diff --git a/.gitignore b/.gitignore index 2c3b047695..d27b3bda05 100644 --- a/.gitignore +++ b/.gitignore @@ -22,6 +22,9 @@ ffplay ffprobe ffserver avconv +doc/avoptions_codec.texi +doc/avoptions_format.texi +doc/print_options libavcodec/*_tablegen libavcodec/*_tables.c libavcodec/*_tables.h diff --git a/configure b/configure index 744a342b6e..868ae2bc00 100755 --- a/configure +++ b/configure @@ -1266,6 +1266,7 @@ CONFIG_EXTRA=" h264chroma h264dsp h264pred + h264qpel huffman lgplv3 lpc @@ -1424,7 +1425,7 @@ h263_vaapi_hwaccel_select="vaapi h263_decoder" h263i_decoder_select="h263_decoder" h263p_encoder_select="h263_encoder" h264_crystalhd_decoder_select="crystalhd h264_mp4toannexb_bsf h264_parser" -h264_decoder_select="golomb h264chroma h264dsp h264pred" +h264_decoder_select="golomb h264chroma h264dsp h264pred h264qpel" h264_dxva2_hwaccel_deps="dxva2api_h" h264_dxva2_hwaccel_select="dxva2 h264_decoder" h264_vaapi_hwaccel_select="vaapi h264_decoder" @@ -1485,8 +1486,8 @@ rv10_decoder_select="h263_decoder" rv10_encoder_select="h263_encoder" rv20_decoder_select="h263_decoder" rv20_encoder_select="h263_encoder" -rv30_decoder_select="golomb h264chroma h264pred" -rv40_decoder_select="golomb h264chroma h264pred" +rv30_decoder_select="golomb h264chroma h264pred h264qpel" +rv40_decoder_select="golomb h264chroma h264pred h264qpel" shorten_decoder_select="golomb" sipr_decoder_select="lsp" snow_decoder_select="dwt" @@ -1495,7 +1496,7 @@ sonic_decoder_select="golomb" sonic_encoder_select="golomb" sonic_ls_encoder_select="golomb" svq1_encoder_select="aandct" -svq3_decoder_select="golomb h264chroma h264dsp h264pred" +svq3_decoder_select="golomb h264chroma h264dsp h264pred h264qpel" svq3_decoder_suggest="zlib" theora_decoder_select="vp3_decoder" tiff_decoder_suggest="zlib" @@ -1504,7 +1505,7 @@ truehd_decoder_select="mlp_decoder" tscc_decoder_select="zlib" twinvq_decoder_select="mdct lsp sinewin" vc1_crystalhd_decoder_select="crystalhd" -vc1_decoder_select="h263_decoder h264chroma" +vc1_decoder_select="h263_decoder h264chroma h264qpel" vc1_dxva2_hwaccel_deps="dxva2api_h" vc1_dxva2_hwaccel_select="dxva2 vc1_decoder" vc1_vaapi_hwaccel_select="vaapi vc1_decoder" @@ -1515,7 +1516,7 @@ vorbis_encoder_select="mdct" vp6_decoder_select="huffman" vp6a_decoder_select="vp6_decoder" vp6f_decoder_select="vp6_decoder" -vp8_decoder_select="h264pred" +vp8_decoder_select="h264pred h264qpel" wmapro_decoder_select="mdct sinewin" wmav1_decoder_select="mdct sinewin" wmav1_encoder_select="mdct sinewin" @@ -1544,7 +1545,7 @@ vda_deps="VideoDecodeAcceleration_VDADecoder_h pthreads" vdpau_deps="vdpau_vdpau_h vdpau_vdpau_x11_h" # parsers -h264_parser_select="golomb h264chroma h264dsp h264pred" +h264_parser_select="golomb h264chroma h264dsp h264pred h264qpel" # external libraries libaacplus_encoder_deps="libaacplus" diff --git a/libavcodec/dpcm.c b/libavcodec/dpcm.c index 4915223bc8..b7e620a3e0 100644 --- a/libavcodec/dpcm.c +++ b/libavcodec/dpcm.c @@ -40,6 +40,7 @@ #include "libavutil/intreadwrite.h" #include "avcodec.h" #include "bytestream.h" +#include "mathops.h" typedef struct DPCMContext { AVFrame frame; @@ -173,20 +174,18 @@ static av_cold int dpcm_decode_init(AVCodecContext *avctx) static int dpcm_decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr, AVPacket *avpkt) { - const uint8_t *buf = avpkt->data; int buf_size = avpkt->size; - const uint8_t *buf_end = buf + buf_size; DPCMContext *s = avctx->priv_data; int out = 0, ret; int predictor[2]; int ch = 0; int stereo = s->channels - 1; - int16_t *output_samples; + int16_t *output_samples, *samples_end; + GetByteContext gb; - if (stereo && (buf_size & 1)) { + if (stereo && (buf_size & 1)) buf_size--; - buf_end--; - } + bytestream2_init(&gb, avpkt->data, buf_size); /* calculate output size */ switch(avctx->codec->id) { @@ -221,22 +220,23 @@ static int dpcm_decode_frame(AVCodecContext *avctx, void *data, return ret; } output_samples = (int16_t *)s->frame.data[0]; + samples_end = output_samples + out; switch(avctx->codec->id) { case CODEC_ID_ROQ_DPCM: - buf += 6; + bytestream2_skipu(&gb, 6); if (stereo) { - predictor[1] = (int16_t)(bytestream_get_byte(&buf) << 8); - predictor[0] = (int16_t)(bytestream_get_byte(&buf) << 8); + predictor[1] = sign_extend(bytestream2_get_byteu(&gb) << 8, 16); + predictor[0] = sign_extend(bytestream2_get_byteu(&gb) << 8, 16); } else { - predictor[0] = (int16_t)bytestream_get_le16(&buf); + predictor[0] = sign_extend(bytestream2_get_le16u(&gb), 16); } /* decode the samples */ - while (buf < buf_end) { - predictor[ch] += s->roq_square_array[*buf++]; + while (output_samples < samples_end) { + predictor[ch] += s->roq_square_array[bytestream2_get_byteu(&gb)]; predictor[ch] = av_clip_int16(predictor[ch]); *output_samples++ = predictor[ch]; @@ -246,16 +246,16 @@ static int dpcm_decode_frame(AVCodecContext *avctx, void *data, break; case CODEC_ID_INTERPLAY_DPCM: - buf += 6; /* skip over the stream mask and stream length */ + bytestream2_skipu(&gb, 6); /* skip over the stream mask and stream length */ for (ch = 0; ch < s->channels; ch++) { - predictor[ch] = (int16_t)bytestream_get_le16(&buf); + predictor[ch] = sign_extend(bytestream2_get_le16u(&gb), 16); *output_samples++ = predictor[ch]; } ch = 0; - while (buf < buf_end) { - predictor[ch] += interplay_delta_table[*buf++]; + while (output_samples < samples_end) { + predictor[ch] += interplay_delta_table[bytestream2_get_byteu(&gb)]; predictor[ch] = av_clip_int16(predictor[ch]); *output_samples++ = predictor[ch]; @@ -269,16 +269,19 @@ static int dpcm_decode_frame(AVCodecContext *avctx, void *data, int shift[2] = { 4, 4 }; for (ch = 0; ch < s->channels; ch++) - predictor[ch] = (int16_t)bytestream_get_le16(&buf); + predictor[ch] = sign_extend(bytestream2_get_le16u(&gb), 16); ch = 0; - while (buf < buf_end) { - uint8_t n = *buf++; - int16_t diff = (n & 0xFC) << 8; - if ((n & 0x03) == 3) + while (output_samples < samples_end) { + int diff = bytestream2_get_byteu(&gb); + int n = diff & 3; + + if (n == 3) shift[ch]++; else - shift[ch] -= (2 * (n & 3)); + shift[ch] -= (2 * n); + diff = sign_extend((diff &~ 3) << 8, 16); + /* saturate the shifter to a lower limit of 0 */ if (shift[ch] < 0) shift[ch] = 0; @@ -296,9 +299,10 @@ static int dpcm_decode_frame(AVCodecContext *avctx, void *data, } case CODEC_ID_SOL_DPCM: if (avctx->codec_tag != 3) { - uint8_t *output_samples_u8 = s->frame.data[0]; - while (buf < buf_end) { - uint8_t n = *buf++; + uint8_t *output_samples_u8 = s->frame.data[0], + *samples_end_u8 = output_samples_u8 + out; + while (output_samples_u8 < samples_end_u8) { + int n = bytestream2_get_byteu(&gb); s->sample[0] += s->sol_table[n >> 4]; s->sample[0] = av_clip_uint8(s->sample[0]); @@ -309,8 +313,8 @@ static int dpcm_decode_frame(AVCodecContext *avctx, void *data, *output_samples_u8++ = s->sample[stereo]; } } else { - while (buf < buf_end) { - uint8_t n = *buf++; + while (output_samples < samples_end) { + int n = bytestream2_get_byteu(&gb); if (n & 0x80) s->sample[ch] -= sol_table_16[n & 0x7F]; else s->sample[ch] += sol_table_16[n & 0x7F]; s->sample[ch] = av_clip_int16(s->sample[ch]); diff --git a/libavcodec/h264_ps.c b/libavcodec/h264_ps.c index 9c39af1f17..d9ed178664 100644 --- a/libavcodec/h264_ps.c +++ b/libavcodec/h264_ps.c @@ -352,9 +352,9 @@ int ff_h264_decode_seq_parameter_set(H264Context *h){ if (sps->chroma_format_idc > 3U) { av_log(h->s.avctx, AV_LOG_ERROR, "chroma_format_idc %d is illegal\n", sps->chroma_format_idc); goto fail; - } - if(sps->chroma_format_idc == 3) + } else if(sps->chroma_format_idc == 3) { sps->residual_color_transform_flag = get_bits1(&s->gb); + } sps->bit_depth_luma = get_ue_golomb(&s->gb) + 8; sps->bit_depth_chroma = get_ue_golomb(&s->gb) + 8; if (sps->bit_depth_luma > 12U || sps->bit_depth_chroma > 12U) { diff --git a/libavcodec/interplayvideo.c b/libavcodec/interplayvideo.c index ca8f041320..3ea7c50f87 100644 --- a/libavcodec/interplayvideo.c +++ b/libavcodec/interplayvideo.c @@ -56,14 +56,8 @@ typedef struct IpvideoContext { const unsigned char *decoding_map; int decoding_map_size; - const unsigned char *buf; - int size; - int is_16bpp; - const unsigned char *stream_ptr; - const unsigned char *stream_end; - const uint8_t *mv_ptr; - const uint8_t *mv_end; + GetByteContext stream_ptr, mv_ptr; unsigned char *pixel_ptr; int line_inc; int stride; @@ -72,13 +66,6 @@ typedef struct IpvideoContext { uint32_t pal[256]; } IpvideoContext; -#define CHECK_STREAM_PTR(stream_ptr, stream_end, n) \ - if (stream_end - stream_ptr < n) { \ - av_log(s->avctx, AV_LOG_ERROR, "stream_ptr out of bounds (%p >= %p)\n", \ - stream_ptr + n, stream_end); \ - return -1; \ - } - static int copy_from(IpvideoContext *s, AVFrame *src, int delta_x, int delta_y) { int current_offset = s->pixel_ptr - s->current_frame.data[0]; @@ -118,11 +105,9 @@ static int ipvideo_decode_block_opcode_0x2(IpvideoContext *s) /* copy block from 2 frames ago using a motion vector; need 1 more byte */ if (!s->is_16bpp) { - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 1); - B = *s->stream_ptr++; + B = bytestream2_get_byte(&s->stream_ptr); } else { - CHECK_STREAM_PTR(s->mv_ptr, s->mv_end, 1); - B = *s->mv_ptr++; + B = bytestream2_get_byte(&s->mv_ptr); } if (B < 56) { @@ -146,11 +131,9 @@ static int ipvideo_decode_block_opcode_0x3(IpvideoContext *s) /* need 1 more byte for motion */ if (!s->is_16bpp) { - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 1); - B = *s->stream_ptr++; + B = bytestream2_get_byte(&s->stream_ptr); } else { - CHECK_STREAM_PTR(s->mv_ptr, s->mv_end, 1); - B = *s->mv_ptr++; + B = bytestream2_get_byte(&s->mv_ptr); } if (B < 56) { @@ -172,11 +155,9 @@ static int ipvideo_decode_block_opcode_0x4(IpvideoContext *s) /* copy a block from the previous frame; need 1 more byte */ if (!s->is_16bpp) { - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 1); - B = *s->stream_ptr++; + B = bytestream2_get_byte(&s->stream_ptr); } else { - CHECK_STREAM_PTR(s->mv_ptr, s->mv_end, 1); - B = *s->mv_ptr++; + B = bytestream2_get_byte(&s->mv_ptr); } BL = B & 0x0F; @@ -194,10 +175,8 @@ static int ipvideo_decode_block_opcode_0x5(IpvideoContext *s) /* copy a block from the previous frame using an expanded range; * need 2 more bytes */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 2); - - x = *s->stream_ptr++; - y = *s->stream_ptr++; + x = bytestream2_get_byte(&s->stream_ptr); + y = bytestream2_get_byte(&s->stream_ptr); av_dlog(s->avctx, "motion bytes = %d, %d\n", x, y); return copy_from(s, &s->last_frame, x, y); @@ -219,18 +198,14 @@ static int ipvideo_decode_block_opcode_0x7(IpvideoContext *s) unsigned int flags; /* 2-color encoding */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 2); - - P[0] = *s->stream_ptr++; - P[1] = *s->stream_ptr++; + P[0] = bytestream2_get_byte(&s->stream_ptr); + P[1] = bytestream2_get_byte(&s->stream_ptr); if (P[0] <= P[1]) { /* need 8 more bytes from the stream */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 8); - for (y = 0; y < 8; y++) { - flags = *s->stream_ptr++ | 0x100; + flags = bytestream2_get_byte(&s->stream_ptr) | 0x100; for (; flags != 1; flags >>= 1) *s->pixel_ptr++ = P[flags & 1]; s->pixel_ptr += s->line_inc; @@ -239,9 +214,7 @@ static int ipvideo_decode_block_opcode_0x7(IpvideoContext *s) } else { /* need 2 more bytes from the stream */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 2); - - flags = bytestream_get_le16(&s->stream_ptr); + flags = bytestream2_get_le16(&s->stream_ptr); for (y = 0; y < 8; y += 2) { for (x = 0; x < 8; x += 2, flags >>= 1) { s->pixel_ptr[x ] = @@ -260,26 +233,23 @@ static int ipvideo_decode_block_opcode_0x7(IpvideoContext *s) static int ipvideo_decode_block_opcode_0x8(IpvideoContext *s) { int x, y; - unsigned char P[2]; + unsigned char P[4]; unsigned int flags = 0; /* 2-color encoding for each 4x4 quadrant, or 2-color encoding on * either top and bottom or left and right halves */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 2); - - P[0] = *s->stream_ptr++; - P[1] = *s->stream_ptr++; + P[0] = bytestream2_get_byte(&s->stream_ptr); + P[1] = bytestream2_get_byte(&s->stream_ptr); if (P[0] <= P[1]) { - - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 14); - s->stream_ptr -= 2; - for (y = 0; y < 16; y++) { // new values for each 4x4 block if (!(y & 3)) { - P[0] = *s->stream_ptr++; P[1] = *s->stream_ptr++; - flags = bytestream_get_le16(&s->stream_ptr); + if (y) { + P[0] = bytestream2_get_byte(&s->stream_ptr); + P[1] = bytestream2_get_byte(&s->stream_ptr); + } + flags = bytestream2_get_le16(&s->stream_ptr); } for (x = 0; x < 4; x++, flags >>= 1) @@ -290,13 +260,11 @@ static int ipvideo_decode_block_opcode_0x8(IpvideoContext *s) } } else { + flags = bytestream2_get_le32(&s->stream_ptr); + P[2] = bytestream2_get_byte(&s->stream_ptr); + P[3] = bytestream2_get_byte(&s->stream_ptr); - /* need 10 more bytes */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 10); - - if (s->stream_ptr[4] <= s->stream_ptr[5]) { - - flags = bytestream_get_le32(&s->stream_ptr); + if (P[2] <= P[3]) { /* vertical split; left & right halves are 2-color encoded */ @@ -307,8 +275,9 @@ static int ipvideo_decode_block_opcode_0x8(IpvideoContext *s) // switch to right half if (y == 7) { s->pixel_ptr -= 8 * s->stride - 4; - P[0] = *s->stream_ptr++; P[1] = *s->stream_ptr++; - flags = bytestream_get_le32(&s->stream_ptr); + P[0] = P[2]; + P[1] = P[3]; + flags = bytestream2_get_le32(&s->stream_ptr); } } @@ -318,12 +287,12 @@ static int ipvideo_decode_block_opcode_0x8(IpvideoContext *s) for (y = 0; y < 8; y++) { if (y == 4) { - P[0] = *s->stream_ptr++; - P[1] = *s->stream_ptr++; + P[0] = P[2]; + P[1] = P[3]; + flags = bytestream2_get_le32(&s->stream_ptr); } - flags = *s->stream_ptr++ | 0x100; - for (; flags != 1; flags >>= 1) + for (x = 0; x < 8; x++, flags >>= 1) *s->pixel_ptr++ = P[flags & 1]; s->pixel_ptr += s->line_inc; } @@ -340,20 +309,15 @@ static int ipvideo_decode_block_opcode_0x9(IpvideoContext *s) unsigned char P[4]; /* 4-color encoding */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 4); - - memcpy(P, s->stream_ptr, 4); - s->stream_ptr += 4; + bytestream2_get_buffer(&s->stream_ptr, P, 4); if (P[0] <= P[1]) { if (P[2] <= P[3]) { /* 1 of 4 colors for each pixel, need 16 more bytes */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 16); - for (y = 0; y < 8; y++) { /* get the next set of 8 2-bit flags */ - int flags = bytestream_get_le16(&s->stream_ptr); + int flags = bytestream2_get_le16(&s->stream_ptr); for (x = 0; x < 8; x++, flags >>= 2) *s->pixel_ptr++ = P[flags & 0x03]; s->pixel_ptr += s->line_inc; @@ -363,9 +327,7 @@ static int ipvideo_decode_block_opcode_0x9(IpvideoContext *s) uint32_t flags; /* 1 of 4 colors for each 2x2 block, need 4 more bytes */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 4); - - flags = bytestream_get_le32(&s->stream_ptr); + flags = bytestream2_get_le32(&s->stream_ptr); for (y = 0; y < 8; y += 2) { for (x = 0; x < 8; x += 2, flags >>= 2) { @@ -382,9 +344,7 @@ static int ipvideo_decode_block_opcode_0x9(IpvideoContext *s) uint64_t flags; /* 1 of 4 colors for each 2x1 or 1x2 block, need 8 more bytes */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 8); - - flags = bytestream_get_le64(&s->stream_ptr); + flags = bytestream2_get_le64(&s->stream_ptr); if (P[2] <= P[3]) { for (y = 0; y < 8; y++) { for (x = 0; x < 8; x += 2, flags >>= 2) { @@ -411,24 +371,21 @@ static int ipvideo_decode_block_opcode_0x9(IpvideoContext *s) static int ipvideo_decode_block_opcode_0xA(IpvideoContext *s) { int x, y; - unsigned char P[4]; + unsigned char P[8]; int flags = 0; + bytestream2_get_buffer(&s->stream_ptr, P, 4); + /* 4-color encoding for each 4x4 quadrant, or 4-color encoding on * either top and bottom or left and right halves */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 24); - - if (s->stream_ptr[0] <= s->stream_ptr[1]) { + if (P[0] <= P[1]) { /* 4-color encoding for each quadrant; need 32 bytes */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 32); - for (y = 0; y < 16; y++) { // new values for each 4x4 block if (!(y & 3)) { - memcpy(P, s->stream_ptr, 4); - s->stream_ptr += 4; - flags = bytestream_get_le32(&s->stream_ptr); + if (y) bytestream2_get_buffer(&s->stream_ptr, P, 4); + flags = bytestream2_get_le32(&s->stream_ptr); } for (x = 0; x < 4; x++, flags >>= 2) @@ -441,20 +398,16 @@ static int ipvideo_decode_block_opcode_0xA(IpvideoContext *s) } else { // vertical split? - int vert = s->stream_ptr[12] <= s->stream_ptr[13]; - uint64_t flags = 0; + int vert; + uint64_t flags = bytestream2_get_le64(&s->stream_ptr); + + bytestream2_get_buffer(&s->stream_ptr, P + 4, 4); + vert = P[4] <= P[5]; /* 4-color encoding for either left and right or top and bottom * halves */ for (y = 0; y < 16; y++) { - // load values for each half - if (!(y & 7)) { - memcpy(P, s->stream_ptr, 4); - s->stream_ptr += 4; - flags = bytestream_get_le64(&s->stream_ptr); - } - for (x = 0; x < 4; x++, flags >>= 2) *s->pixel_ptr++ = P[flags & 0x03]; @@ -463,6 +416,12 @@ static int ipvideo_decode_block_opcode_0xA(IpvideoContext *s) // switch to right half if (y == 7) s->pixel_ptr -= 8 * s->stride - 4; } else if (y & 1) s->pixel_ptr += s->line_inc; + + // load values for second half + if (y == 7) { + memcpy(P, P + 4, 4); + flags = bytestream2_get_le64(&s->stream_ptr); + } } } @@ -475,11 +434,8 @@ static int ipvideo_decode_block_opcode_0xB(IpvideoContext *s) int y; /* 64-color encoding (each pixel in block is a different color) */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 64); - for (y = 0; y < 8; y++) { - memcpy(s->pixel_ptr, s->stream_ptr, 8); - s->stream_ptr += 8; + bytestream2_get_buffer(&s->stream_ptr, s->pixel_ptr, 8); s->pixel_ptr += s->stride; } @@ -492,14 +448,12 @@ static int ipvideo_decode_block_opcode_0xC(IpvideoContext *s) int x, y; /* 16-color block encoding: each 2x2 block is a different color */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 16); - for (y = 0; y < 8; y += 2) { for (x = 0; x < 8; x += 2) { s->pixel_ptr[x ] = s->pixel_ptr[x + 1 ] = s->pixel_ptr[x + s->stride] = - s->pixel_ptr[x + 1 + s->stride] = *s->stream_ptr++; + s->pixel_ptr[x + 1 + s->stride] = bytestream2_get_byte(&s->stream_ptr); } s->pixel_ptr += s->stride * 2; } @@ -514,12 +468,10 @@ static int ipvideo_decode_block_opcode_0xD(IpvideoContext *s) unsigned char P[2]; /* 4-color block encoding: each 4x4 block is a different color */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 4); - for (y = 0; y < 8; y++) { if (!(y & 3)) { - P[0] = *s->stream_ptr++; - P[1] = *s->stream_ptr++; + P[0] = bytestream2_get_byte(&s->stream_ptr); + P[1] = bytestream2_get_byte(&s->stream_ptr); } memset(s->pixel_ptr, P[0], 4); memset(s->pixel_ptr + 4, P[1], 4); @@ -536,8 +488,7 @@ static int ipvideo_decode_block_opcode_0xE(IpvideoContext *s) unsigned char pix; /* 1-color encoding: the whole block is 1 solid color */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 1); - pix = *s->stream_ptr++; + pix = bytestream2_get_byte(&s->stream_ptr); for (y = 0; y < 8; y++) { memset(s->pixel_ptr, pix, 8); @@ -554,9 +505,8 @@ static int ipvideo_decode_block_opcode_0xF(IpvideoContext *s) unsigned char sample[2]; /* dithered encoding */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 2); - sample[0] = *s->stream_ptr++; - sample[1] = *s->stream_ptr++; + sample[0] = bytestream2_get_byte(&s->stream_ptr); + sample[1] = bytestream2_get_byte(&s->stream_ptr); for (y = 0; y < 8; y++) { for (x = 0; x < 8; x += 2) { @@ -575,10 +525,8 @@ static int ipvideo_decode_block_opcode_0x6_16(IpvideoContext *s) signed char x, y; /* copy a block from the second last frame using an expanded range */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 2); - - x = *s->stream_ptr++; - y = *s->stream_ptr++; + x = bytestream2_get_byte(&s->stream_ptr); + y = bytestream2_get_byte(&s->stream_ptr); av_dlog(s->avctx, "motion bytes = %d, %d\n", x, y); return copy_from(s, &s->second_last_frame, x, y); @@ -592,17 +540,13 @@ static int ipvideo_decode_block_opcode_0x7_16(IpvideoContext *s) uint16_t *pixel_ptr = (uint16_t*)s->pixel_ptr; /* 2-color encoding */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 4); - - P[0] = bytestream_get_le16(&s->stream_ptr); - P[1] = bytestream_get_le16(&s->stream_ptr); + P[0] = bytestream2_get_le16(&s->stream_ptr); + P[1] = bytestream2_get_le16(&s->stream_ptr); if (!(P[0] & 0x8000)) { - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 8); - for (y = 0; y < 8; y++) { - flags = *s->stream_ptr++ | 0x100; + flags = bytestream2_get_byte(&s->stream_ptr) | 0x100; for (; flags != 1; flags >>= 1) *pixel_ptr++ = P[flags & 1]; pixel_ptr += s->line_inc; @@ -610,9 +554,7 @@ static int ipvideo_decode_block_opcode_0x7_16(IpvideoContext *s) } else { - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 2); - - flags = bytestream_get_le16(&s->stream_ptr); + flags = bytestream2_get_le16(&s->stream_ptr); for (y = 0; y < 8; y += 2) { for (x = 0; x < 8; x += 2, flags >>= 1) { pixel_ptr[x ] = @@ -630,28 +572,25 @@ static int ipvideo_decode_block_opcode_0x7_16(IpvideoContext *s) static int ipvideo_decode_block_opcode_0x8_16(IpvideoContext *s) { int x, y; - uint16_t P[2]; + uint16_t P[4]; unsigned int flags = 0; uint16_t *pixel_ptr = (uint16_t*)s->pixel_ptr; /* 2-color encoding for each 4x4 quadrant, or 2-color encoding on * either top and bottom or left and right halves */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 4); - - P[0] = bytestream_get_le16(&s->stream_ptr); - P[1] = bytestream_get_le16(&s->stream_ptr); + P[0] = bytestream2_get_le16(&s->stream_ptr); + P[1] = bytestream2_get_le16(&s->stream_ptr); if (!(P[0] & 0x8000)) { - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 24); - s->stream_ptr -= 4; - for (y = 0; y < 16; y++) { // new values for each 4x4 block if (!(y & 3)) { - P[0] = bytestream_get_le16(&s->stream_ptr); - P[1] = bytestream_get_le16(&s->stream_ptr); - flags = bytestream_get_le16(&s->stream_ptr); + if (y) { + P[0] = bytestream2_get_le16(&s->stream_ptr); + P[1] = bytestream2_get_le16(&s->stream_ptr); + } + flags = bytestream2_get_le16(&s->stream_ptr); } for (x = 0; x < 4; x++, flags >>= 1) @@ -663,11 +602,11 @@ static int ipvideo_decode_block_opcode_0x8_16(IpvideoContext *s) } else { - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 12); + flags = bytestream2_get_le32(&s->stream_ptr); + P[2] = bytestream2_get_le16(&s->stream_ptr); + P[3] = bytestream2_get_le16(&s->stream_ptr); - if (!(AV_RL16(s->stream_ptr + 4) & 0x8000)) { - - flags = bytestream_get_le32(&s->stream_ptr); + if (!(P[2] & 0x8000)) { /* vertical split; left & right halves are 2-color encoded */ @@ -678,9 +617,9 @@ static int ipvideo_decode_block_opcode_0x8_16(IpvideoContext *s) // switch to right half if (y == 7) { pixel_ptr -= 8 * s->stride - 4; - P[0] = bytestream_get_le16(&s->stream_ptr); - P[1] = bytestream_get_le16(&s->stream_ptr); - flags = bytestream_get_le32(&s->stream_ptr); + P[0] = P[2]; + P[1] = P[3]; + flags = bytestream2_get_le32(&s->stream_ptr); } } @@ -690,12 +629,12 @@ static int ipvideo_decode_block_opcode_0x8_16(IpvideoContext *s) for (y = 0; y < 8; y++) { if (y == 4) { - P[0] = bytestream_get_le16(&s->stream_ptr); - P[1] = bytestream_get_le16(&s->stream_ptr); + P[0] = P[2]; + P[1] = P[3]; + flags = bytestream2_get_le32(&s->stream_ptr); } - flags = *s->stream_ptr++ | 0x100; - for (; flags != 1; flags >>= 1) + for (x = 0; x < 8; x++, flags >>= 1) *pixel_ptr++ = P[flags & 1]; pixel_ptr += s->line_inc; } @@ -713,20 +652,16 @@ static int ipvideo_decode_block_opcode_0x9_16(IpvideoContext *s) uint16_t *pixel_ptr = (uint16_t*)s->pixel_ptr; /* 4-color encoding */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 8); - for (x = 0; x < 4; x++) - P[x] = bytestream_get_le16(&s->stream_ptr); + P[x] = bytestream2_get_le16(&s->stream_ptr); if (!(P[0] & 0x8000)) { if (!(P[2] & 0x8000)) { /* 1 of 4 colors for each pixel */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 16); - for (y = 0; y < 8; y++) { /* get the next set of 8 2-bit flags */ - int flags = bytestream_get_le16(&s->stream_ptr); + int flags = bytestream2_get_le16(&s->stream_ptr); for (x = 0; x < 8; x++, flags >>= 2) *pixel_ptr++ = P[flags & 0x03]; pixel_ptr += s->line_inc; @@ -736,9 +671,7 @@ static int ipvideo_decode_block_opcode_0x9_16(IpvideoContext *s) uint32_t flags; /* 1 of 4 colors for each 2x2 block */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 4); - - flags = bytestream_get_le32(&s->stream_ptr); + flags = bytestream2_get_le32(&s->stream_ptr); for (y = 0; y < 8; y += 2) { for (x = 0; x < 8; x += 2, flags >>= 2) { @@ -755,9 +688,7 @@ static int ipvideo_decode_block_opcode_0x9_16(IpvideoContext *s) uint64_t flags; /* 1 of 4 colors for each 2x1 or 1x2 block */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 8); - - flags = bytestream_get_le64(&s->stream_ptr); + flags = bytestream2_get_le64(&s->stream_ptr); if (!(P[2] & 0x8000)) { for (y = 0; y < 8; y++) { for (x = 0; x < 8; x += 2, flags >>= 2) { @@ -784,25 +715,25 @@ static int ipvideo_decode_block_opcode_0x9_16(IpvideoContext *s) static int ipvideo_decode_block_opcode_0xA_16(IpvideoContext *s) { int x, y; - uint16_t P[4]; + uint16_t P[8]; int flags = 0; uint16_t *pixel_ptr = (uint16_t*)s->pixel_ptr; + for (x = 0; x < 4; x++) + P[x] = bytestream2_get_le16(&s->stream_ptr); + /* 4-color encoding for each 4x4 quadrant, or 4-color encoding on * either top and bottom or left and right halves */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 24); - - if (!(AV_RL16(s->stream_ptr) & 0x8000)) { + if (!(P[0] & 0x8000)) { /* 4-color encoding for each quadrant */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 48); - for (y = 0; y < 16; y++) { // new values for each 4x4 block if (!(y & 3)) { - for (x = 0; x < 4; x++) - P[x] = bytestream_get_le16(&s->stream_ptr); - flags = bytestream_get_le32(&s->stream_ptr); + if (y) + for (x = 0; x < 4; x++) + P[x] = bytestream2_get_le16(&s->stream_ptr); + flags = bytestream2_get_le32(&s->stream_ptr); } for (x = 0; x < 4; x++, flags >>= 2) @@ -815,20 +746,17 @@ static int ipvideo_decode_block_opcode_0xA_16(IpvideoContext *s) } else { // vertical split? - int vert = !(AV_RL16(s->stream_ptr + 16) & 0x8000); - uint64_t flags = 0; + int vert; + uint64_t flags = bytestream2_get_le64(&s->stream_ptr); + + for (x = 4; x < 8; x++) + P[x] = bytestream2_get_le16(&s->stream_ptr); + vert = !(P[4] & 0x8000); /* 4-color encoding for either left and right or top and bottom * halves */ for (y = 0; y < 16; y++) { - // load values for each half - if (!(y & 7)) { - for (x = 0; x < 4; x++) - P[x] = bytestream_get_le16(&s->stream_ptr); - flags = bytestream_get_le64(&s->stream_ptr); - } - for (x = 0; x < 4; x++, flags >>= 2) *pixel_ptr++ = P[flags & 0x03]; @@ -837,6 +765,12 @@ static int ipvideo_decode_block_opcode_0xA_16(IpvideoContext *s) // switch to right half if (y == 7) pixel_ptr -= 8 * s->stride - 4; } else if (y & 1) pixel_ptr += s->line_inc; + + // load values for second half + if (y == 7) { + memcpy(P, P + 4, 8); + flags = bytestream2_get_le64(&s->stream_ptr); + } } } @@ -850,11 +784,9 @@ static int ipvideo_decode_block_opcode_0xB_16(IpvideoContext *s) uint16_t *pixel_ptr = (uint16_t*)s->pixel_ptr; /* 64-color encoding (each pixel in block is a different color) */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 128); - for (y = 0; y < 8; y++) { for (x = 0; x < 8; x++) - pixel_ptr[x] = bytestream_get_le16(&s->stream_ptr); + pixel_ptr[x] = bytestream2_get_le16(&s->stream_ptr); pixel_ptr += s->stride; } @@ -868,14 +800,12 @@ static int ipvideo_decode_block_opcode_0xC_16(IpvideoContext *s) uint16_t *pixel_ptr = (uint16_t*)s->pixel_ptr; /* 16-color block encoding: each 2x2 block is a different color */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 32); - for (y = 0; y < 8; y += 2) { for (x = 0; x < 8; x += 2) { pixel_ptr[x ] = pixel_ptr[x + 1 ] = pixel_ptr[x + s->stride] = - pixel_ptr[x + 1 + s->stride] = bytestream_get_le16(&s->stream_ptr); + pixel_ptr[x + 1 + s->stride] = bytestream2_get_le16(&s->stream_ptr); } pixel_ptr += s->stride * 2; } @@ -891,12 +821,10 @@ static int ipvideo_decode_block_opcode_0xD_16(IpvideoContext *s) uint16_t *pixel_ptr = (uint16_t*)s->pixel_ptr; /* 4-color block encoding: each 4x4 block is a different color */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 8); - for (y = 0; y < 8; y++) { if (!(y & 3)) { - P[0] = bytestream_get_le16(&s->stream_ptr); - P[1] = bytestream_get_le16(&s->stream_ptr); + P[0] = bytestream2_get_le16(&s->stream_ptr); + P[1] = bytestream2_get_le16(&s->stream_ptr); } for (x = 0; x < 8; x++) pixel_ptr[x] = P[x >> 2]; @@ -914,8 +842,7 @@ static int ipvideo_decode_block_opcode_0xE_16(IpvideoContext *s) uint16_t *pixel_ptr = (uint16_t*)s->pixel_ptr; /* 1-color encoding: the whole block is 1 solid color */ - CHECK_STREAM_PTR(s->stream_ptr, s->stream_end, 2); - pix = bytestream_get_le16(&s->stream_ptr); + pix = bytestream2_get_le16(&s->stream_ptr); for (y = 0; y < 8; y++) { for (x = 0; x < 8; x++) @@ -960,19 +887,16 @@ static void ipvideo_decode_opcodes(IpvideoContext *s) av_dlog(s->avctx, "frame %d\n", frame); frame++; + bytestream2_skip(&s->stream_ptr, 14); /* data starts 14 bytes in */ if (!s->is_16bpp) { /* this is PAL8, so make the palette available */ memcpy(s->current_frame.data[1], s->pal, AVPALETTE_SIZE); s->stride = s->current_frame.linesize[0]; - s->stream_ptr = s->buf + 14; /* data starts 14 bytes in */ - s->stream_end = s->buf + s->size; } else { s->stride = s->current_frame.linesize[0] >> 1; - s->stream_ptr = s->buf + 16; - s->stream_end = - s->mv_ptr = s->buf + 14 + AV_RL16(s->buf+14); - s->mv_end = s->buf + s->size; + s->mv_ptr = s->stream_ptr; + bytestream2_skip(&s->mv_ptr, bytestream2_get_le16(&s->stream_ptr)); } s->line_inc = s->stride - 8; s->upper_motion_limit_offset = (s->avctx->height - 8) * s->current_frame.linesize[0] @@ -1002,9 +926,10 @@ static void ipvideo_decode_opcodes(IpvideoContext *s) } } } - if (s->stream_end - s->stream_ptr > 1) { - av_log(s->avctx, AV_LOG_ERROR, "decode finished with %td bytes left over\n", - s->stream_end - s->stream_ptr); + if (bytestream2_get_bytes_left(&s->stream_ptr) > 1) { + av_log(s->avctx, AV_LOG_ERROR, + "decode finished with %d bytes left over\n", + bytestream2_get_bytes_left(&s->stream_ptr)); } } @@ -1046,8 +971,8 @@ static int ipvideo_decode_frame(AVCodecContext *avctx, return buf_size; s->decoding_map = buf; - s->buf = buf + s->decoding_map_size; - s->size = buf_size - s->decoding_map_size; + bytestream2_init(&s->stream_ptr, buf + s->decoding_map_size, + buf_size - s->decoding_map_size); s->current_frame.reference = 3; if (avctx->get_buffer(avctx, &s->current_frame)) { diff --git a/libavcodec/pthread.c b/libavcodec/pthread.c index e253eb9c36..ccf84efdbe 100644 --- a/libavcodec/pthread.c +++ b/libavcodec/pthread.c @@ -907,9 +907,13 @@ void ff_thread_flush(AVCodecContext *avctx) fctx->next_decoding = fctx->next_finished = 0; fctx->delaying = 1; fctx->prev_thread = NULL; - // Make sure decode flush calls with size=0 won't return old frames - for (i = 0; i < avctx->thread_count; i++) - fctx->threads[i].got_frame = 0; + for (i = 0; i < avctx->thread_count; i++) { + PerThreadContext *p = &fctx->threads[i]; + // Make sure decode flush calls with size=0 won't return old frames + p->got_frame = 0; + + release_delayed_buffers(p); + } } static int *allocate_progress(PerThreadContext *p) diff --git a/libavcodec/utvideo.c b/libavcodec/utvideo.c index cdc688da22..f0161ac9fc 100644 --- a/libavcodec/utvideo.c +++ b/libavcodec/utvideo.c @@ -358,13 +358,12 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, AVPac { const uint8_t *buf = avpkt->data; int buf_size = avpkt->size; - const uint8_t *buf_end = buf + buf_size; UtvideoContext *c = avctx->priv_data; - const uint8_t *ptr; int i, j; const uint8_t *plane_start[5]; int plane_size, max_slice_size = 0, slice_start, slice_end, slice_size; int ret; + GetByteContext gb; if (c->pic.data[0]) ff_thread_release_buffer(avctx, &c->pic); @@ -377,20 +376,21 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, AVPac } /* parse plane structure to retrieve frame flags and validate slice offsets */ - ptr = buf; + bytestream2_init(&gb, buf, buf_size); for (i = 0; i < c->planes; i++) { - plane_start[i] = ptr; - if (buf_end - ptr < 256 + 4 * c->slices) { + plane_start[i] = gb.buffer; + if (bytestream2_get_bytes_left(&gb) < 256 + 4 * c->slices) { av_log(avctx, AV_LOG_ERROR, "Insufficient data for a plane\n"); return AVERROR_INVALIDDATA; } - ptr += 256; + bytestream2_skipu(&gb, 256); slice_start = 0; slice_end = 0; for (j = 0; j < c->slices; j++) { - slice_end = bytestream_get_le32(&ptr); + slice_end = bytestream2_get_le32u(&gb); slice_size = slice_end - slice_start; - if (slice_size < 0) { + if (slice_end <= 0 || slice_size <= 0 || + bytestream2_get_bytes_left(&gb) < slice_end) { av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n"); return AVERROR_INVALIDDATA; } @@ -398,18 +398,14 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, AVPac max_slice_size = FFMAX(max_slice_size, slice_size); } plane_size = slice_end; - if (buf_end - ptr < plane_size) { - av_log(avctx, AV_LOG_ERROR, "Plane size is bigger than available data\n"); - return AVERROR_INVALIDDATA; - } - ptr += plane_size; + bytestream2_skipu(&gb, plane_size); } - plane_start[c->planes] = ptr; - if (buf_end - ptr < c->frame_info_size) { + plane_start[c->planes] = gb.buffer; + if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) { av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n"); return AVERROR_INVALIDDATA; } - c->frame_info = AV_RL32(ptr); + c->frame_info = bytestream2_get_le32u(&gb); av_log(avctx, AV_LOG_DEBUG, "frame information flags %X\n", c->frame_info); c->frame_pred = (c->frame_info >> 8) & 3; diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile index 0f746e289e..267cf241b7 100644 --- a/libavcodec/x86/Makefile +++ b/libavcodec/x86/Makefile @@ -25,6 +25,7 @@ YASM-OBJS-$(CONFIG_H264DSP) += x86/h264_deblock.o \ YASM-OBJS-$(CONFIG_H264PRED) += x86/h264_intrapred.o \ x86/h264_intrapred_10bit.o MMX-OBJS-$(CONFIG_H264PRED) += x86/h264_intrapred_init.o +YASM-OBJS-$(CONFIG_H264QPEL) += x86/h264_qpel_10bit.o MMX-OBJS-$(CONFIG_RV30_DECODER) += x86/rv34dsp_init.o YASM-OBJS-$(CONFIG_RV30_DECODER) += x86/rv34dsp.o @@ -71,7 +72,6 @@ MMX-OBJS-$(CONFIG_VP8_DECODER) += x86/vp8dsp-init.o MMX-OBJS-$(HAVE_YASM) += x86/dsputil_yasm.o \ x86/deinterlace.o \ x86/fmtconvert.o \ - x86/h264_qpel_10bit.o \ $(YASM-OBJS-yes) MMX-OBJS-$(CONFIG_FFT) += x86/fft.o diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c index 976008af9b..c887d6d79c 100644 --- a/libavcodec/x86/dsputil_mmx.c +++ b/libavcodec/x86/dsputil_mmx.c @@ -41,129 +41,129 @@ DECLARE_ALIGNED(8, const uint64_t, ff_bone) = 0x0101010101010101ULL; DECLARE_ALIGNED(8, const uint64_t, ff_wtwo) = 0x0002000200020002ULL; DECLARE_ALIGNED(16, const uint64_t, ff_pdw_80000000)[2] = -{0x8000000080000000ULL, 0x8000000080000000ULL}; + { 0x8000000080000000ULL, 0x8000000080000000ULL }; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_1 ) = {0x0001000100010001ULL, 0x0001000100010001ULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_2 ) = {0x0002000200020002ULL, 0x0002000200020002ULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_3 ) = {0x0003000300030003ULL, 0x0003000300030003ULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_4 ) = {0x0004000400040004ULL, 0x0004000400040004ULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_5 ) = {0x0005000500050005ULL, 0x0005000500050005ULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_8 ) = {0x0008000800080008ULL, 0x0008000800080008ULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_9 ) = {0x0009000900090009ULL, 0x0009000900090009ULL}; -DECLARE_ALIGNED(8, const uint64_t, ff_pw_15 ) = 0x000F000F000F000FULL; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_16 ) = {0x0010001000100010ULL, 0x0010001000100010ULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_17 ) = {0x0011001100110011ULL, 0x0011001100110011ULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_18 ) = {0x0012001200120012ULL, 0x0012001200120012ULL}; -DECLARE_ALIGNED(8, const uint64_t, ff_pw_20 ) = 0x0014001400140014ULL; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_27 ) = {0x001B001B001B001BULL, 0x001B001B001B001BULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_28 ) = {0x001C001C001C001CULL, 0x001C001C001C001CULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_32 ) = {0x0020002000200020ULL, 0x0020002000200020ULL}; -DECLARE_ALIGNED(8, const uint64_t, ff_pw_42 ) = 0x002A002A002A002AULL; -DECLARE_ALIGNED(8, const uint64_t, ff_pw_53 ) = 0x0035003500350035ULL; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_63 ) = {0x003F003F003F003FULL, 0x003F003F003F003FULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_64 ) = {0x0040004000400040ULL, 0x0040004000400040ULL}; -DECLARE_ALIGNED(8, const uint64_t, ff_pw_96 ) = 0x0060006000600060ULL; -DECLARE_ALIGNED(8, const uint64_t, ff_pw_128) = 0x0080008000800080ULL; -DECLARE_ALIGNED(8, const uint64_t, ff_pw_255) = 0x00ff00ff00ff00ffULL; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_512) = {0x0200020002000200ULL, 0x0200020002000200ULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pw_1019)= {0x03FB03FB03FB03FBULL, 0x03FB03FB03FB03FBULL}; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_1) = { 0x0001000100010001ULL, 0x0001000100010001ULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_2) = { 0x0002000200020002ULL, 0x0002000200020002ULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_3) = { 0x0003000300030003ULL, 0x0003000300030003ULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_4) = { 0x0004000400040004ULL, 0x0004000400040004ULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_5) = { 0x0005000500050005ULL, 0x0005000500050005ULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_8) = { 0x0008000800080008ULL, 0x0008000800080008ULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_9) = { 0x0009000900090009ULL, 0x0009000900090009ULL }; +DECLARE_ALIGNED(8, const uint64_t, ff_pw_15) = 0x000F000F000F000FULL; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_16) = { 0x0010001000100010ULL, 0x0010001000100010ULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_17) = { 0x0011001100110011ULL, 0x0011001100110011ULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_18) = { 0x0012001200120012ULL, 0x0012001200120012ULL }; +DECLARE_ALIGNED(8, const uint64_t, ff_pw_20) = 0x0014001400140014ULL; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_27) = { 0x001B001B001B001BULL, 0x001B001B001B001BULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_28) = { 0x001C001C001C001CULL, 0x001C001C001C001CULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_32) = { 0x0020002000200020ULL, 0x0020002000200020ULL }; +DECLARE_ALIGNED(8, const uint64_t, ff_pw_42) = 0x002A002A002A002AULL; +DECLARE_ALIGNED(8, const uint64_t, ff_pw_53) = 0x0035003500350035ULL; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_63) = { 0x003F003F003F003FULL, 0x003F003F003F003FULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_64) = { 0x0040004000400040ULL, 0x0040004000400040ULL }; +DECLARE_ALIGNED(8, const uint64_t, ff_pw_96) = 0x0060006000600060ULL; +DECLARE_ALIGNED(8, const uint64_t, ff_pw_128) = 0x0080008000800080ULL; +DECLARE_ALIGNED(8, const uint64_t, ff_pw_255) = 0x00ff00ff00ff00ffULL; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_512) = { 0x0200020002000200ULL, 0x0200020002000200ULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pw_1019) = { 0x03FB03FB03FB03FBULL, 0x03FB03FB03FB03FBULL }; -DECLARE_ALIGNED(16, const xmm_reg, ff_pb_0 ) = {0x0000000000000000ULL, 0x0000000000000000ULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pb_1 ) = {0x0101010101010101ULL, 0x0101010101010101ULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pb_3 ) = {0x0303030303030303ULL, 0x0303030303030303ULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pb_4 ) = {0x0404040404040404ULL, 0x0404040404040404ULL}; -DECLARE_ALIGNED(8, const uint64_t, ff_pb_7 ) = 0x0707070707070707ULL; -DECLARE_ALIGNED(8, const uint64_t, ff_pb_1F ) = 0x1F1F1F1F1F1F1F1FULL; -DECLARE_ALIGNED(8, const uint64_t, ff_pb_3F ) = 0x3F3F3F3F3F3F3F3FULL; -DECLARE_ALIGNED(16, const xmm_reg, ff_pb_80 ) = {0x8080808080808080ULL, 0x8080808080808080ULL}; -DECLARE_ALIGNED(8, const uint64_t, ff_pb_81 ) = 0x8181818181818181ULL; -DECLARE_ALIGNED(16, const xmm_reg, ff_pb_A1 ) = {0xA1A1A1A1A1A1A1A1ULL, 0xA1A1A1A1A1A1A1A1ULL}; -DECLARE_ALIGNED(16, const xmm_reg, ff_pb_F8 ) = {0xF8F8F8F8F8F8F8F8ULL, 0xF8F8F8F8F8F8F8F8ULL}; -DECLARE_ALIGNED(8, const uint64_t, ff_pb_FC ) = 0xFCFCFCFCFCFCFCFCULL; -DECLARE_ALIGNED(16, const xmm_reg, ff_pb_FE ) = {0xFEFEFEFEFEFEFEFEULL, 0xFEFEFEFEFEFEFEFEULL}; +DECLARE_ALIGNED(16, const xmm_reg, ff_pb_0) = { 0x0000000000000000ULL, 0x0000000000000000ULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pb_1) = { 0x0101010101010101ULL, 0x0101010101010101ULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pb_3) = { 0x0303030303030303ULL, 0x0303030303030303ULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pb_4) = { 0x0404040404040404ULL, 0x0404040404040404ULL }; +DECLARE_ALIGNED(8, const uint64_t, ff_pb_7) = 0x0707070707070707ULL; +DECLARE_ALIGNED(8, const uint64_t, ff_pb_1F) = 0x1F1F1F1F1F1F1F1FULL; +DECLARE_ALIGNED(8, const uint64_t, ff_pb_3F) = 0x3F3F3F3F3F3F3F3FULL; +DECLARE_ALIGNED(16, const xmm_reg, ff_pb_80) = { 0x8080808080808080ULL, 0x8080808080808080ULL }; +DECLARE_ALIGNED(8, const uint64_t, ff_pb_81) = 0x8181818181818181ULL; +DECLARE_ALIGNED(16, const xmm_reg, ff_pb_A1) = { 0xA1A1A1A1A1A1A1A1ULL, 0xA1A1A1A1A1A1A1A1ULL }; +DECLARE_ALIGNED(16, const xmm_reg, ff_pb_F8) = { 0xF8F8F8F8F8F8F8F8ULL, 0xF8F8F8F8F8F8F8F8ULL }; +DECLARE_ALIGNED(8, const uint64_t, ff_pb_FC) = 0xFCFCFCFCFCFCFCFCULL; +DECLARE_ALIGNED(16, const xmm_reg, ff_pb_FE) = { 0xFEFEFEFEFEFEFEFEULL, 0xFEFEFEFEFEFEFEFEULL }; DECLARE_ALIGNED(16, const double, ff_pd_1)[2] = { 1.0, 1.0 }; DECLARE_ALIGNED(16, const double, ff_pd_2)[2] = { 2.0, 2.0 }; -#define JUMPALIGN() __asm__ volatile (".p2align 3"::) -#define MOVQ_ZERO(regd) __asm__ volatile ("pxor %%" #regd ", %%" #regd ::) +#define JUMPALIGN() __asm__ volatile (".p2align 3"::) +#define MOVQ_ZERO(regd) __asm__ volatile ("pxor %%"#regd", %%"#regd ::) -#define MOVQ_BFE(regd) \ - __asm__ volatile ( \ - "pcmpeqd %%" #regd ", %%" #regd " \n\t"\ - "paddb %%" #regd ", %%" #regd " \n\t" ::) +#define MOVQ_BFE(regd) \ + __asm__ volatile ( \ + "pcmpeqd %%"#regd", %%"#regd" \n\t" \ + "paddb %%"#regd", %%"#regd" \n\t" ::) #ifndef PIC -#define MOVQ_BONE(regd) __asm__ volatile ("movq %0, %%" #regd " \n\t" ::"m"(ff_bone)) -#define MOVQ_WTWO(regd) __asm__ volatile ("movq %0, %%" #regd " \n\t" ::"m"(ff_wtwo)) +#define MOVQ_BONE(regd) __asm__ volatile ("movq %0, %%"#regd" \n\t" :: "m"(ff_bone)) +#define MOVQ_WTWO(regd) __asm__ volatile ("movq %0, %%"#regd" \n\t" :: "m"(ff_wtwo)) #else // for shared library it's better to use this way for accessing constants // pcmpeqd -> -1 -#define MOVQ_BONE(regd) \ - __asm__ volatile ( \ - "pcmpeqd %%" #regd ", %%" #regd " \n\t" \ - "psrlw $15, %%" #regd " \n\t" \ - "packuswb %%" #regd ", %%" #regd " \n\t" ::) +#define MOVQ_BONE(regd) \ + __asm__ volatile ( \ + "pcmpeqd %%"#regd", %%"#regd" \n\t" \ + "psrlw $15, %%"#regd" \n\t" \ + "packuswb %%"#regd", %%"#regd" \n\t" ::) -#define MOVQ_WTWO(regd) \ - __asm__ volatile ( \ - "pcmpeqd %%" #regd ", %%" #regd " \n\t" \ - "psrlw $15, %%" #regd " \n\t" \ - "psllw $1, %%" #regd " \n\t"::) +#define MOVQ_WTWO(regd) \ + __asm__ volatile ( \ + "pcmpeqd %%"#regd", %%"#regd" \n\t" \ + "psrlw $15, %%"#regd" \n\t" \ + "psllw $1, %%"#regd" \n\t"::) #endif // using regr as temporary and for the output result // first argument is unmodifed and second is trashed // regfe is supposed to contain 0xfefefefefefefefe -#define PAVGB_MMX_NO_RND(rega, regb, regr, regfe) \ - "movq " #rega ", " #regr " \n\t"\ - "pand " #regb ", " #regr " \n\t"\ - "pxor " #rega ", " #regb " \n\t"\ - "pand " #regfe "," #regb " \n\t"\ - "psrlq $1, " #regb " \n\t"\ - "paddb " #regb ", " #regr " \n\t" +#define PAVGB_MMX_NO_RND(rega, regb, regr, regfe) \ + "movq "#rega", "#regr" \n\t" \ + "pand "#regb", "#regr" \n\t" \ + "pxor "#rega", "#regb" \n\t" \ + "pand "#regfe", "#regb" \n\t" \ + "psrlq $1, "#regb" \n\t" \ + "paddb "#regb", "#regr" \n\t" -#define PAVGB_MMX(rega, regb, regr, regfe) \ - "movq " #rega ", " #regr " \n\t"\ - "por " #regb ", " #regr " \n\t"\ - "pxor " #rega ", " #regb " \n\t"\ - "pand " #regfe "," #regb " \n\t"\ - "psrlq $1, " #regb " \n\t"\ - "psubb " #regb ", " #regr " \n\t" +#define PAVGB_MMX(rega, regb, regr, regfe) \ + "movq "#rega", "#regr" \n\t" \ + "por "#regb", "#regr" \n\t" \ + "pxor "#rega", "#regb" \n\t" \ + "pand "#regfe", "#regb" \n\t" \ + "psrlq $1, "#regb" \n\t" \ + "psubb "#regb", "#regr" \n\t" // mm6 is supposed to contain 0xfefefefefefefefe -#define PAVGBP_MMX_NO_RND(rega, regb, regr, regc, regd, regp) \ - "movq " #rega ", " #regr " \n\t"\ - "movq " #regc ", " #regp " \n\t"\ - "pand " #regb ", " #regr " \n\t"\ - "pand " #regd ", " #regp " \n\t"\ - "pxor " #rega ", " #regb " \n\t"\ - "pxor " #regc ", " #regd " \n\t"\ - "pand %%mm6, " #regb " \n\t"\ - "pand %%mm6, " #regd " \n\t"\ - "psrlq $1, " #regb " \n\t"\ - "psrlq $1, " #regd " \n\t"\ - "paddb " #regb ", " #regr " \n\t"\ - "paddb " #regd ", " #regp " \n\t" +#define PAVGBP_MMX_NO_RND(rega, regb, regr, regc, regd, regp) \ + "movq "#rega", "#regr" \n\t" \ + "movq "#regc", "#regp" \n\t" \ + "pand "#regb", "#regr" \n\t" \ + "pand "#regd", "#regp" \n\t" \ + "pxor "#rega", "#regb" \n\t" \ + "pxor "#regc", "#regd" \n\t" \ + "pand %%mm6, "#regb" \n\t" \ + "pand %%mm6, "#regd" \n\t" \ + "psrlq $1, "#regb" \n\t" \ + "psrlq $1, "#regd" \n\t" \ + "paddb "#regb", "#regr" \n\t" \ + "paddb "#regd", "#regp" \n\t" -#define PAVGBP_MMX(rega, regb, regr, regc, regd, regp) \ - "movq " #rega ", " #regr " \n\t"\ - "movq " #regc ", " #regp " \n\t"\ - "por " #regb ", " #regr " \n\t"\ - "por " #regd ", " #regp " \n\t"\ - "pxor " #rega ", " #regb " \n\t"\ - "pxor " #regc ", " #regd " \n\t"\ - "pand %%mm6, " #regb " \n\t"\ - "pand %%mm6, " #regd " \n\t"\ - "psrlq $1, " #regd " \n\t"\ - "psrlq $1, " #regb " \n\t"\ - "psubb " #regb ", " #regr " \n\t"\ - "psubb " #regd ", " #regp " \n\t" +#define PAVGBP_MMX(rega, regb, regr, regc, regd, regp) \ + "movq "#rega", "#regr" \n\t" \ + "movq "#regc", "#regp" \n\t" \ + "por "#regb", "#regr" \n\t" \ + "por "#regd", "#regp" \n\t" \ + "pxor "#rega", "#regb" \n\t" \ + "pxor "#regc", "#regd" \n\t" \ + "pand %%mm6, "#regb" \n\t" \ + "pand %%mm6, "#regd" \n\t" \ + "psrlq $1, "#regd" \n\t" \ + "psrlq $1, "#regb" \n\t" \ + "psubb "#regb", "#regr" \n\t" \ + "psubb "#regd", "#regp" \n\t" /***********************************/ /* MMX no rounding */ -#define DEF(x, y) x ## _no_rnd_ ## y ##_mmx +#define DEF(x, y) x ## _no_rnd_ ## y ## _mmx #define SET_RND MOVQ_WONE #define PAVGBP(a, b, c, d, e, f) PAVGBP_MMX_NO_RND(a, b, c, d, e, f) #define PAVGB(a, b, c, e) PAVGB_MMX_NO_RND(a, b, c, e) @@ -178,7 +178,7 @@ DECLARE_ALIGNED(16, const double, ff_pd_2)[2] = { 2.0, 2.0 }; /***********************************/ /* MMX rounding */ -#define DEF(x, y) x ## _ ## y ##_mmx +#define DEF(x, y) x ## _ ## y ## _mmx #define SET_RND MOVQ_WTWO #define PAVGBP(a, b, c, d, e, f) PAVGBP_MMX(a, b, c, d, e, f) #define PAVGB(a, b, c, e) PAVGB_MMX(a, b, c, e) @@ -235,537 +235,552 @@ DECLARE_ALIGNED(16, const double, ff_pd_2)[2] = { 2.0, 2.0 }; /***********************************/ /* standard MMX */ -void ff_put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size) +void ff_put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, + int line_size) { const DCTELEM *p; uint8_t *pix; /* read the pixels */ - p = block; + p = block; pix = pixels; /* unrolled loop */ - __asm__ volatile( - "movq %3, %%mm0 \n\t" - "movq 8%3, %%mm1 \n\t" - "movq 16%3, %%mm2 \n\t" - "movq 24%3, %%mm3 \n\t" - "movq 32%3, %%mm4 \n\t" - "movq 40%3, %%mm5 \n\t" - "movq 48%3, %%mm6 \n\t" - "movq 56%3, %%mm7 \n\t" - "packuswb %%mm1, %%mm0 \n\t" - "packuswb %%mm3, %%mm2 \n\t" - "packuswb %%mm5, %%mm4 \n\t" - "packuswb %%mm7, %%mm6 \n\t" - "movq %%mm0, (%0) \n\t" - "movq %%mm2, (%0, %1) \n\t" - "movq %%mm4, (%0, %1, 2) \n\t" - "movq %%mm6, (%0, %2) \n\t" - ::"r" (pix), "r" ((x86_reg)line_size), "r" ((x86_reg)line_size*3), "m"(*p) - :"memory"); - pix += line_size*4; - p += 32; + __asm__ volatile ( + "movq %3, %%mm0 \n\t" + "movq 8%3, %%mm1 \n\t" + "movq 16%3, %%mm2 \n\t" + "movq 24%3, %%mm3 \n\t" + "movq 32%3, %%mm4 \n\t" + "movq 40%3, %%mm5 \n\t" + "movq 48%3, %%mm6 \n\t" + "movq 56%3, %%mm7 \n\t" + "packuswb %%mm1, %%mm0 \n\t" + "packuswb %%mm3, %%mm2 \n\t" + "packuswb %%mm5, %%mm4 \n\t" + "packuswb %%mm7, %%mm6 \n\t" + "movq %%mm0, (%0) \n\t" + "movq %%mm2, (%0, %1) \n\t" + "movq %%mm4, (%0, %1, 2) \n\t" + "movq %%mm6, (%0, %2) \n\t" + :: "r"(pix), "r"((x86_reg)line_size), "r"((x86_reg)line_size * 3), + "m"(*p) + : "memory"); + pix += line_size * 4; + p += 32; // if here would be an exact copy of the code above // compiler would generate some very strange code // thus using "r" - __asm__ volatile( - "movq (%3), %%mm0 \n\t" - "movq 8(%3), %%mm1 \n\t" - "movq 16(%3), %%mm2 \n\t" - "movq 24(%3), %%mm3 \n\t" - "movq 32(%3), %%mm4 \n\t" - "movq 40(%3), %%mm5 \n\t" - "movq 48(%3), %%mm6 \n\t" - "movq 56(%3), %%mm7 \n\t" - "packuswb %%mm1, %%mm0 \n\t" - "packuswb %%mm3, %%mm2 \n\t" - "packuswb %%mm5, %%mm4 \n\t" - "packuswb %%mm7, %%mm6 \n\t" - "movq %%mm0, (%0) \n\t" - "movq %%mm2, (%0, %1) \n\t" - "movq %%mm4, (%0, %1, 2) \n\t" - "movq %%mm6, (%0, %2) \n\t" - ::"r" (pix), "r" ((x86_reg)line_size), "r" ((x86_reg)line_size*3), "r"(p) - :"memory"); + __asm__ volatile ( + "movq (%3), %%mm0 \n\t" + "movq 8(%3), %%mm1 \n\t" + "movq 16(%3), %%mm2 \n\t" + "movq 24(%3), %%mm3 \n\t" + "movq 32(%3), %%mm4 \n\t" + "movq 40(%3), %%mm5 \n\t" + "movq 48(%3), %%mm6 \n\t" + "movq 56(%3), %%mm7 \n\t" + "packuswb %%mm1, %%mm0 \n\t" + "packuswb %%mm3, %%mm2 \n\t" + "packuswb %%mm5, %%mm4 \n\t" + "packuswb %%mm7, %%mm6 \n\t" + "movq %%mm0, (%0) \n\t" + "movq %%mm2, (%0, %1) \n\t" + "movq %%mm4, (%0, %1, 2) \n\t" + "movq %%mm6, (%0, %2) \n\t" + :: "r"(pix), "r"((x86_reg)line_size), "r"((x86_reg)line_size * 3), "r"(p) + : "memory"); } -#define put_signed_pixels_clamped_mmx_half(off) \ - "movq "#off"(%2), %%mm1 \n\t"\ - "movq 16+"#off"(%2), %%mm2 \n\t"\ - "movq 32+"#off"(%2), %%mm3 \n\t"\ - "movq 48+"#off"(%2), %%mm4 \n\t"\ - "packsswb 8+"#off"(%2), %%mm1 \n\t"\ - "packsswb 24+"#off"(%2), %%mm2 \n\t"\ - "packsswb 40+"#off"(%2), %%mm3 \n\t"\ - "packsswb 56+"#off"(%2), %%mm4 \n\t"\ - "paddb %%mm0, %%mm1 \n\t"\ - "paddb %%mm0, %%mm2 \n\t"\ - "paddb %%mm0, %%mm3 \n\t"\ - "paddb %%mm0, %%mm4 \n\t"\ - "movq %%mm1, (%0) \n\t"\ - "movq %%mm2, (%0, %3) \n\t"\ - "movq %%mm3, (%0, %3, 2) \n\t"\ - "movq %%mm4, (%0, %1) \n\t" +#define put_signed_pixels_clamped_mmx_half(off) \ + "movq "#off"(%2), %%mm1 \n\t" \ + "movq 16 + "#off"(%2), %%mm2 \n\t" \ + "movq 32 + "#off"(%2), %%mm3 \n\t" \ + "movq 48 + "#off"(%2), %%mm4 \n\t" \ + "packsswb 8 + "#off"(%2), %%mm1 \n\t" \ + "packsswb 24 + "#off"(%2), %%mm2 \n\t" \ + "packsswb 40 + "#off"(%2), %%mm3 \n\t" \ + "packsswb 56 + "#off"(%2), %%mm4 \n\t" \ + "paddb %%mm0, %%mm1 \n\t" \ + "paddb %%mm0, %%mm2 \n\t" \ + "paddb %%mm0, %%mm3 \n\t" \ + "paddb %%mm0, %%mm4 \n\t" \ + "movq %%mm1, (%0) \n\t" \ + "movq %%mm2, (%0, %3) \n\t" \ + "movq %%mm3, (%0, %3, 2) \n\t" \ + "movq %%mm4, (%0, %1) \n\t" -void ff_put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size) +void ff_put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, + int line_size) { x86_reg line_skip = line_size; x86_reg line_skip3; __asm__ volatile ( - "movq "MANGLE(ff_pb_80)", %%mm0 \n\t" - "lea (%3, %3, 2), %1 \n\t" - put_signed_pixels_clamped_mmx_half(0) - "lea (%0, %3, 4), %0 \n\t" - put_signed_pixels_clamped_mmx_half(64) - :"+&r" (pixels), "=&r" (line_skip3) - :"r" (block), "r"(line_skip) - :"memory"); + "movq "MANGLE(ff_pb_80)", %%mm0 \n\t" + "lea (%3, %3, 2), %1 \n\t" + put_signed_pixels_clamped_mmx_half(0) + "lea (%0, %3, 4), %0 \n\t" + put_signed_pixels_clamped_mmx_half(64) + : "+&r"(pixels), "=&r"(line_skip3) + : "r"(block), "r"(line_skip) + : "memory"); } -void ff_add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size) +void ff_add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, + int line_size) { const DCTELEM *p; uint8_t *pix; int i; /* read the pixels */ - p = block; + p = block; pix = pixels; MOVQ_ZERO(mm7); i = 4; do { - __asm__ volatile( - "movq (%2), %%mm0 \n\t" - "movq 8(%2), %%mm1 \n\t" - "movq 16(%2), %%mm2 \n\t" - "movq 24(%2), %%mm3 \n\t" - "movq %0, %%mm4 \n\t" - "movq %1, %%mm6 \n\t" - "movq %%mm4, %%mm5 \n\t" - "punpcklbw %%mm7, %%mm4 \n\t" - "punpckhbw %%mm7, %%mm5 \n\t" - "paddsw %%mm4, %%mm0 \n\t" - "paddsw %%mm5, %%mm1 \n\t" - "movq %%mm6, %%mm5 \n\t" - "punpcklbw %%mm7, %%mm6 \n\t" - "punpckhbw %%mm7, %%mm5 \n\t" - "paddsw %%mm6, %%mm2 \n\t" - "paddsw %%mm5, %%mm3 \n\t" - "packuswb %%mm1, %%mm0 \n\t" - "packuswb %%mm3, %%mm2 \n\t" - "movq %%mm0, %0 \n\t" - "movq %%mm2, %1 \n\t" - :"+m"(*pix), "+m"(*(pix+line_size)) - :"r"(p) - :"memory"); - pix += line_size*2; - p += 16; + __asm__ volatile ( + "movq (%2), %%mm0 \n\t" + "movq 8(%2), %%mm1 \n\t" + "movq 16(%2), %%mm2 \n\t" + "movq 24(%2), %%mm3 \n\t" + "movq %0, %%mm4 \n\t" + "movq %1, %%mm6 \n\t" + "movq %%mm4, %%mm5 \n\t" + "punpcklbw %%mm7, %%mm4 \n\t" + "punpckhbw %%mm7, %%mm5 \n\t" + "paddsw %%mm4, %%mm0 \n\t" + "paddsw %%mm5, %%mm1 \n\t" + "movq %%mm6, %%mm5 \n\t" + "punpcklbw %%mm7, %%mm6 \n\t" + "punpckhbw %%mm7, %%mm5 \n\t" + "paddsw %%mm6, %%mm2 \n\t" + "paddsw %%mm5, %%mm3 \n\t" + "packuswb %%mm1, %%mm0 \n\t" + "packuswb %%mm3, %%mm2 \n\t" + "movq %%mm0, %0 \n\t" + "movq %%mm2, %1 \n\t" + : "+m"(*pix), "+m"(*(pix + line_size)) + : "r"(p) + : "memory"); + pix += line_size * 2; + p += 16; } while (--i); } -static void put_pixels4_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h) +static void put_pixels4_mmx(uint8_t *block, const uint8_t *pixels, + int line_size, int h) { - __asm__ volatile( - "lea (%3, %3), %%"REG_a" \n\t" - ".p2align 3 \n\t" - "1: \n\t" - "movd (%1), %%mm0 \n\t" - "movd (%1, %3), %%mm1 \n\t" - "movd %%mm0, (%2) \n\t" - "movd %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "movd (%1), %%mm0 \n\t" - "movd (%1, %3), %%mm1 \n\t" - "movd %%mm0, (%2) \n\t" - "movd %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - : "+g"(h), "+r" (pixels), "+r" (block) - : "r"((x86_reg)line_size) - : "%"REG_a, "memory" + __asm__ volatile ( + "lea (%3, %3), %%"REG_a" \n\t" + ".p2align 3 \n\t" + "1: \n\t" + "movd (%1 ), %%mm0 \n\t" + "movd (%1, %3), %%mm1 \n\t" + "movd %%mm0, (%2) \n\t" + "movd %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "movd (%1 ), %%mm0 \n\t" + "movd (%1, %3), %%mm1 \n\t" + "movd %%mm0, (%2) \n\t" + "movd %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + : "+g"(h), "+r"(pixels), "+r"(block) + : "r"((x86_reg)line_size) + : "%"REG_a, "memory" ); } -static void put_pixels8_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h) +static void put_pixels8_mmx(uint8_t *block, const uint8_t *pixels, + int line_size, int h) { - __asm__ volatile( - "lea (%3, %3), %%"REG_a" \n\t" - ".p2align 3 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - : "+g"(h), "+r" (pixels), "+r" (block) - : "r"((x86_reg)line_size) - : "%"REG_a, "memory" + __asm__ volatile ( + "lea (%3, %3), %%"REG_a" \n\t" + ".p2align 3 \n\t" + "1: \n\t" + "movq (%1 ), %%mm0 \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "movq (%1 ), %%mm0 \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + : "+g"(h), "+r"(pixels), "+r"(block) + : "r"((x86_reg)line_size) + : "%"REG_a, "memory" ); } -static void put_pixels16_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h) +static void put_pixels16_mmx(uint8_t *block, const uint8_t *pixels, + int line_size, int h) { - __asm__ volatile( - "lea (%3, %3), %%"REG_a" \n\t" - ".p2align 3 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq 8(%1), %%mm4 \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq 8(%1, %3), %%mm5 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm4, 8(%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "movq %%mm5, 8(%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "movq (%1), %%mm0 \n\t" - "movq 8(%1), %%mm4 \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq 8(%1, %3), %%mm5 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm4, 8(%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "movq %%mm5, 8(%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - : "+g"(h), "+r" (pixels), "+r" (block) - : "r"((x86_reg)line_size) - : "%"REG_a, "memory" + __asm__ volatile ( + "lea (%3, %3), %%"REG_a" \n\t" + ".p2align 3 \n\t" + "1: \n\t" + "movq (%1 ), %%mm0 \n\t" + "movq 8(%1 ), %%mm4 \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq 8(%1, %3), %%mm5 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm4, 8(%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "movq %%mm5, 8(%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "movq (%1 ), %%mm0 \n\t" + "movq 8(%1 ), %%mm4 \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq 8(%1, %3), %%mm5 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm4, 8(%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "movq %%mm5, 8(%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + : "+g"(h), "+r"(pixels), "+r"(block) + : "r"((x86_reg)line_size) + : "%"REG_a, "memory" ); } -static void put_pixels16_sse2(uint8_t *block, const uint8_t *pixels, int line_size, int h) +static void put_pixels16_sse2(uint8_t *block, const uint8_t *pixels, + int line_size, int h) { - __asm__ volatile( - "1: \n\t" - "movdqu (%1), %%xmm0 \n\t" - "movdqu (%1,%3), %%xmm1 \n\t" - "movdqu (%1,%3,2), %%xmm2 \n\t" - "movdqu (%1,%4), %%xmm3 \n\t" - "lea (%1,%3,4), %1 \n\t" - "movdqa %%xmm0, (%2) \n\t" - "movdqa %%xmm1, (%2,%3) \n\t" - "movdqa %%xmm2, (%2,%3,2) \n\t" - "movdqa %%xmm3, (%2,%4) \n\t" - "subl $4, %0 \n\t" - "lea (%2,%3,4), %2 \n\t" - "jnz 1b \n\t" - : "+g"(h), "+r" (pixels), "+r" (block) - : "r"((x86_reg)line_size), "r"((x86_reg)3L*line_size) - : "memory" + __asm__ volatile ( + "1: \n\t" + "movdqu (%1 ), %%xmm0 \n\t" + "movdqu (%1, %3 ), %%xmm1 \n\t" + "movdqu (%1, %3, 2), %%xmm2 \n\t" + "movdqu (%1, %4 ), %%xmm3 \n\t" + "lea (%1, %3, 4), %1 \n\t" + "movdqa %%xmm0, (%2) \n\t" + "movdqa %%xmm1, (%2, %3) \n\t" + "movdqa %%xmm2, (%2, %3, 2) \n\t" + "movdqa %%xmm3, (%2, %4) \n\t" + "subl $4, %0 \n\t" + "lea (%2, %3, 4), %2 \n\t" + "jnz 1b \n\t" + : "+g"(h), "+r"(pixels), "+r"(block) + : "r"((x86_reg)line_size), "r"((x86_reg)3L * line_size) + : "memory" ); } -static void avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels, int line_size, int h) +static void avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels, + int line_size, int h) { - __asm__ volatile( - "1: \n\t" - "movdqu (%1), %%xmm0 \n\t" - "movdqu (%1,%3), %%xmm1 \n\t" - "movdqu (%1,%3,2), %%xmm2 \n\t" - "movdqu (%1,%4), %%xmm3 \n\t" - "lea (%1,%3,4), %1 \n\t" - "pavgb (%2), %%xmm0 \n\t" - "pavgb (%2,%3), %%xmm1 \n\t" - "pavgb (%2,%3,2), %%xmm2 \n\t" - "pavgb (%2,%4), %%xmm3 \n\t" - "movdqa %%xmm0, (%2) \n\t" - "movdqa %%xmm1, (%2,%3) \n\t" - "movdqa %%xmm2, (%2,%3,2) \n\t" - "movdqa %%xmm3, (%2,%4) \n\t" - "subl $4, %0 \n\t" - "lea (%2,%3,4), %2 \n\t" - "jnz 1b \n\t" - : "+g"(h), "+r" (pixels), "+r" (block) - : "r"((x86_reg)line_size), "r"((x86_reg)3L*line_size) - : "memory" + __asm__ volatile ( + "1: \n\t" + "movdqu (%1 ), %%xmm0 \n\t" + "movdqu (%1, %3 ), %%xmm1 \n\t" + "movdqu (%1, %3, 2), %%xmm2 \n\t" + "movdqu (%1, %4 ), %%xmm3 \n\t" + "lea (%1, %3, 4), %1 \n\t" + "pavgb (%2 ), %%xmm0 \n\t" + "pavgb (%2, %3 ), %%xmm1 \n\t" + "pavgb (%2, %3, 2), %%xmm2 \n\t" + "pavgb (%2, %4), %%xmm3 \n\t" + "movdqa %%xmm0, (%2) \n\t" + "movdqa %%xmm1, (%2, %3) \n\t" + "movdqa %%xmm2, (%2, %3, 2) \n\t" + "movdqa %%xmm3, (%2, %4) \n\t" + "subl $4, %0 \n\t" + "lea (%2, %3, 4), %2 \n\t" + "jnz 1b \n\t" + : "+g"(h), "+r"(pixels), "+r"(block) + : "r"((x86_reg)line_size), "r"((x86_reg)3L * line_size) + : "memory" ); } -#define CLEAR_BLOCKS(name,n) \ -static void name(DCTELEM *blocks)\ -{\ - __asm__ volatile(\ - "pxor %%mm7, %%mm7 \n\t"\ - "mov %1, %%"REG_a" \n\t"\ - "1: \n\t"\ - "movq %%mm7, (%0, %%"REG_a") \n\t"\ - "movq %%mm7, 8(%0, %%"REG_a") \n\t"\ - "movq %%mm7, 16(%0, %%"REG_a") \n\t"\ - "movq %%mm7, 24(%0, %%"REG_a") \n\t"\ - "add $32, %%"REG_a" \n\t"\ - " js 1b \n\t"\ - : : "r" (((uint8_t *)blocks)+128*n),\ - "i" (-128*n)\ - : "%"REG_a\ - );\ +#define CLEAR_BLOCKS(name, n) \ +static void name(DCTELEM *blocks) \ +{ \ + __asm__ volatile ( \ + "pxor %%mm7, %%mm7 \n\t" \ + "mov %1, %%"REG_a" \n\t" \ + "1: \n\t" \ + "movq %%mm7, (%0, %%"REG_a") \n\t" \ + "movq %%mm7, 8(%0, %%"REG_a") \n\t" \ + "movq %%mm7, 16(%0, %%"REG_a") \n\t" \ + "movq %%mm7, 24(%0, %%"REG_a") \n\t" \ + "add $32, %%"REG_a" \n\t" \ + "js 1b \n\t" \ + :: "r"(((uint8_t *)blocks) + 128 * n), \ + "i"(-128 * n) \ + : "%"REG_a \ + ); \ } CLEAR_BLOCKS(clear_blocks_mmx, 6) CLEAR_BLOCKS(clear_block_mmx, 1) static void clear_block_sse(DCTELEM *block) { - __asm__ volatile( - "xorps %%xmm0, %%xmm0 \n" - "movaps %%xmm0, (%0) \n" - "movaps %%xmm0, 16(%0) \n" - "movaps %%xmm0, 32(%0) \n" - "movaps %%xmm0, 48(%0) \n" - "movaps %%xmm0, 64(%0) \n" - "movaps %%xmm0, 80(%0) \n" - "movaps %%xmm0, 96(%0) \n" - "movaps %%xmm0, 112(%0) \n" + __asm__ volatile ( + "xorps %%xmm0, %%xmm0 \n" + "movaps %%xmm0, (%0) \n" + "movaps %%xmm0, 16(%0) \n" + "movaps %%xmm0, 32(%0) \n" + "movaps %%xmm0, 48(%0) \n" + "movaps %%xmm0, 64(%0) \n" + "movaps %%xmm0, 80(%0) \n" + "movaps %%xmm0, 96(%0) \n" + "movaps %%xmm0, 112(%0) \n" :: "r"(block) : "memory" ); } static void clear_blocks_sse(DCTELEM *blocks) -{\ - __asm__ volatile( - "xorps %%xmm0, %%xmm0 \n" - "mov %1, %%"REG_a" \n" - "1: \n" - "movaps %%xmm0, (%0, %%"REG_a") \n" - "movaps %%xmm0, 16(%0, %%"REG_a") \n" - "movaps %%xmm0, 32(%0, %%"REG_a") \n" - "movaps %%xmm0, 48(%0, %%"REG_a") \n" - "movaps %%xmm0, 64(%0, %%"REG_a") \n" - "movaps %%xmm0, 80(%0, %%"REG_a") \n" - "movaps %%xmm0, 96(%0, %%"REG_a") \n" - "movaps %%xmm0, 112(%0, %%"REG_a") \n" - "add $128, %%"REG_a" \n" - " js 1b \n" - : : "r" (((uint8_t *)blocks)+128*6), - "i" (-128*6) +{ + __asm__ volatile ( + "xorps %%xmm0, %%xmm0 \n" + "mov %1, %%"REG_a" \n" + "1: \n" + "movaps %%xmm0, (%0, %%"REG_a") \n" + "movaps %%xmm0, 16(%0, %%"REG_a") \n" + "movaps %%xmm0, 32(%0, %%"REG_a") \n" + "movaps %%xmm0, 48(%0, %%"REG_a") \n" + "movaps %%xmm0, 64(%0, %%"REG_a") \n" + "movaps %%xmm0, 80(%0, %%"REG_a") \n" + "movaps %%xmm0, 96(%0, %%"REG_a") \n" + "movaps %%xmm0, 112(%0, %%"REG_a") \n" + "add $128, %%"REG_a" \n" + "js 1b \n" + :: "r"(((uint8_t *)blocks) + 128 * 6), + "i"(-128 * 6) : "%"REG_a ); } -static void add_bytes_mmx(uint8_t *dst, uint8_t *src, int w){ - x86_reg i=0; - __asm__ volatile( - "jmp 2f \n\t" +static void add_bytes_mmx(uint8_t *dst, uint8_t *src, int w) +{ + x86_reg i = 0; + __asm__ volatile ( + "jmp 2f \n\t" "1: \n\t" - "movq (%1, %0), %%mm0 \n\t" - "movq (%2, %0), %%mm1 \n\t" - "paddb %%mm0, %%mm1 \n\t" - "movq %%mm1, (%2, %0) \n\t" - "movq 8(%1, %0), %%mm0 \n\t" - "movq 8(%2, %0), %%mm1 \n\t" - "paddb %%mm0, %%mm1 \n\t" - "movq %%mm1, 8(%2, %0) \n\t" - "add $16, %0 \n\t" + "movq (%1, %0), %%mm0 \n\t" + "movq (%2, %0), %%mm1 \n\t" + "paddb %%mm0, %%mm1 \n\t" + "movq %%mm1, (%2, %0) \n\t" + "movq 8(%1, %0), %%mm0 \n\t" + "movq 8(%2, %0), %%mm1 \n\t" + "paddb %%mm0, %%mm1 \n\t" + "movq %%mm1, 8(%2, %0) \n\t" + "add $16, %0 \n\t" "2: \n\t" - "cmp %3, %0 \n\t" - " js 1b \n\t" - : "+r" (i) - : "r"(src), "r"(dst), "r"((x86_reg)w-15) + "cmp %3, %0 \n\t" + "js 1b \n\t" + : "+r"(i) + : "r"(src), "r"(dst), "r"((x86_reg)w - 15) ); - for(; i= h){ - src_y_add = h-1-src_y; - src_y=h-1; - }else if(src_y<=-block_h){ - src_y_add = 1-block_h-src_y; - src_y=1-block_h; + if (src_y >= h) { + src_y_add = h - 1 - src_y; + src_y = h - 1; + } else if (src_y <= -block_h) { + src_y_add = 1 - block_h - src_y; + src_y = 1 - block_h; } - if(src_x>= w){ - src+= (w-1-src_x); - src_x=w-1; - }else if(src_x<=-block_w){ - src+= (1-block_w-src_x); - src_x=1-block_w; + if (src_x >= w) { + src += w - 1 - src_x; + src_x = w - 1; + } else if (src_x <= -block_w) { + src += 1 - block_w - src_x; + src_x = 1 - block_w; } - start_y= FFMAX(0, -src_y); - start_x= FFMAX(0, -src_x); - end_y= FFMIN(block_h, h-src_y); - end_x= FFMIN(block_w, w-src_x); + start_y = FFMAX(0, -src_y); + start_x = FFMAX(0, -src_x); + end_y = FFMIN(block_h, h-src_y); + end_x = FFMIN(block_w, w-src_x); assert(start_x < end_x && block_w > 0); assert(start_y < end_y && block_h > 0); // fill in the to-be-copied part plus all above/below - src += (src_y_add+start_y)*linesize + start_x; + src += (src_y_add + start_y) * linesize + start_x; buf += start_x; - core_fn(buf, src, linesize, start_y, end_y, block_h, start_x, end_x, block_w); + core_fn(buf, src, linesize, start_y, end_y, + block_h, start_x, end_x, block_w); } #if ARCH_X86_32 -static av_noinline -void emulated_edge_mc_mmx(uint8_t *buf, const uint8_t *src, int linesize, - int block_w, int block_h, - int src_x, int src_y, int w, int h) +static av_noinline void emulated_edge_mc_mmx(uint8_t *buf, const uint8_t *src, + int linesize, + int block_w, int block_h, + int src_x, int src_y, int w, int h) { emulated_edge_mc(buf, src, linesize, block_w, block_h, src_x, src_y, w, h, &ff_emu_edge_core_mmx); } #endif -static av_noinline -void emulated_edge_mc_sse(uint8_t *buf, const uint8_t *src, int linesize, - int block_w, int block_h, - int src_x, int src_y, int w, int h) + +static av_noinline void emulated_edge_mc_sse(uint8_t *buf, const uint8_t *src, + int linesize, + int block_w, int block_h, + int src_x, int src_y, int w, int h) { emulated_edge_mc(buf, src, linesize, block_w, block_h, src_x, src_y, w, h, &ff_emu_edge_core_sse); } #endif /* HAVE_YASM */ -typedef void emulated_edge_mc_func (uint8_t *dst, const uint8_t *src, - int linesize, int block_w, int block_h, - int src_x, int src_y, int w, int h); +typedef void emulated_edge_mc_func(uint8_t *dst, const uint8_t *src, + int linesize, int block_w, int block_h, + int src_x, int src_y, int w, int h); -static av_always_inline -void gmc(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy, - int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height, - emulated_edge_mc_func *emu_edge_fn) +static av_always_inline void gmc(uint8_t *dst, uint8_t *src, + int stride, int h, int ox, int oy, + int dxx, int dxy, int dyx, int dyy, + int shift, int r, int width, int height, + emulated_edge_mc_func *emu_edge_fn) { - const int w = 8; - const int ix = ox>>(16+shift); - const int iy = oy>>(16+shift); - const int oxs = ox>>4; - const int oys = oy>>4; - const int dxxs = dxx>>4; - const int dxys = dxy>>4; - const int dyxs = dyx>>4; - const int dyys = dyy>>4; - const uint16_t r4[4] = {r,r,r,r}; - const uint16_t dxy4[4] = {dxys,dxys,dxys,dxys}; - const uint16_t dyy4[4] = {dyys,dyys,dyys,dyys}; - const uint64_t shift2 = 2*shift; - uint8_t edge_buf[(h+1)*stride]; + const int w = 8; + const int ix = ox >> (16 + shift); + const int iy = oy >> (16 + shift); + const int oxs = ox >> 4; + const int oys = oy >> 4; + const int dxxs = dxx >> 4; + const int dxys = dxy >> 4; + const int dyxs = dyx >> 4; + const int dyys = dyy >> 4; + const uint16_t r4[4] = { r, r, r, r }; + const uint16_t dxy4[4] = { dxys, dxys, dxys, dxys }; + const uint16_t dyy4[4] = { dyys, dyys, dyys, dyys }; + const uint64_t shift2 = 2 * shift; + uint8_t edge_buf[(h + 1) * stride]; int x, y; - const int dxw = (dxx-(1<<(16+shift)))*(w-1); - const int dyh = (dyy-(1<<(16+shift)))*(h-1); - const int dxh = dxy*(h-1); - const int dyw = dyx*(w-1); - if( // non-constant fullpel offset (3% of blocks) - ((ox^(ox+dxw)) | (ox^(ox+dxh)) | (ox^(ox+dxw+dxh)) | - (oy^(oy+dyw)) | (oy^(oy+dyh)) | (oy^(oy+dyw+dyh))) >> (16+shift) + const int dxw = (dxx - (1 << (16 + shift))) * (w - 1); + const int dyh = (dyy - (1 << (16 + shift))) * (h - 1); + const int dxh = dxy * (h - 1); + const int dyw = dyx * (w - 1); + if ( // non-constant fullpel offset (3% of blocks) + ((ox ^ (ox + dxw)) | (ox ^ (ox + dxh)) | (ox ^ (ox + dxw + dxh)) | + (oy ^ (oy + dyw)) | (oy ^ (oy + dyh)) | (oy ^ (oy + dyw + dyh))) >> (16 + shift) // uses more than 16 bits of subpel mv (only at huge resolution) - || (dxx|dxy|dyx|dyy)&15 ) - { - //FIXME could still use mmx for some of the rows - ff_gmc_c(dst, src, stride, h, ox, oy, dxx, dxy, dyx, dyy, shift, r, width, height); + || (dxx | dxy | dyx | dyy) & 15) { + // FIXME could still use mmx for some of the rows + ff_gmc_c(dst, src, stride, h, ox, oy, dxx, dxy, dyx, dyy, + shift, r, width, height); return; } - src += ix + iy*stride; - if( (unsigned)ix >= width-w || - (unsigned)iy >= height-h ) - { - emu_edge_fn(edge_buf, src, stride, w+1, h+1, ix, iy, width, height); + src += ix + iy * stride; + if ((unsigned)ix >= width - w || + (unsigned)iy >= height - h) { + emu_edge_fn(edge_buf, src, stride, w + 1, h + 1, ix, iy, width, height); src = edge_buf; } - __asm__ volatile( - "movd %0, %%mm6 \n\t" - "pxor %%mm7, %%mm7 \n\t" - "punpcklwd %%mm6, %%mm6 \n\t" - "punpcklwd %%mm6, %%mm6 \n\t" + __asm__ volatile ( + "movd %0, %%mm6 \n\t" + "pxor %%mm7, %%mm7 \n\t" + "punpcklwd %%mm6, %%mm6 \n\t" + "punpcklwd %%mm6, %%mm6 \n\t" :: "r"(1<0) & (a ^ sign(m))) - "movq %%mm3, %1 \n\t" - "movq %%mm0, %0 \n\t" - :"+m"(mag[i]), "+m"(ang[i]) - ::"memory" + __asm__ volatile ("pxor %%mm7, %%mm7":); + for (i = 0; i < blocksize; i += 2) { + __asm__ volatile ( + "movq %0, %%mm0 \n\t" + "movq %1, %%mm1 \n\t" + "movq %%mm0, %%mm2 \n\t" + "movq %%mm1, %%mm3 \n\t" + "pfcmpge %%mm7, %%mm2 \n\t" // m <= 0.0 + "pfcmpge %%mm7, %%mm3 \n\t" // a <= 0.0 + "pslld $31, %%mm2 \n\t" // keep only the sign bit + "pxor %%mm2, %%mm1 \n\t" + "movq %%mm3, %%mm4 \n\t" + "pand %%mm1, %%mm3 \n\t" + "pandn %%mm1, %%mm4 \n\t" + "pfadd %%mm0, %%mm3 \n\t" // a = m + ((a < 0) & (a ^ sign(m))) + "pfsub %%mm4, %%mm0 \n\t" // m = m + ((a > 0) & (a ^ sign(m))) + "movq %%mm3, %1 \n\t" + "movq %%mm0, %0 \n\t" + : "+m"(mag[i]), "+m"(ang[i]) + :: "memory" ); } - __asm__ volatile("femms"); + __asm__ volatile ("femms"); } + static void vorbis_inverse_coupling_sse(float *mag, float *ang, int blocksize) { int i; - __asm__ volatile( - "movaps %0, %%xmm5 \n\t" - ::"m"(ff_pdw_80000000[0]) + __asm__ volatile ( + "movaps %0, %%xmm5 \n\t" + :: "m"(ff_pdw_80000000[0]) ); - for(i=0; i0) & (a ^ sign(m))) + "addps %%xmm0, %%xmm3 \n\t" // a = m + ((a < 0) & (a ^ sign(m))) + "subps %%xmm4, %%xmm0 \n\t" // m = m + ((a > 0) & (a ^ sign(m))) "movaps %%xmm3, %1 \n\t" "movaps %%xmm0, %0 \n\t" - :"+m"(mag[i]), "+m"(ang[i]) - ::"memory" + : "+m"(mag[i]), "+m"(ang[i]) + :: "memory" ); } } @@ -2049,97 +2305,105 @@ static void vorbis_inverse_coupling_sse(float *mag, float *ang, int blocksize) #define IF1(x) x #define IF0(x) -#define MIX5(mono,stereo)\ - __asm__ volatile(\ - "movss 0(%2), %%xmm5 \n"\ - "movss 8(%2), %%xmm6 \n"\ - "movss 24(%2), %%xmm7 \n"\ - "shufps $0, %%xmm5, %%xmm5 \n"\ - "shufps $0, %%xmm6, %%xmm6 \n"\ - "shufps $0, %%xmm7, %%xmm7 \n"\ - "1: \n"\ - "movaps (%0,%1), %%xmm0 \n"\ - "movaps 0x400(%0,%1), %%xmm1 \n"\ - "movaps 0x800(%0,%1), %%xmm2 \n"\ - "movaps 0xc00(%0,%1), %%xmm3 \n"\ - "movaps 0x1000(%0,%1), %%xmm4 \n"\ - "mulps %%xmm5, %%xmm0 \n"\ - "mulps %%xmm6, %%xmm1 \n"\ - "mulps %%xmm5, %%xmm2 \n"\ - "mulps %%xmm7, %%xmm3 \n"\ - "mulps %%xmm7, %%xmm4 \n"\ - stereo("addps %%xmm1, %%xmm0 \n")\ - "addps %%xmm1, %%xmm2 \n"\ - "addps %%xmm3, %%xmm0 \n"\ - "addps %%xmm4, %%xmm2 \n"\ - mono("addps %%xmm2, %%xmm0 \n")\ - "movaps %%xmm0, (%0,%1) \n"\ - stereo("movaps %%xmm2, 0x400(%0,%1) \n")\ - "add $16, %0 \n"\ - "jl 1b \n"\ - :"+&r"(i)\ - :"r"(samples[0]+len), "r"(matrix)\ - :XMM_CLOBBERS("%xmm0", "%xmm1", "%xmm2", "%xmm3", \ - "%xmm4", "%xmm5", "%xmm6", "%xmm7",)\ - "memory"\ +#define MIX5(mono, stereo) \ + __asm__ volatile ( \ + "movss 0(%2), %%xmm5 \n" \ + "movss 8(%2), %%xmm6 \n" \ + "movss 24(%2), %%xmm7 \n" \ + "shufps $0, %%xmm5, %%xmm5 \n" \ + "shufps $0, %%xmm6, %%xmm6 \n" \ + "shufps $0, %%xmm7, %%xmm7 \n" \ + "1: \n" \ + "movaps (%0, %1), %%xmm0 \n" \ + "movaps 0x400(%0, %1), %%xmm1 \n" \ + "movaps 0x800(%0, %1), %%xmm2 \n" \ + "movaps 0xc00(%0, %1), %%xmm3 \n" \ + "movaps 0x1000(%0, %1), %%xmm4 \n" \ + "mulps %%xmm5, %%xmm0 \n" \ + "mulps %%xmm6, %%xmm1 \n" \ + "mulps %%xmm5, %%xmm2 \n" \ + "mulps %%xmm7, %%xmm3 \n" \ + "mulps %%xmm7, %%xmm4 \n" \ + stereo("addps %%xmm1, %%xmm0 \n") \ + "addps %%xmm1, %%xmm2 \n" \ + "addps %%xmm3, %%xmm0 \n" \ + "addps %%xmm4, %%xmm2 \n" \ + mono("addps %%xmm2, %%xmm0 \n") \ + "movaps %%xmm0, (%0, %1) \n" \ + stereo("movaps %%xmm2, 0x400(%0, %1) \n") \ + "add $16, %0 \n" \ + "jl 1b \n" \ + : "+&r"(i) \ + : "r"(samples[0] + len), "r"(matrix) \ + : XMM_CLOBBERS("%xmm0", "%xmm1", "%xmm2", "%xmm3", \ + "%xmm4", "%xmm5", "%xmm6", "%xmm7",) \ + "memory" \ ); -#define MIX_MISC(stereo)\ - __asm__ volatile(\ - "1: \n"\ - "movaps (%3,%0), %%xmm0 \n"\ - stereo("movaps %%xmm0, %%xmm1 \n")\ - "mulps %%xmm4, %%xmm0 \n"\ - stereo("mulps %%xmm5, %%xmm1 \n")\ - "lea 1024(%3,%0), %1 \n"\ - "mov %5, %2 \n"\ - "2: \n"\ - "movaps (%1), %%xmm2 \n"\ - stereo("movaps %%xmm2, %%xmm3 \n")\ - "mulps (%4,%2), %%xmm2 \n"\ - stereo("mulps 16(%4,%2), %%xmm3 \n")\ - "addps %%xmm2, %%xmm0 \n"\ - stereo("addps %%xmm3, %%xmm1 \n")\ - "add $1024, %1 \n"\ - "add $32, %2 \n"\ - "jl 2b \n"\ - "movaps %%xmm0, (%3,%0) \n"\ - stereo("movaps %%xmm1, 1024(%3,%0) \n")\ - "add $16, %0 \n"\ - "jl 1b \n"\ - :"+&r"(i), "=&r"(j), "=&r"(k)\ - :"r"(samples[0]+len), "r"(matrix_simd+in_ch), "g"((intptr_t)-32*(in_ch-1))\ - :"memory"\ +#define MIX_MISC(stereo) \ + __asm__ volatile ( \ + "1: \n" \ + "movaps (%3, %0), %%xmm0 \n" \ + stereo("movaps %%xmm0, %%xmm1 \n") \ + "mulps %%xmm4, %%xmm0 \n" \ + stereo("mulps %%xmm5, %%xmm1 \n") \ + "lea 1024(%3, %0), %1 \n" \ + "mov %5, %2 \n" \ + "2: \n" \ + "movaps (%1), %%xmm2 \n" \ + stereo("movaps %%xmm2, %%xmm3 \n") \ + "mulps (%4, %2), %%xmm2 \n" \ + stereo("mulps 16(%4, %2), %%xmm3 \n") \ + "addps %%xmm2, %%xmm0 \n" \ + stereo("addps %%xmm3, %%xmm1 \n") \ + "add $1024, %1 \n" \ + "add $32, %2 \n" \ + "jl 2b \n" \ + "movaps %%xmm0, (%3, %0) \n" \ + stereo("movaps %%xmm1, 1024(%3, %0) \n") \ + "add $16, %0 \n" \ + "jl 1b \n" \ + : "+&r"(i), "=&r"(j), "=&r"(k) \ + : "r"(samples[0] + len), "r"(matrix_simd + in_ch), \ + "g"((intptr_t) - 32 * (in_ch - 1)) \ + : "memory" \ ); -static void ac3_downmix_sse(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len) +static void ac3_downmix_sse(float (*samples)[256], float (*matrix)[2], + int out_ch, int in_ch, int len) { int (*matrix_cmp)[2] = (int(*)[2])matrix; - intptr_t i,j,k; + intptr_t i, j, k; - i = -len*sizeof(float); - if(in_ch == 5 && out_ch == 2 && !(matrix_cmp[0][1]|matrix_cmp[2][0]|matrix_cmp[3][1]|matrix_cmp[4][0]|(matrix_cmp[1][0]^matrix_cmp[1][1])|(matrix_cmp[0][0]^matrix_cmp[2][1]))) { - MIX5(IF0,IF1); - } else if(in_ch == 5 && out_ch == 1 && matrix_cmp[0][0]==matrix_cmp[2][0] && matrix_cmp[3][0]==matrix_cmp[4][0]) { - MIX5(IF1,IF0); + i = -len * sizeof(float); + if (in_ch == 5 && out_ch == 2 && + !(matrix_cmp[0][1] | matrix_cmp[2][0] | + matrix_cmp[3][1] | matrix_cmp[4][0] | + (matrix_cmp[1][0] ^ matrix_cmp[1][1]) | + (matrix_cmp[0][0] ^ matrix_cmp[2][1]))) { + MIX5(IF0, IF1); + } else if (in_ch == 5 && out_ch == 1 && + matrix_cmp[0][0] == matrix_cmp[2][0] && + matrix_cmp[3][0] == matrix_cmp[4][0]) { + MIX5(IF1, IF0); } else { DECLARE_ALIGNED(16, float, matrix_simd)[AC3_MAX_CHANNELS][2][4]; - j = 2*in_ch*sizeof(float); - __asm__ volatile( - "1: \n" - "sub $8, %0 \n" - "movss (%2,%0), %%xmm4 \n" - "movss 4(%2,%0), %%xmm5 \n" - "shufps $0, %%xmm4, %%xmm4 \n" - "shufps $0, %%xmm5, %%xmm5 \n" - "movaps %%xmm4, (%1,%0,4) \n" - "movaps %%xmm5, 16(%1,%0,4) \n" - "jg 1b \n" - :"+&r"(j) - :"r"(matrix_simd), "r"(matrix) - :"memory" + j = 2 * in_ch * sizeof(float); + __asm__ volatile ( + "1: \n" + "sub $8, %0 \n" + "movss (%2, %0), %%xmm4 \n" + "movss 4(%2, %0), %%xmm5 \n" + "shufps $0, %%xmm4, %%xmm4 \n" + "shufps $0, %%xmm5, %%xmm5 \n" + "movaps %%xmm4, (%1, %0, 4) \n" + "movaps %%xmm5, 16(%1, %0, 4) \n" + "jg 1b \n" + : "+&r"(j) + : "r"(matrix_simd), "r"(matrix) + : "memory" ); - if(out_ch == 2) { + if (out_ch == 2) { MIX_MISC(IF1); } else { MIX_MISC(IF0); @@ -2147,216 +2411,232 @@ static void ac3_downmix_sse(float (*samples)[256], float (*matrix)[2], int out_c } } -static void vector_fmul_3dnow(float *dst, const float *src0, const float *src1, int len){ - x86_reg i = (len-4)*4; - __asm__ volatile( - "1: \n\t" - "movq (%2,%0), %%mm0 \n\t" - "movq 8(%2,%0), %%mm1 \n\t" - "pfmul (%3,%0), %%mm0 \n\t" - "pfmul 8(%3,%0), %%mm1 \n\t" - "movq %%mm0, (%1,%0) \n\t" - "movq %%mm1, 8(%1,%0) \n\t" - "sub $16, %0 \n\t" - "jge 1b \n\t" - "femms \n\t" - :"+r"(i) - :"r"(dst), "r"(src0), "r"(src1) - :"memory" - ); -} -static void vector_fmul_sse(float *dst, const float *src0, const float *src1, int len){ - x86_reg i = (len-8)*4; - __asm__ volatile( - "1: \n\t" - "movaps (%2,%0), %%xmm0 \n\t" - "movaps 16(%2,%0), %%xmm1 \n\t" - "mulps (%3,%0), %%xmm0 \n\t" - "mulps 16(%3,%0), %%xmm1 \n\t" - "movaps %%xmm0, (%1,%0) \n\t" - "movaps %%xmm1, 16(%1,%0) \n\t" - "sub $32, %0 \n\t" - "jge 1b \n\t" - :"+r"(i) - :"r"(dst), "r"(src0), "r"(src1) - :"memory" +static void vector_fmul_3dnow(float *dst, const float *src0, const float *src1, + int len) +{ + x86_reg i = (len - 4) * 4; + __asm__ volatile ( + "1: \n\t" + "movq (%2, %0), %%mm0 \n\t" + "movq 8(%2, %0), %%mm1 \n\t" + "pfmul (%3, %0), %%mm0 \n\t" + "pfmul 8(%3, %0), %%mm1 \n\t" + "movq %%mm0, (%1, %0) \n\t" + "movq %%mm1, 8(%1, %0) \n\t" + "sub $16, %0 \n\t" + "jge 1b \n\t" + "femms \n\t" + : "+r"(i) + : "r"(dst), "r"(src0), "r"(src1) + : "memory" ); } -static void vector_fmul_reverse_3dnow2(float *dst, const float *src0, const float *src1, int len){ - x86_reg i = len*4-16; - __asm__ volatile( - "1: \n\t" - "pswapd 8(%1), %%mm0 \n\t" - "pswapd (%1), %%mm1 \n\t" - "pfmul (%3,%0), %%mm0 \n\t" - "pfmul 8(%3,%0), %%mm1 \n\t" - "movq %%mm0, (%2,%0) \n\t" - "movq %%mm1, 8(%2,%0) \n\t" - "add $16, %1 \n\t" - "sub $16, %0 \n\t" - "jge 1b \n\t" - :"+r"(i), "+r"(src1) - :"r"(dst), "r"(src0) - ); - __asm__ volatile("femms"); -} -static void vector_fmul_reverse_sse(float *dst, const float *src0, const float *src1, int len){ - x86_reg i = len*4-32; - __asm__ volatile( - "1: \n\t" - "movaps 16(%1), %%xmm0 \n\t" - "movaps (%1), %%xmm1 \n\t" - "shufps $0x1b, %%xmm0, %%xmm0 \n\t" - "shufps $0x1b, %%xmm1, %%xmm1 \n\t" - "mulps (%3,%0), %%xmm0 \n\t" - "mulps 16(%3,%0), %%xmm1 \n\t" - "movaps %%xmm0, (%2,%0) \n\t" - "movaps %%xmm1, 16(%2,%0) \n\t" - "add $32, %1 \n\t" - "sub $32, %0 \n\t" - "jge 1b \n\t" - :"+r"(i), "+r"(src1) - :"r"(dst), "r"(src0) +static void vector_fmul_sse(float *dst, const float *src0, const float *src1, + int len) +{ + x86_reg i = (len - 8) * 4; + __asm__ volatile ( + "1: \n\t" + "movaps (%2, %0), %%xmm0 \n\t" + "movaps 16(%2, %0), %%xmm1 \n\t" + "mulps (%3, %0), %%xmm0 \n\t" + "mulps 16(%3, %0), %%xmm1 \n\t" + "movaps %%xmm0, (%1, %0) \n\t" + "movaps %%xmm1, 16(%1, %0) \n\t" + "sub $32, %0 \n\t" + "jge 1b \n\t" + : "+r"(i) + : "r"(dst), "r"(src0), "r"(src1) + : "memory" ); } -static void vector_fmul_add_3dnow(float *dst, const float *src0, const float *src1, - const float *src2, int len){ - x86_reg i = (len-4)*4; - __asm__ volatile( - "1: \n\t" - "movq (%2,%0), %%mm0 \n\t" - "movq 8(%2,%0), %%mm1 \n\t" - "pfmul (%3,%0), %%mm0 \n\t" - "pfmul 8(%3,%0), %%mm1 \n\t" - "pfadd (%4,%0), %%mm0 \n\t" - "pfadd 8(%4,%0), %%mm1 \n\t" - "movq %%mm0, (%1,%0) \n\t" - "movq %%mm1, 8(%1,%0) \n\t" - "sub $16, %0 \n\t" - "jge 1b \n\t" - :"+r"(i) - :"r"(dst), "r"(src0), "r"(src1), "r"(src2) - :"memory" +static void vector_fmul_reverse_3dnow2(float *dst, const float *src0, + const float *src1, int len) +{ + x86_reg i = len * 4 - 16; + __asm__ volatile ( + "1: \n\t" + "pswapd 8(%1), %%mm0 \n\t" + "pswapd (%1), %%mm1 \n\t" + "pfmul (%3, %0), %%mm0 \n\t" + "pfmul 8(%3, %0), %%mm1 \n\t" + "movq %%mm0, (%2, %0) \n\t" + "movq %%mm1, 8(%2, %0) \n\t" + "add $16, %1 \n\t" + "sub $16, %0 \n\t" + "jge 1b \n\t" + : "+r"(i), "+r"(src1) + : "r"(dst), "r"(src0) ); - __asm__ volatile("femms"); + __asm__ volatile ("femms"); } -static void vector_fmul_add_sse(float *dst, const float *src0, const float *src1, - const float *src2, int len){ - x86_reg i = (len-8)*4; - __asm__ volatile( - "1: \n\t" - "movaps (%2,%0), %%xmm0 \n\t" - "movaps 16(%2,%0), %%xmm1 \n\t" - "mulps (%3,%0), %%xmm0 \n\t" - "mulps 16(%3,%0), %%xmm1 \n\t" - "addps (%4,%0), %%xmm0 \n\t" - "addps 16(%4,%0), %%xmm1 \n\t" - "movaps %%xmm0, (%1,%0) \n\t" - "movaps %%xmm1, 16(%1,%0) \n\t" - "sub $32, %0 \n\t" - "jge 1b \n\t" - :"+r"(i) - :"r"(dst), "r"(src0), "r"(src1), "r"(src2) - :"memory" + +static void vector_fmul_reverse_sse(float *dst, const float *src0, + const float *src1, int len) +{ + x86_reg i = len * 4 - 32; + __asm__ volatile ( + "1: \n\t" + "movaps 16(%1), %%xmm0 \n\t" + "movaps (%1), %%xmm1 \n\t" + "shufps $0x1b, %%xmm0, %%xmm0 \n\t" + "shufps $0x1b, %%xmm1, %%xmm1 \n\t" + "mulps (%3, %0), %%xmm0 \n\t" + "mulps 16(%3, %0), %%xmm1 \n\t" + "movaps %%xmm0, (%2, %0) \n\t" + "movaps %%xmm1, 16(%2, %0) \n\t" + "add $32, %1 \n\t" + "sub $32, %0 \n\t" + "jge 1b \n\t" + : "+r"(i), "+r"(src1) + : "r"(dst), "r"(src0) + ); +} + +static void vector_fmul_add_3dnow(float *dst, const float *src0, + const float *src1, const float *src2, int len) +{ + x86_reg i = (len - 4) * 4; + __asm__ volatile ( + "1: \n\t" + "movq (%2, %0), %%mm0 \n\t" + "movq 8(%2, %0), %%mm1 \n\t" + "pfmul (%3, %0), %%mm0 \n\t" + "pfmul 8(%3, %0), %%mm1 \n\t" + "pfadd (%4, %0), %%mm0 \n\t" + "pfadd 8(%4, %0), %%mm1 \n\t" + "movq %%mm0, (%1, %0) \n\t" + "movq %%mm1, 8(%1, %0) \n\t" + "sub $16, %0 \n\t" + "jge 1b \n\t" + : "+r"(i) + : "r"(dst), "r"(src0), "r"(src1), "r"(src2) + : "memory" + ); + __asm__ volatile ("femms"); +} + +static void vector_fmul_add_sse(float *dst, const float *src0, + const float *src1, const float *src2, int len) +{ + x86_reg i = (len - 8) * 4; + __asm__ volatile ( + "1: \n\t" + "movaps (%2, %0), %%xmm0 \n\t" + "movaps 16(%2, %0), %%xmm1 \n\t" + "mulps (%3, %0), %%xmm0 \n\t" + "mulps 16(%3, %0), %%xmm1 \n\t" + "addps (%4, %0), %%xmm0 \n\t" + "addps 16(%4, %0), %%xmm1 \n\t" + "movaps %%xmm0, (%1, %0) \n\t" + "movaps %%xmm1, 16(%1, %0) \n\t" + "sub $32, %0 \n\t" + "jge 1b \n\t" + : "+r"(i) + : "r"(dst), "r"(src0), "r"(src1), "r"(src2) + : "memory" ); } #if HAVE_6REGS -static void vector_fmul_window_3dnow2(float *dst, const float *src0, const float *src1, - const float *win, int len){ - x86_reg i = -len*4; - x86_reg j = len*4-8; - __asm__ volatile( - "1: \n" - "pswapd (%5,%1), %%mm1 \n" - "movq (%5,%0), %%mm0 \n" - "pswapd (%4,%1), %%mm5 \n" - "movq (%3,%0), %%mm4 \n" - "movq %%mm0, %%mm2 \n" - "movq %%mm1, %%mm3 \n" - "pfmul %%mm4, %%mm2 \n" // src0[len+i]*win[len+i] - "pfmul %%mm5, %%mm3 \n" // src1[ j]*win[len+j] - "pfmul %%mm4, %%mm1 \n" // src0[len+i]*win[len+j] - "pfmul %%mm5, %%mm0 \n" // src1[ j]*win[len+i] - "pfadd %%mm3, %%mm2 \n" - "pfsub %%mm0, %%mm1 \n" - "pswapd %%mm2, %%mm2 \n" - "movq %%mm1, (%2,%0) \n" - "movq %%mm2, (%2,%1) \n" - "sub $8, %1 \n" - "add $8, %0 \n" - "jl 1b \n" - "femms \n" - :"+r"(i), "+r"(j) - :"r"(dst+len), "r"(src0+len), "r"(src1), "r"(win+len) +static void vector_fmul_window_3dnow2(float *dst, const float *src0, + const float *src1, const float *win, + int len) +{ + x86_reg i = -len * 4; + x86_reg j = len * 4 - 8; + __asm__ volatile ( + "1: \n" + "pswapd (%5, %1), %%mm1 \n" + "movq (%5, %0), %%mm0 \n" + "pswapd (%4, %1), %%mm5 \n" + "movq (%3, %0), %%mm4 \n" + "movq %%mm0, %%mm2 \n" + "movq %%mm1, %%mm3 \n" + "pfmul %%mm4, %%mm2 \n" // src0[len + i] * win[len + i] + "pfmul %%mm5, %%mm3 \n" // src1[j] * win[len + j] + "pfmul %%mm4, %%mm1 \n" // src0[len + i] * win[len + j] + "pfmul %%mm5, %%mm0 \n" // src1[j] * win[len + i] + "pfadd %%mm3, %%mm2 \n" + "pfsub %%mm0, %%mm1 \n" + "pswapd %%mm2, %%mm2 \n" + "movq %%mm1, (%2, %0) \n" + "movq %%mm2, (%2, %1) \n" + "sub $8, %1 \n" + "add $8, %0 \n" + "jl 1b \n" + "femms \n" + : "+r"(i), "+r"(j) + : "r"(dst + len), "r"(src0 + len), "r"(src1), "r"(win + len) ); } -static void vector_fmul_window_sse(float *dst, const float *src0, const float *src1, - const float *win, int len){ - x86_reg i = -len*4; - x86_reg j = len*4-16; - __asm__ volatile( - "1: \n" - "movaps (%5,%1), %%xmm1 \n" - "movaps (%5,%0), %%xmm0 \n" - "movaps (%4,%1), %%xmm5 \n" - "movaps (%3,%0), %%xmm4 \n" - "shufps $0x1b, %%xmm1, %%xmm1 \n" - "shufps $0x1b, %%xmm5, %%xmm5 \n" - "movaps %%xmm0, %%xmm2 \n" - "movaps %%xmm1, %%xmm3 \n" - "mulps %%xmm4, %%xmm2 \n" // src0[len+i]*win[len+i] - "mulps %%xmm5, %%xmm3 \n" // src1[ j]*win[len+j] - "mulps %%xmm4, %%xmm1 \n" // src0[len+i]*win[len+j] - "mulps %%xmm5, %%xmm0 \n" // src1[ j]*win[len+i] - "addps %%xmm3, %%xmm2 \n" - "subps %%xmm0, %%xmm1 \n" - "shufps $0x1b, %%xmm2, %%xmm2 \n" - "movaps %%xmm1, (%2,%0) \n" - "movaps %%xmm2, (%2,%1) \n" - "sub $16, %1 \n" - "add $16, %0 \n" - "jl 1b \n" - :"+r"(i), "+r"(j) - :"r"(dst+len), "r"(src0+len), "r"(src1), "r"(win+len) +static void vector_fmul_window_sse(float *dst, const float *src0, + const float *src1, const float *win, int len) +{ + x86_reg i = -len * 4; + x86_reg j = len * 4 - 16; + __asm__ volatile ( + "1: \n" + "movaps (%5, %1), %%xmm1 \n" + "movaps (%5, %0), %%xmm0 \n" + "movaps (%4, %1), %%xmm5 \n" + "movaps (%3, %0), %%xmm4 \n" + "shufps $0x1b, %%xmm1, %%xmm1 \n" + "shufps $0x1b, %%xmm5, %%xmm5 \n" + "movaps %%xmm0, %%xmm2 \n" + "movaps %%xmm1, %%xmm3 \n" + "mulps %%xmm4, %%xmm2 \n" // src0[len + i] * win[len + i] + "mulps %%xmm5, %%xmm3 \n" // src1[j] * win[len + j] + "mulps %%xmm4, %%xmm1 \n" // src0[len + i] * win[len + j] + "mulps %%xmm5, %%xmm0 \n" // src1[j] * win[len + i] + "addps %%xmm3, %%xmm2 \n" + "subps %%xmm0, %%xmm1 \n" + "shufps $0x1b, %%xmm2, %%xmm2 \n" + "movaps %%xmm1, (%2, %0) \n" + "movaps %%xmm2, (%2, %1) \n" + "sub $16, %1 \n" + "add $16, %0 \n" + "jl 1b \n" + : "+r"(i), "+r"(j) + : "r"(dst + len), "r"(src0 + len), "r"(src1), "r"(win + len) ); } #endif /* HAVE_6REGS */ -static void vector_clipf_sse(float *dst, const float *src, float min, float max, - int len) +static void vector_clipf_sse(float *dst, const float *src, + float min, float max, int len) { - x86_reg i = (len-16)*4; - __asm__ volatile( - "movss %3, %%xmm4 \n" - "movss %4, %%xmm5 \n" - "shufps $0, %%xmm4, %%xmm4 \n" - "shufps $0, %%xmm5, %%xmm5 \n" - "1: \n\t" - "movaps (%2,%0), %%xmm0 \n\t" // 3/1 on intel - "movaps 16(%2,%0), %%xmm1 \n\t" - "movaps 32(%2,%0), %%xmm2 \n\t" - "movaps 48(%2,%0), %%xmm3 \n\t" - "maxps %%xmm4, %%xmm0 \n\t" - "maxps %%xmm4, %%xmm1 \n\t" - "maxps %%xmm4, %%xmm2 \n\t" - "maxps %%xmm4, %%xmm3 \n\t" - "minps %%xmm5, %%xmm0 \n\t" - "minps %%xmm5, %%xmm1 \n\t" - "minps %%xmm5, %%xmm2 \n\t" - "minps %%xmm5, %%xmm3 \n\t" - "movaps %%xmm0, (%1,%0) \n\t" - "movaps %%xmm1, 16(%1,%0) \n\t" - "movaps %%xmm2, 32(%1,%0) \n\t" - "movaps %%xmm3, 48(%1,%0) \n\t" - "sub $64, %0 \n\t" - "jge 1b \n\t" - :"+&r"(i) - :"r"(dst), "r"(src), "m"(min), "m"(max) - :"memory" + x86_reg i = (len - 16) * 4; + __asm__ volatile ( + "movss %3, %%xmm4 \n\t" + "movss %4, %%xmm5 \n\t" + "shufps $0, %%xmm4, %%xmm4 \n\t" + "shufps $0, %%xmm5, %%xmm5 \n\t" + "1: \n\t" + "movaps (%2, %0), %%xmm0 \n\t" // 3/1 on intel + "movaps 16(%2, %0), %%xmm1 \n\t" + "movaps 32(%2, %0), %%xmm2 \n\t" + "movaps 48(%2, %0), %%xmm3 \n\t" + "maxps %%xmm4, %%xmm0 \n\t" + "maxps %%xmm4, %%xmm1 \n\t" + "maxps %%xmm4, %%xmm2 \n\t" + "maxps %%xmm4, %%xmm3 \n\t" + "minps %%xmm5, %%xmm0 \n\t" + "minps %%xmm5, %%xmm1 \n\t" + "minps %%xmm5, %%xmm2 \n\t" + "minps %%xmm5, %%xmm3 \n\t" + "movaps %%xmm0, (%1, %0) \n\t" + "movaps %%xmm1, 16(%1, %0) \n\t" + "movaps %%xmm2, 32(%1, %0) \n\t" + "movaps %%xmm3, 48(%1, %0) \n\t" + "sub $64, %0 \n\t" + "jge 1b \n\t" + : "+&r"(i) + : "r"(dst), "r"(src), "m"(min), "m"(max) + : "memory" ); } @@ -2364,7 +2644,8 @@ void ff_vp3_idct_mmx(int16_t *input_data); void ff_vp3_idct_put_mmx(uint8_t *dest, int line_size, DCTELEM *block); void ff_vp3_idct_add_mmx(uint8_t *dest, int line_size, DCTELEM *block); -void ff_vp3_idct_dc_add_mmx2(uint8_t *dest, int line_size, const DCTELEM *block); +void ff_vp3_idct_dc_add_mmx2(uint8_t *dest, int line_size, + const DCTELEM *block); void ff_vp3_v_loop_filter_mmx2(uint8_t *src, int stride, int *bounding_values); void ff_vp3_h_loop_filter_mmx2(uint8_t *src, int stride, int *bounding_values); @@ -2373,11 +2654,19 @@ void ff_vp3_idct_sse2(int16_t *input_data); void ff_vp3_idct_put_sse2(uint8_t *dest, int line_size, DCTELEM *block); void ff_vp3_idct_add_sse2(uint8_t *dest, int line_size, DCTELEM *block); -int32_t ff_scalarproduct_int16_mmx2(const int16_t *v1, const int16_t *v2, int order, int shift); -int32_t ff_scalarproduct_int16_sse2(const int16_t *v1, const int16_t *v2, int order, int shift); -int32_t ff_scalarproduct_and_madd_int16_mmx2(int16_t *v1, const int16_t *v2, const int16_t *v3, int order, int mul); -int32_t ff_scalarproduct_and_madd_int16_sse2(int16_t *v1, const int16_t *v2, const int16_t *v3, int order, int mul); -int32_t ff_scalarproduct_and_madd_int16_ssse3(int16_t *v1, const int16_t *v2, const int16_t *v3, int order, int mul); +int32_t ff_scalarproduct_int16_mmx2(const int16_t *v1, const int16_t *v2, + int order, int shift); +int32_t ff_scalarproduct_int16_sse2(const int16_t *v1, const int16_t *v2, + int order, int shift); +int32_t ff_scalarproduct_and_madd_int16_mmx2(int16_t *v1, const int16_t *v2, + const int16_t *v3, + int order, int mul); +int32_t ff_scalarproduct_and_madd_int16_sse2(int16_t *v1, const int16_t *v2, + const int16_t *v3, + int order, int mul); +int32_t ff_scalarproduct_and_madd_int16_ssse3(int16_t *v1, const int16_t *v2, + const int16_t *v3, + int order, int mul); void ff_apply_window_int16_mmxext (int16_t *output, const int16_t *input, const int16_t *window, unsigned int len); @@ -2395,27 +2684,32 @@ void ff_apply_window_int16_ssse3_atom(int16_t *output, const int16_t *input, void ff_bswap32_buf_ssse3(uint32_t *dst, const uint32_t *src, int w); void ff_bswap32_buf_sse2(uint32_t *dst, const uint32_t *src, int w); -void ff_add_hfyu_median_prediction_mmx2(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top); -int ff_add_hfyu_left_prediction_ssse3(uint8_t *dst, const uint8_t *src, int w, int left); -int ff_add_hfyu_left_prediction_sse4(uint8_t *dst, const uint8_t *src, int w, int left); +void ff_add_hfyu_median_prediction_mmx2(uint8_t *dst, const uint8_t *top, + const uint8_t *diff, int w, + int *left, int *left_top); +int ff_add_hfyu_left_prediction_ssse3(uint8_t *dst, const uint8_t *src, + int w, int left); +int ff_add_hfyu_left_prediction_sse4(uint8_t *dst, const uint8_t *src, + int w, int left); float ff_scalarproduct_float_sse(const float *v1, const float *v2, int order); -void ff_vector_clip_int32_mmx (int32_t *dst, const int32_t *src, int32_t min, - int32_t max, unsigned int len); -void ff_vector_clip_int32_sse2 (int32_t *dst, const int32_t *src, int32_t min, - int32_t max, unsigned int len); -void ff_vector_clip_int32_int_sse2(int32_t *dst, const int32_t *src, int32_t min, - int32_t max, unsigned int len); -void ff_vector_clip_int32_sse4 (int32_t *dst, const int32_t *src, int32_t min, - int32_t max, unsigned int len); +void ff_vector_clip_int32_mmx (int32_t *dst, const int32_t *src, + int32_t min, int32_t max, unsigned int len); +void ff_vector_clip_int32_sse2 (int32_t *dst, const int32_t *src, + int32_t min, int32_t max, unsigned int len); +void ff_vector_clip_int32_int_sse2(int32_t *dst, const int32_t *src, + int32_t min, int32_t max, unsigned int len); +void ff_vector_clip_int32_sse4 (int32_t *dst, const int32_t *src, + int32_t min, int32_t max, unsigned int len); extern void ff_butterflies_float_interleave_sse(float *dst, const float *src0, const float *src1, int len); extern void ff_butterflies_float_interleave_avx(float *dst, const float *src0, const float *src1, int len); -#define SET_QPEL_FUNCS(PFX, IDX, SIZE, CPU, PREFIX) \ +#define SET_QPEL_FUNCS(PFX, IDX, SIZE, CPU, PREFIX) \ + do { \ c->PFX ## _pixels_tab[IDX][ 0] = PREFIX ## PFX ## SIZE ## _mc00_ ## CPU; \ c->PFX ## _pixels_tab[IDX][ 1] = PREFIX ## PFX ## SIZE ## _mc10_ ## CPU; \ c->PFX ## _pixels_tab[IDX][ 2] = PREFIX ## PFX ## SIZE ## _mc20_ ## CPU; \ @@ -2431,25 +2725,32 @@ extern void ff_butterflies_float_interleave_avx(float *dst, const float *src0, c->PFX ## _pixels_tab[IDX][12] = PREFIX ## PFX ## SIZE ## _mc03_ ## CPU; \ c->PFX ## _pixels_tab[IDX][13] = PREFIX ## PFX ## SIZE ## _mc13_ ## CPU; \ c->PFX ## _pixels_tab[IDX][14] = PREFIX ## PFX ## SIZE ## _mc23_ ## CPU; \ - c->PFX ## _pixels_tab[IDX][15] = PREFIX ## PFX ## SIZE ## _mc33_ ## CPU + c->PFX ## _pixels_tab[IDX][15] = PREFIX ## PFX ## SIZE ## _mc33_ ## CPU; \ + } while (0) -#define SET_HPEL_FUNCS(PFX, IDX, SIZE, CPU) \ - c->PFX ## _pixels_tab[IDX][0] = PFX ## _pixels ## SIZE ## _ ## CPU; \ - c->PFX ## _pixels_tab[IDX][1] = PFX ## _pixels ## SIZE ## _x2_ ## CPU; \ - c->PFX ## _pixels_tab[IDX][2] = PFX ## _pixels ## SIZE ## _y2_ ## CPU; \ - c->PFX ## _pixels_tab[IDX][3] = PFX ## _pixels ## SIZE ## _xy2_ ## CPU +#define SET_HPEL_FUNCS(PFX, IDX, SIZE, CPU) \ + do { \ + c->PFX ## _pixels_tab[IDX][0] = PFX ## _pixels ## SIZE ## _ ## CPU; \ + c->PFX ## _pixels_tab[IDX][1] = PFX ## _pixels ## SIZE ## _x2_ ## CPU; \ + c->PFX ## _pixels_tab[IDX][2] = PFX ## _pixels ## SIZE ## _y2_ ## CPU; \ + c->PFX ## _pixels_tab[IDX][3] = PFX ## _pixels ## SIZE ## _xy2_ ## CPU; \ + } while (0) -#define H264_QPEL_FUNCS(x, y, CPU) \ - c->put_h264_qpel_pixels_tab[0][x+y*4] = put_h264_qpel16_mc##x##y##_##CPU; \ - c->put_h264_qpel_pixels_tab[1][x+y*4] = put_h264_qpel8_mc##x##y##_##CPU; \ - c->avg_h264_qpel_pixels_tab[0][x+y*4] = avg_h264_qpel16_mc##x##y##_##CPU; \ - c->avg_h264_qpel_pixels_tab[1][x+y*4] = avg_h264_qpel8_mc##x##y##_##CPU +#define H264_QPEL_FUNCS(x, y, CPU) \ + do { \ + c->put_h264_qpel_pixels_tab[0][x + y * 4] = put_h264_qpel16_mc ## x ## y ## _ ## CPU; \ + c->put_h264_qpel_pixels_tab[1][x + y * 4] = put_h264_qpel8_mc ## x ## y ## _ ## CPU; \ + c->avg_h264_qpel_pixels_tab[0][x + y * 4] = avg_h264_qpel16_mc ## x ## y ## _ ## CPU; \ + c->avg_h264_qpel_pixels_tab[1][x + y * 4] = avg_h264_qpel8_mc ## x ## y ## _ ## CPU; \ + } while (0) -#define H264_QPEL_FUNCS_10(x, y, CPU) \ - c->put_h264_qpel_pixels_tab[0][x+y*4] = ff_put_h264_qpel16_mc##x##y##_10_##CPU; \ - c->put_h264_qpel_pixels_tab[1][x+y*4] = ff_put_h264_qpel8_mc##x##y##_10_##CPU; \ - c->avg_h264_qpel_pixels_tab[0][x+y*4] = ff_avg_h264_qpel16_mc##x##y##_10_##CPU; \ - c->avg_h264_qpel_pixels_tab[1][x+y*4] = ff_avg_h264_qpel8_mc##x##y##_10_##CPU; +#define H264_QPEL_FUNCS_10(x, y, CPU) \ + do { \ + c->put_h264_qpel_pixels_tab[0][x + y * 4] = ff_put_h264_qpel16_mc ## x ## y ## _10_ ## CPU; \ + c->put_h264_qpel_pixels_tab[1][x + y * 4] = ff_put_h264_qpel8_mc ## x ## y ## _10_ ## CPU; \ + c->avg_h264_qpel_pixels_tab[0][x + y * 4] = ff_avg_h264_qpel16_mc ## x ## y ## _10_ ## CPU; \ + c->avg_h264_qpel_pixels_tab[1][x + y * 4] = ff_avg_h264_qpel8_mc ## x ## y ## _10_ ## CPU; \ + } while (0) static void dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx, int mm_flags) { @@ -2464,18 +2765,18 @@ static void dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx, int mm_flags) c->clear_blocks = clear_blocks_mmx; c->draw_edges = draw_edges_mmx; - SET_HPEL_FUNCS(put, 0, 16, mmx); + SET_HPEL_FUNCS(put, 0, 16, mmx); SET_HPEL_FUNCS(put_no_rnd, 0, 16, mmx); - SET_HPEL_FUNCS(avg, 0, 16, mmx); + SET_HPEL_FUNCS(avg, 0, 16, mmx); SET_HPEL_FUNCS(avg_no_rnd, 0, 16, mmx); - SET_HPEL_FUNCS(put, 1, 8, mmx); - SET_HPEL_FUNCS(put_no_rnd, 1, 8, mmx); - SET_HPEL_FUNCS(avg, 1, 8, mmx); - SET_HPEL_FUNCS(avg_no_rnd, 1, 8, mmx); + SET_HPEL_FUNCS(put, 1, 8, mmx); + SET_HPEL_FUNCS(put_no_rnd, 1, 8, mmx); + SET_HPEL_FUNCS(avg, 1, 8, mmx); + SET_HPEL_FUNCS(avg_no_rnd, 1, 8, mmx); } #if ARCH_X86_32 || !HAVE_YASM - c->gmc= gmc_mmx; + c->gmc = gmc_mmx; #endif #if ARCH_X86_32 && HAVE_YASM if (!high_bit_depth) @@ -2543,47 +2844,48 @@ static void dsputil_init_mmx2(DSPContext *c, AVCodecContext *avctx, c->vp3_h_loop_filter = ff_vp3_h_loop_filter_mmx2; } } - if (CONFIG_VP3_DECODER && HAVE_YASM) { + if (CONFIG_VP3_DECODER && HAVE_YASM) c->vp3_idct_dc_add = ff_vp3_idct_dc_add_mmx2; - } - if (CONFIG_VP3_DECODER - && (avctx->codec_id == CODEC_ID_VP3 || avctx->codec_id == CODEC_ID_THEORA)) { + if (CONFIG_VP3_DECODER && (avctx->codec_id == CODEC_ID_VP3 || + avctx->codec_id == CODEC_ID_THEORA)) { c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_exact_mmx2; c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_exact_mmx2; } - SET_QPEL_FUNCS(put_qpel, 0, 16, mmx2, ); - SET_QPEL_FUNCS(put_qpel, 1, 8, mmx2, ); - SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, mmx2, ); - SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, mmx2, ); - SET_QPEL_FUNCS(avg_qpel, 0, 16, mmx2, ); - SET_QPEL_FUNCS(avg_qpel, 1, 8, mmx2, ); + if (CONFIG_H264QPEL) { + SET_QPEL_FUNCS(put_qpel, 0, 16, mmx2, ); + SET_QPEL_FUNCS(put_qpel, 1, 8, mmx2, ); + SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, mmx2, ); + SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, mmx2, ); + SET_QPEL_FUNCS(avg_qpel, 0, 16, mmx2, ); + SET_QPEL_FUNCS(avg_qpel, 1, 8, mmx2, ); - if (!high_bit_depth) { - SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmx2, ); - SET_QPEL_FUNCS(put_h264_qpel, 1, 8, mmx2, ); - SET_QPEL_FUNCS(put_h264_qpel, 2, 4, mmx2, ); - SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, mmx2, ); - SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, mmx2, ); - SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, mmx2, ); - } else if (bit_depth == 10) { + if (!high_bit_depth) { + SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmx2, ); + SET_QPEL_FUNCS(put_h264_qpel, 1, 8, mmx2, ); + SET_QPEL_FUNCS(put_h264_qpel, 2, 4, mmx2, ); + SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, mmx2, ); + SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, mmx2, ); + SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, mmx2, ); + } else if (bit_depth == 10) { #if HAVE_YASM #if !ARCH_X86_64 - SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_mmxext, ff_); - SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_mmxext, ff_); - SET_QPEL_FUNCS(put_h264_qpel, 1, 8, 10_mmxext, ff_); - SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, 10_mmxext, ff_); + SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_mmxext, ff_); + SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_mmxext, ff_); + SET_QPEL_FUNCS(put_h264_qpel, 1, 8, 10_mmxext, ff_); + SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, 10_mmxext, ff_); #endif - SET_QPEL_FUNCS(put_h264_qpel, 2, 4, 10_mmxext, ff_); - SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, 10_mmxext, ff_); + SET_QPEL_FUNCS(put_h264_qpel, 2, 4, 10_mmxext, ff_); + SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, 10_mmxext, ff_); #endif - } + } - SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, mmx2, ); - SET_QPEL_FUNCS(put_2tap_qpel, 1, 8, mmx2, ); - SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, mmx2, ); - SET_QPEL_FUNCS(avg_2tap_qpel, 1, 8, mmx2, ); + SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, mmx2, ); + SET_QPEL_FUNCS(put_2tap_qpel, 1, 8, mmx2, ); + SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, mmx2, ); + SET_QPEL_FUNCS(avg_2tap_qpel, 1, 8, mmx2, ); + } #if HAVE_YASM if (!high_bit_depth && CONFIG_H264CHROMA) { @@ -2599,7 +2901,7 @@ static void dsputil_init_mmx2(DSPContext *c, AVCodecContext *avctx, c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_10_mmxext; } - c->add_hfyu_median_prediction = ff_add_hfyu_median_prediction_mmx2; + c->add_hfyu_median_prediction = ff_add_hfyu_median_prediction_mmx2; c->scalarproduct_int16 = ff_scalarproduct_int16_mmx2; c->scalarproduct_and_madd_int16 = ff_scalarproduct_and_madd_int16_mmx2; @@ -2645,33 +2947,35 @@ static void dsputil_init_3dnow(DSPContext *c, AVCodecContext *avctx, } } - if (CONFIG_VP3_DECODER - && (avctx->codec_id == CODEC_ID_VP3 || avctx->codec_id == CODEC_ID_THEORA)) { + if (CONFIG_VP3_DECODER && (avctx->codec_id == CODEC_ID_VP3 || + avctx->codec_id == CODEC_ID_THEORA)) { c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_exact_3dnow; c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_exact_3dnow; } - SET_QPEL_FUNCS(put_qpel, 0, 16, 3dnow, ); - SET_QPEL_FUNCS(put_qpel, 1, 8, 3dnow, ); - SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, 3dnow, ); - SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, 3dnow, ); - SET_QPEL_FUNCS(avg_qpel, 0, 16, 3dnow, ); - SET_QPEL_FUNCS(avg_qpel, 1, 8, 3dnow, ); + if (CONFIG_H264QPEL) { + SET_QPEL_FUNCS(put_qpel, 0, 16, 3dnow, ); + SET_QPEL_FUNCS(put_qpel, 1, 8, 3dnow, ); + SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, 3dnow, ); + SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, 3dnow, ); + SET_QPEL_FUNCS(avg_qpel, 0, 16, 3dnow, ); + SET_QPEL_FUNCS(avg_qpel, 1, 8, 3dnow, ); - if (!high_bit_depth) { - SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 3dnow, ); - SET_QPEL_FUNCS(put_h264_qpel, 1, 8, 3dnow, ); - SET_QPEL_FUNCS(put_h264_qpel, 2, 4, 3dnow, ); - SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 3dnow, ); - SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, 3dnow, ); - SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, 3dnow, ); + if (!high_bit_depth) { + SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 3dnow, ); + SET_QPEL_FUNCS(put_h264_qpel, 1, 8, 3dnow, ); + SET_QPEL_FUNCS(put_h264_qpel, 2, 4, 3dnow, ); + SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 3dnow, ); + SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, 3dnow, ); + SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, 3dnow, ); + } + + SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, 3dnow, ); + SET_QPEL_FUNCS(put_2tap_qpel, 1, 8, 3dnow, ); + SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, 3dnow, ); + SET_QPEL_FUNCS(avg_2tap_qpel, 1, 8, 3dnow, ); } - SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, 3dnow, ); - SET_QPEL_FUNCS(put_2tap_qpel, 1, 8, 3dnow, ); - SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, 3dnow, ); - SET_QPEL_FUNCS(avg_2tap_qpel, 1, 8, 3dnow, ); - #if HAVE_YASM if (!high_bit_depth && CONFIG_H264CHROMA) { c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_3dnow_rnd; @@ -2702,7 +3006,7 @@ static void dsputil_init_sse(DSPContext *c, AVCodecContext *avctx, int mm_flags) const int high_bit_depth = avctx->bits_per_raw_sample > 8; if (!high_bit_depth) { - if (!(CONFIG_MPEG_XVMC_DECODER && avctx->xvmc_acceleration > 1)){ + if (!(CONFIG_MPEG_XVMC_DECODER && avctx->xvmc_acceleration > 1)) { /* XvMCCreateBlocks() may not allocate 16-byte aligned blocks */ c->clear_block = clear_block_sse; c->clear_blocks = clear_blocks_sse; @@ -2745,11 +3049,12 @@ static void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx, c->put_pixels_tab[0][0] = put_pixels16_sse2; c->put_no_rnd_pixels_tab[0][0] = put_pixels16_sse2; c->avg_pixels_tab[0][0] = avg_pixels16_sse2; - H264_QPEL_FUNCS(0, 0, sse2); + if (CONFIG_H264QPEL) + H264_QPEL_FUNCS(0, 0, sse2); } } - if (!high_bit_depth) { + if (!high_bit_depth && CONFIG_H264QPEL) { H264_QPEL_FUNCS(0, 1, sse2); H264_QPEL_FUNCS(0, 2, sse2); H264_QPEL_FUNCS(0, 3, sse2); @@ -2766,14 +3071,15 @@ static void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx, #if HAVE_YASM if (bit_depth == 10) { - SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_sse2, ff_); - SET_QPEL_FUNCS(put_h264_qpel, 1, 8, 10_sse2, ff_); - SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_sse2, ff_); - SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, 10_sse2, ff_); - H264_QPEL_FUNCS_10(1, 0, sse2_cache64); - H264_QPEL_FUNCS_10(2, 0, sse2_cache64); - H264_QPEL_FUNCS_10(3, 0, sse2_cache64); - + if (CONFIG_H264QPEL) { + SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_sse2, ff_); + SET_QPEL_FUNCS(put_h264_qpel, 1, 8, 10_sse2, ff_); + SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_sse2, ff_); + SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, 10_sse2, ff_); + H264_QPEL_FUNCS_10(1, 0, sse2_cache64); + H264_QPEL_FUNCS_10(2, 0, sse2_cache64); + H264_QPEL_FUNCS_10(3, 0, sse2_cache64); + } if (CONFIG_H264CHROMA) { c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_10_sse2; c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_10_sse2; @@ -2789,7 +3095,7 @@ static void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx, } if (avctx->flags & CODEC_FLAG_BITEXACT) { c->apply_window_int16 = ff_apply_window_int16_sse2_ba; - } else if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) { + } else if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) { c->apply_window_int16 = ff_apply_window_int16_sse2; } c->bswap_buf = ff_bswap32_buf_sse2; @@ -2803,7 +3109,7 @@ static void dsputil_init_ssse3(DSPContext *c, AVCodecContext *avctx, const int high_bit_depth = avctx->bits_per_raw_sample > 8; const int bit_depth = avctx->bits_per_raw_sample; - if (!high_bit_depth) { + if (!high_bit_depth && CONFIG_H264QPEL) { H264_QPEL_FUNCS(1, 0, ssse3); H264_QPEL_FUNCS(1, 1, ssse3); H264_QPEL_FUNCS(1, 2, ssse3); @@ -2818,7 +3124,7 @@ static void dsputil_init_ssse3(DSPContext *c, AVCodecContext *avctx, H264_QPEL_FUNCS(3, 3, ssse3); } #if HAVE_YASM - else if (bit_depth == 10) { + else if (bit_depth == 10 && CONFIG_H264QPEL) { H264_QPEL_FUNCS_10(1, 0, ssse3_cache64); H264_QPEL_FUNCS_10(2, 0, ssse3_cache64); H264_QPEL_FUNCS_10(3, 0, ssse3_cache64); @@ -2833,14 +3139,12 @@ static void dsputil_init_ssse3(DSPContext *c, AVCodecContext *avctx, if (mm_flags & AV_CPU_FLAG_SSE4) // not really sse4, just slow on Conroe c->add_hfyu_left_prediction = ff_add_hfyu_left_prediction_sse4; - if (mm_flags & AV_CPU_FLAG_ATOM) { + if (mm_flags & AV_CPU_FLAG_ATOM) c->apply_window_int16 = ff_apply_window_int16_ssse3_atom; - } else { + else c->apply_window_int16 = ff_apply_window_int16_ssse3; - } - if (!(mm_flags & (AV_CPU_FLAG_SSE42|AV_CPU_FLAG_3DNOW))) { // cachesplit + if (!(mm_flags & (AV_CPU_FLAG_SSE42|AV_CPU_FLAG_3DNOW))) // cachesplit c->scalarproduct_and_madd_int16 = ff_scalarproduct_and_madd_int16_ssse3; - } c->bswap_buf = ff_bswap32_buf_ssse3; #endif #endif @@ -2862,9 +3166,11 @@ static void dsputil_init_avx(DSPContext *c, AVCodecContext *avctx, int mm_flags) if (bit_depth == 10) { // AVX implies !cache64. // TODO: Port cache(32|64) detection from x264. - H264_QPEL_FUNCS_10(1, 0, sse2); - H264_QPEL_FUNCS_10(2, 0, sse2); - H264_QPEL_FUNCS_10(3, 0, sse2); + if (CONFIG_H264QPEL) { + H264_QPEL_FUNCS_10(1, 0, sse2); + H264_QPEL_FUNCS_10(2, 0, sse2); + H264_QPEL_FUNCS_10(3, 0, sse2); + } if (CONFIG_H264CHROMA) { c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_10_avx; @@ -2875,13 +3181,13 @@ static void dsputil_init_avx(DSPContext *c, AVCodecContext *avctx, int mm_flags) #endif } -void ff_dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) +void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx) { int mm_flags = av_get_cpu_flags(); if (avctx->dsp_mask) { if (avctx->dsp_mask & AV_CPU_FLAG_FORCE) - mm_flags |= (avctx->dsp_mask & 0xffff); + mm_flags |= avctx->dsp_mask & 0xffff; else mm_flags &= ~(avctx->dsp_mask & 0xffff); } @@ -2902,56 +3208,57 @@ void ff_dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) #endif if (mm_flags & AV_CPU_FLAG_MMX) { - const int idct_algo= avctx->idct_algo; + const int idct_algo = avctx->idct_algo; if (avctx->lowres == 0 && avctx->bits_per_raw_sample <= 8) { - if(idct_algo==FF_IDCT_AUTO || idct_algo==FF_IDCT_SIMPLEMMX){ - c->idct_put= ff_simple_idct_put_mmx; - c->idct_add= ff_simple_idct_add_mmx; - c->idct = ff_simple_idct_mmx; - c->idct_permutation_type= FF_SIMPLE_IDCT_PERM; + if (idct_algo == FF_IDCT_AUTO || idct_algo == FF_IDCT_SIMPLEMMX) { + c->idct_put = ff_simple_idct_put_mmx; + c->idct_add = ff_simple_idct_add_mmx; + c->idct = ff_simple_idct_mmx; + c->idct_permutation_type = FF_SIMPLE_IDCT_PERM; #if CONFIG_GPL - }else if(idct_algo==FF_IDCT_LIBMPEG2MMX){ - if(mm_flags & AV_CPU_FLAG_MMX2){ - c->idct_put= ff_libmpeg2mmx2_idct_put; - c->idct_add= ff_libmpeg2mmx2_idct_add; - c->idct = ff_mmxext_idct; - }else{ - c->idct_put= ff_libmpeg2mmx_idct_put; - c->idct_add= ff_libmpeg2mmx_idct_add; - c->idct = ff_mmx_idct; + } else if (idct_algo == FF_IDCT_LIBMPEG2MMX) { + if (mm_flags & AV_CPU_FLAG_MMX2) { + c->idct_put = ff_libmpeg2mmx2_idct_put; + c->idct_add = ff_libmpeg2mmx2_idct_add; + c->idct = ff_mmxext_idct; + } else { + c->idct_put = ff_libmpeg2mmx_idct_put; + c->idct_add = ff_libmpeg2mmx_idct_add; + c->idct = ff_mmx_idct; } - c->idct_permutation_type= FF_LIBMPEG2_IDCT_PERM; + c->idct_permutation_type = FF_LIBMPEG2_IDCT_PERM; #endif - }else if((CONFIG_VP3_DECODER || CONFIG_VP5_DECODER || CONFIG_VP6_DECODER) && - idct_algo==FF_IDCT_VP3 && HAVE_YASM){ - if(mm_flags & AV_CPU_FLAG_SSE2){ - c->idct_put= ff_vp3_idct_put_sse2; - c->idct_add= ff_vp3_idct_add_sse2; - c->idct = ff_vp3_idct_sse2; - c->idct_permutation_type= FF_TRANSPOSE_IDCT_PERM; - }else{ - c->idct_put= ff_vp3_idct_put_mmx; - c->idct_add= ff_vp3_idct_add_mmx; - c->idct = ff_vp3_idct_mmx; - c->idct_permutation_type= FF_PARTTRANS_IDCT_PERM; + } else if ((CONFIG_VP3_DECODER || CONFIG_VP5_DECODER || + CONFIG_VP6_DECODER) && + idct_algo == FF_IDCT_VP3 && HAVE_YASM) { + if (mm_flags & AV_CPU_FLAG_SSE2) { + c->idct_put = ff_vp3_idct_put_sse2; + c->idct_add = ff_vp3_idct_add_sse2; + c->idct = ff_vp3_idct_sse2; + c->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM; + } else { + c->idct_put = ff_vp3_idct_put_mmx; + c->idct_add = ff_vp3_idct_add_mmx; + c->idct = ff_vp3_idct_mmx; + c->idct_permutation_type = FF_PARTTRANS_IDCT_PERM; } - }else if(idct_algo==FF_IDCT_CAVS){ - c->idct_permutation_type= FF_TRANSPOSE_IDCT_PERM; - }else if(idct_algo==FF_IDCT_XVIDMMX){ - if(mm_flags & AV_CPU_FLAG_SSE2){ - c->idct_put= ff_idct_xvid_sse2_put; - c->idct_add= ff_idct_xvid_sse2_add; - c->idct = ff_idct_xvid_sse2; - c->idct_permutation_type= FF_SSE2_IDCT_PERM; - }else if(mm_flags & AV_CPU_FLAG_MMX2){ - c->idct_put= ff_idct_xvid_mmx2_put; - c->idct_add= ff_idct_xvid_mmx2_add; - c->idct = ff_idct_xvid_mmx2; - }else{ - c->idct_put= ff_idct_xvid_mmx_put; - c->idct_add= ff_idct_xvid_mmx_add; - c->idct = ff_idct_xvid_mmx; + } else if (idct_algo == FF_IDCT_CAVS) { + c->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM; + } else if (idct_algo == FF_IDCT_XVIDMMX) { + if (mm_flags & AV_CPU_FLAG_SSE2) { + c->idct_put = ff_idct_xvid_sse2_put; + c->idct_add = ff_idct_xvid_sse2_add; + c->idct = ff_idct_xvid_sse2; + c->idct_permutation_type = FF_SSE2_IDCT_PERM; + } else if (mm_flags & AV_CPU_FLAG_MMX2) { + c->idct_put = ff_idct_xvid_mmx2_put; + c->idct_add = ff_idct_xvid_mmx2_add; + c->idct = ff_idct_xvid_mmx2; + } else { + c->idct_put = ff_idct_xvid_mmx_put; + c->idct_add = ff_idct_xvid_mmx_add; + c->idct = ff_idct_xvid_mmx; } } } @@ -2962,13 +3269,13 @@ void ff_dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) if (mm_flags & AV_CPU_FLAG_MMX2) dsputil_init_mmx2(c, avctx, mm_flags); - if (HAVE_AMD3DNOW && (mm_flags & AV_CPU_FLAG_3DNOW)) + if (mm_flags & AV_CPU_FLAG_3DNOW && HAVE_AMD3DNOW) dsputil_init_3dnow(c, avctx, mm_flags); - if (HAVE_AMD3DNOWEXT && (mm_flags & AV_CPU_FLAG_3DNOWEXT)) + if (mm_flags & AV_CPU_FLAG_3DNOWEXT && HAVE_AMD3DNOWEXT) dsputil_init_3dnow2(c, avctx, mm_flags); - if (HAVE_SSE && (mm_flags & AV_CPU_FLAG_SSE)) + if (mm_flags & AV_CPU_FLAG_SSE && HAVE_SSE) dsputil_init_sse(c, avctx, mm_flags); if (mm_flags & AV_CPU_FLAG_SSE2) diff --git a/libavformat/movenc.c b/libavformat/movenc.c index 9a8c38d097..f0b606e0fc 100644 --- a/libavformat/movenc.c +++ b/libavformat/movenc.c @@ -2209,14 +2209,11 @@ static int mov_write_isml_manifest(AVIOContext *pb, MOVMuxContext *mov) size); av_free(ptr); } - } else { - param_write_hex(pb, "CodecPrivateData", track->enc->extradata, - track->enc->extradata_size); - } - if (track->enc->codec_id == CODEC_ID_H264) { param_write_string(pb, "FourCC", "H264"); } else if (track->enc->codec_id == CODEC_ID_VC1) { param_write_string(pb, "FourCC", "WVC1"); + param_write_hex(pb, "CodecPrivateData", track->enc->extradata, + track->enc->extradata_size); } param_write_int(pb, "MaxWidth", track->enc->width); param_write_int(pb, "MaxHeight", track->enc->height);