Le 16/03/2021 à 19:46, Ezequiel Garcia a écrit :
Hi Benjamin, The series is looking really good. Some comments below. On Wed, 2021-03-03 at 12:39 +0100, Benjamin Gaignard wrote:Implement all the logic to get G2 hardware decoding HEVC frames. It support up level 5.1 HEVC stream. It doesn't support yet 10 bits formats or scaling feature. Add HANTRO HEVC dedicated control to skip some bits at the beginning of the slice header. That is very specific to this hardware so can't go into uapi structures. Compute the needed value is complex and require information from the stream that only the userland knows so let it provide the correct value to the driver. Signed-off-by: Benjamin Gaignard <benjamin.gaignard@xxxxxxxxxxxxx> --- version 4: - fix Ezequiel comments - use dedicated control as an integer - change hantro_g2_hevc_dec_run prototype to return errors version 2: - squash multiple commits in this one. - fix the comments done by Ezequiel about dma_alloc_coherent usage - fix Dan's comments about control copy, reverse the test logic in tile_buffer_reallocate, rework some goto and return cases. drivers/staging/media/hantro/Makefile | 2 + drivers/staging/media/hantro/hantro.h | 18 + drivers/staging/media/hantro/hantro_drv.c | 53 ++ .../staging/media/hantro/hantro_g2_hevc_dec.c | 587 ++++++++++++++++++ drivers/staging/media/hantro/hantro_g2_regs.h | 198 ++++++ drivers/staging/media/hantro/hantro_hevc.c | 321 ++++++++++ drivers/staging/media/hantro/hantro_hw.h | 49 ++ 7 files changed, 1228 insertions(+) create mode 100644 drivers/staging/media/hantro/hantro_g2_hevc_dec.c create mode 100644 drivers/staging/media/hantro/hantro_g2_regs.h create mode 100644 drivers/staging/media/hantro/hantro_hevc.c diff --git a/drivers/staging/media/hantro/Makefile b/drivers/staging/media/hantro/Makefile index 743ce08eb184..0357f1772267 100644 --- a/drivers/staging/media/hantro/Makefile +++ b/drivers/staging/media/hantro/Makefile @@ -9,12 +9,14 @@ hantro-vpu-y += \ hantro_h1_jpeg_enc.o \ hantro_g1_h264_dec.o \ hantro_g1_mpeg2_dec.o \ + hantro_g2_hevc_dec.o \ hantro_g1_vp8_dec.o \ rk3399_vpu_hw_jpeg_enc.o \ rk3399_vpu_hw_mpeg2_dec.o \ rk3399_vpu_hw_vp8_dec.o \ hantro_jpeg.o \ hantro_h264.o \ + hantro_hevc.o \ hantro_mpeg2.o \ hantro_vp8.odiff --git a/drivers/staging/media/hantro/hantro.h b/drivers/staging/media/hantro/hantro.hindex 05876e426419..a9b80b2c9124 100644 --- a/drivers/staging/media/hantro/hantro.h +++ b/drivers/staging/media/hantro/hantro.h @@ -225,6 +225,7 @@ struct hantro_dev { * @jpeg_enc: JPEG-encoding context. * @mpeg2_dec: MPEG-2-decoding context. * @vp8_dec: VP8-decoding context. + * @hevc_dec: HEVC-decoding context. */ struct hantro_ctx { struct hantro_dev *dev; @@ -251,6 +252,7 @@ struct hantro_ctx { struct hantro_jpeg_enc_hw_ctx jpeg_enc; struct hantro_mpeg2_dec_hw_ctx mpeg2_dec; struct hantro_vp8_dec_hw_ctx vp8_dec; + struct hantro_hevc_dec_hw_ctx hevc_dec; }; };@@ -428,6 +430,22 @@ hantro_get_dec_buf_addr(struct hantro_ctx *ctx, struct vb2_buffer *vb)return vb2_dma_contig_plane_dma_addr(vb, 0); }+static inline size_t+hantro_get_dec_buf_size(struct hantro_ctx *ctx, struct vb2_buffer *vb) +{ + if (hantro_needs_postproc(ctx, ctx->vpu_dst_fmt)) + return ctx->postproc.dec_q[vb->index].size; + return vb2_plane_size(vb, 0); +} + +static inline void * +hantro_get_dec_buf(struct hantro_ctx *ctx, struct vb2_buffer *vb) +{ + if (hantro_needs_postproc(ctx, ctx->vpu_dst_fmt)) + return ctx->postproc.dec_q[vb->index].cpu; + return vb2_plane_vaddr(vb, 0); +} +Seems hantro_get_dec_buf_size and hantro_get_dec_buf are not used?
You are right I will remove them
void hantro_postproc_disable(struct hantro_ctx *ctx); void hantro_postproc_enable(struct hantro_ctx *ctx); void hantro_postproc_free(struct hantro_ctx *ctx); diff --git a/drivers/staging/media/hantro/hantro_drv.c b/drivers/staging/media/hantro/hantro_drv.c index e3e6df28f470..bc90a52f4d3d 100644 --- a/drivers/staging/media/hantro/hantro_drv.c +++ b/drivers/staging/media/hantro/hantro_drv.c @@ -30,6 +30,13 @@#define DRIVER_NAME "hantro-vpu" +/*+ * V4L2_CID_HANTRO_HEVC_SLICE_HEADER_SKIP - + * the number of data (in bits) to skip in the + * slice segment header syntax after 'slice type' token + */I think we need to document this better, so applications can correctly use the control. From i.MX reference code, it seems this needs to be used as follows: If non-IDR, the bits to be skipped go from syntax element "pic_output_flag" to before syntax element "slice_temporal_mvp_enabled_flag". If IDR, the skipped bits are just "pic_output_flag" (separate_colour_plane_flag is not supported). And it seems this needs to be passed parsing only the first slice, given this syntax remains invariant across all the slices.
Ok I will add your description in the next version.
+#define V4L2_CID_HANTRO_HEVC_SLICE_HEADER_SKIP (V4L2_CID_USER_HANTRO_BASE + 0) + int hantro_debug; module_param_named(debug, hantro_debug, int, 0644); MODULE_PARM_DESC(debug, @@ -281,6 +288,26 @@ static int hantro_jpeg_s_ctrl(struct v4l2_ctrl *ctrl) return 0; }+static int hantro_hevc_s_ctrl(struct v4l2_ctrl *ctrl)+{ + struct hantro_ctx *ctx; + + ctx = container_of(ctrl->handler, + struct hantro_ctx, ctrl_handler); + + vpu_debug(1, "s_ctrl: id = %d, val = %d\n", ctrl->id, ctrl->val); + + switch (ctrl->id) { + case V4L2_CID_HANTRO_HEVC_SLICE_HEADER_SKIP: + ctx->hevc_dec.ctrls.hevc_hdr_skip_length = ctrl->val; + break; + default: + return -EINVAL; + } + + return 0; +} + static const struct v4l2_ctrl_ops hantro_ctrl_ops = { .try_ctrl = hantro_try_ctrl, }; @@ -289,6 +316,10 @@ static const struct v4l2_ctrl_ops hantro_jpeg_ctrl_ops = { .s_ctrl = hantro_jpeg_s_ctrl, };+static const struct v4l2_ctrl_ops hantro_hevc_ctrl_ops = {+ .s_ctrl = hantro_hevc_s_ctrl, +}; + static const struct hantro_ctrl controls[] = { { .codec = HANTRO_JPEG_ENCODER, @@ -409,6 +440,28 @@ static const struct hantro_ctrl controls[] = { .cfg = { .id = V4L2_CID_MPEG_VIDEO_HEVC_DECODE_PARAMS, }, + }, { + .codec = HANTRO_HEVC_DECODER, + .cfg = { + .id = V4L2_CID_HANTRO_HEVC_SLICE_HEADER_SKIP, + .name = "Hantro HEVC slice header skip bytes", + .type = V4L2_CTRL_TYPE_INTEGER, + .min = 0, + .def = 0, + .max = 0x7fffffff, + .step = 1, + .ops = &hantro_hevc_ctrl_ops, + }, + }, { + .codec = HANTRO_JPEG_ENCODER | HANTRO_MPEG2_DECODER | + HANTRO_VP8_DECODER | HANTRO_H264_DECODER | + HANTRO_HEVC_DECODER, + .cfg = { + .id = V4L2_CID_USER_CLASS,This shouldn't be here, is this V4L2_CID_USER_CLASS required by v4l2-compliance or by the spec?
It is required by v4l2-compliance.
+ .name = "HANTRO controls", + .type = V4L2_CTRL_TYPE_CTRL_CLASS, + .flags = V4L2_CTRL_FLAG_READ_ONLY | V4L2_CTRL_FLAG_WRITE_ONLY, + }, }, };diff --git a/drivers/staging/media/hantro/hantro_g2_hevc_dec.c b/drivers/staging/media/hantro/hantro_g2_hevc_dec.cnew file mode 100644 index 000000000000..5d75b36bc40c --- /dev/null +++ b/drivers/staging/media/hantro/hantro_g2_hevc_dec.c @@ -0,0 +1,587 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Hantro VPU HEVC codec driver + * + * Copyright (C) 2020 Safran Passenger Innovations LLC + */ + +#include "hantro_hw.h" +#include "hantro_g2_regs.h" + +#define HEVC_DEC_MODE 0xC + +#define BUS_WIDTH_32 0 +#define BUS_WIDTH_64 1 +#define BUS_WIDTH_128 2 +#define BUS_WIDTH_256 3 + +static inline void hantro_write_addr(struct hantro_dev *vpu, + unsigned long offset, + dma_addr_t addr) +{ + vdpu_write(vpu, addr & 0xffffffff, offset); +} + +static void prepare_tile_info_buffer(struct hantro_ctx *ctx) +{ + struct hantro_dev *vpu = ctx->dev; + const struct hantro_hevc_dec_ctrls *ctrls = &ctx->hevc_dec.ctrls; + const struct v4l2_ctrl_hevc_pps *pps = ctrls->pps; + const struct v4l2_ctrl_hevc_sps *sps = ctrls->sps; + u16 *p = (u16 *)((u8 *)ctx->hevc_dec.tile_sizes.cpu); + unsigned int num_tile_rows = pps->num_tile_rows_minus1 + 1; + unsigned int num_tile_cols = pps->num_tile_columns_minus1 + 1; + unsigned int pic_width_in_ctbs, pic_height_in_ctbs; + unsigned int max_log2_ctb_size, ctb_size; + bool tiles_enabled, uniform_spacing; + u32 no_chroma = 0; + + tiles_enabled = !!(pps->flags & V4L2_HEVC_PPS_FLAG_TILES_ENABLED); + uniform_spacing = !!(pps->flags & V4L2_HEVC_PPS_FLAG_UNIFORM_SPACING); + + hantro_reg_write(vpu, hevc_tile_e, tiles_enabled); + + max_log2_ctb_size = sps->log2_min_luma_coding_block_size_minus3 + 3 + + sps->log2_diff_max_min_luma_coding_block_size; + pic_width_in_ctbs = (sps->pic_width_in_luma_samples + + (1 << max_log2_ctb_size) - 1) >> max_log2_ctb_size; + pic_height_in_ctbs = (sps->pic_height_in_luma_samples + (1 << max_log2_ctb_size) - 1) + >> max_log2_ctb_size; + ctb_size = 1 << max_log2_ctb_size; + + vpu_debug(1, "Preparing tile sizes buffer for %dx%d CTBs (CTB size %d)\n", + pic_width_in_ctbs, pic_height_in_ctbs, ctb_size); + + if (tiles_enabled) { + unsigned int i, j, h; + + vpu_debug(1, "Tiles enabled! %dx%d\n", num_tile_cols, num_tile_rows); + + hantro_reg_write(vpu, hevc_num_tile_rows, num_tile_rows); + hantro_reg_write(vpu, hevc_num_tile_cols, num_tile_cols); + + /* write width + height for each tile in pic */ + if (!uniform_spacing) { + u32 tmp_w = 0, tmp_h = 0; + + for (i = 0; i < num_tile_rows; i++) { + if (i == num_tile_rows - 1) + h = pic_height_in_ctbs - tmp_h; + else + h = pps->row_height_minus1[i] + 1; + tmp_h += h; + if (i == 0 && h == 1 && ctb_size == 16) + no_chroma = 1; + for (j = 0, tmp_w = 0; j < num_tile_cols - 1; j++) { + tmp_w += pps->column_width_minus1[j] + 1; + *p++ = pps->column_width_minus1[j + 1]; + *p++ = h; + if (i == 0 && h == 1 && ctb_size == 16) + no_chroma = 1; + } + /* last column */ + *p++ = pic_width_in_ctbs - tmp_w; + *p++ = h; + } + } else { /* uniform spacing */ + u32 tmp, prev_h, prev_w; + + for (i = 0, prev_h = 0; i < num_tile_rows; i++) { + tmp = (i + 1) * pic_height_in_ctbs / num_tile_rows; + h = tmp - prev_h; + prev_h = tmp; + if (i == 0 && h == 1 && ctb_size == 16) + no_chroma = 1; + for (j = 0, prev_w = 0; j < num_tile_cols; j++) { + tmp = (j + 1) * pic_width_in_ctbs / num_tile_cols; + *p++ = tmp - prev_w; + *p++ = h; + if (j == 0 && + (pps->column_width_minus1[0] + 1) == 1 && + ctb_size == 16) + no_chroma = 1; + prev_w = tmp; + } + } + } + } else { + hantro_reg_write(vpu, hevc_num_tile_rows, 1); + hantro_reg_write(vpu, hevc_num_tile_cols, 1); + + /* There's one tile, with dimensions equal to pic size. */ + p[0] = pic_width_in_ctbs; + p[1] = pic_height_in_ctbs; + } + + if (no_chroma) + vpu_debug(1, "%s: no chroma!\n", __func__); +} + +static void set_params(struct hantro_ctx *ctx) +{ + const struct hantro_hevc_dec_ctrls *ctrls = &ctx->hevc_dec.ctrls; + const struct v4l2_ctrl_hevc_sps *sps = ctrls->sps; + const struct v4l2_ctrl_hevc_pps *pps = ctrls->pps; + const struct v4l2_ctrl_hevc_decode_params *decode_params = ctrls->decode_params; + struct hantro_dev *vpu = ctx->dev; + u32 min_log2_cb_size, max_log2_ctb_size, min_cb_size, max_ctb_size; + u32 pic_width_in_min_cbs, pic_height_in_min_cbs; + u32 pic_width_aligned, pic_height_aligned; + u32 partial_ctb_x, partial_ctb_y; + + hantro_reg_write(vpu, hevc_bit_depth_y_minus8, sps->bit_depth_luma_minus8); + hantro_reg_write(vpu, hevc_bit_depth_c_minus8, sps->bit_depth_chroma_minus8); + + hantro_reg_write(vpu, hevc_output_8_bits, 0); + + hantro_reg_write(vpu, hevc_hdr_skip_length, ctrls->hevc_hdr_skip_length); + + min_log2_cb_size = sps->log2_min_luma_coding_block_size_minus3 + 3; + max_log2_ctb_size = min_log2_cb_size + sps->log2_diff_max_min_luma_coding_block_size; + + hantro_reg_write(vpu, hevc_min_cb_size, min_log2_cb_size); + hantro_reg_write(vpu, hevc_max_cb_size, max_log2_ctb_size); + + min_cb_size = 1 << min_log2_cb_size; + max_ctb_size = 1 << max_log2_ctb_size; + + pic_width_in_min_cbs = sps->pic_width_in_luma_samples / min_cb_size; + pic_height_in_min_cbs = sps->pic_height_in_luma_samples / min_cb_size; + pic_width_aligned = ALIGN(sps->pic_width_in_luma_samples, max_ctb_size); + pic_height_aligned = ALIGN(sps->pic_height_in_luma_samples, max_ctb_size); + + partial_ctb_x = !!(sps->pic_width_in_luma_samples != pic_width_aligned); + partial_ctb_y = !!(sps->pic_height_in_luma_samples != pic_height_aligned); + + hantro_reg_write(vpu, hevc_partial_ctb_x, partial_ctb_x); + hantro_reg_write(vpu, hevc_partial_ctb_y, partial_ctb_y); + + hantro_reg_write(vpu, hevc_pic_width_in_cbs, pic_width_in_min_cbs); + hantro_reg_write(vpu, hevc_pic_height_in_cbs, pic_height_in_min_cbs); + + hantro_reg_write(vpu, hevc_pic_width_4x4, + (pic_width_in_min_cbs * min_cb_size) / 4); + hantro_reg_write(vpu, hevc_pic_height_4x4, + (pic_height_in_min_cbs * min_cb_size) / 4); + + hantro_reg_write(vpu, hevc_max_inter_hierdepth, + sps->max_transform_hierarchy_depth_inter); + hantro_reg_write(vpu, hevc_max_intra_hierdepth, + sps->max_transform_hierarchy_depth_intra); + hantro_reg_write(vpu, hevc_min_trb_size, + sps->log2_min_luma_transform_block_size_minus2 + 2); + hantro_reg_write(vpu, hevc_max_trb_size, + sps->log2_min_luma_transform_block_size_minus2 + 2 + + sps->log2_diff_max_min_luma_transform_block_size); + + hantro_reg_write(vpu, hevc_tempor_mvp_e, + !!(sps->flags & V4L2_HEVC_SPS_FLAG_SPS_TEMPORAL_MVP_ENABLED) && + !(decode_params->flags & V4L2_HEVC_DECODE_PARAM_FLAG_IDR_PIC)); + hantro_reg_write(vpu, hevc_strong_smooth_e, + !!(sps->flags & V4L2_HEVC_SPS_FLAG_STRONG_INTRA_SMOOTHING_ENABLED)); + hantro_reg_write(vpu, hevc_asym_pred_e, + !!(sps->flags & V4L2_HEVC_SPS_FLAG_AMP_ENABLED)); + hantro_reg_write(vpu, hevc_sao_e, + !!(sps->flags & V4L2_HEVC_SPS_FLAG_SAMPLE_ADAPTIVE_OFFSET)); + hantro_reg_write(vpu, hevc_sign_data_hide, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_SIGN_DATA_HIDING_ENABLED)); + + if (pps->flags & V4L2_HEVC_PPS_FLAG_CU_QP_DELTA_ENABLED) { + hantro_reg_write(vpu, hevc_cu_qpd_e, 1); + hantro_reg_write(vpu, hevc_max_cu_qpd_depth, pps->diff_cu_qp_delta_depth); + } else { + hantro_reg_write(vpu, hevc_cu_qpd_e, 0); + hantro_reg_write(vpu, hevc_max_cu_qpd_depth, 0); + } + + if (pps->flags & V4L2_HEVC_PPS_FLAG_PPS_SLICE_CHROMA_QP_OFFSETS_PRESENT) { + hantro_reg_write(vpu, hevc_cb_qp_offset, pps->pps_cb_qp_offset); + hantro_reg_write(vpu, hevc_cr_qp_offset, pps->pps_cr_qp_offset); + } else { + hantro_reg_write(vpu, hevc_cb_qp_offset, 0); + hantro_reg_write(vpu, hevc_cr_qp_offset, 0); + } + + hantro_reg_write(vpu, hevc_filt_offset_beta, pps->pps_beta_offset_div2); + hantro_reg_write(vpu, hevc_filt_offset_tc, pps->pps_tc_offset_div2); + hantro_reg_write(vpu, hevc_slice_hdr_ext_e, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_SLICE_SEGMENT_HEADER_EXTENSION_PRESENT)); + hantro_reg_write(vpu, hevc_slice_hdr_ext_bits, pps->num_extra_slice_header_bits); + hantro_reg_write(vpu, hevc_slice_chqp_present, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_PPS_SLICE_CHROMA_QP_OFFSETS_PRESENT)); + hantro_reg_write(vpu, hevc_weight_bipr_idc, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_WEIGHTED_BIPRED)); + hantro_reg_write(vpu, hevc_transq_bypass, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_TRANSQUANT_BYPASS_ENABLED)); + hantro_reg_write(vpu, hevc_list_mod_e, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_LISTS_MODIFICATION_PRESENT)); + hantro_reg_write(vpu, hevc_entropy_sync_e, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_ENTROPY_CODING_SYNC_ENABLED)); + hantro_reg_write(vpu, hevc_cabac_init_present, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_CABAC_INIT_PRESENT)); + hantro_reg_write(vpu, hevc_idr_pic_e, + !!(decode_params->flags & V4L2_HEVC_DECODE_PARAM_FLAG_IRAP_PIC)); + hantro_reg_write(vpu, hevc_parallel_merge, + pps->log2_parallel_merge_level_minus2 + 2); + hantro_reg_write(vpu, hevc_pcm_filt_d, + !!(sps->flags & V4L2_HEVC_SPS_FLAG_PCM_LOOP_FILTER_DISABLED)); + hantro_reg_write(vpu, hevc_pcm_e, + !!(sps->flags & V4L2_HEVC_SPS_FLAG_PCM_ENABLED)); + if (sps->flags & V4L2_HEVC_SPS_FLAG_PCM_ENABLED) { + hantro_reg_write(vpu, hevc_max_pcm_size, + sps->log2_diff_max_min_pcm_luma_coding_block_size + + sps->log2_min_pcm_luma_coding_block_size_minus3 + 3); + hantro_reg_write(vpu, hevc_min_pcm_size, + sps->log2_min_pcm_luma_coding_block_size_minus3 + 3); + hantro_reg_write(vpu, hevc_bit_depth_pcm_y, + sps->pcm_sample_bit_depth_luma_minus1 + 1); + hantro_reg_write(vpu, hevc_bit_depth_pcm_c, + sps->pcm_sample_bit_depth_chroma_minus1 + 1); + } else { + hantro_reg_write(vpu, hevc_max_pcm_size, 0); + hantro_reg_write(vpu, hevc_min_pcm_size, 0); + hantro_reg_write(vpu, hevc_bit_depth_pcm_y, 0); + hantro_reg_write(vpu, hevc_bit_depth_pcm_c, 0); + } + + hantro_reg_write(vpu, hevc_start_code_e, 1); + hantro_reg_write(vpu, hevc_init_qp, pps->init_qp_minus26 + 26); + hantro_reg_write(vpu, hevc_weight_pred_e, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_WEIGHTED_PRED)); + hantro_reg_write(vpu, hevc_cabac_init_present, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_CABAC_INIT_PRESENT)); + hantro_reg_write(vpu, hevc_const_intra_e, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_CONSTRAINED_INTRA_PRED)); + hantro_reg_write(vpu, hevc_transform_skip, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_TRANSFORM_SKIP_ENABLED)); + hantro_reg_write(vpu, hevc_out_filtering_dis, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_PPS_DISABLE_DEBLOCKING_FILTER)); + hantro_reg_write(vpu, hevc_filt_ctrl_pres, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_DEBLOCKING_FILTER_CONTROL_PRESENT)); + hantro_reg_write(vpu, hevc_dependent_slice, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_DEPENDENT_SLICE_SEGMENT)); + hantro_reg_write(vpu, hevc_filter_override, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_DEBLOCKING_FILTER_OVERRIDE_ENABLED)); + hantro_reg_write(vpu, hevc_refidx0_active, + pps->num_ref_idx_l0_default_active_minus1 + 1); + hantro_reg_write(vpu, hevc_refidx1_active, + pps->num_ref_idx_l1_default_active_minus1 + 1); + hantro_reg_write(vpu, hevc_apf_threshold, 8); +} + +static int find_ref_pic_index(const struct v4l2_hevc_dpb_entry *dpb, int pic_order_cnt) +{ + int i; + + for (i = 0; i < V4L2_HEVC_DPB_ENTRIES_NUM_MAX; i++) { + if (dpb[i].pic_order_cnt[0] == pic_order_cnt) + return i; + } + + return 0x0; +} + +static void set_ref_pic_list(struct hantro_ctx *ctx) +{ + const struct hantro_hevc_dec_ctrls *ctrls = &ctx->hevc_dec.ctrls; + struct hantro_dev *vpu = ctx->dev; + const struct v4l2_ctrl_hevc_decode_params *decode_params = ctrls->decode_params; + const struct v4l2_hevc_dpb_entry *dpb = decode_params->dpb; + u32 list0[V4L2_HEVC_DPB_ENTRIES_NUM_MAX] = {0}; + u32 list1[V4L2_HEVC_DPB_ENTRIES_NUM_MAX] = {0}; + const struct hantro_reg *ref_pic_regs0[] = { + hevc_rlist_f0, + hevc_rlist_f1, + hevc_rlist_f2, + hevc_rlist_f3, + hevc_rlist_f4, + hevc_rlist_f5, + hevc_rlist_f6, + hevc_rlist_f7, + hevc_rlist_f8, + hevc_rlist_f9, + hevc_rlist_f10, + hevc_rlist_f11, + hevc_rlist_f12, + hevc_rlist_f13, + hevc_rlist_f14, + hevc_rlist_f15, + }; + const struct hantro_reg *ref_pic_regs1[] = { + hevc_rlist_b0, + hevc_rlist_b1, + hevc_rlist_b2, + hevc_rlist_b3, + hevc_rlist_b4, + hevc_rlist_b5, + hevc_rlist_b6, + hevc_rlist_b7, + hevc_rlist_b8, + hevc_rlist_b9, + hevc_rlist_b10, + hevc_rlist_b11, + hevc_rlist_b12, + hevc_rlist_b13, + hevc_rlist_b14, + hevc_rlist_b15, + }; + unsigned int i, j; + + /* List 0 contains: short term before, short term after and long term */ + j = 0; + for (i = 0; i < decode_params->num_poc_st_curr_before && j < ARRAY_SIZE(list0); i++) + list0[j++] = find_ref_pic_index(dpb, decode_params->poc_st_curr_before[i]); + for (i = 0; i < decode_params->num_poc_st_curr_after && j < ARRAY_SIZE(list0); i++) + list0[j++] = find_ref_pic_index(dpb, decode_params->poc_st_curr_after[i]); + for (i = 0; i < decode_params->num_poc_lt_curr && j < ARRAY_SIZE(list0); i++) + list0[j++] = find_ref_pic_index(dpb, decode_params->poc_lt_curr[i]); + + /* Fill the list, copying over and over */ + i = 0; + while (j < ARRAY_SIZE(list0)) + list0[j++] = list0[i++]; + + j = 0; + for (i = 0; i < decode_params->num_poc_st_curr_after && j < ARRAY_SIZE(list1); i++) + list1[j++] = find_ref_pic_index(dpb, decode_params->poc_st_curr_after[i]); + for (i = 0; i < decode_params->num_poc_st_curr_before && j < ARRAY_SIZE(list1); i++) + list1[j++] = find_ref_pic_index(dpb, decode_params->poc_st_curr_before[i]); + for (i = 0; i < decode_params->num_poc_lt_curr && j < ARRAY_SIZE(list1); i++) + list1[j++] = find_ref_pic_index(dpb, decode_params->poc_lt_curr[i]); + + i = 0; + while (j < ARRAY_SIZE(list1)) + list1[j++] = list1[i++]; + + for (i = 0; i < V4L2_HEVC_DPB_ENTRIES_NUM_MAX; i++) { + hantro_reg_write(vpu, ref_pic_regs0[i], list0[i]); + hantro_reg_write(vpu, ref_pic_regs1[i], list1[i]); + } +} + +static int set_ref(struct hantro_ctx *ctx) +{ + const struct hantro_hevc_dec_ctrls *ctrls = &ctx->hevc_dec.ctrls; + const struct v4l2_ctrl_hevc_sps *sps = ctrls->sps; + const struct v4l2_ctrl_hevc_pps *pps = ctrls->pps; + const struct v4l2_ctrl_hevc_decode_params *decode_params = ctrls->decode_params; + const struct v4l2_hevc_dpb_entry *dpb = decode_params->dpb; + dma_addr_t luma_addr, chroma_addr, mv_addr = 0; + struct hantro_dev *vpu = ctx->dev; + size_t cr_offset = hantro_hevc_chroma_offset(sps); + size_t mv_offset = hantro_hevc_motion_vectors_offset(sps); + u32 max_ref_frames; + u16 dpb_longterm_e; + + const struct hantro_reg *cur_poc[] = { + hevc_cur_poc_00, + hevc_cur_poc_01, + hevc_cur_poc_02, + hevc_cur_poc_03, + hevc_cur_poc_04, + hevc_cur_poc_05, + hevc_cur_poc_06, + hevc_cur_poc_07, + hevc_cur_poc_08, + hevc_cur_poc_09, + hevc_cur_poc_10, + hevc_cur_poc_11, + hevc_cur_poc_12, + hevc_cur_poc_13, + hevc_cur_poc_14, + hevc_cur_poc_15, + }; + unsigned int i; + + max_ref_frames = decode_params->num_poc_lt_curr + + decode_params->num_poc_st_curr_before + + decode_params->num_poc_st_curr_after; + /* + * Set max_ref_frames to non-zero to avoid HW hang when decoding + * badly marked I-frames. + */ + max_ref_frames = max_ref_frames ? max_ref_frames : 1; + hantro_reg_write(vpu, hevc_num_ref_frames, max_ref_frames); + hantro_reg_write(vpu, hevc_filter_over_slices, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_PPS_LOOP_FILTER_ACROSS_SLICES_ENABLED)); + hantro_reg_write(vpu, hevc_filter_over_tiles, + !!(pps->flags & V4L2_HEVC_PPS_FLAG_LOOP_FILTER_ACROSS_TILES_ENABLED)); + + /* + * Write POC count diff from current pic. For frame decoding only compute + * pic_order_cnt[0] and ignore pic_order_cnt[1] used in field-coding. + */ + for (i = 0; i < decode_params->num_active_dpb_entries && i < ARRAY_SIZE(cur_poc); i++) { + char poc_diff = decode_params->pic_order_cnt_val - dpb[i].pic_order_cnt[0]; + + hantro_reg_write(vpu, cur_poc[i], poc_diff); + } + + if (i < ARRAY_SIZE(cur_poc)) { + /* + * After the references, fill one entry pointing to itself, + * i.e. difference is zero. + */ + hantro_reg_write(vpu, cur_poc[i], 0); + i++; + } + + /* Fill the rest with the current picture */ + for (; i < ARRAY_SIZE(cur_poc); i++) + hantro_reg_write(vpu, cur_poc[i], decode_params->pic_order_cnt_val); + + set_ref_pic_list(ctx); + + /* We will only keep the references picture that are still used */ + ctx->hevc_dec.ref_bufs_used = 0; + + /* Set up addresses of DPB buffers */ + dpb_longterm_e = 0; + for (i = 0; i < decode_params->num_active_dpb_entries && + i < (V4L2_HEVC_DPB_ENTRIES_NUM_MAX - 1); i++) { + luma_addr = hantro_hevc_get_ref_buf(ctx, dpb[i].pic_order_cnt[0]); + if (!luma_addr) + return -ENOMEM; + + chroma_addr = luma_addr + cr_offset; + mv_addr = luma_addr + mv_offset; + + if (dpb[i].rps == V4L2_HEVC_DPB_ENTRY_RPS_LT_CURR) + dpb_longterm_e |= BIT(V4L2_HEVC_DPB_ENTRIES_NUM_MAX - 1 - i); + + hantro_write_addr(vpu, HEVC_REG_ADDR_REF(i), luma_addr); + hantro_write_addr(vpu, HEVC_REG_CHR_REF(i), chroma_addr); + hantro_write_addr(vpu, HEVC_REG_DMV_REF(i), mv_addr); + } + + luma_addr = hantro_hevc_get_ref_buf(ctx, decode_params->pic_order_cnt_val); + if (!luma_addr) + return -ENOMEM; + + chroma_addr = luma_addr + cr_offset; + mv_addr = luma_addr + mv_offset; + + hantro_write_addr(vpu, HEVC_REG_ADDR_REF(i), luma_addr); + hantro_write_addr(vpu, HEVC_REG_CHR_REF(i), chroma_addr); + hantro_write_addr(vpu, HEVC_REG_DMV_REF(i++), mv_addr); + + hantro_write_addr(vpu, HEVC_ADDR_DST, luma_addr); + hantro_write_addr(vpu, HEVC_ADDR_DST_CHR, chroma_addr); + hantro_write_addr(vpu, HEVC_ADDR_DST_MV, mv_addr); + + hantro_hevc_ref_remove_unused(ctx); + + for (; i < V4L2_HEVC_DPB_ENTRIES_NUM_MAX; i++) { + hantro_write_addr(vpu, HEVC_REG_ADDR_REF(i), 0); + hantro_write_addr(vpu, HEVC_REG_CHR_REF(i), 0); + hantro_write_addr(vpu, HEVC_REG_DMV_REF(i), 0); + } + + hantro_reg_write(vpu, hevc_refer_lterm_e, dpb_longterm_e); + + return 0; +} + +static void set_buffers(struct hantro_ctx *ctx) +{ + struct vb2_v4l2_buffer *src_buf, *dst_buf; + struct hantro_dev *vpu = ctx->dev; + const struct hantro_hevc_dec_ctrls *ctrls = &ctx->hevc_dec.ctrls; + const struct v4l2_ctrl_hevc_sps *sps = ctrls->sps; + size_t cr_offset = hantro_hevc_chroma_offset(sps); + dma_addr_t src_dma, dst_dma; + u32 src_len, src_buf_len; + + src_buf = hantro_get_src_buf(ctx); + dst_buf = hantro_get_dst_buf(ctx); + + /* Source (stream) buffer. */ + src_dma = vb2_dma_contig_plane_dma_addr(&src_buf->vb2_buf, 0); + src_len = vb2_get_plane_payload(&src_buf->vb2_buf, 0); + src_buf_len = vb2_plane_size(&src_buf->vb2_buf, 0); + + hantro_write_addr(vpu, HEVC_ADDR_STR, src_dma); + hantro_reg_write(vpu, hevc_stream_len, src_len); + hantro_reg_write(vpu, hevc_strm_buffer_len, src_buf_len); + hantro_reg_write(vpu, hevc_strm_start_offset, 0); + hantro_reg_write(vpu, hevc_write_mvs_e, 1); + + /* Destination (decoded frame) buffer. */ + dst_dma = hantro_get_dec_buf_addr(ctx, &dst_buf->vb2_buf); + + hantro_write_addr(vpu, HEVC_RASTER_SCAN, dst_dma); + hantro_write_addr(vpu, HEVC_RASTER_SCAN_CHR, dst_dma + cr_offset); + hantro_write_addr(vpu, HEVC_ADDR_TILE_SIZE, ctx->hevc_dec.tile_sizes.dma); + hantro_write_addr(vpu, HEVC_TILE_FILTER, ctx->hevc_dec.tile_filter.dma); + hantro_write_addr(vpu, HEVC_TILE_SAO, ctx->hevc_dec.tile_sao.dma); + hantro_write_addr(vpu, HEVC_TILE_BSD, ctx->hevc_dec.tile_bsd.dma); +} + +void hantro_g2_check_idle(struct hantro_dev *vpu) +{ + int i; + + for (i = 0; i < 3; i++) { + u32 status; + + /* Make sure the VPU is idle */ + status = vdpu_read(vpu, HEVC_REG_INTERRUPT); + if (status & HEVC_REG_INTERRUPT_DEC_E) { + pr_warn("%s: still enabled!!! resetting.\n", __func__);How about we clean this pr_warn: use either v4l2_warn or dev_warn and make the warning "device still running, aborting" (I personally dislike the abort metaphor, but guess it's OK here).
Ok
+ status |= HEVC_REG_INTERRUPT_DEC_ABORT_E | HEVC_REG_INTERRUPT_DEC_IRQ_DIS; + vdpu_write(vpu, status, HEVC_REG_INTERRUPT); + } + } +} + +int hantro_g2_hevc_dec_run(struct hantro_ctx *ctx) +{ + struct hantro_dev *vpu = ctx->dev; + int ret; + + hantro_g2_check_idle(vpu); + + /* Prepare HEVC decoder context. */ + ret = hantro_hevc_dec_prepare_run(ctx); + if (ret) + return ret; + + /* Configure hardware registers. */ + set_params(ctx); + + /* set reference pictures */ + ret = set_ref(ctx); + if (ret) + return ret; + + set_buffers(ctx); + prepare_tile_info_buffer(ctx); + + hantro_end_prepare_run(ctx); + + hantro_reg_write(vpu, hevc_mode, HEVC_DEC_MODE); + hantro_reg_write(vpu, hevc_clk_gate_e, 1); + + /* Don't disable output */ + hantro_reg_write(vpu, hevc_out_dis, 0); + + /* Don't compress buffers */ + hantro_reg_write(vpu, hevc_ref_compress_bypass, 1); + + /* use NV12 as output format */ + hantro_reg_write(vpu, hevc_out_rs_e, 1); + + /* Bus width and max burst */ + hantro_reg_write(vpu, hevc_buswidth, BUS_WIDTH_128); + hantro_reg_write(vpu, hevc_max_burst, 16); + + /* Swap */ + hantro_reg_write(vpu, hevc_strm_swap, 0xf); + hantro_reg_write(vpu, hevc_dirmv_swap, 0xf); + hantro_reg_write(vpu, hevc_compress_swap, 0xf); + + /* Start decoding! */ + vdpu_write(vpu, HEVC_REG_INTERRUPT_DEC_E, HEVC_REG_INTERRUPT); + + return 0; +} diff --git a/drivers/staging/media/hantro/hantro_g2_regs.h b/drivers/staging/media/hantro/hantro_g2_regs.h new file mode 100644 index 000000000000..a361c9ba911d --- /dev/null +++ b/drivers/staging/media/hantro/hantro_g2_regs.h @@ -0,0 +1,198 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * Copyright (c) 2021, Collabora + * + * Author: Benjamin Gaignard <benjamin.gaignard@xxxxxxxxxxxxx> + */ + +#ifndef HANTRO_G2_REGS_H_ +#define HANTRO_G2_REGS_H_ + +#include "hantro.h" + +#define G2_SWREG(nr) ((nr) * 4) + +#define HEVC_DEC_REG(name, base, shift, mask) \ + static const struct hantro_reg _hevc_##name[] = { \ + { G2_SWREG(base), (shift), (mask) } \ + }; \ + static const struct hantro_reg __maybe_unused *hevc_##name = &_hevc_##name[0]; + +#define HEVC_REG_VERSION G2_SWREG(0) + +#define HEVC_REG_INTERRUPT G2_SWREG(1) +#define HEVC_REG_INTERRUPT_DEC_RDY_INT BIT(12) +#define HEVC_REG_INTERRUPT_DEC_ABORT_E BIT(5) +#define HEVC_REG_INTERRUPT_DEC_IRQ_DIS BIT(4) +#define HEVC_REG_INTERRUPT_DEC_E BIT(0) + +HEVC_DEC_REG(strm_swap, 2, 28, 0xf) +HEVC_DEC_REG(dirmv_swap, 2, 20, 0xf) + +HEVC_DEC_REG(mode, 3, 27, 0x1f) +HEVC_DEC_REG(compress_swap, 3, 20, 0xf) +HEVC_DEC_REG(ref_compress_bypass, 3, 17, 0x1) +HEVC_DEC_REG(out_rs_e, 3, 16, 0x1) +HEVC_DEC_REG(out_dis, 3, 15, 0x1) +HEVC_DEC_REG(out_filtering_dis, 3, 14, 0x1) +HEVC_DEC_REG(write_mvs_e, 3, 12, 0x1) + +HEVC_DEC_REG(pic_width_in_cbs, 4, 19, 0x1ff) +HEVC_DEC_REG(pic_height_in_cbs, 4, 6, 0x1ff) +HEVC_DEC_REG(num_ref_frames, 4, 0, 0x1f) + +HEVC_DEC_REG(scaling_list_e, 5, 24, 0x1) +HEVC_DEC_REG(cb_qp_offset, 5, 19, 0x1f) +HEVC_DEC_REG(cr_qp_offset, 5, 14, 0x1f) +HEVC_DEC_REG(sign_data_hide, 5, 12, 0x1) +HEVC_DEC_REG(tempor_mvp_e, 5, 11, 0x1) +HEVC_DEC_REG(max_cu_qpd_depth, 5, 5, 0x3f) +HEVC_DEC_REG(cu_qpd_e, 5, 4, 0x1) + +HEVC_DEC_REG(stream_len, 6, 0, 0xffffffff) + +HEVC_DEC_REG(cabac_init_present, 7, 31, 0x1) +HEVC_DEC_REG(weight_pred_e, 7, 28, 0x1) +HEVC_DEC_REG(weight_bipr_idc, 7, 26, 0x3) +HEVC_DEC_REG(filter_over_slices, 7, 25, 0x1) +HEVC_DEC_REG(filter_over_tiles, 7, 24, 0x1) +HEVC_DEC_REG(asym_pred_e, 7, 23, 0x1) +HEVC_DEC_REG(sao_e, 7, 22, 0x1) +HEVC_DEC_REG(pcm_filt_d, 7, 21, 0x1) +HEVC_DEC_REG(slice_chqp_present, 7, 20, 0x1) +HEVC_DEC_REG(dependent_slice, 7, 19, 0x1) +HEVC_DEC_REG(filter_override, 7, 18, 0x1) +HEVC_DEC_REG(strong_smooth_e, 7, 17, 0x1) +HEVC_DEC_REG(filt_offset_beta, 7, 12, 0x1f) +HEVC_DEC_REG(filt_offset_tc, 7, 7, 0x1f) +HEVC_DEC_REG(slice_hdr_ext_e, 7, 6, 0x1) +HEVC_DEC_REG(slice_hdr_ext_bits, 7, 3, 0x7) + +HEVC_DEC_REG(const_intra_e, 8, 31, 0x1) +HEVC_DEC_REG(filt_ctrl_pres, 8, 30, 0x1) +HEVC_DEC_REG(idr_pic_e, 8, 16, 0x1) +HEVC_DEC_REG(bit_depth_pcm_y, 8, 12, 0xf) +HEVC_DEC_REG(bit_depth_pcm_c, 8, 8, 0xf) +HEVC_DEC_REG(bit_depth_y_minus8, 8, 6, 0x3) +HEVC_DEC_REG(bit_depth_c_minus8, 8, 4, 0x3) +HEVC_DEC_REG(output_8_bits, 8, 3, 0x1) + +HEVC_DEC_REG(refidx1_active, 9, 19, 0x1f) +HEVC_DEC_REG(refidx0_active, 9, 14, 0x1f) +HEVC_DEC_REG(hdr_skip_length, 9, 0, 0x3fff) + +HEVC_DEC_REG(start_code_e, 10, 31, 0x1) +HEVC_DEC_REG(init_qp, 10, 24, 0x3f) +HEVC_DEC_REG(num_tile_cols, 10, 19, 0x1f) +HEVC_DEC_REG(num_tile_rows, 10, 14, 0x1f) +HEVC_DEC_REG(tile_e, 10, 1, 0x1) +HEVC_DEC_REG(entropy_sync_e, 10, 0, 0x1) + +HEVC_DEC_REG(refer_lterm_e, 12, 16, 0xffff) +HEVC_DEC_REG(min_cb_size, 12, 13, 0x7) +HEVC_DEC_REG(max_cb_size, 12, 10, 0x7) +HEVC_DEC_REG(min_pcm_size, 12, 7, 0x7) +HEVC_DEC_REG(max_pcm_size, 12, 4, 0x7) +HEVC_DEC_REG(pcm_e, 12, 3, 0x1) +HEVC_DEC_REG(transform_skip, 12, 2, 0x1) +HEVC_DEC_REG(transq_bypass, 12, 1, 0x1) +HEVC_DEC_REG(list_mod_e, 12, 0, 0x1) + +HEVC_DEC_REG(min_trb_size, 13, 13, 0x7) +HEVC_DEC_REG(max_trb_size, 13, 10, 0x7) +HEVC_DEC_REG(max_intra_hierdepth, 13, 7, 0x7) +HEVC_DEC_REG(max_inter_hierdepth, 13, 4, 0x7) +HEVC_DEC_REG(parallel_merge, 13, 0, 0xf) + +HEVC_DEC_REG(rlist_f0, 14, 0, 0x1f) +HEVC_DEC_REG(rlist_f1, 14, 10, 0x1f) +HEVC_DEC_REG(rlist_f2, 14, 20, 0x1f) +HEVC_DEC_REG(rlist_b0, 14, 5, 0x1f) +HEVC_DEC_REG(rlist_b1, 14, 15, 0x1f) +HEVC_DEC_REG(rlist_b2, 14, 25, 0x1f) + +HEVC_DEC_REG(rlist_f3, 15, 0, 0x1f) +HEVC_DEC_REG(rlist_f4, 15, 10, 0x1f) +HEVC_DEC_REG(rlist_f5, 15, 20, 0x1f) +HEVC_DEC_REG(rlist_b3, 15, 5, 0x1f) +HEVC_DEC_REG(rlist_b4, 15, 15, 0x1f) +HEVC_DEC_REG(rlist_b5, 15, 25, 0x1f) + +HEVC_DEC_REG(rlist_f6, 16, 0, 0x1f) +HEVC_DEC_REG(rlist_f7, 16, 10, 0x1f) +HEVC_DEC_REG(rlist_f8, 16, 20, 0x1f) +HEVC_DEC_REG(rlist_b6, 16, 5, 0x1f) +HEVC_DEC_REG(rlist_b7, 16, 15, 0x1f) +HEVC_DEC_REG(rlist_b8, 16, 25, 0x1f) + +HEVC_DEC_REG(rlist_f9, 17, 0, 0x1f) +HEVC_DEC_REG(rlist_f10, 17, 10, 0x1f) +HEVC_DEC_REG(rlist_f11, 17, 20, 0x1f) +HEVC_DEC_REG(rlist_b9, 17, 5, 0x1f) +HEVC_DEC_REG(rlist_b10, 17, 15, 0x1f) +HEVC_DEC_REG(rlist_b11, 17, 25, 0x1f) + +HEVC_DEC_REG(rlist_f12, 18, 0, 0x1f) +HEVC_DEC_REG(rlist_f13, 18, 10, 0x1f) +HEVC_DEC_REG(rlist_f14, 18, 20, 0x1f) +HEVC_DEC_REG(rlist_b12, 18, 5, 0x1f) +HEVC_DEC_REG(rlist_b13, 18, 15, 0x1f) +HEVC_DEC_REG(rlist_b14, 18, 25, 0x1f) + +HEVC_DEC_REG(rlist_f15, 19, 0, 0x1f) +HEVC_DEC_REG(rlist_b15, 19, 5, 0x1f) + +HEVC_DEC_REG(partial_ctb_x, 20, 31, 0x1) +HEVC_DEC_REG(partial_ctb_y, 20, 30, 0x1) +HEVC_DEC_REG(pic_width_4x4, 20, 16, 0xfff) +HEVC_DEC_REG(pic_height_4x4, 20, 0, 0xfff) + +HEVC_DEC_REG(cur_poc_00, 46, 24, 0xff) +HEVC_DEC_REG(cur_poc_01, 46, 16, 0xff) +HEVC_DEC_REG(cur_poc_02, 46, 8, 0xff) +HEVC_DEC_REG(cur_poc_03, 46, 0, 0xff) + +HEVC_DEC_REG(cur_poc_04, 47, 24, 0xff) +HEVC_DEC_REG(cur_poc_05, 47, 16, 0xff) +HEVC_DEC_REG(cur_poc_06, 47, 8, 0xff) +HEVC_DEC_REG(cur_poc_07, 47, 0, 0xff) + +HEVC_DEC_REG(cur_poc_08, 48, 24, 0xff) +HEVC_DEC_REG(cur_poc_09, 48, 16, 0xff) +HEVC_DEC_REG(cur_poc_10, 48, 8, 0xff) +HEVC_DEC_REG(cur_poc_11, 48, 0, 0xff) + +HEVC_DEC_REG(cur_poc_12, 49, 24, 0xff) +HEVC_DEC_REG(cur_poc_13, 49, 16, 0xff) +HEVC_DEC_REG(cur_poc_14, 49, 8, 0xff) +HEVC_DEC_REG(cur_poc_15, 49, 0, 0xff) + +HEVC_DEC_REG(apf_threshold, 55, 0, 0xffff) + +HEVC_DEC_REG(clk_gate_e, 58, 16, 0x1) +HEVC_DEC_REG(buswidth, 58, 8, 0x7) +HEVC_DEC_REG(max_burst, 58, 0, 0xff) + +#define HEVC_REG_CONFIG G2_SWREG(58) +#define HEVC_REG_CONFIG_DEC_CLK_GATE_E BIT(16) +#define HEVC_REG_CONFIG_DEC_CLK_GATE_IDLE_E BIT(17) + +#define HEVC_ADDR_DST (G2_SWREG(65)) +#define HEVC_REG_ADDR_REF(i) (G2_SWREG(67) + ((i) * 0x8)) +#define HEVC_ADDR_DST_CHR (G2_SWREG(99)) +#define HEVC_REG_CHR_REF(i) (G2_SWREG(101) + ((i) * 0x8)) +#define HEVC_ADDR_DST_MV (G2_SWREG(133)) +#define HEVC_REG_DMV_REF(i) (G2_SWREG(135) + ((i) * 0x8)) +#define HEVC_ADDR_TILE_SIZE (G2_SWREG(167)) +#define HEVC_ADDR_STR (G2_SWREG(169)) +#define HEVC_SCALING_LIST (G2_SWREG(171)) +#define HEVC_RASTER_SCAN (G2_SWREG(175)) +#define HEVC_RASTER_SCAN_CHR (G2_SWREG(177)) +#define HEVC_TILE_FILTER (G2_SWREG(179)) +#define HEVC_TILE_SAO (G2_SWREG(181)) +#define HEVC_TILE_BSD (G2_SWREG(183)) + +HEVC_DEC_REG(strm_buffer_len, 258, 0, 0xffffffff) +HEVC_DEC_REG(strm_start_offset, 259, 0, 0xffffffff) + +#endif diff --git a/drivers/staging/media/hantro/hantro_hevc.c b/drivers/staging/media/hantro/hantro_hevc.c new file mode 100644 index 000000000000..8e319a837ff3 --- /dev/null +++ b/drivers/staging/media/hantro/hantro_hevc.c @@ -0,0 +1,321 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Hantro VPU HEVC codec driver + * + * Copyright (C) 2020 Safran Passenger Innovations LLC + */ + +#include <linux/types.h> +#include <media/v4l2-mem2mem.h> + +#include "hantro.h" +#include "hantro_hw.h" + +#define VERT_FILTER_RAM_SIZE 8 /* bytes per pixel row */ +/* + * BSD control data of current picture at tile border + * 128 bits per 4x4 tile = 128/(8*4) bytes per row + */ +#define BSD_CTRL_RAM_SIZE 4 /* bytes per pixel row */ +/* tile border coefficients of filter */ +#define VERT_SAO_RAM_SIZE 48 /* bytes per pixel */ + +#define MAX_TILE_COLS 20 +#define MAX_TILE_ROWS 22 + +#define UNUSED_REF -1 + +#define G2_ALIGN 16 +#define MC_WORD_SIZE 32 + +size_t hantro_hevc_chroma_offset(const struct v4l2_ctrl_hevc_sps *sps) +{ + int bytes_per_pixel = sps->bit_depth_luma_minus8 == 0 ? 1 : 2; + + return sps->pic_width_in_luma_samples * + sps->pic_height_in_luma_samples * bytes_per_pixel; +} + +size_t hantro_hevc_motion_vectors_offset(const struct v4l2_ctrl_hevc_sps *sps) +{ + size_t cr_offset = hantro_hevc_chroma_offset(sps); + + return ALIGN((cr_offset * 3) / 2, G2_ALIGN) + MC_WORD_SIZE; +} + +static size_t hantro_hevc_mv_size(const struct v4l2_ctrl_hevc_sps *sps) +{ + u32 pic_width_in_ctb64 = (sps->pic_width_in_luma_samples + (1 << 8) - 1) >> 8; + u32 pic_height_in_ctb64 = (sps->pic_height_in_luma_samples + (1 << 8) - 1) >> 8; + size_t mv_size; + + mv_size = (pic_width_in_ctb64 * pic_height_in_ctb64 * + (1 << (2 * (8 - 4))) * 16) + 32; + + vpu_debug(4, "%dx%d (CTBs) %lu MV bytes\n", + pic_width_in_ctb64, pic_height_in_ctb64, mv_size); + + return mv_size; +} + +static size_t hantro_hevc_ref_size(struct hantro_ctx *ctx) +{ + const struct hantro_hevc_dec_ctrls *ctrls = &ctx->hevc_dec.ctrls; + const struct v4l2_ctrl_hevc_sps *sps = ctrls->sps; + + return hantro_hevc_motion_vectors_offset(sps) + hantro_hevc_mv_size(sps); +} + +static void hantro_hevc_ref_free(struct hantro_ctx *ctx) +{ + struct hantro_hevc_dec_hw_ctx *hevc_dec = &ctx->hevc_dec; + struct hantro_dev *vpu = ctx->dev; + int i; + + /* Just tag buffer as unused, do not free them */This comment seems wrong.
You are right I will remove it.
+ for (i = 0; i < NUM_REF_PICTURES; i++) { + if (hevc_dec->ref_bufs[i].cpu) { + memset(hevc_dec->ref_bufs[i].cpu, 0, hantro_hevc_ref_size(ctx));Is this memset clearing the buffer required? If we're getting artifacts from previous decodes, then that would be more of a bug somewhere.
Clear is done after allocating/reused the buffer I can remove this one.
+ dma_free_coherent(vpu->dev, hevc_dec->ref_bufs[i].size, + hevc_dec->ref_bufs[i].cpu, + hevc_dec->ref_bufs[i].dma); + } + } +} + +static void hantro_hevc_ref_init(struct hantro_ctx *ctx) +{ + struct hantro_hevc_dec_hw_ctx *hevc_dec = &ctx->hevc_dec; + int i; + + for (i = 0; i < NUM_REF_PICTURES; i++) + hevc_dec->ref_bufs_poc[i] = UNUSED_REF; +} + +dma_addr_t hantro_hevc_get_ref_buf(struct hantro_ctx *ctx, + int poc) +{ + struct hantro_hevc_dec_hw_ctx *hevc_dec = &ctx->hevc_dec; + int i; + + /* Find the reference buffer in already know ones */ + for (i = 0; i < NUM_REF_PICTURES; i++) { + if (hevc_dec->ref_bufs_poc[i] == poc) { + hevc_dec->ref_bufs_used |= 1 << i; + return hevc_dec->ref_bufs[i].dma; + } + } + + /* Allocate a new reference buffer */ + for (i = 0; i < NUM_REF_PICTURES; i++) { + if (hevc_dec->ref_bufs_poc[i] == UNUSED_REF) { + if (!hevc_dec->ref_bufs[i].cpu) { + struct hantro_dev *vpu = ctx->dev; + + hevc_dec->ref_bufs[i].cpu = + dma_alloc_coherent(vpu->dev, + hantro_hevc_ref_size(ctx), + &hevc_dec->ref_bufs[i].dma, + GFP_KERNEL);Is there any reason why we need to allocate reference buffers and MV contiguously?
It is done like that in IMX reference code and makes the management of reference frame and MV more simple.
+ if (!hevc_dec->ref_bufs[i].cpu) + return 0; + + hevc_dec->ref_bufs[i].size = hantro_hevc_ref_size(ctx); + } + hevc_dec->ref_bufs_used |= 1 << i; + memset(hevc_dec->ref_bufs[i].cpu, 0, hantro_hevc_ref_size(ctx));I believe the coherent allocation is to be able to clear each reference, but is this really needed? I recall maybe only the MV buffer needs clearing, maybe you can try that? Also, if that's the case, then allocating the MV buffer separatedly will allow to not allocate the reference buffers coherently (note that we use NO_MAPPING in the vb2_queue, so the vb2_buffers shouldn't be coherent).
That sound like good possible optimizations but I'm not at this stage. I would rather keep it in this fairly functional state and improve it later. I think the patches are already enough larges and complexes like that. Benjamin
Thanks, Ezequiel
