[PATCH v1 1/1] doc: announce change in bbdev api related to operation extension
Chautru, Nicolas
nicolas.chautru at intel.com
Mon Jun 5 22:08:41 CEST 2023
Hi Maxime,
So basically the fft structure change is okay since these are still marked as rte_experimental (it got reported in the ABI report though).
Wrt the MLD functions: these are new into the related serie but still the break the ABI since the struct rte_bbdev includes these functions hence causing offset changes.
Should I then just rephrase as:
+* bbdev: Will extend the API to support the new operation type
+``RTE_BBDEV_OP_MLDTS`` as per
+ this `v1
+<https://patches.dpdk.org/project/dpdk/list/?series=28192>`. This
+ will notably introduce
+ new symbols for ``rte_bbdev_dequeue_mldts_ops``,
+``rte_bbdev_enqueue_mldts_ops`` into the stuct rte_bbdev.
Pasting below the ABI results for reference
[C] 'function rte_bbdev* rte_bbdev_allocate(const char*)' at rte_bbdev.c:174:1 has some indirect sub-type changes:
return type changed:
in pointed to type 'struct rte_bbdev' at rte_bbdev.h:498:1:
type size hasn't changed
2 data member insertions:
'rte_bbdev_enqueue_mldts_ops_t rte_bbdev::enqueue_mldts_ops', at offset 640 (in bits) at rte_bbdev.h:520:1
'rte_bbdev_dequeue_mldts_ops_t rte_bbdev::dequeue_mldts_ops', at offset 704 (in bits) at rte_bbdev.h:522:1
7 data member changes (9 filtered):
type of 'rte_bbdev_dequeue_fft_ops_t rte_bbdev::dequeue_fft_ops' changed:
underlying type 'typedef uint16_t (rte_bbdev_queue_data*, rte_bbdev_fft_op**, typedef uint16_t)*' changed:
in pointed to type 'function type typedef uint16_t (rte_bbdev_queue_data*, rte_bbdev_fft_op**, typedef uint16_t)':
parameter 2 of type 'rte_bbdev_fft_op**' has sub-type changes:
in pointed to type 'rte_bbdev_fft_op*':
in pointed to type 'struct rte_bbdev_fft_op' at rte_bbdev_op.h:978:1:
type size changed from 832 to 1664 (in bits)
1 data member change:
type of 'rte_bbdev_op_fft rte_bbdev_fft_op::fft' changed:
type size changed from 640 to 1472 (in bits)
6 data member insertions:
'rte_bbdev_op_data rte_bbdev_op_fft::dewindowing_input', at offset 256 (in bits) at rte_bbdev_op.h:771:1
'int8_t rte_bbdev_op_fft::freq_resample_mode', at offset 768 (in bits) at rte_bbdev_op.h:807:1
'uint16_t rte_bbdev_op_fft::output_depadded_size', at offset 784 (in bits) at rte_bbdev_op.h:809:1
'uint16_t rte_bbdev_op_fft::cs_theta_0[12]', at offset 800 (in bits) at rte_bbdev_op.h:811:1
'uint32_t rte_bbdev_op_fft::cs_theta_d[12]', at offset 992 (in bits) at rte_bbdev_op.h:813:1
'int8_t rte_bbdev_op_fft::time_offset[12]', at offset 1376 (in bits) at rte_bbdev_op.h:815:1
17 data member changes:
'rte_bbdev_op_data rte_bbdev_op_fft::power_meas_output' offset changed from 256 to 384 (in bits) (by +128 bits)
'uint32_t rte_bbdev_op_fft::op_flags' offset changed from 384 to 512 (in bits) (by +128 bits)
'uint16_t rte_bbdev_op_fft::input_sequence_size' offset changed from 416 to 544 (in bits) (by +128 bits)
'uint16_t rte_bbdev_op_fft::input_leading_padding' offset changed from 432 to 560 (in bits) (by +128 bits)
'uint16_t rte_bbdev_op_fft::output_sequence_size' offset changed from 448 to 576 (in bits) (by +128 bits)
'uint16_t rte_bbdev_op_fft::output_leading_depadding' offset changed from 464 to 592 (in bits) (by +128 bits)
'uint8_t rte_bbdev_op_fft::window_index[6]' offset changed from 480 to 608 (in bits) (by +128 bits)
'uint16_t rte_bbdev_op_fft::cs_bitmap' offset changed from 528 to 656 (in bits) (by +128 bits)
'uint8_t rte_bbdev_op_fft::num_antennas_log2' offset changed from 544 to 672 (in bits) (by +128 bits)
'uint8_t rte_bbdev_op_fft::idft_log2' offset changed from 552 to 680 (in bits) (by +128 bits)
'uint8_t rte_bbdev_op_fft::dft_log2' offset changed from 560 to 688 (in bits) (by +128 bits)
'int8_t rte_bbdev_op_fft::cs_time_adjustment' offset changed from 568 to 696 (in bits) (by +128 bits)
'int8_t rte_bbdev_op_fft::idft_shift' offset changed from 576 to 704 (in bits) (by +128 bits)
'int8_t rte_bbdev_op_fft::dft_shift' offset changed from 584 to 712 (in bits) (by +128 bits)
'uint16_t rte_bbdev_op_fft::ncs_reciprocal' offset changed from 592 to 720 (in bits) (by +128 bits)
'uint16_t rte_bbdev_op_fft::power_shift' offset changed from 608 to 736 (in bits) (by +128 bits)
'uint16_t rte_bbdev_op_fft::fp16_exp_adjust' offset changed from 624 to 752 (in bits) (by +128 bits)
'const rte_bbdev_ops* rte_bbdev::dev_ops' offset changed from 640 to 768 (in bits) (by +128 bits)
'rte_bbdev_data* rte_bbdev::data' offset changed from 704 to 832 (in bits) (by +128 bits)
'rte_bbdev_state rte_bbdev::state' offset changed from 768 to 896 (in bits) (by +128 bits)
'rte_device* rte_bbdev::device' offset changed from 832 to 960 (in bits) (by +128 bits)
'rte_bbdev_cb_list rte_bbdev::list_cbs' offset changed from 896 to 1024 (in bits) (by +128 bits)
'rte_intr_handle* rte_bbdev::intr_handle' offset changed from 1024 to 1152 (in bits) (by +128 bits)
Thanks
Nic
> -----Original Message-----
> From: Maxime Coquelin <maxime.coquelin at redhat.com>
> Sent: Monday, June 5, 2023 12:08 PM
> To: Stephen Hemminger <stephen at networkplumber.org>; Chautru, Nicolas
> <nicolas.chautru at intel.com>
> Cc: dev at dpdk.org; Rix, Tom <trix at redhat.com>; hemant.agrawal at nxp.com;
> david.marchand at redhat.com; Vargas, Hernan <hernan.vargas at intel.com>
> Subject: Re: [PATCH v1 1/1] doc: announce change in bbdev api related to
> operation extension
>
>
>
> On 5/26/23 05:47, Stephen Hemminger wrote:
> > On Fri, 26 May 2023 02:11:32 +0000
> > Nicolas Chautru <nicolas.chautru at intel.com> wrote:
> >
> >> +
> >> +* bbdev: Will extend the API to support the new operation type
> >> +``RTE_BBDEV_OP_MLDTS`` as per
> >> + this `v1
> >> +<https://patches.dpdk.org/project/dpdk/list/?series=28192>`. This
> >> +will also introduce
> >> + new symbols for ``rte_bbdev_dequeue_mldts_ops``,
> >> +``rte_bbdev_enqueue_mldts_ops``,
> >> + ``rte_bbdev_mldts_op_alloc_bulk`` and
> >> +``rte_bbdev_mldts_op_free_bulk``. This will also extend
> >> + the API related to the FFT operation in ``rte_bbdev_op_fft``.
> >> --
> >
> > New symbols do not require a deprecation notice.
> > Only changes and removal.
> >
> I agree with Stephen.
> There is some changes in struct rte_bbdev_op_fft, but the related API are
> experimental, so I think it is not needed to have a deprecation notice.
>
> Regards,
> Maxime
More information about the dev
mailing list