Re: [PATCH v1 1/2] baseband/acc: FFT support in VRB2 PRQ device

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

Hi Nicolas,

On 10/15/24 00:30, Nicolas Chautru wrote:
> Supporting recent change in the device to
> extend FFT capability processing in latest stepping.
> Also including cosmetic change to VRB2 register definition.
> 
> Signed-off-by: Nicolas Chautru <nicolas.chautru@intel.com>
> ---
>   drivers/baseband/acc/acc_common.h   |  2 +-
>   drivers/baseband/acc/rte_vrb_pmd.c  | 30 +++++++++++++++++++++++++----
>   drivers/baseband/acc/vrb2_vf_enum.h |  4 ++--
>   3 files changed, 29 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/baseband/acc/acc_common.h b/drivers/baseband/acc/acc_common.h
> index 0c249d5b93..4c60b7896b 100644
> --- a/drivers/baseband/acc/acc_common.h
> +++ b/drivers/baseband/acc/acc_common.h
> @@ -106,7 +106,7 @@
>   #define ACC_MAX_FCW_SIZE              128
>   #define ACC_IQ_SIZE                    4
>   
> -#define ACC_FCW_FFT_BLEN_3             28
> +#define ACC_FCW_FFT_BLEN_VRB2         128
>   
>   /* Constants from K0 computation from 3GPP 38.212 Table 5.4.2.1-2 */
>   #define ACC_N_ZC_1 66 /* N = 66 Zc for BG 1 */
> diff --git a/drivers/baseband/acc/rte_vrb_pmd.c b/drivers/baseband/acc/rte_vrb_pmd.c
> index 0455320c2a..5eb3e8dd48 100644
> --- a/drivers/baseband/acc/rte_vrb_pmd.c
> +++ b/drivers/baseband/acc/rte_vrb_pmd.c
> @@ -1006,7 +1006,7 @@ vrb_queue_setup(struct rte_bbdev *dev, uint16_t queue_id,
>   	case RTE_BBDEV_OP_FFT:
>   		fcw_len = ACC_FCW_FFT_BLEN;
>   		if (q->d->device_variant == VRB2_VARIANT)
> -			fcw_len = ACC_FCW_FFT_BLEN_3;
> +			fcw_len = ACC_FCW_FFT_BLEN_VRB2;
>   		break;
>   	case RTE_BBDEV_OP_MLDTS:
>   		fcw_len = ACC_FCW_MLDTS_BLEN;
> @@ -1402,7 +1402,11 @@ vrb_dev_info_get(struct rte_bbdev *dev, struct rte_bbdev_driver_info *dev_info)
>   						RTE_BBDEV_FFT_FP16_INPUT |
>   						RTE_BBDEV_FFT_FP16_OUTPUT |
>   						RTE_BBDEV_FFT_POWER_MEAS |
> -						RTE_BBDEV_FFT_WINDOWING_BYPASS,
> +						RTE_BBDEV_FFT_WINDOWING_BYPASS |
> +						RTE_BBDEV_FFT_TIMING_OFFSET_PER_CS |
> +						RTE_BBDEV_FFT_TIMING_ERROR |
> +						RTE_BBDEV_FFT_DEWINDOWING |
> +						RTE_BBDEV_FFT_FREQ_RESAMPLING,
>   				.num_buffers_src = 1,
>   				.num_buffers_dst = 1,
>   				.fft_windows_num = ACC_MAX_FFT_WIN,
> @@ -3725,6 +3729,8 @@ vrb1_fcw_fft_fill(struct rte_bbdev_fft_op *op, struct acc_fcw_fft *fcw)
>   static inline void
>   vrb2_fcw_fft_fill(struct rte_bbdev_fft_op *op, struct acc_fcw_fft_3 *fcw)
>   {
> +	uint8_t cs;
> +
>   	fcw->in_frame_size = op->fft.input_sequence_size;
>   	fcw->leading_pad_size = op->fft.input_leading_padding;
>   	fcw->out_frame_size = op->fft.output_sequence_size;
> @@ -3760,6 +3766,16 @@ vrb2_fcw_fft_fill(struct rte_bbdev_fft_op *op, struct acc_fcw_fft_3 *fcw)
>   		fcw->bypass = 3;
>   	else
>   		fcw->bypass = 0;
> +
> +	fcw->enable_dewin = check_bit(op->fft.op_flags, RTE_BBDEV_FFT_DEWINDOWING);
> +	fcw->freq_resample_mode = op->fft.freq_resample_mode;
> +	fcw->depad_output_size = fcw->freq_resample_mode == 0 ?
> +			op->fft.output_sequence_size : op->fft.output_depadded_size;
> +	for (cs = 0; cs < RTE_BBDEV_MAX_CS; cs++) {
> +		fcw->cs_theta_0[cs] = op->fft.cs_theta_0[cs];
> +		fcw->cs_theta_d[cs] = op->fft.cs_theta_d[cs];
> +		fcw->cs_time_offset[cs] = op->fft.time_offset[cs];
> +	}
>   }
>   
>   static inline int
> @@ -3782,8 +3798,14 @@ vrb_dma_desc_fft_fill(struct rte_bbdev_fft_op *op,
>   	/* FCW already done */
>   	acc_header_init(desc);
>   
> -	RTE_SET_USED(win_input);
> -	RTE_SET_USED(win_offset);
> +	if (win_en && win_input) {
> +		desc->data_ptrs[bd_idx].address = rte_pktmbuf_iova_offset(win_input, *win_offset);
> +		desc->data_ptrs[bd_idx].blen = op->fft.output_depadded_size * 2;
> +		desc->data_ptrs[bd_idx].blkid = ACC_DMA_BLKID_DEWIN_IN;
> +		desc->data_ptrs[bd_idx].last = 0;
> +		desc->data_ptrs[bd_idx].dma_ext = 0;
> +		bd_idx++;
> +	}
>   
>   	desc->data_ptrs[bd_idx].address = rte_pktmbuf_iova_offset(input, *in_offset);
>   	desc->data_ptrs[bd_idx].blen = op->fft.input_sequence_size * ACC_IQ_SIZE;
> diff --git a/drivers/baseband/acc/vrb2_vf_enum.h b/drivers/baseband/acc/vrb2_vf_enum.h
> index 9c6e451010..1cc6986c67 100644
> --- a/drivers/baseband/acc/vrb2_vf_enum.h
> +++ b/drivers/baseband/acc/vrb2_vf_enum.h
> @@ -18,8 +18,8 @@ enum {
>   	VRB2_VfHiInfoRingIntWrEnVf       = 0x00000020,
>   	VRB2_VfHiInfoRingPf2VfWrEnVf     = 0x00000024,
>   	VRB2_VfHiMsixVectorMapperVf      = 0x00000060,
> -	VRB2_VfHiDeviceStatus            = 0x00000068,
> -	VRB2_VfHiInterruptSrc            = 0x00000070,
> +	VRB2_VfHiDeviceStatus            = 0x00000064,
> +	VRB2_VfHiInterruptSrc            = 0x00000068,

The offset of the registers change, is that what you describe as
cosmetic change?

Does it have an impact on older DPDK versions? i.e. should it be
backported?


Thanks,
Maxime

>   	VRB2_VfDmaFec5GulDescBaseLoRegVf = 0x00000120,
>   	VRB2_VfDmaFec5GulDescBaseHiRegVf = 0x00000124,
>   	VRB2_VfDmaFec5GulRespPtrLoRegVf  = 0x00000128,