From ec910d81405c736a4490383a250299a7837c2e64 Mon Sep 17 00:00:00 2001 From: trav90 Date: Thu, 18 Oct 2018 21:53:44 -0500 Subject: Update aom to commit id e87fb2378f01103d5d6e477a4ef6892dc714e614 --- third_party/aom/av1/common/resize.c | 614 ++++++++++++++++++++++-------------- 1 file changed, 379 insertions(+), 235 deletions(-) (limited to 'third_party/aom/av1/common/resize.c') diff --git a/third_party/aom/av1/common/resize.c b/third_party/aom/av1/common/resize.c index 8ddca0acb..b0f303e35 100644 --- a/third_party/aom/av1/common/resize.c +++ b/third_party/aom/av1/common/resize.c @@ -32,7 +32,7 @@ #define INTERP_TAPS 8 #define SUBPEL_BITS_RS 6 #define SUBPEL_MASK_RS ((1 << SUBPEL_BITS_RS) - 1) -#define INTERP_PRECISION_BITS 32 +#define INTERP_PRECISION_BITS 16 #define SUBPEL_INTERP_EXTRA_BITS (INTERP_PRECISION_BITS - SUBPEL_BITS_RS) #define SUBPEL_INTERP_EXTRA_OFF (1 << (SUBPEL_INTERP_EXTRA_BITS - 1)) @@ -40,24 +40,6 @@ typedef int16_t interp_kernel[INTERP_TAPS]; // Filters for interpolation (0.5-band) - note this also filters integer pels. static const interp_kernel filteredinterp_filters500[(1 << SUBPEL_BITS_RS)] = { -#if SUBPEL_BITS_RS == 5 - { -3, 0, 35, 64, 35, 0, -3, 0 }, { -3, -1, 34, 64, 36, 1, -3, 0 }, - { -3, -1, 32, 64, 38, 1, -3, 0 }, { -2, -2, 31, 63, 39, 2, -3, 0 }, - { -2, -2, 29, 63, 41, 2, -3, 0 }, { -2, -2, 28, 63, 42, 3, -4, 0 }, - { -2, -3, 27, 63, 43, 4, -4, 0 }, { -2, -3, 25, 62, 45, 5, -4, 0 }, - { -2, -3, 24, 62, 46, 5, -4, 0 }, { -2, -3, 23, 61, 47, 6, -4, 0 }, - { -2, -3, 21, 60, 49, 7, -4, 0 }, { -1, -4, 20, 60, 50, 8, -4, -1 }, - { -1, -4, 19, 59, 51, 9, -4, -1 }, { -1, -4, 17, 58, 52, 10, -4, 0 }, - { -1, -4, 16, 57, 53, 12, -4, -1 }, { -1, -4, 15, 56, 54, 13, -4, -1 }, - { -1, -4, 14, 55, 55, 14, -4, -1 }, { -1, -4, 13, 54, 56, 15, -4, -1 }, - { -1, -4, 12, 53, 57, 16, -4, -1 }, { 0, -4, 10, 52, 58, 17, -4, -1 }, - { -1, -4, 9, 51, 59, 19, -4, -1 }, { -1, -4, 8, 50, 60, 20, -4, -1 }, - { 0, -4, 7, 49, 60, 21, -3, -2 }, { 0, -4, 6, 47, 61, 23, -3, -2 }, - { 0, -4, 5, 46, 62, 24, -3, -2 }, { 0, -4, 5, 45, 62, 25, -3, -2 }, - { 0, -4, 4, 43, 63, 27, -3, -2 }, { 0, -4, 3, 42, 63, 28, -2, -2 }, - { 0, -3, 2, 41, 63, 29, -2, -2 }, { 0, -3, 2, 39, 63, 31, -2, -2 }, - { 0, -3, 1, 38, 64, 32, -1, -3 }, { 0, -3, 1, 36, 64, 34, -1, -3 }, -#elif SUBPEL_BITS_RS == 6 { -3, 0, 35, 64, 35, 0, -3, 0 }, { -3, 0, 34, 64, 36, 0, -3, 0 }, { -3, -1, 34, 64, 36, 1, -3, 0 }, { -3, -1, 33, 64, 37, 1, -3, 0 }, { -3, -1, 32, 64, 38, 1, -3, 0 }, { -3, -1, 31, 64, 39, 1, -3, 0 }, @@ -90,29 +72,10 @@ static const interp_kernel filteredinterp_filters500[(1 << SUBPEL_BITS_RS)] = { { 0, -3, 2, 39, 63, 31, -1, -3 }, { 0, -3, 1, 39, 64, 31, -1, -3 }, { 0, -3, 1, 38, 64, 32, -1, -3 }, { 0, -3, 1, 37, 64, 33, -1, -3 }, { 0, -3, 1, 36, 64, 34, -1, -3 }, { 0, -3, 0, 36, 64, 34, 0, -3 }, -#endif // SUBPEL_BITS_RS == 5 }; // Filters for interpolation (0.625-band) - note this also filters integer pels. static const interp_kernel filteredinterp_filters625[(1 << SUBPEL_BITS_RS)] = { -#if SUBPEL_BITS_RS == 5 - { -1, -8, 33, 80, 33, -8, -1, 0 }, { -1, -8, 30, 80, 35, -8, -1, 1 }, - { -1, -8, 28, 80, 37, -7, -2, 1 }, { 0, -8, 26, 79, 39, -7, -2, 1 }, - { 0, -8, 24, 79, 41, -7, -2, 1 }, { 0, -8, 22, 78, 43, -6, -2, 1 }, - { 0, -8, 20, 78, 45, -5, -3, 1 }, { 0, -8, 18, 77, 48, -5, -3, 1 }, - { 0, -8, 16, 76, 50, -4, -3, 1 }, { 0, -8, 15, 75, 52, -3, -4, 1 }, - { 0, -7, 13, 74, 54, -3, -4, 1 }, { 0, -7, 11, 73, 56, -2, -4, 1 }, - { 0, -7, 10, 71, 58, -1, -4, 1 }, { 1, -7, 8, 70, 60, 0, -5, 1 }, - { 1, -6, 6, 68, 62, 1, -5, 1 }, { 1, -6, 5, 67, 63, 2, -5, 1 }, - { 1, -6, 4, 65, 65, 4, -6, 1 }, { 1, -5, 2, 63, 67, 5, -6, 1 }, - { 1, -5, 1, 62, 68, 6, -6, 1 }, { 1, -5, 0, 60, 70, 8, -7, 1 }, - { 1, -4, -1, 58, 71, 10, -7, 0 }, { 1, -4, -2, 56, 73, 11, -7, 0 }, - { 1, -4, -3, 54, 74, 13, -7, 0 }, { 1, -4, -3, 52, 75, 15, -8, 0 }, - { 1, -3, -4, 50, 76, 16, -8, 0 }, { 1, -3, -5, 48, 77, 18, -8, 0 }, - { 1, -3, -5, 45, 78, 20, -8, 0 }, { 1, -2, -6, 43, 78, 22, -8, 0 }, - { 1, -2, -7, 41, 79, 24, -8, 0 }, { 1, -2, -7, 39, 79, 26, -8, 0 }, - { 1, -2, -7, 37, 80, 28, -8, -1 }, { 1, -1, -8, 35, 80, 30, -8, -1 }, -#elif SUBPEL_BITS_RS == 6 { -1, -8, 33, 80, 33, -8, -1, 0 }, { -1, -8, 31, 80, 34, -8, -1, 1 }, { -1, -8, 30, 80, 35, -8, -1, 1 }, { -1, -8, 29, 80, 36, -7, -2, 1 }, { -1, -8, 28, 80, 37, -7, -2, 1 }, { -1, -8, 27, 80, 38, -7, -2, 1 }, @@ -145,29 +108,10 @@ static const interp_kernel filteredinterp_filters625[(1 << SUBPEL_BITS_RS)] = { { 1, -2, -7, 39, 79, 26, -8, 0 }, { 1, -2, -7, 38, 80, 27, -8, -1 }, { 1, -2, -7, 37, 80, 28, -8, -1 }, { 1, -2, -7, 36, 80, 29, -8, -1 }, { 1, -1, -8, 35, 80, 30, -8, -1 }, { 1, -1, -8, 34, 80, 31, -8, -1 }, -#endif // SUBPEL_BITS_RS == 5 }; // Filters for interpolation (0.75-band) - note this also filters integer pels. static const interp_kernel filteredinterp_filters750[(1 << SUBPEL_BITS_RS)] = { -#if SUBPEL_BITS_RS == 5 - { 2, -11, 25, 96, 25, -11, 2, 0 }, { 2, -11, 22, 96, 28, -11, 2, 0 }, - { 2, -10, 19, 95, 31, -11, 2, 0 }, { 2, -10, 17, 95, 34, -12, 2, 0 }, - { 2, -9, 14, 94, 37, -12, 2, 0 }, { 2, -8, 12, 93, 40, -12, 1, 0 }, - { 2, -8, 9, 92, 43, -12, 1, 1 }, { 2, -7, 7, 91, 46, -12, 1, 0 }, - { 2, -7, 5, 90, 49, -12, 1, 0 }, { 2, -6, 3, 88, 52, -12, 0, 1 }, - { 2, -5, 1, 86, 55, -12, 0, 1 }, { 2, -5, -1, 84, 58, -11, 0, 1 }, - { 2, -4, -2, 82, 61, -11, -1, 1 }, { 2, -4, -4, 80, 64, -10, -1, 1 }, - { 1, -3, -5, 77, 67, -9, -1, 1 }, { 1, -3, -6, 75, 70, -8, -2, 1 }, - { 1, -2, -7, 72, 72, -7, -2, 1 }, { 1, -2, -8, 70, 75, -6, -3, 1 }, - { 1, -1, -9, 67, 77, -5, -3, 1 }, { 1, -1, -10, 64, 80, -4, -4, 2 }, - { 1, -1, -11, 61, 82, -2, -4, 2 }, { 1, 0, -11, 58, 84, -1, -5, 2 }, - { 1, 0, -12, 55, 86, 1, -5, 2 }, { 1, 0, -12, 52, 88, 3, -6, 2 }, - { 0, 1, -12, 49, 90, 5, -7, 2 }, { 0, 1, -12, 46, 91, 7, -7, 2 }, - { 1, 1, -12, 43, 92, 9, -8, 2 }, { 0, 1, -12, 40, 93, 12, -8, 2 }, - { 0, 2, -12, 37, 94, 14, -9, 2 }, { 0, 2, -12, 34, 95, 17, -10, 2 }, - { 0, 2, -11, 31, 95, 19, -10, 2 }, { 0, 2, -11, 28, 96, 22, -11, 2 }, -#elif SUBPEL_BITS_RS == 6 { 2, -11, 25, 96, 25, -11, 2, 0 }, { 2, -11, 24, 96, 26, -11, 2, 0 }, { 2, -11, 22, 96, 28, -11, 2, 0 }, { 2, -10, 21, 96, 29, -12, 2, 0 }, { 2, -10, 19, 96, 31, -12, 2, 0 }, { 2, -10, 18, 95, 32, -11, 2, 0 }, @@ -200,29 +144,10 @@ static const interp_kernel filteredinterp_filters750[(1 << SUBPEL_BITS_RS)] = { { 0, 2, -12, 34, 95, 17, -10, 2 }, { 0, 2, -11, 32, 95, 18, -10, 2 }, { 0, 2, -12, 31, 96, 19, -10, 2 }, { 0, 2, -12, 29, 96, 21, -10, 2 }, { 0, 2, -11, 28, 96, 22, -11, 2 }, { 0, 2, -11, 26, 96, 24, -11, 2 }, -#endif // SUBPEL_BITS_RS == 5 }; // Filters for interpolation (0.875-band) - note this also filters integer pels. static const interp_kernel filteredinterp_filters875[(1 << SUBPEL_BITS_RS)] = { -#if SUBPEL_BITS_RS == 5 - { 3, -8, 13, 112, 13, -8, 3, 0 }, { 3, -7, 10, 112, 17, -9, 3, -1 }, - { 2, -6, 7, 111, 21, -9, 3, -1 }, { 2, -5, 4, 111, 24, -10, 3, -1 }, - { 2, -4, 1, 110, 28, -11, 3, -1 }, { 1, -3, -1, 108, 32, -12, 4, -1 }, - { 1, -2, -3, 106, 36, -13, 4, -1 }, { 1, -1, -6, 105, 40, -14, 4, -1 }, - { 1, -1, -7, 102, 44, -14, 4, -1 }, { 1, 0, -9, 100, 48, -15, 4, -1 }, - { 1, 1, -11, 97, 53, -16, 4, -1 }, { 0, 1, -12, 95, 57, -16, 4, -1 }, - { 0, 2, -13, 91, 61, -16, 4, -1 }, { 0, 2, -14, 88, 65, -16, 4, -1 }, - { 0, 3, -15, 84, 69, -17, 4, 0 }, { 0, 3, -16, 81, 73, -16, 3, 0 }, - { 0, 3, -16, 77, 77, -16, 3, 0 }, { 0, 3, -16, 73, 81, -16, 3, 0 }, - { 0, 4, -17, 69, 84, -15, 3, 0 }, { -1, 4, -16, 65, 88, -14, 2, 0 }, - { -1, 4, -16, 61, 91, -13, 2, 0 }, { -1, 4, -16, 57, 95, -12, 1, 0 }, - { -1, 4, -16, 53, 97, -11, 1, 1 }, { -1, 4, -15, 48, 100, -9, 0, 1 }, - { -1, 4, -14, 44, 102, -7, -1, 1 }, { -1, 4, -14, 40, 105, -6, -1, 1 }, - { -1, 4, -13, 36, 106, -3, -2, 1 }, { -1, 4, -12, 32, 108, -1, -3, 1 }, - { -1, 3, -11, 28, 110, 1, -4, 2 }, { -1, 3, -10, 24, 111, 4, -5, 2 }, - { -1, 3, -9, 21, 111, 7, -6, 2 }, { -1, 3, -9, 17, 112, 10, -7, 3 }, -#elif SUBPEL_BITS_RS == 6 { 3, -8, 13, 112, 13, -8, 3, 0 }, { 2, -7, 12, 112, 15, -8, 3, -1 }, { 3, -7, 10, 112, 17, -9, 3, -1 }, { 2, -6, 8, 112, 19, -9, 3, -1 }, { 2, -6, 7, 112, 21, -10, 3, -1 }, { 2, -5, 6, 111, 22, -10, 3, -1 }, @@ -255,29 +180,10 @@ static const interp_kernel filteredinterp_filters875[(1 << SUBPEL_BITS_RS)] = { { -1, 3, -10, 24, 111, 4, -5, 2 }, { -1, 3, -10, 22, 111, 6, -5, 2 }, { -1, 3, -10, 21, 112, 7, -6, 2 }, { -1, 3, -9, 19, 112, 8, -6, 2 }, { -1, 3, -9, 17, 112, 10, -7, 3 }, { -1, 3, -8, 15, 112, 12, -7, 2 }, -#endif // SUBPEL_BITS_RS == 5 }; // Filters for interpolation (full-band) - no filtering for integer pixels static const interp_kernel filteredinterp_filters1000[(1 << SUBPEL_BITS_RS)] = { -#if SUBPEL_BITS_RS == 5 - { 0, 0, 0, 128, 0, 0, 0, 0 }, { 0, 1, -3, 128, 3, -1, 0, 0 }, - { -1, 2, -6, 127, 7, -2, 1, 0 }, { -1, 3, -9, 126, 12, -4, 1, 0 }, - { -1, 4, -12, 125, 16, -5, 1, 0 }, { -1, 4, -14, 123, 20, -6, 2, 0 }, - { -1, 5, -15, 120, 25, -8, 2, 0 }, { -1, 5, -17, 118, 30, -9, 3, -1 }, - { -1, 6, -18, 114, 35, -10, 3, -1 }, { -1, 6, -19, 111, 41, -12, 3, -1 }, - { -1, 6, -20, 107, 46, -13, 4, -1 }, { -1, 6, -21, 103, 52, -14, 4, -1 }, - { -1, 6, -21, 99, 57, -16, 5, -1 }, { -1, 6, -21, 94, 63, -17, 5, -1 }, - { -1, 6, -20, 89, 68, -18, 5, -1 }, { -1, 6, -20, 84, 73, -19, 6, -1 }, - { -1, 6, -20, 79, 79, -20, 6, -1 }, { -1, 6, -19, 73, 84, -20, 6, -1 }, - { -1, 5, -18, 68, 89, -20, 6, -1 }, { -1, 5, -17, 63, 94, -21, 6, -1 }, - { -1, 5, -16, 57, 99, -21, 6, -1 }, { -1, 4, -14, 52, 103, -21, 6, -1 }, - { -1, 4, -13, 46, 107, -20, 6, -1 }, { -1, 3, -12, 41, 111, -19, 6, -1 }, - { -1, 3, -10, 35, 114, -18, 6, -1 }, { -1, 3, -9, 30, 118, -17, 5, -1 }, - { 0, 2, -8, 25, 120, -15, 5, -1 }, { 0, 2, -6, 20, 123, -14, 4, -1 }, - { 0, 1, -5, 16, 125, -12, 4, -1 }, { 0, 1, -4, 12, 126, -9, 3, -1 }, - { 0, 1, -2, 7, 127, -6, 2, -1 }, { 0, 0, -1, 3, 128, -3, 1, 0 }, -#elif SUBPEL_BITS_RS == 6 { 0, 0, 0, 128, 0, 0, 0, 0 }, { 0, 0, -1, 128, 2, -1, 0, 0 }, { 0, 1, -3, 127, 4, -2, 1, 0 }, { 0, 1, -4, 127, 6, -3, 1, 0 }, { 0, 2, -6, 126, 8, -3, 1, 0 }, { 0, 2, -7, 125, 11, -4, 1, 0 }, @@ -310,9 +216,86 @@ static const interp_kernel filteredinterp_filters1000[(1 << SUBPEL_BITS_RS)] = { { 0, 2, -5, 13, 125, -8, 2, -1 }, { 0, 1, -4, 11, 125, -7, 2, 0 }, { 0, 1, -3, 8, 126, -6, 2, 0 }, { 0, 1, -3, 6, 127, -4, 1, 0 }, { 0, 1, -2, 4, 127, -3, 1, 0 }, { 0, 0, -1, 2, 128, -1, 0, 0 }, -#endif // SUBPEL_BITS_RS == 5 }; +#if CONFIG_FRAME_SUPERRES && CONFIG_LOOP_RESTORATION +#define INTERP_SIMPLE_TAPS 4 +static const int16_t filter_simple[(1 + << SUBPEL_BITS_RS)][INTERP_SIMPLE_TAPS] = { +#if INTERP_SIMPLE_TAPS == 2 + { 128, 0 }, { 126, 2 }, { 124, 4 }, { 122, 6 }, { 120, 8 }, { 118, 10 }, + { 116, 12 }, { 114, 14 }, { 112, 16 }, { 110, 18 }, { 108, 20 }, { 106, 22 }, + { 104, 24 }, { 102, 26 }, { 100, 28 }, { 98, 30 }, { 96, 32 }, { 94, 34 }, + { 92, 36 }, { 90, 38 }, { 88, 40 }, { 86, 42 }, { 84, 44 }, { 82, 46 }, + { 80, 48 }, { 78, 50 }, { 76, 52 }, { 74, 54 }, { 72, 56 }, { 70, 58 }, + { 68, 60 }, { 66, 62 }, { 64, 64 }, { 62, 66 }, { 60, 68 }, { 58, 70 }, + { 56, 72 }, { 54, 74 }, { 52, 76 }, { 50, 78 }, { 48, 80 }, { 46, 82 }, + { 44, 84 }, { 42, 86 }, { 40, 88 }, { 38, 90 }, { 36, 92 }, { 34, 94 }, + { 32, 96 }, { 30, 98 }, { 28, 100 }, { 26, 102 }, { 24, 104 }, { 22, 106 }, + { 20, 108 }, { 18, 110 }, { 16, 112 }, { 14, 114 }, { 12, 116 }, { 10, 118 }, + { 8, 120 }, { 6, 122 }, { 4, 124 }, { 2, 126 }, +#elif INTERP_SIMPLE_TAPS == 4 + { 0, 128, 0, 0 }, { -1, 128, 2, -1 }, { -2, 127, 4, -1 }, + { -3, 126, 7, -2 }, { -4, 125, 9, -2 }, { -5, 125, 11, -3 }, + { -6, 124, 13, -3 }, { -7, 123, 16, -4 }, { -7, 122, 18, -5 }, + { -8, 121, 20, -5 }, { -9, 120, 23, -6 }, { -9, 118, 25, -6 }, + { -10, 117, 28, -7 }, { -11, 116, 30, -7 }, { -11, 114, 33, -8 }, + { -12, 113, 35, -8 }, { -12, 111, 38, -9 }, { -13, 109, 41, -9 }, + { -13, 108, 43, -10 }, { -13, 106, 45, -10 }, { -13, 104, 48, -11 }, + { -14, 102, 51, -11 }, { -14, 100, 53, -11 }, { -14, 98, 56, -12 }, + { -14, 96, 58, -12 }, { -14, 94, 61, -13 }, { -15, 92, 64, -13 }, + { -15, 90, 66, -13 }, { -15, 87, 69, -13 }, { -14, 85, 71, -14 }, + { -14, 83, 73, -14 }, { -14, 80, 76, -14 }, { -14, 78, 78, -14 }, + { -14, 76, 80, -14 }, { -14, 73, 83, -14 }, { -14, 71, 85, -14 }, + { -13, 69, 87, -15 }, { -13, 66, 90, -15 }, { -13, 64, 92, -15 }, + { -13, 61, 94, -14 }, { -12, 58, 96, -14 }, { -12, 56, 98, -14 }, + { -11, 53, 100, -14 }, { -11, 51, 102, -14 }, { -11, 48, 104, -13 }, + { -10, 45, 106, -13 }, { -10, 43, 108, -13 }, { -9, 41, 109, -13 }, + { -9, 38, 111, -12 }, { -8, 35, 113, -12 }, { -8, 33, 114, -11 }, + { -7, 30, 116, -11 }, { -7, 28, 117, -10 }, { -6, 25, 118, -9 }, + { -6, 23, 120, -9 }, { -5, 20, 121, -8 }, { -5, 18, 122, -7 }, + { -4, 16, 123, -7 }, { -3, 13, 124, -6 }, { -3, 11, 125, -5 }, + { -2, 9, 125, -4 }, { -2, 7, 126, -3 }, { -1, 4, 127, -2 }, + { -1, 2, 128, -1 }, +#elif INTERP_SIMPLE_TAPS == 6 + { 0, 0, 128, 0, 0, 0 }, { 0, -1, 128, 2, -1, 0 }, + { 1, -3, 127, 4, -2, 1 }, { 1, -4, 127, 6, -3, 1 }, + { 2, -6, 126, 8, -3, 1 }, { 2, -7, 125, 11, -4, 1 }, + { 2, -9, 125, 13, -5, 2 }, { 3, -10, 124, 15, -6, 2 }, + { 3, -11, 123, 18, -7, 2 }, { 3, -12, 122, 20, -8, 3 }, + { 4, -13, 121, 22, -9, 3 }, { 4, -14, 119, 25, -9, 3 }, + { 4, -15, 118, 27, -10, 4 }, { 4, -16, 117, 30, -11, 4 }, + { 5, -17, 116, 32, -12, 4 }, { 5, -17, 114, 35, -13, 4 }, + { 5, -18, 112, 37, -13, 5 }, { 5, -19, 111, 40, -14, 5 }, + { 6, -19, 109, 42, -15, 5 }, { 6, -20, 107, 45, -15, 5 }, + { 6, -20, 105, 48, -16, 5 }, { 6, -21, 103, 51, -17, 6 }, + { 6, -21, 101, 53, -17, 6 }, { 6, -21, 99, 56, -18, 6 }, + { 7, -22, 97, 58, -18, 6 }, { 7, -22, 95, 61, -19, 6 }, + { 7, -22, 93, 63, -19, 6 }, { 7, -22, 91, 66, -20, 6 }, + { 7, -22, 88, 69, -20, 6 }, { 7, -22, 86, 71, -21, 7 }, + { 7, -22, 83, 74, -21, 7 }, { 7, -22, 81, 76, -21, 7 }, + { 7, -22, 79, 79, -22, 7 }, { 7, -21, 76, 81, -22, 7 }, + { 7, -21, 74, 83, -22, 7 }, { 7, -21, 71, 86, -22, 7 }, + { 6, -20, 69, 88, -22, 7 }, { 6, -20, 66, 91, -22, 7 }, + { 6, -19, 63, 93, -22, 7 }, { 6, -19, 61, 95, -22, 7 }, + { 6, -18, 58, 97, -22, 7 }, { 6, -18, 56, 99, -21, 6 }, + { 6, -17, 53, 101, -21, 6 }, { 6, -17, 51, 103, -21, 6 }, + { 5, -16, 48, 105, -20, 6 }, { 5, -15, 45, 107, -20, 6 }, + { 5, -15, 42, 109, -19, 6 }, { 5, -14, 40, 111, -19, 5 }, + { 5, -13, 37, 112, -18, 5 }, { 4, -13, 35, 114, -17, 5 }, + { 4, -12, 32, 116, -17, 5 }, { 4, -11, 30, 117, -16, 4 }, + { 4, -10, 27, 118, -15, 4 }, { 3, -9, 25, 119, -14, 4 }, + { 3, -9, 22, 121, -13, 4 }, { 3, -8, 20, 122, -12, 3 }, + { 2, -7, 18, 123, -11, 3 }, { 2, -6, 15, 124, -10, 3 }, + { 2, -5, 13, 125, -9, 2 }, { 1, -4, 11, 125, -7, 2 }, + { 1, -3, 8, 126, -6, 2 }, { 1, -3, 6, 127, -4, 1 }, + { 1, -2, 4, 127, -3, 1 }, { 0, -1, 2, 128, -1, 0 }, +#else +#error "Invalid value of INTERP_SIMPLE_TAPS" +#endif // INTERP_SIMPLE_TAPS == 2 +}; +#endif // CONFIG_FRAME_SUPERRES && CONFIG_LOOP_RESTORATION + // Filters for factor of 2 downsampling. static const int16_t av1_down2_symeven_half_filter[] = { 56, 12, -3, -1 }; static const int16_t av1_down2_symodd_half_filter[] = { 64, 35, 0, -3 }; @@ -331,33 +314,34 @@ static const interp_kernel *choose_interp_filter(int inlength, int outlength) { return filteredinterp_filters500; } -static void interpolate(const uint8_t *const input, int inlength, - uint8_t *output, int outlength) { - const int64_t delta = - (((uint64_t)inlength << 32) + outlength / 2) / outlength; - const int64_t offset = +static void interpolate_core(const uint8_t *const input, int inlength, + uint8_t *output, int outlength, + const int16_t *interp_filters, int interp_taps) { + const int32_t delta = + (((uint32_t)inlength << INTERP_PRECISION_BITS) + outlength / 2) / + outlength; + const int32_t offset = inlength > outlength - ? (((int64_t)(inlength - outlength) << 31) + outlength / 2) / + ? (((int32_t)(inlength - outlength) << (INTERP_PRECISION_BITS - 1)) + + outlength / 2) / outlength - : -(((int64_t)(outlength - inlength) << 31) + outlength / 2) / + : -(((int32_t)(outlength - inlength) << (INTERP_PRECISION_BITS - 1)) + + outlength / 2) / outlength; uint8_t *optr = output; int x, x1, x2, sum, k, int_pel, sub_pel; - int64_t y; - - const interp_kernel *interp_filters = - choose_interp_filter(inlength, outlength); + int32_t y; x = 0; y = offset + SUBPEL_INTERP_EXTRA_OFF; - while ((y >> INTERP_PRECISION_BITS) < (INTERP_TAPS / 2 - 1)) { + while ((y >> INTERP_PRECISION_BITS) < (interp_taps / 2 - 1)) { x++; y += delta; } x1 = x; x = outlength - 1; y = delta * x + offset + SUBPEL_INTERP_EXTRA_OFF; - while ((y >> INTERP_PRECISION_BITS) + (int64_t)(INTERP_TAPS / 2) >= + while ((y >> INTERP_PRECISION_BITS) + (int32_t)(interp_taps / 2) >= inlength) { x--; y -= delta; @@ -366,13 +350,12 @@ static void interpolate(const uint8_t *const input, int inlength, if (x1 > x2) { for (x = 0, y = offset + SUBPEL_INTERP_EXTRA_OFF; x < outlength; ++x, y += delta) { - const int16_t *filter; int_pel = y >> INTERP_PRECISION_BITS; sub_pel = (y >> SUBPEL_INTERP_EXTRA_BITS) & SUBPEL_MASK_RS; - filter = interp_filters[sub_pel]; + const int16_t *filter = &interp_filters[sub_pel * interp_taps]; sum = 0; - for (k = 0; k < INTERP_TAPS; ++k) { - const int pk = int_pel - INTERP_TAPS / 2 + 1 + k; + for (k = 0; k < interp_taps; ++k) { + const int pk = int_pel - interp_taps / 2 + 1 + k; sum += filter[k] * input[AOMMAX(AOMMIN(pk, inlength - 1), 0)]; } *optr++ = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS)); @@ -380,41 +363,55 @@ static void interpolate(const uint8_t *const input, int inlength, } else { // Initial part. for (x = 0, y = offset + SUBPEL_INTERP_EXTRA_OFF; x < x1; ++x, y += delta) { - const int16_t *filter; int_pel = y >> INTERP_PRECISION_BITS; sub_pel = (y >> SUBPEL_INTERP_EXTRA_BITS) & SUBPEL_MASK_RS; - filter = interp_filters[sub_pel]; + const int16_t *filter = &interp_filters[sub_pel * interp_taps]; sum = 0; - for (k = 0; k < INTERP_TAPS; ++k) - sum += filter[k] * input[AOMMAX(int_pel - INTERP_TAPS / 2 + 1 + k, 0)]; + for (k = 0; k < interp_taps; ++k) + sum += filter[k] * input[AOMMAX(int_pel - interp_taps / 2 + 1 + k, 0)]; *optr++ = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS)); } // Middle part. for (; x <= x2; ++x, y += delta) { - const int16_t *filter; int_pel = y >> INTERP_PRECISION_BITS; sub_pel = (y >> SUBPEL_INTERP_EXTRA_BITS) & SUBPEL_MASK_RS; - filter = interp_filters[sub_pel]; + const int16_t *filter = &interp_filters[sub_pel * interp_taps]; sum = 0; - for (k = 0; k < INTERP_TAPS; ++k) - sum += filter[k] * input[int_pel - INTERP_TAPS / 2 + 1 + k]; + for (k = 0; k < interp_taps; ++k) + sum += filter[k] * input[int_pel - interp_taps / 2 + 1 + k]; *optr++ = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS)); } // End part. for (; x < outlength; ++x, y += delta) { - const int16_t *filter; int_pel = y >> INTERP_PRECISION_BITS; sub_pel = (y >> SUBPEL_INTERP_EXTRA_BITS) & SUBPEL_MASK_RS; - filter = interp_filters[sub_pel]; + const int16_t *filter = &interp_filters[sub_pel * interp_taps]; sum = 0; - for (k = 0; k < INTERP_TAPS; ++k) + for (k = 0; k < interp_taps; ++k) sum += filter[k] * - input[AOMMIN(int_pel - INTERP_TAPS / 2 + 1 + k, inlength - 1)]; + input[AOMMIN(int_pel - interp_taps / 2 + 1 + k, inlength - 1)]; *optr++ = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS)); } } } +static void interpolate(const uint8_t *const input, int inlength, + uint8_t *output, int outlength) { + const interp_kernel *interp_filters = + choose_interp_filter(inlength, outlength); + + interpolate_core(input, inlength, output, outlength, &interp_filters[0][0], + INTERP_TAPS); +} + +#if CONFIG_FRAME_SUPERRES && CONFIG_LOOP_RESTORATION +static void interpolate_simple(const uint8_t *const input, int inlength, + uint8_t *output, int outlength) { + interpolate_core(input, inlength, output, outlength, &filter_simple[0][0], + INTERP_SIMPLE_TAPS); +} +#endif // CONFIG_FRAME_SUPERRES && CONFIG_LOOP_RESTORATION + #ifndef __clang_analyzer__ static void down2_symeven(const uint8_t *const input, int length, uint8_t *output) { @@ -596,14 +593,15 @@ static void fill_arr_to_col(uint8_t *img, int stride, int len, uint8_t *arr) { } } -void av1_resize_plane(const uint8_t *const input, int height, int width, - int in_stride, uint8_t *output, int height2, int width2, - int out_stride) { +static void resize_plane(const uint8_t *const input, int height, int width, + int in_stride, uint8_t *output, int height2, + int width2, int out_stride) { int i; - uint8_t *intbuf = (uint8_t *)malloc(sizeof(uint8_t) * width2 * height); - uint8_t *tmpbuf = (uint8_t *)malloc(sizeof(uint8_t) * AOMMAX(width, height)); - uint8_t *arrbuf = (uint8_t *)malloc(sizeof(uint8_t) * height); - uint8_t *arrbuf2 = (uint8_t *)malloc(sizeof(uint8_t) * height2); + uint8_t *intbuf = (uint8_t *)aom_malloc(sizeof(uint8_t) * width2 * height); + uint8_t *tmpbuf = + (uint8_t *)aom_malloc(sizeof(uint8_t) * AOMMAX(width, height)); + uint8_t *arrbuf = (uint8_t *)aom_malloc(sizeof(uint8_t) * height); + uint8_t *arrbuf2 = (uint8_t *)aom_malloc(sizeof(uint8_t) * height2); if (intbuf == NULL || tmpbuf == NULL || arrbuf == NULL || arrbuf2 == NULL) goto Error; assert(width > 0); @@ -620,40 +618,80 @@ void av1_resize_plane(const uint8_t *const input, int height, int width, } Error: - free(intbuf); - free(tmpbuf); - free(arrbuf); - free(arrbuf2); + aom_free(intbuf); + aom_free(tmpbuf); + aom_free(arrbuf); + aom_free(arrbuf2); +} + +#if CONFIG_FRAME_SUPERRES +static void upscale_normative(const uint8_t *const input, int length, + uint8_t *output, int olength) { +#if CONFIG_LOOP_RESTORATION + interpolate_simple(input, length, output, olength); +#else + interpolate(input, length, output, olength); +#endif // CONFIG_LOOP_RESTORATION } +static void upscale_normative_plane(const uint8_t *const input, int height, + int width, int in_stride, uint8_t *output, + int height2, int width2, int out_stride) { + int i; + uint8_t *intbuf = (uint8_t *)aom_malloc(sizeof(uint8_t) * width2 * height); + uint8_t *arrbuf = (uint8_t *)aom_malloc(sizeof(uint8_t) * height); + uint8_t *arrbuf2 = (uint8_t *)aom_malloc(sizeof(uint8_t) * height2); + if (intbuf == NULL || arrbuf == NULL || arrbuf2 == NULL) goto Error; + assert(width > 0); + assert(height > 0); + assert(width2 > 0); + assert(height2 > 0); + for (i = 0; i < height; ++i) + upscale_normative(input + in_stride * i, width, intbuf + width2 * i, + width2); + for (i = 0; i < width2; ++i) { + fill_col_to_arr(intbuf + i, width2, height, arrbuf); + upscale_normative(arrbuf, height, arrbuf2, height2); + fill_arr_to_col(output + i, out_stride, height2, arrbuf2); + } + +Error: + aom_free(intbuf); + aom_free(arrbuf); + aom_free(arrbuf2); +} +#endif // CONFIG_FRAME_SUPERRES + #if CONFIG_HIGHBITDEPTH -static void highbd_interpolate(const uint16_t *const input, int inlength, - uint16_t *output, int outlength, int bd) { - const int64_t delta = - (((uint64_t)inlength << 32) + outlength / 2) / outlength; - const int64_t offset = +static void highbd_interpolate_core(const uint16_t *const input, int inlength, + uint16_t *output, int outlength, int bd, + const int16_t *interp_filters, + int interp_taps) { + const int32_t delta = + (((uint32_t)inlength << INTERP_PRECISION_BITS) + outlength / 2) / + outlength; + const int32_t offset = inlength > outlength - ? (((int64_t)(inlength - outlength) << 31) + outlength / 2) / + ? (((int32_t)(inlength - outlength) << (INTERP_PRECISION_BITS - 1)) + + outlength / 2) / outlength - : -(((int64_t)(outlength - inlength) << 31) + outlength / 2) / + : -(((int32_t)(outlength - inlength) << (INTERP_PRECISION_BITS - 1)) + + outlength / 2) / outlength; uint16_t *optr = output; int x, x1, x2, sum, k, int_pel, sub_pel; - int64_t y; - - const interp_kernel *interp_filters = - choose_interp_filter(inlength, outlength); + int32_t y; x = 0; y = offset + SUBPEL_INTERP_EXTRA_OFF; - while ((y >> INTERP_PRECISION_BITS) < (INTERP_TAPS / 2 - 1)) { + while ((y >> INTERP_PRECISION_BITS) < (interp_taps / 2 - 1)) { x++; y += delta; } x1 = x; x = outlength - 1; y = delta * x + offset + SUBPEL_INTERP_EXTRA_OFF; - while ((y >> INTERP_PRECISION_BITS) + (int64_t)(INTERP_TAPS / 2) >= + while ((y >> INTERP_PRECISION_BITS) + (int32_t)(interp_taps / 2) >= inlength) { x--; y -= delta; @@ -662,13 +700,12 @@ static void highbd_interpolate(const uint16_t *const input, int inlength, if (x1 > x2) { for (x = 0, y = offset + SUBPEL_INTERP_EXTRA_OFF; x < outlength; ++x, y += delta) { - const int16_t *filter; int_pel = y >> INTERP_PRECISION_BITS; sub_pel = (y >> SUBPEL_INTERP_EXTRA_BITS) & SUBPEL_MASK_RS; - filter = interp_filters[sub_pel]; + const int16_t *filter = &interp_filters[sub_pel * interp_taps]; sum = 0; - for (k = 0; k < INTERP_TAPS; ++k) { - const int pk = int_pel - INTERP_TAPS / 2 + 1 + k; + for (k = 0; k < interp_taps; ++k) { + const int pk = int_pel - interp_taps / 2 + 1 + k; sum += filter[k] * input[AOMMAX(AOMMIN(pk, inlength - 1), 0)]; } *optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd); @@ -676,41 +713,55 @@ static void highbd_interpolate(const uint16_t *const input, int inlength, } else { // Initial part. for (x = 0, y = offset + SUBPEL_INTERP_EXTRA_OFF; x < x1; ++x, y += delta) { - const int16_t *filter; int_pel = y >> INTERP_PRECISION_BITS; sub_pel = (y >> SUBPEL_INTERP_EXTRA_BITS) & SUBPEL_MASK_RS; - filter = interp_filters[sub_pel]; + const int16_t *filter = &interp_filters[sub_pel * interp_taps]; sum = 0; - for (k = 0; k < INTERP_TAPS; ++k) - sum += filter[k] * input[AOMMAX(int_pel - INTERP_TAPS / 2 + 1 + k, 0)]; + for (k = 0; k < interp_taps; ++k) + sum += filter[k] * input[AOMMAX(int_pel - interp_taps / 2 + 1 + k, 0)]; *optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd); } // Middle part. for (; x <= x2; ++x, y += delta) { - const int16_t *filter; int_pel = y >> INTERP_PRECISION_BITS; sub_pel = (y >> SUBPEL_INTERP_EXTRA_BITS) & SUBPEL_MASK_RS; - filter = interp_filters[sub_pel]; + const int16_t *filter = &interp_filters[sub_pel * interp_taps]; sum = 0; - for (k = 0; k < INTERP_TAPS; ++k) - sum += filter[k] * input[int_pel - INTERP_TAPS / 2 + 1 + k]; + for (k = 0; k < interp_taps; ++k) + sum += filter[k] * input[int_pel - interp_taps / 2 + 1 + k]; *optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd); } // End part. for (; x < outlength; ++x, y += delta) { - const int16_t *filter; int_pel = y >> INTERP_PRECISION_BITS; sub_pel = (y >> SUBPEL_INTERP_EXTRA_BITS) & SUBPEL_MASK_RS; - filter = interp_filters[sub_pel]; + const int16_t *filter = &interp_filters[sub_pel * interp_taps]; sum = 0; - for (k = 0; k < INTERP_TAPS; ++k) + for (k = 0; k < interp_taps; ++k) sum += filter[k] * - input[AOMMIN(int_pel - INTERP_TAPS / 2 + 1 + k, inlength - 1)]; + input[AOMMIN(int_pel - interp_taps / 2 + 1 + k, inlength - 1)]; *optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd); } } } +static void highbd_interpolate(const uint16_t *const input, int inlength, + uint16_t *output, int outlength, int bd) { + const interp_kernel *interp_filters = + choose_interp_filter(inlength, outlength); + + highbd_interpolate_core(input, inlength, output, outlength, bd, + &interp_filters[0][0], INTERP_TAPS); +} + +#if CONFIG_FRAME_SUPERRES && CONFIG_LOOP_RESTORATION +static void highbd_interpolate_simple(const uint16_t *const input, int inlength, + uint16_t *output, int outlength, int bd) { + highbd_interpolate_core(input, inlength, output, outlength, bd, + &filter_simple[0][0], INTERP_SIMPLE_TAPS); +} +#endif // CONFIG_FRAME_SUPERRES && CONFIG_LOOP_RESTORATION + #ifndef __clang_analyzer__ static void highbd_down2_symeven(const uint16_t *const input, int length, uint16_t *output, int bd) { @@ -877,15 +928,16 @@ static void highbd_fill_arr_to_col(uint16_t *img, int stride, int len, } } -void av1_highbd_resize_plane(const uint8_t *const input, int height, int width, - int in_stride, uint8_t *output, int height2, - int width2, int out_stride, int bd) { +static void highbd_resize_plane(const uint8_t *const input, int height, + int width, int in_stride, uint8_t *output, + int height2, int width2, int out_stride, + int bd) { int i; - uint16_t *intbuf = (uint16_t *)malloc(sizeof(uint16_t) * width2 * height); + uint16_t *intbuf = (uint16_t *)aom_malloc(sizeof(uint16_t) * width2 * height); uint16_t *tmpbuf = - (uint16_t *)malloc(sizeof(uint16_t) * AOMMAX(width, height)); - uint16_t *arrbuf = (uint16_t *)malloc(sizeof(uint16_t) * height); - uint16_t *arrbuf2 = (uint16_t *)malloc(sizeof(uint16_t) * height2); + (uint16_t *)aom_malloc(sizeof(uint16_t) * AOMMAX(width, height)); + uint16_t *arrbuf = (uint16_t *)aom_malloc(sizeof(uint16_t) * height); + uint16_t *arrbuf2 = (uint16_t *)aom_malloc(sizeof(uint16_t) * height2); if (intbuf == NULL || tmpbuf == NULL || arrbuf == NULL || arrbuf2 == NULL) goto Error; for (i = 0; i < height; ++i) { @@ -900,11 +952,49 @@ void av1_highbd_resize_plane(const uint8_t *const input, int height, int width, } Error: - free(intbuf); - free(tmpbuf); - free(arrbuf); - free(arrbuf2); + aom_free(intbuf); + aom_free(tmpbuf); + aom_free(arrbuf); + aom_free(arrbuf2); } + +#if CONFIG_FRAME_SUPERRES +static void highbd_upscale_normative(const uint16_t *const input, int length, + uint16_t *output, int olength, int bd) { +#if CONFIG_LOOP_RESTORATION + highbd_interpolate_simple(input, length, output, olength, bd); +#else + highbd_interpolate(input, length, output, olength, bd); +#endif // CONFIG_LOOP_RESTORATION +} + +static void highbd_upscale_normative_plane(const uint8_t *const input, + int height, int width, int in_stride, + uint8_t *output, int height2, + int width2, int out_stride, int bd) { + int i; + uint16_t *intbuf = (uint16_t *)aom_malloc(sizeof(uint16_t) * width2 * height); + uint16_t *arrbuf = (uint16_t *)aom_malloc(sizeof(uint16_t) * height); + uint16_t *arrbuf2 = (uint16_t *)aom_malloc(sizeof(uint16_t) * height2); + if (intbuf == NULL || arrbuf == NULL || arrbuf2 == NULL) goto Error; + for (i = 0; i < height; ++i) { + highbd_upscale_normative(CONVERT_TO_SHORTPTR(input + in_stride * i), width, + intbuf + width2 * i, width2, bd); + } + for (i = 0; i < width2; ++i) { + highbd_fill_col_to_arr(intbuf + i, width2, height, arrbuf); + highbd_upscale_normative(arrbuf, height, arrbuf2, height2, bd); + highbd_fill_arr_to_col(CONVERT_TO_SHORTPTR(output + i), out_stride, height2, + arrbuf2); + } + +Error: + aom_free(intbuf); + aom_free(arrbuf); + aom_free(arrbuf2); +} +#endif // CONFIG_FRAME_SUPERRES + #endif // CONFIG_HIGHBITDEPTH void av1_resize_frame420(const uint8_t *const y, int y_stride, @@ -912,11 +1002,11 @@ void av1_resize_frame420(const uint8_t *const y, int y_stride, int uv_stride, int height, int width, uint8_t *oy, int oy_stride, uint8_t *ou, uint8_t *ov, int ouv_stride, int oheight, int owidth) { - av1_resize_plane(y, height, width, y_stride, oy, oheight, owidth, oy_stride); - av1_resize_plane(u, height / 2, width / 2, uv_stride, ou, oheight / 2, - owidth / 2, ouv_stride); - av1_resize_plane(v, height / 2, width / 2, uv_stride, ov, oheight / 2, - owidth / 2, ouv_stride); + resize_plane(y, height, width, y_stride, oy, oheight, owidth, oy_stride); + resize_plane(u, height / 2, width / 2, uv_stride, ou, oheight / 2, owidth / 2, + ouv_stride); + resize_plane(v, height / 2, width / 2, uv_stride, ov, oheight / 2, owidth / 2, + ouv_stride); } void av1_resize_frame422(const uint8_t *const y, int y_stride, @@ -924,11 +1014,11 @@ void av1_resize_frame422(const uint8_t *const y, int y_stride, int uv_stride, int height, int width, uint8_t *oy, int oy_stride, uint8_t *ou, uint8_t *ov, int ouv_stride, int oheight, int owidth) { - av1_resize_plane(y, height, width, y_stride, oy, oheight, owidth, oy_stride); - av1_resize_plane(u, height, width / 2, uv_stride, ou, oheight, owidth / 2, - ouv_stride); - av1_resize_plane(v, height, width / 2, uv_stride, ov, oheight, owidth / 2, - ouv_stride); + resize_plane(y, height, width, y_stride, oy, oheight, owidth, oy_stride); + resize_plane(u, height, width / 2, uv_stride, ou, oheight, owidth / 2, + ouv_stride); + resize_plane(v, height, width / 2, uv_stride, ov, oheight, owidth / 2, + ouv_stride); } void av1_resize_frame444(const uint8_t *const y, int y_stride, @@ -936,11 +1026,9 @@ void av1_resize_frame444(const uint8_t *const y, int y_stride, int uv_stride, int height, int width, uint8_t *oy, int oy_stride, uint8_t *ou, uint8_t *ov, int ouv_stride, int oheight, int owidth) { - av1_resize_plane(y, height, width, y_stride, oy, oheight, owidth, oy_stride); - av1_resize_plane(u, height, width, uv_stride, ou, oheight, owidth, - ouv_stride); - av1_resize_plane(v, height, width, uv_stride, ov, oheight, owidth, - ouv_stride); + resize_plane(y, height, width, y_stride, oy, oheight, owidth, oy_stride); + resize_plane(u, height, width, uv_stride, ou, oheight, owidth, ouv_stride); + resize_plane(v, height, width, uv_stride, ov, oheight, owidth, ouv_stride); } #if CONFIG_HIGHBITDEPTH @@ -950,12 +1038,12 @@ void av1_highbd_resize_frame420(const uint8_t *const y, int y_stride, uint8_t *oy, int oy_stride, uint8_t *ou, uint8_t *ov, int ouv_stride, int oheight, int owidth, int bd) { - av1_highbd_resize_plane(y, height, width, y_stride, oy, oheight, owidth, - oy_stride, bd); - av1_highbd_resize_plane(u, height / 2, width / 2, uv_stride, ou, oheight / 2, - owidth / 2, ouv_stride, bd); - av1_highbd_resize_plane(v, height / 2, width / 2, uv_stride, ov, oheight / 2, - owidth / 2, ouv_stride, bd); + highbd_resize_plane(y, height, width, y_stride, oy, oheight, owidth, + oy_stride, bd); + highbd_resize_plane(u, height / 2, width / 2, uv_stride, ou, oheight / 2, + owidth / 2, ouv_stride, bd); + highbd_resize_plane(v, height / 2, width / 2, uv_stride, ov, oheight / 2, + owidth / 2, ouv_stride, bd); } void av1_highbd_resize_frame422(const uint8_t *const y, int y_stride, @@ -964,12 +1052,12 @@ void av1_highbd_resize_frame422(const uint8_t *const y, int y_stride, uint8_t *oy, int oy_stride, uint8_t *ou, uint8_t *ov, int ouv_stride, int oheight, int owidth, int bd) { - av1_highbd_resize_plane(y, height, width, y_stride, oy, oheight, owidth, - oy_stride, bd); - av1_highbd_resize_plane(u, height, width / 2, uv_stride, ou, oheight, - owidth / 2, ouv_stride, bd); - av1_highbd_resize_plane(v, height, width / 2, uv_stride, ov, oheight, - owidth / 2, ouv_stride, bd); + highbd_resize_plane(y, height, width, y_stride, oy, oheight, owidth, + oy_stride, bd); + highbd_resize_plane(u, height, width / 2, uv_stride, ou, oheight, owidth / 2, + ouv_stride, bd); + highbd_resize_plane(v, height, width / 2, uv_stride, ov, oheight, owidth / 2, + ouv_stride, bd); } void av1_highbd_resize_frame444(const uint8_t *const y, int y_stride, @@ -978,12 +1066,12 @@ void av1_highbd_resize_frame444(const uint8_t *const y, int y_stride, uint8_t *oy, int oy_stride, uint8_t *ou, uint8_t *ov, int ouv_stride, int oheight, int owidth, int bd) { - av1_highbd_resize_plane(y, height, width, y_stride, oy, oheight, owidth, - oy_stride, bd); - av1_highbd_resize_plane(u, height, width, uv_stride, ou, oheight, owidth, - ouv_stride, bd); - av1_highbd_resize_plane(v, height, width, uv_stride, ov, oheight, owidth, - ouv_stride, bd); + highbd_resize_plane(y, height, width, y_stride, oy, oheight, owidth, + oy_stride, bd); + highbd_resize_plane(u, height, width, uv_stride, ou, oheight, owidth, + ouv_stride, bd); + highbd_resize_plane(v, height, width, uv_stride, ov, oheight, owidth, + ouv_stride, bd); } #endif // CONFIG_HIGHBITDEPTH @@ -1013,30 +1101,56 @@ void av1_resize_and_extend_frame(const YV12_BUFFER_CONFIG *src, for (i = 0; i < MAX_MB_PLANE; ++i) { #if CONFIG_HIGHBITDEPTH if (src->flags & YV12_FLAG_HIGHBITDEPTH) - av1_highbd_resize_plane(srcs[i], src_heights[i], src_widths[i], - src_strides[i], dsts[i], dst_heights[i], - dst_widths[i], dst_strides[i], bd); + highbd_resize_plane(srcs[i], src_heights[i], src_widths[i], + src_strides[i], dsts[i], dst_heights[i], + dst_widths[i], dst_strides[i], bd); else #endif // CONFIG_HIGHBITDEPTH - av1_resize_plane(srcs[i], src_heights[i], src_widths[i], src_strides[i], - dsts[i], dst_heights[i], dst_widths[i], dst_strides[i]); + resize_plane(srcs[i], src_heights[i], src_widths[i], src_strides[i], + dsts[i], dst_heights[i], dst_widths[i], dst_strides[i]); } aom_extend_frame_borders(dst); } -YV12_BUFFER_CONFIG *av1_scale_if_required_fast(AV1_COMMON *cm, - YV12_BUFFER_CONFIG *unscaled, - YV12_BUFFER_CONFIG *scaled) { - if (cm->width != unscaled->y_crop_width || - cm->height != unscaled->y_crop_height) { - // For 2x2 scaling down. - aom_scale_frame(unscaled, scaled, unscaled->y_buffer, 9, 2, 1, 2, 1, 0); - aom_extend_frame_borders(scaled); - return scaled; - } else { - return unscaled; +#if CONFIG_FRAME_SUPERRES +#if CONFIG_HIGHBITDEPTH +void av1_upscale_normative_and_extend_frame(const YV12_BUFFER_CONFIG *src, + YV12_BUFFER_CONFIG *dst, int bd) { +#else +void av1_upscale_normative_and_extend_frame(const YV12_BUFFER_CONFIG *src, + YV12_BUFFER_CONFIG *dst) { +#endif // CONFIG_HIGHBITDEPTH + // TODO(dkovalev): replace YV12_BUFFER_CONFIG with aom_image_t + int i; + const uint8_t *const srcs[3] = { src->y_buffer, src->u_buffer, + src->v_buffer }; + const int src_strides[3] = { src->y_stride, src->uv_stride, src->uv_stride }; + const int src_widths[3] = { src->y_crop_width, src->uv_crop_width, + src->uv_crop_width }; + const int src_heights[3] = { src->y_crop_height, src->uv_crop_height, + src->uv_crop_height }; + uint8_t *const dsts[3] = { dst->y_buffer, dst->u_buffer, dst->v_buffer }; + const int dst_strides[3] = { dst->y_stride, dst->uv_stride, dst->uv_stride }; + const int dst_widths[3] = { dst->y_crop_width, dst->uv_crop_width, + dst->uv_crop_width }; + const int dst_heights[3] = { dst->y_crop_height, dst->uv_crop_height, + dst->uv_crop_height }; + + for (i = 0; i < MAX_MB_PLANE; ++i) { +#if CONFIG_HIGHBITDEPTH + if (src->flags & YV12_FLAG_HIGHBITDEPTH) + highbd_upscale_normative_plane(srcs[i], src_heights[i], src_widths[i], + src_strides[i], dsts[i], dst_heights[i], + dst_widths[i], dst_strides[i], bd); + else +#endif // CONFIG_HIGHBITDEPTH + upscale_normative_plane(srcs[i], src_heights[i], src_widths[i], + src_strides[i], dsts[i], dst_heights[i], + dst_widths[i], dst_strides[i]); } + aom_extend_frame_borders(dst); } +#endif // CONFIG_FRAME_SUPERRES YV12_BUFFER_CONFIG *av1_scale_if_required(AV1_COMMON *cm, YV12_BUFFER_CONFIG *unscaled, @@ -1054,17 +1168,45 @@ YV12_BUFFER_CONFIG *av1_scale_if_required(AV1_COMMON *cm, } } -void av1_calculate_scaled_size(int *width, int *height, int num) { - if (num != SCALE_DENOMINATOR) { - *width = *width * num / SCALE_DENOMINATOR; - *height = *height * num / SCALE_DENOMINATOR; - // Make width and height even - *width += *width & 1; - *height += *height & 1; +// Calculates scaled dimensions given original dimensions and the scale +// denominator. If 'scale_height' is 1, both width and height are scaled; +// otherwise, only the width is scaled. +static void calculate_scaled_size_helper(int *width, int *height, int denom, + int scale_height) { + if (denom != SCALE_NUMERATOR) { + *width = *width * SCALE_NUMERATOR / denom; + *width += *width & 1; // Make it even. + if (scale_height) { + *height = *height * SCALE_NUMERATOR / denom; + *height += *height & 1; // Make it even. + } } } +void av1_calculate_scaled_size(int *width, int *height, int resize_denom) { + calculate_scaled_size_helper(width, height, resize_denom, 1); +} + #if CONFIG_FRAME_SUPERRES +void av1_calculate_scaled_superres_size(int *width, int *height, + int superres_denom) { + calculate_scaled_size_helper(width, height, superres_denom, + !CONFIG_HORZONLY_FRAME_SUPERRES); +} + +void av1_calculate_unscaled_superres_size(int *width, int *height, int denom) { + if (denom != SCALE_NUMERATOR) { + // Note: av1_calculate_scaled_superres_size() rounds *up* after division + // when the resulting dimensions are odd. So here, we round *down*. + *width = *width * denom / SCALE_NUMERATOR; +#if CONFIG_HORZONLY_FRAME_SUPERRES + (void)height; +#else + *height = *height * denom / SCALE_NUMERATOR; +#endif // CONFIG_HORZONLY_FRAME_SUPERRES + } +} + // TODO(afergs): Look for in-place upscaling // TODO(afergs): aom_ vs av1_ functions? Which can I use? // Upscale decoded image. @@ -1138,11 +1280,13 @@ void av1_superres_upscale(AV1_COMMON *cm, BufferPool *const pool) { // Scale up and back into frame_to_show. assert(frame_to_show->y_crop_width != cm->width); - assert(frame_to_show->y_crop_height != cm->height); + assert(IMPLIES(!CONFIG_HORZONLY_FRAME_SUPERRES, + frame_to_show->y_crop_height != cm->height)); #if CONFIG_HIGHBITDEPTH - av1_resize_and_extend_frame(©_buffer, frame_to_show, (int)cm->bit_depth); + av1_upscale_normative_and_extend_frame(©_buffer, frame_to_show, + (int)cm->bit_depth); #else - av1_resize_and_extend_frame(©_buffer, frame_to_show); + av1_upscale_normative_and_extend_frame(©_buffer, frame_to_show); #endif // CONFIG_HIGHBITDEPTH // Free the copy buffer -- cgit v1.2.3