diff --git a/libavcodec/armv4l/mpegvideo_arm.c b/libavcodec/armv4l/mpegvideo_arm.c index d54bb7a37e..afb7a41c9d 100644 --- a/libavcodec/armv4l/mpegvideo_arm.c +++ b/libavcodec/armv4l/mpegvideo_arm.c @@ -44,7 +44,6 @@ void MPV_common_init_armv4l(MpegEncContext *s) if(idct_algo==FF_IDCT_AUTO || idct_algo==FF_IDCT_ARM){ s->idct_put= arm_idct_put; s->idct_add= arm_idct_add; - for(i=0; i<64; i++) - s->idct_permutation[i]= i; + s->idct_permutation_type= FF_NO_IDCT_PERM; } } diff --git a/libavcodec/i386/mpegvideo_mmx.c b/libavcodec/i386/mpegvideo_mmx.c index 48d6fc5b60..8645fa3709 100644 --- a/libavcodec/i386/mpegvideo_mmx.c +++ b/libavcodec/i386/mpegvideo_mmx.c @@ -25,18 +25,6 @@ #include "../avcodec.h" #include "../simple_idct.h" -/* Input permutation for the simple_idct_mmx */ -static UINT8 simple_mmx_permutation[64]={ - 0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D, - 0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D, - 0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D, - 0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F, - 0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F, - 0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D, - 0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F, - 0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F, -}; - extern UINT8 zigzag_direct_noperm[64]; extern UINT16 inv_zigzag_direct16[64]; extern UINT32 inverse[256]; @@ -559,12 +547,11 @@ void MPV_common_init_mmx(MpegEncContext *s) s->dct_quantize= dct_quantize_MMX; } } - + if(idct_algo==FF_IDCT_AUTO || idct_algo==FF_IDCT_SIMPLEMMX){ s->idct_put= ff_simple_idct_put_mmx; s->idct_add= ff_simple_idct_add_mmx; - for(i=0; i<64; i++) - s->idct_permutation[i]= simple_mmx_permutation[i]; + s->idct_permutation_type= FF_SIMPLE_IDCT_PERM; }else if(idct_algo==FF_IDCT_LIBMPEG2MMX){ if(mm_flags & MM_MMXEXT){ s->idct_put= ff_libmpeg2mmx2_idct_put; @@ -573,8 +560,7 @@ void MPV_common_init_mmx(MpegEncContext *s) s->idct_put= ff_libmpeg2mmx_idct_put; s->idct_add= ff_libmpeg2mmx_idct_add; } - for(i=0; i<64; i++) - s->idct_permutation[i]= (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2); + s->idct_permutation_type= FF_LIBMPEG2_IDCT_PERM; } } } diff --git a/libavcodec/mlib/dsputil_mlib.c b/libavcodec/mlib/dsputil_mlib.c index 00ffe1e27e..445500c50d 100644 --- a/libavcodec/mlib/dsputil_mlib.c +++ b/libavcodec/mlib/dsputil_mlib.c @@ -260,8 +260,6 @@ void MPV_common_init_mlib(MpegEncContext *s) if(s->avctx->idct_algo==FF_IDCT_AUTO || s->avctx->idct_algo==FF_IDCT_MLIB){ s->idct_put= ff_idct_put_mlib; s->idct_add= ff_idct_add_mlib; - for(i=0; i<64; i++) - s->idct_permutation[i]= i; + s->idct_permutation_type= FF_NO_IDCT_PERM; } - } diff --git a/libavcodec/mpegvideo.c b/libavcodec/mpegvideo.c index 1bf2cb4496..2ec1ea397b 100644 --- a/libavcodec/mpegvideo.c +++ b/libavcodec/mpegvideo.c @@ -69,6 +69,18 @@ static const unsigned short aanscales[64] = { 4520, 6270, 5906, 5315, 4520, 3552, 2446, 1247 }; +/* Input permutation for the simple_idct_mmx */ +static const UINT8 simple_mmx_permutation[64]={ + 0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D, + 0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D, + 0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D, + 0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F, + 0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F, + 0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D, + 0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F, + 0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F, +}; + static UINT8 h263_chroma_roundtab[16] = { 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, }; @@ -187,13 +199,11 @@ int DCT_common_init(MpegEncContext *s) if(s->avctx->idct_algo==FF_IDCT_INT){ s->idct_put= ff_jref_idct_put; s->idct_add= ff_jref_idct_add; - for(i=0; i<64; i++) - s->idct_permutation[i]= (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2); + s->idct_permutation_type= FF_LIBMPEG2_IDCT_PERM; }else{ //accurate/default s->idct_put= simple_idct_put; s->idct_add= simple_idct_add; - for(i=0; i<64; i++) - s->idct_permutation[i]= i; + s->idct_permutation_type= FF_NO_IDCT_PERM; } #ifdef HAVE_MMX @@ -212,6 +222,28 @@ int DCT_common_init(MpegEncContext *s) MPV_common_init_armv4l(); #endif + switch(s->idct_permutation_type){ + case FF_NO_IDCT_PERM: + for(i=0; i<64; i++) + s->idct_permutation[i]= i; + break; + case FF_LIBMPEG2_IDCT_PERM: + for(i=0; i<64; i++) + s->idct_permutation[i]= (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2); + break; + case FF_SIMPLE_IDCT_PERM: + for(i=0; i<64; i++) + s->idct_permutation[i]= simple_mmx_permutation[i]; + break; + case FF_TRANSPOSE_IDCT_PERM: + for(i=0; i<64; i++) + s->idct_permutation[i]= ((i&7)<<3) | (i>>3); + break; + default: + fprintf(stderr, "Internal error, IDCT permutation not set\n"); + return -1; + } + /* load & permutate scantables note: only wmv uses differnt ones diff --git a/libavcodec/mpegvideo.h b/libavcodec/mpegvideo.h index dc5af27294..28ef946c73 100644 --- a/libavcodec/mpegvideo.h +++ b/libavcodec/mpegvideo.h @@ -298,6 +298,11 @@ typedef struct MpegEncContext { ScanTable intra_v_scantable; ScanTable inter_scantable; // if inter == intra then intra should be used to reduce tha cache usage UINT8 idct_permutation[64]; + int idct_permutation_type; +#define FF_NO_IDCT_PERM 1 +#define FF_LIBMPEG2_IDCT_PERM 2 +#define FF_SIMPLE_IDCT_PERM 3 +#define FF_TRANSPOSE_IDCT_PERM 4 void *opaque; /* private data for the user */ diff --git a/libavcodec/ps2/mpegvideo_mmi.c b/libavcodec/ps2/mpegvideo_mmi.c index f12640cb8b..322ac2c7b1 100644 --- a/libavcodec/ps2/mpegvideo_mmi.c +++ b/libavcodec/ps2/mpegvideo_mmi.c @@ -85,14 +85,12 @@ static void dct_unquantize_h263_mmi(MpegEncContext *s, void MPV_common_init_mmi(MpegEncContext *s) { int i; -// const int dct_algo = s->avctx->dct_algo; const int idct_algo= s->avctx->idct_algo; - if(idct_algo==FF_IDCT_AUTO){ + if(idct_algo==FF_IDCT_AUTO || idct_algo==FF_IDCT_PS2){ s->idct_put= ff_mmi_idct_put; s->idct_add= ff_mmi_idct_add; - for(i=0; i<64; i++) - s->idct_permutation[i]= (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2); + s->idct_permutation_type= FF_LIBMPEG2_IDCT_PERM; } s->dct_unquantize_h263 = dct_unquantize_h263_mmi; }