Re: [PATCH v1 04/13] test/bbdev: early termination not explicit set

[Maxime Coquelin <maxime.coquelin@redhat.com>]

On 2/10/23 18:15, Vargas, Hernan wrote:
> Hi Maxime,
> 
> -----Original Message-----
> From: Maxime Coquelin <maxime.coquelin@redhat.com>
> Sent: Tuesday, January 31, 2023 4:04 AM
> To: Vargas, Hernan <hernan.vargas@intel.com>; dev@dpdk.org; gakhil@marvell.com; Rix, Tom <trix@redhat.com>
> Cc: Chautru, Nicolas <nicolas.chautru@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>
> Subject: Re: [PATCH v1 04/13] test/bbdev: early termination not explicit set
> 
> 
> 
> On 1/17/23 17:50, Hernan Vargas wrote:
>> Early termination needs to be explicitly enabled.
>>
>> Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
>> ---
>>    app/test-bbdev/test_bbdev_perf.c | 11 +++++++++--
>>    1 file changed, 9 insertions(+), 2 deletions(-)
>>
>> diff --git a/app/test-bbdev/test_bbdev_perf.c b/app/test-bbdev/test_bbdev_perf.c
>> index 14fe76eb64..12fa988da6 100644
>> --- a/app/test-bbdev/test_bbdev_perf.c
>> +++ b/app/test-bbdev/test_bbdev_perf.c
>> @@ -4463,7 +4463,7 @@ latency_test_dec(struct rte_mempool *mempool,
>>    		struct test_buffers *bufs, struct rte_bbdev_dec_op *ref_op,
>>    		int vector_mask, 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)
>> +		uint64_t *total_time, uint64_t *min_time, uint64_t *max_time, bool disable_et)
>>    {
>>    	int ret = TEST_SUCCESS;
>>    	uint16_t i, j, dequeued;
>> @@ -4481,6 +4481,13 @@ latency_test_dec(struct rte_mempool *mempool,
>>    		ret = rte_bbdev_dec_op_alloc_bulk(mempool, ops_enq, burst_sz);
>>    		TEST_ASSERT_SUCCESS(ret,
>>    				"rte_bbdev_dec_op_alloc_bulk() failed");
>> +		ref_op->turbo_dec.iter_max = get_iter_max();
>> +		/* For validation tests we want to enable early termination */
>> +		if (!disable_et && !check_bit(ref_op->turbo_dec.op_flags,
>> +				RTE_BBDEV_TURBO_EARLY_TERMINATION))
>> +			ref_op->turbo_dec.op_flags |=
>> +					RTE_BBDEV_TURBO_EARLY_TERMINATION;
>> +
>>    		if (test_vector.op_type != RTE_BBDEV_OP_NONE)
>>    			copy_reference_dec_op(ops_enq, burst_sz, dequeued,
>>    					bufs->inputs,
>> @@ -4873,7 +4880,7 @@ validation_latency_test(struct active_device *ad,
>>    		iter = latency_test_dec(op_params->mp, bufs,
>>    				op_params->ref_dec_op, op_params->vector_mask,
>>    				ad->dev_id, queue_id, num_to_process,
>> -				burst_sz, &total_time, &min_time, &max_time);
>> +				burst_sz, &total_time, &min_time, &max_time, latency_flag);
>>    	else if (op_type == RTE_BBDEV_OP_LDPC_ENC)
>>    		iter = latency_test_ldpc_enc(op_params->mp, bufs,
>>    				op_params->ref_enc_op, ad->dev_id, queue_id,
> 
> Not clear to me whether it should be a fix. Any thoughts?
> 
> Thanks,
> Maxime
> 
> This is not a fix, it's just an enhancement to enable early termination for validation tests if the device supports it.
> The validation tests currently run iter_max number of iterations for decoding, with this enhancement the decoding could terminate early.
> 
> Thanks,
> Hernan

Could you please fix your email client?
The message you are replying to is not quoted.

Thanks,
Maxime