From 30bfbb3f97bd64b7838bcb55c98fa698b1bcc9d2 Mon Sep 17 00:00:00 2001 From: trav90 Date: Sun, 4 Feb 2018 13:19:22 -0600 Subject: Update FFmpeg code to n3.2-65-gee56777 --- media/ffvpx/libavcodec/allcodecs.c | 28 +- media/ffvpx/libavcodec/avcodec.h | 213 +++++++++++--- media/ffvpx/libavcodec/avpacket.c | 20 +- media/ffvpx/libavcodec/bsf.h | 11 + media/ffvpx/libavcodec/codec_desc.c | 2 +- media/ffvpx/libavcodec/dummy_funcs.c | 28 ++ media/ffvpx/libavcodec/h264dsp.h | 5 +- media/ffvpx/libavcodec/imgconvert.c | 8 - media/ffvpx/libavcodec/internal.h | 1 + media/ffvpx/libavcodec/me_cmp.h | 1 + media/ffvpx/libavcodec/mpegvideo.h | 3 +- media/ffvpx/libavcodec/parser.c | 5 + media/ffvpx/libavcodec/raw.c | 10 + media/ffvpx/libavcodec/utils.c | 145 ++++++++-- media/ffvpx/libavcodec/version.h | 2 +- media/ffvpx/libavcodec/vp9.c | 5 + media/ffvpx/libavcodec/x86/h264_i386.h | 12 +- media/ffvpx/libavcodec/x86/vp9dsp_init.c | 10 + media/ffvpx/libavcodec/x86/vp9itxfm.asm | 434 +++++++++++++++++++++++++++- media/ffvpx/libavcodec/x86/vp9lpf.asm | 238 +++++++++------ media/ffvpx/libavcodec/x86/vp9lpf_16bpp.asm | 2 +- 21 files changed, 993 insertions(+), 190 deletions(-) (limited to 'media/ffvpx/libavcodec') diff --git a/media/ffvpx/libavcodec/allcodecs.c b/media/ffvpx/libavcodec/allcodecs.c index 54efaad34..b592aa3b2 100644 --- a/media/ffvpx/libavcodec/allcodecs.c +++ b/media/ffvpx/libavcodec/allcodecs.c @@ -67,11 +67,13 @@ void avcodec_register_all(void) initialized = 1; /* hardware accelerators */ + REGISTER_HWACCEL(H263_CUVID, h263_cuvid); REGISTER_HWACCEL(H263_VAAPI, h263_vaapi); REGISTER_HWACCEL(H263_VIDEOTOOLBOX, h263_videotoolbox); REGISTER_HWACCEL(H264_CUVID, h264_cuvid); REGISTER_HWACCEL(H264_D3D11VA, h264_d3d11va); REGISTER_HWACCEL(H264_DXVA2, h264_dxva2); + REGISTER_HWACCEL(H264_MEDIACODEC, h264_mediacodec); REGISTER_HWACCEL(H264_MMAL, h264_mmal); REGISTER_HWACCEL(H264_QSV, h264_qsv); REGISTER_HWACCEL(H264_VAAPI, h264_vaapi); @@ -82,12 +84,16 @@ void avcodec_register_all(void) REGISTER_HWACCEL(HEVC_CUVID, hevc_cuvid); REGISTER_HWACCEL(HEVC_D3D11VA, hevc_d3d11va); REGISTER_HWACCEL(HEVC_DXVA2, hevc_dxva2); + REGISTER_HWACCEL(HEVC_MEDIACODEC, hevc_mediacodec); REGISTER_HWACCEL(HEVC_QSV, hevc_qsv); REGISTER_HWACCEL(HEVC_VAAPI, hevc_vaapi); REGISTER_HWACCEL(HEVC_VDPAU, hevc_vdpau); + REGISTER_HWACCEL(MJPEG_CUVID, mjpeg_cuvid); + REGISTER_HWACCEL(MPEG1_CUVID, mpeg1_cuvid); REGISTER_HWACCEL(MPEG1_XVMC, mpeg1_xvmc); REGISTER_HWACCEL(MPEG1_VDPAU, mpeg1_vdpau); REGISTER_HWACCEL(MPEG1_VIDEOTOOLBOX, mpeg1_videotoolbox); + REGISTER_HWACCEL(MPEG2_CUVID, mpeg2_cuvid); REGISTER_HWACCEL(MPEG2_XVMC, mpeg2_xvmc); REGISTER_HWACCEL(MPEG2_D3D11VA, mpeg2_d3d11va); REGISTER_HWACCEL(MPEG2_DXVA2, mpeg2_dxva2); @@ -96,6 +102,8 @@ void avcodec_register_all(void) REGISTER_HWACCEL(MPEG2_VAAPI, mpeg2_vaapi); REGISTER_HWACCEL(MPEG2_VDPAU, mpeg2_vdpau); REGISTER_HWACCEL(MPEG2_VIDEOTOOLBOX, mpeg2_videotoolbox); + REGISTER_HWACCEL(MPEG4_CUVID, mpeg4_cuvid); + REGISTER_HWACCEL(MPEG4_MEDIACODEC, mpeg4_mediacodec); REGISTER_HWACCEL(MPEG4_MMAL, mpeg4_mmal); REGISTER_HWACCEL(MPEG4_VAAPI, mpeg4_vaapi); REGISTER_HWACCEL(MPEG4_VDPAU, mpeg4_vdpau); @@ -108,9 +116,11 @@ void avcodec_register_all(void) REGISTER_HWACCEL(VC1_MMAL, vc1_mmal); REGISTER_HWACCEL(VC1_QSV, vc1_qsv); REGISTER_HWACCEL(VP8_CUVID, vp8_cuvid); + REGISTER_HWACCEL(VP8_MEDIACODEC, vp8_mediacodec); REGISTER_HWACCEL(VP9_CUVID, vp9_cuvid); REGISTER_HWACCEL(VP9_D3D11VA, vp9_d3d11va); REGISTER_HWACCEL(VP9_DXVA2, vp9_dxva2); + REGISTER_HWACCEL(VP9_MEDIACODEC, vp9_mediacodec); REGISTER_HWACCEL(VP9_VAAPI, vp9_vaapi); REGISTER_HWACCEL(WMV3_D3D11VA, wmv3_d3d11va); REGISTER_HWACCEL(WMV3_DXVA2, wmv3_dxva2); @@ -418,7 +428,7 @@ void avcodec_register_all(void) REGISTER_DECODER(MACE3, mace3); REGISTER_DECODER(MACE6, mace6); REGISTER_DECODER(METASOUND, metasound); - REGISTER_DECODER(MLP, mlp); + REGISTER_ENCDEC (MLP, mlp); REGISTER_DECODER(MP1, mp1); REGISTER_DECODER(MP1FLOAT, mp1float); REGISTER_ENCDEC (MP2, mp2); @@ -447,7 +457,7 @@ void avcodec_register_all(void) REGISTER_ENCDEC (SONIC, sonic); REGISTER_ENCODER(SONIC_LS, sonic_ls); REGISTER_DECODER(TAK, tak); - REGISTER_DECODER(TRUEHD, truehd); + REGISTER_ENCDEC (TRUEHD, truehd); REGISTER_DECODER(TRUESPEECH, truespeech); REGISTER_ENCDEC (TTA, tta); REGISTER_DECODER(TWINVQ, twinvq); @@ -486,6 +496,8 @@ void avcodec_register_all(void) REGISTER_ENCDEC (PCM_S32BE, pcm_s32be); REGISTER_ENCDEC (PCM_S32LE, pcm_s32le); REGISTER_ENCDEC (PCM_S32LE_PLANAR, pcm_s32le_planar); + REGISTER_ENCDEC (PCM_S64BE, pcm_s64be); + REGISTER_ENCDEC (PCM_S64LE, pcm_s64le); REGISTER_ENCDEC (PCM_U8, pcm_u8); REGISTER_ENCDEC (PCM_U16BE, pcm_u16be); REGISTER_ENCDEC (PCM_U16LE, pcm_u16le); @@ -585,7 +597,6 @@ void avcodec_register_all(void) REGISTER_DECODER(QDMC_AT, qdmc_at); REGISTER_DECODER(QDM2_AT, qdm2_at); REGISTER_DECODER(LIBCELT, libcelt); - REGISTER_ENCODER(LIBFAAC, libfaac); REGISTER_ENCDEC (LIBFDK_AAC, libfdk_aac); REGISTER_ENCDEC (LIBGSM, libgsm); REGISTER_ENCDEC (LIBGSM_MS, libgsm_ms); @@ -622,7 +633,8 @@ void avcodec_register_all(void) /* external libraries, that shouldn't be used by default if one of the * above is available */ - REGISTER_ENCODER(LIBOPENH264, libopenh264); + REGISTER_ENCDEC (LIBOPENH264, libopenh264); + REGISTER_DECODER(H263_CUVID, h263_cuvid); REGISTER_DECODER(H264_CUVID, h264_cuvid); REGISTER_ENCODER(H264_NVENC, h264_nvenc); REGISTER_ENCODER(H264_OMX, h264_omx); @@ -635,15 +647,23 @@ void avcodec_register_all(void) REGISTER_ENCODER(NVENC_HEVC, nvenc_hevc); #endif REGISTER_DECODER(HEVC_CUVID, hevc_cuvid); + REGISTER_DECODER(HEVC_MEDIACODEC, hevc_mediacodec); REGISTER_ENCODER(HEVC_NVENC, hevc_nvenc); REGISTER_ENCODER(HEVC_QSV, hevc_qsv); REGISTER_ENCODER(HEVC_VAAPI, hevc_vaapi); REGISTER_ENCODER(LIBKVAZAAR, libkvazaar); + REGISTER_DECODER(MJPEG_CUVID, mjpeg_cuvid); REGISTER_ENCODER(MJPEG_VAAPI, mjpeg_vaapi); + REGISTER_DECODER(MPEG1_CUVID, mpeg1_cuvid); + REGISTER_DECODER(MPEG2_CUVID, mpeg2_cuvid); REGISTER_ENCODER(MPEG2_QSV, mpeg2_qsv); + REGISTER_DECODER(MPEG4_CUVID, mpeg4_cuvid); + REGISTER_DECODER(MPEG4_MEDIACODEC, mpeg4_mediacodec); REGISTER_DECODER(VC1_CUVID, vc1_cuvid); REGISTER_DECODER(VP8_CUVID, vp8_cuvid); + REGISTER_DECODER(VP8_MEDIACODEC, vp8_mediacodec); REGISTER_DECODER(VP9_CUVID, vp9_cuvid); + REGISTER_DECODER(VP9_MEDIACODEC, vp9_mediacodec); /* parsers */ REGISTER_PARSER(AAC, aac); diff --git a/media/ffvpx/libavcodec/avcodec.h b/media/ffvpx/libavcodec/avcodec.h index 39713ed76..e5e7f4225 100644 --- a/media/ffvpx/libavcodec/avcodec.h +++ b/media/ffvpx/libavcodec/avcodec.h @@ -43,7 +43,9 @@ #include "version.h" /** - * @defgroup libavc Encoding/Decoding Library + * @defgroup libavc libavcodec + * Encoding/Decoding Library + * * @{ * * @defgroup lavc_decoding Decoding @@ -443,9 +445,9 @@ enum AVCodecID { AV_CODEC_ID_PCM_S24LE_PLANAR, AV_CODEC_ID_PCM_S32LE_PLANAR, AV_CODEC_ID_PCM_S16BE_PLANAR, - /* new PCM "codecs" should be added right below this line starting with - * an explicit value of for example 0x10800 - */ + + AV_CODEC_ID_PCM_S64LE = 0x10800, + AV_CODEC_ID_PCM_S64BE, /* various ADPCM codecs */ AV_CODEC_ID_ADPCM_IMA_QT = 0x11000, @@ -629,6 +631,7 @@ enum AVCodecID { AV_CODEC_ID_FIRST_UNKNOWN = 0x18000, ///< A dummy ID pointing at the start of various fake codecs. AV_CODEC_ID_TTF = 0x18000, + AV_CODEC_ID_SCTE_35, ///< Contain timestamp estimated through PCR of program stream. AV_CODEC_ID_BINTEXT = 0x18800, AV_CODEC_ID_XBIN, AV_CODEC_ID_IDF, @@ -1033,6 +1036,16 @@ typedef struct RcOverride{ * Audio encoder supports receiving a different number of samples in each call. */ #define AV_CODEC_CAP_VARIABLE_FRAME_SIZE (1 << 16) +/** + * Decoder is not a preferred choice for probing. + * This indicates that the decoder is not a good choice for probing. + * It could for example be an expensive to spin up hardware decoder, + * or it could simply not provide a lot of useful information about + * the stream. + * A decoder marked with this flag should only be used as last resort + * choice for probing. + */ +#define AV_CODEC_CAP_AVOID_PROBING (1 << 17) /** * Codec is intra only. */ @@ -1348,6 +1361,14 @@ typedef struct AVCPBProperties { */ enum AVPacketSideDataType { AV_PKT_DATA_PALETTE, + + /** + * The AV_PKT_DATA_NEW_EXTRADATA is used to notify the codec or the format + * that the extradata buffer was changed and the receiving side should + * act upon it appropriately. The new extradata is embedded in the side + * data buffer and should be immediately used for processing the current + * frame or packet. + */ AV_PKT_DATA_NEW_EXTRADATA, /** @@ -1611,6 +1632,12 @@ typedef struct AVPacket { } AVPacket; #define AV_PKT_FLAG_KEY 0x0001 ///< The packet contains a keyframe #define AV_PKT_FLAG_CORRUPT 0x0002 ///< The packet content is corrupted +/** + * Flag is used to discard packets which are required to maintain valid + * decoder state but are not required for output and should be dropped + * after decoding. + **/ +#define AV_PKT_FLAG_DISCARD 0x0004 enum AVSideDataParamChangeFlags { AV_SIDE_DATA_PARAM_CHANGE_CHANNEL_COUNT = 0x0001, @@ -2083,22 +2110,23 @@ typedef struct AVCodecContext { * - decoding: unused */ int ildct_cmp; -#define FF_CMP_SAD 0 -#define FF_CMP_SSE 1 -#define FF_CMP_SATD 2 -#define FF_CMP_DCT 3 -#define FF_CMP_PSNR 4 -#define FF_CMP_BIT 5 -#define FF_CMP_RD 6 -#define FF_CMP_ZERO 7 -#define FF_CMP_VSAD 8 -#define FF_CMP_VSSE 9 -#define FF_CMP_NSSE 10 -#define FF_CMP_W53 11 -#define FF_CMP_W97 12 -#define FF_CMP_DCTMAX 13 -#define FF_CMP_DCT264 14 -#define FF_CMP_CHROMA 256 +#define FF_CMP_SAD 0 +#define FF_CMP_SSE 1 +#define FF_CMP_SATD 2 +#define FF_CMP_DCT 3 +#define FF_CMP_PSNR 4 +#define FF_CMP_BIT 5 +#define FF_CMP_RD 6 +#define FF_CMP_ZERO 7 +#define FF_CMP_VSAD 8 +#define FF_CMP_VSSE 9 +#define FF_CMP_NSSE 10 +#define FF_CMP_W53 11 +#define FF_CMP_W97 12 +#define FF_CMP_DCTMAX 13 +#define FF_CMP_DCT264 14 +#define FF_CMP_MEDIAN_SAD 15 +#define FF_CMP_CHROMA 256 /** * ME diamond size & shape @@ -2850,6 +2878,7 @@ typedef struct AVCodecContext { #define FF_BUG_DC_CLIP 4096 #define FF_BUG_MS 8192 ///< Work around various bugs in Microsoft's broken decoders. #define FF_BUG_TRUNCATED 16384 +#define FF_BUG_IEDGE 32768 /** * strictly follow the standard (MPEG-4, ...). @@ -3165,6 +3194,13 @@ typedef struct AVCodecContext { #define FF_PROFILE_MPEG2_AAC_LOW 128 #define FF_PROFILE_MPEG2_AAC_HE 131 +#define FF_PROFILE_DNXHD 0 +#define FF_PROFILE_DNXHR_LB 1 +#define FF_PROFILE_DNXHR_SQ 2 +#define FF_PROFILE_DNXHR_HQ 3 +#define FF_PROFILE_DNXHR_HQX 4 +#define FF_PROFILE_DNXHR_444 5 + #define FF_PROFILE_DTS 20 #define FF_PROFILE_DTS_ES 30 #define FF_PROFILE_DTS_96_24 40 @@ -3189,8 +3225,10 @@ typedef struct AVCodecContext { #define FF_PROFILE_H264_HIGH 100 #define FF_PROFILE_H264_HIGH_10 110 #define FF_PROFILE_H264_HIGH_10_INTRA (110|FF_PROFILE_H264_INTRA) +#define FF_PROFILE_H264_MULTIVIEW_HIGH 118 #define FF_PROFILE_H264_HIGH_422 122 #define FF_PROFILE_H264_HIGH_422_INTRA (122|FF_PROFILE_H264_INTRA) +#define FF_PROFILE_H264_STEREO_HIGH 128 #define FF_PROFILE_H264_HIGH_444 144 #define FF_PROFILE_H264_HIGH_444_PREDICTIVE 244 #define FF_PROFILE_H264_HIGH_444_INTRA (244|FF_PROFILE_H264_INTRA) @@ -3482,15 +3520,25 @@ typedef struct AVCodecContext { int nb_coded_side_data; /** - * Encoding only. + * A reference to the AVHWFramesContext describing the input (for encoding) + * or output (decoding) frames. The reference is set by the caller and + * afterwards owned (and freed) by libavcodec. + * + * - decoding: This field should be set by the caller from the get_format() + * callback. The previous reference (if any) will always be + * unreffed by libavcodec before the get_format() call. + * + * If the default get_buffer2() is used with a hwaccel pixel + * format, then this AVHWFramesContext will be used for + * allocating the frame buffers. * - * For hardware encoders configured to use a hwaccel pixel format, this - * field should be set by the caller to a reference to the AVHWFramesContext - * describing input frames. AVHWFramesContext.format must be equal to - * AVCodecContext.pix_fmt. + * - encoding: For hardware encoders configured to use a hwaccel pixel + * format, this field should be set by the caller to a reference + * to the AVHWFramesContext describing input frames. + * AVHWFramesContext.format must be equal to + * AVCodecContext.pix_fmt. * - * This field should be set before avcodec_open2() is called and is - * afterwards owned and managed by libavcodec. + * This field should be set before avcodec_open2() is called. */ AVBufferRef *hw_frames_ctx; @@ -3505,6 +3553,17 @@ typedef struct AVCodecContext { #define FF_SUB_TEXT_FMT_ASS_WITH_TIMINGS 1 #endif + /** + * Audio only. The amount of padding (in samples) appended by the encoder to + * the end of the audio. I.e. this number of decoded samples must be + * discarded by the caller from the end of the stream to get the original + * audio without any trailing padding. + * + * - decoding: unused + * - encoding: unused + */ + int trailing_padding; + } AVCodecContext; AVRational av_codec_get_pkt_timebase (const AVCodecContext *avctx); @@ -5115,7 +5174,10 @@ AVCodecParserContext *av_parser_init(int codec_id); * @param poutbuf set to pointer to parsed buffer or NULL if not yet finished. * @param poutbuf_size set to size of parsed buffer or zero if not yet finished. * @param buf input buffer. - * @param buf_size input length, to signal EOF, this should be 0 (so that the last frame can be output). + * @param buf_size buffer size in bytes without the padding. I.e. the full buffer + size is assumed to be buf_size + AV_INPUT_BUFFER_PADDING_SIZE. + To signal EOF, this should be 0 (so that the last frame + can be output). * @param pts input presentation timestamp. * @param dts input decoding timestamp. * @param pos input byte position in stream. @@ -5505,15 +5567,8 @@ enum AVPixelFormat avcodec_find_best_pix_fmt_of_2(enum AVPixelFormat dst_pix_fmt enum AVPixelFormat src_pix_fmt, int has_alpha, int *loss_ptr); attribute_deprecated -#if AV_HAVE_INCOMPATIBLE_LIBAV_ABI -enum AVPixelFormat avcodec_find_best_pix_fmt2(const enum AVPixelFormat *pix_fmt_list, - enum AVPixelFormat src_pix_fmt, - int has_alpha, int *loss_ptr); -#else enum AVPixelFormat avcodec_find_best_pix_fmt2(enum AVPixelFormat dst_pix_fmt1, enum AVPixelFormat dst_pix_fmt2, enum AVPixelFormat src_pix_fmt, int has_alpha, int *loss_ptr); -#endif - enum AVPixelFormat avcodec_default_get_format(struct AVCodecContext *s, const enum AVPixelFormat * fmt); @@ -5881,7 +5936,8 @@ int av_bsf_init(AVBSFContext *ctx); * av_bsf_receive_packet() repeatedly until it returns AVERROR(EAGAIN) or * AVERROR_EOF. * - * @param pkt the packet to filter. The bitstream filter will take ownership of + * @param pkt the packet to filter. pkt must contain some payload (i.e data or + * side data must be present in pkt). The bitstream filter will take ownership of * the packet and reset the contents of pkt. pkt is not touched if an error occurs. * This parameter may be NULL, which signals the end of the stream (i.e. no more * packets will be sent). That will cause the filter to output any packets it @@ -5931,6 +5987,91 @@ void av_bsf_free(AVBSFContext **ctx); */ const AVClass *av_bsf_get_class(void); +/** + * Structure for chain/list of bitstream filters. + * Empty list can be allocated by av_bsf_list_alloc(). + */ +typedef struct AVBSFList AVBSFList; + +/** + * Allocate empty list of bitstream filters. + * The list must be later freed by av_bsf_list_free() + * or finalized by av_bsf_list_finalize(). + * + * @return Pointer to @ref AVBSFList on success, NULL in case of failure + */ +AVBSFList *av_bsf_list_alloc(void); + +/** + * Free list of bitstream filters. + * + * @param lst Pointer to pointer returned by av_bsf_list_alloc() + */ +void av_bsf_list_free(AVBSFList **lst); + +/** + * Append bitstream filter to the list of bitstream filters. + * + * @param lst List to append to + * @param bsf Filter context to be appended + * + * @return >=0 on success, negative AVERROR in case of failure + */ +int av_bsf_list_append(AVBSFList *lst, AVBSFContext *bsf); + +/** + * Construct new bitstream filter context given it's name and options + * and append it to the list of bitstream filters. + * + * @param lst List to append to + * @param bsf_name Name of the bitstream filter + * @param options Options for the bitstream filter, can be set to NULL + * + * @return >=0 on success, negative AVERROR in case of failure + */ +int av_bsf_list_append2(AVBSFList *lst, const char * bsf_name, AVDictionary **options); +/** + * Finalize list of bitstream filters. + * + * This function will transform @ref AVBSFList to single @ref AVBSFContext, + * so the whole chain of bitstream filters can be treated as single filter + * freshly allocated by av_bsf_alloc(). + * If the call is successful, @ref AVBSFList structure is freed and lst + * will be set to NULL. In case of failure, caller is responsible for + * freeing the structure by av_bsf_list_free() + * + * @param lst Filter list structure to be transformed + * @param[out] bsf Pointer to be set to newly created @ref AVBSFContext structure + * representing the chain of bitstream filters + * + * @return >=0 on success, negative AVERROR in case of failure + */ +int av_bsf_list_finalize(AVBSFList **lst, AVBSFContext **bsf); + +/** + * Parse string describing list of bitstream filters and create single + * @ref AVBSFContext describing the whole chain of bitstream filters. + * Resulting @ref AVBSFContext can be treated as any other @ref AVBSFContext freshly + * allocated by av_bsf_alloc(). + * + * @param str String describing chain of bitstream filters in format + * `bsf1[=opt1=val1:opt2=val2][,bsf2]` + * @param[out] bsf Pointer to be set to newly created @ref AVBSFContext structure + * representing the chain of bitstream filters + * + * @return >=0 on success, negative AVERROR in case of failure + */ +int av_bsf_list_parse_str(const char *str, AVBSFContext **bsf); + +/** + * Get null/pass-through bitstream filter. + * + * @param[out] bsf Pointer to be set to new instance of pass-through bitstream filter + * + * @return + */ +int av_bsf_get_null_filter(AVBSFContext **bsf); + /* memory */ /** diff --git a/media/ffvpx/libavcodec/avpacket.c b/media/ffvpx/libavcodec/avpacket.c index 921868923..e5a8bdbe4 100644 --- a/media/ffvpx/libavcodec/avpacket.c +++ b/media/ffvpx/libavcodec/avpacket.c @@ -139,7 +139,8 @@ int av_grow_packet(AVPacket *pkt, int grow_by) pkt->buf = av_buffer_alloc(new_size); if (!pkt->buf) return AVERROR(ENOMEM); - memcpy(pkt->buf->data, pkt->data, pkt->size); + if (pkt->size > 0) + memcpy(pkt->buf->data, pkt->data, pkt->size); pkt->data = pkt->buf->data; } pkt->size += grow_by; @@ -198,6 +199,7 @@ static int copy_packet_data(AVPacket *pkt, const AVPacket *src, int dup) { pkt->data = NULL; pkt->side_data = NULL; + pkt->side_data_elems = 0; if (pkt->buf) { AVBufferRef *ref = av_buffer_ref(src->buf); if (!ref) @@ -207,9 +209,11 @@ static int copy_packet_data(AVPacket *pkt, const AVPacket *src, int dup) } else { DUP_DATA(pkt->data, src->data, pkt->size, 1, ALLOC_BUF); } - if (pkt->side_data_elems && dup) + if (src->side_data_elems && dup) { pkt->side_data = src->side_data; - if (pkt->side_data_elems && !dup) { + pkt->side_data_elems = src->side_data_elems; + } + if (src->side_data_elems && !dup) { return av_copy_packet_side_data(pkt, src); } return 0; @@ -291,16 +295,17 @@ FF_ENABLE_DEPRECATION_WARNINGS int av_packet_add_side_data(AVPacket *pkt, enum AVPacketSideDataType type, uint8_t *data, size_t size) { + AVPacketSideData *tmp; int elems = pkt->side_data_elems; if ((unsigned)elems + 1 > INT_MAX / sizeof(*pkt->side_data)) return AVERROR(ERANGE); - pkt->side_data = av_realloc(pkt->side_data, - (elems + 1) * sizeof(*pkt->side_data)); - if (!pkt->side_data) + tmp = av_realloc(pkt->side_data, (elems + 1) * sizeof(*tmp)); + if (!tmp) return AVERROR(ENOMEM); + pkt->side_data = tmp; pkt->side_data[elems].data = data; pkt->side_data[elems].size = size; pkt->side_data[elems].type = type; @@ -581,7 +586,8 @@ int av_packet_ref(AVPacket *dst, const AVPacket *src) ret = packet_alloc(&dst->buf, src->size); if (ret < 0) goto fail; - memcpy(dst->buf->data, src->data, src->size); + if (src->size) + memcpy(dst->buf->data, src->data, src->size); dst->data = dst->buf->data; } else { diff --git a/media/ffvpx/libavcodec/bsf.h b/media/ffvpx/libavcodec/bsf.h index 3435df5d8..af035eee4 100644 --- a/media/ffvpx/libavcodec/bsf.h +++ b/media/ffvpx/libavcodec/bsf.h @@ -28,6 +28,17 @@ */ int ff_bsf_get_packet(AVBSFContext *ctx, AVPacket **pkt); +/** + * Called by bitstream filters to get packet for filtering. + * The reference to packet is moved to provided packet structure. + * + * @param ctx pointer to AVBSFContext of filter + * @param pkt pointer to packet to move reference to + * + * @return 0>= on success, negative AVERROR in case of failure + */ +int ff_bsf_get_packet_ref(AVBSFContext *ctx, AVPacket *pkt); + const AVClass *ff_bsf_child_class_next(const AVClass *prev); #endif /* AVCODEC_BSF_H */ diff --git a/media/ffvpx/libavcodec/codec_desc.c b/media/ffvpx/libavcodec/codec_desc.c index 9d94b72ed..d862cc8f6 100644 --- a/media/ffvpx/libavcodec/codec_desc.c +++ b/media/ffvpx/libavcodec/codec_desc.c @@ -2281,7 +2281,7 @@ static const AVCodecDescriptor codec_descriptors[] = { .id = AV_CODEC_ID_COOK, .type = AVMEDIA_TYPE_AUDIO, .name = "cook", - .long_name = NULL_IF_CONFIG_SMALL("Cook / Cooker / Gecko (RealAudio G2)"), + .long_name = NULL_IF_CONFIG_SMALL("Cook / Cooker / Goanna (RealAudio G2)"), .props = AV_CODEC_PROP_LOSSY, }, { diff --git a/media/ffvpx/libavcodec/dummy_funcs.c b/media/ffvpx/libavcodec/dummy_funcs.c index a065c3576..200e1d266 100644 --- a/media/ffvpx/libavcodec/dummy_funcs.c +++ b/media/ffvpx/libavcodec/dummy_funcs.c @@ -62,6 +62,17 @@ AVHWAccel ff_vp8_cuvid_hwaccel; AVHWAccel ff_vc1_cuvid_hwaccel; AVHWAccel ff_hevc_cuvid_hwaccel; AVHWAccel ff_h264_cuvid_hwaccel; +/* Added by FFmpeg 3.2 */ +AVHWAccel ff_h263_cuvid_hwaccel; +AVHWAccel ff_mjpeg_cuvid_hwaccel; +AVHWAccel ff_mpeg1_cuvid_hwaccel; +AVHWAccel ff_mpeg2_cuvid_hwaccel; +AVHWAccel ff_mpeg4_cuvid_hwaccel; +AVHWAccel ff_h264_mediacodec_hwaccel; +AVHWAccel ff_hevc_mediacodec_hwaccel; +AVHWAccel ff_mpeg4_mediacodec_hwaccel; +AVHWAccel ff_vp8_mediacodec_hwaccel; +AVHWAccel ff_vp9_mediacodec_hwaccel; AVCodec ff_a64multi_encoder; AVCodec ff_a64multi5_encoder; @@ -713,6 +724,23 @@ AVCodec ff_sheervideo_decoder; AVCodec ff_magicyuv_decoder; AVCodec ff_m101_decoder; AVCodec ff_h264_mediacodec_decoder; +/* Added by FFmpeg 3.2 */ +AVCodec ff_vp9_mediacodec_decoder; +AVCodec ff_vp8_mediacodec_decoder; +AVCodec ff_mpeg4_mediacodec_decoder; +AVCodec ff_mpeg4_cuvid_decoder; +AVCodec ff_mpeg2_cuvid_decoder; +AVCodec ff_mpeg1_cuvid_decoder; +AVCodec ff_mjpeg_cuvid_decoder; +AVCodec ff_hevc_mediacodec_decoder; +AVCodec ff_h263_cuvid_decoder; +AVCodec ff_libopenh264_decoder; +AVCodec ff_pcm_s64le_decoder; +AVCodec ff_pcm_s64le_encoder; +AVCodec ff_pcm_s64be_decoder; +AVCodec ff_pcm_s64be_encoder; +AVCodec ff_truehd_encoder; +AVCodec ff_mlp_encoder; AVCodecParser ff_aac_parser; AVCodecParser ff_aac_latm_parser; diff --git a/media/ffvpx/libavcodec/h264dsp.h b/media/ffvpx/libavcodec/h264dsp.h index 7f24376b8..bcd76abcc 100644 --- a/media/ffvpx/libavcodec/h264dsp.h +++ b/media/ffvpx/libavcodec/h264dsp.h @@ -28,11 +28,12 @@ #define AVCODEC_H264DSP_H #include +#include -typedef void (*h264_weight_func)(uint8_t *block, int stride, int height, +typedef void (*h264_weight_func)(uint8_t *block, ptrdiff_t stride, int height, int log2_denom, int weight, int offset); typedef void (*h264_biweight_func)(uint8_t *dst, uint8_t *src, - int stride, int height, int log2_denom, + ptrdiff_t stride, int height, int log2_denom, int weightd, int weights, int offset); /** diff --git a/media/ffvpx/libavcodec/imgconvert.c b/media/ffvpx/libavcodec/imgconvert.c index 46fa7809b..1547f1896 100644 --- a/media/ffvpx/libavcodec/imgconvert.c +++ b/media/ffvpx/libavcodec/imgconvert.c @@ -57,19 +57,11 @@ enum AVPixelFormat avcodec_find_best_pix_fmt_of_2(enum AVPixelFormat dst_pix_fmt return av_find_best_pix_fmt_of_2(dst_pix_fmt1, dst_pix_fmt2, src_pix_fmt, has_alpha, loss_ptr); } -#if AV_HAVE_INCOMPATIBLE_LIBAV_ABI -enum AVPixelFormat avcodec_find_best_pix_fmt2(const enum AVPixelFormat *pix_fmt_list, - enum AVPixelFormat src_pix_fmt, - int has_alpha, int *loss_ptr){ - return avcodec_find_best_pix_fmt_of_list(pix_fmt_list, src_pix_fmt, has_alpha, loss_ptr); -} -#else enum AVPixelFormat avcodec_find_best_pix_fmt2(enum AVPixelFormat dst_pix_fmt1, enum AVPixelFormat dst_pix_fmt2, enum AVPixelFormat src_pix_fmt, int has_alpha, int *loss_ptr) { return avcodec_find_best_pix_fmt_of_2(dst_pix_fmt1, dst_pix_fmt2, src_pix_fmt, has_alpha, loss_ptr); } -#endif enum AVPixelFormat avcodec_find_best_pix_fmt_of_list(const enum AVPixelFormat *pix_fmt_list, enum AVPixelFormat src_pix_fmt, diff --git a/media/ffvpx/libavcodec/internal.h b/media/ffvpx/libavcodec/internal.h index 000fe263c..35b9630b5 100644 --- a/media/ffvpx/libavcodec/internal.h +++ b/media/ffvpx/libavcodec/internal.h @@ -173,6 +173,7 @@ typedef struct AVCodecInternal { int buffer_pkt_valid; // encoding: packet without data can be valid AVFrame *buffer_frame; int draining_done; + int showed_multi_packet_warning; } AVCodecInternal; struct AVCodecDefault { diff --git a/media/ffvpx/libavcodec/me_cmp.h b/media/ffvpx/libavcodec/me_cmp.h index a3603ec2c..5666f59ad 100644 --- a/media/ffvpx/libavcodec/me_cmp.h +++ b/media/ffvpx/libavcodec/me_cmp.h @@ -76,6 +76,7 @@ typedef struct MECmpContext { me_cmp_func frame_skip_cmp[6]; // only width 8 used me_cmp_func pix_abs[2][4]; + me_cmp_func median_sad[2]; } MECmpContext; void ff_me_cmp_init_static(void); diff --git a/media/ffvpx/libavcodec/mpegvideo.h b/media/ffvpx/libavcodec/mpegvideo.h index a1f3d4bd9..c82fa3e1a 100644 --- a/media/ffvpx/libavcodec/mpegvideo.h +++ b/media/ffvpx/libavcodec/mpegvideo.h @@ -594,7 +594,8 @@ enum rc_strategy { { "nsse", "Noise preserving sum of squared differences", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_NSSE }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \ { "dct264", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_DCT264 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \ { "dctmax", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_DCTMAX }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \ -{ "chroma", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_CHROMA }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" } +{ "chroma", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_CHROMA }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \ +{ "msad", "Sum of absolute differences, median predicted", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_MEDIAN_SAD }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" } #ifndef FF_MPV_OFFSET #define FF_MPV_OFFSET(x) offsetof(MpegEncContext, x) diff --git a/media/ffvpx/libavcodec/parser.c b/media/ffvpx/libavcodec/parser.c index 2c8fc6904..30cfc55cb 100644 --- a/media/ffvpx/libavcodec/parser.c +++ b/media/ffvpx/libavcodec/parser.c @@ -182,6 +182,11 @@ int av_parser_parse2(AVCodecParserContext *s, AVCodecContext *avctx, index = s->parser->parser_parse(s, avctx, (const uint8_t **) poutbuf, poutbuf_size, buf, buf_size); av_assert0(index > -0x20000000); // The API does not allow returning AVERROR codes +#define FILL(name) if(s->name > 0 && avctx->name <= 0) avctx->name = s->name + if (avctx->codec_type == AVMEDIA_TYPE_VIDEO) { + FILL(field_order); + } + /* update the file pointer */ if (*poutbuf_size) { /* fill the data for the current frame */ diff --git a/media/ffvpx/libavcodec/raw.c b/media/ffvpx/libavcodec/raw.c index bfa2537b5..f73a134a9 100644 --- a/media/ffvpx/libavcodec/raw.c +++ b/media/ffvpx/libavcodec/raw.c @@ -31,6 +31,7 @@ const PixelFormatTag ff_raw_pix_fmt_tags[] = { { AV_PIX_FMT_YUV420P, MKTAG('I', '4', '2', '0') }, /* Planar formats */ { AV_PIX_FMT_YUV420P, MKTAG('I', 'Y', 'U', 'V') }, + { AV_PIX_FMT_YUV420P, MKTAG('y', 'v', '1', '2') }, { AV_PIX_FMT_YUV420P, MKTAG('Y', 'V', '1', '2') }, { AV_PIX_FMT_YUV410P, MKTAG('Y', 'U', 'V', '9') }, { AV_PIX_FMT_YUV410P, MKTAG('Y', 'V', 'U', '9') }, @@ -187,6 +188,14 @@ const PixelFormatTag ff_raw_pix_fmt_tags[] = { { AV_PIX_FMT_GBRP16LE, MKTAG('G', '3', 00 , 16 ) }, { AV_PIX_FMT_GBRP16BE, MKTAG(16 , 00 , '3', 'G') }, + { AV_PIX_FMT_GBRAP, MKTAG('G', '4', 00 , 8 ) }, + { AV_PIX_FMT_GBRAP10LE, MKTAG('G', '4', 00 , 10 ) }, + { AV_PIX_FMT_GBRAP10BE, MKTAG(10 , 00 , '4', 'G') }, + { AV_PIX_FMT_GBRAP12LE, MKTAG('G', '4', 00 , 12 ) }, + { AV_PIX_FMT_GBRAP12BE, MKTAG(12 , 00 , '4', 'G') }, + { AV_PIX_FMT_GBRAP16LE, MKTAG('G', '4', 00 , 16 ) }, + { AV_PIX_FMT_GBRAP16BE, MKTAG(16 , 00 , '4', 'G') }, + { AV_PIX_FMT_XYZ12LE, MKTAG('X', 'Y', 'Z' , 36 ) }, { AV_PIX_FMT_XYZ12BE, MKTAG(36 , 'Z' , 'Y', 'X') }, @@ -224,6 +233,7 @@ const PixelFormatTag ff_raw_pix_fmt_tags[] = { { AV_PIX_FMT_ABGR, MKTAG('A', 'B', 'G', 'R') }, { AV_PIX_FMT_GRAY16BE,MKTAG('b', '1', '6', 'g') }, { AV_PIX_FMT_RGB48BE, MKTAG('b', '4', '8', 'r') }, + { AV_PIX_FMT_RGBA64BE,MKTAG('b', '6', '4', 'a') }, /* vlc */ { AV_PIX_FMT_YUV410P, MKTAG('I', '4', '1', '0') }, diff --git a/media/ffvpx/libavcodec/utils.c b/media/ffvpx/libavcodec/utils.c index f7adb525f..87de15fc6 100644 --- a/media/ffvpx/libavcodec/utils.c +++ b/media/ffvpx/libavcodec/utils.c @@ -724,6 +724,9 @@ int avcodec_default_get_buffer2(AVCodecContext *avctx, AVFrame *frame, int flags { int ret; + if (avctx->hw_frames_ctx) + return av_hwframe_get_buffer(avctx->hw_frames_ctx, frame, 0); + if ((ret = update_frame_pool(avctx, frame)) < 0) return ret; @@ -765,7 +768,12 @@ int ff_init_buffer_info(AVCodecContext *avctx, AVFrame *frame) }; if (pkt) { + frame->pts = pkt->pts; +#if FF_API_PKT_PTS +FF_DISABLE_DEPRECATION_WARNINGS frame->pkt_pts = pkt->pts; +FF_ENABLE_DEPRECATION_WARNINGS +#endif av_frame_set_pkt_pos (frame, pkt->pos); av_frame_set_pkt_duration(frame, pkt->duration); av_frame_set_pkt_size (frame, pkt->size); @@ -784,8 +792,19 @@ int ff_init_buffer_info(AVCodecContext *avctx, AVFrame *frame) } } add_metadata_from_side_data(pkt, frame); + + if (pkt->flags & AV_PKT_FLAG_DISCARD) { + frame->flags |= AV_FRAME_FLAG_DISCARD; + } else { + frame->flags = (frame->flags & ~AV_FRAME_FLAG_DISCARD); + } } else { + frame->pts = AV_NOPTS_VALUE; +#if FF_API_PKT_PTS +FF_DISABLE_DEPRECATION_WARNINGS frame->pkt_pts = AV_NOPTS_VALUE; +FF_ENABLE_DEPRECATION_WARNINGS +#endif av_frame_set_pkt_pos (frame, -1); av_frame_set_pkt_duration(frame, 0); av_frame_set_pkt_size (frame, -1); @@ -991,6 +1010,7 @@ int avcodec_default_execute(AVCodecContext *c, int (*func)(AVCodecContext *c2, v if (ret) ret[i] = r; } + emms_c(); return 0; } @@ -1003,6 +1023,7 @@ int avcodec_default_execute2(AVCodecContext *c, int (*func)(AVCodecContext *c2, if (ret) ret[i] = r; } + emms_c(); return 0; } @@ -1113,6 +1134,8 @@ int ff_get_format(AVCodecContext *avctx, const enum AVPixelFormat *fmt) av_freep(&avctx->internal->hwaccel_priv_data); avctx->hwaccel = NULL; + av_buffer_unref(&avctx->hw_frames_ctx); + ret = avctx->get_format(avctx, choices); desc = av_pix_fmt_desc_get(ret); @@ -1128,6 +1151,16 @@ int ff_get_format(AVCodecContext *avctx, const enum AVPixelFormat *fmt) break; #endif + if (avctx->hw_frames_ctx) { + AVHWFramesContext *hw_frames_ctx = (AVHWFramesContext*)avctx->hw_frames_ctx->data; + if (hw_frames_ctx->format != ret) { + av_log(avctx, AV_LOG_ERROR, "Format returned from get_buffer() " + "does not match the format of provided AVHWFramesContext\n"); + ret = AV_PIX_FMT_NONE; + break; + } + } + if (!setup_hwaccel(avctx, ret, desc->name)) break; @@ -1389,10 +1422,9 @@ int attribute_align_arg avcodec_open2(AVCodecContext *avctx, const AVCodec *code avctx->thread_count = 1; if (avctx->codec->max_lowres < avctx->lowres || avctx->lowres < 0) { - av_log(avctx, AV_LOG_ERROR, "The maximum value for lowres supported by the decoder is %d\n", + av_log(avctx, AV_LOG_WARNING, "The maximum value for lowres supported by the decoder is %d\n", avctx->codec->max_lowres); - ret = AVERROR(EINVAL); - goto free_and_end; + avctx->lowres = avctx->codec->max_lowres; } #if FF_API_VISMV @@ -1962,6 +1994,8 @@ int attribute_align_arg avcodec_encode_video2(AVCodecContext *avctx, ret = avctx->codec->encode2(avctx, avpkt, frame, got_packet_ptr); av_assert0(ret <= 0); + emms_c(); + if (avpkt->data && avpkt->data == avctx->internal->byte_buffer) { needs_realloc = 0; if (user_pkt.data) { @@ -1999,7 +2033,6 @@ int attribute_align_arg avcodec_encode_video2(AVCodecContext *avctx, if (ret < 0 || !*got_packet_ptr) av_packet_unref(avpkt); - emms_c(); return ret; } @@ -2023,7 +2056,7 @@ int avcodec_encode_subtitle(AVCodecContext *avctx, uint8_t *buf, int buf_size, * which case the output will as well. * * @param pts the pts field of the decoded AVPacket, as passed through - * AVFrame.pkt_pts + * AVFrame.pts * @param dts the dts field of the decoded AVPacket * @return one of the input values, may be AV_NOPTS_VALUE */ @@ -2248,7 +2281,9 @@ fail: if(ret == tmp.size) ret = avpkt->size; } - + if (picture->flags & AV_FRAME_FLAG_DISCARD) { + *got_picture_ptr = 0; + } if (*got_picture_ptr) { if (!avctx->refcounted_frames) { int err = unrefcount_frame(avci, picture); @@ -2259,7 +2294,7 @@ fail: avctx->frame_number++; av_frame_set_best_effort_timestamp(picture, guess_correct_pts(avctx, - picture->pkt_pts, + picture->pts, picture->pkt_dts)); } else av_frame_unref(picture); @@ -2332,7 +2367,7 @@ int attribute_align_arg avcodec_decode_audio4(AVCodecContext *avctx, avctx->frame_number++; av_frame_set_best_effort_timestamp(frame, guess_correct_pts(avctx, - frame->pkt_pts, + frame->pts, frame->pkt_dts)); if (frame->format == AV_SAMPLE_FMT_NONE) frame->format = avctx->sample_fmt; @@ -2353,6 +2388,13 @@ int attribute_align_arg avcodec_decode_audio4(AVCodecContext *avctx, skip_reason = AV_RL8(side + 8); discard_reason = AV_RL8(side + 9); } + + if ((frame->flags & AV_FRAME_FLAG_DISCARD) && *got_frame_ptr && + !(avctx->flags2 & AV_CODEC_FLAG2_SKIP_MANUAL)) { + avctx->internal->skip_samples -= frame->nb_samples; + *got_frame_ptr = 0; + } + if (avctx->internal->skip_samples > 0 && *got_frame_ptr && !(avctx->flags2 & AV_CODEC_FLAG2_SKIP_MANUAL)) { if(frame->nb_samples <= avctx->internal->skip_samples){ @@ -2367,8 +2409,14 @@ int attribute_align_arg avcodec_decode_audio4(AVCodecContext *avctx, int64_t diff_ts = av_rescale_q(avctx->internal->skip_samples, (AVRational){1, avctx->sample_rate}, avctx->pkt_timebase); + if(frame->pts!=AV_NOPTS_VALUE) + frame->pts += diff_ts; +#if FF_API_PKT_PTS +FF_DISABLE_DEPRECATION_WARNINGS if(frame->pkt_pts!=AV_NOPTS_VALUE) frame->pkt_pts += diff_ts; +FF_ENABLE_DEPRECATION_WARNINGS +#endif if(frame->pkt_dts!=AV_NOPTS_VALUE) frame->pkt_dts += diff_ts; if (av_frame_get_pkt_duration(frame) >= diff_ts) @@ -2432,6 +2480,12 @@ fail: av_assert0(ret <= avpkt->size); + if (!avci->showed_multi_packet_warning && + ret >= 0 && ret != avpkt->size && !(avctx->codec->capabilities & AV_CODEC_CAP_SUBFRAMES)) { + av_log(avctx, AV_LOG_WARNING, "Multiple frames in a packet.\n"); + avci->showed_multi_packet_warning = 1; + } + return ret; } @@ -2784,6 +2838,9 @@ int attribute_align_arg avcodec_send_packet(AVCodecContext *avctx, const AVPacke if (avctx->internal->draining) return AVERROR_EOF; + if (avpkt && !avpkt->size && avpkt->data) + return AVERROR(EINVAL); + if (!avpkt || !avpkt->size) { avctx->internal->draining = 1; avpkt = NULL; @@ -2832,7 +2889,14 @@ int attribute_align_arg avcodec_receive_frame(AVCodecContext *avctx, AVFrame *fr if (avctx->codec->receive_frame) { if (avctx->internal->draining && !(avctx->codec->capabilities & AV_CODEC_CAP_DELAY)) return AVERROR_EOF; - return avctx->codec->receive_frame(avctx, frame); + ret = avctx->codec->receive_frame(avctx, frame); + if (ret >= 0) { + if (av_frame_get_best_effort_timestamp(frame) == AV_NOPTS_VALUE) { + av_frame_set_best_effort_timestamp(frame, + guess_correct_pts(avctx, frame->pts, frame->pkt_dts)); + } + } + return ret; } // Emulation via old API. @@ -3190,7 +3254,21 @@ void avcodec_string(char *buf, int buf_size, AVCodecContext *enc, int encode) av_get_colorspace_name(enc->colorspace)); } - if (av_log_get_level() >= AV_LOG_DEBUG && + if (enc->field_order != AV_FIELD_UNKNOWN) { + const char *field_order = "progressive"; + if (enc->field_order == AV_FIELD_TT) + field_order = "top first"; + else if (enc->field_order == AV_FIELD_BB) + field_order = "bottom first"; + else if (enc->field_order == AV_FIELD_TB) + field_order = "top coded first (swapped)"; + else if (enc->field_order == AV_FIELD_BT) + field_order = "bottom coded first (swapped)"; + + av_strlcatf(detail, sizeof(detail), "%s, ", field_order); + } + + if (av_log_get_level() >= AV_LOG_VERBOSE && enc->chroma_sample_location != AVCHROMA_LOC_UNSPECIFIED) av_strlcatf(detail, sizeof(detail), "%s, ", av_chroma_location_name(enc->chroma_sample_location)); @@ -3259,6 +3337,14 @@ void avcodec_string(char *buf, int buf_size, AVCodecContext *enc, int encode) && enc->bits_per_raw_sample != av_get_bytes_per_sample(enc->sample_fmt) * 8) snprintf(buf + strlen(buf), buf_size - strlen(buf), " (%d bit)", enc->bits_per_raw_sample); + if (av_log_get_level() >= AV_LOG_VERBOSE) { + if (enc->initial_padding) + snprintf(buf + strlen(buf), buf_size - strlen(buf), + ", delay %d", enc->initial_padding); + if (enc->trailing_padding) + snprintf(buf + strlen(buf), buf_size - strlen(buf), + ", padding %d", enc->trailing_padding); + } break; case AVMEDIA_TYPE_DATA: if (av_log_get_level() >= AV_LOG_DEBUG) { @@ -3416,6 +3502,8 @@ int av_get_exact_bits_per_sample(enum AVCodecID codec_id) return 32; case AV_CODEC_ID_PCM_F64BE: case AV_CODEC_ID_PCM_F64LE: + case AV_CODEC_ID_PCM_S64BE: + case AV_CODEC_ID_PCM_S64LE: return 64; default: return 0; @@ -3433,6 +3521,7 @@ enum AVCodecID av_get_pcm_codec(enum AVSampleFormat fmt, int be) [AV_SAMPLE_FMT_U8P ] = { AV_CODEC_ID_PCM_U8, AV_CODEC_ID_PCM_U8 }, [AV_SAMPLE_FMT_S16P] = { AV_CODEC_ID_PCM_S16LE, AV_CODEC_ID_PCM_S16BE }, [AV_SAMPLE_FMT_S32P] = { AV_CODEC_ID_PCM_S32LE, AV_CODEC_ID_PCM_S32BE }, + [AV_SAMPLE_FMT_S64P] = { AV_CODEC_ID_PCM_S64LE, AV_CODEC_ID_PCM_S64BE }, [AV_SAMPLE_FMT_FLTP] = { AV_CODEC_ID_PCM_F32LE, AV_CODEC_ID_PCM_F32BE }, [AV_SAMPLE_FMT_DBLP] = { AV_CODEC_ID_PCM_F64LE, AV_CODEC_ID_PCM_F64BE }, }; @@ -4104,14 +4193,15 @@ int avcodec_parameters_from_context(AVCodecParameters *par, par->video_delay = codec->has_b_frames; break; case AVMEDIA_TYPE_AUDIO: - par->format = codec->sample_fmt; - par->channel_layout = codec->channel_layout; - par->channels = codec->channels; - par->sample_rate = codec->sample_rate; - par->block_align = codec->block_align; - par->frame_size = codec->frame_size; - par->initial_padding = codec->initial_padding; - par->seek_preroll = codec->seek_preroll; + par->format = codec->sample_fmt; + par->channel_layout = codec->channel_layout; + par->channels = codec->channels; + par->sample_rate = codec->sample_rate; + par->block_align = codec->block_align; + par->frame_size = codec->frame_size; + par->initial_padding = codec->initial_padding; + par->trailing_padding = codec->trailing_padding; + par->seek_preroll = codec->seek_preroll; break; case AVMEDIA_TYPE_SUBTITLE: par->width = codec->width; @@ -4158,15 +4248,16 @@ int avcodec_parameters_to_context(AVCodecContext *codec, codec->has_b_frames = par->video_delay; break; case AVMEDIA_TYPE_AUDIO: - codec->sample_fmt = par->format; - codec->channel_layout = par->channel_layout; - codec->channels = par->channels; - codec->sample_rate = par->sample_rate; - codec->block_align = par->block_align; - codec->frame_size = par->frame_size; - codec->delay = - codec->initial_padding = par->initial_padding; - codec->seek_preroll = par->seek_preroll; + codec->sample_fmt = par->format; + codec->channel_layout = par->channel_layout; + codec->channels = par->channels; + codec->sample_rate = par->sample_rate; + codec->block_align = par->block_align; + codec->frame_size = par->frame_size; + codec->delay = + codec->initial_padding = par->initial_padding; + codec->trailing_padding = par->trailing_padding; + codec->seek_preroll = par->seek_preroll; break; case AVMEDIA_TYPE_SUBTITLE: codec->width = par->width; diff --git a/media/ffvpx/libavcodec/version.h b/media/ffvpx/libavcodec/version.h index 4f6423b15..ec8837a4e 100644 --- a/media/ffvpx/libavcodec/version.h +++ b/media/ffvpx/libavcodec/version.h @@ -28,7 +28,7 @@ #include "libavutil/version.h" #define LIBAVCODEC_VERSION_MAJOR 57 -#define LIBAVCODEC_VERSION_MINOR 48 +#define LIBAVCODEC_VERSION_MINOR 64 #define LIBAVCODEC_VERSION_MICRO 101 #define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \ diff --git a/media/ffvpx/libavcodec/vp9.c b/media/ffvpx/libavcodec/vp9.c index 3b721495d..1aab6ba03 100644 --- a/media/ffvpx/libavcodec/vp9.c +++ b/media/ffvpx/libavcodec/vp9.c @@ -3991,7 +3991,12 @@ static int vp9_decode_frame(AVCodecContext *ctx, void *frame, } if ((res = av_frame_ref(frame, s->s.refs[ref].f)) < 0) return res; + ((AVFrame *)frame)->pts = pkt->pts; +#if FF_API_PKT_PTS +FF_DISABLE_DEPRECATION_WARNINGS ((AVFrame *)frame)->pkt_pts = pkt->pts; +FF_ENABLE_DEPRECATION_WARNINGS +#endif ((AVFrame *)frame)->pkt_dts = pkt->dts; for (i = 0; i < 8; i++) { if (s->next_refs[i].f->buf[0]) diff --git a/media/ffvpx/libavcodec/x86/h264_i386.h b/media/ffvpx/libavcodec/x86/h264_i386.h index 4dfbc3093..19cd12838 100644 --- a/media/ffvpx/libavcodec/x86/h264_i386.h +++ b/media/ffvpx/libavcodec/x86/h264_i386.h @@ -91,13 +91,13 @@ static int decode_significance_x86(CABACContext *c, int max_coeff, "sub %10, %1 \n\t" "mov %2, %0 \n\t" "movl %7, %%ecx \n\t" - "add %1, %%"REG_c" \n\t" + "add %1, %%"FF_REG_c" \n\t" "movl %%ecx, (%0) \n\t" "test $1, %4 \n\t" " jnz 5f \n\t" - "add"OPSIZE" $4, %2 \n\t" + "add"FF_OPSIZE" $4, %2 \n\t" "4: \n\t" "add $1, %1 \n\t" @@ -105,7 +105,7 @@ static int decode_significance_x86(CABACContext *c, int max_coeff, " jb 3b \n\t" "mov %2, %0 \n\t" "movl %7, %%ecx \n\t" - "add %1, %%"REG_c" \n\t" + "add %1, %%"FF_REG_c" \n\t" "movl %%ecx, (%0) \n\t" "5: \n\t" "add %9, %k0 \n\t" @@ -116,7 +116,7 @@ static int decode_significance_x86(CABACContext *c, int max_coeff, "i"(offsetof(CABACContext, bytestream)), "i"(offsetof(CABACContext, bytestream_end)) TABLES_ARG - : "%"REG_c, "memory" + : "%"FF_REG_c, "memory" ); return coeff_count; } @@ -183,7 +183,7 @@ static int decode_significance_8x8_x86(CABACContext *c, "test $1, %4 \n\t" " jnz 5f \n\t" - "add"OPSIZE" $4, %2 \n\t" + "add"FF_OPSIZE" $4, %2 \n\t" "4: \n\t" "add $1, %6 \n\t" @@ -202,7 +202,7 @@ static int decode_significance_8x8_x86(CABACContext *c, "i"(offsetof(CABACContext, bytestream)), "i"(offsetof(CABACContext, bytestream_end)), "i"(H264_LAST_COEFF_FLAG_OFFSET_8x8_OFFSET) TABLES_ARG - : "%"REG_c, "memory" + : "%"FF_REG_c, "memory" ); return coeff_count; } diff --git a/media/ffvpx/libavcodec/x86/vp9dsp_init.c b/media/ffvpx/libavcodec/x86/vp9dsp_init.c index 469a66171..cc781a009 100644 --- a/media/ffvpx/libavcodec/x86/vp9dsp_init.c +++ b/media/ffvpx/libavcodec/x86/vp9dsp_init.c @@ -114,6 +114,8 @@ itxfm_func(idct, idct, 32, sse2); itxfm_func(idct, idct, 32, ssse3); itxfm_func(idct, idct, 32, avx); itxfm_func(iwht, iwht, 4, mmx); +itxfm_func(idct, idct, 16, avx2); +itxfm_func(idct, idct, 32, avx2); #undef itxfm_func #undef itxfm_funcs @@ -124,6 +126,8 @@ void ff_vp9_loop_filter_v_##size1##_##size2##_##opt(uint8_t *dst, ptrdiff_t stri void ff_vp9_loop_filter_h_##size1##_##size2##_##opt(uint8_t *dst, ptrdiff_t stride, \ int E, int I, int H) +lpf_funcs(4, 8, mmxext); +lpf_funcs(8, 8, mmxext); lpf_funcs(16, 16, sse2); lpf_funcs(16, 16, ssse3); lpf_funcs(16, 16, avx); @@ -279,6 +283,10 @@ av_cold void ff_vp9dsp_init_x86(VP9DSPContext *dsp, int bpp, int bitexact) } if (EXTERNAL_MMXEXT(cpu_flags)) { + dsp->loop_filter_8[0][0] = ff_vp9_loop_filter_h_4_8_mmxext; + dsp->loop_filter_8[0][1] = ff_vp9_loop_filter_v_4_8_mmxext; + dsp->loop_filter_8[1][0] = ff_vp9_loop_filter_h_8_8_mmxext; + dsp->loop_filter_8[1][1] = ff_vp9_loop_filter_v_8_8_mmxext; init_subpel2(4, 0, 4, put, 8, mmxext); init_subpel2(4, 1, 4, avg, 8, mmxext); init_fpel_func(4, 1, 4, avg, _8, mmxext); @@ -382,6 +390,8 @@ av_cold void ff_vp9dsp_init_x86(VP9DSPContext *dsp, int bpp, int bitexact) init_fpel_func(0, 1, 64, avg, _8, avx2); if (ARCH_X86_64) { #if ARCH_X86_64 && HAVE_AVX2_EXTERNAL + dsp->itxfm_add[TX_16X16][DCT_DCT] = ff_vp9_idct_idct_16x16_add_avx2; + dsp->itxfm_add[TX_32X32][DCT_DCT] = ff_vp9_idct_idct_32x32_add_avx2; init_subpel3_32_64(0, put, 8, avx2); init_subpel3_32_64(1, avg, 8, avx2); #endif diff --git a/media/ffvpx/libavcodec/x86/vp9itxfm.asm b/media/ffvpx/libavcodec/x86/vp9itxfm.asm index 6d5008e33..57d6d353b 100644 --- a/media/ffvpx/libavcodec/x86/vp9itxfm.asm +++ b/media/ffvpx/libavcodec/x86/vp9itxfm.asm @@ -24,36 +24,36 @@ %include "libavutil/x86/x86util.asm" %include "vp9itxfm_template.asm" -SECTION_RODATA +SECTION_RODATA 32 %macro VP9_IDCT_COEFFS 2-3 0 const pw_m%1_%2 -times 4 dw -%1, %2 +times 8 dw -%1, %2 const pw_%2_%1 -times 4 dw %2, %1 +times 8 dw %2, %1 %if %3 == 1 const pw_m%2_m%1 -times 4 dw -%2, -%1 +times 8 dw -%2, -%1 %if %1 != %2 const pw_m%2_%1 -times 4 dw -%2, %1 +times 8 dw -%2, %1 const pw_%1_%2 -times 4 dw %1, %2 +times 8 dw %1, %2 %endif %endif %if %1 < 11585 -pw_m%1x2: times 8 dw -%1*2 +pw_m%1x2: times 16 dw -%1*2 %elif %1 > 11585 -pw_%1x2: times 8 dw %1*2 +pw_%1x2: times 16 dw %1*2 %else const pw_%1x2 -times 8 dw %1*2 +times 16 dw %1*2 %endif %if %2 != %1 -pw_%2x2: times 8 dw %2*2 +pw_%2x2: times 16 dw %2*2 %endif %endmacro @@ -127,16 +127,33 @@ SECTION .text %endmacro %macro VP9_STORE_2X 5-6 dstq ; reg1, reg2, tmp1, tmp2, zero, dst +%if mmsize == 32 + pmovzxbw m%3, [%6] + pmovzxbw m%4, [%6+strideq] +%else movh m%3, [%6] movh m%4, [%6+strideq] punpcklbw m%3, m%5 punpcklbw m%4, m%5 +%endif paddw m%3, m%1 paddw m%4, m%2 +%if mmsize == 32 + packuswb m%3, m%4 + ; Intel... + vpermq m%3, m%3, q3120 + mova [%6], xm%3 + vextracti128 [%6+strideq], m%3, 1 +%elif mmsize == 16 + packuswb m%3, m%4 + movh [%6], m%3 + movhps [%6+strideq], m%3 +%else packuswb m%3, m%5 packuswb m%4, m%5 movh [%6], m%3 movh [%6+strideq], m%4 +%endif %endmacro %macro ZERO_BLOCK 4 ; mem, stride, nnzcpl, zero_reg @@ -1421,6 +1438,180 @@ VP9_IDCT_IDCT_16x16_ADD_XMM sse2 VP9_IDCT_IDCT_16x16_ADD_XMM ssse3 VP9_IDCT_IDCT_16x16_ADD_XMM avx +%macro VP9_IDCT16_YMM_1D 0 + VP9_UNPACK_MULSUB_2W_4X 1, 15, 16305, 1606, [pd_8192], 0, 4 ; t8, t15 + VP9_UNPACK_MULSUB_2W_4X 9, 7, 10394, 12665, [pd_8192], 0, 4 ; t9, t14 + + SUMSUB_BA w, 9, 1, 0 ; t8, t9 + SUMSUB_BA w, 7, 15, 0 ; t15, t14 + + VP9_UNPACK_MULSUB_2W_4X 15, 1, 15137, 6270, [pd_8192], 0, 4 ; t9, t14 + + VP9_UNPACK_MULSUB_2W_4X 5, 11, 14449, 7723, [pd_8192], 0, 4 ; t10, t13 + VP9_UNPACK_MULSUB_2W_4X 13, 3, 4756, 15679, [pd_8192], 0, 4 ; t11, t12 + + SUMSUB_BA w, 5, 13, 0 ; t11, t10 + SUMSUB_BA w, 11, 3, 0 ; t12, t13 + + VP9_UNPACK_MULSUB_2W_4X 3, 13, 6270, m15137, [pd_8192], 0, 4 ; t10, t13 + + SUMSUB_BA w, 5, 9, 0 ; t8, t11 + SUMSUB_BA w, 3, 15, 0 ; t9, t10 + SUMSUB_BA w, 11, 7, 0 ; t15, t12 + SUMSUB_BA w, 13, 1, 0 ; t14, t13 + + SUMSUB_BA w, 15, 1, 0 + SUMSUB_BA w, 9, 7, 0 + pmulhrsw m1, [pw_11585x2] ; t10 + pmulhrsw m7, [pw_11585x2] ; t11 + pmulhrsw m9, [pw_11585x2] ; t12 + pmulhrsw m15, [pw_11585x2] ; t13 + + ; even (tx8x8) + mova m4, [blockq+128] + mova [blockq+128], m5 + VP9_UNPACK_MULSUB_2W_4X 4, 12, 15137, 6270, [pd_8192], 0, 5 ; t2, t3 + VP9_UNPACK_MULSUB_2W_4X 2, 14, 16069, 3196, [pd_8192], 0, 5 ; t4, t7 + VP9_UNPACK_MULSUB_2W_4X 10, 6, 9102, 13623, [pd_8192], 0, 5 ; t5, t6 + mova m0, [blockq+ 0] + SUMSUB_BA w, 8, 0, 5 + pmulhrsw m8, [pw_11585x2] ; t0 + pmulhrsw m0, [pw_11585x2] ; t1 + + SUMSUB_BA w, 10, 2, 5 ; t4, t5 + SUMSUB_BA w, 6, 14, 5 ; t7, t6 + SUMSUB_BA w, 12, 8, 5 ; t0, t3 + SUMSUB_BA w, 4, 0, 5 ; t1, t2 + + SUMSUB_BA w, 2, 14, 5 + pmulhrsw m14, [pw_11585x2] ; t5 + pmulhrsw m2, [pw_11585x2] ; t6 + + SUMSUB_BA w, 6, 12, 5 ; t0, t7 + SUMSUB_BA w, 2, 4, 5 ; t1, t6 + SUMSUB_BA w, 14, 0, 5 ; t2, t5 + SUMSUB_BA w, 10, 8, 5 ; t3, t4 + + ; final stage + SUMSUB_BA w, 11, 6, 5 ; out0, out15 + SUMSUB_BA w, 13, 2, 5 ; out1, out14 + SUMSUB_BA w, 15, 14, 5 ; out2, out13 + SUMSUB_BA w, 9, 10, 5 ; out3, out12 + SUMSUB_BA w, 7, 8, 5 ; out4, out11 + SUMSUB_BA w, 1, 0, 5 ; out5, out10 + SUMSUB_BA w, 3, 4, 5 ; out6, out9 + mova m5, [blockq+128] + mova [blockq+192], m3 + SUMSUB_BA w, 5, 12, 3 ; out7, out8 + + SWAP 0, 11, 8, 12, 10 + SWAP 1, 13, 14, 2, 15, 6, 3, 9, 4, 7, 5 +%endmacro + +; this is almost identical to VP9_STORE_2X, but it does two rows +; for slightly improved interleaving, and it omits vpermq since the +; input is DC so all values are identical +%macro VP9_STORE_YMM_DC_4X 6 ; reg, tmp1, tmp2, tmp3, tmp4, zero + mova xm%2, [dstq] + mova xm%4, [dstq+strideq*2] + vinserti128 m%2, m%2, [dstq+strideq], 1 + vinserti128 m%4, m%4, [dstq+stride3q], 1 + punpckhbw m%3, m%2, m%6 + punpcklbw m%2, m%6 + punpckhbw m%5, m%4, m%6 + punpcklbw m%4, m%6 + paddw m%3, m%1 + paddw m%2, m%1 + paddw m%5, m%1 + paddw m%4, m%1 + packuswb m%2, m%3 + packuswb m%4, m%5 + mova [dstq], xm%2 + mova [dstq+strideq*2], xm%4 + vextracti128 [dstq+strideq], m%2, 1 + vextracti128 [dstq+stride3q], m%4, 1 +%endmacro + +%if ARCH_X86_64 && HAVE_AVX2_EXTERNAL +INIT_YMM avx2 +cglobal vp9_idct_idct_16x16_add, 4, 4, 16, dst, stride, block, eob + cmp eobd, 1 ; faster path for when only DC is set + jg .idctfull + + ; dc-only + mova m1, [pw_11585x2] + vpbroadcastw m0, [blockq] + pmulhrsw m0, m1 + pmulhrsw m0, m1 + pxor m5, m5 + pmulhrsw m0, [pw_512] + movd [blockq], xm5 + + DEFINE_ARGS dst, stride, stride3, cnt + mov cntd, 4 + lea stride3q, [strideq*3] +.loop_dc: + VP9_STORE_YMM_DC_4X 0, 1, 2, 3, 4, 5 + lea dstq, [dstq+4*strideq] + dec cntd + jg .loop_dc + RET + + DEFINE_ARGS dst, stride, block, eob +.idctfull: + mova m1, [blockq+ 32] + mova m2, [blockq+ 64] + mova m3, [blockq+ 96] + mova m5, [blockq+160] + mova m6, [blockq+192] + mova m7, [blockq+224] + mova m8, [blockq+256] + mova m9, [blockq+288] + mova m10, [blockq+320] + mova m11, [blockq+352] + mova m12, [blockq+384] + mova m13, [blockq+416] + mova m14, [blockq+448] + mova m15, [blockq+480] + + VP9_IDCT16_YMM_1D + TRANSPOSE16x16W 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, \ + [blockq+192], [blockq+128], 1 + mova [blockq+ 0], m0 + VP9_IDCT16_YMM_1D + + mova [blockq+224], m7 + mova [blockq+480], m15 + pxor m15, m15 + + ; store + VP9_IDCT8_WRITEx2 0, 1, 6, 7, 15, [pw_512], 6 + lea dstq, [dstq+2*strideq] + VP9_IDCT8_WRITEx2 2, 3, 6, 7, 15, [pw_512], 6 + lea dstq, [dstq+2*strideq] + VP9_IDCT8_WRITEx2 4, 5, 6, 7, 15, [pw_512], 6 + lea dstq, [dstq+2*strideq] + mova m6, [blockq+192] + mova m7, [blockq+224] + SWAP 0, 15 + mova m15, [blockq+480] + VP9_IDCT8_WRITEx2 6, 7, 1, 2, 0, [pw_512], 6 + lea dstq, [dstq+2*strideq] + VP9_IDCT8_WRITEx2 8, 9, 1, 2, 0, [pw_512], 6 + lea dstq, [dstq+2*strideq] + VP9_IDCT8_WRITEx2 10, 11, 1, 2, 0, [pw_512], 6 + lea dstq, [dstq+2*strideq] + VP9_IDCT8_WRITEx2 12, 13, 1, 2, 0, [pw_512], 6 + lea dstq, [dstq+2*strideq] + VP9_IDCT8_WRITEx2 14, 15, 1, 2, 0, [pw_512], 6 + lea dstq, [dstq+2*strideq] + + ; at the end of the loop, m0 should still be zero + ; use that to zero out block coefficients + ZERO_BLOCK blockq, 32, 16, m0 + RET +%endif + ;--------------------------------------------------------------------------------------------- ; void vp9_iadst_iadst_16x16_add_(uint8_t *dst, ptrdiff_t stride, int16_t *block, int eob); ;--------------------------------------------------------------------------------------------- @@ -1801,7 +1992,12 @@ IADST16_FN iadst, IADST16, iadst, IADST16, avx ;--------------------------------------------------------------------------------------------- %macro VP9_IDCT32_1D 2-3 32 ; src, pass, nnzc -%assign %%str 16*%2*%2 +%if %2 == 1 +%assign %%str mmsize +%else +%assign %%str 64 +%endif + ; first do t0-15, this can be done identical to idct16x16 VP9_IDCT16_1D_START %1, %3/2, 64*2, tmpq, 2*%%str, 1 @@ -2096,17 +2292,125 @@ IADST16_FN iadst, IADST16, iadst, IADST16, avx mova m3, [tmpq+20*%%str] ; t5 mova m13, [tmpq+24*%%str] ; t6 - SUMSUB_BA w, 6, 8, 10 + SUMSUB_BA w, 6, 8, 10 mova [tmpq+ 3*%%str], m8 ; t15 - mova m10, [tmpq+28*%%str] ; t7 SUMSUB_BA w, 0, 9, 8 SUMSUB_BA w, 15, 12, 8 SUMSUB_BA w, 14, 11, 8 SUMSUB_BA w, 1, 2, 8 SUMSUB_BA w, 7, 3, 8 SUMSUB_BA w, 5, 13, 8 + mova m10, [tmpq+28*%%str] ; t7 SUMSUB_BA w, 4, 10, 8 +%if cpuflag(avx2) + ; the "shitty" about this idct is that the final pass does the outermost + ; interleave sumsubs (t0/31, t1/30, etc) but the tN for the 16x16 need + ; to be sequential, which means I need to load/store half of the sumsub + ; intermediates back to/from memory to get a 16x16 transpose going... + ; This would be easier if we had more (e.g. 32) YMM regs here. + mova [tmpq+ 7*%%str], m9 + mova [tmpq+11*%%str], m12 + mova [tmpq+15*%%str], m11 + mova [tmpq+19*%%str], m2 + mova [tmpq+23*%%str], m3 + mova [tmpq+27*%%str], m13 + mova [tmpq+31*%%str], m10 + mova [tmpq+12*%%str], m5 + + mova m13, [tmpq+30*%%str] ; t8 + mova m12, [tmpq+26*%%str] ; t9 + mova m11, [tmpq+22*%%str] ; t10 + mova m10, [tmpq+18*%%str] ; t11 + mova m9, [tmpq+17*%%str] ; t20 + mova m8, [tmpq+ 1*%%str] ; t21 + mova m3, [tmpq+25*%%str] ; t22 + mova m2, [tmpq+ 5*%%str] ; t23 + + SUMSUB_BA w, 9, 10, 5 + SUMSUB_BA w, 8, 11, 5 + SUMSUB_BA w, 3, 12, 5 + SUMSUB_BA w, 2, 13, 5 + mova [tmpq+ 1*%%str], m10 + mova [tmpq+ 5*%%str], m11 + mova [tmpq+17*%%str], m12 + mova [tmpq+25*%%str], m13 + + mova m13, [tmpq+14*%%str] ; t12 + mova m12, [tmpq+10*%%str] ; t13 + mova m11, [tmpq+ 9*%%str] ; t18 + mova m10, [tmpq+13*%%str] ; t19 + + SUMSUB_BA w, 11, 12, 5 + SUMSUB_BA w, 10, 13, 5 + mova [tmpq+ 9*%%str], m13 + mova [tmpq+13*%%str], m12 + mova [tmpq+10*%%str], m10 + mova [tmpq+14*%%str], m11 + mova m13, [tmpq+ 6*%%str] ; t14 + mova m12, [tmpq+ 2*%%str] ; t15 + mova m11, [tmpq+21*%%str] ; t16 + mova m10, [tmpq+29*%%str] ; t17 + SUMSUB_BA w, 11, 12, 5 + SUMSUB_BA w, 10, 13, 5 + mova [tmpq+21*%%str], m12 + mova [tmpq+29*%%str], m13 + mova m12, [tmpq+10*%%str] + mova m13, [tmpq+14*%%str] + + TRANSPOSE16x16W 6, 0, 15, 14, 1, 7, 5, 4, \ + 2, 3, 8, 9, 12, 13, 10, 11, \ + [tmpq+12*%%str], [tmpq+ 8*%%str], 1 + mova [tmpq+ 0*%%str], m6 + mova [tmpq+ 2*%%str], m0 + mova [tmpq+ 4*%%str], m15 + mova [tmpq+ 6*%%str], m14 + mova [tmpq+10*%%str], m7 + mova [tmpq+12*%%str], m5 + mova [tmpq+14*%%str], m4 + mova [tmpq+16*%%str], m2 + mova [tmpq+18*%%str], m3 + mova [tmpq+20*%%str], m8 + mova [tmpq+22*%%str], m9 + mova [tmpq+24*%%str], m12 + mova [tmpq+26*%%str], m13 + mova [tmpq+28*%%str], m10 + mova [tmpq+30*%%str], m11 + + mova m0, [tmpq+21*%%str] + mova m1, [tmpq+29*%%str] + mova m2, [tmpq+13*%%str] + mova m3, [tmpq+ 9*%%str] + mova m4, [tmpq+ 1*%%str] + mova m5, [tmpq+ 5*%%str] + mova m7, [tmpq+25*%%str] + mova m8, [tmpq+31*%%str] + mova m9, [tmpq+27*%%str] + mova m10, [tmpq+23*%%str] + mova m11, [tmpq+19*%%str] + mova m12, [tmpq+15*%%str] + mova m13, [tmpq+11*%%str] + mova m14, [tmpq+ 7*%%str] + mova m15, [tmpq+ 3*%%str] + TRANSPOSE16x16W 0, 1, 2, 3, 4, 5, 6, 7, \ + 8, 9, 10, 11, 12, 13, 14, 15, \ + [tmpq+17*%%str], [tmpq+ 9*%%str], 1 + mova [tmpq+ 1*%%str], m0 + mova [tmpq+ 3*%%str], m1 + mova [tmpq+ 5*%%str], m2 + mova [tmpq+ 7*%%str], m3 + mova [tmpq+11*%%str], m5 + mova [tmpq+13*%%str], m6 + mova [tmpq+15*%%str], m7 + mova [tmpq+17*%%str], m8 + mova [tmpq+19*%%str], m9 + mova [tmpq+21*%%str], m10 + mova [tmpq+23*%%str], m11 + mova [tmpq+25*%%str], m12 + mova [tmpq+27*%%str], m13 + mova [tmpq+29*%%str], m14 + mova [tmpq+31*%%str], m15 +%else ; !avx2 TRANSPOSE8x8W 6, 0, 15, 14, 1, 7, 5, 4, 8 mova [tmpq+ 0*%%str], m6 mova [tmpq+ 4*%%str], m0 @@ -2175,6 +2479,7 @@ IADST16_FN iadst, IADST16, iadst, IADST16, avx mova [tmpq+22*%%str], m13 mova [tmpq+26*%%str], m14 mova [tmpq+30*%%str], m15 +%endif ; avx2 %else mova m2, [tmpq+24*%%str] ; t6 mova m3, [tmpq+28*%%str] ; t7 @@ -2623,3 +2928,106 @@ cglobal vp9_idct_idct_32x32_add, 0, 6 + ARCH_X86_64 * 3, 16, 2048, dst, stride, VP9_IDCT_IDCT_32x32_ADD_XMM sse2 VP9_IDCT_IDCT_32x32_ADD_XMM ssse3 VP9_IDCT_IDCT_32x32_ADD_XMM avx + +; this is almost identical to VP9_STORE_2X, but it does two rows +; for slightly improved interleaving, and it omits vpermq since the +; input is DC so all values are identical +%macro VP9_STORE_YMM_DC_2X2 6 ; reg, tmp1, tmp2, tmp3, tmp4, zero + mova m%2, [dstq] + mova m%4, [dstq+strideq] + punpckhbw m%3, m%2, m%6 + punpcklbw m%2, m%6 + punpckhbw m%5, m%4, m%6 + punpcklbw m%4, m%6 + paddw m%3, m%1 + paddw m%2, m%1 + paddw m%5, m%1 + paddw m%4, m%1 + packuswb m%2, m%3 + packuswb m%4, m%5 + mova [dstq+strideq*0], m%2 + mova [dstq+strideq*1], m%4 +%endmacro + +%if ARCH_X86_64 && HAVE_AVX2_EXTERNAL +INIT_YMM avx2 +cglobal vp9_idct_idct_32x32_add, 4, 9, 16, 2048, dst, stride, block, eob + cmp eobd, 135 + jg .idctfull + cmp eobd, 1 + jg .idct16x16 + + ; dc-only case + mova m1, [pw_11585x2] + vpbroadcastw m0, [blockq] + pmulhrsw m0, m1 + pmulhrsw m0, m1 + pxor m5, m5 + pmulhrsw m0, [pw_512] + movd [blockq], xm5 + + DEFINE_ARGS dst, stride, cnt + mov cntd, 16 +.loop_dc: + VP9_STORE_YMM_DC_2X2 0, 1, 2, 3, 4, 5 + lea dstq, [dstq+2*strideq] + dec cntd + jg .loop_dc + RET + + DEFINE_ARGS dst_bak, stride, block, cnt, dst, stride30, dst_end, stride2, tmp +.idct16x16: + mov tmpq, rsp + VP9_IDCT32_1D blockq, 1, 16 + + mov stride30q, strideq ; stride + lea stride2q, [strideq*2] ; stride*2 + shl stride30q, 5 ; stride*32 + mov cntd, 2 + sub stride30q, stride2q ; stride*30 +.loop2_16x16: + mov dstq, dst_bakq + lea dst_endq, [dstq+stride30q] + VP9_IDCT32_1D tmpq, 2, 16 + add dst_bakq, 16 + add tmpq, 32 + dec cntd + jg .loop2_16x16 + + ; at the end of the loop, m1 should still be zero + ; use that to zero out block coefficients + ZERO_BLOCK blockq, 64, 16, m1 + RET + +.idctfull: + mov cntd, 2 + mov tmpq, rsp +.loop1_full: + VP9_IDCT32_1D blockq, 1 + add blockq, 32 + add tmpq, 1024 + dec cntd + jg .loop1_full + + sub blockq, 64 + + mov stride30q, strideq ; stride + lea stride2q, [strideq*2] ; stride*2 + shl stride30q, 5 ; stride*32 + mov cntd, 2 + mov tmpq, rsp + sub stride30q, stride2q ; stride*30 +.loop2_full: + mov dstq, dst_bakq + lea dst_endq, [dstq+stride30q] + VP9_IDCT32_1D tmpq, 2 + add dst_bakq, 16 + add tmpq, 32 + dec cntd + jg .loop2_full + + ; at the end of the loop, m1 should still be zero + ; use that to zero out block coefficients + ZERO_BLOCK blockq, 64, 32, m1 + RET +%endif diff --git a/media/ffvpx/libavcodec/x86/vp9lpf.asm b/media/ffvpx/libavcodec/x86/vp9lpf.asm index 2c4fe214d..4e7ede223 100644 --- a/media/ffvpx/libavcodec/x86/vp9lpf.asm +++ b/media/ffvpx/libavcodec/x86/vp9lpf.asm @@ -52,7 +52,7 @@ mask_mix48: times 8 db 0x00 SECTION .text %macro SCRATCH 3 -%if ARCH_X86_64 +%ifdef m8 SWAP %1, %2 %else mova [%3], m%1 @@ -60,7 +60,7 @@ SECTION .text %endmacro %macro UNSCRATCH 3 -%if ARCH_X86_64 +%ifdef m8 SWAP %1, %2 %else mova m%1, [%3] @@ -69,7 +69,7 @@ SECTION .text ; %1 = abs(%2-%3) %macro ABSSUB 4 ; dst, src1 (RO), src2 (RO), tmp -%if ARCH_X86_64 +%ifdef m8 psubusb %1, %3, %2 psubusb %4, %2, %3 %else @@ -102,7 +102,7 @@ SECTION .text %endmacro %macro UNPACK 4 -%if ARCH_X86_64 +%ifdef m8 punpck%1bw %2, %3, %4 %else mova %2, %3 @@ -112,27 +112,27 @@ SECTION .text %macro FILTER_SUBx2_ADDx2 11 ; %1=dst %2=h/l %3=cache %4=stack_off %5=sub1 %6=sub2 %7=add1 ; %8=add2 %9=rshift, [unpack], [unpack_is_mem_on_x86_32] - psubw %3, [rsp+%4+%5*32] - psubw %3, [rsp+%4+%6*32] - paddw %3, [rsp+%4+%7*32] + psubw %3, [rsp+%4+%5*mmsize*2] + psubw %3, [rsp+%4+%6*mmsize*2] + paddw %3, [rsp+%4+%7*mmsize*2] %ifnidn %10, "" %if %11 == 0 punpck%2bw %1, %10, m0 %else UNPACK %2, %1, %10, m0 %endif - mova [rsp+%4+%8*32], %1 + mova [rsp+%4+%8*mmsize*2], %1 paddw %3, %1 %else - paddw %3, [rsp+%4+%8*32] + paddw %3, [rsp+%4+%8*mmsize*2] %endif psraw %1, %3, %9 %endmacro ; FIXME interleave l/h better (for instruction pairing) %macro FILTER_INIT 9 ; tmp1, tmp2, cacheL, cacheH, dstp, stack_off, filterid, mask, source - FILTER%7_INIT %1, l, %3, %6 + 0 - FILTER%7_INIT %2, h, %4, %6 + 16 + FILTER%7_INIT %1, l, %3, %6 + 0 + FILTER%7_INIT %2, h, %4, %6 + mmsize packuswb %1, %2 MASK_APPLY %1, %9, %8, %2 mova %5, %1 @@ -147,8 +147,8 @@ SECTION .text mova %14, %15 %endif %endif - FILTER_SUBx2_ADDx2 %1, l, %3, %6 + 0, %7, %8, %9, %10, %11, %14, %16 - FILTER_SUBx2_ADDx2 %2, h, %4, %6 + 16, %7, %8, %9, %10, %11, %14, %16 + FILTER_SUBx2_ADDx2 %1, l, %3, %6 + 0, %7, %8, %9, %10, %11, %14, %16 + FILTER_SUBx2_ADDx2 %2, h, %4, %6 + mmsize, %7, %8, %9, %10, %11, %14, %16 packuswb %1, %2 %ifnidn %13, "" MASK_APPLY %1, %13, %12, %2 @@ -195,21 +195,21 @@ SECTION .text %macro FILTER6_INIT 4 ; %1=dst %2=h/l %3=cache, %4=stack_off UNPACK %2, %1, rp3, m0 ; p3: B->W - mova [rsp+%4+0*32], %1 + mova [rsp+%4+0*mmsize*2], %1 paddw %3, %1, %1 ; p3*2 paddw %3, %1 ; p3*3 punpck%2bw %1, m1, m0 ; p2: B->W - mova [rsp+%4+1*32], %1 + mova [rsp+%4+1*mmsize*2], %1 paddw %3, %1 ; p3*3 + p2 paddw %3, %1 ; p3*3 + p2*2 UNPACK %2, %1, rp1, m0 ; p1: B->W - mova [rsp+%4+2*32], %1 + mova [rsp+%4+2*mmsize*2], %1 paddw %3, %1 ; p3*3 + p2*2 + p1 UNPACK %2, %1, rp0, m0 ; p0: B->W - mova [rsp+%4+3*32], %1 + mova [rsp+%4+3*mmsize*2], %1 paddw %3, %1 ; p3*3 + p2*2 + p1 + p0 UNPACK %2, %1, rq0, m0 ; q0: B->W - mova [rsp+%4+4*32], %1 + mova [rsp+%4+4*mmsize*2], %1 paddw %3, %1 ; p3*3 + p2*2 + p1 + p0 + q0 paddw %3, [pw_4] ; p3*3 + p2*2 + p1 + p0 + q0 + 4 psraw %1, %3, 3 ; (p3*3 + p2*2 + p1 + p0 + q0 + 4) >> 3 @@ -217,24 +217,24 @@ SECTION .text %macro FILTER14_INIT 4 ; %1=dst %2=h/l %3=cache, %4=stack_off punpck%2bw %1, m2, m0 ; p7: B->W - mova [rsp+%4+ 8*32], %1 + mova [rsp+%4+ 8*mmsize*2], %1 psllw %3, %1, 3 ; p7*8 psubw %3, %1 ; p7*7 punpck%2bw %1, m3, m0 ; p6: B->W - mova [rsp+%4+ 9*32], %1 + mova [rsp+%4+ 9*mmsize*2], %1 paddw %3, %1 ; p7*7 + p6 paddw %3, %1 ; p7*7 + p6*2 UNPACK %2, %1, rp5, m0 ; p5: B->W - mova [rsp+%4+10*32], %1 + mova [rsp+%4+10*mmsize*2], %1 paddw %3, %1 ; p7*7 + p6*2 + p5 UNPACK %2, %1, rp4, m0 ; p4: B->W - mova [rsp+%4+11*32], %1 + mova [rsp+%4+11*mmsize*2], %1 paddw %3, %1 ; p7*7 + p6*2 + p5 + p4 - paddw %3, [rsp+%4+ 0*32] ; p7*7 + p6*2 + p5 + p4 + p3 - paddw %3, [rsp+%4+ 1*32] ; p7*7 + p6*2 + p5 + .. + p2 - paddw %3, [rsp+%4+ 2*32] ; p7*7 + p6*2 + p5 + .. + p1 - paddw %3, [rsp+%4+ 3*32] ; p7*7 + p6*2 + p5 + .. + p0 - paddw %3, [rsp+%4+ 4*32] ; p7*7 + p6*2 + p5 + .. + p0 + q0 + paddw %3, [rsp+%4+ 0*mmsize*2] ; p7*7 + p6*2 + p5 + p4 + p3 + paddw %3, [rsp+%4+ 1*mmsize*2] ; p7*7 + p6*2 + p5 + .. + p2 + paddw %3, [rsp+%4+ 2*mmsize*2] ; p7*7 + p6*2 + p5 + .. + p1 + paddw %3, [rsp+%4+ 3*mmsize*2] ; p7*7 + p6*2 + p5 + .. + p0 + paddw %3, [rsp+%4+ 4*mmsize*2] ; p7*7 + p6*2 + p5 + .. + p0 + q0 paddw %3, [pw_8] ; p7*7 + p6*2 + p5 + .. + p0 + q0 + 8 psraw %1, %3, 4 ; (p7*7 + p6*2 + p5 + .. + p0 + q0 + 8) >> 4 %endmacro @@ -334,22 +334,24 @@ SECTION .text %endmacro %macro DEFINE_TRANSPOSED_P7_TO_Q7 0-1 0 -%define P3 rsp + 0 + %1 -%define P2 rsp + 16 + %1 -%define P1 rsp + 32 + %1 -%define P0 rsp + 48 + %1 -%define Q0 rsp + 64 + %1 -%define Q1 rsp + 80 + %1 -%define Q2 rsp + 96 + %1 -%define Q3 rsp + 112 + %1 -%define P7 rsp + 128 + %1 -%define P6 rsp + 144 + %1 -%define P5 rsp + 160 + %1 -%define P4 rsp + 176 + %1 -%define Q4 rsp + 192 + %1 -%define Q5 rsp + 208 + %1 -%define Q6 rsp + 224 + %1 -%define Q7 rsp + 240 + %1 +%define P3 rsp + 0*mmsize + %1 +%define P2 rsp + 1*mmsize + %1 +%define P1 rsp + 2*mmsize + %1 +%define P0 rsp + 3*mmsize + %1 +%define Q0 rsp + 4*mmsize + %1 +%define Q1 rsp + 5*mmsize + %1 +%define Q2 rsp + 6*mmsize + %1 +%define Q3 rsp + 7*mmsize + %1 +%if mmsize == 16 +%define P7 rsp + 8*mmsize + %1 +%define P6 rsp + 9*mmsize + %1 +%define P5 rsp + 10*mmsize + %1 +%define P4 rsp + 11*mmsize + %1 +%define Q4 rsp + 12*mmsize + %1 +%define Q5 rsp + 13*mmsize + %1 +%define Q6 rsp + 14*mmsize + %1 +%define Q7 rsp + 15*mmsize + %1 +%endif %endmacro ; ..............AB -> AAAAAAAABBBBBBBB @@ -363,14 +365,19 @@ SECTION .text %endif %endmacro -%macro LOOPFILTER 5 ; %1=v/h %2=size1 %3+%4=stack, %5=32bit stack only +%macro LOOPFILTER 5 ; %1=v/h %2=size1 %3+%4=stack, %5=mmx/32bit stack only +%assign %%ext 0 +%if ARCH_X86_32 || mmsize == 8 +%assign %%ext %5 +%endif + %if UNIX64 -cglobal vp9_loop_filter_%1_%2_16, 5, 9, 16, %3 + %4, dst, stride, E, I, H, mstride, dst2, stride3, mstride3 +cglobal vp9_loop_filter_%1_%2_ %+ mmsize, 5, 9, 16, %3 + %4 + %%ext, dst, stride, E, I, H, mstride, dst2, stride3, mstride3 %else %if WIN64 -cglobal vp9_loop_filter_%1_%2_16, 4, 8, 16, %3 + %4, dst, stride, E, I, mstride, dst2, stride3, mstride3 +cglobal vp9_loop_filter_%1_%2_ %+ mmsize, 4, 8, 16, %3 + %4 + %%ext, dst, stride, E, I, mstride, dst2, stride3, mstride3 %else -cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, dst2, stride3, mstride3 +cglobal vp9_loop_filter_%1_%2_ %+ mmsize, 2, 6, 16, %3 + %4 + %%ext, dst, stride, mstride, dst2, stride3, mstride3 %define Ed dword r2m %define Id dword r3m %endif @@ -384,18 +391,22 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, lea mstride3q, [mstrideq*3] %ifidn %1, h -%if %2 > 16 +%if %2 != 16 +%if mmsize == 16 %define movx movh +%else +%define movx mova +%endif lea dstq, [dstq + 4*strideq - 4] %else %define movx movu lea dstq, [dstq + 4*strideq - 8] ; go from top center (h pos) to center left (v pos) %endif - lea dst2q, [dstq + 8*strideq] %else lea dstq, [dstq + 4*mstrideq] - lea dst2q, [dstq + 8*strideq] %endif + ; FIXME we shouldn't need two dts registers if mmsize == 8 + lea dst2q, [dstq + 8*strideq] DEFINE_REAL_P7_TO_Q7 @@ -406,11 +417,11 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, movx m3, [P4] movx m4, [P3] movx m5, [P2] -%if ARCH_X86_64 || %2 != 16 +%if (ARCH_X86_64 && mmsize == 16) || %2 > 16 movx m6, [P1] %endif movx m7, [P0] -%if ARCH_X86_64 +%ifdef m8 movx m8, [Q0] movx m9, [Q1] movx m10, [Q2] @@ -502,7 +513,7 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, movhps [Q5], m6 movhps [Q7], m7 DEFINE_TRANSPOSED_P7_TO_Q7 -%else ; %2 == 44/48/84/88 +%elif %2 > 16 ; %2 == 44/48/84/88 punpcklbw m0, m1 punpcklbw m2, m3 punpcklbw m4, m5 @@ -529,12 +540,31 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, mova [Q1], m5 mova [Q2], m7 mova [Q3], m3 +%else ; %2 == 4 || %2 == 8 + SBUTTERFLY bw, 0, 1, 6 + SBUTTERFLY bw, 2, 3, 6 + SBUTTERFLY bw, 4, 5, 6 + mova [rsp+4*mmsize], m5 + mova m6, [P1] + SBUTTERFLY bw, 6, 7, 5 + DEFINE_TRANSPOSED_P7_TO_Q7 + TRANSPOSE4x4W 0, 2, 4, 6, 5 + mova [P3], m0 + mova [P2], m2 + mova [P1], m4 + mova [P0], m6 + mova m5, [rsp+4*mmsize] + TRANSPOSE4x4W 1, 3, 5, 7, 0 + mova [Q0], m1 + mova [Q1], m3 + mova [Q2], m5 + mova [Q3], m7 %endif ; %2 %endif ; x86-32/64 %endif ; %1 == h ; calc fm mask -%if %2 == 16 +%if %2 == 16 || mmsize == 8 %if cpuflag(ssse3) pxor m0, m0 %endif @@ -552,7 +582,7 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, mova m0, [pb_80] pxor m2, m0 pxor m3, m0 -%if ARCH_X86_64 +%ifdef m8 %ifidn %1, v mova m8, [P3] mova m9, [P2] @@ -613,10 +643,10 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, ; (m3: fm, m8..15: p3 p2 p1 p0 q0 q1 q2 q3) ; calc flat8in (if not 44_16) and hev masks -%if %2 != 44 +%if %2 != 44 && %2 != 4 mova m6, [pb_81] ; [1 1 1 1 ...] ^ 0x80 ABSSUB_GT m2, rp3, rp0, m6, m5 ; abs(p3 - p0) <= 1 -%if ARCH_X86_64 +%ifdef m8 mova m8, [pb_80] %define rb80 m8 %else @@ -625,7 +655,7 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, ABSSUB_GT m1, rp2, rp0, m6, m5, rb80 ; abs(p2 - p0) <= 1 por m2, m1 ABSSUB m4, rp1, rp0, m5 ; abs(p1 - p0) -%if %2 == 16 +%if %2 <= 16 %if cpuflag(ssse3) pxor m0, m0 %endif @@ -655,8 +685,15 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, %endif %else mova m6, [pb_80] +%if %2 == 44 movd m7, Hd SPLATB_MIX m7 +%else +%if cpuflag(ssse3) + pxor m0, m0 +%endif + SPLATB_REG m7, H, m0 ; H H H H ... +%endif pxor m7, m6 ABSSUB m4, rp1, rp0, m1 ; abs(p1 - p0) pxor m4, m6 @@ -670,7 +707,7 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, %if %2 == 16 ; (m0: hev, m2: flat8in, m3: fm, m6: pb_81, m9..15: p2 p1 p0 q0 q1 q2 q3) ; calc flat8out mask -%if ARCH_X86_64 +%ifdef m8 mova m8, [P7] mova m9, [P6] %define rp7 m8 @@ -682,7 +719,7 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, ABSSUB_GT m1, rp7, rp0, m6, m5 ; abs(p7 - p0) <= 1 ABSSUB_GT m7, rp6, rp0, m6, m5 ; abs(p6 - p0) <= 1 por m1, m7 -%if ARCH_X86_64 +%ifdef m8 mova m8, [P5] mova m9, [P4] %define rp5 m8 @@ -695,7 +732,7 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, por m1, m7 ABSSUB_GT m7, rp4, rp0, m6, m5 ; abs(p4 - p0) <= 1 por m1, m7 -%if ARCH_X86_64 +%ifdef m8 mova m14, [Q4] mova m15, [Q5] %define rq4 m14 @@ -708,7 +745,7 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, por m1, m7 ABSSUB_GT m7, rq5, rq0, m6, m5 ; abs(q5 - q0) <= 1 por m1, m7 -%if ARCH_X86_64 +%ifdef m8 mova m14, [Q6] mova m15, [Q7] %define rq6 m14 @@ -738,7 +775,7 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, ; (m0: hev, [m1: flat8out], [m2: flat8in], m3: fm, m8..15: p5 p4 p1 p0 q0 q1 q6 q7) ; filter2() -%if %2 != 44 +%if %2 != 44 && %2 != 4 mova m6, [pb_80] ; already in m6 if 44_16 SCRATCH 2, 15, rsp+%3+%4 %if %2 == 16 @@ -756,7 +793,7 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, paddsb m4, m2 ; 3*(q0 - p0) + (p1 - q1) paddsb m6, m4, [pb_4] ; m6: f1 = clip(f + 4, 127) paddsb m4, [pb_3] ; m4: f2 = clip(f + 3, 127) -%if ARCH_X86_64 +%ifdef m8 mova m14, [pb_10] ; will be reused in filter4() %define rb10 m14 %else @@ -765,8 +802,8 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, SRSHIFT3B_2X m6, m4, rb10, m7 ; f1 and f2 sign byte shift by 3 SIGN_SUB m7, rq0, m6, m5 ; m7 = q0 - f1 SIGN_ADD m1, rp0, m4, m5 ; m1 = p0 + f2 -%if %2 != 44 -%if ARCH_X86_64 +%if %2 != 44 && %2 != 4 +%ifdef m8 pandn m6, m15, m3 ; ~mask(in) & mask(fm) %else mova m6, [rsp+%3+%4] @@ -787,8 +824,8 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, paddsb m6, m2, [pb_4] ; m6: f1 = clip(f + 4, 127) paddsb m2, [pb_3] ; m2: f2 = clip(f + 3, 127) SRSHIFT3B_2X m6, m2, rb10, m4 ; f1 and f2 sign byte shift by 3 -%if %2 != 44 -%if ARCH_X86_64 +%if %2 != 44 && %2 != 4 +%ifdef m8 pandn m5, m15, m3 ; ~mask(in) & mask(fm) %else mova m5, [rsp+%3+%4] @@ -815,26 +852,26 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, mova [P1], m1 mova [Q1], m4 -%if %2 != 44 +%if %2 != 44 && %2 != 4 UNSCRATCH 2, 15, rsp+%3+%4 %endif ; ([m1: flat8out], m2: flat8in, m3: fm, m10..13: p1 p0 q0 q1) ; filter6() -%if %2 != 44 +%if %2 != 44 && %2 != 4 pxor m0, m0 -%if %2 > 16 +%if %2 != 16 pand m3, m2 %else pand m2, m3 ; mask(fm) & mask(in) -%if ARCH_X86_64 +%ifdef m8 pandn m3, m8, m2 ; ~mask(out) & (mask(fm) & mask(in)) %else mova m3, [rsp+%3+%4+16] pandn m3, m2 %endif %endif -%if ARCH_X86_64 +%ifdef m8 mova m14, [P3] mova m9, [Q3] %define rp3 m14 @@ -882,7 +919,7 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, pand m1, m2 ; mask(out) & (mask(fm) & mask(in)) mova m2, [P7] mova m3, [P6] -%if ARCH_X86_64 +%ifdef m8 mova m8, [P5] mova m9, [P4] %define rp5 m8 @@ -1008,7 +1045,7 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, movhps [Q5], m6 movhps [Q7], m7 %endif -%elif %2 == 44 +%elif %2 == 44 || %2 == 4 SWAP 0, 1 ; m0 = p1 SWAP 1, 7 ; m1 = p0 SWAP 2, 5 ; m2 = q0 @@ -1018,6 +1055,7 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, SBUTTERFLY bw, 2, 3, 4 SBUTTERFLY wd, 0, 2, 4 SBUTTERFLY wd, 1, 3, 4 +%if mmsize == 16 movd [P7], m0 movd [P3], m2 movd [Q0], m1 @@ -1046,6 +1084,20 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, movd [P0], m2 movd [Q3], m1 movd [Q7], m3 +%else + movd [P7], m0 + movd [P5], m2 + movd [P3], m1 + movd [P1], m3 + psrlq m0, 32 + psrlq m2, 32 + psrlq m1, 32 + psrlq m3, 32 + movd [P6], m0 + movd [P4], m2 + movd [P2], m1 + movd [P0], m3 +%endif %else ; the following code do a transpose of 8 full lines to 16 half ; lines (high part). It is inlined to avoid the need of a staging area @@ -1055,12 +1107,12 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, mova m3, [P0] mova m4, [Q0] mova m5, [Q1] -%if ARCH_X86_64 +%ifdef m8 mova m6, [Q2] %endif mova m7, [Q3] DEFINE_REAL_P7_TO_Q7 -%if ARCH_X86_64 +%ifdef m8 SBUTTERFLY bw, 0, 1, 8 SBUTTERFLY bw, 2, 3, 8 SBUTTERFLY bw, 4, 5, 8 @@ -1075,27 +1127,32 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, SBUTTERFLY dq, 3, 7, 8 %else SBUTTERFLY bw, 0, 1, 6 - mova [rsp+64], m1 - mova m6, [rsp+96] + mova [rsp+mmsize*4], m1 + mova m6, [rsp+mmsize*6] SBUTTERFLY bw, 2, 3, 1 SBUTTERFLY bw, 4, 5, 1 SBUTTERFLY bw, 6, 7, 1 SBUTTERFLY wd, 0, 2, 1 - mova [rsp+96], m2 - mova m1, [rsp+64] + mova [rsp+mmsize*6], m2 + mova m1, [rsp+mmsize*4] SBUTTERFLY wd, 1, 3, 2 SBUTTERFLY wd, 4, 6, 2 SBUTTERFLY wd, 5, 7, 2 SBUTTERFLY dq, 0, 4, 2 SBUTTERFLY dq, 1, 5, 2 +%if mmsize == 16 movh [Q0], m1 movhps [Q1], m1 - mova m2, [rsp+96] +%else + mova [P3], m1 +%endif + mova m2, [rsp+mmsize*6] SBUTTERFLY dq, 2, 6, 1 SBUTTERFLY dq, 3, 7, 1 %endif SWAP 3, 6 SWAP 1, 4 +%if mmsize == 16 movh [P7], m0 movhps [P6], m0 movh [P5], m1 @@ -1104,7 +1161,7 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, movhps [P2], m2 movh [P1], m3 movhps [P0], m3 -%if ARCH_X86_64 +%ifdef m8 movh [Q0], m4 movhps [Q1], m4 %endif @@ -1114,6 +1171,15 @@ cglobal vp9_loop_filter_%1_%2_16, 2, 6, 16, %3 + %4 + %5, dst, stride, mstride, movhps [Q5], m6 movh [Q6], m7 movhps [Q7], m7 +%else + mova [P7], m0 + mova [P6], m1 + mova [P5], m2 + mova [P4], m3 + mova [P2], m5 + mova [P1], m6 + mova [P0], m7 +%endif %endif %endif @@ -1137,3 +1203,9 @@ LPF_16_VH_ALL_OPTS 44, 0, 128, 0 LPF_16_VH_ALL_OPTS 48, 256, 128, 16 LPF_16_VH_ALL_OPTS 84, 256, 128, 16 LPF_16_VH_ALL_OPTS 88, 256, 128, 16 + +INIT_MMX mmxext +LOOPFILTER v, 4, 0, 0, 0 +LOOPFILTER h, 4, 0, 64, 0 +LOOPFILTER v, 8, 128, 0, 8 +LOOPFILTER h, 8, 128, 64, 8 diff --git a/media/ffvpx/libavcodec/x86/vp9lpf_16bpp.asm b/media/ffvpx/libavcodec/x86/vp9lpf_16bpp.asm index c15437b8b..c0888170c 100644 --- a/media/ffvpx/libavcodec/x86/vp9lpf_16bpp.asm +++ b/media/ffvpx/libavcodec/x86/vp9lpf_16bpp.asm @@ -78,7 +78,7 @@ SECTION .text %endif %endmacro -; calulate p or q portion of flat8out +; calculate p or q portion of flat8out %macro FLAT8OUT_HALF 0 psubw m4, m0 ; q4-q0 psubw m5, m0 ; q5-q0 -- cgit v1.2.3