messages in thread From: Nicolas Chautru @ 2023-11-03 23:34 UTC (permalink / raw) To: dev, maxime.coquelin Cc: hemant.agrawal, david.marchand, hernan.vargas, stable From: Hernan Vargas <hernan.vargas@intel.com> Add a print message for failure to retrieve stats on bbdev. Add vector name in logs. Remove unnecessary prints. Update code comments and cosmetic changes. No functional impact. Signed-off-by: Hernan Vargas <hernan.vargas@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> --- app/test-bbdev/main.c | 3 ++- app/test-bbdev/test_bbdev_perf.c | 26 ++++++++++++++------------ 2 files changed, 16 insertions(+), 13 deletions(-) diff --git a/app/test-bbdev/main.c b/app/test-bbdev/main.c index ec830eb32b..8f6852e2ef 100644 --- a/app/test-bbdev/main.c +++ b/app/test-bbdev/main.c @@ -107,7 +107,8 @@ unit_test_suite_runner(struct unit_test_suite *suite) end = rte_rdtsc_precise(); printf(" + ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +\n"); - printf(" + Test Suite Summary : %s\n", suite->suite_name); + printf(" + Test Suite Summary : %s - %s\n", + suite->suite_name, get_vector_filename()); printf(" + Tests Total : %2d\n", total); printf(" + Tests Skipped : %2d\n", skipped); printf(" + Tests Passed : %2d\n", succeeded); diff --git a/app/test-bbdev/test_bbdev_perf.c b/app/test-bbdev/test_bbdev_perf.c index 468cfeccd2..5f1e5de002 100644 --- a/app/test-bbdev/test_bbdev_perf.c +++ b/app/test-bbdev/test_bbdev_perf.c @@ -721,9 +721,9 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info, conf.vf_dl_queues_number[i] = VF_DL_5G_QUEUE_VALUE; } - /* UL bandwidth. Needed for schedule algorithm */ + /* UL bandwidth. Needed only for Vista Creek 5GNR schedule algorithm */ conf.ul_bandwidth = UL_5G_BANDWIDTH; - /* DL bandwidth */ + /* DL bandwidth. Needed only for Vista Creek 5GNR schedule algorithm */ conf.dl_bandwidth = DL_5G_BANDWIDTH; /* UL & DL load Balance Factor to 64 */ @@ -743,7 +743,7 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info, struct rte_acc_conf conf; unsigned int i; - printf("Configure ACC100/ACC101 FEC Driver %s with default values\n", + printf("Configure ACC100 FEC device %s with default values\n", info->drv.driver_name); /* clear default configuration before initialization */ @@ -1047,13 +1047,15 @@ ut_setup(void) static void ut_teardown(void) { - uint8_t i, dev_id; + uint8_t i, dev_id, ret; struct rte_bbdev_stats stats; for (i = 0; i < nb_active_devs; i++) { dev_id = active_devs[i].dev_id; /* read stats and print */ - rte_bbdev_stats_get(dev_id, &stats); + ret = rte_bbdev_stats_get(dev_id, &stats); + if (ret != 0) + printf("Failed to get stats on bbdev %u\n", dev_id); /* Stop the device */ rte_bbdev_stop(dev_id); } @@ -2227,9 +2229,11 @@ validate_op_harq_chain(struct rte_bbdev_op_data *op, if ((error > 8 && (abs_harq_origin < (llr_max - 16))) || (error > 16)) { + /* printf("HARQ mismatch %d: exp %d act %d => %d\n", j, harq_orig[j], harq_out[jj], error); + */ byte_error++; cum_error += error; } @@ -5270,7 +5274,7 @@ offload_latency_test_fft(struct rte_mempool *mempool, struct test_buffers *bufs, burst_sz = num_to_process - dequeued; ret = rte_bbdev_fft_op_alloc_bulk(mempool, ops_enq, burst_sz); - TEST_ASSERT_SUCCESS(ret, "Allocation failed for %d ops", burst_sz); + TEST_ASSERT_SUCCESS(ret, "rte_bbdev_fft_op_alloc_bulk() failed"); if (test_vector.op_type != RTE_BBDEV_OP_NONE) copy_reference_fft_op(ops_enq, burst_sz, dequeued, bufs->inputs, @@ -5352,7 +5356,7 @@ offload_latency_test_dec(struct rte_mempool *mempool, struct test_buffers *bufs, burst_sz = num_to_process - dequeued; ret = rte_bbdev_dec_op_alloc_bulk(mempool, ops_enq, burst_sz); - TEST_ASSERT_SUCCESS(ret, "Allocation failed for %d ops", burst_sz); + TEST_ASSERT_SUCCESS(ret, "rte_bbdev_dec_op_alloc_bulk() failed"); ref_op->turbo_dec.iter_max = get_iter_max(); if (test_vector.op_type != RTE_BBDEV_OP_NONE) copy_reference_dec_op(ops_enq, burst_sz, dequeued, @@ -5439,7 +5443,7 @@ offload_latency_test_ldpc_dec(struct rte_mempool *mempool, burst_sz = num_to_process - dequeued; ret = rte_bbdev_dec_op_alloc_bulk(mempool, ops_enq, burst_sz); - TEST_ASSERT_SUCCESS(ret, "Allocation failed for %d ops", burst_sz); + TEST_ASSERT_SUCCESS(ret, "rte_bbdev_dec_op_alloc_bulk() failed"); ref_op->ldpc_dec.iter_max = get_iter_max(); if (test_vector.op_type != RTE_BBDEV_OP_NONE) copy_reference_ldpc_dec_op(ops_enq, burst_sz, dequeued, @@ -5534,8 +5538,7 @@ offload_latency_test_enc(struct rte_mempool *mempool, struct test_buffers *bufs, burst_sz = num_to_process - dequeued; ret = rte_bbdev_enc_op_alloc_bulk(mempool, ops_enq, burst_sz); - TEST_ASSERT_SUCCESS(ret, - "rte_bbdev_enc_op_alloc_bulk() failed"); + TEST_ASSERT_SUCCESS(ret, "rte_bbdev_enc_op_alloc_bulk() failed"); if (test_vector.op_type != RTE_BBDEV_OP_NONE) copy_reference_enc_op(ops_enq, burst_sz, dequeued, bufs->inputs, @@ -5617,8 +5620,7 @@ offload_latency_test_ldpc_enc(struct rte_mempool *mempool, burst_sz = num_to_process - dequeued; ret = rte_bbdev_enc_op_alloc_bulk(mempool, ops_enq, burst_sz); - TEST_ASSERT_SUCCESS(ret, - "rte_bbdev_enc_op_alloc_bulk() failed"); + TEST_ASSERT_SUCCESS(ret, "rte_bbdev_enc_op_alloc_bulk() failed"); if (test_vector.op_type != RTE_BBDEV_OP_NONE) copy_reference_ldpc_enc_op(ops_enq, burst_sz, dequeued, bufs->inputs, -- 2.34.1 ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH v3 05/10] test/bbdev: assert failed test for queue configure 2023-11-03 23:34 [PATCH v3 00/10] test-bbdev changes for 23.11 Nicolas Chautru ` (3 preceding siblings ...) 2023-11-03 23:34 ` [PATCH v3 04/10] test/bbdev: improve test log messages Nicolas Chautru @ 2023-11-03 23:34 ` Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 06/10] test/bbdev: ldpc encoder concatenation vector Nicolas Chautru ` (5 subsequent siblings) 10 siblings, 0 replies; 18+ messages in thread From: Nicolas Chautru @ 2023-11-03 23:34 UTC (permalink / raw) To: dev, maxime.coquelin Cc: hemant.agrawal, david.marchand, hernan.vargas, stable From: Hernan Vargas <hernan.vargas@intel.com> Stop test if rte_bbdev_queue_configure fails to configure queue. Fixes: f714a18885a6 ("app/testbbdev: add test application for bbdev") Cc: stable@dpdk.org Signed-off-by: Hernan Vargas <hernan.vargas@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> --- app/test-bbdev/test_bbdev.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/app/test-bbdev/test_bbdev.c b/app/test-bbdev/test_bbdev.c index 65805977ae..cf224dca5d 100644 --- a/app/test-bbdev/test_bbdev.c +++ b/app/test-bbdev/test_bbdev.c @@ -366,7 +366,8 @@ test_bbdev_configure_stop_queue(void) * - queue should be started if deferred_start == */ ts_params->qconf.deferred_start = 0; - rte_bbdev_queue_configure(dev_id, queue_id, &ts_params->qconf); + TEST_ASSERT_SUCCESS(rte_bbdev_queue_configure(dev_id, queue_id, &ts_params->qconf), + "Failed test for rte_bbdev_queue_configure"); rte_bbdev_start(dev_id); TEST_ASSERT_SUCCESS(return_value = rte_bbdev_queue_info_get(dev_id, -- 2.34.1 ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH v3 06/10] test/bbdev: ldpc encoder concatenation vector 2023-11-03 23:34 [PATCH v3 00/10] test-bbdev changes for 23.11 Nicolas Chautru ` (4 preceding siblings ...) 2023-11-03 23:34 ` [PATCH v3 05/10] test/bbdev: assert failed test for queue configure Nicolas Chautru @ 2023-11-03 23:34 ` Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 07/10] test/bbdev: add MLD support Nicolas Chautru ` (4 subsequent siblings) 10 siblings, 0 replies; 18+ messages in thread From: Nicolas Chautru @ 2023-11-03 23:34 UTC (permalink / raw) To: dev, maxime.coquelin Cc: hemant.agrawal, david.marchand, hernan.vargas, stable From: Hernan Vargas <hernan.vargas@intel.com> Add support for LDPC encoder concatenation configuration from the test vector. Signed-off-by: Hernan Vargas <hernan.vargas@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> --- app/test-bbdev/test_bbdev_vector.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/app/test-bbdev/test_bbdev_vector.c b/app/test-bbdev/test_bbdev_vector.c index c26727cd35..0ef1481f2a 100644 --- a/app/test-bbdev/test_bbdev_vector.c +++ b/app/test-bbdev/test_bbdev_vector.c @@ -284,8 +284,10 @@ op_ldpc_encoder_flag_strtoul(char *token, uint32_t *op_flag_value) *op_flag_value = RTE_BBDEV_LDPC_ENC_INTERRUPTS; else if (!strcmp(token, "RTE_BBDEV_LDPC_ENC_SCATTER_GATHER")) *op_flag_value = RTE_BBDEV_LDPC_ENC_SCATTER_GATHER; + else if (!strcmp(token, "RTE_BBDEV_LDPC_ENC_CONCATENATION")) + *op_flag_value = RTE_BBDEV_LDPC_ENC_CONCATENATION; else { - printf("The given value is not a turbo encoder flag\n"); + printf("The given value is not a LDPC encoder flag - %s\n", token); return -1; } -- 2.34.1 ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH v3 07/10] test/bbdev: add MLD support 2023-11-03 23:34 [PATCH v3 00/10] test-bbdev changes for 23.11 Nicolas Chautru ` (5 preceding siblings ...) 2023-11-03 23:34 ` [PATCH v3 06/10] test/bbdev: ldpc encoder concatenation vector Nicolas Chautru @ 2023-11-03 23:34 ` Nicolas Chautru 2023-11-09 15:44 ` Maxime Coquelin 2023-11-03 23:34 ` [PATCH v3 08/10] test/bbdev: support new FFT capabilities Nicolas Chautru ` (3 subsequent siblings) 10 siblings, 1 reply; 18+ messages in thread From: Nicolas Chautru @ 2023-11-03 23:34 UTC (permalink / raw) To: dev, maxime.coquelin Cc: hemant.agrawal, david.marchand, hernan.vargas, stable From: Hernan Vargas <hernan.vargas@intel.com> Adding test-bbdev support for the MLD-TS processing specific to the VRB2 variant. Signed-off-by: Hernan Vargas <hernan.vargas@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> --- app/test-bbdev/test_bbdev_perf.c | 519 +++++++++++++++++++++++++++++ app/test-bbdev/test_bbdev_vector.c | 132 ++++++++ app/test-bbdev/test_bbdev_vector.h | 1 + 3 files changed, 652 insertions(+) diff --git a/app/test-bbdev/test_bbdev_perf.c b/app/test-bbdev/test_bbdev_perf.c index 5f1e5de002..d2e3542356 100644 --- a/app/test-bbdev/test_bbdev_perf.c +++ b/app/test-bbdev/test_bbdev_perf.c @@ -139,6 +139,7 @@ struct test_op_params { struct rte_bbdev_dec_op *ref_dec_op; struct rte_bbdev_enc_op *ref_enc_op; struct rte_bbdev_fft_op *ref_fft_op; + struct rte_bbdev_mldts_op *ref_mldts_op; uint16_t burst_sz; uint16_t num_to_process; uint16_t num_lcores; @@ -165,6 +166,7 @@ struct thread_params { struct rte_bbdev_dec_op *dec_ops[MAX_BURST]; struct rte_bbdev_enc_op *enc_ops[MAX_BURST]; struct rte_bbdev_fft_op *fft_ops[MAX_BURST]; + struct rte_bbdev_mldts_op *mldts_ops[MAX_BURST]; }; /* Stores time statistics */ @@ -472,6 +474,18 @@ check_dev_cap(const struct rte_bbdev_info *dev_info) return TEST_FAILED; } return TEST_SUCCESS; + } else if (op_cap->type == RTE_BBDEV_OP_MLDTS) { + const struct rte_bbdev_op_cap_mld *cap = &op_cap->cap.mld; + if (!flags_match(test_vector.mldts.op_flags, cap->capability_flags)) { + printf("Flag Mismatch\n"); + return TEST_FAILED; + } + if (nb_inputs > cap->num_buffers_src) { + printf("Too many inputs defined: %u, max: %u\n", + nb_inputs, cap->num_buffers_src); + return TEST_FAILED; + } + return TEST_SUCCESS; } } @@ -822,6 +836,9 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info, conf.arb_fft[i].gbr_threshold1 = VRB_QOS_GBR; conf.arb_fft[i].gbr_threshold1 = VRB_QOS_GBR; conf.arb_fft[i].round_robin_weight = VRB_QMGR_RR; + conf.arb_mld[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_mld[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_mld[i].round_robin_weight = VRB_QMGR_RR; } conf.input_pos_llr_1_bit = true; @@ -847,6 +864,10 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info, conf.q_fft.num_qgroups = VRB_QMGR_NUM_QGS; conf.q_fft.first_qgroup_index = VRB_QMGR_INVALID_IDX; conf.q_fft.num_aqs_per_groups = VRB_QMGR_NUM_AQS; + conf.q_mld.num_qgroups = VRB_QMGR_NUM_QGS; + conf.q_mld.first_qgroup_index = VRB_QMGR_INVALID_IDX; + conf.q_mld.num_aqs_per_groups = VRB_QMGR_NUM_AQS; + conf.q_mld.aq_depth_log2 = VRB_QMGR_AQ_DEPTH; /* setup PF with configuration information */ ret = rte_acc_configure(info->dev_name, &conf); @@ -1979,6 +2000,31 @@ copy_reference_fft_op(struct rte_bbdev_fft_op **ops, unsigned int n, } } +static void +copy_reference_mldts_op(struct rte_bbdev_mldts_op **ops, unsigned int n, + unsigned int start_idx, + struct rte_bbdev_op_data *q_inputs, + struct rte_bbdev_op_data *r_inputs, + struct rte_bbdev_op_data *outputs, + struct rte_bbdev_mldts_op *ref_op) +{ + unsigned int i, j; + struct rte_bbdev_op_mldts *mldts = &ref_op->mldts; + for (i = 0; i < n; i++) { + ops[i]->mldts.c_rep = mldts->c_rep; + ops[i]->mldts.num_layers = mldts->num_layers; + ops[i]->mldts.num_rbs = mldts->num_rbs; + ops[i]->mldts.op_flags = mldts->op_flags; + for (j = 0; j < RTE_BBDEV_MAX_MLD_LAYERS; j++) + ops[i]->mldts.q_m[j] = mldts->q_m[j]; + ops[i]->mldts.r_rep = mldts->r_rep; + ops[i]->mldts.c_rep = mldts->c_rep; + ops[i]->mldts.r_input = r_inputs[start_idx + i]; + ops[i]->mldts.qhy_input = q_inputs[start_idx + i]; + ops[i]->mldts.output = outputs[start_idx + i]; + } +} + static int check_dec_status_and_ordering(struct rte_bbdev_dec_op *op, unsigned int order_idx, const int expected_status) @@ -2039,6 +2085,21 @@ check_fft_status_and_ordering(struct rte_bbdev_fft_op *op, return TEST_SUCCESS; } +static int +check_mldts_status_and_ordering(struct rte_bbdev_mldts_op *op, + unsigned int order_idx, const int expected_status) +{ + TEST_ASSERT(op->status == expected_status, + "op_status (%d) != expected_status (%d)", + op->status, expected_status); + + TEST_ASSERT((void *)(uintptr_t)order_idx == op->opaque_data, + "Ordering error, expected %p, got %p", + (void *)(uintptr_t)order_idx, op->opaque_data); + + return TEST_SUCCESS; +} + static inline int validate_op_chain(struct rte_bbdev_op_data *op, struct op_data_entries *orig_op) @@ -2554,6 +2615,57 @@ validate_op_fft_chain(struct rte_bbdev_op_data *op, struct op_data_entries *orig return TEST_SUCCESS; } +static inline int +validate_op_mldts_chain(struct rte_bbdev_op_data *op, + struct op_data_entries *orig_op) +{ + uint8_t i; + struct rte_mbuf *m = op->data; + uint8_t nb_dst_segments = orig_op->nb_segments; + /*the result is not bit exact*/ + int16_t thres_hold = 3; + int16_t delt, abs_delt; + uint32_t j, data_len_iq; + uint32_t error_num; + int8_t *ref_out; + int8_t *op_out; + + TEST_ASSERT(nb_dst_segments == m->nb_segs, + "Number of segments differ in original (%u) and filled (%u) op mldts", + nb_dst_segments, m->nb_segs); + + /* Due to size limitation of mbuf, MLDTS doesn't use real mbuf. */ + for (i = 0; i < nb_dst_segments; ++i) { + uint16_t offset = (i == 0) ? op->offset : 0; + uint32_t data_len = op->length; + + TEST_ASSERT(orig_op->segments[i].length == data_len, + "Length of segment differ in original (%u) and filled (%u) op mldts", + orig_op->segments[i].length, data_len); + data_len_iq = data_len; + ref_out = (int8_t *)(orig_op->segments[i].addr); + op_out = rte_pktmbuf_mtod_offset(m, int8_t *, offset), + error_num = 0; + for (j = 0; j < data_len_iq; j++) { + + delt = ref_out[j] - op_out[j]; + abs_delt = delt > 0 ? delt : -delt; + error_num += (abs_delt > thres_hold ? 1 : 0); + if (error_num > 0) + printf("MLD Error %d: Exp %x %d Actual %x %d Diff %d\n", + j, ref_out[j], ref_out[j], op_out[j], op_out[j], + delt); + } + TEST_ASSERT(error_num == 0, + "MLDTS Output are not matched total (%u) errors (%u)", + data_len_iq, error_num); + + m = m->next; + } + + return TEST_SUCCESS; +} + static int validate_fft_op(struct rte_bbdev_fft_op **ops, const uint16_t n, struct rte_bbdev_fft_op *ref_op) @@ -2578,6 +2690,28 @@ validate_fft_op(struct rte_bbdev_fft_op **ops, const uint16_t n, return TEST_SUCCESS; } +static int +validate_mldts_op(struct rte_bbdev_mldts_op **ops, const uint16_t n, + struct rte_bbdev_mldts_op *ref_op) +{ + unsigned int i; + int ret; + struct op_data_entries *mldts_data_orig = + &test_vector.entries[DATA_HARD_OUTPUT]; + for (i = 0; i < n; ++i) { + ret = check_mldts_status_and_ordering(ops[i], i, ref_op->status); + TEST_ASSERT_SUCCESS(ret, + "Checking status and ordering for MLDTS failed"); + TEST_ASSERT_SUCCESS(validate_op_mldts_chain( + &ops[i]->mldts.output, + mldts_data_orig), + "MLDTS Output buffers (op=%u) are not matched", + i); + } + + return TEST_SUCCESS; +} + static void create_reference_dec_op(struct rte_bbdev_dec_op *op) { @@ -2622,6 +2756,20 @@ create_reference_fft_op(struct rte_bbdev_fft_op *op) op->fft.base_input.length += entry->segments[i].length; } +static void +create_reference_mldts_op(struct rte_bbdev_mldts_op *op) +{ + unsigned int i; + struct op_data_entries *entry; + op->mldts = test_vector.mldts; + entry = &test_vector.entries[DATA_INPUT]; + for (i = 0; i < entry->nb_segments; ++i) + op->mldts.qhy_input.length += entry->segments[i].length; + entry = &test_vector.entries[DATA_HARQ_INPUT]; + for (i = 0; i < entry->nb_segments; ++i) + op->mldts.r_input.length += entry->segments[i].length; +} + static void create_reference_enc_op(struct rte_bbdev_enc_op *op) { @@ -2730,6 +2878,14 @@ calc_fft_size(struct rte_bbdev_fft_op *op) return output_size; } +static uint32_t +calc_mldts_size(struct rte_bbdev_mldts_op *op) +{ + uint32_t output_size; + output_size = op->mldts.num_layers * op->mldts.num_rbs * op->mldts.c_rep; + return output_size; +} + static int init_test_op_params(struct test_op_params *op_params, enum rte_bbdev_op_type op_type, const int expected_status, @@ -2744,6 +2900,9 @@ init_test_op_params(struct test_op_params *op_params, else if (op_type == RTE_BBDEV_OP_FFT) ret = rte_bbdev_fft_op_alloc_bulk(ops_mp, &op_params->ref_fft_op, 1); + else if (op_type == RTE_BBDEV_OP_MLDTS) + ret = rte_bbdev_mldts_op_alloc_bulk(ops_mp, + &op_params->ref_mldts_op, 1); else ret = rte_bbdev_enc_op_alloc_bulk(ops_mp, &op_params->ref_enc_op, 1); @@ -2763,6 +2922,8 @@ init_test_op_params(struct test_op_params *op_params, op_params->ref_enc_op->status = expected_status; else if (op_type == RTE_BBDEV_OP_FFT) op_params->ref_fft_op->status = expected_status; + else if (op_type == RTE_BBDEV_OP_MLDTS) + op_params->ref_mldts_op->status = expected_status; return 0; } @@ -2831,6 +2992,8 @@ run_test_case_on_device(test_case_function *test_case_func, uint8_t dev_id, create_reference_ldpc_dec_op(op_params->ref_dec_op); else if (test_vector.op_type == RTE_BBDEV_OP_FFT) create_reference_fft_op(op_params->ref_fft_op); + else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) + create_reference_mldts_op(op_params->ref_mldts_op); for (i = 0; i < ad->nb_queues; ++i) { f_ret = fill_queue_buffers(op_params, @@ -3047,6 +3210,11 @@ dequeue_event_callback(uint16_t dev_id, &tp->fft_ops[ __atomic_load_n(&tp->nb_dequeued, __ATOMIC_RELAXED)], burst_sz); + else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) + deq = rte_bbdev_dequeue_mldts_ops(dev_id, queue_id, + &tp->mldts_ops[ + __atomic_load_n(&tp->nb_dequeued, __ATOMIC_RELAXED)], + burst_sz); else /*RTE_BBDEV_OP_TURBO_ENC*/ deq = rte_bbdev_dequeue_enc_ops(dev_id, queue_id, &tp->enc_ops[ @@ -3093,6 +3261,10 @@ dequeue_event_callback(uint16_t dev_id, struct rte_bbdev_fft_op *ref_op = tp->op_params->ref_fft_op; ret = validate_fft_op(tp->fft_ops, num_ops, ref_op); rte_bbdev_fft_op_free_bulk(tp->fft_ops, deq); + } else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) { + struct rte_bbdev_mldts_op *ref_op = tp->op_params->ref_mldts_op; + ret = validate_mldts_op(tp->mldts_ops, num_ops, ref_op); + rte_bbdev_mldts_op_free_bulk(tp->mldts_ops, deq); } else if (test_vector.op_type == RTE_BBDEV_OP_LDPC_DEC) { struct rte_bbdev_dec_op *ref_op = tp->op_params->ref_dec_op; ret = validate_ldpc_dec_op(tp->dec_ops, num_ops, ref_op, @@ -3118,6 +3290,9 @@ dequeue_event_callback(uint16_t dev_id, case RTE_BBDEV_OP_FFT: tb_len_bits = calc_fft_size(tp->op_params->ref_fft_op); break; + case RTE_BBDEV_OP_MLDTS: + tb_len_bits = calc_mldts_size(tp->op_params->ref_mldts_op); + break; case RTE_BBDEV_OP_LDPC_ENC: tb_len_bits = calc_ldpc_enc_TB_size(tp->op_params->ref_enc_op); break; @@ -3593,6 +3768,88 @@ throughput_intr_lcore_fft(void *arg) return TEST_SUCCESS; } +static int +throughput_intr_lcore_mldts(void *arg) +{ + struct thread_params *tp = arg; + unsigned int enqueued; + const uint16_t queue_id = tp->queue_id; + const uint16_t burst_sz = tp->op_params->burst_sz; + const uint16_t num_to_process = tp->op_params->num_to_process; + struct rte_bbdev_mldts_op *ops[num_to_process]; + struct test_buffers *bufs = NULL; + struct rte_bbdev_info info; + int ret, i, j; + uint16_t num_to_enq, enq; + + TEST_ASSERT_SUCCESS((burst_sz > MAX_BURST), "BURST_SIZE should be <= %u", MAX_BURST); + + TEST_ASSERT_SUCCESS(rte_bbdev_queue_intr_enable(tp->dev_id, queue_id), + "Failed to enable interrupts for dev: %u, queue_id: %u", + tp->dev_id, queue_id); + + rte_bbdev_info_get(tp->dev_id, &info); + + TEST_ASSERT_SUCCESS((num_to_process > info.drv.queue_size_lim), + "NUM_OPS cannot exceed %u for this device", + info.drv.queue_size_lim); + + bufs = &tp->op_params->q_bufs[GET_SOCKET(info.socket_id)][queue_id]; + + __atomic_store_n(&tp->processing_status, 0, __ATOMIC_RELAXED); + __atomic_store_n(&tp->nb_dequeued, 0, __ATOMIC_RELAXED); + + rte_wait_until_equal_16(&tp->op_params->sync, SYNC_START, __ATOMIC_RELAXED); + + ret = rte_bbdev_mldts_op_alloc_bulk(tp->op_params->mp, ops, num_to_process); + TEST_ASSERT_SUCCESS(ret, "Allocation failed for %d ops", num_to_process); + if (test_vector.op_type != RTE_BBDEV_OP_NONE) + copy_reference_mldts_op(ops, num_to_process, 0, bufs->inputs, bufs->harq_inputs, + bufs->hard_outputs, tp->op_params->ref_mldts_op); + + /* Set counter to validate the ordering */ + for (j = 0; j < num_to_process; ++j) + ops[j]->opaque_data = (void *)(uintptr_t)j; + + for (j = 0; j < TEST_REPETITIONS; ++j) { + for (i = 0; i < num_to_process; ++i) + mbuf_reset(ops[i]->mldts.output.data); + + tp->start_time = rte_rdtsc_precise(); + for (enqueued = 0; enqueued < num_to_process;) { + num_to_enq = burst_sz; + + if (unlikely(num_to_process - enqueued < num_to_enq)) + num_to_enq = num_to_process - enqueued; + + enq = 0; + do { + enq += rte_bbdev_enqueue_mldts_ops(tp->dev_id, + queue_id, &ops[enqueued], num_to_enq); + } while (unlikely(enq != num_to_enq)); + enqueued += enq; + + /* Write to thread burst_sz current number of enqueued + * descriptors. It ensures that proper number of + * descriptors will be dequeued in callback + * function - needed for last batch in case where + * the number of operations is not a multiple of + * burst size. + */ + __atomic_store_n(&tp->burst_sz, num_to_enq, __ATOMIC_RELAXED); + + /* Wait until processing of previous batch is + * completed + */ + rte_wait_until_equal_16(&tp->nb_dequeued, enqueued, __ATOMIC_RELAXED); + } + if (j != TEST_REPETITIONS - 1) + __atomic_store_n(&tp->nb_dequeued, 0, __ATOMIC_RELAXED); + } + + return TEST_SUCCESS; +} + static int throughput_pmd_lcore_dec(void *arg) { @@ -4403,6 +4660,104 @@ throughput_pmd_lcore_fft(void *arg) return TEST_SUCCESS; } +static int +throughput_pmd_lcore_mldts(void *arg) +{ + struct thread_params *tp = arg; + uint16_t enq, deq; + uint64_t total_time = 0, start_time; + const uint16_t queue_id = tp->queue_id; + const uint16_t burst_sz = tp->op_params->burst_sz; + const uint16_t num_ops = tp->op_params->num_to_process; + struct rte_bbdev_mldts_op *ops_enq[num_ops]; + struct rte_bbdev_mldts_op *ops_deq[num_ops]; + struct rte_bbdev_mldts_op *ref_op = tp->op_params->ref_mldts_op; + struct test_buffers *bufs = NULL; + int i, j, ret; + struct rte_bbdev_info info; + uint16_t num_to_enq; + + TEST_ASSERT_SUCCESS((burst_sz > MAX_BURST), "BURST_SIZE should be <= %u", MAX_BURST); + + rte_bbdev_info_get(tp->dev_id, &info); + + TEST_ASSERT_SUCCESS((num_ops > info.drv.queue_size_lim), + "NUM_OPS cannot exceed %u for this device", + info.drv.queue_size_lim); + + bufs = &tp->op_params->q_bufs[GET_SOCKET(info.socket_id)][queue_id]; + + rte_wait_until_equal_16(&tp->op_params->sync, SYNC_START, __ATOMIC_RELAXED); + + ret = rte_bbdev_mldts_op_alloc_bulk(tp->op_params->mp, ops_enq, num_ops); + TEST_ASSERT_SUCCESS(ret, "Allocation failed for %d ops", num_ops); + + if (test_vector.op_type != RTE_BBDEV_OP_NONE) + copy_reference_mldts_op(ops_enq, num_ops, 0, bufs->inputs, bufs->harq_inputs, + bufs->hard_outputs, ref_op); + + /* Set counter to validate the ordering */ + for (j = 0; j < num_ops; ++j) + ops_enq[j]->opaque_data = (void *)(uintptr_t)j; + + for (i = 0; i < TEST_REPETITIONS; ++i) { + uint32_t time_out = 0; + for (j = 0; j < num_ops; ++j) + mbuf_reset(ops_enq[j]->mldts.output.data); + + start_time = rte_rdtsc_precise(); + + for (enq = 0, deq = 0; enq < num_ops;) { + num_to_enq = burst_sz; + + if (unlikely(num_ops - enq < num_to_enq)) + num_to_enq = num_ops - enq; + + enq += rte_bbdev_enqueue_mldts_ops(tp->dev_id, + queue_id, &ops_enq[enq], num_to_enq); + + deq += rte_bbdev_dequeue_mldts_ops(tp->dev_id, + queue_id, &ops_deq[deq], enq - deq); + time_out++; + if (time_out >= TIME_OUT_POLL) { + timeout_exit(tp->dev_id); + TEST_ASSERT_SUCCESS(TEST_FAILED, "Enqueue timeout!"); + } + } + + /* dequeue the remaining */ + time_out = 0; + while (deq < enq) { + deq += rte_bbdev_dequeue_mldts_ops(tp->dev_id, + queue_id, &ops_deq[deq], enq - deq); + time_out++; + if (time_out >= TIME_OUT_POLL) { + timeout_exit(tp->dev_id); + TEST_ASSERT_SUCCESS(TEST_FAILED, "Dequeue timeout!"); + } + } + + total_time += rte_rdtsc_precise() - start_time; + } + + if (test_vector.op_type != RTE_BBDEV_OP_NONE) { + ret = validate_mldts_op(ops_deq, num_ops, ref_op); + TEST_ASSERT_SUCCESS(ret, "Validation failed!"); + } + + rte_bbdev_mldts_op_free_bulk(ops_enq, num_ops); + + double tb_len_bits = calc_mldts_size(ref_op); + + tp->ops_per_sec = ((double)num_ops * TEST_REPETITIONS) / + ((double)total_time / (double)rte_get_tsc_hz()); + tp->mbps = (((double)(num_ops * TEST_REPETITIONS * tb_len_bits)) / + 1000000.0) / ((double)total_time / + (double)rte_get_tsc_hz()); + + return TEST_SUCCESS; +} + static void print_enc_throughput(struct thread_params *t_params, unsigned int used_cores) { @@ -4624,6 +4979,8 @@ throughput_test(struct active_device *ad, throughput_function = throughput_intr_lcore_ldpc_enc; else if (test_vector.op_type == RTE_BBDEV_OP_FFT) throughput_function = throughput_intr_lcore_fft; + else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) + throughput_function = throughput_intr_lcore_mldts; else throughput_function = throughput_intr_lcore_enc; @@ -4646,6 +5003,8 @@ throughput_test(struct active_device *ad, throughput_function = throughput_pmd_lcore_ldpc_enc; else if (test_vector.op_type == RTE_BBDEV_OP_FFT) throughput_function = throughput_pmd_lcore_fft; + else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) + throughput_function = throughput_pmd_lcore_mldts; else throughput_function = throughput_pmd_lcore_enc; } @@ -5139,6 +5498,77 @@ latency_test_fft(struct rte_mempool *mempool, return i; } +static int +latency_test_mldts(struct rte_mempool *mempool, + struct test_buffers *bufs, struct rte_bbdev_mldts_op *ref_op, + uint16_t dev_id, uint16_t queue_id, + const uint16_t num_to_process, uint16_t burst_sz, + uint64_t *total_time, uint64_t *min_time, uint64_t *max_time) +{ + int ret = TEST_SUCCESS; + uint16_t i, j, dequeued; + struct rte_bbdev_mldts_op *ops_enq[MAX_BURST], *ops_deq[MAX_BURST]; + uint64_t start_time = 0, last_time = 0; + + for (i = 0, dequeued = 0; dequeued < num_to_process; ++i) { + uint16_t enq = 0, deq = 0; + uint32_t time_out = 0; + bool first_time = true; + last_time = 0; + + if (unlikely(num_to_process - dequeued < burst_sz)) + burst_sz = num_to_process - dequeued; + + ret = rte_bbdev_mldts_op_alloc_bulk(mempool, ops_enq, burst_sz); + TEST_ASSERT_SUCCESS(ret, "rte_bbdev_mldts_op_alloc_bulk() failed"); + if (test_vector.op_type != RTE_BBDEV_OP_NONE) + copy_reference_mldts_op(ops_enq, burst_sz, dequeued, + bufs->inputs, bufs->harq_inputs, + bufs->hard_outputs, + ref_op); + + /* Set counter to validate the ordering */ + for (j = 0; j < burst_sz; ++j) + ops_enq[j]->opaque_data = (void *)(uintptr_t)j; + + start_time = rte_rdtsc_precise(); + + enq = rte_bbdev_enqueue_mldts_ops(dev_id, queue_id, &ops_enq[enq], burst_sz); + TEST_ASSERT(enq == burst_sz, + "Error enqueueing burst, expected %u, got %u", + burst_sz, enq); + + /* Dequeue */ + do { + deq += rte_bbdev_dequeue_mldts_ops(dev_id, queue_id, + &ops_deq[deq], burst_sz - deq); + if (likely(first_time && (deq > 0))) { + last_time += rte_rdtsc_precise() - start_time; + first_time = false; + } + time_out++; + if (time_out >= TIME_OUT_POLL) { + timeout_exit(dev_id); + TEST_ASSERT_SUCCESS(TEST_FAILED, "Dequeue timeout!"); + } + } while (unlikely(burst_sz != deq)); + + *max_time = RTE_MAX(*max_time, last_time); + *min_time = RTE_MIN(*min_time, last_time); + *total_time += last_time; + + if (test_vector.op_type != RTE_BBDEV_OP_NONE) { + ret = validate_mldts_op(ops_deq, burst_sz, ref_op); + TEST_ASSERT_SUCCESS(ret, "Validation failed!"); + } + + rte_bbdev_mldts_op_free_bulk(ops_enq, deq); + dequeued += deq; + } + + return i; +} + /* Common function for running validation and latency test cases */ static int validation_latency_test(struct active_device *ad, @@ -5196,6 +5626,12 @@ validation_latency_test(struct active_device *ad, ad->dev_id, queue_id, num_to_process, burst_sz, &total_time, &min_time, &max_time); + else if (op_type == RTE_BBDEV_OP_MLDTS) + iter = latency_test_mldts(op_params->mp, bufs, + op_params->ref_mldts_op, + ad->dev_id, queue_id, + num_to_process, burst_sz, &total_time, + &min_time, &max_time); else /* RTE_BBDEV_OP_TURBO_ENC */ iter = latency_test_enc(op_params->mp, bufs, op_params->ref_enc_op, @@ -5337,6 +5773,85 @@ offload_latency_test_fft(struct rte_mempool *mempool, struct test_buffers *bufs, return i; } +static int +offload_latency_test_mldts(struct rte_mempool *mempool, struct test_buffers *bufs, + struct rte_bbdev_mldts_op *ref_op, uint16_t dev_id, + uint16_t queue_id, const uint16_t num_to_process, + uint16_t burst_sz, struct test_time_stats *time_st) +{ + int i, dequeued, ret; + struct rte_bbdev_mldts_op *ops_enq[MAX_BURST], *ops_deq[MAX_BURST]; + uint64_t enq_start_time, deq_start_time; + uint64_t enq_sw_last_time, deq_last_time; + struct rte_bbdev_stats stats; + + for (i = 0, dequeued = 0; dequeued < num_to_process; ++i) { + uint16_t enq = 0, deq = 0; + + if (unlikely(num_to_process - dequeued < burst_sz)) + burst_sz = num_to_process - dequeued; + + ret = rte_bbdev_mldts_op_alloc_bulk(mempool, ops_enq, burst_sz); + TEST_ASSERT_SUCCESS(ret, "rte_bbdev_mldts_op_alloc_bulk() failed"); + if (test_vector.op_type != RTE_BBDEV_OP_NONE) + copy_reference_mldts_op(ops_enq, burst_sz, dequeued, + bufs->inputs, bufs->harq_inputs, + bufs->hard_outputs, + ref_op); + + /* Start time meas for enqueue function offload latency */ + enq_start_time = rte_rdtsc_precise(); + do { + enq += rte_bbdev_enqueue_mldts_ops(dev_id, queue_id, + &ops_enq[enq], burst_sz - enq); + } while (unlikely(burst_sz != enq)); + + ret = get_bbdev_queue_stats(dev_id, queue_id, &stats); + TEST_ASSERT_SUCCESS(ret, + "Failed to get stats for queue (%u) of device (%u)", + queue_id, dev_id); + + enq_sw_last_time = rte_rdtsc_precise() - enq_start_time - + stats.acc_offload_cycles; + time_st->enq_sw_max_time = RTE_MAX(time_st->enq_sw_max_time, + enq_sw_last_time); + time_st->enq_sw_min_time = RTE_MIN(time_st->enq_sw_min_time, + enq_sw_last_time); + time_st->enq_sw_total_time += enq_sw_last_time; + + time_st->enq_acc_max_time = RTE_MAX(time_st->enq_acc_max_time, + stats.acc_offload_cycles); + time_st->enq_acc_min_time = RTE_MIN(time_st->enq_acc_min_time, + stats.acc_offload_cycles); + time_st->enq_acc_total_time += stats.acc_offload_cycles; + + /* give time for device to process ops */ + rte_delay_us(WAIT_OFFLOAD_US); + + /* Start time meas for dequeue function offload latency */ + deq_start_time = rte_rdtsc_precise(); + /* Dequeue one operation */ + do { + deq += rte_bbdev_dequeue_mldts_ops(dev_id, queue_id, &ops_deq[deq], enq); + } while (unlikely(deq == 0)); + + deq_last_time = rte_rdtsc_precise() - deq_start_time; + time_st->deq_max_time = RTE_MAX(time_st->deq_max_time, deq_last_time); + time_st->deq_min_time = RTE_MIN(time_st->deq_min_time, deq_last_time); + time_st->deq_total_time += deq_last_time; + + /* Dequeue remaining operations if needed*/ + while (burst_sz != deq) + deq += rte_bbdev_dequeue_mldts_ops(dev_id, queue_id, + &ops_deq[deq], burst_sz - deq); + + rte_bbdev_mldts_op_free_bulk(ops_enq, deq); + dequeued += deq; + } + + return i; +} + static int offload_latency_test_dec(struct rte_mempool *mempool, struct test_buffers *bufs, struct rte_bbdev_dec_op *ref_op, uint16_t dev_id, @@ -5734,6 +6249,10 @@ offload_cost_test(struct active_device *ad, iter = offload_latency_test_fft(op_params->mp, bufs, op_params->ref_fft_op, ad->dev_id, queue_id, num_to_process, burst_sz, &time_st); + else if (op_type == RTE_BBDEV_OP_MLDTS) + iter = offload_latency_test_mldts(op_params->mp, bufs, + op_params->ref_mldts_op, ad->dev_id, queue_id, + num_to_process, burst_sz, &time_st); else iter = offload_latency_test_enc(op_params->mp, bufs, op_params->ref_enc_op, ad->dev_id, queue_id, diff --git a/app/test-bbdev/test_bbdev_vector.c b/app/test-bbdev/test_bbdev_vector.c index 0ef1481f2a..8f464db838 100644 --- a/app/test-bbdev/test_bbdev_vector.c +++ b/app/test-bbdev/test_bbdev_vector.c @@ -244,6 +244,20 @@ op_fft_flag_strtoul(char *token, uint32_t *op_flag_value) return 0; } +/* convert MLD flag from string to unsigned long int*/ +static int +op_mld_flag_strtoul(char *token, uint32_t *op_flag_value) +{ + if (!strcmp(token, "RTE_BBDEV_MLDTS_REP")) + *op_flag_value = RTE_BBDEV_MLDTS_REP; + else { + printf("The given value is not a MLD flag\n"); + return -1; + } + + return 0; +} + /* convert turbo encoder flag from string to unsigned long int*/ static int op_encoder_flag_strtoul(char *token, uint32_t *op_flag_value) @@ -326,6 +340,10 @@ parse_turbo_flags(char *tokens, uint32_t *op_flags, if (op_fft_flag_strtoul(tok, &op_flag_value) == -1) return -1; + } else if (op_type == RTE_BBDEV_OP_MLDTS) { + if (op_mld_flag_strtoul(tok, &op_flag_value) + == -1) + return -1; } else { return -1; } @@ -355,6 +373,8 @@ op_turbo_type_strtol(char *token, enum rte_bbdev_op_type *op_type) *op_type = RTE_BBDEV_OP_LDPC_DEC; else if (!strcmp(token, "RTE_BBDEV_OP_FFT")) *op_type = RTE_BBDEV_OP_FFT; + else if (!strcmp(token, "RTE_BBDEV_OP_MLDTS")) + *op_type = RTE_BBDEV_OP_MLDTS; else if (!strcmp(token, "RTE_BBDEV_OP_NONE")) *op_type = RTE_BBDEV_OP_NONE; else { @@ -992,6 +1012,73 @@ parse_fft_params(const char *key_token, char *token, return 0; } +/* parses MLD parameters and assigns to global variable */ +static int +parse_mld_params(const char *key_token, char *token, + struct test_bbdev_vector *vector) +{ + int ret = 0, status = 0; + uint32_t op_flags = 0; + char *err = NULL; + + struct rte_bbdev_op_mldts *mld = &vector->mldts; + + if (starts_with(key_token, "qhy_input")) { + ret = parse_data_entry(key_token, token, vector, + DATA_INPUT, "qhy_input"); + } else if (starts_with(key_token, "r_input")) { + ret = parse_data_entry(key_token, token, vector, + DATA_HARQ_INPUT, "r_input"); + } else if (starts_with(key_token, "output")) { + ret = parse_data_entry(key_token, token, vector, + DATA_HARD_OUTPUT, "output"); + } else if (!strcmp(key_token, "layers")) { + mld->num_layers = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "layer1")) { + mld->q_m[0] = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "layer2")) { + mld->q_m[1] = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "layer3")) { + mld->q_m[2] = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "layer4")) { + mld->q_m[3] = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "crep")) { + mld->c_rep = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "rrep")) { + mld->r_rep = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "rbs")) { + mld->num_rbs = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "op_flags")) { + vector->mask |= TEST_BBDEV_VF_OP_FLAGS; + ret = parse_turbo_flags(token, &op_flags, vector->op_type); + if (!ret) + mld->op_flags = op_flags; + } else if (!strcmp(key_token, "expected_status")) { + vector->mask |= TEST_BBDEV_VF_EXPECTED_STATUS; + ret = parse_expected_status(token, &status, vector->op_type); + if (!ret) + vector->expected_status = status; + } else { + printf("Not valid mld key: '%s'\n", key_token); + return -1; + } + + if (ret != 0) { + printf("Failed with convert '%s\t%s'\n", key_token, token); + return -1; + } + + return 0; +} + /* checks the type of key and assigns data */ static int parse_entry(char *entry, struct test_bbdev_vector *vector) @@ -1046,6 +1133,9 @@ parse_entry(char *entry, struct test_bbdev_vector *vector) } else if (vector->op_type == RTE_BBDEV_OP_FFT) { if (parse_fft_params(key_token, token, vector) == -1) return -1; + } else if (vector->op_type == RTE_BBDEV_OP_MLDTS) { + if (parse_mld_params(key_token, token, vector) == -1) + return -1; } return 0; @@ -1132,6 +1222,25 @@ check_fft_segments(struct test_bbdev_vector *vector) return 0; } +static int +check_mld_segments(struct test_bbdev_vector *vector) +{ + unsigned char i; + + for (i = 0; i < vector->entries[DATA_INPUT].nb_segments; i++) + if (vector->entries[DATA_INPUT].segments[i].addr == NULL) + return -1; + + for (i = 0; i < vector->entries[DATA_HARQ_INPUT].nb_segments; i++) + if (vector->entries[DATA_HARQ_INPUT].segments[i].addr == NULL) + return -1; + + for (i = 0; i < vector->entries[DATA_HARD_OUTPUT].nb_segments; i++) + if (vector->entries[DATA_HARD_OUTPUT].segments[i].addr == NULL) + return -1; + return 0; +} + static int check_decoder_llr_spec(struct test_bbdev_vector *vector) { @@ -1359,6 +1468,26 @@ check_fft(struct test_bbdev_vector *vector) return 0; } +/* checks mld parameters */ +static int +check_mld(struct test_bbdev_vector *vector) +{ + const int mask = vector->mask; + + if (check_mld_segments(vector) < 0) + return -1; + + /* Check which params were set */ + if (!(mask & TEST_BBDEV_VF_OP_FLAGS)) { + printf( + "WARNING: op_flags was not specified in vector file and capabilities will not be validated\n"); + } + if (!(mask & TEST_BBDEV_VF_EXPECTED_STATUS)) + printf( + "WARNING: expected_status was not specified in vector file and will be set to 0\n"); + return 0; +} + /* checks encoder parameters */ static int check_encoder(struct test_bbdev_vector *vector) @@ -1520,6 +1649,9 @@ bbdev_check_vector(struct test_bbdev_vector *vector) } else if (vector->op_type == RTE_BBDEV_OP_FFT) { if (check_fft(vector) == -1) return -1; + } else if (vector->op_type == RTE_BBDEV_OP_MLDTS) { + if (check_mld(vector) == -1) + return -1; } else if (vector->op_type != RTE_BBDEV_OP_NONE) { printf("Vector was not filled\n"); return -1; diff --git a/app/test-bbdev/test_bbdev_vector.h b/app/test-bbdev/test_bbdev_vector.h index 2ea271ffb7..14b8ef2764 100644 --- a/app/test-bbdev/test_bbdev_vector.h +++ b/app/test-bbdev/test_bbdev_vector.h @@ -65,6 +65,7 @@ struct test_bbdev_vector { struct rte_bbdev_op_ldpc_dec ldpc_dec; struct rte_bbdev_op_ldpc_enc ldpc_enc; struct rte_bbdev_op_fft fft; + struct rte_bbdev_op_mldts mldts; }; /* Additional storage for op data entries */ struct op_data_entries entries[DATA_NUM_TYPES]; -- 2.34.1 ^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [PATCH v3 07/10] test/bbdev: add MLD support 2023-11-03 23:34 ` [PATCH v3 07/10] test/bbdev: add MLD support Nicolas Chautru @ 2023-11-09 15:44 ` Maxime Coquelin 0 siblings, 0 replies; 18+ messages in thread From: Maxime Coquelin @ 2023-11-09 15:44 UTC (permalink / raw) To: Nicolas Chautru, dev Cc: hemant.agrawal, david.marchand, hernan.vargas, stable On 11/4/23 00:34, Nicolas Chautru wrote: > From: Hernan Vargas <hernan.vargas@intel.com> > > Adding test-bbdev support for the MLD-TS processing specific to the VRB2 > variant. > > Signed-off-by: Hernan Vargas <hernan.vargas@intel.com> > Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> > --- > app/test-bbdev/test_bbdev_perf.c | 519 +++++++++++++++++++++++++++++ > app/test-bbdev/test_bbdev_vector.c | 132 ++++++++ > app/test-bbdev/test_bbdev_vector.h | 1 + > 3 files changed, 652 insertions(+) > > diff --git a/app/test-bbdev/test_bbdev_perf.c b/app/test-bbdev/test_bbdev_perf.c > index 5f1e5de002..d2e3542356 100644 > --- a/app/test-bbdev/test_bbdev_perf.c > +++ b/app/test-bbdev/test_bbdev_perf.c > @@ -139,6 +139,7 @@ struct test_op_params { > struct rte_bbdev_dec_op *ref_dec_op; > struct rte_bbdev_enc_op *ref_enc_op; > struct rte_bbdev_fft_op *ref_fft_op; > + struct rte_bbdev_mldts_op *ref_mldts_op; > uint16_t burst_sz; > uint16_t num_to_process; > uint16_t num_lcores; > @@ -165,6 +166,7 @@ struct thread_params { > struct rte_bbdev_dec_op *dec_ops[MAX_BURST]; > struct rte_bbdev_enc_op *enc_ops[MAX_BURST]; > struct rte_bbdev_fft_op *fft_ops[MAX_BURST]; > + struct rte_bbdev_mldts_op *mldts_ops[MAX_BURST]; > }; > > /* Stores time statistics */ > @@ -472,6 +474,18 @@ check_dev_cap(const struct rte_bbdev_info *dev_info) > return TEST_FAILED; > } > return TEST_SUCCESS; > + } else if (op_cap->type == RTE_BBDEV_OP_MLDTS) { > + const struct rte_bbdev_op_cap_mld *cap = &op_cap->cap.mld; > + if (!flags_match(test_vector.mldts.op_flags, cap->capability_flags)) { > + printf("Flag Mismatch\n"); > + return TEST_FAILED; > + } > + if (nb_inputs > cap->num_buffers_src) { > + printf("Too many inputs defined: %u, max: %u\n", > + nb_inputs, cap->num_buffers_src); > + return TEST_FAILED; > + } > + return TEST_SUCCESS; > } > } > > @@ -822,6 +836,9 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info, > conf.arb_fft[i].gbr_threshold1 = VRB_QOS_GBR; > conf.arb_fft[i].gbr_threshold1 = VRB_QOS_GBR; > conf.arb_fft[i].round_robin_weight = VRB_QMGR_RR; > + conf.arb_mld[i].gbr_threshold1 = VRB_QOS_GBR; > + conf.arb_mld[i].gbr_threshold1 = VRB_QOS_GBR; > + conf.arb_mld[i].round_robin_weight = VRB_QMGR_RR; > } > > conf.input_pos_llr_1_bit = true; > @@ -847,6 +864,10 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info, > conf.q_fft.num_qgroups = VRB_QMGR_NUM_QGS; > conf.q_fft.first_qgroup_index = VRB_QMGR_INVALID_IDX; > conf.q_fft.num_aqs_per_groups = VRB_QMGR_NUM_AQS; > + conf.q_mld.num_qgroups = VRB_QMGR_NUM_QGS; > + conf.q_mld.first_qgroup_index = VRB_QMGR_INVALID_IDX; > + conf.q_mld.num_aqs_per_groups = VRB_QMGR_NUM_AQS; > + conf.q_mld.aq_depth_log2 = VRB_QMGR_AQ_DEPTH; > > /* setup PF with configuration information */ > ret = rte_acc_configure(info->dev_name, &conf); > @@ -1979,6 +2000,31 @@ copy_reference_fft_op(struct rte_bbdev_fft_op **ops, unsigned int n, > } > } > > +static void > +copy_reference_mldts_op(struct rte_bbdev_mldts_op **ops, unsigned int n, > + unsigned int start_idx, > + struct rte_bbdev_op_data *q_inputs, > + struct rte_bbdev_op_data *r_inputs, > + struct rte_bbdev_op_data *outputs, > + struct rte_bbdev_mldts_op *ref_op) > +{ > + unsigned int i, j; > + struct rte_bbdev_op_mldts *mldts = &ref_op->mldts; > + for (i = 0; i < n; i++) { > + ops[i]->mldts.c_rep = mldts->c_rep; > + ops[i]->mldts.num_layers = mldts->num_layers; > + ops[i]->mldts.num_rbs = mldts->num_rbs; > + ops[i]->mldts.op_flags = mldts->op_flags; > + for (j = 0; j < RTE_BBDEV_MAX_MLD_LAYERS; j++) > + ops[i]->mldts.q_m[j] = mldts->q_m[j]; > + ops[i]->mldts.r_rep = mldts->r_rep; > + ops[i]->mldts.c_rep = mldts->c_rep; > + ops[i]->mldts.r_input = r_inputs[start_idx + i]; > + ops[i]->mldts.qhy_input = q_inputs[start_idx + i]; > + ops[i]->mldts.output = outputs[start_idx + i]; > + } > +} > + > static int > check_dec_status_and_ordering(struct rte_bbdev_dec_op *op, > unsigned int order_idx, const int expected_status) > @@ -2039,6 +2085,21 @@ check_fft_status_and_ordering(struct rte_bbdev_fft_op *op, > return TEST_SUCCESS; > } > > +static int > +check_mldts_status_and_ordering(struct rte_bbdev_mldts_op *op, > + unsigned int order_idx, const int expected_status) > +{ > + TEST_ASSERT(op->status == expected_status, > + "op_status (%d) != expected_status (%d)", > + op->status, expected_status); > + > + TEST_ASSERT((void *)(uintptr_t)order_idx == op->opaque_data, > + "Ordering error, expected %p, got %p", > + (void *)(uintptr_t)order_idx, op->opaque_data); > + > + return TEST_SUCCESS; > +} > + > static inline int > validate_op_chain(struct rte_bbdev_op_data *op, > struct op_data_entries *orig_op) > @@ -2554,6 +2615,57 @@ validate_op_fft_chain(struct rte_bbdev_op_data *op, struct op_data_entries *orig > return TEST_SUCCESS; > } > > +static inline int > +validate_op_mldts_chain(struct rte_bbdev_op_data *op, > + struct op_data_entries *orig_op) > +{ > + uint8_t i; > + struct rte_mbuf *m = op->data; > + uint8_t nb_dst_segments = orig_op->nb_segments; > + /*the result is not bit exact*/ > + int16_t thres_hold = 3; > + int16_t delt, abs_delt; > + uint32_t j, data_len_iq; > + uint32_t error_num; > + int8_t *ref_out; > + int8_t *op_out; > + > + TEST_ASSERT(nb_dst_segments == m->nb_segs, > + "Number of segments differ in original (%u) and filled (%u) op mldts", > + nb_dst_segments, m->nb_segs); > + > + /* Due to size limitation of mbuf, MLDTS doesn't use real mbuf. */ > + for (i = 0; i < nb_dst_segments; ++i) { > + uint16_t offset = (i == 0) ? op->offset : 0; > + uint32_t data_len = op->length; > + > + TEST_ASSERT(orig_op->segments[i].length == data_len, > + "Length of segment differ in original (%u) and filled (%u) op mldts", > + orig_op->segments[i].length, data_len); > + data_len_iq = data_len; > + ref_out = (int8_t *)(orig_op->segments[i].addr); > + op_out = rte_pktmbuf_mtod_offset(m, int8_t *, offset), I missed it while reviewing, but checkpatch did: ### [PATCH 08/13] test/bbdev: add MLD support WARNING:SUSPECT_COMMA_SEMICOLON: Possible comma where semicolon could be used #164: FILE: app/test-bbdev/test_bbdev_perf.c:2647: + op_out = rte_pktmbuf_mtod_offset(m, int8_t *, offset), + error_num = 0; total: 0 errors, 1 warnings, 856 lines checked I fix it directly in my tree, no need to submit new version. > + error_num = 0; > + for (j = 0; j < data_len_iq; j++) { > + > + delt = ref_out[j] - op_out[j]; > + abs_delt = delt > 0 ? delt : -delt; > + error_num += (abs_delt > thres_hold ? 1 : 0); > + if (error_num > 0) > + printf("MLD Error %d: Exp %x %d Actual %x %d Diff %d\n", > + j, ref_out[j], ref_out[j], op_out[j], op_out[j], > + delt); > + } > + TEST_ASSERT(error_num == 0, > + "MLDTS Output are not matched total (%u) errors (%u)", > + data_len_iq, error_num); > + > + m = m->next; > + } > + > + return TEST_SUCCESS; > +} > + > static int > validate_fft_op(struct rte_bbdev_fft_op **ops, const uint16_t n, > struct rte_bbdev_fft_op *ref_op) > @@ -2578,6 +2690,28 @@ validate_fft_op(struct rte_bbdev_fft_op **ops, const uint16_t n, > return TEST_SUCCESS; > } > > +static int > +validate_mldts_op(struct rte_bbdev_mldts_op **ops, const uint16_t n, > + struct rte_bbdev_mldts_op *ref_op) > +{ > + unsigned int i; > + int ret; > + struct op_data_entries *mldts_data_orig = > + &test_vector.entries[DATA_HARD_OUTPUT]; > + for (i = 0; i < n; ++i) { > + ret = check_mldts_status_and_ordering(ops[i], i, ref_op->status); > + TEST_ASSERT_SUCCESS(ret, > + "Checking status and ordering for MLDTS failed"); > + TEST_ASSERT_SUCCESS(validate_op_mldts_chain( > + &ops[i]->mldts.output, > + mldts_data_orig), > + "MLDTS Output buffers (op=%u) are not matched", > + i); > + } > + > + return TEST_SUCCESS; > +} > + > static void > create_reference_dec_op(struct rte_bbdev_dec_op *op) > { > @@ -2622,6 +2756,20 @@ create_reference_fft_op(struct rte_bbdev_fft_op *op) > op->fft.base_input.length += entry->segments[i].length; > } > > +static void > +create_reference_mldts_op(struct rte_bbdev_mldts_op *op) > +{ > + unsigned int i; > + struct op_data_entries *entry; > + op->mldts = test_vector.mldts; > + entry = &test_vector.entries[DATA_INPUT]; > + for (i = 0; i < entry->nb_segments; ++i) > + op->mldts.qhy_input.length += entry->segments[i].length; > + entry = &test_vector.entries[DATA_HARQ_INPUT]; > + for (i = 0; i < entry->nb_segments; ++i) > + op->mldts.r_input.length += entry->segments[i].length; > +} > + > static void > create_reference_enc_op(struct rte_bbdev_enc_op *op) > { > @@ -2730,6 +2878,14 @@ calc_fft_size(struct rte_bbdev_fft_op *op) > return output_size; > } > > +static uint32_t > +calc_mldts_size(struct rte_bbdev_mldts_op *op) > +{ > + uint32_t output_size; > + output_size = op->mldts.num_layers * op->mldts.num_rbs * op->mldts.c_rep; > + return output_size; > +} > + > static int > init_test_op_params(struct test_op_params *op_params, > enum rte_bbdev_op_type op_type, const int expected_status, > @@ -2744,6 +2900,9 @@ init_test_op_params(struct test_op_params *op_params, > else if (op_type == RTE_BBDEV_OP_FFT) > ret = rte_bbdev_fft_op_alloc_bulk(ops_mp, > &op_params->ref_fft_op, 1); > + else if (op_type == RTE_BBDEV_OP_MLDTS) > + ret = rte_bbdev_mldts_op_alloc_bulk(ops_mp, > + &op_params->ref_mldts_op, 1); > else > ret = rte_bbdev_enc_op_alloc_bulk(ops_mp, > &op_params->ref_enc_op, 1); > @@ -2763,6 +2922,8 @@ init_test_op_params(struct test_op_params *op_params, > op_params->ref_enc_op->status = expected_status; > else if (op_type == RTE_BBDEV_OP_FFT) > op_params->ref_fft_op->status = expected_status; > + else if (op_type == RTE_BBDEV_OP_MLDTS) > + op_params->ref_mldts_op->status = expected_status; > return 0; > } > > @@ -2831,6 +2992,8 @@ run_test_case_on_device(test_case_function *test_case_func, uint8_t dev_id, > create_reference_ldpc_dec_op(op_params->ref_dec_op); > else if (test_vector.op_type == RTE_BBDEV_OP_FFT) > create_reference_fft_op(op_params->ref_fft_op); > + else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) > + create_reference_mldts_op(op_params->ref_mldts_op); > > for (i = 0; i < ad->nb_queues; ++i) { > f_ret = fill_queue_buffers(op_params, > @@ -3047,6 +3210,11 @@ dequeue_event_callback(uint16_t dev_id, > &tp->fft_ops[ > __atomic_load_n(&tp->nb_dequeued, __ATOMIC_RELAXED)], > burst_sz); > + else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) > + deq = rte_bbdev_dequeue_mldts_ops(dev_id, queue_id, > + &tp->mldts_ops[ > + __atomic_load_n(&tp->nb_dequeued, __ATOMIC_RELAXED)], > + burst_sz); > else /*RTE_BBDEV_OP_TURBO_ENC*/ > deq = rte_bbdev_dequeue_enc_ops(dev_id, queue_id, > &tp->enc_ops[ > @@ -3093,6 +3261,10 @@ dequeue_event_callback(uint16_t dev_id, > struct rte_bbdev_fft_op *ref_op = tp->op_params->ref_fft_op; > ret = validate_fft_op(tp->fft_ops, num_ops, ref_op); > rte_bbdev_fft_op_free_bulk(tp->fft_ops, deq); > + } else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) { > + struct rte_bbdev_mldts_op *ref_op = tp->op_params->ref_mldts_op; > + ret = validate_mldts_op(tp->mldts_ops, num_ops, ref_op); > + rte_bbdev_mldts_op_free_bulk(tp->mldts_ops, deq); > } else if (test_vector.op_type == RTE_BBDEV_OP_LDPC_DEC) { > struct rte_bbdev_dec_op *ref_op = tp->op_params->ref_dec_op; > ret = validate_ldpc_dec_op(tp->dec_ops, num_ops, ref_op, > @@ -3118,6 +3290,9 @@ dequeue_event_callback(uint16_t dev_id, > case RTE_BBDEV_OP_FFT: > tb_len_bits = calc_fft_size(tp->op_params->ref_fft_op); > break; > + case RTE_BBDEV_OP_MLDTS: > + tb_len_bits = calc_mldts_size(tp->op_params->ref_mldts_op); > + break; > case RTE_BBDEV_OP_LDPC_ENC: > tb_len_bits = calc_ldpc_enc_TB_size(tp->op_params->ref_enc_op); > break; > @@ -3593,6 +3768,88 @@ throughput_intr_lcore_fft(void *arg) > return TEST_SUCCESS; > } > > +static int > +throughput_intr_lcore_mldts(void *arg) > +{ > + struct thread_params *tp = arg; > + unsigned int enqueued; > + const uint16_t queue_id = tp->queue_id; > + const uint16_t burst_sz = tp->op_params->burst_sz; > + const uint16_t num_to_process = tp->op_params->num_to_process; > + struct rte_bbdev_mldts_op *ops[num_to_process]; > + struct test_buffers *bufs = NULL; > + struct rte_bbdev_info info; > + int ret, i, j; > + uint16_t num_to_enq, enq; > + > + TEST_ASSERT_SUCCESS((burst_sz > MAX_BURST), "BURST_SIZE should be <= %u", MAX_BURST); > + > + TEST_ASSERT_SUCCESS(rte_bbdev_queue_intr_enable(tp->dev_id, queue_id), > + "Failed to enable interrupts for dev: %u, queue_id: %u", > + tp->dev_id, queue_id); > + > + rte_bbdev_info_get(tp->dev_id, &info); > + > + TEST_ASSERT_SUCCESS((num_to_process > info.drv.queue_size_lim), > + "NUM_OPS cannot exceed %u for this device", > + info.drv.queue_size_lim); > + > + bufs = &tp->op_params->q_bufs[GET_SOCKET(info.socket_id)][queue_id]; > + > + __atomic_store_n(&tp->processing_status, 0, __ATOMIC_RELAXED); > + __atomic_store_n(&tp->nb_dequeued, 0, __ATOMIC_RELAXED); > + > + rte_wait_until_equal_16(&tp->op_params->sync, SYNC_START, __ATOMIC_RELAXED); > + > + ret = rte_bbdev_mldts_op_alloc_bulk(tp->op_params->mp, ops, num_to_process); > + TEST_ASSERT_SUCCESS(ret, "Allocation failed for %d ops", num_to_process); > + if (test_vector.op_type != RTE_BBDEV_OP_NONE) > + copy_reference_mldts_op(ops, num_to_process, 0, bufs->inputs, bufs->harq_inputs, > + bufs->hard_outputs, tp->op_params->ref_mldts_op); > + > + /* Set counter to validate the ordering */ > + for (j = 0; j < num_to_process; ++j) > + ops[j]->opaque_data = (void *)(uintptr_t)j; > + > + for (j = 0; j < TEST_REPETITIONS; ++j) { > + for (i = 0; i < num_to_process; ++i) > + mbuf_reset(ops[i]->mldts.output.data); > + > + tp->start_time = rte_rdtsc_precise(); > + for (enqueued = 0; enqueued < num_to_process;) { > + num_to_enq = burst_sz; > + > + if (unlikely(num_to_process - enqueued < num_to_enq)) > + num_to_enq = num_to_process - enqueued; > + > + enq = 0; > + do { > + enq += rte_bbdev_enqueue_mldts_ops(tp->dev_id, > + queue_id, &ops[enqueued], num_to_enq); > + } while (unlikely(enq != num_to_enq)); > + enqueued += enq; > + > + /* Write to thread burst_sz current number of enqueued > + * descriptors. It ensures that proper number of > + * descriptors will be dequeued in callback > + * function - needed for last batch in case where > + * the number of operations is not a multiple of > + * burst size. > + */ > + __atomic_store_n(&tp->burst_sz, num_to_enq, __ATOMIC_RELAXED); > + > + /* Wait until processing of previous batch is > + * completed > + */ > + rte_wait_until_equal_16(&tp->nb_dequeued, enqueued, __ATOMIC_RELAXED); > + } > + if (j != TEST_REPETITIONS - 1) > + __atomic_store_n(&tp->nb_dequeued, 0, __ATOMIC_RELAXED); > + } > + > + return TEST_SUCCESS; > +} > + > static int > throughput_pmd_lcore_dec(void *arg) > { > @@ -4403,6 +4660,104 @@ throughput_pmd_lcore_fft(void *arg) > return TEST_SUCCESS; > } > > +static int > +throughput_pmd_lcore_mldts(void *arg) > +{ > + struct thread_params *tp = arg; > + uint16_t enq, deq; > + uint64_t total_time = 0, start_time; > + const uint16_t queue_id = tp->queue_id; > + const uint16_t burst_sz = tp->op_params->burst_sz; > + const uint16_t num_ops = tp->op_params->num_to_process; > + struct rte_bbdev_mldts_op *ops_enq[num_ops]; > + struct rte_bbdev_mldts_op *ops_deq[num_ops]; > + struct rte_bbdev_mldts_op *ref_op = tp->op_params->ref_mldts_op; > + struct test_buffers *bufs = NULL; > + int i, j, ret; > + struct rte_bbdev_info info; > + uint16_t num_to_enq; > + > + TEST_ASSERT_SUCCESS((burst_sz > MAX_BURST), "BURST_SIZE should be <= %u", MAX_BURST); > + > + rte_bbdev_info_get(tp->dev_id, &info); > + > + TEST_ASSERT_SUCCESS((num_ops > info.drv.queue_size_lim), > + "NUM_OPS cannot exceed %u for this device", > + info.drv.queue_size_lim); > + > + bufs = &tp->op_params->q_bufs[GET_SOCKET(info.socket_id)][queue_id]; > + > + rte_wait_until_equal_16(&tp->op_params->sync, SYNC_START, __ATOMIC_RELAXED); > + > + ret = rte_bbdev_mldts_op_alloc_bulk(tp->op_params->mp, ops_enq, num_ops); > + TEST_ASSERT_SUCCESS(ret, "Allocation failed for %d ops", num_ops); > + > + if (test_vector.op_type != RTE_BBDEV_OP_NONE) > + copy_reference_mldts_op(ops_enq, num_ops, 0, bufs->inputs, bufs->harq_inputs, > + bufs->hard_outputs, ref_op); > + > + /* Set counter to validate the ordering */ > + for (j = 0; j < num_ops; ++j) > + ops_enq[j]->opaque_data = (void *)(uintptr_t)j; > + > + for (i = 0; i < TEST_REPETITIONS; ++i) { > + uint32_t time_out = 0; > + for (j = 0; j < num_ops; ++j) > + mbuf_reset(ops_enq[j]->mldts.output.data); > + > + start_time = rte_rdtsc_precise(); > + > + for (enq = 0, deq = 0; enq < num_ops;) { > + num_to_enq = burst_sz; > + > + if (unlikely(num_ops - enq < num_to_enq)) > + num_to_enq = num_ops - enq; > + > + enq += rte_bbdev_enqueue_mldts_ops(tp->dev_id, > + queue_id, &ops_enq[enq], num_to_enq); > + > + deq += rte_bbdev_dequeue_mldts_ops(tp->dev_id, > + queue_id, &ops_deq[deq], enq - deq); > + time_out++; > + if (time_out >= TIME_OUT_POLL) { > + timeout_exit(tp->dev_id); > + TEST_ASSERT_SUCCESS(TEST_FAILED, "Enqueue timeout!"); > + } > + } > + > + /* dequeue the remaining */ > + time_out = 0; > + while (deq < enq) { > + deq += rte_bbdev_dequeue_mldts_ops(tp->dev_id, > + queue_id, &ops_deq[deq], enq - deq); > + time_out++; > + if (time_out >= TIME_OUT_POLL) { > + timeout_exit(tp->dev_id); > + TEST_ASSERT_SUCCESS(TEST_FAILED, "Dequeue timeout!"); > + } > + } > + > + total_time += rte_rdtsc_precise() - start_time; > + } > + > + if (test_vector.op_type != RTE_BBDEV_OP_NONE) { > + ret = validate_mldts_op(ops_deq, num_ops, ref_op); > + TEST_ASSERT_SUCCESS(ret, "Validation failed!"); > + } > + > + rte_bbdev_mldts_op_free_bulk(ops_enq, num_ops); > + > + double tb_len_bits = calc_mldts_size(ref_op); > + > + tp->ops_per_sec = ((double)num_ops * TEST_REPETITIONS) / > + ((double)total_time / (double)rte_get_tsc_hz()); > + tp->mbps = (((double)(num_ops * TEST_REPETITIONS * tb_len_bits)) / > + 1000000.0) / ((double)total_time / > + (double)rte_get_tsc_hz()); > + > + return TEST_SUCCESS; > +} > + > static void > print_enc_throughput(struct thread_params *t_params, unsigned int used_cores) > { > @@ -4624,6 +4979,8 @@ throughput_test(struct active_device *ad, > throughput_function = throughput_intr_lcore_ldpc_enc; > else if (test_vector.op_type == RTE_BBDEV_OP_FFT) > throughput_function = throughput_intr_lcore_fft; > + else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) > + throughput_function = throughput_intr_lcore_mldts; > else > throughput_function = throughput_intr_lcore_enc; > > @@ -4646,6 +5003,8 @@ throughput_test(struct active_device *ad, > throughput_function = throughput_pmd_lcore_ldpc_enc; > else if (test_vector.op_type == RTE_BBDEV_OP_FFT) > throughput_function = throughput_pmd_lcore_fft; > + else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) > + throughput_function = throughput_pmd_lcore_mldts; > else > throughput_function = throughput_pmd_lcore_enc; > } > @@ -5139,6 +5498,77 @@ latency_test_fft(struct rte_mempool *mempool, > return i; > } > > +static int > +latency_test_mldts(struct rte_mempool *mempool, > + struct test_buffers *bufs, struct rte_bbdev_mldts_op *ref_op, > + uint16_t dev_id, uint16_t queue_id, > + const uint16_t num_to_process, uint16_t burst_sz, > + uint64_t *total_time, uint64_t *min_time, uint64_t *max_time) > +{ > + int ret = TEST_SUCCESS; > + uint16_t i, j, dequeued; > + struct rte_bbdev_mldts_op *ops_enq[MAX_BURST], *ops_deq[MAX_BURST]; > + uint64_t start_time = 0, last_time = 0; > + > + for (i = 0, dequeued = 0; dequeued < num_to_process; ++i) { > + uint16_t enq = 0, deq = 0; > + uint32_t time_out = 0; > + bool first_time = true; > + last_time = 0; > + > + if (unlikely(num_to_process - dequeued < burst_sz)) > + burst_sz = num_to_process - dequeued; > + > + ret = rte_bbdev_mldts_op_alloc_bulk(mempool, ops_enq, burst_sz); > + TEST_ASSERT_SUCCESS(ret, "rte_bbdev_mldts_op_alloc_bulk() failed"); > + if (test_vector.op_type != RTE_BBDEV_OP_NONE) > + copy_reference_mldts_op(ops_enq, burst_sz, dequeued, > + bufs->inputs, bufs->harq_inputs, > + bufs->hard_outputs, > + ref_op); > + > + /* Set counter to validate the ordering */ > + for (j = 0; j < burst_sz; ++j) > + ops_enq[j]->opaque_data = (void *)(uintptr_t)j; > + > + start_time = rte_rdtsc_precise(); > + > + enq = rte_bbdev_enqueue_mldts_ops(dev_id, queue_id, &ops_enq[enq], burst_sz); > + TEST_ASSERT(enq == burst_sz, > + "Error enqueueing burst, expected %u, got %u", > + burst_sz, enq); > + > + /* Dequeue */ > + do { > + deq += rte_bbdev_dequeue_mldts_ops(dev_id, queue_id, > + &ops_deq[deq], burst_sz - deq); > + if (likely(first_time && (deq > 0))) { > + last_time += rte_rdtsc_precise() - start_time; > + first_time = false; > + } > + time_out++; > + if (time_out >= TIME_OUT_POLL) { > + timeout_exit(dev_id); > + TEST_ASSERT_SUCCESS(TEST_FAILED, "Dequeue timeout!"); > + } > + } while (unlikely(burst_sz != deq)); > + > + *max_time = RTE_MAX(*max_time, last_time); > + *min_time = RTE_MIN(*min_time, last_time); > + *total_time += last_time; > + > + if (test_vector.op_type != RTE_BBDEV_OP_NONE) { > + ret = validate_mldts_op(ops_deq, burst_sz, ref_op); > + TEST_ASSERT_SUCCESS(ret, "Validation failed!"); > + } > + > + rte_bbdev_mldts_op_free_bulk(ops_enq, deq); > + dequeued += deq; > + } > + > + return i; > +} > + > /* Common function for running validation and latency test cases */ > static int > validation_latency_test(struct active_device *ad, > @@ -5196,6 +5626,12 @@ validation_latency_test(struct active_device *ad, > ad->dev_id, queue_id, > num_to_process, burst_sz, &total_time, > &min_time, &max_time); > + else if (op_type == RTE_BBDEV_OP_MLDTS) > + iter = latency_test_mldts(op_params->mp, bufs, > + op_params->ref_mldts_op, > + ad->dev_id, queue_id, > + num_to_process, burst_sz, &total_time, > + &min_time, &max_time); > else /* RTE_BBDEV_OP_TURBO_ENC */ > iter = latency_test_enc(op_params->mp, bufs, > op_params->ref_enc_op, > @@ -5337,6 +5773,85 @@ offload_latency_test_fft(struct rte_mempool *mempool, struct test_buffers *bufs, > return i; > } > > +static int > +offload_latency_test_mldts(struct rte_mempool *mempool, struct test_buffers *bufs, > + struct rte_bbdev_mldts_op *ref_op, uint16_t dev_id, > + uint16_t queue_id, const uint16_t num_to_process, > + uint16_t burst_sz, struct test_time_stats *time_st) > +{ > + int i, dequeued, ret; > + struct rte_bbdev_mldts_op *ops_enq[MAX_BURST], *ops_deq[MAX_BURST]; > + uint64_t enq_start_time, deq_start_time; > + uint64_t enq_sw_last_time, deq_last_time; > + struct rte_bbdev_stats stats; > + > + for (i = 0, dequeued = 0; dequeued < num_to_process; ++i) { > + uint16_t enq = 0, deq = 0; > + > + if (unlikely(num_to_process - dequeued < burst_sz)) > + burst_sz = num_to_process - dequeued; > + > + ret = rte_bbdev_mldts_op_alloc_bulk(mempool, ops_enq, burst_sz); > + TEST_ASSERT_SUCCESS(ret, "rte_bbdev_mldts_op_alloc_bulk() failed"); > + if (test_vector.op_type != RTE_BBDEV_OP_NONE) > + copy_reference_mldts_op(ops_enq, burst_sz, dequeued, > + bufs->inputs, bufs->harq_inputs, > + bufs->hard_outputs, > + ref_op); > + > + /* Start time meas for enqueue function offload latency */ > + enq_start_time = rte_rdtsc_precise(); > + do { > + enq += rte_bbdev_enqueue_mldts_ops(dev_id, queue_id, > + &ops_enq[enq], burst_sz - enq); > + } while (unlikely(burst_sz != enq)); > + > + ret = get_bbdev_queue_stats(dev_id, queue_id, &stats); > + TEST_ASSERT_SUCCESS(ret, > + "Failed to get stats for queue (%u) of device (%u)", > + queue_id, dev_id); > + > + enq_sw_last_time = rte_rdtsc_precise() - enq_start_time - > + stats.acc_offload_cycles; > + time_st->enq_sw_max_time = RTE_MAX(time_st->enq_sw_max_time, > + enq_sw_last_time); > + time_st->enq_sw_min_time = RTE_MIN(time_st->enq_sw_min_time, > + enq_sw_last_time); > + time_st->enq_sw_total_time += enq_sw_last_time; > + > + time_st->enq_acc_max_time = RTE_MAX(time_st->enq_acc_max_time, > + stats.acc_offload_cycles); > + time_st->enq_acc_min_time = RTE_MIN(time_st->enq_acc_min_time, > + stats.acc_offload_cycles); > + time_st->enq_acc_total_time += stats.acc_offload_cycles; > + > + /* give time for device to process ops */ > + rte_delay_us(WAIT_OFFLOAD_US); > + > + /* Start time meas for dequeue function offload latency */ > + deq_start_time = rte_rdtsc_precise(); > + /* Dequeue one operation */ > + do { > + deq += rte_bbdev_dequeue_mldts_ops(dev_id, queue_id, &ops_deq[deq], enq); > + } while (unlikely(deq == 0)); > + > + deq_last_time = rte_rdtsc_precise() - deq_start_time; > + time_st->deq_max_time = RTE_MAX(time_st->deq_max_time, deq_last_time); > + time_st->deq_min_time = RTE_MIN(time_st->deq_min_time, deq_last_time); > + time_st->deq_total_time += deq_last_time; > + > + /* Dequeue remaining operations if needed*/ > + while (burst_sz != deq) > + deq += rte_bbdev_dequeue_mldts_ops(dev_id, queue_id, > + &ops_deq[deq], burst_sz - deq); > + > + rte_bbdev_mldts_op_free_bulk(ops_enq, deq); > + dequeued += deq; > + } > + > + return i; > +} > + > static int > offload_latency_test_dec(struct rte_mempool *mempool, struct test_buffers *bufs, > struct rte_bbdev_dec_op *ref_op, uint16_t dev_id, > @@ -5734,6 +6249,10 @@ offload_cost_test(struct active_device *ad, > iter = offload_latency_test_fft(op_params->mp, bufs, > op_params->ref_fft_op, ad->dev_id, queue_id, > num_to_process, burst_sz, &time_st); > + else if (op_type == RTE_BBDEV_OP_MLDTS) > + iter = offload_latency_test_mldts(op_params->mp, bufs, > + op_params->ref_mldts_op, ad->dev_id, queue_id, > + num_to_process, burst_sz, &time_st); > else > iter = offload_latency_test_enc(op_params->mp, bufs, > op_params->ref_enc_op, ad->dev_id, queue_id, > diff --git a/app/test-bbdev/test_bbdev_vector.c b/app/test-bbdev/test_bbdev_vector.c > index 0ef1481f2a..8f464db838 100644 > --- a/app/test-bbdev/test_bbdev_vector.c > +++ b/app/test-bbdev/test_bbdev_vector.c > @@ -244,6 +244,20 @@ op_fft_flag_strtoul(char *token, uint32_t *op_flag_value) > return 0; > } > > +/* convert MLD flag from string to unsigned long int*/ > +static int > +op_mld_flag_strtoul(char *token, uint32_t *op_flag_value) > +{ > + if (!strcmp(token, "RTE_BBDEV_MLDTS_REP")) > + *op_flag_value = RTE_BBDEV_MLDTS_REP; > + else { > + printf("The given value is not a MLD flag\n"); > + return -1; > + } > + > + return 0; > +} > + > /* convert turbo encoder flag from string to unsigned long int*/ > static int > op_encoder_flag_strtoul(char *token, uint32_t *op_flag_value) > @@ -326,6 +340,10 @@ parse_turbo_flags(char *tokens, uint32_t *op_flags, > if (op_fft_flag_strtoul(tok, &op_flag_value) > == -1) > return -1; > + } else if (op_type == RTE_BBDEV_OP_MLDTS) { > + if (op_mld_flag_strtoul(tok, &op_flag_value) > + == -1) > + return -1; > } else { > return -1; > } > @@ -355,6 +373,8 @@ op_turbo_type_strtol(char *token, enum rte_bbdev_op_type *op_type) > *op_type = RTE_BBDEV_OP_LDPC_DEC; > else if (!strcmp(token, "RTE_BBDEV_OP_FFT")) > *op_type = RTE_BBDEV_OP_FFT; > + else if (!strcmp(token, "RTE_BBDEV_OP_MLDTS")) > + *op_type = RTE_BBDEV_OP_MLDTS; > else if (!strcmp(token, "RTE_BBDEV_OP_NONE")) > *op_type = RTE_BBDEV_OP_NONE; > else { > @@ -992,6 +1012,73 @@ parse_fft_params(const char *key_token, char *token, > return 0; > } > > +/* parses MLD parameters and assigns to global variable */ > +static int > +parse_mld_params(const char *key_token, char *token, > + struct test_bbdev_vector *vector) > +{ > + int ret = 0, status = 0; > + uint32_t op_flags = 0; > + char *err = NULL; > + > + struct rte_bbdev_op_mldts *mld = &vector->mldts; > + > + if (starts_with(key_token, "qhy_input")) { > + ret = parse_data_entry(key_token, token, vector, > + DATA_INPUT, "qhy_input"); > + } else if (starts_with(key_token, "r_input")) { > + ret = parse_data_entry(key_token, token, vector, > + DATA_HARQ_INPUT, "r_input"); > + } else if (starts_with(key_token, "output")) { > + ret = parse_data_entry(key_token, token, vector, > + DATA_HARD_OUTPUT, "output"); > + } else if (!strcmp(key_token, "layers")) { > + mld->num_layers = (uint32_t) strtoul(token, &err, 0); > + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; > + } else if (!strcmp(key_token, "layer1")) { > + mld->q_m[0] = (uint32_t) strtoul(token, &err, 0); > + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; > + } else if (!strcmp(key_token, "layer2")) { > + mld->q_m[1] = (uint32_t) strtoul(token, &err, 0); > + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; > + } else if (!strcmp(key_token, "layer3")) { > + mld->q_m[2] = (uint32_t) strtoul(token, &err, 0); > + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; > + } else if (!strcmp(key_token, "layer4")) { > + mld->q_m[3] = (uint32_t) strtoul(token, &err, 0); > + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; > + } else if (!strcmp(key_token, "crep")) { > + mld->c_rep = (uint32_t) strtoul(token, &err, 0); > + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; > + } else if (!strcmp(key_token, "rrep")) { > + mld->r_rep = (uint32_t) strtoul(token, &err, 0); > + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; > + } else if (!strcmp(key_token, "rbs")) { > + mld->num_rbs = (uint32_t) strtoul(token, &err, 0); > + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; > + } else if (!strcmp(key_token, "op_flags")) { > + vector->mask |= TEST_BBDEV_VF_OP_FLAGS; > + ret = parse_turbo_flags(token, &op_flags, vector->op_type); > + if (!ret) > + mld->op_flags = op_flags; > + } else if (!strcmp(key_token, "expected_status")) { > + vector->mask |= TEST_BBDEV_VF_EXPECTED_STATUS; > + ret = parse_expected_status(token, &status, vector->op_type); > + if (!ret) > + vector->expected_status = status; > + } else { > + printf("Not valid mld key: '%s'\n", key_token); > + return -1; > + } > + > + if (ret != 0) { > + printf("Failed with convert '%s\t%s'\n", key_token, token); > + return -1; > + } > + > + return 0; > +} > + > /* checks the type of key and assigns data */ > static int > parse_entry(char *entry, struct test_bbdev_vector *vector) > @@ -1046,6 +1133,9 @@ parse_entry(char *entry, struct test_bbdev_vector *vector) > } else if (vector->op_type == RTE_BBDEV_OP_FFT) { > if (parse_fft_params(key_token, token, vector) == -1) > return -1; > + } else if (vector->op_type == RTE_BBDEV_OP_MLDTS) { > + if (parse_mld_params(key_token, token, vector) == -1) > + return -1; > } > > return 0; > @@ -1132,6 +1222,25 @@ check_fft_segments(struct test_bbdev_vector *vector) > return 0; > } > > +static int > +check_mld_segments(struct test_bbdev_vector *vector) > +{ > + unsigned char i; > + > + for (i = 0; i < vector->entries[DATA_INPUT].nb_segments; i++) > + if (vector->entries[DATA_INPUT].segments[i].addr == NULL) > + return -1; > + > + for (i = 0; i < vector->entries[DATA_HARQ_INPUT].nb_segments; i++) > + if (vector->entries[DATA_HARQ_INPUT].segments[i].addr == NULL) > + return -1; > + > + for (i = 0; i < vector->entries[DATA_HARD_OUTPUT].nb_segments; i++) > + if (vector->entries[DATA_HARD_OUTPUT].segments[i].addr == NULL) > + return -1; > + return 0; > +} > + > static int > check_decoder_llr_spec(struct test_bbdev_vector *vector) > { > @@ -1359,6 +1468,26 @@ check_fft(struct test_bbdev_vector *vector) > return 0; > } > > +/* checks mld parameters */ > +static int > +check_mld(struct test_bbdev_vector *vector) > +{ > + const int mask = vector->mask; > + > + if (check_mld_segments(vector) < 0) > + return -1; > + > + /* Check which params were set */ > + if (!(mask & TEST_BBDEV_VF_OP_FLAGS)) { > + printf( > + "WARNING: op_flags was not specified in vector file and capabilities will not be validated\n"); > + } > + if (!(mask & TEST_BBDEV_VF_EXPECTED_STATUS)) > + printf( > + "WARNING: expected_status was not specified in vector file and will be set to 0\n"); > + return 0; > +} > + > /* checks encoder parameters */ > static int > check_encoder(struct test_bbdev_vector *vector) > @@ -1520,6 +1649,9 @@ bbdev_check_vector(struct test_bbdev_vector *vector) > } else if (vector->op_type == RTE_BBDEV_OP_FFT) { > if (check_fft(vector) == -1) > return -1; > + } else if (vector->op_type == RTE_BBDEV_OP_MLDTS) { > + if (check_mld(vector) == -1) > + return -1; > } else if (vector->op_type != RTE_BBDEV_OP_NONE) { > printf("Vector was not filled\n"); > return -1; > diff --git a/app/test-bbdev/test_bbdev_vector.h b/app/test-bbdev/test_bbdev_vector.h > index 2ea271ffb7..14b8ef2764 100644 > --- a/app/test-bbdev/test_bbdev_vector.h > +++ b/app/test-bbdev/test_bbdev_vector.h > @@ -65,6 +65,7 @@ struct test_bbdev_vector { > struct rte_bbdev_op_ldpc_dec ldpc_dec; > struct rte_bbdev_op_ldpc_enc ldpc_enc; > struct rte_bbdev_op_fft fft; > + struct rte_bbdev_op_mldts mldts; > }; > /* Additional storage for op data entries */ > struct op_data_entries entries[DATA_NUM_TYPES]; ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH v3 08/10] test/bbdev: support new FFT capabilities 2023-11-03 23:34 [PATCH v3 00/10] test-bbdev changes for 23.11 Nicolas Chautru ` (6 preceding siblings ...) 2023-11-03 23:34 ` [PATCH v3 07/10] test/bbdev: add MLD support Nicolas Chautru @ 2023-11-03 23:34 ` Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 09/10] test/bbdev: support 4 bit LLR compression Nicolas Chautru ` (2 subsequent siblings) 10 siblings, 0 replies; 18+ messages in thread From: Nicolas Chautru @ 2023-11-03 23:34 UTC (permalink / raw) To: dev, maxime.coquelin Cc: hemant.agrawal, david.marchand, hernan.vargas, stable From: Hernan Vargas <hernan.vargas@intel.com> Adding support to test new FFT capabilities. Optional frequency domain dewindowing, frequency resampling, timing error correction and time offset per CS. Signed-off-by: Hernan Vargas <hernan.vargas@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> --- app/test-bbdev/test_bbdev_perf.c | 26 ++++++++++--- app/test-bbdev/test_bbdev_vector.c | 61 ++++++++++++++++++++++++++++-- 2 files changed, 78 insertions(+), 9 deletions(-) diff --git a/app/test-bbdev/test_bbdev_perf.c b/app/test-bbdev/test_bbdev_perf.c index d2e3542356..84596e28da 100644 --- a/app/test-bbdev/test_bbdev_perf.c +++ b/app/test-bbdev/test_bbdev_perf.c @@ -864,6 +864,7 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info, conf.q_fft.num_qgroups = VRB_QMGR_NUM_QGS; conf.q_fft.first_qgroup_index = VRB_QMGR_INVALID_IDX; conf.q_fft.num_aqs_per_groups = VRB_QMGR_NUM_AQS; + conf.q_fft.aq_depth_log2 = VRB_QMGR_AQ_DEPTH; conf.q_mld.num_qgroups = VRB_QMGR_NUM_QGS; conf.q_mld.first_qgroup_index = VRB_QMGR_INVALID_IDX; conf.q_mld.num_aqs_per_groups = VRB_QMGR_NUM_AQS; @@ -1970,7 +1971,7 @@ static void copy_reference_fft_op(struct rte_bbdev_fft_op **ops, unsigned int n, unsigned int start_idx, struct rte_bbdev_op_data *inputs, struct rte_bbdev_op_data *outputs, struct rte_bbdev_op_data *pwrouts, - struct rte_bbdev_fft_op *ref_op) + struct rte_bbdev_op_data *win_inputs, struct rte_bbdev_fft_op *ref_op) { unsigned int i, j; struct rte_bbdev_op_fft *fft = &ref_op->fft; @@ -1982,6 +1983,11 @@ copy_reference_fft_op(struct rte_bbdev_fft_op **ops, unsigned int n, fft->output_leading_depadding; for (j = 0; j < RTE_BBDEV_MAX_CS_2; j++) ops[i]->fft.window_index[j] = fft->window_index[j]; + for (j = 0; j < RTE_BBDEV_MAX_CS; j++) { + ops[i]->fft.cs_theta_0[j] = fft->cs_theta_0[j]; + ops[i]->fft.cs_theta_d[j] = fft->cs_theta_d[j]; + ops[i]->fft.time_offset[j] = fft->time_offset[j]; + } ops[i]->fft.cs_bitmap = fft->cs_bitmap; ops[i]->fft.num_antennas_log2 = fft->num_antennas_log2; ops[i]->fft.idft_log2 = fft->idft_log2; @@ -1992,8 +1998,12 @@ copy_reference_fft_op(struct rte_bbdev_fft_op **ops, unsigned int n, ops[i]->fft.ncs_reciprocal = fft->ncs_reciprocal; ops[i]->fft.power_shift = fft->power_shift; ops[i]->fft.fp16_exp_adjust = fft->fp16_exp_adjust; + ops[i]->fft.output_depadded_size = fft->output_depadded_size; + ops[i]->fft.freq_resample_mode = fft->freq_resample_mode; ops[i]->fft.base_output = outputs[start_idx + i]; ops[i]->fft.base_input = inputs[start_idx + i]; + if (win_inputs != NULL) + ops[i]->fft.dewindowing_input = win_inputs[start_idx + i]; if (pwrouts != NULL) ops[i]->fft.power_meas_output = pwrouts[start_idx + i]; ops[i]->fft.op_flags = fft->op_flags; @@ -2575,7 +2585,7 @@ validate_op_fft_chain(struct rte_bbdev_op_data *op, struct op_data_entries *orig { struct rte_mbuf *m = op->data; uint8_t i, nb_dst_segments = orig_op->nb_segments; - int16_t delt, abs_delt, thres_hold = 3; + int16_t delt, abs_delt, thres_hold = 4; uint32_t j, data_len_iq, error_num; int16_t *ref_out, *op_out; @@ -2754,6 +2764,9 @@ create_reference_fft_op(struct rte_bbdev_fft_op *op) entry = &test_vector.entries[DATA_INPUT]; for (i = 0; i < entry->nb_segments; ++i) op->fft.base_input.length += entry->segments[i].length; + entry = &test_vector.entries[DATA_HARQ_INPUT]; + for (i = 0; i < entry->nb_segments; ++i) + op->fft.dewindowing_input.length += entry->segments[i].length; } static void @@ -3722,7 +3735,8 @@ throughput_intr_lcore_fft(void *arg) num_to_process); if (test_vector.op_type != RTE_BBDEV_OP_NONE) copy_reference_fft_op(ops, num_to_process, 0, bufs->inputs, - bufs->hard_outputs, bufs->soft_outputs, tp->op_params->ref_fft_op); + bufs->hard_outputs, bufs->soft_outputs, bufs->harq_inputs, + tp->op_params->ref_fft_op); /* Set counter to validate the ordering */ for (j = 0; j < num_to_process; ++j) @@ -4596,7 +4610,7 @@ throughput_pmd_lcore_fft(void *arg) if (test_vector.op_type != RTE_BBDEV_OP_NONE) copy_reference_fft_op(ops_enq, num_ops, 0, bufs->inputs, - bufs->hard_outputs, bufs->soft_outputs, ref_op); + bufs->hard_outputs, bufs->soft_outputs, bufs->harq_inputs, ref_op); /* Set counter to validate the ordering */ for (j = 0; j < num_ops; ++j) @@ -5452,7 +5466,7 @@ latency_test_fft(struct rte_mempool *mempool, if (test_vector.op_type != RTE_BBDEV_OP_NONE) copy_reference_fft_op(ops_enq, burst_sz, dequeued, bufs->inputs, - bufs->hard_outputs, bufs->soft_outputs, + bufs->hard_outputs, bufs->soft_outputs, bufs->harq_inputs, ref_op); /* Set counter to validate the ordering */ @@ -5714,7 +5728,7 @@ offload_latency_test_fft(struct rte_mempool *mempool, struct test_buffers *bufs, if (test_vector.op_type != RTE_BBDEV_OP_NONE) copy_reference_fft_op(ops_enq, burst_sz, dequeued, bufs->inputs, - bufs->hard_outputs, bufs->soft_outputs, + bufs->hard_outputs, bufs->soft_outputs, bufs->harq_inputs, ref_op); /* Start time meas for enqueue function offload latency */ diff --git a/app/test-bbdev/test_bbdev_vector.c b/app/test-bbdev/test_bbdev_vector.c index 8f464db838..56b882533c 100644 --- a/app/test-bbdev/test_bbdev_vector.c +++ b/app/test-bbdev/test_bbdev_vector.c @@ -215,7 +215,6 @@ op_ldpc_decoder_flag_strtoul(char *token, uint32_t *op_flag_value) return 0; } - /* Convert FFT flag from string to unsigned long int. */ static int op_fft_flag_strtoul(char *token, uint32_t *op_flag_value) @@ -236,6 +235,14 @@ op_fft_flag_strtoul(char *token, uint32_t *op_flag_value) *op_flag_value = RTE_BBDEV_FFT_FP16_INPUT; else if (!strcmp(token, "RTE_BBDEV_FFT_FP16_OUTPUT")) *op_flag_value = RTE_BBDEV_FFT_FP16_OUTPUT; + else if (!strcmp(token, "RTE_BBDEV_FFT_TIMING_OFFSET_PER_CS")) + *op_flag_value = RTE_BBDEV_FFT_TIMING_OFFSET_PER_CS; + else if (!strcmp(token, "RTE_BBDEV_FFT_TIMING_ERROR")) + *op_flag_value = RTE_BBDEV_FFT_TIMING_ERROR; + else if (!strcmp(token, "RTE_BBDEV_FFT_DEWINDOWING")) + *op_flag_value = RTE_BBDEV_FFT_DEWINDOWING; + else if (!strcmp(token, "RTE_BBDEV_FFT_FREQ_RESAMPLING")) + *op_flag_value = RTE_BBDEV_FFT_FREQ_RESAMPLING; else { printf("The given value is not a FFT flag\n"); return -1; @@ -907,8 +914,7 @@ parse_ldpc_decoder_params(const char *key_token, char *token, return 0; } - -/* Parse FFT parameters and assigns to global variable. */ +/* Parses FFT parameters and assigns to global variable. */ static int parse_fft_params(const char *key_token, char *token, struct test_bbdev_vector *vector) @@ -923,6 +929,10 @@ parse_fft_params(const char *key_token, char *token, ret = parse_data_entry(key_token, token, vector, DATA_INPUT, op_data_prefixes[DATA_INPUT]); + } else if (starts_with(key_token, "dewin_input")) { + ret = parse_data_entry(key_token, token, vector, + DATA_HARQ_INPUT, + "dewin_input"); } else if (starts_with(key_token, "output")) { ret = parse_data_entry(key_token, token, vector, DATA_HARD_OUTPUT, @@ -989,6 +999,51 @@ parse_fft_params(const char *key_token, char *token, fft->fp16_exp_adjust = (uint32_t) strtoul(token, &err, 0); printf("%d\n", fft->fp16_exp_adjust); ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "freq_resample_mode")) { + fft->freq_resample_mode = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "out_depadded_size")) { + fft->output_depadded_size = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "cs_theta_0")) { + tok = strtok(token, VALUE_DELIMITER); + if (tok == NULL) + return -1; + for (i = 0; i < FFT_WIN_SIZE; i++) { + fft->cs_theta_0[i] = (uint32_t) strtoul(tok, &err, 0); + if (i < (FFT_WIN_SIZE - 1)) { + tok = strtok(NULL, VALUE_DELIMITER); + if (tok == NULL) + return -1; + } + } + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "cs_theta_d")) { + tok = strtok(token, VALUE_DELIMITER); + if (tok == NULL) + return -1; + for (i = 0; i < FFT_WIN_SIZE; i++) { + fft->cs_theta_d[i] = (uint32_t) strtoul(tok, &err, 0); + if (i < (FFT_WIN_SIZE - 1)) { + tok = strtok(NULL, VALUE_DELIMITER); + if (tok == NULL) + return -1; + } + } + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "time_offset")) { + tok = strtok(token, VALUE_DELIMITER); + if (tok == NULL) + return -1; + for (i = 0; i < FFT_WIN_SIZE; i++) { + fft->time_offset[i] = (uint32_t) strtoul(tok, &err, 0); + if (i < (FFT_WIN_SIZE - 1)) { + tok = strtok(NULL, VALUE_DELIMITER); + if (tok == NULL) + return -1; + } + } + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; } else if (!strcmp(key_token, "op_flags")) { vector->mask |= TEST_BBDEV_VF_OP_FLAGS; ret = parse_turbo_flags(token, &op_flags, vector->op_type); -- 2.34.1 ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH v3 09/10] test/bbdev: support 4 bit LLR compression 2023-11-03 23:34 [PATCH v3 00/10] test-bbdev changes for 23.11 Nicolas Chautru ` (7 preceding siblings ...) 2023-11-03 23:34 ` [PATCH v3 08/10] test/bbdev: support new FFT capabilities Nicolas Chautru @ 2023-11-03 23:34 ` Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 10/10] test/bbdev: update python script parameters Nicolas Chautru 2023-11-09 15:51 ` [PATCH v3 00/10] test-bbdev changes for 23.11 Maxime Coquelin 10 siblings, 0 replies; 18+ messages in thread From: Nicolas Chautru @ 2023-11-03 23:34 UTC (permalink / raw) To: dev, maxime.coquelin Cc: hemant.agrawal, david.marchand, hernan.vargas, stable From: Hernan Vargas <hernan.vargas@intel.com> Add support to test LDPC UL operation for new capability. Option to compress HARQ memory to 4 bits per LLR. Signed-off-by: Hernan Vargas <hernan.vargas@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> --- app/test-bbdev/test_bbdev_perf.c | 3 ++- app/test-bbdev/test_bbdev_vector.c | 2 ++ 2 files changed, 4 insertions(+), 1 deletion(-) diff --git a/app/test-bbdev/test_bbdev_perf.c b/app/test-bbdev/test_bbdev_perf.c index 84596e28da..426c24a536 100644 --- a/app/test-bbdev/test_bbdev_perf.c +++ b/app/test-bbdev/test_bbdev_perf.c @@ -2240,7 +2240,8 @@ validate_op_harq_chain(struct rte_bbdev_op_data *op, /* Cannot compare HARQ output data for such cases */ if ((ldpc_llr_decimals > 1) && ((ops_ld->op_flags & RTE_BBDEV_LDPC_LLR_COMPRESSION) - || (ops_ld->op_flags & RTE_BBDEV_LDPC_HARQ_6BIT_COMPRESSION))) + || (ops_ld->op_flags & RTE_BBDEV_LDPC_HARQ_6BIT_COMPRESSION) + || (ops_ld->op_flags & RTE_BBDEV_LDPC_HARQ_4BIT_COMPRESSION))) break; if (!(ldpc_cap_flags & diff --git a/app/test-bbdev/test_bbdev_vector.c b/app/test-bbdev/test_bbdev_vector.c index 56b882533c..42fa630041 100644 --- a/app/test-bbdev/test_bbdev_vector.c +++ b/app/test-bbdev/test_bbdev_vector.c @@ -196,6 +196,8 @@ op_ldpc_decoder_flag_strtoul(char *token, uint32_t *op_flag_value) *op_flag_value = RTE_BBDEV_LDPC_DEC_SCATTER_GATHER; else if (!strcmp(token, "RTE_BBDEV_LDPC_HARQ_6BIT_COMPRESSION")) *op_flag_value = RTE_BBDEV_LDPC_HARQ_6BIT_COMPRESSION; + else if (!strcmp(token, "RTE_BBDEV_LDPC_HARQ_4BIT_COMPRESSION")) + *op_flag_value = RTE_BBDEV_LDPC_HARQ_4BIT_COMPRESSION; else if (!strcmp(token, "RTE_BBDEV_LDPC_LLR_COMPRESSION")) *op_flag_value = RTE_BBDEV_LDPC_LLR_COMPRESSION; else if (!strcmp(token, -- 2.34.1 ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH v3 10/10] test/bbdev: update python script parameters 2023-11-03 23:34 [PATCH v3 00/10] test-bbdev changes for 23.11 Nicolas Chautru ` (8 preceding siblings ...) 2023-11-03 23:34 ` [PATCH v3 09/10] test/bbdev: support 4 bit LLR compression Nicolas Chautru @ 2023-11-03 23:34 ` Nicolas Chautru 2023-11-09 10:16 ` Maxime Coquelin 2023-11-09 15:51 ` [PATCH v3 00/10] test-bbdev changes for 23.11 Maxime Coquelin 10 siblings, 1 reply; 18+ messages in thread From: Nicolas Chautru @ 2023-11-03 23:34 UTC (permalink / raw) To: dev, maxime.coquelin Cc: hemant.agrawal, david.marchand, hernan.vargas, stable, Nicolas Chautru Update the timeout argument and default values. Update EAL help message and default value. Add iter_max and snr arguments. Until next release we keep -t as a possible command-line argument for time-out, until -t becomes the argument for iter-max so that to match the binary command-line argument. Signed-off-by: Nicolas Chautru <nicolas.chautru@intel.com> --- app/test-bbdev/test-bbdev.py | 26 +++++++++++++++++++++++--- app/test-bbdev/test_bbdev_perf.c | 2 +- 2 files changed, 24 insertions(+), 4 deletions(-) diff --git a/app/test-bbdev/test-bbdev.py b/app/test-bbdev/test-bbdev.py index 02c678a360..463de9ce05 100755 --- a/app/test-bbdev/test-bbdev.py +++ b/app/test-bbdev/test-bbdev.py @@ -25,12 +25,18 @@ def kill(process): help="specifies path to the bbdev test app", default=dpdk_path + "/" + dpdk_target + "/app/dpdk-test-bbdev") parser.add_argument("-e", "--eal-params", - help="EAL arguments which are passed to the test app", - default="--vdev=baseband_null0") + help="EAL arguments which must be passed to the test app", + default="--vdev=baseband_null0 -a00:00.0") +# Until deprecated in next release keep -t as an valid argument for timeout, then use -T parser.add_argument("-t", "--timeout", type=int, help="Timeout in seconds", - default=300) + default=600) +# This will become -t option for iter_max in next release +parser.add_argument("-x", "--iter-max", + type=int, + help="Max iterations", + default=6) parser.add_argument("-c", "--test-cases", nargs="+", help="Defines test cases to run. Run all if not specified") @@ -48,6 +54,10 @@ def kill(process): type=int, help="Operations enqueue/dequeue burst size.", default=[32]) +parser.add_argument("-s", "--snr", + type=int, + help="SNR in dB for BLER tests", + default=0) parser.add_argument("-l", "--num-lcores", type=int, help="Number of lcores to run.", @@ -68,6 +78,16 @@ def kill(process): params.extend(["--"]) +if args.snr: + params.extend(["-s", str(args.snr)]) + +if args.iter_max: + params.extend(["-t", str(args.iter_max)]) + print("The argument for iter_max will be -t in next release") + +if args.timeout: + print("The argument for timeout will be -T in next release") + if args.num_ops: params.extend(["-n", str(args.num_ops)]) diff --git a/app/test-bbdev/test_bbdev_perf.c b/app/test-bbdev/test_bbdev_perf.c index 426c24a536..67cd429cef 100644 --- a/app/test-bbdev/test_bbdev_perf.c +++ b/app/test-bbdev/test_bbdev_perf.c @@ -26,7 +26,7 @@ #define MAX_QUEUES RTE_MAX_LCORE #define TEST_REPETITIONS 100 -#define TIME_OUT_POLL 1e8 +#define TIME_OUT_POLL 1e9 #define WAIT_OFFLOAD_US 1000 #ifdef RTE_BASEBAND_FPGA_LTE_FEC -- 2.34.1 ^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [PATCH v3 10/10] test/bbdev: update python script parameters 2023-11-03 23:34 ` [PATCH v3 10/10] test/bbdev: update python script parameters Nicolas Chautru @ 2023-11-09 10:16 ` Maxime Coquelin 2023-11-09 15:37 ` Chautru, Nicolas 0 siblings, 1 reply; 18+ messages in thread From: Maxime Coquelin @ 2023-11-09 10:16 UTC (permalink / raw) To: Nicolas Chautru, dev Cc: hemant.agrawal, david.marchand, hernan.vargas, stable On 11/4/23 00:34, Nicolas Chautru wrote: > +# This will become -t option for iter_max in next release > +parser.add_argument("-x", "--iter-max", > + type=int, > + help="Max iterations", > + default=6) No need for "-x", it is possible to only have the long name: parser.add_argument("--iter-max", type=int, help="Max iterations", default=6) If that works for you, I can do the change while applying. Regards, Maxime ^ permalink raw reply [flat|nested] 18+ messages in thread
* RE: [PATCH v3 10/10] test/bbdev: update python script parameters 2023-11-09 10:16 ` Maxime Coquelin @ 2023-11-09 15:37 ` Chautru, Nicolas 0 siblings, 0 replies; 18+ messages in thread From: Chautru, Nicolas @ 2023-11-09 15:37 UTC (permalink / raw) To: Coquelin, Maxime, dev Cc: hemant.agrawal, Marchand, David, Vargas, Hernan, stable OK with me Maxime, thanks. > -----Original Message----- > From: Maxime Coquelin <maxime.coquelin@redhat.com> > Sent: Thursday, November 9, 2023 2:16 AM > To: Chautru, Nicolas <nicolas.chautru@intel.com>; dev@dpdk.org > Cc: hemant.agrawal@nxp.com; Marchand, David > <david.marchand@redhat.com>; Vargas, Hernan > <hernan.vargas@intel.com>; stable@dpdk.org > Subject: Re: [PATCH v3 10/10] test/bbdev: update python script parameters > > > > On 11/4/23 00:34, Nicolas Chautru wrote: > > +# This will become -t option for iter_max in next release > > +parser.add_argument("-x", "--iter-max", > > + type=int, > > + help="Max iterations", > > + default=6) > > No need for "-x", it is possible to only have the long name: > > parser.add_argument("--iter-max", > type=int, > help="Max iterations", > default=6) > > If that works for you, I can do the change while applying. > > Regards, > Maxime ^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [PATCH v3 00/10] test-bbdev changes for 23.11 2023-11-03 23:34 [PATCH v3 00/10] test-bbdev changes for 23.11 Nicolas Chautru ` (9 preceding siblings ...) 2023-11-03 23:34 ` [PATCH v3 10/10] test/bbdev: update python script parameters Nicolas Chautru @ 2023-11-09 15:51 ` Maxime Coquelin 10 siblings, 0 replies; 18+ messages in thread From: Maxime Coquelin @ 2023-11-09 15:51 UTC (permalink / raw) To: Nicolas Chautru, dev Cc: hemant.agrawal, david.marchand, hernan.vargas, stable On 11/4/23 00:34, Nicolas Chautru wrote: > v3: python script argument fix marked as deprecation first, with > formal fix in following release. > Maxime let me know if you want to something more formal in the > release notes to call it out, currently embedded in script and > commit message. > typo correction. > v2: adding fixes for some of the commits requested by Maxime. > > Update test-bbdev for 23.11. > > Hernan Vargas (8): > test/bbdev: fix python script subprocess > test/bbdev: handle exception for LLR generation > test/bbdev: improve test log messages > test/bbdev: assert failed test for queue configure > test/bbdev: ldpc encoder concatenation vector > test/bbdev: add MLD support > test/bbdev: support new FFT capabilities > test/bbdev: support 4 bit LLR compression > > Nicolas Chautru (2): > test/bbdev: rename macros from acc200 to vrb > test/bbdev: update python script parameters > > app/test-bbdev/main.c | 3 +- > app/test-bbdev/test-bbdev.py | 55 ++- > app/test-bbdev/test_bbdev.c | 3 +- > app/test-bbdev/test_bbdev_perf.c | 673 ++++++++++++++++++++++++++--- > app/test-bbdev/test_bbdev_vector.c | 199 ++++++++- > app/test-bbdev/test_bbdev_vector.h | 1 + > 6 files changed, 843 insertions(+), 91 deletions(-) > Applied to next-baseband/for-main with agreed changes. Thanks, Maxime ^ permalink raw reply [flat|nested] 18+ messages in thread
end of thread, other threads:[~2023-11-09 15:51 UTC | newest] Thread overview: 18+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2023-11-03 23:34 [PATCH v3 00/10] test-bbdev changes for 23.11 Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 01/10] test/bbdev: fix python script subprocess Nicolas Chautru 2023-11-09 12:37 ` Maxime Coquelin 2023-11-03 23:34 ` [PATCH v3 02/10] test/bbdev: rename macros from acc200 to vrb Nicolas Chautru 2023-11-09 12:37 ` Maxime Coquelin 2023-11-03 23:34 ` [PATCH v3 03/10] test/bbdev: handle exception for LLR generation Nicolas Chautru 2023-11-09 12:46 ` Maxime Coquelin 2023-11-03 23:34 ` [PATCH v3 04/10] test/bbdev: improve test log messages Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 05/10] test/bbdev: assert failed test for queue configure Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 06/10] test/bbdev: ldpc encoder concatenation vector Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 07/10] test/bbdev: add MLD support Nicolas Chautru 2023-11-09 15:44 ` Maxime Coquelin 2023-11-03 23:34 ` [PATCH v3 08/10] test/bbdev: support new FFT capabilities Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 09/10] test/bbdev: support 4 bit LLR compression Nicolas Chautru 2023-11-03 23:34 ` [PATCH v3 10/10] test/bbdev: update python script parameters Nicolas Chautru 2023-11-09 10:16 ` Maxime Coquelin 2023-11-09 15:37 ` Chautru, Nicolas 2023-11-09 15:51 ` [PATCH v3 00/10] test-bbdev changes for 23.11 Maxime Coquelin
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).