diff --git a/driver/variant/iris2/inc/hfi_buffer_iris2.h b/driver/variant/iris2/inc/hfi_buffer_iris2.h index eabf036ffd..858bb75a0f 100644 --- a/driver/variant/iris2/inc/hfi_buffer_iris2.h +++ b/driver/variant/iris2/inc/hfi_buffer_iris2.h @@ -3,11 +3,1798 @@ * Copyright (c) 2020, The Linux Foundation. All rights reserved. */ +#ifndef __HFI_BUFFER_IRIS2__ +#define __HFI_BUFFER_IRIS2__ -#define HFI_BUFFER_ARP_ENC (204800) -#define HFI_BUFFER_BIN_ENC (15 * 1000 * 1000) -#define HFI_BUFFER_LINE_ENC (200 * 1000) -#define HFI_BUFFER_COMV_ENC (1 * 1000 * 1000) -#define HFI_BUFFER_NON_COMV_ENC (500 * 1000) -#define HFI_BUFFER_DPB_ENC (3 * 1000 * 1000) -#define HFI_BUFFER_VPSS_ENC (2 * 1000 * 1000) +#ifndef MIN +#define MIN(x, y) (((x) < (y)) ? (x) : (y)) +#endif + +#ifndef MAX +#define MAX(x, y) (((x) > (y)) ? (x) : (y)) +#endif + +/* + * Default buffer size alignment value + */ +#define HFI_ALIGNMENT_4096 (4096) + +/* + * Macro to align a to b + */ +#define HFI_ALIGN(a, b) (((b) & ((b) - 1)) ? (((a) + (b) - 1) / \ + (b) * (b)) : (((a) + (b) - 1) & (~((b) - 1)))) + +/* + * Default ubwc metadata buffer stride and height alignment values + */ +#define HFI_DEFAULT_METADATA_STRIDE_MULTIPLE (64) +#define HFI_DEFAULT_METADATA_BUFFERHEIGHT_MULTIPLE (16) + +/* + * Level 2 Comment: "Default Parameters for Firmware " + * This section defines all the default constants used by Firmware + * for any encoding session: + * 1. Bitstream Restriction VUI: TRUE + * 2. Picture Order Count: 2 (for all profiles except B frame case) + * 3. Constrained intra pred flag : TRUE (if Intra-refresh (IR) is enabled) + */ + +/* + * Level 2 Comment: "Tile dimensions(in pixels) macros for + * different color formats" + * @datatypes + * @sa + */ +#define HFI_COLOR_FORMAT_YUV420_NV12_UBWC_Y_TILE_HEIGHT (8) +#define HFI_COLOR_FORMAT_YUV420_NV12_UBWC_Y_TILE_WIDTH (32) +#define HFI_COLOR_FORMAT_YUV420_NV12_UBWC_UV_TILE_HEIGHT (8) +#define HFI_COLOR_FORMAT_YUV420_NV12_UBWC_UV_TILE_WIDTH (16) +#define HFI_COLOR_FORMAT_YUV420_TP10_UBWC_Y_TILE_HEIGHT (4) +#define HFI_COLOR_FORMAT_YUV420_TP10_UBWC_Y_TILE_WIDTH (48) +#define HFI_COLOR_FORMAT_YUV420_TP10_UBWC_UV_TILE_HEIGHT (4) +#define HFI_COLOR_FORMAT_YUV420_TP10_UBWC_UV_TILE_WIDTH (24) +#define HFI_COLOR_FORMAT_RGBA8888_UBWC_TILE_HEIGHT (4) +#define HFI_COLOR_FORMAT_RGBA8888_UBWC_TILE_WIDTH (16) + +/* + * Level 2 Comment: "Macros to calculate YUV strides and size" + * @datatypes + * HFI_UNCOMPRESSED_FORMAT_SUPPORTED_TYPE + * HFI_UNCOMPRESSED_PLANE_INFO_TYPE + * HFI_UNCOMPRESSED_PLANE_CONSTRAINTS_TYPE + * @sa + * HFI_PROPERTY_PARAM_UNCOMPRESSED_FORMAT_SUPPORTED + */ + +/* + * Luma stride calculation for YUV420, NV12/NV21, NV12_UBWC color format + * Stride arrived at here is the minimum required stride. Host may + * set a stride higher than the one calculated here, till the stride + * is a multiple of "nStrideMultiples" in + * HFI_UNCOMPRESSED_PLANE_CONSTRAINTS_TYPE + */ +#define HFI_NV12_IL_CALC_Y_STRIDE(stride, frame_width, stride_multiple) \ + stride = HFI_ALIGN(frame_width, stride_multiple) + +/* + * Luma plane height calculation for YUV420 NV12/NV21, NV12_UBWC color format + * Luma plane height used by the host needs to be either equal + * to higher than the value calculated here + */ +#define HFI_NV12_IL_CALC_Y_BUFHEIGHT(buf_height, frame_height, \ + min_buf_height_multiple) buf_height = HFI_ALIGN(frame_height, \ + min_buf_height_multiple) + +/* + * Chroma stride calculation for NV12/NV21, NV12_UBWC color format + */ +#define HFI_NV12_IL_CALC_UV_STRIDE(stride, frame_width, stride_multiple) \ + stride = HFI_ALIGN(frame_width, stride_multiple) + +/* + * Chroma plane height calculation for NV12/NV21, NV12_UBWC color format + */ +#define HFI_NV12_IL_CALC_UV_BUFHEIGHT(buf_height, frame_height, \ + min_buf_height_multiple) buf_height = HFI_ALIGN(((frame_height + 1) \ + >> 1), min_buf_height_multiple) + +/* + * Minimum buffer size that needs to be allocated for current + * frame dimensions for NV12/N21 Linear format + * (calcualtion includes both luma and chroma plane) + */ +#define HFI_NV12_IL_CALC_BUF_SIZE(buf_size, y_bufSize, y_stride, y_buf_height, \ + uv_buf_size, uv_stride, uv_buf_height) \ + y_bufSize = (y_stride * y_buf_height); \ + uv_buf_size = (uv_stride * uv_buf_height); \ + buf_size = HFI_ALIGN(y_bufSize + uv_buf_size, HFI_ALIGNMENT_4096) + +/* + * Minimum Luma buffer size that needs to be allocated for current + * frame dimensions NV12_UBWC format + */ +#define HFI_NV12_UBWC_IL_CALC_Y_BUF_SIZE(y_bufSize, y_stride, y_buf_height) \ + y_bufSize = HFI_ALIGN(y_stride * y_buf_height, HFI_ALIGNMENT_4096) + +/* + * Minimum chroma buffer size that needs to be allocated for current + * frame dimensions NV12_UBWC format + */ +#define HFI_NV12_UBWC_IL_CALC_UV_BUF_SIZE(uv_buf_size, \ + uv_stride, uv_buf_height) \ + uv_buf_size = HFI_ALIGN(uv_stride * uv_buf_height, HFI_ALIGNMENT_4096) + +/* + * This is for sdm845 onwards: + * Ver2.0: Minimum buffer size that needs to be allocated for current + * frame dimensions NV12_UBWC format (including interlace UBWC) + * (calculation includes all data & metadata planes) + */ +#define HFI_NV12_UBWC_IL_CALC_BUF_SIZE_V2 \ + (buf_size, frame_width, frame_height, y_stride_multiple,\ + y_buffer_height_multiple, uv_stride_multiple, \ + uv_buffer_height_multiple, y_metadata_stride_multiple, \ + y_metadata_buffer_height_multiple, \ + uv_metadata_stride_multiple, uv_metadata_buffer_height_multiple) \ + do \ + { \ + UInt32 y_buf_size, uv_buf_size, y_meta_size, uv_meta_size; \ + UInt32 stride, height; \ + UInt32 halfHeight = (frame_height + 1) >> 1; \ + HFI_NV12_IL_CALC_Y_STRIDE(stride, frame_width,\ + y_stride_multiple); \ + HFI_NV12_IL_CALC_Y_BUFHEIGHT(height, half_height,\ + y_buffer_height_multiple); \ + HFI_NV12_UBWC_IL_CALC_Y_BUF_SIZE(y_buf_size, stride, height);\ + HFI_NV12_IL_CALC_UV_STRIDE(stride, frame_width, \ + uv_stride_multiple); \ + HFI_NV12_IL_CALC_UV_BUFHEIGHT(height, half_height, \ + uv_buffer_height_multiple); \ + HFI_NV12_UBWC_IL_CALC_UV_BUF_SIZE(uv_buf_size, stride, height);\ + HFI_UBWC_CALC_METADATA_PLANE_STRIDE(stride, frame_width,\ + y_metadata_stride_multiple, \ + HFI_COLOR_FORMAT_YUV420_NV12_UBWC_Y_TILE_WIDTH);\ + HFI_UBWC_METADATA_PLANE_BUFHEIGHT(height, half_height, \ + y_metadata_buffer_height_multiple,\ + HFI_COLOR_FORMAT_YUV420_NV12_UBWC_Y_TILE_HEIGHT);\ + HFI_UBWC_METADATA_PLANE_BUFFER_SIZE(y_meta_size, stride, \ + height); \ + HFI_UBWC_UV_METADATA_PLANE_STRIDE(stride, frame_width,\ + uv_metadata_stride_multiple, \ + HFI_COLOR_FORMAT_YUV420_NV12_UBWC_UV_TILE_WIDTH); \ + HFI_UBWC_UV_METADATA_PLANE_BUFHEIGHT(height, half_height,\ + uv_metadata_buffer_height_multiple,\ + HFI_COLOR_FORMAT_YUV420_NV12_UBWC_UV_TILE_HEIGHT);\ + HFI_UBWC_METADATA_PLANE_BUFFER_SIZE(uv_meta_size, stride, \ + height); \ + bufSize = (y_buf_size + uv_buf_size + y_meta_size + \ + uv_meta_size) << 1;\ + } while (0) + +/* + * Luma stride calculation for YUV420_TP10 color format + * Stride arrived at here is the minimum required stride. Host may + * set a stride higher than the one calculated here, till the stride + * is a multiple of "nStrideMultiples" in + * HFI_UNCOMPRESSED_PLANE_CONSTRAINTS_TYPE + */ +#define HFI_YUV420_TP10_CALC_Y_STRIDE(stride, frame_width, stride_multiple) \ + stride = HFI_ALIGN(frame_width, 192); \ + stride = HFI_ALIGN(stride * 4 / 3, stride_multiple) + +/* + * Luma plane height calculation for YUV420_TP10 linear & UBWC color format + * Luma plane height used by the host needs to be either equal + * to higher than the value calculated here + */ +#define HFI_YUV420_TP10_CALC_Y_BUFHEIGHT(buf_height, frame_height, \ + min_buf_height_multiple) \ + buf_height = HFI_ALIGN(frame_height, min_buf_height_multiple) + +/* + * Chroma stride calculation for YUV420_TP10 linear & UBWC color format + */ +#define HFI_YUV420_TP10_CALC_UV_STRIDE(stride, frame_width, stride_multiple) \ + stride = HFI_ALIGN(frame_width, 192); \ + stride = HFI_ALIGN(stride * 4 / 3, stride_multiple) + +/* + * Chroma plane height calculation for YUV420_TP10 linear & UBWC color format + */ +#define HFI_YUV420_TP10_CALC_UV_BUFHEIGHT(buf_height, frame_height, \ + min_buf_height_multiple) \ + buf_height = HFI_ALIGN(((frame_height + 1) >> 1), \ + min_buf_height_multiple) + +/* + * Minimum buffer size that needs to be allocated for current + * frame dimensions for YUV420_TP10 linear format + * (calcualtion includes both luma and chroma plane) + */ +#define HFI_YUV420_TP10_CALC_BUF_SIZE(buf_size, y_buf_size, y_stride,\ + y_buf_height, uv_buf_size, uv_stride, uv_buf_height) \ + y_buf_size = (y_stride * y_buf_height); \ + uv_buf_size = (uv_stride * uv_buf_height); \ + buf_size = y_buf_size + uv_buf_size + +/* + * Minimum Luma data buffer size that needs to be allocated for current + * frame dimensions YUV420_TP10_UBWC format + */ +#define HFI_YUV420_TP10_UBWC_CALC_Y_BUF_SIZE(y_buf_size, y_stride, \ + y_buf_height) \ + y_buf_size = HFI_ALIGN(y_stride * y_buf_height, HFI_ALIGNMENT_4096) + +/* + * Minimum chroma data buffer size that needs to be allocated for current + * frame dimensions YUV420_TP10_UBWC format + */ +#define HFI_YUV420_TP10_UBWC_CALC_UV_BUF_SIZE(uv_buf_size, uv_stride, \ + uv_buf_height) \ + uv_buf_size = HFI_ALIGN(uv_stride * uv_buf_height, HFI_ALIGNMENT_4096) + +/* + * Minimum buffer size that needs to be allocated for current + * frame dimensions NV12_UBWC format + * (calculation includes all data & metadata planes) + */ +#define HFI_YUV420_TP10_UBWC_CALC_BUF_SIZE(buf_size, y_stride, y_buf_height, \ + uv_stride, uv_buf_height, y_md_stride, y_md_height, uv_md_stride, \ + uv_md_height)\ + do \ + { \ + HFI_U32 y_data_size, uv_data_size, y_md_size, uv_md_size; \ + HFI_YUV420_TP10_UBWC_CALC_Y_BUF_SIZE(y_data_size, y_stride,\ + y_buf_height); \ + HFI_YUV420_TP10_UBWC_CALC_UV_BUF_SIZE(uv_data_size, uv_stride, \ + uv_buf_height); \ + HFI_UBWC_METADATA_PLANE_BUFFER_SIZE(y_md_size, y_md_stride, \ + y_md_height); \ + HFI_UBWC_METADATA_PLANE_BUFFER_SIZE(uv_md_size, uv_md_stride, \ + uv_md_height); \ + buf_size = y_data_size + uv_data_size + y_md_size + \ + uv_md_size; \ + } while (0) + +/* + * Luma stride calculation for YUV420_P010 color format + * Stride arrived at here is the minimum required stride. Host may + * set a stride higher than the one calculated here, till the stride + * is a multiple of "nStrideMultiples" in + * HFI_UNCOMPRESSED_PLANE_CONSTRAINTS_TYPE + */ + +#define HFI_YUV420_P010_CALC_Y_STRIDE(stride, frame_width, stride_multiple) \ + stride = HFI_ALIGN(frame_width * 2, stride_multiple) + +/* + * Luma plane height calculation for YUV420_P010 linear color format + * Luma plane height used by the host needs to be either equal + * to higher than the value calculated here + */ +#define HFI_YUV420_P010_CALC_Y_BUFHEIGHT(buf_height, frame_height, \ + min_buf_height_multiple) \ + buf_height = HFI_ALIGN(frame_height, min_buf_height_multiple) + +/* + * Chroma stride calculation for YUV420_P010 linear color format + */ +#define HFI_YUV420_P010_CALC_UV_STRIDE(stride, frame_width, stride_multiple) \ + stride = HFI_ALIGN(frame_width * 2, stride_multiple) + +/* + * Chroma plane height calculation for YUV420_P010 linear color format + */ +#define HFI_YUV420_P010_CALC_UV_BUFHEIGHT(buf_height, frame_height, \ + min_buf_height_multiple) \ + buf_height = HFI_ALIGN(((frame_height + 1) >> 1), \ + min_buf_height_multiple) + +/* + * Minimum buffer size that needs to be allocated for current + * frame dimensions for YUV420_P010 linear format + * (calculation includes both luma and chroma plane) + */ +#define HFI_YUV420_P010_CALC_BUF_SIZE(buf_size, y_data_size, y_stride, \ + y_buf_height, uv_data_size, uv_stride, uv_buf_height) \ + do \ + { \ + y_data_size = HFI_ALIGN(y_stride * y_buf_height, \ + HFI_ALIGNMENT_4096);\ + uv_data_size = HFI_ALIGN(uv_stride * uv_buf_height, \ + HFI_ALIGNMENT_4096); \ + buf_size = y_data_size + uv_data_size; \ + } while (0) + +/* + * Plane stride calculation for RGB888/BGR888 color format + * Stride arrived at here is the minimum required stride. Host may + * set a stride higher than the one calculated here, till the stride + * is a multiple of "nStrideMultiples" in + * HFI_UNCOMPRESSED_PLANE_CONSTRAINTS_TYPE + */ +#define HFI_RGB888_CALC_STRIDE(stride, frame_width, stride_multiple) \ + stride = ((frame_width * 3) + stride_multiple - 1) & \ + (0xffffffff - (stride_multiple - 1)) + +/* + * Plane height calculation for RGB888/BGR888 color format + * Luma plane height used by the host needs to be either equal + * to higher than the value calculated here + */ +#define HFI_RGB888_CALC_BUFHEIGHT(buf_height, frame_height, \ + min_buf_height_multiple) \ + buf_height = ((frame_height + min_buf_height_multiple - 1) & \ + (0xffffffff - (min_buf_height_multiple - 1))) + +/* + * Minimum buffer size that needs to be allocated for current + * frame dimensions for RGB888/BGR888 format + */ +#define HFI_RGB888_CALC_BUF_SIZE(buf_size, stride, buf_height) \ + buf_size = ((stride) * (buf_height)) + +/* + * Plane stride calculation for RGBA8888 color format + * Stride arrived at here is the minimum required stride. Host may + * set a stride higher than the one calculated here, till the stride + * is a multiple of "nStrideMultiples" in + * HFI_UNCOMPRESSED_PLANE_CONSTRAINTS_TYPE + */ +#define HFI_RGBA8888_CALC_STRIDE(stride, frame_width, stride_multiple) \ + stride = HFI_ALIGN((frame_width << 2), stride_multiple) +/* + * Plane height calculation for RGBA8888 color format + * Luma plane height used by the host needs to be either equal + * to higher than the value calculated here + */ +#define HFI_RGBA8888_CALC_BUFHEIGHT(buf_height, frame_height, \ + min_buf_height_multiple) \ + buf_height = HFI_ALIGN(frame_height, min_buf_height_multiple) +/* + * Minimum buffer size that needs to be allocated for current + * frame dimensions for RGBA8888 format + */ +#define HFI_RGBA8888_CALC_BUF_SIZE(buf_size, stride, buf_height) \ + buf_size = (stride) * (buf_height) +/* + * Minimum buffer size that needs to be allocated for current + * frame dimensions for data plane of RGBA8888_UBWC format + */ +#define HFI_RGBA8888_UBWC_CALC_DATA_PLANE_BUF_SIZE(buf_size, stride, \ + buf_height) \ + buf_size = HFI_ALIGN((stride) * (buf_height), HFI_ALIGNMENT_4096) + +/* + * Minimum buffer size that needs to be allocated for current + * frame dimensions for of RGBA8888_UBWC format + */ +#define HFI_RGBA8888_UBWC_BUF_SIZE(buf_size, data_buf_size, \ + metadata_buffer_size, stride, buf_height, _metadata_tride, \ + _metadata_buf_height) \ + HFI_RGBA8888_UBWC_CALC_DATA_PLANE_BUF_SIZE(data_buf_size, \ + stride, buf_height); \ + HFI_UBWC_METADATA_PLANE_BUFFER_SIZE(metadata_buffer_size, \ + _metadata_tride, _metadata_buf_height); \ + buf_size = data_buf_size + metadata_buffer_size + +/* + * Metadata plane stride calculation for all UBWC color formats + * Should be used for Y metadata Plane & all single plane color + * formats. Stride arrived at here is the minimum required + * stride. Host may set a stride higher than the one calculated + * here, till the stride is a multiple of + * "_metadata_trideMultiple" in + * HFI_UNCOMPRESSED_PLANE_CONSTRAINTS_TYPE Default + * metadataDataStrideMultiple = 64 + */ +#define HFI_UBWC_CALC_METADATA_PLANE_STRIDE(metadata_stride, frame_width,\ + metadata_stride_multiple, tile_width_in_pels) \ + metadata_stride = HFI_ALIGN(((frame_width + (tile_width_in_pels - 1)) /\ + tile_width_in_pels), metadata_stride_multiple) + +/* + * Metadata plane height calculation for all UBWC color formats + * Should be used for Y metadata Plane & all single plane color + * formats. Plane height used by the host needs to be either + * equal to higher than the value calculated here + * Default metadataHeightMultiple = 16 + */ +#define HFI_UBWC_METADATA_PLANE_BUFHEIGHT(metadata_buf_height, frame_height, \ + metadata_height_multiple, tile_height_in_pels) \ + metadata_buf_height = HFI_ALIGN(((frame_height + \ + (tile_height_in_pels - 1)) / tile_height_in_pels), \ + metadata_height_multiple) + +/* + * UV Metadata plane stride calculation for NV12_UBWC color + * format. Stride arrived at here is the minimum required + * stride. Host may set a stride higher than the one calculated + * here, till the stride is a multiple of + * "_metadata_trideMultiple" in + * HFI_UNCOMPRESSED_PLANE_CONSTRAINTS_TYPE Default + * metadataDataStrideMultiple = 64 + */ +#define HFI_UBWC_UV_METADATA_PLANE_STRIDE(metadata_stride, frame_width, \ + metadata_stride_multiple, tile_width_in_pels) \ + metadata_stride = HFI_ALIGN(((((frame_width + 1) >> 1) +\ + (tile_width_in_pels - 1)) / tile_width_in_pels), \ + metadata_stride_multiple) + +/* + * UV Metadata plane height calculation for NV12_UBWC color + * format. Plane height used by the host needs to be either + * equal to higher than the value calculated here Default + * metadata_height_multiple = 16 + */ +#define HFI_UBWC_UV_METADATA_PLANE_BUFHEIGHT(metadata_buf_height, frame_height,\ + metadata_height_multiple, tile_height_in_pels) \ + metadata_buf_height = HFI_ALIGN(((((frame_height + 1) >> 1) + \ + (tile_height_in_pels - 1)) / tile_height_in_pels), \ + metadata_height_multiple) + +/* + * Minimum metadata buffer size that needs to be allocated for + * current frame dimensions for each metadata plane. + * This macro applies to all UBWC color format + */ +#define HFI_UBWC_METADATA_PLANE_BUFFER_SIZE(buffer_size, _metadata_tride, \ + _metadata_buf_height) \ + buffer_size = HFI_ALIGN(_metadata_tride * _metadata_buf_height, \ + HFI_ALIGNMENT_4096) + + + +#define BUFFER_ALIGNMENT_512_BYTES 512 +#define BUFFER_ALIGNMENT_256_BYTES 256 +#define BUFFER_ALIGNMENT_128_BYTES 128 +#define BUFFER_ALIGNMENT_64_BYTES 64 +#define BUFFER_ALIGNMENT_32_BYTES 32 +#define BUFFER_ALIGNMENT_16_BYTES 16 +#define BUFFER_ALIGNMENT_8_BYTES 8 +#define BUFFER_ALIGNMENT_4_BYTES 4 + +#define VENUS_DMA_ALIGNMENT BUFFER_ALIGNMENT_256_BYTES + +#define MAX_FE_NBR_CTRL_LCU64_LINE_BUFFER_SIZE 64 +#define MAX_FE_NBR_CTRL_LCU32_LINE_BUFFER_SIZE 64 +#define MAX_FE_NBR_CTRL_LCU16_LINE_BUFFER_SIZE 64 +#define MAX_FE_NBR_DATA_LUMA_LINE_BUFFER_SIZE 640 +#define MAX_FE_NBR_DATA_CB_LINE_BUFFER_SIZE 320 +#define MAX_FE_NBR_DATA_CR_LINE_BUFFER_SIZE 320 + +#define MAX_SE_NBR_CTRL_LCU64_LINE_BUFFER_SIZE (128 / 8) +#define MAX_SE_NBR_CTRL_LCU32_LINE_BUFFER_SIZE (128 / 8) +#define MAX_SE_NBR_CTRL_LCU16_LINE_BUFFER_SIZE (128 / 8) + +#define MAX_PE_NBR_DATA_LCU64_LINE_BUFFER_SIZE (64 * 2 * 3) +#define MAX_PE_NBR_DATA_LCU32_LINE_BUFFER_SIZE (32 * 2 * 3) +#define MAX_PE_NBR_DATA_LCU16_LINE_BUFFER_SIZE (16 * 2 * 3) + +/* Begin of IRIS2 */ + +/* + * VPSS internal buffer definition + * Only for 1:1 DS ratio case + */ +#define MAX_TILE_COLUMNS 32 /* 8K/256 */ + +/* + * For all buffer size calculators using num_vpp_pipes, + * use per chipset "static" pipe count. + * Note that this applies to all use-cases, + * e.g. buffer sizes for interlace decode + * will be calculated using 4 vpp pipes on Kona, + * even though interlace decode uses single vpp pipe. + * __________________________________________________________ + * |_____________Target_____________|_______numVPPpipes_______| + * | IRIS2(e.g. Kona) | 4 | + * | IRIS2(e.g. Cedros) | 2 | + * |_______IRIS2(e.g. Bitra)________|___________1_____________| + */ + +#define SIZE_VPSS_LB(Size, frame_width, frame_height, num_vpp_pipes) \ + do \ + { \ + HFI_U32 vpss_4tap_top_buffer_size, vpss_div2_top_buffer_size, \ + vpss_4tap_left_buffer_size, vpss_div2_left_buffer_size; \ + HFI_U32 opb_wr_top_line_luma_buffer_size, \ + opb_wr_top_line_chroma_buffer_size, \ + opb_lb_wr_llb_y_buffer_size,\ + opb_lb_wr_llb_uv_buffer_size; \ + HFI_U32 lcu_size_pels, macrotiling_size, padding_size; \ + vpss_4tap_top_buffer_size = vpss_div2_top_buffer_size = \ + vpss_4tap_left_buffer_size = vpss_div2_left_buffer_size = 0; \ + macrotiling_size = 32; \ + opb_wr_top_line_luma_buffer_size = HFI_ALIGN(frame_width, \ + macrotiling_size) / macrotiling_size * 256; \ + opb_wr_top_line_luma_buffer_size = \ + HFI_ALIGN(opb_wr_top_line_luma_buffer_size, \ + VENUS_DMA_ALIGNMENT) + (MAX_TILE_COLUMNS - 1) * 256; \ + opb_wr_top_line_luma_buffer_size = \ + MAX(opb_wr_top_line_luma_buffer_size, (32 * \ + HFI_ALIGN(frame_height, 8))); \ + opb_wr_top_line_chroma_buffer_size = \ + opb_wr_top_line_luma_buffer_size;\ + opb_lb_wr_llb_uv_buffer_size = opb_lb_wr_llb_y_buffer_size = \ + HFI_ALIGN((HFI_ALIGN(frame_height, 8) / (4 / 2)) * 64,\ + BUFFER_ALIGNMENT_32_BYTES); \ + Size = num_vpp_pipes * 2 * (vpss_4tap_top_buffer_size + \ + vpss_div2_top_buffer_size) + \ + 2 * (vpss_4tap_left_buffer_size + \ + vpss_div2_left_buffer_size) + \ + opb_wr_top_line_luma_buffer_size + \ + opb_wr_top_line_chroma_buffer_size + \ + opb_lb_wr_llb_uv_buffer_size + \ + opb_lb_wr_llb_y_buffer_size; \ + } while (0) + +/* + * H264d internal buffer definition + */ +#define VPP_CMD_MAX_SIZE (1 << 20) +#define NUM_HW_PIC_BUF 32 +#define BIN_BUFFER_THRESHOLD (1280 * 736) +#define H264D_MAX_SLICE 1800 +#define SIZE_H264D_BUFTAB_T (256) /* sizeof(h264d_buftab_t) aligned to 256 */ +#define SIZE_H264D_HW_PIC_T (1 << 11) /* sizeof(h264d_hw_pic_t) 32 aligned */ +#define SIZE_H264D_BSE_CMD_PER_BUF (32 * 4) +#define SIZE_H264D_VPP_CMD_PER_BUF (512) + +/* Line Buffer definitions */ +/* one for luma and 1/2 for each chroma */ +#define SIZE_H264D_LB_FE_TOP_DATA(frame_width, frame_height) \ + (MAX_FE_NBR_DATA_LUMA_LINE_BUFFER_SIZE * HFI_ALIGN(frame_width, 16) * 3) + +#define SIZE_H264D_LB_FE_TOP_CTRL(frame_width, frame_height) \ + (MAX_FE_NBR_CTRL_LCU64_LINE_BUFFER_SIZE * ((frame_width + 15) >> 4)) + +#define SIZE_H264D_LB_FE_LEFT_CTRL(frame_width, frame_height) \ + (MAX_FE_NBR_CTRL_LCU64_LINE_BUFFER_SIZE * ((frame_height + 15) >> 4)) + +#define SIZE_H264D_LB_SE_TOP_CTRL(frame_width, frame_height) \ + (MAX_SE_NBR_CTRL_LCU64_LINE_BUFFER_SIZE * ((frame_width + 15) >> 4)) + +#define SIZE_H264D_LB_SE_LEFT_CTRL(frame_width, frame_height) \ + (MAX_SE_NBR_CTRL_LCU64_LINE_BUFFER_SIZE * ((frame_height + 15) >> 4)) + +#define SIZE_H264D_LB_PE_TOP_DATA(frame_width, frame_height) \ + (MAX_PE_NBR_DATA_LCU64_LINE_BUFFER_SIZE * ((frame_width + 15) >> 4)) + +#define SIZE_H264D_LB_VSP_TOP(frame_width, frame_height) \ + ((((frame_width + 15) >> 4) << 7)) + +#define SIZE_H264D_LB_RECON_DMA_METADATA_WR(frame_width, frame_height) \ + (HFI_ALIGN(frame_height, 16) * 32) + +#define SIZE_H264D_QP(frame_width, frame_height) \ + (((frame_width + 63) >> 6) * ((frame_height + 63) >> 6) * 128) + +#define SIZE_HW_PIC(size_per_buf) \ + (NUM_HW_PIC_BUF * size_per_buf) + +#define SIZE_H264D_BSE_CMD_BUF(_size, frame_width, frame_height) \ + do \ + { /* this could change alignment */ \ + UInt32 height = HFI_ALIGN(frame_height, \ + BUFFER_ALIGNMENT_32_BYTES); \ + _size = MIN((((height + 15) >> 4) * 3 * 4), H264D_MAX_SLICE) *\ + SIZE_H264D_BSE_CMD_PER_BUF; \ + } while (0) + +#define SIZE_H264D_VPP_CMD_BUF(_size, frame_width, frame_height) \ + do \ + { /* this could change alignment */ \ + UInt32 height = HFI_ALIGN(frame_height, \ + BUFFER_ALIGNMENT_32_BYTES); \ + _size = MIN((((height + 15) >> 4) * 3 * 4), H264D_MAX_SLICE) * \ + SIZE_H264D_VPP_CMD_PER_BUF; \ + if (_size > VPP_CMD_MAX_SIZE) { _size = VPP_CMD_MAX_SIZE; } \ + } while (0) + +#define HFI_BUFFER_COMV_H264D(coMV_size, frame_width, \ + frame_height, _yuv_bufcount_min) \ + do \ + { \ + HFI_U32 frame_width_in_mbs = ((frame_width + 15) >> 4); \ + HFI_U32 frame_height_in_mbs = ((frame_height + 15) >> 4); \ + HFI_U32 col_mv_aligned_width = (frame_width_in_mbs << 7); \ + HFI_U32 col_zero_aligned_width = (frame_width_in_mbs << 2); \ + HFI_U32 col_zero_size = 0, size_colloc = 0; \ + col_mv_aligned_width = HFI_ALIGN(col_mv_aligned_width, \ + BUFFER_ALIGNMENT_16_BYTES); \ + col_zero_aligned_width = HFI_ALIGN(col_zero_aligned_width, \ + BUFFER_ALIGNMENT_16_BYTES); \ + col_zero_size = col_zero_aligned_width * \ + ((frame_height_in_mbs + 1) >> 1); \ + col_zero_size = HFI_ALIGN(col_zero_size, \ + BUFFER_ALIGNMENT_64_BYTES); \ + col_zero_size <<= 1; \ + col_zero_size = HFI_ALIGN(col_zero_size, \ + BUFFER_ALIGNMENT_512_BYTES); \ + size_colloc = col_mv_aligned_width * ((frame_height_in_mbs + \ + 1) >> 1); \ + size_colloc = HFI_ALIGN(size_colloc, \ + BUFFER_ALIGNMENT_64_BYTES); \ + size_colloc <<= 1; \ + size_colloc = HFI_ALIGN(size_colloc, \ + BUFFER_ALIGNMENT_512_BYTES); \ + size_colloc += (col_zero_size + SIZE_H264D_BUFTAB_T * 2); \ + coMV_size = size_colloc * (_yuv_bufcount_min); \ + coMV_size += BUFFER_ALIGNMENT_512_BYTES; \ + } while (0) + +#define HFI_BUFFER_NON_COMV_H264D(_size, frame_width, frame_height, \ + num_vpp_pipes) \ + do \ + { \ + HFI_U32 _size_bse, _size_vpp; \ + SIZE_H264D_BSE_CMD_BUF(_size_bse, frame_width, frame_height); \ + SIZE_H264D_VPP_CMD_BUF(_size_vpp, frame_width, frame_height); \ + _size = HFI_ALIGN(_size_bse, VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(_size_vpp, VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_HW_PIC(SIZE_H264D_HW_PIC_T), \ + VENUS_DMA_ALIGNMENT)+ HFI_ALIGN(SIZE_H264D_QP\ + (frame_width, frame_height), VENUS_DMA_ALIGNMENT); \ + _size = HFI_ALIGN(_size, VENUS_DMA_ALIGNMENT); \ + } while (0) + +/* + * _yuv_bufcount_min = MAX(Min YUV Buffer count, + * (HFI_PROPERTY_PARAM_VDEC_VPP_DELAY + 1)) + */ +#define HFI_BUFFER_LINE_H264D(_size, frame_width, frame_height, \ + _yuv_bufcount_min, is_opb, num_vpp_pipes) \ + do \ + { \ + HFI_U32 vpss_lb_size = 0; \ + _size = HFI_ALIGN(SIZE_H264D_LB_FE_TOP_DATA(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_H264D_LB_FE_TOP_CTRL(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_H264D_LB_FE_LEFT_CTRL(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) * num_vpp_pipes + \ + HFI_ALIGN(SIZE_H264D_LB_SE_TOP_CTRL(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_H264D_LB_SE_LEFT_CTRL(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) * \ + num_vpp_pipes + \ + HFI_ALIGN(SIZE_H264D_LB_PE_TOP_DATA(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_H264D_LB_VSP_TOP(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_H264D_LB_RECON_DMA_METADATA_WR\ + (frame_width, frame_height), \ + VENUS_DMA_ALIGNMENT) * 2; \ + _size = HFI_ALIGN(_size, VENUS_DMA_ALIGNMENT); \ + if (is_opb) \ + { \ + SIZE_VPSS_LB(vpss_lb_size, frame_width, frame_height, \ + num_vpp_pipes); \ + } \ + _size = HFI_ALIGN((_size + vpss_lb_size), \ + VENUS_DMA_ALIGNMENT); \ + } while (0) + +#define H264_CABAC_HDR_RATIO_HD_TOT 1 +#define H264_CABAC_RES_RATIO_HD_TOT 3 +/* + * some content need more bin buffer, + * but limit buffer size for high resolution + */ +#define SIZE_H264D_HW_BIN_BUFFER(_size, frame_width, frame_height, \ + delay, num_vpp_pipes) \ + do \ + { \ + HFI_U32 size_yuv, size_bin_hdr, size_bin_res; \ + size_yuv = ((frame_width * frame_height) <= \ + BIN_BUFFER_THRESHOLD) ?\ + ((BIN_BUFFER_THRESHOLD * 3) >> 1) : \ + ((frame_width * frame_height * 3) >> 1); \ + size_bin_hdr = size_yuv * H264_CABAC_HDR_RATIO_HD_TOT; \ + size_bin_res = size_yuv * H264_CABAC_RES_RATIO_HD_TOT; \ + size_bin_hdr = size_bin_hdr * (((((HFI_U32)(delay)) & 31) /\ + 10) + 2) / 2; \ + size_bin_res = size_bin_res * (((((HFI_U32)(delay)) & 31) /\ + 10) + 2) / 2; \ + size_bin_hdr = HFI_ALIGN(size_bin_hdr / num_vpp_pipes,\ + VENUS_DMA_ALIGNMENT) * num_vpp_pipes; \ + size_bin_res = HFI_ALIGN(size_bin_res / num_vpp_pipes, \ + VENUS_DMA_ALIGNMENT) * num_vpp_pipes; \ + _size = size_bin_hdr + size_bin_res; \ + } while (0) + +#define HFI_BUFFER_BIN_H264D(_size, frame_width, frame_height, is_interlaced, \ + delay, num_vpp_pipes) \ + do \ + { \ + HFI_U32 n_aligned_w = HFI_ALIGN(frame_width, \ + BUFFER_ALIGNMENT_16_BYTES);\ + HFI_U32 n_aligned_h = HFI_ALIGN(frame_height, \ + BUFFER_ALIGNMENT_16_BYTES); \ + if (!is_interlaced) \ + { \ + SIZE_H264D_HW_BIN_BUFFER(_size, n_aligned_w, \ + n_aligned_h, delay, num_vpp_pipes); \ + } \ + else \ + { \ + _size = 0; \ + } \ + } while (0) + +#define NUM_SLIST_BUF_H264 (256 + 32) +#define SIZE_SLIST_BUF_H264 (512) +#define SIZE_SEI_USERDATA (4096) +#define HFI_BUFFER_PERSIST_H264D(_size) \ + _size = HFI_ALIGN((SIZE_SLIST_BUF_H264 * NUM_SLIST_BUF_H264 + \ + NUM_HW_PIC_BUF * SIZE_SEI_USERDATA), VENUS_DMA_ALIGNMENT) + +/* + * H265d internal buffer definition + */ +#define LCU_MAX_SIZE_PELS 64 +#define LCU_MIN_SIZE_PELS 16 + +#define H265D_MAX_SLICE 1200 +#define SIZE_H265D_HW_PIC_T SIZE_H264D_HW_PIC_T +#define SIZE_H265D_BSE_CMD_PER_BUF (16 * sizeof(HFI_U32)) +#define SIZE_H265D_VPP_CMD_PER_BUF (256) + +#define SIZE_H265D_LB_FE_TOP_DATA(frame_width, frame_height) \ + (MAX_FE_NBR_DATA_LUMA_LINE_BUFFER_SIZE * \ + (HFI_ALIGN(frame_width, 64) + 8) * 2) + +#define SIZE_H265D_LB_FE_TOP_CTRL(frame_width, frame_height) \ + (MAX_FE_NBR_CTRL_LCU64_LINE_BUFFER_SIZE * \ + (HFI_ALIGN(frame_width, LCU_MAX_SIZE_PELS) / LCU_MIN_SIZE_PELS)) + +#define SIZE_H265D_LB_FE_LEFT_CTRL(frame_width, frame_height) \ + (MAX_FE_NBR_CTRL_LCU64_LINE_BUFFER_SIZE * \ + (HFI_ALIGN(frame_height, LCU_MAX_SIZE_PELS) / LCU_MIN_SIZE_PELS)) + +#define SIZE_H265D_LB_SE_TOP_CTRL(frame_width, frame_height) \ + ((LCU_MAX_SIZE_PELS / 8 * (128 / 8)) * ((frame_width + 15) >> 4)) + +#define SIZE_H265D_LB_SE_LEFT_CTRL(frame_width, frame_height) \ + (MAX(((frame_height + 16 - 1) / 8) * \ + MAX_SE_NBR_CTRL_LCU16_LINE_BUFFER_SIZE, \ + MAX(((frame_height + 32 - 1) / 8) * \ + MAX_SE_NBR_CTRL_LCU32_LINE_BUFFER_SIZE, \ + ((frame_height + 64 - 1) / 8) * \ + MAX_SE_NBR_CTRL_LCU64_LINE_BUFFER_SIZE))) + +#define SIZE_H265D_LB_PE_TOP_DATA(frame_width, frame_height) \ + (MAX_PE_NBR_DATA_LCU64_LINE_BUFFER_SIZE * (HFI_ALIGN(frame_width, \ + LCU_MIN_SIZE_PELS) / LCU_MIN_SIZE_PELS)) + +#define SIZE_H265D_LB_VSP_TOP(frame_width, frame_height) \ + (((frame_width + 63) >> 6) * 128) + +#define SIZE_H265D_LB_VSP_LEFT(frame_width, frame_height) \ + (((frame_height + 63) >> 6) * 128) + +#define SIZE_H265D_LB_RECON_DMA_METADATA_WR(frame_width, frame_height) \ + SIZE_H264D_LB_RECON_DMA_METADATA_WR(frame_width, frame_height) + +#define SIZE_H265D_QP(frame_width, frame_height) \ + SIZE_H264D_QP(frame_width, frame_height) + +#define SIZE_H265D_BSE_CMD_BUF(_size, frame_width, frame_height)\ + do \ + { \ + _size = HFI_ALIGN(((HFI_ALIGN(frame_width, \ + LCU_MAX_SIZE_PELS) / LCU_MIN_SIZE_PELS) * \ + (HFI_ALIGN(frame_height, LCU_MAX_SIZE_PELS) /\ + LCU_MIN_SIZE_PELS)) * NUM_HW_PIC_BUF, VENUS_DMA_ALIGNMENT); \ + _size = MIN(_size, H265D_MAX_SLICE + 1); \ + _size = 2 * _size * SIZE_H265D_BSE_CMD_PER_BUF; \ + } while (0) + +#define SIZE_H265D_VPP_CMD_BUF(_size, frame_width, frame_height) \ + do \ + { \ + _size = HFI_ALIGN(((HFI_ALIGN(frame_width, LCU_MAX_SIZE_PELS) /\ + LCU_MIN_SIZE_PELS) * (HFI_ALIGN(frame_height, \ + LCU_MAX_SIZE_PELS) / LCU_MIN_SIZE_PELS)) * \ + NUM_HW_PIC_BUF, VENUS_DMA_ALIGNMENT); \ + _size = MIN(_size, H265D_MAX_SLICE + 1); \ + _size = HFI_ALIGN(_size, 4); \ + _size = 2 * _size * SIZE_H265D_VPP_CMD_PER_BUF; \ + if (_size > VPP_CMD_MAX_SIZE) \ + { \ + _size = VPP_CMD_MAX_SIZE; \ + } \ + } while (0) + +#define HFI_BUFFER_COMV_H265D(_size, frame_width, frame_height, \ + _yuv_bufcount_min) \ + do \ + { \ + _size = HFI_ALIGN(((((frame_width + 15) >> 4) * \ + ((frame_height + 15) >> 4)) << 8), \ + BUFFER_ALIGNMENT_512_BYTES); \ + _size *= _yuv_bufcount_min; \ + _size += BUFFER_ALIGNMENT_512_BYTES; \ + } while (0) + +/* c2 divide this into NON_COMV and LINE */ +#define HFI_BUFFER_NON_COMV_H265D(_size, frame_width, frame_height, \ + num_vpp_pipes) \ + do \ + { \ + HFI_U32 _size_bse, _size_vpp; \ + SIZE_H265D_BSE_CMD_BUF(_size_bse, frame_width, \ + frame_height); \ + SIZE_H265D_VPP_CMD_BUF(_size_vpp, frame_width, \ + frame_height); \ + _size = HFI_ALIGN(_size_bse, VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(_size_vpp, VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(NUM_HW_PIC_BUF * 20 * 22 * 4, \ + VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(2 * sizeof(HFI_U16) * \ + (HFI_ALIGN(frame_width, LCU_MAX_SIZE_PELS) / \ + LCU_MIN_SIZE_PELS) * (HFI_ALIGN(frame_height, \ + LCU_MAX_SIZE_PELS) / LCU_MIN_SIZE_PELS), \ + VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_HW_PIC(SIZE_H265D_HW_PIC_T), \ + VENUS_DMA_ALIGNMENT)\ + + HDR10_HIST_EXTRADATA_SIZE; \ + _size = HFI_ALIGN(_size, VENUS_DMA_ALIGNMENT); \ + } while (0) + +#define HFI_BUFFER_LINE_H265D(_size, frame_width, frame_height, \ + is_opb, num_vpp_pipes) \ + do \ + { \ + HFI_U32 vpssLB_size = 0; \ + _size = HFI_ALIGN(SIZE_H265D_LB_FE_TOP_DATA(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_H265D_LB_FE_TOP_CTRL(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_H265D_LB_FE_LEFT_CTRL(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) * num_vpp_pipes + \ + HFI_ALIGN(SIZE_H265D_LB_SE_LEFT_CTRL(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) * num_vpp_pipes + \ + HFI_ALIGN(SIZE_H265D_LB_SE_TOP_CTRL(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_H265D_LB_PE_TOP_DATA(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_H265D_LB_VSP_TOP(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_H265D_LB_VSP_LEFT(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) * num_vpp_pipes + \ + HFI_ALIGN(SIZE_H265D_LB_RECON_DMA_METADATA_WR\ + (frame_width, frame_height), \ + VENUS_DMA_ALIGNMENT) * 4 + \ + HFI_ALIGN(SIZE_H265D_QP(frame_width, frame_height),\ + VENUS_DMA_ALIGNMENT); \ + if (is_opb) \ + { \ + SIZE_VPSS_LB(vpss_lb_size, frame_width, frame_height,\ + num_vpp_pipes); \ + } \ + _size = HFI_ALIGN((_size + vpss_lb_size), \ + VENUS_DMA_ALIGNMENT); \ + } while (0) + +#define H265_CABAC_HDR_RATIO_HD_TOT 2 +#define H265_CABAC_RES_RATIO_HD_TOT 2 +/* + * some content need more bin buffer, + * but limit buffer size for high resolution + */ +#define SIZE_H265D_HW_BIN_BUFFER(_size, frame_width, frame_height, \ + delay, num_vpp_pipes) \ + do \ + { \ + HFI_U32 size_yuv, size_bin_hdr, size_bin_res; \ + size_yuv = ((frame_width * frame_height) <= \ + BIN_BUFFER_THRESHOLD) ? \ + ((BIN_BUFFER_THRESHOLD * 3) >> 1) : \ + ((frame_width * frame_height * 3) >> 1); \ + size_bin_hdr = size_yuv * H265_CABAC_HDR_RATIO_HD_TOT; \ + size_bin_res = size_yuv * H265_CABAC_RES_RATIO_HD_TOT; \ + size_bin_hdr = size_bin_hdr * \ + (((((HFI_U32)(delay)) & 31) / 10) + 2) / 2; \ + size_bin_res = size_bin_res * \ + (((((HFI_U32)(delay)) & 31) / 10) + 2) / 2; \ + size_bin_hdr = HFI_ALIGN(size_bin_hdr / \ + num_vpp_pipes, VENUS_DMA_ALIGNMENT) * \ + num_vpp_pipes; \ + size_bin_res = HFI_ALIGN(size_bin_res / num_vpp_pipes,\ + VENUS_DMA_ALIGNMENT) * num_vpp_pipes; \ + _size = size_bin_hdr + size_bin_res; \ + } while (0) + +#define HFI_BUFFER_BIN_H265D(_size, frame_width, frame_height, \ + is_interlaced, delay, num_vpp_pipes) \ + do \ + { \ + HFI_U32 n_aligned_w = HFI_ALIGN(frame_width, \ + BUFFER_ALIGNMENT_16_BYTES); \ + HFI_U32 n_aligned_h = HFI_ALIGN(frame_height, \ + BUFFER_ALIGNMENT_16_BYTES); \ + if (!is_interlaced) \ + { \ + SIZE_H265D_HW_BIN_BUFFER(_size, n_aligned_w, \ + n_aligned_h, delay, num_vpp_pipes); \ + } \ + else \ + { \ + _size = 0; \ + } \ + } while (0) + +#define SIZE_SLIST_BUF_H265 (1 << 10) +#define NUM_SLIST_BUF_H265 (80 + 20) +#define H265_NUM_TILE_COL 32 +#define H265_NUM_TILE_ROW 128 +#define H265_NUM_TILE (H265_NUM_TILE_ROW * H265_NUM_TILE_COL + 1) +#define HFI_BUFFER_PERSIST_H265D(_size) \ + _size = HFI_ALIGN((SIZE_SLIST_BUF_H265 * NUM_SLIST_BUF_H265 + \ + H265_NUM_TILE * sizeof(Int32) + NUM_HW_PIC_BUF * SIZE_SEI_USERDATA),\ + VENUS_DMA_ALIGNMENT) + +/* + * VPxd internal buffer definition + */ +#define SIZE_VPXD_LB_FE_LEFT_CTRL(frame_width, frame_height) \ + MAX(((frame_height + 15) >> 4) * \ + MAX_FE_NBR_CTRL_LCU16_LINE_BUFFER_SIZE, \ + MAX(((frame_height + 31) >> 5) * \ + MAX_FE_NBR_CTRL_LCU32_LINE_BUFFER_SIZE, \ + ((frame_height + 63) >> 6) * MAX_FE_NBR_CTRL_LCU64_LINE_BUFFER_SIZE)) +#define SIZE_VPXD_LB_FE_TOP_CTRL(frame_width, frame_height) \ + (((HFI_ALIGN(frame_width, 64) + 8) * 10 * 2)) /* + small line */ +#define SIZE_VPXD_LB_SE_TOP_CTRL(frame_width, frame_height) \ + (((frame_width + 15) >> 4) * MAX_FE_NBR_CTRL_LCU16_LINE_BUFFER_SIZE) +#define SIZE_VPXD_LB_SE_LEFT_CTRL(frame_width, frame_height) \ + MAX(((frame_height + 15) >> 4) * \ + MAX_SE_NBR_CTRL_LCU16_LINE_BUFFER_SIZE,\ + MAX(((frame_height + 31) >> 5) * \ + MAX_SE_NBR_CTRL_LCU32_LINE_BUFFER_SIZE, \ + ((frame_height + 63) >> 6) * MAX_SE_NBR_CTRL_LCU64_LINE_BUFFER_SIZE)) +#define SIZE_VPXD_LB_RECON_DMA_METADATA_WR(frame_width, frame_height) \ + HFI_ALIGN((HFI_ALIGN(frame_height, 8) / (4 / 2)) * 64,\ + BUFFER_ALIGNMENT_32_BYTES) +#define SIZE_MP2D_LB_FE_TOP_DATA(frame_width, frame_height) \ + ((HFI_ALIGN(frame_width, 16) + 8) * 10 * 2) +#define SIZE_VP9D_LB_FE_TOP_DATA(frame_width, frame_height) \ + ((HFI_ALIGN(HFI_ALIGN(frame_width, 8), 64) + 8) * 10 * 2) +#define SIZE_MP2D_LB_PE_TOP_DATA(frame_width, frame_height) \ + ((HFI_ALIGN(frame_width, 16) >> 4) * 64) +#define SIZE_VP9D_LB_PE_TOP_DATA(frame_width, frame_height) \ + ((HFI_ALIGN(HFI_ALIGN(frame_width, 8), 64) >> 6) * 176) +#define SIZE_MP2D_LB_VSP_TOP(frame_width, frame_height) \ + (((HFI_ALIGN(frame_width, 16) >> 4) * 64 / 2) + 256) +#define SIZE_VP9D_LB_VSP_TOP(frame_width, frame_height) \ + ((((HFI_ALIGN(HFI_ALIGN(frame_width, 8), 64) >> 6) * 64 * 8) + 256)) + +/* sizeof(VP9_COL_MV_BUFFER) */ +#define HFI_IRIS2_VP9D_COMV_SIZE \ + ((((8192 + 63) >> 6) * ((4320 + 63) >> 6) * 8 * 8 * 2 * 8)) + +#define HFI_IRIS2_VP9D_LB_SIZE(_size, frame_width, frame_height, num_vpp_pipes)\ + do \ + { \ + _size = HFI_ALIGN(SIZE_VPXD_LB_FE_LEFT_CTRL(frame_width, \ + frame_height),VENUS_DMA_ALIGNMENT) * num_vpp_pipes + \ + HFI_ALIGN(SIZE_VPXD_LB_SE_LEFT_CTRL(frame_width, frame_height),\ + VENUS_DMA_ALIGNMENT) * num_vpp_pipes + \ + HFI_ALIGN(SIZE_VP9D_LB_VSP_TOP(frame_width, frame_height), \ + VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_VPXD_LB_FE_TOP_CTRL(frame_width, frame_height), \ + VENUS_DMA_ALIGNMENT) + 2 * \ + HFI_ALIGN(SIZE_VPXD_LB_RECON_DMA_METADATA_WR \ + (frame_width, frame_height), VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_VPXD_LB_SE_TOP_CTRL(frame_width, frame_height), \ + VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_VP9D_LB_PE_TOP_DATA(frame_width, frame_height), \ + VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_VP9D_LB_FE_TOP_DATA(frame_width, frame_height), \ + VENUS_DMA_ALIGNMENT); \ + } while (0) + +/* _yuv_bufcount_min = MAX(Min YUV Buffer count, + * (HFI_PROPERTY_PARAM_VDEC_VPP_DELAY + 1)) + */ +#define HFI_BUFFER_LINE_VP9D(_size, frame_width, frame_height, \ + _yuv_bufcount_min, is_opb, num_vpp_pipes) \ + do \ + { \ + HFI_U32 _lb_size = 0; \ + HFI_U32 vpss_lb_size = 0; \ + HFI_IRIS2_VP9D_LB_SIZE(_lb_size, frame_width, frame_height,\ + num_vpp_pipes); \ + if (is_opb) \ + { \ + SIZE_VPSS_LB(vpss_lb_size, frame_width, frame_height, \ + num_vpp_pipes); \ + } \ + _size = _lb_size + vpss_lb_size + HDR10_HIST_EXTRADATA_SIZE; \ + } while (0) + +#define VPX_DECODER_FRAME_CONCURENCY_LVL (2) +#define VPX_DECODER_FRAME_BIN_HDR_BUDGET_RATIO 1 / 2 +#define VPX_DECODER_FRAME_BIN_RES_BUDGET_RATIO 3 / 2 + +#define HFI_BUFFER_BIN_VP9D(_size, frame_width, frame_height, \ + is_interlaced, num_vpp_pipes) \ + do \ + { \ + HFI_U32 _size_yuv = HFI_ALIGN(frame_width, \ + BUFFER_ALIGNMENT_16_BYTES) *\ + HFI_ALIGN(frame_height, BUFFER_ALIGNMENT_16_BYTES) * 3 / 2; \ + if (!is_interlaced) \ + { \ + /* binbuffer1_size + binbufer2_size */ \ + _size = HFI_ALIGN(((MAX(_size_yuv, \ + ((BIN_BUFFER_THRESHOLD * 3) >> 1)) * \ + VPX_DECODER_FRAME_BIN_HDR_BUDGET_RATIO * \ + VPX_DECODER_FRAME_CONCURENCY_LVL) / num_vpp_pipes), \ + VENUS_DMA_ALIGNMENT) + HFI_ALIGN(((MAX(_size_yuv, \ + ((BIN_BUFFER_THRESHOLD * 3) >> 1)) * \ + VPX_DECODER_FRAME_BIN_RES_BUDGET_RATIO * \ + VPX_DECODER_FRAME_CONCURENCY_LVL) / num_vpp_pipes), \ + VENUS_DMA_ALIGNMENT); \ + _size = _size * num_vpp_pipes; \ + } \ + else \ + { \ + _size = 0; \ + } \ + } while (0) + +#define VP9_NUM_FRAME_INFO_BUF 32 +#define VP9_NUM_PROBABILITY_TABLE_BUF (VP9_NUM_FRAME_INFO_BUF + 4) +#define VP9_PROB_TABLE_SIZE (3840) + +#define VP9_UDC_HEADER_BUF_SIZE (3 * 128) +#define MAX_SUPERFRAME_HEADER_LEN (34) +#define CCE_TILE_OFFSET_SIZE HFI_ALIGN(32 * 4 * 4, BUFFER_ALIGNMENT_32_BYTES) + +#define HFI_BUFFER_PERSIST_VP9D(_size) \ + _size = HFI_ALIGN(VP9_NUM_PROBABILITY_TABLE_BUF * VP9_PROB_TABLE_SIZE, \ + VENUS_DMA_ALIGNMENT) + HFI_ALIGN(HFI_IRIS2_VP9D_COMV_SIZE, \ + VENUS_DMA_ALIGNMENT) + HFI_ALIGN(MAX_SUPERFRAME_HEADER_LEN, \ + VENUS_DMA_ALIGNMENT) + HFI_ALIGN(VP9_UDC_HEADER_BUF_SIZE, \ + VENUS_DMA_ALIGNMENT) + HFI_ALIGN(VP9_NUM_FRAME_INFO_BUF * \ + CCE_TILE_OFFSET_SIZE, VENUS_DMA_ALIGNMENT) + +/* + * MP2d internal buffer definition + */ +#define HFI_BUFFER_LINE_MP2D(_size, frame_width, frame_height, \ +_yuv_bufcount_min, is_opb, num_vpp_pipes) \ + do \ + { \ + HFI_U32 vpss_lb_size = 0; \ + _size = HFI_ALIGN(SIZE_VPXD_LB_FE_LEFT_CTRL(frame_width, \ + frame_height), VENUS_DMA_ALIGNMENT) * num_vpp_pipes + \ + HFI_ALIGN(SIZE_VPXD_LB_SE_LEFT_CTRL(frame_width, frame_height),\ + VENUS_DMA_ALIGNMENT) * num_vpp_pipes + \ + HFI_ALIGN(SIZE_MP2D_LB_VSP_TOP(frame_width, frame_height),\ + VENUS_DMA_ALIGNMENT) + HFI_ALIGN(SIZE_VPXD_LB_FE_TOP_CTRL\ + (frame_width, frame_height), VENUS_DMA_ALIGNMENT) + \ + 2 * HFI_ALIGN(SIZE_VPXD_LB_RECON_DMA_METADATA_WR(frame_width,\ + frame_height), VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_VPXD_LB_SE_TOP_CTRL(frame_width, frame_height),\ + VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_MP2D_LB_PE_TOP_DATA(frame_width, frame_height), \ + VENUS_DMA_ALIGNMENT) + \ + HFI_ALIGN(SIZE_MP2D_LB_FE_TOP_DATA(frame_width, frame_height), \ + VENUS_DMA_ALIGNMENT); \ + if (is_opb) \ + { \ + SIZE_VPSS_LB(vpss_lb_size, frame_width, frame_height, \ + num_vpp_pipes); \ + } \ + _size += vpss_lb_size; \ + } while (0) + +#define HFI_BUFFER_BIN_MP2D(_size, frame_width, frame_height, is_interlaced) 0 + +#define QMATRIX_SIZE (sizeof(HFI_U32) * 128 + 256) +#define MP2D_QPDUMP_SIZE 115200 +#define HFI_BUFFER_PERSIST_MP2D(_size) \ + _size = QMATRIX_SIZE + MP2D_QPDUMP_SIZE; + +/* Begin of IRIS2 Encoder */ + +/* + * Encoder Output Bitstream Buffer definition + * To match with driver Calculation, + * the output bitstream = YUV/4 for larger than 4K size. + */ +#define HFI_BUFFER_BITSTREAM_ENC(size, frame_width, frame_height, \ + rc_type, is_ten_bit) \ + do \ + { \ + HFI_U32 aligned_width, aligned_height, bitstream_size; \ + aligned_width = HFI_ALIGN(frame_width, 32); \ + aligned_height = HFI_ALIGN(frame_height, 32); \ + bitstream_size = aligned_width * aligned_height * 3; \ + if (aligned_width * aligned_height > (4096 * 2176)) \ + { \ + bitstream_size = (bitstream_size >> 3); \ + } \ + else if (bitstream_size > (1280 * 720)) \ + { \ + bitstream_size = (bitstream_size >> 2); \ + } \ + else \ + { \ + bitstream_size = (bitstream_size << 1);\ + } \ + if ((rc_type == HFI_RC_CQ) || (rc_type == HFI_RC_OFF)) \ + { \ + bitstream_size = (bitstream_size << 1);\ + } \ + if (is_ten_bit) \ + { \ + bitstream_size = (bitstream_size) + \ + (bitstream_size >> 2); \ + } \ + size = HFI_ALIGN(bitstream_size, HFI_ALIGNMENT_4096); \ + } while (0) + +#define SIZE_ROI_METADATA_ENC(size_roi, frame_width, frame_height, lcu_size)\ + do \ + { \ + HFI_U32 width_in_lcus = 0, height_in_lcus = 0, n_shift = 0; \ + while (lcu_size && !(lcu_size & 0x1)) \ + { \ + n_shift++; \ + lcu_size = lcu_size >> 1; \ + } \ + width_in_lcus = (frame_width + (lcu_size - 1)) >> n_shift; \ + height_in_lcus = (frame_height + (lcu_size - 1)) >> n_shift; \ + size_roi = (((width_in_lcus + 7) >> 3) << 3) * \ + height_in_lcus * 2 + 256; \ + } while (0) + +/* + * Encoder Input Extradata Buffer definition + */ +#define HFI_BUFFER_INPUT_METADATA_ENC(size, frame_width, frame_height, \ + is_roi_enabled, lcu_size) \ + do \ + { \ + HFI_U32 roi_size = 0; \ + if (is_roi_enabled) \ + { \ + SIZE_ROI_METADATA_ENC(roi_size, frame_width, \ + frame_height, lcu_size); \ + } \ + size = roi_size + 16384; \ + size = HFI_ALIGN(size, HFI_ALIGNMENT_4096); \ + } while (0) + +#define HFI_BUFFER_INPUT_METADATA_H264E(size_metadata, frame_width, \ + frame_height, is_roi_enabled) \ + do \ + { \ + HFI_BUFFER_INPUT_METADATA_ENC(size_metadata, frame_width, \ + frame_height, is_roi_enabled, 16); \ + }while (0) + +#define HFI_BUFFER_INPUT_METADATA_H265E(size_metadata, frame_width, \ + frame_height, is_roi_enabled) \ + do \ + { \ + HFI_BUFFER_INPUT_METADATA_ENC(size_metadata, frame_width, \ + frame_height, is_roi_enabled, 32); \ + } while (0) + +#define HFI_BUFFER_ARP_ENC(size) \ + do \ + { \ + size = 204800; \ + } while (0) + +/* + * Encoder Scratch Buffer definition + */ +#define HFI_MAX_COL_FRAME 6 +#define HFI_VENUS_VENC_TRE_WB_BUFF_SIZE (65 << 4) // bytes +#define HFI_VENUS_VENC_DB_LINE_BUFF_PER_MB 512 +#define HFI_VENUS_VPPSG_MAX_REGISTERS 2048 +#define HFI_VENUS_WIDTH_ALIGNMENT 128 +#define HFI_VENUS_WIDTH_TEN_BIT_ALIGNMENT 192 +#define HFI_VENUS_HEIGHT_ALIGNMENT 32 +#define VENUS_METADATA_STRIDE_MULTIPLE 64 +#define VENUS_METADATA_HEIGHT_MULTIPLE 16 + +#ifndef SYSTEM_LAL_TILE10 +#define SYSTEM_LAL_TILE10 192 +#endif + +/* + * Host uses following macro to calculate num_ref for encoder + * Here: _total_hp_layers = HFI_PROPERTY_PARAM_VENC_HIER_P_MAX_NUM_ENH_LAYER + 1 + * Here: _total_hb_layers = HFI_PROPERTY_PARAM_VENC_HIER_B_MAX_NUM_ENH_LAYER + 1 + */ +#define HFI_IRIS2_ENC_RECON_BUF_COUNT(num_ref, n_bframe, ltr_count, \ + _total_hp_layers, _total_hb_layers, hybrid_hp, codec_standard) \ + do \ + { \ + num_ref = 1; \ + if (n_bframe) \ + num_ref = n_bframe + 1; \ + if (ltr_count) \ + /* B and LTR can't be at same time */\ + num_ref = num_ref + ltr_count; \ + if (_total_hp_layers) \ + { \ + if (hybrid_hp) \ + /* LTR and B-frame not supported with hybrid HP */\ + num_ref = (_total_hp_layers - 1); \ + if (codec_standard == HFI_VIDEO_CODEC_HEVC) \ + num_ref = (_total_hp_layers + 1) / 2 + \ + ltr_count; \ + else if (codec_standard == HFI_VIDEO_CODEC_H264 && \ + _total_hp_layers <= 4) \ + num_ref = (2 ^ (_total_hp_layers - 1)) - 1 + \ + ltr_count; \ + else \ + /* AVC normal HP and TotalHPLayer>4.*/ \ + /* This is NPOR. uses MMCO. */ \ + num_ref = (_total_hp_layers + 1) / 2 + \ + ltr_count; \ + } \ + if (_total_hb_layers >= 2) \ + { \ + num_ref = (2 ^ (_total_hb_layers - 1)) / 2 + 1; \ + } \ + } while (0) + +#define SIZE_BIN_BITSTREAM_ENC(_size, frame_width, frame_height, \ + work_mode, lcu_size) \ + do \ + { \ + HFI_U32 size_aligned_width = 0, size_aligned_height = 0; \ + HFI_U32 bitstream_size_eval = 0; \ + size_aligned_width = HFI_ALIGN((frame_width), lcu_size); \ + size_aligned_height = HFI_ALIGN((frame_height), lcu_size); \ + if (work_mode == HFI_WORKMODE_2) \ + { \ + bitstream_size_eval = (((size_aligned_width) * \ + (size_aligned_height)*3 * 5) >> 2); \ + if (size_aligned_width * size_aligned_height > \ + (4096 * 2176)) \ + { \ + bitstream_size_eval = \ + (bitstream_size_eval >> 3); \ + } \ + else if (bitstream_size_eval > (352 * 288 * 4)) \ + { \ + bitstream_size_eval = \ + (bitstream_size_eval >> 2); \ + } \ + } \ + else \ + { \ + bitstream_size_eval = size_aligned_width * \ + size_aligned_height * 3; \ + } \ + _size = HFI_ALIGN(bitstream_size_eval, VENUS_DMA_ALIGNMENT); \ + } while (0) + +#define SIZE_ENC_SINGLE_PIPE(size, bitbin_size, num_vpp_pipes, \ + frame_height, frame_width) \ + do \ + { \ + HFI_U32 size_single_pipe_eval = 0, sao_bin_buffer_size = 0, \ + _padded_bin_sz = 0; \ + if (num_vpp_pipes > 2) \ + { \ + size_single_pipe_eval = bitbin_size / 2; \ + } \ + else \ + { \ + size_single_pipe_eval = bitbin_size; \ + } \ + size_single_pipe_eval = HFI_ALIGN(size_single_pipe_eval, \ + VENUS_DMA_ALIGNMENT); \ + sao_bin_buffer_size = (64 * ((((frame_width) + \ + BUFFER_ALIGNMENT_32_BYTES) * ((frame_height) +\ + BUFFER_ALIGNMENT_32_BYTES)) >> 10)) + 384; \ + _padded_bin_sz = HFI_ALIGN(size_single_pipe_eval, \ + VENUS_DMA_ALIGNMENT);\ + size_single_pipe_eval = sao_bin_buffer_size + _padded_bin_sz; \ + size_single_pipe_eval = HFI_ALIGN(size_single_pipe_eval, \ + VENUS_DMA_ALIGNMENT); \ + size = size_single_pipe_eval; \ + } while (0) + +#define HFI_BUFFER_BIN_ENC(_size, frame_width, frame_height, lcu_size, \ + work_mode, num_vpp_pipes) \ + do \ + { \ + HFI_U32 bitstream_size = 0, total_bitbin_buffers = 0, \ + size_single_pipe = 0, bitbin_size = 0; \ + HFI_U32 sao_bin_buffer_size = 0, _padded_bin_sz = 0; \ + SIZE_BIN_BITSTREAM_ENC(bitstream_size, frame_width, \ + frame_height, work_mode, lcu_size); \ + if (work_mode == HFI_WORKMODE_2) \ + { \ + total_bitbin_buffers = 3; \ + bitbin_size = bitstream_size * 17 / 10; \ + bitbin_size = HFI_ALIGN(bitbin_size, \ + VENUS_DMA_ALIGNMENT); \ + } \ + else \ + { \ + total_bitbin_buffers = 1; \ + bitbin_size = bitstream_size; \ + } \ + SIZE_ENC_SINGLE_PIPE(size_single_pipe, bitbin_size, \ + num_vpp_pipes, frame_height, frame_width); \ + bitbin_size = size_single_pipe * num_vpp_pipes; \ + _size = HFI_ALIGN(bitbin_size, VENUS_DMA_ALIGNMENT) * \ + total_bitbin_buffers + 512; \ + } while (0) + +#define HFI_BUFFER_BIN_H264E(_size, frame_width, frame_height, \ + work_mode, num_vpp_pipes) \ + do \ + { \ + HFI_BUFFER_BIN_ENC(_size, frame_width, frame_height, 16, \ + work_mode, num_vpp_pipes); \ + } while (0) + +#define HFI_BUFFER_BIN_H265E(_size, frame_width, frame_height, \ + work_mode, num_vpp_pipes) \ + do \ + { \ + HFI_BUFFER_BIN_ENC(_size, frame_width, frame_height, 32,\ + work_mode, num_vpp_pipes); \ + } while (0) + +#define SIZE_ENC_SLICE_INFO_BUF(num_lcu_in_frame) HFI_ALIGN((256 + \ + (num_lcu_in_frame << 4)), VENUS_DMA_ALIGNMENT) +#define SIZE_LINE_BUF_CTRL(frame_width_coded) \ + HFI_ALIGN(frame_width_coded, VENUS_DMA_ALIGNMENT) +#define SIZE_LINE_BUF_CTRL_ID2(frame_width_coded) \ + HFI_ALIGN(frame_width_coded, VENUS_DMA_ALIGNMENT) + +#define SIZE_LINEBUFF_DATA(_size, is_ten_bit, frame_width_coded) \ + do \ + { \ + _size = is_ten_bit ? (((((10 * (frame_width_coded) +\ + 1024) + (VENUS_DMA_ALIGNMENT - 1)) & \ + (~(VENUS_DMA_ALIGNMENT - 1))) * 1) + \ + (((((10 * (frame_width_coded) + 1024) >> 1) + \ + (VENUS_DMA_ALIGNMENT - 1)) & (~(VENUS_DMA_ALIGNMENT - 1))) * \ + 2)) : (((((8 * (frame_width_coded) + 1024) + \ + (VENUS_DMA_ALIGNMENT - 1)) \ + & (~(VENUS_DMA_ALIGNMENT - 1))) * 1) + \ + (((((8 * (frame_width_coded) +\ + 1024) >> 1) + (VENUS_DMA_ALIGNMENT - 1)) & \ + (~(VENUS_DMA_ALIGNMENT - 1))) * 2)); \ + } while (0) + +#define SIZE_LEFT_LINEBUFF_CTRL(_size, standard, frame_height_coded, \ + num_vpp_pipes_enc) \ + do \ + { \ + _size = (standard == HFI_VIDEO_CODEC_HEVC) ? \ + (((frame_height_coded) + \ + (BUF_SIZE_ALIGN_32)) / BUF_SIZE_ALIGN_32 * 4 * 16) : \ + (((frame_height_coded) + 15) / 16 * 5 * 16); \ + if ((num_vpp_pipes_enc) > 1) \ + { \ + _size += BUFFER_ALIGNMENT_512_BYTES; \ + _size = HFI_ALIGN(_size, BUFFER_ALIGNMENT_512_BYTES) *\ + (num_vpp_pipes_enc); \ + } \ + _size = HFI_ALIGN(_size, VENUS_DMA_ALIGNMENT); \ + } while (0) + +#define SIZE_LEFT_LINEBUFF_RECON_PIX(_size, is_ten_bit, frame_height_coded, \ + num_vpp_pipes_enc) \ + do \ + { \ + _size = (((is_ten_bit + 1) * 2 * (frame_height_coded) + \ + VENUS_DMA_ALIGNMENT) + \ + (VENUS_DMA_ALIGNMENT << (num_vpp_pipes_enc - 1)) - 1) & \ + (~((VENUS_DMA_ALIGNMENT << (num_vpp_pipes_enc - 1)) - 1)) * 1; \ + } while (0) + +#define SIZE_TOP_LINEBUFF_CTRL_FE(_size, frame_width_coded, standard) \ + do \ + { \ + _size = (standard == HFI_VIDEO_CODEC_HEVC) ? (64 * \ + ((frame_width_coded) >> 5)) : (VENUS_DMA_ALIGNMENT + 16 * \ + ((frame_width_coded) >> 4)); \ + _size = HFI_ALIGN(_size, VENUS_DMA_ALIGNMENT); \ + } while (0) + +#define SIZE_LEFT_LINEBUFF_CTRL_FE(frame_height_coded, num_vpp_pipes_enc) \ + ((((VENUS_DMA_ALIGNMENT + 64 * ((frame_height_coded) >> 4)) + \ + (VENUS_DMA_ALIGNMENT << (num_vpp_pipes_enc - 1)) - 1) & \ + (~((VENUS_DMA_ALIGNMENT << (num_vpp_pipes_enc - 1)) - 1)) * 1) * \ + num_vpp_pipes_enc) + +#define SIZE_LEFT_LINEBUFF_METADATA_RECON_Y(_size, frame_height_coded, \ + is_ten_bit, num_vpp_pipes_enc) \ + do \ + { \ + _size = ((VENUS_DMA_ALIGNMENT + 64 * ((frame_height_coded) / \ + (8 * (is_ten_bit ? 4 : 8))))); \ + _size = HFI_ALIGN(_size, VENUS_DMA_ALIGNMENT); \ + _size = (_size * num_vpp_pipes_enc); \ + } while (0) + +#define SIZE_LEFT_LINEBUFF_METADATA_RECON_UV(_size, frame_height_coded, \ + is_ten_bit, num_vpp_pipes_enc) \ + do \ + { \ + _size = ((VENUS_DMA_ALIGNMENT + 64 * ((frame_height_coded) / \ + (4 * (is_ten_bit ? 4 : 8))))); \ + _size = HFI_ALIGN(_size, VENUS_DMA_ALIGNMENT); \ + _size = (_size * num_vpp_pipes_enc); \ + } while (0) + +#define SIZE_LINEBUFF_RECON_PIX(_size, is_ten_bit, frame_width_coded) \ + do \ + { \ + _size = ((is_ten_bit ? 3 : 2) * (frame_width_coded)); \ + _size = HFI_ALIGN(_size, VENUS_DMA_ALIGNMENT); \ + } while (0) + +#define SIZE_SLICE_CMD_BUFFER (HFI_ALIGN(20480, VENUS_DMA_ALIGNMENT)) +#define SIZE_SPS_PPS_SLICE_HDR (2048 + 4096) + +#define SIZE_FRAME_RC_BUF_SIZE(_size, standard, frame_height_coded, \ + num_vpp_pipes_enc) \ + do \ + { \ + _size = (standard == HFI_VIDEO_CODEC_HEVC) ? (256 + 16 * \ + (14 + ((((frame_height_coded) >> 5) + 7) >> 3))) : \ + (256 + 16 * (14 + ((((frame_height_coded) >> 4) + 7) >> 3))); \ + _size *= 6; /* multiply by max numtilescol*/ \ + if (num_vpp_pipes_enc > 1) \ + { \ + _size = HFI_ALIGN(_size, VENUS_DMA_ALIGNMENT) \ + * num_vpp_pipes_enc;\ + } \ + _size = HFI_ALIGN(_size, BUFFER_ALIGNMENT_512_BYTES) * \ + HFI_MAX_COL_FRAME; \ + } while (0) + +#define ENC_BITCNT_BUF_SIZE(num_lcu_in_frame) HFI_ALIGN((256 + \ + (4 * (num_lcu_in_frame))), VENUS_DMA_ALIGNMENT) +#define ENC_BITMAP_BUF_SIZE(num_lcu_in_frame) HFI_ALIGN((256 + \ + ((num_lcu_in_frame) >> 3)), VENUS_DMA_ALIGNMENT) +#define SIZE_LINE_BUF_SDE(frame_width_coded) HFI_ALIGN((256 + \ + (16 * ((frame_width_coded) >> 4))), VENUS_DMA_ALIGNMENT) + +#define SIZE_SE_STATS_BUF(_size, frame_width_coded, frame_height_coded, \ + num_lcu_in_frame) \ + do \ + { \ + if (((frame_width_coded) * (frame_height_coded)) > \ + (4096 * 2160)) \ + { \ + _size = 0; \ + } \ + else if (((frame_width_coded) * (frame_height_coded)) > \ + (1920 * 1088)) \ + { \ + _size = (40 * 4 * num_lcu_in_frame + 256 + 256); \ + } \ + else \ + { \ + _size = (1024 * num_lcu_in_frame + 256 + 256); \ + } \ + _size = HFI_ALIGN(se_stats_buf_size, VENUS_DMA_ALIGNMENT) * 2; \ + } while (0) + +#define SIZE_BSE_SLICE_CMD_BUF ((((8192 << 2) + 7) & (~7)) * 6) +#define SIZE_BSE_REG_BUF ((((512 << 3) + 7) & (~7)) * 4) +#define SIZE_VPP_REG_BUF ((((HFI_VENUS_VPPSG_MAX_REGISTERS << 3) +\ + 31) & (~31)) * 10) +#define SIZE_LAMBDA_LUT (256 * 11) +#define SIZE_OVERRIDE_BUF(num_lcumb) (HFI_ALIGN(((16 * (((num_lcumb) + 7)\ + >> 3))), VENUS_DMA_ALIGNMENT) * 2) +#define SIZE_IR_BUF(num_lcu_in_frame) (((((num_lcu_in_frame) << 1) + 7) &\ + (~7)) * 3) + +#define SIZE_VPSS_LINE_BUF(num_vpp_pipes_enc, frame_height_coded, \ + frame_width_coded) \ + ((((((8192) >> 2) << 5) * (num_vpp_pipes_enc)) + 64) + \ + (((((MAX((frame_width_coded), (frame_height_coded)) + 3) >> 2) << 5) +\ + 256) * 16)) + +#define SIZE_TOP_LINE_BUF_FIRST_STG_SAO(frame_width_coded) \ + HFI_ALIGN((16 * ((frame_width_coded) >> 5)), VENUS_DMA_ALIGNMENT) + +#define HFI_BUFFER_LINE_ENC(_size, frame_width, frame_height, is_ten_bit, \ + num_vpp_pipes_enc, lcu_size, standard) \ + do \ + { \ + HFI_U32 width_in_lcus = 0, height_in_lcus = 0, \ + frame_width_coded = 0, frame_height_coded = 0; \ + HFI_U32 line_buff_data_size = 0, left_line_buff_ctrl_size = 0, \ + left_line_buff_recon_pix_size = 0, \ + top_line_buff_ctrl_fe_size = 0; \ + HFI_U32 left_line_buff_metadata_recon__y__size = 0, \ + left_line_buff_metadata_recon__uv__size = 0, \ + line_buff_recon_pix_size = 0; \ + width_in_lcus = ((frame_width) + (lcu_size)-1) / (lcu_size); \ + height_in_lcus = ((frame_height) + (lcu_size)-1) / (lcu_size); \ + frame_width_coded = width_in_lcus * (lcu_size); \ + frame_height_coded = height_in_lcus * (lcu_size); \ + SIZE_LINEBUFF_DATA(line_buff_data_size, is_ten_bit, \ + frame_width_coded);\ + SIZE_LEFT_LINEBUFF_CTRL(left_line_buff_ctrl_size, standard, \ + frame_height_coded, num_vpp_pipes_enc); \ + SIZE_LEFT_LINEBUFF_RECON_PIX(left_line_buff_recon_pix_size, \ + is_ten_bit, frame_height_coded, num_vpp_pipes_enc); \ + SIZE_TOP_LINEBUFF_CTRL_FE(top_line_buff_ctrl_fe_size, \ + frame_width_coded, standard); \ + SIZE_LEFT_LINEBUFF_METADATA_RECON_Y\ + (left_line_buff_metadata_recon__y__size, \ + frame_height_coded, is_ten_bit, num_vpp_pipes_enc); \ + SIZE_LEFT_LINEBUFF_METADATA_RECON_UV\ + (left_line_buff_metadata_recon__uv__size, \ + frame_height_coded, is_ten_bit, num_vpp_pipes_enc); \ + SIZE_LINEBUFF_RECON_PIX(line_buff_recon_pix_size, is_ten_bit,\ + frame_width_coded); \ + _size = SIZE_LINE_BUF_CTRL(frame_width_coded) + \ + SIZE_LINE_BUF_CTRL_ID2(frame_width_coded) + \ + line_buff_data_size + \ + left_line_buff_ctrl_size + \ + left_line_buff_recon_pix_size + \ + top_line_buff_ctrl_fe_size + \ + left_line_buff_metadata_recon__y__size + \ + left_line_buff_metadata_recon__uv__size + \ + line_buff_recon_pix_size + \ + SIZE_LEFT_LINEBUFF_CTRL_FE(frame_height_coded, \ + num_vpp_pipes_enc) + SIZE_LINE_BUF_SDE(frame_width_coded) + \ + SIZE_VPSS_LINE_BUF(num_vpp_pipes_enc, frame_height_coded, \ + frame_width_coded) + \ + SIZE_TOP_LINE_BUF_FIRST_STG_SAO(frame_width_coded); \ + } while (0) + +#define HFI_BUFFER_LINE_H264E(_size, frame_width, frame_height, is_ten_bit, \ + num_vpp_pipes) \ + do \ + { \ + HFI_BUFFER_LINE_ENC(_size, frame_width, frame_height, 0, \ + num_vpp_pipes, 16, HFI_VIDEO_CODEC_H264); \ + } while (0) + +#define HFI_BUFFER_LINE_H265E(_size, frame_width, frame_height, is_ten_bit, \ + num_vpp_pipes) \ + do \ + { \ + HFI_BUFFER_LINE_ENC(_size, frame_width, frame_height, \ + is_ten_bit, num_vpp_pipes, 32, HFI_VIDEO_CODEC_HEVC); \ + } while (0) + +#define HFI_BUFFER_COMV_ENC(_size, frame_width, frame_height, lcu_size, \ + num_ref, standard) \ + do \ + { \ + HFI_U32 size_colloc_mv = 0, size_colloc_rc = 0; \ + HFI_U32 mb_width = ((frame_width) + 15) >> 4; \ + HFI_U32 mb_height = ((frame_height) + 15) >> 4; \ + HFI_U32 width_in_lcus = ((frame_width) + (lcu_size)-1) /\ + (lcu_size); \ + HFI_U32 height_in_lcus = ((frame_height) + (lcu_size)-1) / \ + (lcu_size); \ + HFI_U32 num_lcu_in_frame = width_in_lcus * height_in_lcus; \ + size_colloc_mv = (standard == HFI_VIDEO_CODEC_HEVC) ? \ + (16 * ((num_lcu_in_frame << 2) + BUFFER_ALIGNMENT_32_BYTES)) : \ + (3 * 16 * (width_in_lcus * height_in_lcus +\ + BUFFER_ALIGNMENT_32_BYTES)); \ + size_colloc_mv = HFI_ALIGN(size_colloc_mv, \ + VENUS_DMA_ALIGNMENT) * (num_ref + 1); \ + size_colloc_rc = (((mb_width + 7) >> 3) * 16 * 2 * mb_height); \ + size_colloc_rc = HFI_ALIGN(size_colloc_rc, \ + VENUS_DMA_ALIGNMENT) * HFI_MAX_COL_FRAME; \ + _size = size_colloc_mv + size_colloc_rc; \ + } while (0) + +#define HFI_BUFFER_COMV_H264E(_size, frame_width, frame_height, num_ref) \ + do \ + { \ + HFI_BUFFER_COMV_ENC(_size, frame_width, frame_height, 16, \ + num_ref, HFI_VIDEO_CODEC_H264); \ + } while (0) + +#define HFI_BUFFER_COMV_H265E(_size, frame_width, frame_height, num_ref) \ + do \ + { \ + HFI_BUFFER_COMV_ENC(_size, frame_width, frame_height, 32,\ + num_ref, HFI_VIDEO_CODEC_HEVC); \ + } while (0) + +#define HFI_BUFFER_NON_COMV_ENC(_size, frame_width, frame_height, \ + num_vpp_pipes_enc, lcu_size, standard) \ + do \ + { \ + HFI_U32 width_in_lcus = 0, height_in_lcus = 0, \ + frame_width_coded = 0, frame_height_coded = 0, \ + num_lcu_in_frame = 0, num_lcumb = 0; \ + HFI_U32 left_line_buff_metadata_recon__y__size = 0, \ + left_line_buff_metadata_recon__uv__size = 0, \ + line_buff_recon_pix_size = 0, frame_rc_buf_size = 0, \ + se_stats_buf_size = 0; \ + width_in_lcus = ((frame_width) + (lcu_size)-1) / (lcu_size); \ + height_in_lcus = ((frame_height) + (lcu_size)-1) / (lcu_size); \ + num_lcu_in_frame = width_in_lcus * height_in_lcus; \ + frame_width_coded = width_in_lcus * (lcu_size); \ + frame_height_coded = height_in_lcus * (lcu_size); \ + num_lcumb = (frame_height_coded / lcu_size) * \ + ((frame_width_coded + lcu_size * 8) / lcu_size); \ + SIZE_FRAME_RC_BUF_SIZE(frame_rc_buf_size, standard, \ + frame_height_coded, num_vpp_pipes_enc); \ + SIZE_SE_STATS_BUF(se_stats_buf_size, frame_width_coded, \ + frame_height_coded, num_lcu_in_frame); \ + _size = SIZE_ENC_SLICE_INFO_BUF(num_lcu_in_frame) + \ + SIZE_SLICE_CMD_BUFFER + \ + SIZE_SPS_PPS_SLICE_HDR + \ + frame_rc_buf_size + \ + ENC_BITCNT_BUF_SIZE(num_lcu_in_frame) + \ + ENC_BITMAP_BUF_SIZE(num_lcu_in_frame) + \ + se_stats_buf_size + \ + SIZE_BSE_SLICE_CMD_BUF + \ + SIZE_BSE_REG_BUF + \ + SIZE_VPP_REG_BUF + \ + SIZE_LAMBDA_LUT + \ + SIZE_OVERRIDE_BUF(num_lcumb) + \ + SIZE_IR_BUF(num_lcu_in_frame); \ + } while (0) + +#define HFI_BUFFER_NON_COMV_H264E(_size, frame_width, frame_height, \ + num_vpp_pipes_enc) \ + do \ + { \ + HFI_BUFFER_NON_COMV_ENC(_size, frame_width, frame_height, \ + num_vpp_pipes_enc, 16, HFI_VIDEO_CODEC_H264); \ + } while (0) + +#define HFI_BUFFER_NON_COMV_H265E(_size, frame_width, frame_height, \ + num_vpp_pipes_enc) \ + do \ + { \ + HFI_BUFFER_NON_COMV_ENC(_size, frame_width, frame_height, \ + num_vpp_pipes_enc, 32, HFI_VIDEO_CODEC_HEVC); \ + } while (0) + +#define SIZE_ENC_REF_BUFFER(size, frame_width, frame_height) \ + do \ + { \ + HFI_U32 u_buffer_width = 0, u_buffer_height = 0, \ + u_chroma_buffer_height = 0; \ + u_buffer_height = HFI_ALIGN(frame_height, \ + HFI_VENUS_HEIGHT_ALIGNMENT); \ + u_chroma_buffer_height = frame_height >> 1; \ + u_chroma_buffer_height = HFI_ALIGN(u_chroma_buffer_height, \ + HFI_VENUS_HEIGHT_ALIGNMENT); \ + u_buffer_width = HFI_ALIGN(frame_width, \ + HFI_VENUS_WIDTH_ALIGNMENT); \ + size = (u_buffer_height + u_chroma_buffer_height) * \ + u_buffer_width; \ + } while (0) + +#define SIZE_ENC_TEN_BIT_REF_BUFFER(size, frame_width, frame_height) \ + do \ + { \ + HFI_U32 ref_buf_height = 0, ref_luma_stride_in_bytes = 0, \ + u_ref_stride = 0, luma_size = 0, ref_chrm_height_in_bytes = 0, \ + chroma_size = 0, ref_buf_size = 0; \ + ref_buf_height = (frame_height + \ + (HFI_VENUS_HEIGHT_ALIGNMENT - 1)) \ + & (~(HFI_VENUS_HEIGHT_ALIGNMENT - 1)); \ + ref_luma_stride_in_bytes = ((frame_width + \ + SYSTEM_LAL_TILE10 - 1) / SYSTEM_LAL_TILE10) * \ + SYSTEM_LAL_TILE10; \ + u_ref_stride = 4 * (ref_luma_stride_in_bytes / 3); \ + u_ref_stride = (u_ref_stride + (BUF_SIZE_ALIGN_128 - 1)) &\ + (~(BUF_SIZE_ALIGN_128 - 1)); \ + luma_size = ref_buf_height * u_ref_stride; \ + ref_chrm_height_in_bytes = (((frame_height + 1) >> 1) + \ + (BUF_SIZE_ALIGN_32 - 1)) & (~(BUF_SIZE_ALIGN_32 - 1)); \ + chroma_size = u_ref_stride * ref_chrm_height_in_bytes; \ + luma_size = (luma_size + (BUF_SIZE_ALIGN_4096 - 1)) & \ + (~(BUF_SIZE_ALIGN_4096 - 1)); \ + chroma_size = (chroma_size + (BUF_SIZE_ALIGN_4096 - 1)) & \ + (~(BUF_SIZE_ALIGN_4096 - 1)); \ + ref_buf_size = luma_size + chroma_size; \ + size = ref_buf_size; \ + } while (0) + +/* + * Encoder Scratch2 Buffer definition + */ +#define HFI_BUFFER_DPB_ENC(_size, frame_width, frame_height, is_ten_bit) \ + do \ + { \ + HFI_U32 metadata_stride, metadata_buf_height, meta_size_y, \ + meta_size_c;\ + HFI_U32 ten_bit_ref_buf_size = 0, ref_buf_size = 0; \ + if (!is_ten_bit) \ + { \ + SIZE_ENC_REF_BUFFER(ref_buf_size, frame_width, \ + frame_height); \ + HFI_UBWC_CALC_METADATA_PLANE_STRIDE(metadata_stride, \ + (frame_width),\ + 64, HFI_COLOR_FORMAT_YUV420_NV12_UBWC_Y_TILE_WIDTH); \ + HFI_UBWC_METADATA_PLANE_BUFHEIGHT(metadata_buf_height, \ + (frame_height), 16, \ + HFI_COLOR_FORMAT_YUV420_NV12_UBWC_Y_TILE_HEIGHT); \ + HFI_UBWC_METADATA_PLANE_BUFFER_SIZE(meta_size_y, \ + metadata_stride, metadata_buf_height); \ + HFI_UBWC_METADATA_PLANE_BUFFER_SIZE(meta_size_c, \ + metadata_stride, metadata_buf_height); \ + _size = ref_buf_size + meta_size_y + meta_size_c; \ + } \ + else \ + { \ + SIZE_ENC_TEN_BIT_REF_BUFFER(ten_bit_ref_buf_size, \ + frame_width, frame_height); \ + HFI_UBWC_CALC_METADATA_PLANE_STRIDE(metadata_stride, \ + frame_width, VENUS_METADATA_STRIDE_MULTIPLE, \ + HFI_COLOR_FORMAT_YUV420_TP10_UBWC_Y_TILE_WIDTH); \ + HFI_UBWC_METADATA_PLANE_BUFHEIGHT(metadata_buf_height, \ + frame_height, VENUS_METADATA_HEIGHT_MULTIPLE, \ + HFI_COLOR_FORMAT_YUV420_TP10_UBWC_Y_TILE_HEIGHT); \ + HFI_UBWC_METADATA_PLANE_BUFFER_SIZE(meta_size_y, \ + metadata_stride, metadata_buf_height); \ + HFI_UBWC_METADATA_PLANE_BUFFER_SIZE(meta_size_c, \ + metadata_stride, metadata_buf_height); \ + _size = ten_bit_ref_buf_size + meta_size_y + \ + meta_size_c; \ + } \ + } while (0) + +#define HFI_BUFFER_DPB_H264E(_size, frame_width, frame_height) \ + do \ + { \ + HFI_BUFFER_DPB_ENC(_size, frame_width, frame_height, 0); \ + } while (0) + +#define HFI_BUFFER_DPB_H265E(_size, frame_width, frame_height, is_ten_bit) \ + do \ + { \ + HFI_BUFFER_DPB_ENC(_size, frame_width, frame_height, \ + is_ten_bit); \ + } while (0) + +#define HFI_BUFFER_VPSS_ENC(vpss_size, frame_width, frame_height, ds_enable, \ + rot_enable, flip_enable, is_ten_bit) \ + do \ + { \ + HFI_U32 vpss_size = 0; \ + if (ds_enable) \ + { \ + if (rot_enable) \ + { \ + HFI_BUFFER_DPB_ENC(vpss_size, frame_height, \ + frame_width, is_ten_bit); \ + } \ + else if (flip_enable) \ + { \ + HFI_BUFFER_DPB_ENC(vpss_size, frame_width, \ + frame_height, is_ten_bit); \ + } \ + else \ + { \ + vpss_size = 0; \ + } \ + vpss_size = vpss_size; \ + } \ + else \ + { \ + vpss_size = vpss_size; \ + } \ + } while (0) + +/* End of IRIS2 */ +#endif /* __HFI_BUFFER_IRIS2__ */