From patchwork Sat Sep 30 15:01:15 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Andreas Rheinhardt X-Patchwork-Id: 44049 Delivered-To: ffmpegpatchwork2@gmail.com Received: by 2002:a05:6a20:1204:b0:15d:8365:d4b8 with SMTP id v4csp432698pzf; Sat, 30 Sep 2023 08:00:26 -0700 (PDT) X-Google-Smtp-Source: AGHT+IF1MOh3IgKngSCs1h9CSOlTvfqYbOisBIlUSt+HNK/Pfzz+e8jJkbD+yzOyOHbsgWst/2EX X-Received: by 2002:a17:907:75c4:b0:9ae:6196:5094 with SMTP id jl4-20020a17090775c400b009ae61965094mr6775654ejc.30.1696086025945; Sat, 30 Sep 2023 08:00:25 -0700 (PDT) Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org. [79.124.17.100]) by mx.google.com with ESMTP id bq18-20020a170906d0d200b00992cd4e0b83si17995724ejb.426.2023.09.30.08.00.21; Sat, 30 Sep 2023 08:00:25 -0700 (PDT) Received-SPF: pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) client-ip=79.124.17.100; Authentication-Results: mx.google.com; dkim=neutral (body hash did not verify) header.i=@outlook.com header.s=selector1 header.b=p0rhR4Vm; arc=fail (body hash mismatch); spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=QUARANTINE dis=NONE) header.from=outlook.com Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id E62B968CC03; Sat, 30 Sep 2023 18:00:16 +0300 (EEST) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from EUR04-HE1-obe.outbound.protection.outlook.com (mail-he1eur04olkn2021.outbound.protection.outlook.com [40.92.73.21]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id 9DB8268CB32 for ; Sat, 30 Sep 2023 18:00:10 +0300 (EEST) ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=NTdQzHMTFFMVgduDyAV/Kh4G15ysbse7PuyoQ3Uw6dTMwmJAL7sH/3o3w0klr5TmyK8A6hACrTvrRnTXC3Vsxe6Kn7HuPjzw8r9MLYzdAEIw1MX6cVwtZHdprlWcuMHw+FWybHLDRJNzmX+9Je7NVYT0Szjk9ctYF8vQarjPy03ubb4sOr9sB8QFrQiQ5xFDePR9rK0ZvLNMT9gOtCpHtG/SoEzbvGkkHoaLxsuG/Kfi8BgIVDclIzCDWVNTldQ6v/r0iIJttFE9iOQZjdKff2AfjIqERYaAtXmqwH10AkwYxSuuWV2KxxEPax9yDXNpG9lKNTp2Y5gWN8HYjiRd8Q== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=gX2D21mUYtjYH24VWyDgEnDu7HQ56WbrETrI/xzyQAo=; b=V74YlDOJNBJTGFF2DIij4V13/eMqDzOGXJ6otYeh6t1/hs0l6bbjJp2gjuCi3kifPij9KuPbhiHSLKwpSRWg7983bOmunuFX/jvah89ATaBTJS/sLNH9w9LLeUvGTl60U+T162lerNxJG1p+3LSu5pnT3dcOnPcngsj3WHM6F4sICBkDHPlLKyrseKyLr6tfK8H0YhN5gNkDiohy0yDKNZmF0poUNLJk1+7exTPqZoAQ169qW0gmzCvsvhYyH2FuxlSsDdjex9XD3XCMuUdswYWZaxtC0/YmucvZOAte/qEIhMmO4Au7EGw0hdzWkwWDbRjEecUx99It6adFdwCp+A== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=none; dmarc=none; dkim=none; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=outlook.com; s=selector1; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=gX2D21mUYtjYH24VWyDgEnDu7HQ56WbrETrI/xzyQAo=; b=p0rhR4Vml/yBbGO7TVe0FQrzKDUN+J4gymp2r45zfi/riVJFBE/ShADP/hwFqNY9inBKjyOu4eLuart3cpSFDNzA6iIsUFxY6aTAueVf7w36vf05a0P2489cZD0eiRYzmj8+l05Leql1QTOlVtk5vT3OjB0nVWezu7TPUs11D6d7TDyG7i9dF47GzpWlp2zaopHOzq07rLElNfXbUyIStw5uZcTxX1cBlB7PLkQNOsJchwwDepCtU8PcmdyAsRkPa2Dfz7E4w3vCjoEc/1+9tuttYwlgi/n3zPmejs2rKJG3yqjXfN3L6GViPRAN75chUUoqJHcat7AzottY1bZYAQ== Received: from AS8P250MB0744.EURP250.PROD.OUTLOOK.COM (2603:10a6:20b:541::14) by AS8P250MB0234.EURP250.PROD.OUTLOOK.COM (2603:10a6:20b:379::6) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.6838.28; Sat, 30 Sep 2023 15:00:07 +0000 Received: from AS8P250MB0744.EURP250.PROD.OUTLOOK.COM ([fe80::5e01:aea5:d3a8:cafa]) by AS8P250MB0744.EURP250.PROD.OUTLOOK.COM ([fe80::5e01:aea5:d3a8:cafa%3]) with mapi id 15.20.6792.026; Sat, 30 Sep 2023 15:00:07 +0000 From: Andreas Rheinhardt To: ffmpeg-devel@ffmpeg.org Date: Sat, 30 Sep 2023 17:01:15 +0200 Message-ID: X-Mailer: git-send-email 2.34.1 X-TMN: [e9v14bxOqhDAi5jtgiizDOeN2OIvue5L] X-ClientProxiedBy: ZR0P278CA0113.CHEP278.PROD.OUTLOOK.COM (2603:10a6:910:20::10) To AS8P250MB0744.EURP250.PROD.OUTLOOK.COM (2603:10a6:20b:541::14) X-Microsoft-Original-Message-ID: <20230930150116.2602009-1-andreas.rheinhardt@outlook.com> MIME-Version: 1.0 X-MS-Exchange-MessageSentRepresentingType: 1 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: AS8P250MB0744:EE_|AS8P250MB0234:EE_ X-MS-Office365-Filtering-Correlation-Id: cc067bef-ffe8-4a15-9845-08dbc1c5ef30 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: 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 X-OriginatorOrg: outlook.com X-MS-Exchange-CrossTenant-Network-Message-Id: cc067bef-ffe8-4a15-9845-08dbc1c5ef30 X-MS-Exchange-CrossTenant-AuthSource: AS8P250MB0744.EURP250.PROD.OUTLOOK.COM X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 30 Sep 2023 15:00:07.2022 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 84df9e7f-e9f6-40af-b435-aaaaaaaaaaaa X-MS-Exchange-CrossTenant-RMS-PersistedConsumerOrg: 00000000-0000-0000-0000-000000000000 X-MS-Exchange-Transport-CrossTenantHeadersStamped: AS8P250MB0234 Subject: [FFmpeg-devel] [PATCH 1/2] avcodec/snow: Move encoder-only stuff out of SnowContext X-BeenThere: ffmpeg-devel@ffmpeg.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: FFmpeg development discussions and patches List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Reply-To: FFmpeg development discussions and patches Cc: Andreas Rheinhardt Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" X-TUID: 1lQLRmVMVMsp Put it into an encoder-specific context with a SnowContext at its front. This also avoids having to include mpegvideo.h in snow.c and snowdec.c. Signed-off-by: Andreas Rheinhardt --- libavcodec/snow.h | 27 +-- libavcodec/snowenc.c | 448 ++++++++++++++++++++++++------------------- 2 files changed, 254 insertions(+), 221 deletions(-) diff --git a/libavcodec/snow.h b/libavcodec/snow.h index 1c87577699..2e61154d0c 100644 --- a/libavcodec/snow.h +++ b/libavcodec/snow.h @@ -24,21 +24,18 @@ #include "libavutil/motion_vector.h" +#include "avcodec.h" #include "hpeldsp.h" -#include "me_cmp.h" -#include "qpeldsp.h" #include "snow_dwt.h" #include "rangecoder.h" #include "mathops.h" -#include "mpegvideo.h" #include "h264qpel.h" +#include "videodsp.h" #define SNOW_MAX_PLANES 4 -#define FF_ME_ITER 3 - #define MID_STATE 128 #define MAX_PLANES 4 @@ -117,12 +114,9 @@ typedef struct SnowContext{ AVClass *class; AVCodecContext *avctx; RangeCoder c; - MECmpContext mecc; HpelDSPContext hdsp; - QpelDSPContext qdsp; VideoDSPContext vdsp; H264QpelContext h264qpel; - MpegvideoEncDSPContext mpvencdsp; SnowDWTContext dwt; AVFrame *input_picture; ///< new_picture with the internal linesizes AVFrame *current_picture; @@ -155,9 +149,6 @@ typedef struct SnowContext{ int spatial_scalability; int qlog; int last_qlog; - int lambda; - int lambda2; - int pass1_rc; int mv_scale; int last_mv_scale; int qbias; @@ -170,18 +161,7 @@ typedef struct SnowContext{ int nb_planes; Plane plane[MAX_PLANES]; BlockNode *block; -#define ME_CACHE_SIZE 1024 - unsigned me_cache[ME_CACHE_SIZE]; - unsigned me_cache_generation; slice_buffer sb; - int memc_only; - int no_bitstream; - int intra_penalty; - int motion_est; - int iterative_dia_size; - int scenechange_threshold; - - MpegEncContext m; // needed for motion estimation, should not be used for anything else, the idea is to eventually make the motion estimation independent of MpegEncContext, so this will be removed then (FIXME/XXX) uint8_t *scratchbuf; uint8_t *emu_edge_buffer; @@ -189,9 +169,6 @@ typedef struct SnowContext{ AVMotionVector *avmv; unsigned avmv_size; int avmv_index; - uint64_t encoding_error[SNOW_MAX_PLANES]; - - int pred; }SnowContext; /* Tables */ diff --git a/libavcodec/snowenc.c b/libavcodec/snowenc.c index ab15c4e108..c215215346 100644 --- a/libavcodec/snowenc.c +++ b/libavcodec/snowenc.c @@ -30,6 +30,7 @@ #include "internal.h" //For AVCodecInternal.recon_frame #include "me_cmp.h" #include "packet_internal.h" +#include "qpeldsp.h" #include "snow_dwt.h" #include "snow.h" @@ -39,6 +40,34 @@ #include "mpegvideo.h" #include "h263enc.h" +#define FF_ME_ITER 3 + +typedef struct SnowEncContext { + SnowContext com; + QpelDSPContext qdsp; + MpegvideoEncDSPContext mpvencdsp; + + int lambda; + int lambda2; + int pass1_rc; + + int pred; + int memc_only; + int no_bitstream; + int intra_penalty; + int motion_est; + int iterative_dia_size; + int scenechange_threshold; + + MECmpContext mecc; + MpegEncContext m; // needed for motion estimation, should not be used for anything else, the idea is to eventually make the motion estimation independent of MpegEncContext, so this will be removed then (FIXME/XXX) +#define ME_CACHE_SIZE 1024 + unsigned me_cache[ME_CACHE_SIZE]; + unsigned me_cache_generation; + + uint64_t encoding_error[SNOW_MAX_PLANES]; +} SnowEncContext; + static void init_ref(MotionEstContext *c, const uint8_t *const src[3], uint8_t *const ref[3], uint8_t *const ref2[3], int x, int y, int ref_index) @@ -127,18 +156,20 @@ static int get_encode_buffer(SnowContext *s, AVFrame *frame) static av_cold int encode_init(AVCodecContext *avctx) { - SnowContext *s = avctx->priv_data; + SnowEncContext *const enc = avctx->priv_data; + SnowContext *const s = &enc->com; + MpegEncContext *const mpv = &enc->m; int plane_index, ret; int i; - if(s->pred == DWT_97 + if (enc->pred == DWT_97 && (avctx->flags & AV_CODEC_FLAG_QSCALE) && avctx->global_quality == 0){ av_log(avctx, AV_LOG_ERROR, "The 9/7 wavelet is incompatible with lossless mode.\n"); return AVERROR(EINVAL); } - s->spatial_decomposition_type= s->pred; //FIXME add decorrelator type r transform_type + s->spatial_decomposition_type = enc->pred; //FIXME add decorrelator type r transform_type s->mv_scale = (avctx->flags & AV_CODEC_FLAG_QPEL) ? 2 : 4; s->block_max_depth= (avctx->flags & AV_CODEC_FLAG_4MV ) ? 1 : 0; @@ -159,11 +190,11 @@ static av_cold int encode_init(AVCodecContext *avctx) } #define mcf(dx,dy)\ - s->qdsp.put_qpel_pixels_tab [0][dy+dx/4]=\ - s->qdsp.put_no_rnd_qpel_pixels_tab[0][dy+dx/4]=\ + enc->qdsp.put_qpel_pixels_tab [0][dy+dx/4]=\ + enc->qdsp.put_no_rnd_qpel_pixels_tab[0][dy+dx/4]=\ s->h264qpel.put_h264_qpel_pixels_tab[0][dy+dx/4];\ - s->qdsp.put_qpel_pixels_tab [1][dy+dx/4]=\ - s->qdsp.put_no_rnd_qpel_pixels_tab[1][dy+dx/4]=\ + enc->qdsp.put_qpel_pixels_tab [1][dy+dx/4]=\ + enc->qdsp.put_no_rnd_qpel_pixels_tab[1][dy+dx/4]=\ s->h264qpel.put_h264_qpel_pixels_tab[1][dy+dx/4]; mcf( 0, 0) @@ -183,28 +214,28 @@ static av_cold int encode_init(AVCodecContext *avctx) mcf( 8,12) mcf(12,12) - ff_me_cmp_init(&s->mecc, avctx); - ff_mpegvideoencdsp_init(&s->mpvencdsp, avctx); + ff_me_cmp_init(&enc->mecc, avctx); + ff_mpegvideoencdsp_init(&enc->mpvencdsp, avctx); ff_snow_alloc_blocks(s); s->version=0; - s->m.avctx = avctx; - s->m.bit_rate= avctx->bit_rate; - s->m.lmin = avctx->mb_lmin; - s->m.lmax = avctx->mb_lmax; - s->m.mb_num = (avctx->width * avctx->height + 255) / 256; // For ratecontrol - - s->m.me.temp = - s->m.me.scratchpad = av_calloc(avctx->width + 64, 2*16*2*sizeof(uint8_t)); - s->m.sc.obmc_scratchpad= av_mallocz(MB_SIZE*MB_SIZE*12*sizeof(uint32_t)); - s->m.me.map = av_mallocz(2 * ME_MAP_SIZE * sizeof(*s->m.me.map)); - if (!s->m.me.scratchpad || !s->m.me.map || !s->m.sc.obmc_scratchpad) + mpv->avctx = avctx; + mpv->bit_rate= avctx->bit_rate; + mpv->lmin = avctx->mb_lmin; + mpv->lmax = avctx->mb_lmax; + mpv->mb_num = (avctx->width * avctx->height + 255) / 256; // For ratecontrol + + mpv->me.temp = + mpv->me.scratchpad = av_calloc(avctx->width + 64, 2*16*2*sizeof(uint8_t)); + mpv->sc.obmc_scratchpad= av_mallocz(MB_SIZE*MB_SIZE*12*sizeof(uint32_t)); + mpv->me.map = av_mallocz(2 * ME_MAP_SIZE * sizeof(*mpv->me.map)); + if (!mpv->me.scratchpad || !mpv->me.map || !mpv->sc.obmc_scratchpad) return AVERROR(ENOMEM); - s->m.me.score_map = s->m.me.map + ME_MAP_SIZE; + mpv->me.score_map = mpv->me.map + ME_MAP_SIZE; - ff_h263_encode_init(&s->m); //mv_penalty + ff_h263_encode_init(mpv); //mv_penalty s->max_ref_frames = av_clip(avctx->refs, 1, MAX_REF_FRAMES); @@ -216,11 +247,11 @@ static av_cold int encode_init(AVCodecContext *avctx) return AVERROR(ENOMEM); } if((avctx->flags&AV_CODEC_FLAG_PASS2) || !(avctx->flags&AV_CODEC_FLAG_QSCALE)){ - ret = ff_rate_control_init(&s->m); + ret = ff_rate_control_init(mpv); if(ret < 0) return ret; } - s->pass1_rc= !(avctx->flags & (AV_CODEC_FLAG_QSCALE|AV_CODEC_FLAG_PASS2)); + enc->pass1_rc = !(avctx->flags & (AV_CODEC_FLAG_QSCALE|AV_CODEC_FLAG_PASS2)); switch(avctx->pix_fmt){ case AV_PIX_FMT_YUV444P: @@ -245,8 +276,8 @@ static av_cold int encode_init(AVCodecContext *avctx) if (ret) return ret; - ret = ff_set_cmp(&s->mecc, s->mecc.me_cmp, s->avctx->me_cmp); - ret |= ff_set_cmp(&s->mecc, s->mecc.me_sub_cmp, s->avctx->me_sub_cmp); + ret = ff_set_cmp(&enc->mecc, enc->mecc.me_cmp, s->avctx->me_cmp); + ret |= ff_set_cmp(&enc->mecc, enc->mecc.me_sub_cmp, s->avctx->me_sub_cmp); if (ret < 0) return AVERROR(EINVAL); @@ -257,7 +288,7 @@ static av_cold int encode_init(AVCodecContext *avctx) if ((ret = get_encode_buffer(s, s->input_picture)) < 0) return ret; - if(s->motion_est == FF_ME_ITER){ + if (enc->motion_est == FF_ME_ITER) { int size= s->b_width * s->b_height << 2*s->block_max_depth; for(i=0; imax_ref_frames; i++){ s->ref_mvs[i] = av_calloc(size, sizeof(*s->ref_mvs[i])); @@ -335,7 +366,10 @@ static inline int get_penalty_factor(int lambda, int lambda2, int type){ #define P_MV1 P[9] #define FLAG_QPEL 1 //must be 1 -static int encode_q_branch(SnowContext *s, int level, int x, int y){ +static int encode_q_branch(SnowEncContext *enc, int level, int x, int y) +{ + SnowContext *const s = &enc->com; + MotionEstContext *const c = &enc->m.me; uint8_t p_buffer[1024]; uint8_t i_buffer[1024]; uint8_t p_state[sizeof(s->block_state)]; @@ -372,7 +406,6 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ int16_t last_mv[3][2]; int qpel= !!(s->avctx->flags & AV_CODEC_FLAG_QPEL); //unused const int shift= 1+qpel; - MotionEstContext *c= &s->m.me; int ref_context= av_log2(2*left->ref) + av_log2(2*top->ref); int mx_context= av_log2(2*FFABS(left->mx - top->mx)); int my_context= av_log2(2*FFABS(left->my - top->my)); @@ -401,18 +434,18 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ last_mv[2][0]= bottom->mx; last_mv[2][1]= bottom->my; - s->m.mb_stride=2; - s->m.mb_x= - s->m.mb_y= 0; + enc->m.mb_stride = 2; + enc->m.mb_x = + enc->m.mb_y = 0; c->skip= 0; av_assert1(c-> stride == stride); av_assert1(c->uvstride == uvstride); - c->penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_cmp); - c->sub_penalty_factor= get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_sub_cmp); - c->mb_penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->mb_cmp); - c->current_mv_penalty= c->mv_penalty[s->m.f_code=1] + MAX_DMV; + c->penalty_factor = get_penalty_factor(enc->lambda, enc->lambda2, c->avctx->me_cmp); + c->sub_penalty_factor= get_penalty_factor(enc->lambda, enc->lambda2, c->avctx->me_sub_cmp); + c->mb_penalty_factor = get_penalty_factor(enc->lambda, enc->lambda2, c->avctx->mb_cmp); + c->current_mv_penalty = c->mv_penalty[enc->m.f_code=1] + MAX_DMV; c->xmin = - x*block_w - 16+3; c->ymin = - y*block_w - 16+3; @@ -443,7 +476,7 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ for(ref=0; refref_frames; ref++){ init_ref(c, current_data, s->last_picture[ref]->data, NULL, block_w*x, block_w*y, 0); - ref_score= ff_epzs_motion_search(&s->m, &ref_mx, &ref_my, P, 0, /*ref_index*/ 0, last_mv, + ref_score= ff_epzs_motion_search(&enc->m, &ref_mx, &ref_my, P, 0, /*ref_index*/ 0, last_mv, (1<<16)>>shift, level-LOG2_MB_SIZE+4, block_w); av_assert2(ref_mx >= c->xmin); @@ -451,8 +484,8 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ av_assert2(ref_my >= c->ymin); av_assert2(ref_my <= c->ymax); - ref_score= c->sub_motion_search(&s->m, &ref_mx, &ref_my, ref_score, 0, 0, level-LOG2_MB_SIZE+4, block_w); - ref_score= ff_get_mb_score(&s->m, ref_mx, ref_my, 0, 0, level-LOG2_MB_SIZE+4, block_w, 0); + ref_score= c->sub_motion_search(&enc->m, &ref_mx, &ref_my, ref_score, 0, 0, level-LOG2_MB_SIZE+4, block_w); + ref_score= ff_get_mb_score(&enc->m, ref_mx, ref_my, 0, 0, level-LOG2_MB_SIZE+4, block_w, 0); ref_score+= 2*av_log2(2*ref)*c->penalty_factor; if(s->ref_mvs[ref]){ s->ref_mvs[ref][index][0]= ref_mx; @@ -484,7 +517,7 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ put_symbol(&pc, &p_state[128 + 32*(mx_context + 16*!!best_ref)], mx - pmx, 1); put_symbol(&pc, &p_state[128 + 32*(my_context + 16*!!best_ref)], my - pmy, 1); p_len= pc.bytestream - pc.bytestream_start; - score += (s->lambda2*(get_rac_count(&pc)-base_bits))>>FF_LAMBDA_SHIFT; + score += (enc->lambda2*(get_rac_count(&pc)-base_bits))>>FF_LAMBDA_SHIFT; block_s= block_w*block_w; sum = pix_sum(current_data[0], stride, block_w, block_w); @@ -515,9 +548,9 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ put_symbol(&ic, &i_state[96], cr-pcr, 1); } i_len= ic.bytestream - ic.bytestream_start; - iscore += (s->lambda2*(get_rac_count(&ic)-base_bits))>>FF_LAMBDA_SHIFT; + iscore += (enc->lambda2*(get_rac_count(&ic)-base_bits))>>FF_LAMBDA_SHIFT; - av_assert1(iscore < 255*255*256 + s->lambda2*10); + av_assert1(iscore < 255*255*256 + enc->lambda2*10); av_assert1(iscore >= 0); av_assert1(l>=0 && l<=255); av_assert1(pl>=0 && pl<=255); @@ -528,16 +561,16 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ if (vard <= 64 || vard < varc) c->scene_change_score+= ff_sqrt(vard) - ff_sqrt(varc); else - c->scene_change_score+= s->m.qscale; + c->scene_change_score += enc->m.qscale; } if(level!=s->block_max_depth){ put_rac(&s->c, &s->block_state[4 + s_context], 0); - score2 = encode_q_branch(s, level+1, 2*x+0, 2*y+0); - score2+= encode_q_branch(s, level+1, 2*x+1, 2*y+0); - score2+= encode_q_branch(s, level+1, 2*x+0, 2*y+1); - score2+= encode_q_branch(s, level+1, 2*x+1, 2*y+1); - score2+= s->lambda2>>FF_LAMBDA_SHIFT; //FIXME exact split overhead + score2 = encode_q_branch(enc, level+1, 2*x+0, 2*y+0); + score2+= encode_q_branch(enc, level+1, 2*x+1, 2*y+0); + score2+= encode_q_branch(enc, level+1, 2*x+0, 2*y+1); + score2+= encode_q_branch(enc, level+1, 2*x+1, 2*y+1); + score2+= enc->lambda2>>FF_LAMBDA_SHIFT; //FIXME exact split overhead if(score2 < score && score2 < iscore) return score2; @@ -619,7 +652,9 @@ static void encode_q_branch2(SnowContext *s, int level, int x, int y){ } } -static int get_dc(SnowContext *s, int mb_x, int mb_y, int plane_index){ +static int get_dc(SnowEncContext *enc, int mb_x, int mb_y, int plane_index) +{ + SnowContext *const s = &enc->com; int i, x2, y2; Plane *p= &s->plane[plane_index]; const int block_size = MB_SIZE >> s->block_max_depth; @@ -629,7 +664,7 @@ static int get_dc(SnowContext *s, int mb_x, int mb_y, int plane_index){ const int obmc_stride= plane_index ? (2*block_size)>>s->chroma_h_shift : 2*block_size; const int ref_stride= s->current_picture->linesize[plane_index]; const uint8_t *src = s->input_picture->data[plane_index]; - IDWTELEM *dst= (IDWTELEM*)s->m.sc.obmc_scratchpad + plane_index*block_size*block_size*4; //FIXME change to unsigned + IDWTELEM *dst= (IDWTELEM*)enc->m.sc.obmc_scratchpad + plane_index*block_size*block_size*4; //FIXME change to unsigned const int b_stride = s->b_width << s->block_max_depth; const int w= p->width; const int h= p->height; @@ -715,7 +750,10 @@ static inline int get_block_bits(SnowContext *s, int x, int y, int w){ } } -static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, uint8_t (*obmc_edged)[MB_SIZE * 2]){ +static int get_block_rd(SnowEncContext *enc, int mb_x, int mb_y, + int plane_index, uint8_t (*obmc_edged)[MB_SIZE * 2]) +{ + SnowContext *const s = &enc->com; Plane *p= &s->plane[plane_index]; const int block_size = MB_SIZE >> s->block_max_depth; const int block_w = plane_index ? block_size>>s->chroma_h_shift : block_size; @@ -724,7 +762,7 @@ static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, uin const int ref_stride= s->current_picture->linesize[plane_index]; uint8_t *dst= s->current_picture->data[plane_index]; const uint8_t *src = s->input_picture->data[plane_index]; - IDWTELEM *pred= (IDWTELEM*)s->m.sc.obmc_scratchpad + plane_index*block_size*block_size*4; + IDWTELEM *pred= (IDWTELEM*)enc->m.sc.obmc_scratchpad + plane_index*block_size*block_size*4; uint8_t *cur = s->scratchbuf; uint8_t *tmp = s->emu_edge_buffer; const int b_stride = s->b_width << s->block_max_depth; @@ -733,7 +771,7 @@ static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, uin const int h= p->height; int distortion; int rate= 0; - const int penalty_factor= get_penalty_factor(s->lambda, s->lambda2, s->avctx->me_cmp); + const int penalty_factor = get_penalty_factor(enc->lambda, enc->lambda2, s->avctx->me_cmp); int sx= block_w*mb_x - block_w/2; int sy= block_h*mb_y - block_h/2; int x0= FFMAX(0,-sx); @@ -787,19 +825,19 @@ static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, uin * to improve the score of the whole frame, thus iterative motion * estimation does not always converge. */ if(s->avctx->me_cmp == FF_CMP_W97) - distortion = ff_w97_32_c(&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, 32); + distortion = ff_w97_32_c(&enc->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, 32); else if(s->avctx->me_cmp == FF_CMP_W53) - distortion = ff_w53_32_c(&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, 32); + distortion = ff_w53_32_c(&enc->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, 32); else{ distortion = 0; for(i=0; i<4; i++){ int off = sx+16*(i&1) + (sy+16*(i>>1))*ref_stride; - distortion += s->mecc.me_cmp[0](&s->m, src + off, dst + off, ref_stride, 16); + distortion += enc->mecc.me_cmp[0](&enc->m, src + off, dst + off, ref_stride, 16); } } }else{ av_assert2(block_w==8); - distortion = s->mecc.me_cmp[0](&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, block_w*2); + distortion = enc->mecc.me_cmp[0](&enc->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, block_w*2); } if(plane_index==0){ @@ -816,7 +854,9 @@ static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, uin return distortion + rate*penalty_factor; } -static int get_4block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index){ +static int get_4block_rd(SnowEncContext *enc, int mb_x, int mb_y, int plane_index) +{ + SnowContext *const s = &enc->com; int i, y2; Plane *p= &s->plane[plane_index]; const int block_size = MB_SIZE >> s->block_max_depth; @@ -835,7 +875,7 @@ static int get_4block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index){ const int h= p->height; int distortion= 0; int rate= 0; - const int penalty_factor= get_penalty_factor(s->lambda, s->lambda2, s->avctx->me_cmp); + const int penalty_factor= get_penalty_factor(enc->lambda, enc->lambda2, s->avctx->me_cmp); av_assert2(s->chroma_h_shift == s->chroma_v_shift); //obmc and square assumtions below @@ -863,7 +903,7 @@ static int get_4block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index){ } av_assert1(block_w== 8 || block_w==16); - distortion += s->mecc.me_cmp[block_w==8](&s->m, src + x + y*ref_stride, dst + x + y*ref_stride, ref_stride, block_h); + distortion += enc->mecc.me_cmp[block_w==8](&enc->m, src + x + y*ref_stride, dst + x + y*ref_stride, ref_stride, block_h); } if(plane_index==0){ @@ -1012,9 +1052,10 @@ static int encode_subband(SnowContext *s, SubBand *b, const IDWTELEM *src, const // encode_subband_dzr(s, b, src, parent, stride, orientation); } -static av_always_inline int check_block_intra(SnowContext *s, int mb_x, int mb_y, int p[3], +static av_always_inline int check_block_intra(SnowEncContext *enc, int mb_x, int mb_y, int p[3], uint8_t (*obmc_edged)[MB_SIZE * 2], int *best_rd) { + SnowContext *const s = &enc->com; const int b_stride= s->b_width << s->block_max_depth; BlockNode *block= &s->block[mb_x + mb_y * b_stride]; BlockNode backup= *block; @@ -1028,7 +1069,7 @@ static av_always_inline int check_block_intra(SnowContext *s, int mb_x, int mb_y block->color[2] = p[2]; block->type |= BLOCK_INTRA; - rd = get_block_rd(s, mb_x, mb_y, 0, obmc_edged) + s->intra_penalty; + rd = get_block_rd(enc, mb_x, mb_y, 0, obmc_edged) + enc->intra_penalty; //FIXME chroma if(rd < *best_rd){ @@ -1042,7 +1083,11 @@ static av_always_inline int check_block_intra(SnowContext *s, int mb_x, int mb_y /* special case for int[2] args we discard afterwards, * fixes compilation problem with gcc 2.95 */ -static av_always_inline int check_block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, uint8_t (*obmc_edged)[MB_SIZE * 2], int *best_rd){ +static av_always_inline int check_block_inter(SnowEncContext *enc, + int mb_x, int mb_y, int p0, int p1, + uint8_t (*obmc_edged)[MB_SIZE * 2], int *best_rd) +{ + SnowContext *const s = &enc->com; const int b_stride = s->b_width << s->block_max_depth; BlockNode *block = &s->block[mb_x + mb_y * b_stride]; BlockNode backup = *block; @@ -1053,16 +1098,16 @@ static av_always_inline int check_block_inter(SnowContext *s, int mb_x, int mb_y av_assert2(mb_x < b_stride); index = (p0 + 31 * p1) & (ME_CACHE_SIZE-1); - value = s->me_cache_generation + (p0 >> 10) + p1 * (1 << 6) + (block->ref << 12); - if (s->me_cache[index] == value) + value = enc->me_cache_generation + (p0 >> 10) + p1 * (1 << 6) + (block->ref << 12); + if (enc->me_cache[index] == value) return 0; - s->me_cache[index] = value; + enc->me_cache[index] = value; block->mx = p0; block->my = p1; block->type &= ~BLOCK_INTRA; - rd = get_block_rd(s, mb_x, mb_y, 0, obmc_edged); + rd = get_block_rd(enc, mb_x, mb_y, 0, obmc_edged); //FIXME chroma if (rd < *best_rd) { @@ -1074,7 +1119,10 @@ static av_always_inline int check_block_inter(SnowContext *s, int mb_x, int mb_y } } -static av_always_inline int check_4block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, int ref, int *best_rd){ +static av_always_inline int check_4block_inter(SnowEncContext *enc, int mb_x, int mb_y, + int p0, int p1, int ref, int *best_rd) +{ + SnowContext *const s = &enc->com; const int b_stride= s->b_width << s->block_max_depth; BlockNode *block= &s->block[mb_x + mb_y * b_stride]; BlockNode backup[4]; @@ -1094,10 +1142,10 @@ static av_always_inline int check_4block_inter(SnowContext *s, int mb_x, int mb_ av_assert2(((mb_x|mb_y)&1) == 0); index= (p0 + 31*p1) & (ME_CACHE_SIZE-1); - value= s->me_cache_generation + (p0>>10) + (p1<<6) + (block->ref<<12); - if(s->me_cache[index] == value) + value = enc->me_cache_generation + (p0>>10) + (p1<<6) + (block->ref<<12); + if (enc->me_cache[index] == value) return 0; - s->me_cache[index]= value; + enc->me_cache[index] = value; block->mx= p0; block->my= p1; @@ -1105,7 +1153,7 @@ static av_always_inline int check_4block_inter(SnowContext *s, int mb_x, int mb_ block->type &= ~BLOCK_INTRA; block[1]= block[b_stride]= block[b_stride+1]= *block; - rd= get_4block_rd(s, mb_x, mb_y, 0); + rd = get_4block_rd(enc, mb_x, mb_y, 0); //FIXME chroma if(rd < *best_rd){ @@ -1120,7 +1168,9 @@ static av_always_inline int check_4block_inter(SnowContext *s, int mb_x, int mb_ } } -static void iterative_me(SnowContext *s){ +static void iterative_me(SnowEncContext *enc) +{ + SnowContext *const s = &enc->com; int pass, mb_x, mb_y; const int b_width = s->b_width << s->block_max_depth; const int b_height= s->b_height << s->block_max_depth; @@ -1133,7 +1183,7 @@ static void iterative_me(SnowContext *s){ memcpy(state, s->block_state, sizeof(s->block_state)); for(mb_y= 0; mb_yb_height; mb_y++) for(mb_x= 0; mb_xb_width; mb_x++) - encode_q_branch(s, 0, mb_x, mb_y); + encode_q_branch(enc, 0, mb_x, mb_y); s->c = r; memcpy(s->block_state, state, sizeof(s->block_state)); } @@ -1165,9 +1215,9 @@ static void iterative_me(SnowContext *s){ backup= *block; - if(!s->me_cache_generation) - memset(s->me_cache, 0, sizeof(s->me_cache)); - s->me_cache_generation += 1<<22; + if (!enc->me_cache_generation) + memset(enc->me_cache, 0, sizeof(enc->me_cache)); + enc->me_cache_generation += 1<<22; //FIXME precalculate { @@ -1223,14 +1273,14 @@ static void iterative_me(SnowContext *s){ // intra(black) = neighbors' contribution to the current block for(i=0; i < s->nb_planes; i++) - color[i]= get_dc(s, mb_x, mb_y, i); + color[i]= get_dc(enc, mb_x, mb_y, i); // get previous score (cannot be cached due to OBMC) if(pass > 0 && (block->type&BLOCK_INTRA)){ int color0[3]= {block->color[0], block->color[1], block->color[2]}; - check_block_intra(s, mb_x, mb_y, color0, obmc_edged, &best_rd); + check_block_intra(enc, mb_x, mb_y, color0, obmc_edged, &best_rd); }else - check_block_inter(s, mb_x, mb_y, block->mx, block->my, obmc_edged, &best_rd); + check_block_inter(enc, mb_x, mb_y, block->mx, block->my, obmc_edged, &best_rd); ref_b= *block; ref_rd= best_rd; @@ -1241,30 +1291,30 @@ static void iterative_me(SnowContext *s){ block->ref= ref; best_rd= INT_MAX; - check_block_inter(s, mb_x, mb_y, mvr[0][0], mvr[0][1], obmc_edged, &best_rd); - check_block_inter(s, mb_x, mb_y, 0, 0, obmc_edged, &best_rd); + check_block_inter(enc, mb_x, mb_y, mvr[0][0], mvr[0][1], obmc_edged, &best_rd); + check_block_inter(enc, mb_x, mb_y, 0, 0, obmc_edged, &best_rd); if(tb) - check_block_inter(s, mb_x, mb_y, mvr[-b_stride][0], mvr[-b_stride][1], obmc_edged, &best_rd); + check_block_inter(enc, mb_x, mb_y, mvr[-b_stride][0], mvr[-b_stride][1], obmc_edged, &best_rd); if(lb) - check_block_inter(s, mb_x, mb_y, mvr[-1][0], mvr[-1][1], obmc_edged, &best_rd); + check_block_inter(enc, mb_x, mb_y, mvr[-1][0], mvr[-1][1], obmc_edged, &best_rd); if(rb) - check_block_inter(s, mb_x, mb_y, mvr[1][0], mvr[1][1], obmc_edged, &best_rd); + check_block_inter(enc, mb_x, mb_y, mvr[1][0], mvr[1][1], obmc_edged, &best_rd); if(bb) - check_block_inter(s, mb_x, mb_y, mvr[b_stride][0], mvr[b_stride][1], obmc_edged, &best_rd); + check_block_inter(enc, mb_x, mb_y, mvr[b_stride][0], mvr[b_stride][1], obmc_edged, &best_rd); /* fullpel ME */ //FIXME avoid subpel interpolation / round to nearest integer do{ int newx = block->mx; int newy = block->my; - int dia_size = s->iterative_dia_size ? s->iterative_dia_size : FFMAX(s->avctx->dia_size, 1); + int dia_size = enc->iterative_dia_size ? enc->iterative_dia_size : FFMAX(s->avctx->dia_size, 1); dia_change=0; for(i=0; i < dia_size; i++){ for(j=0; jmx+square[i][0], block->my+square[i][1], obmc_edged, &best_rd); + dia_change |= check_block_inter(enc, mb_x, mb_y, block->mx+square[i][0], block->my+square[i][1], obmc_edged, &best_rd); }while(dia_change); //FIXME or try the standard 2 pass qpel or similar @@ -1286,7 +1336,7 @@ static void iterative_me(SnowContext *s){ } best_rd= ref_rd; *block= ref_b; - check_block_intra(s, mb_x, mb_y, color, obmc_edged, &best_rd); + check_block_intra(enc, mb_x, mb_y, color, obmc_edged, &best_rd); //FIXME RD style color selection if(!same_block(block, &backup)){ if(tb ) tb ->type &= ~BLOCK_OPT; @@ -1324,20 +1374,20 @@ static void iterative_me(SnowContext *s){ same_block(b[0], b[3])) continue; - if(!s->me_cache_generation) - memset(s->me_cache, 0, sizeof(s->me_cache)); - s->me_cache_generation += 1<<22; + if (!enc->me_cache_generation) + memset(enc->me_cache, 0, sizeof(enc->me_cache)); + enc->me_cache_generation += 1<<22; - init_rd= best_rd= get_4block_rd(s, mb_x, mb_y, 0); + init_rd = best_rd = get_4block_rd(enc, mb_x, mb_y, 0); //FIXME more multiref search? - check_4block_inter(s, mb_x, mb_y, + check_4block_inter(enc, mb_x, mb_y, (b[0]->mx + b[1]->mx + b[2]->mx + b[3]->mx + 2) >> 2, (b[0]->my + b[1]->my + b[2]->my + b[3]->my + 2) >> 2, 0, &best_rd); for(i=0; i<4; i++) if(!(b[i]->type&BLOCK_INTRA)) - check_4block_inter(s, mb_x, mb_y, b[i]->mx, b[i]->my, b[i]->ref, &best_rd); + check_4block_inter(enc, mb_x, mb_y, b[i]->mx, b[i]->my, b[i]->ref, &best_rd); if(init_rd != best_rd) change++; @@ -1347,13 +1397,15 @@ static void iterative_me(SnowContext *s){ } } -static void encode_blocks(SnowContext *s, int search){ +static void encode_blocks(SnowEncContext *enc, int search) +{ + SnowContext *const s = &enc->com; int x, y; int w= s->b_width; int h= s->b_height; - if(s->motion_est == FF_ME_ITER && !s->keyframe && search) - iterative_me(s); + if (enc->motion_est == FF_ME_ITER && !s->keyframe && search) + iterative_me(enc); for(y=0; yc.bytestream_end - s->c.bytestream < w*MB_SIZE*MB_SIZE*3){ //FIXME nicer limit @@ -1361,10 +1413,10 @@ static void encode_blocks(SnowContext *s, int search){ return; } for(x=0; xmotion_est == FF_ME_ITER || !search) + if (enc->motion_est == FF_ME_ITER || !search) encode_q_branch2(s, 0, x, y); else - encode_q_branch (s, 0, x, y); + encode_q_branch (enc, 0, x, y); } } } @@ -1611,8 +1663,9 @@ static int qscale2qlog(int qscale){ + 61*QROOT/8; ///< 64 > 60 } -static int ratecontrol_1pass(SnowContext *s, AVFrame *pict) +static int ratecontrol_1pass(SnowEncContext *enc, AVFrame *pict) { + SnowContext *const s = &enc->com; /* Estimate the frame's complexity as a sum of weighted dwt coefficients. * FIXME we know exact mv bits at this point, * but ratecontrol isn't set up to include them. */ @@ -1647,17 +1700,17 @@ static int ratecontrol_1pass(SnowContext *s, AVFrame *pict) coef_sum = (uint64_t)coef_sum * coef_sum >> 16; if(pict->pict_type == AV_PICTURE_TYPE_I){ - s->m.mb_var_sum = coef_sum; - s->m.mc_mb_var_sum = 0; + enc->m.mb_var_sum = coef_sum; + enc->m.mc_mb_var_sum = 0; }else{ - s->m.mc_mb_var_sum = coef_sum; - s->m.mb_var_sum = 0; + enc->m.mc_mb_var_sum = coef_sum; + enc->m.mb_var_sum = 0; } - pict->quality= ff_rate_estimate_qscale(&s->m, 1); + pict->quality= ff_rate_estimate_qscale(&enc->m, 1); if (pict->quality < 0) return INT_MIN; - s->lambda= pict->quality * 3/2; + enc->lambda= pict->quality * 3/2; delta_qlog= qscale2qlog(pict->quality) - s->qlog; s->qlog+= delta_qlog; return delta_qlog; @@ -1696,7 +1749,9 @@ static void calculate_visual_weight(SnowContext *s, Plane *p){ static int encode_frame(AVCodecContext *avctx, AVPacket *pkt, const AVFrame *pict, int *got_packet) { - SnowContext *s = avctx->priv_data; + SnowEncContext *const enc = avctx->priv_data; + SnowContext *const s = &enc->com; + MpegEncContext *const mpv = &enc->m; RangeCoder * const c= &s->c; AVCodecInternal *avci = avctx->internal; AVFrame *pic; @@ -1719,7 +1774,7 @@ static int encode_frame(AVCodecContext *avctx, AVPacket *pkt, memcpy(&s->input_picture->data[i][y * s->input_picture->linesize[i]], &pict->data[i][y * pict->linesize[i]], AV_CEIL_RSHIFT(width, hshift)); - s->mpvencdsp.draw_edges(s->input_picture->data[i], s->input_picture->linesize[i], + enc->mpvencdsp.draw_edges(s->input_picture->data[i], s->input_picture->linesize[i], AV_CEIL_RSHIFT(width, hshift), AV_CEIL_RSHIFT(height, vshift), EDGE_WIDTH >> hshift, EDGE_WIDTH >> vshift, EDGE_TOP | EDGE_BOTTOM); @@ -1730,45 +1785,45 @@ static int encode_frame(AVCodecContext *avctx, AVPacket *pkt, pic->pict_type = pict->pict_type; pic->quality = pict->quality; - s->m.picture_number= avctx->frame_num; + mpv->picture_number = avctx->frame_num; if(avctx->flags&AV_CODEC_FLAG_PASS2){ - s->m.pict_type = pic->pict_type = s->m.rc_context.entry[avctx->frame_num].new_pict_type; + mpv->pict_type = pic->pict_type = mpv->rc_context.entry[avctx->frame_num].new_pict_type; s->keyframe = pic->pict_type == AV_PICTURE_TYPE_I; if(!(avctx->flags&AV_CODEC_FLAG_QSCALE)) { - pic->quality = ff_rate_estimate_qscale(&s->m, 0); + pic->quality = ff_rate_estimate_qscale(mpv, 0); if (pic->quality < 0) return -1; } }else{ s->keyframe= avctx->gop_size==0 || avctx->frame_num % avctx->gop_size == 0; - s->m.pict_type = pic->pict_type = s->keyframe ? AV_PICTURE_TYPE_I : AV_PICTURE_TYPE_P; + mpv->pict_type = pic->pict_type = s->keyframe ? AV_PICTURE_TYPE_I : AV_PICTURE_TYPE_P; } - if(s->pass1_rc && avctx->frame_num == 0) + if (enc->pass1_rc && avctx->frame_num == 0) pic->quality = 2*FF_QP2LAMBDA; if (pic->quality) { s->qlog = qscale2qlog(pic->quality); - s->lambda = pic->quality * 3/2; + enc->lambda = pic->quality * 3/2; } if (s->qlog < 0 || (!pic->quality && (avctx->flags & AV_CODEC_FLAG_QSCALE))) { s->qlog= LOSSLESS_QLOG; - s->lambda = 0; + enc->lambda = 0; }//else keep previous frame's qlog until after motion estimation if (s->current_picture->data[0]) { int w = s->avctx->width; int h = s->avctx->height; - s->mpvencdsp.draw_edges(s->current_picture->data[0], - s->current_picture->linesize[0], w , h , - EDGE_WIDTH , EDGE_WIDTH , EDGE_TOP | EDGE_BOTTOM); + enc->mpvencdsp.draw_edges(s->current_picture->data[0], + s->current_picture->linesize[0], w , h , + EDGE_WIDTH , EDGE_WIDTH , EDGE_TOP | EDGE_BOTTOM); if (s->current_picture->data[2]) { - s->mpvencdsp.draw_edges(s->current_picture->data[1], - s->current_picture->linesize[1], w>>s->chroma_h_shift, h>>s->chroma_v_shift, - EDGE_WIDTH>>s->chroma_h_shift, EDGE_WIDTH>>s->chroma_v_shift, EDGE_TOP | EDGE_BOTTOM); - s->mpvencdsp.draw_edges(s->current_picture->data[2], - s->current_picture->linesize[2], w>>s->chroma_h_shift, h>>s->chroma_v_shift, - EDGE_WIDTH>>s->chroma_h_shift, EDGE_WIDTH>>s->chroma_v_shift, EDGE_TOP | EDGE_BOTTOM); + enc->mpvencdsp.draw_edges(s->current_picture->data[1], + s->current_picture->linesize[1], w>>s->chroma_h_shift, h>>s->chroma_v_shift, + EDGE_WIDTH>>s->chroma_h_shift, EDGE_WIDTH>>s->chroma_v_shift, EDGE_TOP | EDGE_BOTTOM); + enc->mpvencdsp.draw_edges(s->current_picture->data[2], + s->current_picture->linesize[2], w>>s->chroma_h_shift, h>>s->chroma_v_shift, + EDGE_WIDTH>>s->chroma_h_shift, EDGE_WIDTH>>s->chroma_v_shift, EDGE_TOP | EDGE_BOTTOM); } emms_c(); } @@ -1778,9 +1833,9 @@ static int encode_frame(AVCodecContext *avctx, AVPacket *pkt, if (ret < 0) return ret; - s->m.current_picture_ptr= &s->m.current_picture; - s->m.current_picture.f = s->current_picture; - s->m.current_picture.f->pts = pict->pts; + mpv->current_picture_ptr = &mpv->current_picture; + mpv->current_picture.f = s->current_picture; + mpv->current_picture.f->pts = pict->pts; if(pic->pict_type == AV_PICTURE_TYPE_P){ int block_width = (width +15)>>4; int block_height= (height+15)>>4; @@ -1789,40 +1844,40 @@ static int encode_frame(AVCodecContext *avctx, AVPacket *pkt, av_assert0(s->current_picture->data[0]); av_assert0(s->last_picture[0]->data[0]); - s->m.avctx= s->avctx; - s->m. last_picture.f = s->last_picture[0]; - s->m. new_picture = s->input_picture; - s->m. last_picture_ptr= &s->m. last_picture; - s->m.linesize = stride; - s->m.uvlinesize= s->current_picture->linesize[1]; - s->m.width = width; - s->m.height= height; - s->m.mb_width = block_width; - s->m.mb_height= block_height; - s->m.mb_stride= s->m.mb_width+1; - s->m.b8_stride= 2*s->m.mb_width+1; - s->m.f_code=1; - s->m.pict_type = pic->pict_type; - s->m.motion_est= s->motion_est; - s->m.me.scene_change_score=0; - s->m.me.dia_size = avctx->dia_size; - s->m.quarter_sample= (s->avctx->flags & AV_CODEC_FLAG_QPEL)!=0; - s->m.out_format= FMT_H263; - s->m.unrestricted_mv= 1; - - s->m.lambda = s->lambda; - s->m.qscale= (s->m.lambda*139 + FF_LAMBDA_SCALE*64) >> (FF_LAMBDA_SHIFT + 7); - s->lambda2= s->m.lambda2= (s->m.lambda*s->m.lambda + FF_LAMBDA_SCALE/2) >> FF_LAMBDA_SHIFT; - - s->m.mecc= s->mecc; //move - s->m.qdsp= s->qdsp; //move - s->m.hdsp = s->hdsp; - ff_init_me(&s->m); - s->hdsp = s->m.hdsp; - s->mecc= s->m.mecc; - } - - if(s->pass1_rc){ + mpv->avctx = s->avctx; + mpv->last_picture.f = s->last_picture[0]; + mpv-> new_picture = s->input_picture; + mpv->last_picture_ptr = &mpv->last_picture; + mpv->linesize = stride; + mpv->uvlinesize = s->current_picture->linesize[1]; + mpv->width = width; + mpv->height = height; + mpv->mb_width = block_width; + mpv->mb_height = block_height; + mpv->mb_stride = mpv->mb_width + 1; + mpv->b8_stride = 2 * mpv->mb_width + 1; + mpv->f_code = 1; + mpv->pict_type = pic->pict_type; + mpv->motion_est = enc->motion_est; + mpv->me.scene_change_score = 0; + mpv->me.dia_size = avctx->dia_size; + mpv->quarter_sample = (s->avctx->flags & AV_CODEC_FLAG_QPEL)!=0; + mpv->out_format = FMT_H263; + mpv->unrestricted_mv = 1; + + mpv->lambda = enc->lambda; + mpv->qscale = (mpv->lambda*139 + FF_LAMBDA_SCALE*64) >> (FF_LAMBDA_SHIFT + 7); + enc->lambda2 = mpv->lambda2 = (mpv->lambda*mpv->lambda + FF_LAMBDA_SCALE/2) >> FF_LAMBDA_SHIFT; + + mpv->mecc = enc->mecc; //move + mpv->qdsp = enc->qdsp; //move + mpv->hdsp = s->hdsp; + ff_init_me(&enc->m); + s->hdsp = mpv->hdsp; + enc->mecc = mpv->mecc; + } + + if (enc->pass1_rc) { memcpy(rc_header_bak, s->header_state, sizeof(s->header_state)); memcpy(rc_block_bak, s->block_state, sizeof(s->block_state)); } @@ -1840,7 +1895,7 @@ redo_frame: return AVERROR(EINVAL); } - s->m.pict_type = pic->pict_type; + mpv->pict_type = pic->pict_type; s->qbias = pic->pict_type == AV_PICTURE_TYPE_P ? 2 : 0; ff_snow_common_init_after_header(avctx); @@ -1852,9 +1907,9 @@ redo_frame: } encode_header(s); - s->m.misc_bits = 8*(s->c.bytestream - s->c.bytestream_start); - encode_blocks(s, 1); - s->m.mv_bits = 8*(s->c.bytestream - s->c.bytestream_start) - s->m.misc_bits; + mpv->misc_bits = 8 * (s->c.bytestream - s->c.bytestream_start); + encode_blocks(enc, 1); + mpv->mv_bits = 8 * (s->c.bytestream - s->c.bytestream_start) - mpv->misc_bits; for(plane_index=0; plane_index < s->nb_planes; plane_index++){ Plane *p= &s->plane[plane_index]; @@ -1863,7 +1918,7 @@ redo_frame: int x, y; // int bits= put_bits_count(&s->c.pb); - if (!s->memc_only) { + if (!enc->memc_only) { //FIXME optimize if(pict->data[plane_index]) //FIXME gray hack for(y=0; ypict_type == AV_PICTURE_TYPE_P && !(avctx->flags&AV_CODEC_FLAG_PASS2) - && s->m.me.scene_change_score > s->scenechange_threshold){ + && mpv->me.scene_change_score > enc->scenechange_threshold) { ff_init_range_encoder(c, pkt->data, pkt->size); ff_build_rac_states(c, (1LL<<32)/20, 256-8); pic->pict_type= AV_PICTURE_TYPE_I; @@ -1901,8 +1956,8 @@ redo_frame: ff_spatial_dwt(s->spatial_dwt_buffer, s->temp_dwt_buffer, w, h, w, s->spatial_decomposition_type, s->spatial_decomposition_count); - if(s->pass1_rc && plane_index==0){ - int delta_qlog = ratecontrol_1pass(s, pic); + if (enc->pass1_rc && plane_index==0) { + int delta_qlog = ratecontrol_1pass(enc, pic); if (delta_qlog <= INT_MIN) return -1; if(delta_qlog){ @@ -1911,7 +1966,7 @@ redo_frame: memcpy(s->header_state, rc_header_bak, sizeof(s->header_state)); memcpy(s->block_state, rc_block_bak, sizeof(s->block_state)); encode_header(s); - encode_blocks(s, 0); + encode_blocks(enc, 0); } } @@ -1922,7 +1977,7 @@ redo_frame: quantize(s, b, b->ibuf, b->buf, b->stride, s->qbias); if(orientation==0) decorrelate(s, b, b->ibuf, b->stride, pic->pict_type == AV_PICTURE_TYPE_P, 0); - if (!s->no_bitstream) + if (!enc->no_bitstream) encode_subband(s, b, b->ibuf, b->parent ? b->parent->ibuf : NULL, b->stride, orientation); av_assert0(b->parent==NULL || b->parent->stride == b->stride*2); if(orientation==0) @@ -1972,7 +2027,7 @@ redo_frame: } } s->avctx->error[plane_index] += error; - s->encoding_error[plane_index] = error; + enc->encoding_error[plane_index] = error; } } @@ -1984,23 +2039,23 @@ redo_frame: s->current_picture->pict_type = pic->pict_type; s->current_picture->quality = pic->quality; - s->m.frame_bits = 8*(s->c.bytestream - s->c.bytestream_start); - s->m.p_tex_bits = s->m.frame_bits - s->m.misc_bits - s->m.mv_bits; - s->m.current_picture.display_picture_number = - s->m.current_picture.coded_picture_number = avctx->frame_num; - s->m.current_picture.f->quality = pic->quality; - s->m.total_bits += 8*(s->c.bytestream - s->c.bytestream_start); - if(s->pass1_rc) - if (ff_rate_estimate_qscale(&s->m, 0) < 0) + mpv->frame_bits = 8 * (s->c.bytestream - s->c.bytestream_start); + mpv->p_tex_bits = mpv->frame_bits - mpv->misc_bits - mpv->mv_bits; + mpv->total_bits += 8*(s->c.bytestream - s->c.bytestream_start); + mpv->current_picture.display_picture_number = + mpv->current_picture.coded_picture_number = avctx->frame_num; + mpv->current_picture.f->quality = pic->quality; + if (enc->pass1_rc) + if (ff_rate_estimate_qscale(mpv, 0) < 0) return -1; if(avctx->flags&AV_CODEC_FLAG_PASS1) - ff_write_pass1_stats(&s->m); - s->m.last_pict_type = s->m.pict_type; + ff_write_pass1_stats(mpv); + mpv->last_pict_type = mpv->pict_type; emms_c(); ff_side_data_set_encoder_stats(pkt, s->current_picture->quality, - s->encoding_error, + enc->encoding_error, (s->avctx->flags&AV_CODEC_FLAG_PSNR) ? SNOW_MAX_PLANES : 0, s->current_picture->pict_type); if (s->avctx->flags & AV_CODEC_FLAG_RECON_FRAME) { @@ -2017,10 +2072,11 @@ redo_frame: static av_cold int encode_end(AVCodecContext *avctx) { - SnowContext *s = avctx->priv_data; + SnowEncContext *const enc = avctx->priv_data; + SnowContext *const s = &enc->com; ff_snow_common_end(s); - ff_rate_control_uninit(&s->m); + ff_rate_control_uninit(&enc->m); av_frame_free(&s->input_picture); for (int i = 0; i < MAX_REF_FRAMES; i++) { @@ -2028,17 +2084,17 @@ static av_cold int encode_end(AVCodecContext *avctx) av_freep(&s->ref_scores[i]); } - s->m.me.temp = NULL; - av_freep(&s->m.me.scratchpad); - av_freep(&s->m.me.map); - av_freep(&s->m.sc.obmc_scratchpad); + enc->m.me.temp = NULL; + av_freep(&enc->m.me.scratchpad); + av_freep(&enc->m.me.map); + av_freep(&enc->m.sc.obmc_scratchpad); av_freep(&avctx->stats_out); return 0; } -#define OFFSET(x) offsetof(SnowContext, x) +#define OFFSET(x) offsetof(SnowEncContext, x) #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM static const AVOption options[] = { {"motion_est", "motion estimation algorithm", OFFSET(motion_est), AV_OPT_TYPE_INT, {.i64 = FF_ME_EPZS }, FF_ME_ZERO, FF_ME_ITER, VE, "motion_est" }, @@ -2077,7 +2133,7 @@ const FFCodec ff_snow_encoder = { .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_ENCODER_REORDERED_OPAQUE | AV_CODEC_CAP_ENCODER_RECON_FRAME, - .priv_data_size = sizeof(SnowContext), + .priv_data_size = sizeof(SnowEncContext), .init = encode_init, FF_CODEC_ENCODE_CB(encode_frame), .close = encode_end, From patchwork Sat Sep 30 15:07:31 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Andreas Rheinhardt X-Patchwork-Id: 44050 Delivered-To: ffmpegpatchwork2@gmail.com Received: by 2002:a05:6a20:1204:b0:15d:8365:d4b8 with SMTP id v4csp437146pzf; Sat, 30 Sep 2023 08:06:33 -0700 (PDT) X-Google-Smtp-Source: AGHT+IG3i7gfdD/o3JFM06U8Xu7fJv4PiMkQegv+IC+kpNJAk1X0rpYn8IaNx5TNNGxJufcDR2ZH X-Received: by 2002:a17:906:7496:b0:9b2:74a1:6b30 with SMTP id e22-20020a170906749600b009b274a16b30mr7320433ejl.33.1696086392642; Sat, 30 Sep 2023 08:06:32 -0700 (PDT) Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org. [79.124.17.100]) by mx.google.com with ESMTP id n11-20020a170906b30b00b009a223e65d0asi7148641ejz.107.2023.09.30.08.06.32; Sat, 30 Sep 2023 08:06:32 -0700 (PDT) Received-SPF: pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) client-ip=79.124.17.100; Authentication-Results: mx.google.com; dkim=neutral (body hash did not verify) header.i=@outlook.com header.s=selector1 header.b=Jk7hAIaw; arc=fail (body hash mismatch); spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=QUARANTINE dis=NONE) header.from=outlook.com Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 2B4A368CC9F; Sat, 30 Sep 2023 18:06:29 +0300 (EEST) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from EUR01-VE1-obe.outbound.protection.outlook.com (mail-ve1eur01olkn2093.outbound.protection.outlook.com [40.92.66.93]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id 9F2B268CC9F for ; Sat, 30 Sep 2023 18:06:22 +0300 (EEST) ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=fbnGRqIgK3tGUWydaLPrjii+OGfNTed1/RxHEGyBjQ5CCqyc15sqp1PCs7pMCeEQecsE7RkrZ3qqawfqQ7nbl2rlPOWwuGDs6qPnhQvPHPLMNGNClIT1UrFYcMjTuHKoZTBX50eF8xL8gJM4Q/TcRgRSm4eLk+iklMh1K5eJd+a4HYTR3winqdMf9j9TldoZ+MHpSS+alb+axvZja5wNIFz3Qmj18lYjo7y8L/2j6Supoj/zQttoKasGpV+JSPLuL+SI/qgtI9cQcFiBlEBA0HcyNNzOehh/Q4ElPLy83LjvL10Mry54dHss2Wj+aZMp2paCMQocokPGX3xaJG07uw== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=38IKIHPqRyOAPbwgS+uTOzgcsvjaIOk6zQFzsVPPvEE=; b=MG6MILZgiTyY/wZu7wFC/SOoqt+orJuvH+4EFCVHbcv65Eg8rRzUYxo2A31TU/mSGVeUwQzonmAuzzFlNOGei1wVioQxYikR3ARpfD+q2Hfv6qsJO7iqcF7stwbAWPgzmqmLGHiAup1Hp2bxOzZ5DsTMkm5wa4khLNWyYl+pM06/kgDcC503gHSnz1mO59hwyBoJCFhC/09pQ5j7dNCw/ms2NdEx57f9Q+a+b+AVzwyvnpqvh4RR0WSO1QHZKPFR18klAQ3izC29UmX2ghBoWKH0F2tBTO+2iL5cdORWU58brmg0a9fuCaIbi6tc5Jjf04jttBuWfaJq9NxNa8kqtg== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=none; dmarc=none; dkim=none; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=outlook.com; s=selector1; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=38IKIHPqRyOAPbwgS+uTOzgcsvjaIOk6zQFzsVPPvEE=; b=Jk7hAIawoaloyxps6q6J2Noal8Pf3Vp7fr6veKKzINXzORNn1jR6YNa2zi3dIV2s48XAhRyz1q+OcPADCjZDSYdHcH1mEuztnn9eeOdnyR2k+Jnm9HbvU9BjRb1wZxIR3QbIHu+PUVvEdA3fqqwxTNjTGAKz3kKX2XcBrzFKRigBvbm5zc/y7OevTgMzMYPIgbhAeRZuBefc0OxICwMGlT72OrXJ7dDdfsEjdfD3LwFIs/x09KSvk+hShxnymsklaroSFHVGeto4mIWrnGcv90Dfv/UHefLYnr6VNptVKN/OuWPMkIShyF5G5WoHcO+E/vS1ocqn1mOHoqIi34QVXQ== Received: from AS8P250MB0744.EURP250.PROD.OUTLOOK.COM (2603:10a6:20b:541::14) by PR3P250MB0051.EURP250.PROD.OUTLOOK.COM (2603:10a6:102:14e::22) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.6838.28; Sat, 30 Sep 2023 15:06:21 +0000 Received: from AS8P250MB0744.EURP250.PROD.OUTLOOK.COM ([fe80::5e01:aea5:d3a8:cafa]) by AS8P250MB0744.EURP250.PROD.OUTLOOK.COM ([fe80::5e01:aea5:d3a8:cafa%3]) with mapi id 15.20.6792.026; Sat, 30 Sep 2023 15:06:21 +0000 From: Andreas Rheinhardt To: ffmpeg-devel@ffmpeg.org Date: Sat, 30 Sep 2023 17:07:31 +0200 Message-ID: X-Mailer: git-send-email 2.34.1 In-Reply-To: References: X-TMN: [tRN0Pudtz67o6nKu9SCjT4sOjflSXvnM] X-ClientProxiedBy: ZR0P278CA0206.CHEP278.PROD.OUTLOOK.COM (2603:10a6:910:6a::23) To AS8P250MB0744.EURP250.PROD.OUTLOOK.COM (2603:10a6:20b:541::14) X-Microsoft-Original-Message-ID: <20230930150731.2605013-1-andreas.rheinhardt@outlook.com> MIME-Version: 1.0 X-MS-Exchange-MessageSentRepresentingType: 1 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: AS8P250MB0744:EE_|PR3P250MB0051:EE_ X-MS-Office365-Filtering-Correlation-Id: 9da52bbd-318a-455a-4c05-08dbc1c6ce46 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: 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 X-OriginatorOrg: outlook.com X-MS-Exchange-CrossTenant-Network-Message-Id: 9da52bbd-318a-455a-4c05-08dbc1c6ce46 X-MS-Exchange-CrossTenant-AuthSource: AS8P250MB0744.EURP250.PROD.OUTLOOK.COM X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 30 Sep 2023 15:06:21.3742 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 84df9e7f-e9f6-40af-b435-aaaaaaaaaaaa X-MS-Exchange-CrossTenant-RMS-PersistedConsumerOrg: 00000000-0000-0000-0000-000000000000 X-MS-Exchange-Transport-CrossTenantHeadersStamped: PR3P250MB0051 Subject: [FFmpeg-devel] [PATCH 2/2] avcodec/snow: Move dsp helper functions to snow_dwt.h X-BeenThere: ffmpeg-devel@ffmpeg.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: FFmpeg development discussions and patches List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Reply-To: FFmpeg development discussions and patches Cc: Andreas Rheinhardt Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" X-TUID: +sURXlZe3c1R Signed-off-by: Andreas Rheinhardt --- libavcodec/snow.h | 38 -------------------------------------- libavcodec/snow_dwt.h | 40 ++++++++++++++++++++++++++++++++++++++++ libavcodec/x86/snowdsp.c | 1 - 3 files changed, 40 insertions(+), 39 deletions(-) diff --git a/libavcodec/snow.h b/libavcodec/snow.h index 2e61154d0c..a5e2c138cb 100644 --- a/libavcodec/snow.h +++ b/libavcodec/snow.h @@ -176,44 +176,6 @@ extern const uint8_t * const ff_obmc_tab[4]; extern const uint8_t ff_qexp[QROOT]; extern int ff_scale_mv_ref[MAX_REF_FRAMES][MAX_REF_FRAMES]; -/* C bits used by mmx/sse2/altivec */ - -static av_always_inline void snow_interleave_line_header(int * i, int width, IDWTELEM * low, IDWTELEM * high){ - (*i) = (width) - 2; - - if (width & 1){ - low[(*i)+1] = low[((*i)+1)>>1]; - (*i)--; - } -} - -static av_always_inline void snow_interleave_line_footer(int * i, IDWTELEM * low, IDWTELEM * high){ - for (; (*i)>=0; (*i)-=2){ - low[(*i)+1] = high[(*i)>>1]; - low[*i] = low[(*i)>>1]; - } -} - -static av_always_inline void snow_horizontal_compose_lift_lead_out(int i, IDWTELEM * dst, IDWTELEM * src, IDWTELEM * ref, int width, int w, int lift_high, int mul, int add, int shift){ - for(; i> shift); - } - - if((width^lift_high)&1){ - dst[w] = src[w] - ((mul * 2 * ref[w] + add) >> shift); - } -} - -static av_always_inline void snow_horizontal_compose_liftS_lead_out(int i, IDWTELEM * dst, IDWTELEM * src, IDWTELEM * ref, int width, int w){ - for(; i> W_BS); - } - - if(width&1){ - dst[w] = src[w] + ((2 * ref[w] + W_BO + 4 * src[w]) >> W_BS); - } -} - /* common code */ int ff_snow_common_init(AVCodecContext *avctx); diff --git a/libavcodec/snow_dwt.h b/libavcodec/snow_dwt.h index 15b8a3007b..6e7d22c71a 100644 --- a/libavcodec/snow_dwt.h +++ b/libavcodec/snow_dwt.h @@ -24,6 +24,8 @@ #include #include +#include "libavutil/attributes.h" + struct MpegEncContext; typedef int DWTELEM; @@ -91,6 +93,44 @@ typedef struct SnowDWTContext { : ff_slice_buffer_load_line((slice_buf), \ (line_num))) +/* C bits used by mmx/sse2/altivec */ + +static av_always_inline void snow_interleave_line_header(int *i, int width, IDWTELEM *low, IDWTELEM *high) +{ + *i = width - 2; + + if (width & 1) { + low[*i + 1] = low[(*i + 1)>>1]; + (*i)--; + } +} + +static av_always_inline void snow_interleave_line_footer(int *i, IDWTELEM *low, const IDWTELEM *high) +{ + for (; *i >= 0; *i -= 2) { + low[*i + 1] = high[*i >> 1]; + low[*i] = low[*i >> 1]; + } +} + +static av_always_inline void snow_horizontal_compose_lift_lead_out(int i, IDWTELEM *dst, const IDWTELEM *src, const IDWTELEM *ref, int width, int w, int lift_high, int mul, int add, int shift) +{ + for (; i < w; i++) + dst[i] = src[i] - ((mul * (ref[i] + ref[i + 1]) + add) >> shift); + + if ((width ^ lift_high) & 1) + dst[w] = src[w] - ((mul * 2 * ref[w] + add) >> shift); +} + +static av_always_inline void snow_horizontal_compose_liftS_lead_out(int i, IDWTELEM *dst, const IDWTELEM *src, const IDWTELEM *ref, int width, int w) +{ + for (; i < w; i++) + dst[i] = src[i] + ((ref[i] + ref[(i+1)]+W_BO + 4 * src[i]) >> W_BS); + + if (width & 1) + dst[w] = src[w] + ((2 * ref[w] + W_BO + 4 * src[w]) >> W_BS); +} + int ff_slice_buffer_init(slice_buffer *buf, int line_count, int max_allocated_lines, int line_width, IDWTELEM *base_buffer); diff --git a/libavcodec/x86/snowdsp.c b/libavcodec/x86/snowdsp.c index bca1f9bd2e..bd0aa766e5 100644 --- a/libavcodec/x86/snowdsp.c +++ b/libavcodec/x86/snowdsp.c @@ -24,7 +24,6 @@ #include "libavutil/attributes.h" #include "libavutil/cpu.h" #include "libavutil/x86/asm.h" -#include "libavcodec/snow.h" #include "libavcodec/snow_dwt.h" #if HAVE_INLINE_ASM