diff options
28 files changed, 169 insertions, 1278 deletions
diff --git a/vp9/common/vp9_blockd.c b/vp9/common/vp9_blockd.c index 4ae8132bb..1eda3cc38 100644 --- a/vp9/common/vp9_blockd.c +++ b/vp9/common/vp9_blockd.c @@ -12,15 +12,15 @@ #include "vp9/common/vp9_blockd.h" #include "vpx_mem/vpx_mem.h" -const uint8_t vp9_block2left[TX_SIZE_MAX_SB][25] = { - {0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8}, - {0, 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 4, 4, 4, 4, 6, 6, 6, 6, 8}, - {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 6, 6, 6, 6, 8}, - {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 6, 6, 6, 6, 8} +const uint8_t vp9_block2left[TX_SIZE_MAX_SB][24] = { + {0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7}, + {0, 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 4, 4, 4, 4, 6, 6, 6, 6}, + {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 6, 6, 6, 6}, + {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 6, 6, 6, 6} }; -const uint8_t vp9_block2above[TX_SIZE_MAX_SB][25] = { - {0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 4, 5, 4, 5, 6, 7, 6, 7, 8}, - {0, 0, 0, 0, 2, 2, 2, 2, 0, 0, 0, 0, 2, 2, 2, 2, 4, 4, 4, 4, 6, 6, 6, 6, 8}, - {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 6, 6, 6, 6, 8}, - {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 6, 6, 6, 6, 8} +const uint8_t vp9_block2above[TX_SIZE_MAX_SB][24] = { + {0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 4, 5, 4, 5, 6, 7, 6, 7}, + {0, 0, 0, 0, 2, 2, 2, 2, 0, 0, 0, 0, 2, 2, 2, 2, 4, 4, 4, 4, 6, 6, 6, 6}, + {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 6, 6, 6, 6}, + {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 6, 6, 6, 6} }; diff --git a/vp9/common/vp9_blockd.h b/vp9/common/vp9_blockd.h index 054d58dba..d77bcb07b 100644 --- a/vp9/common/vp9_blockd.h +++ b/vp9/common/vp9_blockd.h @@ -52,8 +52,8 @@ typedef struct { } POS; typedef enum PlaneType { - PLANE_TYPE_Y_NO_DC = 0, - PLANE_TYPE_Y2, + PLANE_TYPE_Y_NO_DC_UNUSED = 0, + PLANE_TYPE_Y2_UNUSED, PLANE_TYPE_UV, PLANE_TYPE_Y_WITH_DC, } PLANE_TYPE; @@ -63,7 +63,6 @@ typedef struct { ENTROPY_CONTEXT y1[4]; ENTROPY_CONTEXT u[2]; ENTROPY_CONTEXT v[2]; - ENTROPY_CONTEXT y2; } ENTROPY_CONTEXT_PLANES; #define VP9_COMBINEENTROPYCONTEXTS( Dest, A, B) \ @@ -144,7 +143,6 @@ typedef enum { #if CONFIG_LOSSLESS #define WHT_UPSCALE_FACTOR 3 -#define Y2_WHT_UPSCALE_FACTOR 2 #endif typedef enum { @@ -288,23 +286,23 @@ typedef struct blockd { } BLOCKD; typedef struct superblockd { - /* 32x32 Y and 16x16 U/V. No 2nd order transform yet. */ + /* 32x32 Y and 16x16 U/V */ DECLARE_ALIGNED(16, int16_t, diff[32*32+16*16*2]); DECLARE_ALIGNED(16, int16_t, qcoeff[32*32+16*16*2]); DECLARE_ALIGNED(16, int16_t, dqcoeff[32*32+16*16*2]); } SUPERBLOCKD; typedef struct macroblockd { - DECLARE_ALIGNED(16, int16_t, diff[400]); /* from idct diff */ + DECLARE_ALIGNED(16, int16_t, diff[384]); /* from idct diff */ DECLARE_ALIGNED(16, uint8_t, predictor[384]); - DECLARE_ALIGNED(16, int16_t, qcoeff[400]); - DECLARE_ALIGNED(16, int16_t, dqcoeff[400]); - DECLARE_ALIGNED(16, uint16_t, eobs[25]); + DECLARE_ALIGNED(16, int16_t, qcoeff[384]); + DECLARE_ALIGNED(16, int16_t, dqcoeff[384]); + DECLARE_ALIGNED(16, uint16_t, eobs[24]); SUPERBLOCKD sb_coeff_data; - /* 16 Y blocks, 4 U, 4 V, 1 DC 2nd order block, each with 16 entries. */ - BLOCKD block[25]; + /* 16 Y blocks, 4 U, 4 V, each with 16 entries. */ + BLOCKD block[24]; int fullpixel_mask; YV12_BUFFER_CONFIG pre; /* Filtered copy of previous frame reconstruction */ @@ -321,7 +319,7 @@ typedef struct macroblockd { int left_available; int right_available; - /* Y,U,V,Y2 */ + /* Y,U,V */ ENTROPY_CONTEXT_PLANES *above_context; ENTROPY_CONTEXT_PLANES *left_context; @@ -377,17 +375,10 @@ typedef struct macroblockd { /* Inverse transform function pointers. */ void (*inv_txm4x4_1)(int16_t *input, int16_t *output, int pitch); void (*inv_txm4x4)(int16_t *input, int16_t *output, int pitch); - void (*inv_2ndtxm4x4_1)(int16_t *in, int16_t *out); - void (*inv_2ndtxm4x4)(int16_t *in, int16_t *out); void (*itxm_add)(int16_t *input, const int16_t *dq, uint8_t *pred, uint8_t *output, int pitch, int stride); - void (*dc_itxm_add)(int16_t *input, const int16_t *dq, - uint8_t *pred, uint8_t *output, int pitch, int stride, int dc); void (*dc_only_itxm_add)(int input_dc, uint8_t *pred_ptr, uint8_t *dst_ptr, int pitch, int stride); - void (*dc_itxm_add_y_block)(int16_t *q, const int16_t *dq, - uint8_t *pre, uint8_t *dst, int stride, uint16_t *eobs, - const int16_t *dc); void (*itxm_add_y_block)(int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dst, int stride, uint16_t *eobs); void (*itxm_add_uv_block)(int16_t *q, const int16_t *dq, @@ -488,8 +479,8 @@ static TX_TYPE txfm_map(B_PREDICTION_MODE bmode) { return tx_type; } -extern const uint8_t vp9_block2left[TX_SIZE_MAX_SB][25]; -extern const uint8_t vp9_block2above[TX_SIZE_MAX_SB][25]; +extern const uint8_t vp9_block2left[TX_SIZE_MAX_SB][24]; +extern const uint8_t vp9_block2above[TX_SIZE_MAX_SB][24]; #define USE_ADST_FOR_I16X16_8X8 0 #define USE_ADST_FOR_I16X16_4X4 0 @@ -623,20 +614,6 @@ static TX_TYPE get_tx_type(const MACROBLOCKD *xd, const BLOCKD *b) { return tx_type; } -static int get_2nd_order_usage(const MACROBLOCKD *xd) { -#if 1 - return 0; -#else - int has_2nd_order = (xd->mode_info_context->mbmi.mode != SPLITMV && - xd->mode_info_context->mbmi.mode != I8X8_PRED && - xd->mode_info_context->mbmi.mode != B_PRED && - xd->mode_info_context->mbmi.txfm_size != TX_16X16); - if (has_2nd_order) - has_2nd_order = (get_tx_type(xd, xd->block) == DCT_DCT); - return has_2nd_order; -#endif -} - extern void vp9_build_block_doffsets(MACROBLOCKD *xd); extern void vp9_setup_block_dptrs(MACROBLOCKD *xd); diff --git a/vp9/common/vp9_idctllm.c b/vp9/common/vp9_idctllm.c index 2fec98e50..263b98597 100644 --- a/vp9/common/vp9_idctllm.c +++ b/vp9/common/vp9_idctllm.c @@ -298,122 +298,7 @@ void vp9_ihtllm_c(const int16_t *input, int16_t *output, int pitch, } } -void vp9_short_inv_walsh4x4_c(int16_t *input, int16_t *output) { - int i; - int a1, b1, c1, d1; - int16_t *ip = input; - int16_t *op = output; - - for (i = 0; i < 4; i++) { - a1 = ((ip[0] + ip[3])); - b1 = ((ip[1] + ip[2])); - c1 = ((ip[1] - ip[2])); - d1 = ((ip[0] - ip[3])); - - op[0] = (a1 + b1 + 1) >> 1; - op[1] = (c1 + d1) >> 1; - op[2] = (a1 - b1) >> 1; - op[3] = (d1 - c1) >> 1; - - ip += 4; - op += 4; - } - - ip = output; - op = output; - for (i = 0; i < 4; i++) { - a1 = ip[0] + ip[12]; - b1 = ip[4] + ip[8]; - c1 = ip[4] - ip[8]; - d1 = ip[0] - ip[12]; - op[0] = (a1 + b1 + 1) >> 1; - op[4] = (c1 + d1) >> 1; - op[8] = (a1 - b1) >> 1; - op[12] = (d1 - c1) >> 1; - ip++; - op++; - } -} - -void vp9_short_inv_walsh4x4_1_c(int16_t *in, int16_t *out) { - int i; - int16_t tmp[4]; - int16_t *ip = in; - int16_t *op = tmp; - - op[0] = (ip[0] + 1) >> 1; - op[1] = op[2] = op[3] = (ip[0] >> 1); - - ip = tmp; - op = out; - for (i = 0; i < 4; i++) { - op[0] = (ip[0] + 1) >> 1; - op[4] = op[8] = op[12] = (ip[0] >> 1); - ip++; - op++; - } -} - #if CONFIG_LOSSLESS -void vp9_short_inv_walsh4x4_lossless_c(int16_t *input, int16_t *output) { - int i; - int a1, b1, c1, d1; - int16_t *ip = input; - int16_t *op = output; - - for (i = 0; i < 4; i++) { - a1 = ((ip[0] + ip[3])) >> Y2_WHT_UPSCALE_FACTOR; - b1 = ((ip[1] + ip[2])) >> Y2_WHT_UPSCALE_FACTOR; - c1 = ((ip[1] - ip[2])) >> Y2_WHT_UPSCALE_FACTOR; - d1 = ((ip[0] - ip[3])) >> Y2_WHT_UPSCALE_FACTOR; - - op[0] = (a1 + b1 + 1) >> 1; - op[1] = (c1 + d1) >> 1; - op[2] = (a1 - b1) >> 1; - op[3] = (d1 - c1) >> 1; - - ip += 4; - op += 4; - } - - ip = output; - op = output; - for (i = 0; i < 4; i++) { - a1 = ip[0] + ip[12]; - b1 = ip[4] + ip[8]; - c1 = ip[4] - ip[8]; - d1 = ip[0] - ip[12]; - - - op[0] = ((a1 + b1 + 1) >> 1) << Y2_WHT_UPSCALE_FACTOR; - op[4] = ((c1 + d1) >> 1) << Y2_WHT_UPSCALE_FACTOR; - op[8] = ((a1 - b1) >> 1) << Y2_WHT_UPSCALE_FACTOR; - op[12] = ((d1 - c1) >> 1) << Y2_WHT_UPSCALE_FACTOR; - - ip++; - op++; - } -} - -void vp9_short_inv_walsh4x4_1_lossless_c(int16_t *in, int16_t *out) { - int i; - int16_t tmp[4]; - int16_t *ip = in; - int16_t *op = tmp; - - op[0] = ((ip[0] >> Y2_WHT_UPSCALE_FACTOR) + 1) >> 1; - op[1] = op[2] = op[3] = ((ip[0] >> Y2_WHT_UPSCALE_FACTOR) >> 1); - - ip = tmp; - op = out; - for (i = 0; i < 4; i++) { - op[0] = ((ip[0] + 1) >> 1) << Y2_WHT_UPSCALE_FACTOR; - op[4] = op[8] = op[12] = ((ip[0] >> 1)) << Y2_WHT_UPSCALE_FACTOR; - ip++; - op++; - } -} - void vp9_short_inv_walsh4x4_x8_c(int16_t *input, int16_t *output, int pitch) { int i; int a1, b1, c1, d1; @@ -913,20 +798,6 @@ void vp9_short_idct1_8x8_c(int16_t *input, int16_t *output) { *output = (out + 16) >> 5; } -void vp9_short_ihaar2x2_c(int16_t *input, int16_t *output, int pitch) { - int i; - int16_t *ip = input; // 0, 1, 4, 8 - int16_t *op = output; - for (i = 0; i < 16; i++) { - op[i] = 0; - } - - op[0] = (ip[0] + ip[1] + ip[4] + ip[8] + 1) >> 1; - op[1] = (ip[0] - ip[1] + ip[4] - ip[8]) >> 1; - op[4] = (ip[0] + ip[1] - ip[4] - ip[8]) >> 1; - op[8] = (ip[0] - ip[1] - ip[4] + ip[8]) >> 1; -} - void idct16_1d(int16_t *input, int16_t *output) { int16_t step1[16], step2[16]; int temp1, temp2; diff --git a/vp9/common/vp9_invtrans.c b/vp9/common/vp9_invtrans.c index cb9a3db63..591cea8a2 100644 --- a/vp9/common/vp9_invtrans.c +++ b/vp9/common/vp9_invtrans.c @@ -11,24 +11,6 @@ #include "vp9/common/vp9_invtrans.h" #include "./vp9_rtcd.h" -static void recon_dcblock(MACROBLOCKD *xd) { - BLOCKD *b = &xd->block[24]; - int i; - - for (i = 0; i < 16; i++) { - xd->block[i].dqcoeff[0] = b->diff[i]; - } -} - -static void recon_dcblock_8x8(MACROBLOCKD *xd) { - BLOCKD *b = &xd->block[24]; // for coeff 0, 2, 8, 10 - - xd->block[0].dqcoeff[0] = b->diff[0]; - xd->block[4].dqcoeff[0] = b->diff[1]; - xd->block[8].dqcoeff[0] = b->diff[4]; - xd->block[12].dqcoeff[0] = b->diff[8]; -} - void vp9_inverse_transform_b_4x4(MACROBLOCKD *xd, int block, int pitch) { BLOCKD *b = &xd->block[block]; if (b->eob <= 1) @@ -39,14 +21,6 @@ void vp9_inverse_transform_b_4x4(MACROBLOCKD *xd, int block, int pitch) { void vp9_inverse_transform_mby_4x4(MACROBLOCKD *xd) { int i; - BLOCKD *blockd = xd->block; - int has_2nd_order = get_2nd_order_usage(xd); - - if (has_2nd_order) { - /* do 2nd order transform on the dc block */ - xd->inv_2ndtxm4x4(blockd[24].dqcoeff, blockd[24].diff); - recon_dcblock(xd); - } for (i = 0; i < 16; i++) { TX_TYPE tx_type = get_tx_type_4x4(xd, &xd->block[i]); @@ -85,13 +59,6 @@ void vp9_inverse_transform_b_8x8(int16_t *input_dqcoeff, int16_t *output_coeff, void vp9_inverse_transform_mby_8x8(MACROBLOCKD *xd) { int i; BLOCKD *blockd = xd->block; - int has_2nd_order = get_2nd_order_usage(xd); - - if (has_2nd_order) { - // do 2nd order transform on the dc block - vp9_short_ihaar2x2(blockd[24].dqcoeff, blockd[24].diff, 8); - recon_dcblock_8x8(xd); // need to change for 8x8 - } for (i = 0; i < 9; i += 8) { TX_TYPE tx_type = get_tx_type_8x8(xd, &xd->block[i]); diff --git a/vp9/common/vp9_mbpitch.c b/vp9/common/vp9_mbpitch.c index e94144813..ed96292a4 100644 --- a/vp9/common/vp9_mbpitch.c +++ b/vp9/common/vp9_mbpitch.c @@ -102,9 +102,7 @@ void vp9_setup_block_dptrs(MACROBLOCKD *xd) { } } - blockd[24].diff = &xd->diff[384]; - - for (r = 0; r < 25; r++) { + for (r = 0; r < 24; r++) { blockd[r].qcoeff = xd->qcoeff + r * 16; blockd[r].dqcoeff = xd->dqcoeff + r * 16; } diff --git a/vp9/common/vp9_onyxc_int.h b/vp9/common/vp9_onyxc_int.h index 6295514ea..8329d739f 100644 --- a/vp9/common/vp9_onyxc_int.h +++ b/vp9/common/vp9_onyxc_int.h @@ -133,7 +133,6 @@ typedef struct VP9Common { struct vpx_internal_error_info error; DECLARE_ALIGNED(16, int16_t, Y1dequant[QINDEX_RANGE][16]); - DECLARE_ALIGNED(16, int16_t, Y2dequant[QINDEX_RANGE][16]); DECLARE_ALIGNED(16, int16_t, UVdequant[QINDEX_RANGE][16]); int Width; @@ -184,8 +183,8 @@ typedef struct VP9Common { int last_kf_gf_q; /* Q used on the last GF or KF */ int y1dc_delta_q; - int y2dc_delta_q; - int y2ac_delta_q; + int y2dc_delta_q_unused; + int y2ac_delta_q_unused; int uvdc_delta_q; int uvac_delta_q; @@ -217,7 +216,7 @@ typedef struct VP9Common { int ref_frame_sign_bias[MAX_REF_FRAMES]; /* Two state 0, 1 */ - /* Y,U,V,Y2 */ + /* Y,U,V */ ENTROPY_CONTEXT_PLANES *above_context; /* row of context for each plane */ ENTROPY_CONTEXT_PLANES left_context[4]; /* (up to) 4 contexts "" */ diff --git a/vp9/common/vp9_rtcd_defs.sh b/vp9/common/vp9_rtcd_defs.sh index 113df4665..a8fd9a90f 100644 --- a/vp9/common/vp9_rtcd_defs.sh +++ b/vp9/common/vp9_rtcd_defs.sh @@ -29,12 +29,6 @@ forward_decls vp9_common_forward_decls prototype void vp9_dequantize_b "struct blockd *x" specialize vp9_dequantize_b -prototype void vp9_dequantize_b_2x2 "struct blockd *x" -specialize vp9_dequantize_b_2x2 - -prototype void vp9_dequant_dc_idct_add_y_block_8x8 "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dst, int stride, uint16_t *eobs, const int16_t *dc, struct macroblockd *xd" -specialize vp9_dequant_dc_idct_add_y_block_8x8 - prototype void vp9_dequant_idct_add_y_block_8x8 "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dst, int stride, uint16_t *eobs, struct macroblockd *xd" specialize vp9_dequant_idct_add_y_block_8x8 @@ -44,18 +38,12 @@ specialize vp9_dequant_idct_add_uv_block_8x8 prototype void vp9_dequant_idct_add_16x16 "int16_t *input, const int16_t *dq, uint8_t *pred, uint8_t *dest, int pitch, int stride, int eob" specialize vp9_dequant_idct_add_16x16 -prototype void vp9_dequant_idct_add_8x8 "int16_t *input, const int16_t *dq, uint8_t *pred, uint8_t *dest, int pitch, int stride, int dc, int eob" +prototype void vp9_dequant_idct_add_8x8 "int16_t *input, const int16_t *dq, uint8_t *pred, uint8_t *dest, int pitch, int stride, int eob" specialize vp9_dequant_idct_add_8x8 prototype void vp9_dequant_idct_add "int16_t *input, const int16_t *dq, uint8_t *pred, uint8_t *dest, int pitch, int stride" specialize vp9_dequant_idct_add -prototype void vp9_dequant_dc_idct_add "int16_t *input, const int16_t *dq, uint8_t *pred, uint8_t *dest, int pitch, int stride, int dc" -specialize vp9_dequant_dc_idct_add - -prototype void vp9_dequant_dc_idct_add_y_block "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dst, int stride, uint16_t *eobs, const int16_t *dcs" -specialize vp9_dequant_dc_idct_add_y_block - prototype void vp9_dequant_idct_add_y_block "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dst, int stride, uint16_t *eobs" specialize vp9_dequant_idct_add_y_block @@ -280,9 +268,6 @@ specialize vp9_short_idct10_8x8 prototype void vp9_short_idct1_8x8 "int16_t *input, int16_t *output" specialize vp9_short_idct1_8x8 -prototype void vp9_short_ihaar2x2 "int16_t *input, int16_t *output, int pitch" -specialize vp9_short_ihaar2x2 - prototype void vp9_short_idct16x16 "int16_t *input, int16_t *output, int pitch" specialize vp9_short_idct16x16 @@ -312,15 +297,6 @@ specialize vp9_short_iht4x4 prototype void vp9_ihtllm "const int16_t *input, int16_t *output, int pitch, int tx_type, int tx_dim, int16_t eobs" specialize vp9_ihtllm -# -# 2nd order -# -prototype void vp9_short_inv_walsh4x4_1 "int16_t *in, int16_t *out" -specialize vp9_short_inv_walsh4x4_1 - -prototype void vp9_short_inv_walsh4x4 "int16_t *in, int16_t *out" -specialize vp9_short_inv_walsh4x4_ - # dct and add @@ -334,10 +310,6 @@ prototype void vp9_short_inv_walsh4x4_x8 "int16_t *input, int16_t *output, int p specialize vp9_short_inv_walsh4x4_x8 prototype void vp9_dc_only_inv_walsh_add "int input_dc, uint8_t *pred_ptr, uint8_t *dst_ptr, int pitch, int stride" specialize vp9_dc_only_inv_walsh_add -prototype void vp9_short_inv_walsh4x4_1_lossless "int16_t *in, int16_t *out" -specialize vp9_short_inv_walsh4x4_1_lossless -prototype void vp9_short_inv_walsh4x4_lossless "int16_t *in, int16_t *out" -specialize vp9_short_inv_walsh4x4_lossless fi prototype unsigned int vp9_sad32x3 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, int max_sad" @@ -552,7 +524,7 @@ specialize vp9_sub_pixel_mse32x32 prototype unsigned int vp9_get_mb_ss "const int16_t *" specialize vp9_get_mb_ss mmx sse2 # ENCODEMB INVOKE -prototype int vp9_mbblock_error "struct macroblock *mb, int dc" +prototype int vp9_mbblock_error "struct macroblock *mb" specialize vp9_mbblock_error mmx sse2 vp9_mbblock_error_sse2=vp9_mbblock_error_xmm @@ -596,27 +568,18 @@ specialize vp9_fht prototype void vp9_short_fdct8x8 "int16_t *InputData, int16_t *OutputData, int pitch" specialize vp9_short_fdct8x8 -prototype void vp9_short_fhaar2x2 "int16_t *InputData, int16_t *OutputData, int pitch" -specialize vp9_short_fhaar2x2 - prototype void vp9_short_fdct4x4 "int16_t *InputData, int16_t *OutputData, int pitch" specialize vp9_short_fdct4x4 prototype void vp9_short_fdct8x4 "int16_t *InputData, int16_t *OutputData, int pitch" specialize vp9_short_fdct8x4 -prototype void vp9_short_walsh4x4 "int16_t *InputData, int16_t *OutputData, int pitch" -specialize vp9_short_walsh4x4 - prototype void vp9_short_fdct32x32 "int16_t *InputData, int16_t *OutputData, int pitch" specialize vp9_short_fdct32x32 prototype void vp9_short_fdct16x16 "int16_t *InputData, int16_t *OutputData, int pitch" specialize vp9_short_fdct16x16 -prototype void vp9_short_walsh4x4_lossless "int16_t *InputData, int16_t *OutputData, int pitch" -specialize vp9_short_walsh4x4_lossless - prototype void vp9_short_walsh4x4_x8 "int16_t *InputData, int16_t *OutputData, int pitch" specialize vp9_short_walsh4x4_x8 diff --git a/vp9/decoder/vp9_decodframe.c b/vp9/decoder/vp9_decodframe.c index facd761f0..abc9bf6b9 100644 --- a/vp9/decoder/vp9_decodframe.c +++ b/vp9/decoder/vp9_decodframe.c @@ -83,7 +83,6 @@ void vp9_init_de_quantizer(VP9D_COMP *pbi) { for (Q = 0; Q < QINDEX_RANGE; Q++) { pc->Y1dequant[Q][0] = (int16_t)vp9_dc_quant(Q, pc->y1dc_delta_q); - pc->Y2dequant[Q][0] = (int16_t)vp9_dc2quant(Q, pc->y2dc_delta_q); pc->UVdequant[Q][0] = (int16_t)vp9_dc_uv_quant(Q, pc->uvdc_delta_q); /* all the ac values =; */ @@ -91,7 +90,6 @@ void vp9_init_de_quantizer(VP9D_COMP *pbi) { int rc = vp9_default_zig_zag1d_4x4[i]; pc->Y1dequant[Q][rc] = (int16_t)vp9_ac_yquant(Q); - pc->Y2dequant[Q][rc] = (int16_t)vp9_ac2quant(Q, pc->y2ac_delta_q); pc->UVdequant[Q][rc] = (int16_t)vp9_ac_uv_quant(Q, pc->uvac_delta_q); } } @@ -126,12 +124,8 @@ static void mb_init_dequantizer(VP9D_COMP *pbi, MACROBLOCKD *xd) { xd->inv_txm4x4_1 = vp9_short_idct4x4llm_1; xd->inv_txm4x4 = vp9_short_idct4x4llm; - xd->inv_2ndtxm4x4_1 = vp9_short_inv_walsh4x4_1; - xd->inv_2ndtxm4x4 = vp9_short_inv_walsh4x4; xd->itxm_add = vp9_dequant_idct_add; xd->dc_only_itxm_add = vp9_dc_only_idct_add_c; - xd->dc_itxm_add = vp9_dequant_dc_idct_add; - xd->dc_itxm_add_y_block = vp9_dequant_dc_idct_add_y_block; xd->itxm_add_y_block = vp9_dequant_idct_add_y_block; xd->itxm_add_uv_block = vp9_dequant_idct_add_uv_block; #if CONFIG_LOSSLESS @@ -139,12 +133,8 @@ static void mb_init_dequantizer(VP9D_COMP *pbi, MACROBLOCKD *xd) { assert(QIndex == 0); xd->inv_txm4x4_1 = vp9_short_inv_walsh4x4_1_x8; xd->inv_txm4x4 = vp9_short_inv_walsh4x4_x8; - xd->inv_2ndtxm4x4_1 = vp9_short_inv_walsh4x4_1_lossless; - xd->inv_2ndtxm4x4 = vp9_short_inv_walsh4x4_lossless; xd->itxm_add = vp9_dequant_idct_add_lossless_c; xd->dc_only_itxm_add = vp9_dc_only_inv_walsh_add_c; - xd->dc_itxm_add = vp9_dequant_dc_idct_add_lossless_c; - xd->dc_itxm_add_y_block = vp9_dequant_dc_idct_add_y_block_lossless_c; xd->itxm_add_y_block = vp9_dequant_idct_add_y_block_lossless_c; xd->itxm_add_uv_block = vp9_dequant_idct_add_uv_block_lossless_c; } @@ -153,9 +143,6 @@ static void mb_init_dequantizer(VP9D_COMP *pbi, MACROBLOCKD *xd) { for (i = 16; i < 24; i++) { xd->block[i].dequant = pc->UVdequant[QIndex]; } - - xd->block[24].dequant = pc->Y2dequant[QIndex]; - } /* skip_recon_mb() is Modified: Instead of writing the result to predictor buffer and then copying it @@ -222,7 +209,6 @@ static void decode_16x16(VP9D_COMP *pbi, MACROBLOCKD *xd, BOOL_DECODER* const bc) { BLOCKD *bd = &xd->block[0]; TX_TYPE tx_type = get_tx_type_16x16(xd, bd); - assert(get_2nd_order_usage(xd) == 0); #ifdef DEC_DEBUG if (dec_debug) { int i; @@ -274,7 +260,6 @@ static void decode_8x8(VP9D_COMP *pbi, MACROBLOCKD *xd, #endif if (tx_type != DCT_DCT || xd->mode_info_context->mbmi.mode == I8X8_PRED) { int i; - assert(get_2nd_order_usage(xd) == 0); for (i = 0; i < 4; i++) { int ib = vp9_i8x8_block[i]; int idx = (ib & 0x02) ? (ib + 2) : ib; @@ -294,39 +279,16 @@ static void decode_8x8(VP9D_COMP *pbi, MACROBLOCKD *xd, xd->eobs[idx]); } else { vp9_dequant_idct_add_8x8_c(q, dq, pre, dst, 16, stride, - 0, xd->eobs[idx]); + xd->eobs[idx]); } } - } else if (xd->mode_info_context->mbmi.mode == SPLITMV || - get_2nd_order_usage(xd) == 0) { - assert(get_2nd_order_usage(xd) == 0); + } else { vp9_dequant_idct_add_y_block_8x8(xd->qcoeff, xd->block[0].dequant, xd->predictor, xd->dst.y_buffer, xd->dst.y_stride, xd->eobs, xd); - } else { - BLOCKD *b = &xd->block[24]; - assert(get_2nd_order_usage(xd) == 1); - vp9_dequantize_b_2x2(b); - vp9_short_ihaar2x2(&b->dqcoeff[0], b->diff, 8); - ((int *)b->qcoeff)[0] = 0; // 2nd order block are set to 0 after idct - ((int *)b->qcoeff)[1] = 0; - ((int *)b->qcoeff)[2] = 0; - ((int *)b->qcoeff)[3] = 0; - ((int *)b->qcoeff)[4] = 0; - ((int *)b->qcoeff)[5] = 0; - ((int *)b->qcoeff)[6] = 0; - ((int *)b->qcoeff)[7] = 0; - vp9_dequant_dc_idct_add_y_block_8x8(xd->qcoeff, - xd->block[0].dequant, - xd->predictor, - xd->dst.y_buffer, - xd->dst.y_stride, - xd->eobs, - xd->block[24].diff, - xd); } // Now do UV @@ -374,7 +336,6 @@ static void decode_4x4(VP9D_COMP *pbi, MACROBLOCKD *xd, int i, eobtotal = 0; MB_PREDICTION_MODE mode = xd->mode_info_context->mbmi.mode; if (mode == I8X8_PRED) { - assert(get_2nd_order_usage(xd) == 0); for (i = 0; i < 4; i++) { int ib = vp9_i8x8_block[i]; const int iblock[4] = {0, 1, 4, 5}; @@ -407,7 +368,6 @@ static void decode_4x4(VP9D_COMP *pbi, MACROBLOCKD *xd, *(b->base_dst) + b->dst, 8, b->dst_stride); } } else if (mode == B_PRED) { - assert(get_2nd_order_usage(xd) == 0); for (i = 0; i < 16; i++) { int b_mode; BLOCKD *b = &xd->block[i]; @@ -434,8 +394,6 @@ static void decode_4x4(VP9D_COMP *pbi, MACROBLOCKD *xd, if (!xd->mode_info_context->mbmi.mb_skip_coeff) { vp9_decode_mb_tokens_4x4_uv(pbi, xd, bc); } - xd->above_context->y2 = 0; - xd->left_context->y2 = 0; vp9_build_intra_predictors_mbuv(xd); xd->itxm_add_uv_block(xd->qcoeff + 16 * 16, xd->block[16].dequant, @@ -444,8 +402,7 @@ static void decode_4x4(VP9D_COMP *pbi, MACROBLOCKD *xd, xd->dst.v_buffer, xd->dst.uv_stride, xd->eobs + 16); - } else if (mode == SPLITMV || get_2nd_order_usage(xd) == 0) { - assert(get_2nd_order_usage(xd) == 0); + } else if (mode == SPLITMV || get_tx_type_4x4(xd, &xd->block[0]) == DCT_DCT) { xd->itxm_add_y_block(xd->qcoeff, xd->block[0].dequant, xd->predictor, @@ -477,48 +434,18 @@ static void decode_4x4(VP9D_COMP *pbi, MACROBLOCKD *xd, } } #endif - tx_type = get_tx_type_4x4(xd, &xd->block[0]); - if (tx_type != DCT_DCT) { - assert(get_2nd_order_usage(xd) == 0); - for (i = 0; i < 16; i++) { - BLOCKD *b = &xd->block[i]; - tx_type = get_tx_type_4x4(xd, b); - if (tx_type != DCT_DCT) { - vp9_ht_dequant_idct_add_c(tx_type, b->qcoeff, - b->dequant, b->predictor, - *(b->base_dst) + b->dst, 16, - b->dst_stride, b->eob); - } else { - xd->itxm_add(b->qcoeff, b->dequant, b->predictor, - *(b->base_dst) + b->dst, 16, b->dst_stride); - } - } - } else { - BLOCKD *b = &xd->block[24]; - assert(get_2nd_order_usage(xd) == 1); - vp9_dequantize_b(b); - if (xd->eobs[24] > 1) { - xd->inv_2ndtxm4x4(&b->dqcoeff[0], b->diff); - ((int *)b->qcoeff)[0] = 0; - ((int *)b->qcoeff)[1] = 0; - ((int *)b->qcoeff)[2] = 0; - ((int *)b->qcoeff)[3] = 0; - ((int *)b->qcoeff)[4] = 0; - ((int *)b->qcoeff)[5] = 0; - ((int *)b->qcoeff)[6] = 0; - ((int *)b->qcoeff)[7] = 0; + for (i = 0; i < 16; i++) { + BLOCKD *b = &xd->block[i]; + tx_type = get_tx_type_4x4(xd, b); + if (tx_type != DCT_DCT) { + vp9_ht_dequant_idct_add_c(tx_type, b->qcoeff, + b->dequant, b->predictor, + *(b->base_dst) + b->dst, 16, + b->dst_stride, b->eob); } else { - xd->inv_2ndtxm4x4_1(&b->dqcoeff[0], b->diff); - ((int *)b->qcoeff)[0] = 0; + xd->itxm_add(b->qcoeff, b->dequant, b->predictor, + *(b->base_dst) + b->dst, 16, b->dst_stride); } - vp9_dequantize_b(b); - xd->dc_itxm_add_y_block(xd->qcoeff, - xd->block[0].dequant, - xd->predictor, - xd->dst.y_buffer, - xd->dst.y_stride, - xd->eobs, - xd->block[24].diff); } xd->itxm_add_uv_block(xd->qcoeff + 16 * 16, xd->block[16].dequant, @@ -560,7 +487,6 @@ static void decode_8x8_sb(VP9D_COMP *pbi, MACROBLOCKD *xd, BOOL_DECODER* const bc, int n, int maska, int shiftb) { int x_idx = n & maska, y_idx = n >> shiftb; - BLOCKD *b = &xd->block[24]; TX_TYPE tx_type = get_tx_type_8x8(xd, &xd->block[0]); if (tx_type != DCT_DCT) { int i; @@ -587,24 +513,9 @@ static void decode_8x8_sb(VP9D_COMP *pbi, MACROBLOCKD *xd, + x_idx * 16 + (i & 1) * 8, xd->dst.y_buffer + (y_idx * 16 + (i / 2) * 8) * xd->dst.y_stride + x_idx * 16 + (i & 1) * 8, - stride, stride, 0, b->eob); + stride, stride, b->eob); } } - } else if (get_2nd_order_usage(xd) == 1) { - vp9_dequantize_b_2x2(b); - vp9_short_ihaar2x2(&b->dqcoeff[0], b->diff, 8); - ((int *)b->qcoeff)[0] = 0; // 2nd order block are set to 0 after idct - ((int *)b->qcoeff)[1] = 0; - ((int *)b->qcoeff)[2] = 0; - ((int *)b->qcoeff)[3] = 0; - ((int *)b->qcoeff)[4] = 0; - ((int *)b->qcoeff)[5] = 0; - ((int *)b->qcoeff)[6] = 0; - ((int *)b->qcoeff)[7] = 0; - vp9_dequant_dc_idct_add_y_block_8x8_inplace_c( - xd->qcoeff, xd->block[0].dequant, - xd->dst.y_buffer + y_idx * 16 * xd->dst.y_stride + x_idx * 16, - xd->dst.y_stride, xd->eobs, xd->block[24].diff, xd); } else { vp9_dequant_idct_add_y_block_8x8_inplace_c( xd->qcoeff, xd->block[0].dequant, @@ -622,7 +533,6 @@ static void decode_4x4_sb(VP9D_COMP *pbi, MACROBLOCKD *xd, BOOL_DECODER* const bc, int n, int maska, int shiftb) { int x_idx = n & maska, y_idx = n >> shiftb; - BLOCKD *b = &xd->block[24]; TX_TYPE tx_type = get_tx_type_4x4(xd, &xd->block[0]); if (tx_type != DCT_DCT) { int i; @@ -647,26 +557,6 @@ static void decode_4x4_sb(VP9D_COMP *pbi, MACROBLOCKD *xd, xd->dst.y_stride, xd->dst.y_stride); } } - } else if (get_2nd_order_usage(xd) == 1) { - vp9_dequantize_b(b); - if (xd->eobs[24] > 1) { - xd->inv_2ndtxm4x4(&b->dqcoeff[0], b->diff); - ((int *)b->qcoeff)[0] = 0; - ((int *)b->qcoeff)[1] = 0; - ((int *)b->qcoeff)[2] = 0; - ((int *)b->qcoeff)[3] = 0; - ((int *)b->qcoeff)[4] = 0; - ((int *)b->qcoeff)[5] = 0; - ((int *)b->qcoeff)[6] = 0; - ((int *)b->qcoeff)[7] = 0; - } else { - xd->inv_2ndtxm4x4_1(&b->dqcoeff[0], b->diff); - ((int *)b->qcoeff)[0] = 0; - } - vp9_dequant_dc_idct_add_y_block_4x4_inplace_c( - xd->qcoeff, xd->block[0].dequant, - xd->dst.y_buffer + y_idx * 16 * xd->dst.y_stride + x_idx * 16, - xd->dst.y_stride, xd->eobs, xd->block[24].diff, xd); } else { vp9_dequant_idct_add_y_block_4x4_inplace_c( xd->qcoeff, xd->block[0].dequant, @@ -780,7 +670,7 @@ static void decode_superblock64(VP9D_COMP *pbi, MACROBLOCKD *xd, xd->above_context = pc->above_context + mb_col + x_idx; xd->left_context = pc->left_context + y_idx; xd->mode_info_context = orig_mi + x_idx + y_idx * mis; - for (i = 0; i < 25; i++) { + for (i = 0; i < 24; i++) { xd->block[i].eob = 0; xd->eobs[i] = 0; } @@ -885,7 +775,7 @@ static void decode_superblock32(VP9D_COMP *pbi, MACROBLOCKD *xd, xd->above_context = pc->above_context + mb_col + x_idx; xd->left_context = pc->left_context + y_idx + (mb_row & 2); xd->mode_info_context = orig_mi + x_idx + y_idx * mis; - for (i = 0; i < 25; i++) { + for (i = 0; i < 24; i++) { xd->block[i].eob = 0; xd->eobs[i] = 0; } @@ -931,7 +821,7 @@ static void decode_macroblock(VP9D_COMP *pbi, MACROBLOCKD *xd, if (xd->mode_info_context->mbmi.mb_skip_coeff) { vp9_reset_mb_tokens_context(xd); } else if (!bool_error(bc)) { - for (i = 0; i < 25; i++) { + for (i = 0; i < 24; i++) { xd->block[i].eob = 0; xd->eobs[i] = 0; } @@ -1594,8 +1484,10 @@ int vp9_decode_frame(VP9D_COMP *pbi, const unsigned char **p_data_end) { q_update = 0; /* AC 1st order Q = default */ pc->y1dc_delta_q = get_delta_q(&header_bc, pc->y1dc_delta_q, &q_update); - pc->y2dc_delta_q = get_delta_q(&header_bc, pc->y2dc_delta_q, &q_update); - pc->y2ac_delta_q = get_delta_q(&header_bc, pc->y2ac_delta_q, &q_update); + pc->y2dc_delta_q_unused = get_delta_q(&header_bc, pc->y2dc_delta_q_unused, + &q_update); + pc->y2ac_delta_q_unused = get_delta_q(&header_bc, pc->y2ac_delta_q_unused, + &q_update); pc->uvdc_delta_q = get_delta_q(&header_bc, pc->uvdc_delta_q, &q_update); pc->uvac_delta_q = get_delta_q(&header_bc, pc->uvac_delta_q, &q_update); diff --git a/vp9/decoder/vp9_dequantize.c b/vp9/decoder/vp9_dequantize.c index 92a9df84c..fe1474ce9 100644 --- a/vp9/decoder/vp9_dequantize.c +++ b/vp9/decoder/vp9_dequantize.c @@ -185,20 +185,9 @@ void vp9_dequant_dc_idct_add_lossless_c(int16_t *input, const int16_t *dq, } #endif -void vp9_dequantize_b_2x2_c(BLOCKD *d) { - int i; - int16_t *DQ = d->dqcoeff; - const int16_t *Q = d->qcoeff; - const int16_t *DQC = d->dequant; - - for (i = 0; i < 16; i++) { - DQ[i] = (int16_t)((Q[i] * DQC[i])); - } -} - void vp9_dequant_idct_add_8x8_c(int16_t *input, const int16_t *dq, uint8_t *pred, uint8_t *dest, int pitch, - int stride, int dc, int eob) { + int stride, int eob) { int16_t output[64]; int16_t *diff_ptr = output; int i; @@ -206,8 +195,7 @@ void vp9_dequant_idct_add_8x8_c(int16_t *input, const int16_t *dq, /* If dc is 1, then input[0] is the reconstructed value, do not need * dequantization. Also, when dc is 1, dc is counted in eobs, namely eobs >=1. */ - if (!dc) - input[0] *= dq[0]; + input[0] *= dq[0]; /* The calculation can be simplified if there are not many non-zero dct * coefficients. Use eobs to decide what to do. diff --git a/vp9/decoder/vp9_detokenize.c b/vp9/decoder/vp9_detokenize.c index b6b6849f9..0786e2965 100644 --- a/vp9/decoder/vp9_detokenize.c +++ b/vp9/decoder/vp9_detokenize.c @@ -95,8 +95,7 @@ static int decode_coefs(VP9D_COMP *dx, const MACROBLOCKD *xd, ENTROPY_CONTEXT above_ec = A0[aidx] != 0, left_ec = L0[lidx] != 0; FRAME_CONTEXT *const fc = &dx->common.fc; int recent_energy = 0; - int nodc = (type == PLANE_TYPE_Y_NO_DC); - int pt, c = nodc; + int pt, c = 0; vp9_coeff_probs *coef_probs; vp9_prob *prob; vp9_coeff_count *coef_counts; @@ -121,10 +120,8 @@ static int decode_coefs(VP9D_COMP *dx, const MACROBLOCKD *xd, coef_counts = fc->hybrid_coef_counts_8x8; } #if CONFIG_CNVCONTEXT - if (type != PLANE_TYPE_Y2) { - above_ec = (A0[aidx] + A0[aidx + 1]) != 0; - left_ec = (L0[lidx] + L0[lidx + 1]) != 0; - } + above_ec = (A0[aidx] + A0[aidx + 1]) != 0; + left_ec = (L0[lidx] + L0[lidx + 1]) != 0; #endif break; case TX_16X16: @@ -141,7 +138,7 @@ static int decode_coefs(VP9D_COMP *dx, const MACROBLOCKD *xd, ENTROPY_CONTEXT *L1 = (ENTROPY_CONTEXT *) (xd->left_context + 1); above_ec = (A0[aidx] + A0[aidx + 1] + A1[aidx] + A1[aidx + 1]) != 0; left_ec = (L0[lidx] + L0[lidx + 1] + L1[lidx] + L1[lidx + 1]) != 0; - } else if (type != PLANE_TYPE_Y2) { + } else { above_ec = (A0[aidx] + A0[aidx + 1] + A0[aidx + 2] + A0[aidx + 3]) != 0; left_ec = (L0[lidx] + L0[lidx + 1] + L0[lidx + 2] + L0[lidx + 3]) != 0; } @@ -162,7 +159,7 @@ static int decode_coefs(VP9D_COMP *dx, const MACROBLOCKD *xd, A2[aidx] + A2[aidx + 1] + A3[aidx] + A3[aidx + 1]) != 0; left_ec = (L0[lidx] + L0[lidx + 1] + L1[lidx] + L1[lidx + 1] + L2[lidx] + L2[lidx + 1] + L3[lidx] + L3[lidx + 1]) != 0; - } else if (type != PLANE_TYPE_Y2) { + } else { ENTROPY_CONTEXT *A1 = (ENTROPY_CONTEXT *) (xd->above_context + 1); ENTROPY_CONTEXT *L1 = (ENTROPY_CONTEXT *) (xd->left_context + 1); above_ec = (A0[aidx] + A0[aidx + 1] + A0[aidx + 2] + A0[aidx + 3] + @@ -253,8 +250,8 @@ SKIP_START: if (c < seg_eob) coef_counts[type][get_coef_band(c)][pt][DCT_EOB_TOKEN]++; - A0[aidx] = L0[lidx] = (c > !type); - if (txfm_size >= TX_8X8 && type != PLANE_TYPE_Y2) { + A0[aidx] = L0[lidx] = (c > 0); + if (txfm_size >= TX_8X8) { A0[aidx + 1] = L0[lidx + 1] = A0[aidx]; if (txfm_size >= TX_16X16) { if (type == PLANE_TYPE_UV) { @@ -298,10 +295,6 @@ static int get_eob(MACROBLOCKD* const xd, int segment_id, int eob_max) { int vp9_decode_sb_tokens(VP9D_COMP* const pbi, MACROBLOCKD* const xd, BOOL_DECODER* const bc) { - ENTROPY_CONTEXT* const A0 = (ENTROPY_CONTEXT *) xd->above_context; - ENTROPY_CONTEXT* const L0 = (ENTROPY_CONTEXT *) xd->left_context; - ENTROPY_CONTEXT* const A1 = (ENTROPY_CONTEXT *) (xd->above_context + 1); - ENTROPY_CONTEXT* const L1 = (ENTROPY_CONTEXT *) (xd->left_context + 1); uint16_t *const eobs = xd->eobs; const int segment_id = xd->mode_info_context->mbmi.segment_id; int c, i, eobtotal = 0, seg_eob; @@ -324,17 +317,12 @@ int vp9_decode_sb_tokens(VP9D_COMP* const pbi, eobtotal += c; } - // no Y2 block - A0[8] = L0[8] = A1[8] = L1[8] = 0; - return eobtotal; } static int vp9_decode_mb_tokens_16x16(VP9D_COMP* const pbi, MACROBLOCKD* const xd, BOOL_DECODER* const bc) { - ENTROPY_CONTEXT* const A = (ENTROPY_CONTEXT *)xd->above_context; - ENTROPY_CONTEXT* const L = (ENTROPY_CONTEXT *)xd->left_context; uint16_t *const eobs = xd->eobs; const int segment_id = xd->mode_info_context->mbmi.segment_id; int c, i, eobtotal = 0, seg_eob; @@ -356,8 +344,6 @@ static int vp9_decode_mb_tokens_16x16(VP9D_COMP* const pbi, TX_8X8); eobtotal += c; } - A[8] = 0; - L[8] = 0; return eobtotal; } @@ -365,32 +351,14 @@ static int vp9_decode_mb_tokens_8x8(VP9D_COMP* const pbi, MACROBLOCKD* const xd, BOOL_DECODER* const bc) { uint16_t *const eobs = xd->eobs; - PLANE_TYPE type; int c, i, eobtotal = 0, seg_eob; const int segment_id = xd->mode_info_context->mbmi.segment_id; - int has_2nd_order = get_2nd_order_usage(xd); - // 2nd order DC block - if (has_2nd_order) { - eobs[24] = c = decode_coefs(pbi, xd, bc, 24, PLANE_TYPE_Y2, - DCT_DCT, get_eob(xd, segment_id, 4), - xd->block[24].qcoeff, - vp9_default_zig_zag1d_4x4, TX_8X8); - eobtotal += c - 4; - type = PLANE_TYPE_Y_NO_DC; - } else { - xd->above_context->y2 = 0; - xd->left_context->y2 = 0; - eobs[24] = 0; - type = PLANE_TYPE_Y_WITH_DC; - } - // luma blocks seg_eob = get_eob(xd, segment_id, 64); for (i = 0; i < 16; i += 4) { - eobs[i] = c = decode_coefs(pbi, xd, bc, i, type, - type == PLANE_TYPE_Y_WITH_DC ? - get_tx_type(xd, xd->block + i) : DCT_DCT, + eobs[i] = c = decode_coefs(pbi, xd, bc, i, PLANE_TYPE_Y_WITH_DC, + get_tx_type(xd, xd->block + i), seg_eob, xd->block[i].qcoeff, vp9_default_zig_zag1d_8x8, TX_8X8); eobtotal += c; @@ -492,26 +460,13 @@ static int vp9_decode_mb_tokens_4x4(VP9D_COMP* const dx, MACROBLOCKD* const xd, BOOL_DECODER* const bc) { int i, eobtotal = 0; - PLANE_TYPE type; const int segment_id = xd->mode_info_context->mbmi.segment_id; const int seg_eob = get_eob(xd, segment_id, 16); - const int has_2nd_order = get_2nd_order_usage(xd); - - // 2nd order DC block - if (has_2nd_order) { - eobtotal += decode_coefs_4x4(dx, xd, bc, PLANE_TYPE_Y2, 24, seg_eob, - DCT_DCT, vp9_default_zig_zag1d_4x4) - 16; - type = PLANE_TYPE_Y_NO_DC; - } else { - xd->above_context->y2 = 0; - xd->left_context->y2 = 0; - xd->eobs[24] = 0; - type = PLANE_TYPE_Y_WITH_DC; - } // luma blocks for (i = 0; i < 16; ++i) { - eobtotal += decode_coefs_4x4_y(dx, xd, bc, type, i, seg_eob); + eobtotal += decode_coefs_4x4_y(dx, xd, bc, + PLANE_TYPE_Y_WITH_DC, i, seg_eob); } // chroma blocks diff --git a/vp9/decoder/vp9_idct_blk.c b/vp9/decoder/vp9_idct_blk.c index 80b301931..a00ff7306 100644 --- a/vp9/decoder/vp9_idct_blk.c +++ b/vp9/decoder/vp9_idct_blk.c @@ -14,56 +14,6 @@ #include "vp9/decoder/vp9_dequantize.h" #endif -void vp9_dequant_dc_idct_add_y_block_c(int16_t *q, const int16_t *dq, - uint8_t *pre, - uint8_t *dst, - int stride, uint16_t *eobs, - const int16_t *dc) { - int i, j; - - for (i = 0; i < 4; i++) { - for (j = 0; j < 4; j++) { - if (*eobs++ > 1) - vp9_dequant_dc_idct_add_c(q, dq, pre, dst, 16, stride, dc[0]); - else - vp9_dc_only_idct_add_c(dc[0], pre, dst, 16, stride); - - q += 16; - pre += 4; - dst += 4; - dc++; - } - - pre += 64 - 16; - dst += 4 * stride - 16; - } -} - -void vp9_dequant_dc_idct_add_y_block_4x4_inplace_c(int16_t *q, - const int16_t *dq, - uint8_t *dst, - int stride, - uint16_t *eobs, - const int16_t *dc, - MACROBLOCKD *xd) { - int i, j; - - for (i = 0; i < 4; i++) { - for (j = 0; j < 4; j++) { - if (*eobs++ > 1) - xd->dc_itxm_add(q, dq, dst, dst, stride, stride, dc[0]); - else - xd->dc_only_itxm_add(dc[0], dst, dst, stride, stride); - - q += 16; - dst += 4; - dc++; - } - - dst += 4 * stride - 16; - } -} - void vp9_dequant_idct_add_y_block_4x4_inplace_c(int16_t *q, const int16_t *dq, uint8_t *dst, @@ -198,71 +148,23 @@ void vp9_dequant_idct_add_uv_block_4x4_inplace_c(int16_t *q, const int16_t *dq, } } -void vp9_dequant_dc_idct_add_y_block_8x8_c(int16_t *q, const int16_t *dq, - uint8_t *pre, - uint8_t *dst, - int stride, uint16_t *eobs, - const int16_t *dc, - MACROBLOCKD *xd) { - q[0] = dc[0]; - vp9_dequant_idct_add_8x8_c(q, dq, pre, dst, 16, stride, 1, xd->eobs[0]); - - q[64] = dc[1]; - vp9_dequant_idct_add_8x8_c(&q[64], dq, pre + 8, dst + 8, 16, stride, 1, - xd->eobs[4]); - - q[128] = dc[4]; - vp9_dequant_idct_add_8x8_c(&q[128], dq, pre + 8 * 16, - dst + 8 * stride, 16, stride, 1, xd->eobs[8]); - - q[192] = dc[8]; - vp9_dequant_idct_add_8x8_c(&q[192], dq, pre + 8 * 16 + 8, - dst + 8 * stride + 8, 16, stride, 1, - xd->eobs[12]); -} - -void vp9_dequant_dc_idct_add_y_block_8x8_inplace_c(int16_t *q, - const int16_t *dq, - uint8_t *dst, - int stride, - uint16_t *eobs, - const int16_t *dc, - MACROBLOCKD *xd) { - q[0] = dc[0]; - vp9_dequant_idct_add_8x8_c(q, dq, dst, dst, stride, stride, 1, xd->eobs[0]); - - q[64] = dc[1]; - vp9_dequant_idct_add_8x8_c(&q[64], dq, dst + 8, - dst + 8, stride, stride, 1, xd->eobs[4]); - - q[128] = dc[4]; - vp9_dequant_idct_add_8x8_c(&q[128], dq, dst + 8 * stride, - dst + 8 * stride, stride, stride, 1, - xd->eobs[8]); - - q[192] = dc[8]; - vp9_dequant_idct_add_8x8_c(&q[192], dq, dst + 8 * stride + 8, - dst + 8 * stride + 8, stride, stride, 1, - xd->eobs[12]); -} - void vp9_dequant_idct_add_y_block_8x8_inplace_c(int16_t *q, const int16_t *dq, uint8_t *dst, int stride, uint16_t *eobs, MACROBLOCKD *xd) { - vp9_dequant_idct_add_8x8_c(q, dq, dst, dst, stride, stride, 0, xd->eobs[0]); + vp9_dequant_idct_add_8x8_c(q, dq, dst, dst, stride, stride, xd->eobs[0]); vp9_dequant_idct_add_8x8_c(&q[64], dq, dst + 8, - dst + 8, stride, stride, 0, xd->eobs[4]); + dst + 8, stride, stride, xd->eobs[4]); vp9_dequant_idct_add_8x8_c(&q[128], dq, dst + 8 * stride, - dst + 8 * stride, stride, stride, 0, + dst + 8 * stride, stride, stride, xd->eobs[8]); vp9_dequant_idct_add_8x8_c(&q[192], dq, dst + 8 * stride + 8, - dst + 8 * stride + 8, stride, stride, 0, + dst + 8 * stride + 8, stride, stride, xd->eobs[12]); } @@ -274,13 +176,13 @@ void vp9_dequant_idct_add_y_block_8x8_c(int16_t *q, const int16_t *dq, uint8_t *origdest = dst; uint8_t *origpred = pre; - vp9_dequant_idct_add_8x8_c(q, dq, pre, dst, 16, stride, 0, xd->eobs[0]); + vp9_dequant_idct_add_8x8_c(q, dq, pre, dst, 16, stride, xd->eobs[0]); vp9_dequant_idct_add_8x8_c(&q[64], dq, origpred + 8, - origdest + 8, 16, stride, 0, xd->eobs[4]); + origdest + 8, 16, stride, xd->eobs[4]); vp9_dequant_idct_add_8x8_c(&q[128], dq, origpred + 8 * 16, - origdest + 8 * stride, 16, stride, 0, xd->eobs[8]); + origdest + 8 * stride, 16, stride, xd->eobs[8]); vp9_dequant_idct_add_8x8_c(&q[192], dq, origpred + 8 * 16 + 8, - origdest + 8 * stride + 8, 16, stride, 0, + origdest + 8 * stride + 8, 16, stride, xd->eobs[12]); } @@ -290,12 +192,12 @@ void vp9_dequant_idct_add_uv_block_8x8_c(int16_t *q, const int16_t *dq, uint8_t *dstv, int stride, uint16_t *eobs, MACROBLOCKD *xd) { - vp9_dequant_idct_add_8x8_c(q, dq, pre, dstu, 8, stride, 0, xd->eobs[16]); + vp9_dequant_idct_add_8x8_c(q, dq, pre, dstu, 8, stride, xd->eobs[16]); q += 64; pre += 64; - vp9_dequant_idct_add_8x8_c(q, dq, pre, dstv, 8, stride, 0, xd->eobs[20]); + vp9_dequant_idct_add_8x8_c(q, dq, pre, dstv, 8, stride, xd->eobs[20]); } void vp9_dequant_idct_add_uv_block_8x8_inplace_c(int16_t *q, const int16_t *dq, @@ -304,41 +206,15 @@ void vp9_dequant_idct_add_uv_block_8x8_inplace_c(int16_t *q, const int16_t *dq, int stride, uint16_t *eobs, MACROBLOCKD *xd) { - vp9_dequant_idct_add_8x8_c(q, dq, dstu, dstu, stride, stride, 0, + vp9_dequant_idct_add_8x8_c(q, dq, dstu, dstu, stride, stride, xd->eobs[16]); q += 64; - vp9_dequant_idct_add_8x8_c(q, dq, dstv, dstv, stride, stride, 0, + vp9_dequant_idct_add_8x8_c(q, dq, dstv, dstv, stride, stride, xd->eobs[20]); } #if CONFIG_LOSSLESS -void vp9_dequant_dc_idct_add_y_block_lossless_c(int16_t *q, const int16_t *dq, - uint8_t *pre, - uint8_t *dst, - int stride, - uint16_t *eobs, - const int16_t *dc) { - int i, j; - - for (i = 0; i < 4; i++) { - for (j = 0; j < 4; j++) { - if (*eobs++ > 1) - vp9_dequant_dc_idct_add_lossless_c(q, dq, pre, dst, 16, stride, dc[0]); - else - vp9_dc_only_inv_walsh_add_c(dc[0], pre, dst, 16, stride); - - q += 16; - pre += 4; - dst += 4; - dc++; - } - - pre += 64 - 16; - dst += 4 * stride - 16; - } -} - void vp9_dequant_idct_add_y_block_lossless_c(int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dst, diff --git a/vp9/decoder/vp9_onyxd_int.h b/vp9/decoder/vp9_onyxd_int.h index 0b0b90356..ddb01feed 100644 --- a/vp9/decoder/vp9_onyxd_int.h +++ b/vp9/decoder/vp9_onyxd_int.h @@ -49,7 +49,7 @@ typedef struct { vp9_prob const *coef_probs_8x8[BLOCK_TYPES_8X8]; vp9_prob const *coef_probs_16X16[BLOCK_TYPES_16X16]; - uint8_t eob[25]; + uint8_t eob[24]; } DETOK; diff --git a/vp9/encoder/vp9_bitstream.c b/vp9/encoder/vp9_bitstream.c index 257ddb2c5..f0550b2a2 100644 --- a/vp9/encoder/vp9_bitstream.c +++ b/vp9/encoder/vp9_bitstream.c @@ -1786,8 +1786,8 @@ void vp9_pack_bitstream(VP9_COMP *cpi, unsigned char *dest, // Transmit Dc, Second order and Uv quantizer delta information put_delta_q(&header_bc, pc->y1dc_delta_q); - put_delta_q(&header_bc, pc->y2dc_delta_q); - put_delta_q(&header_bc, pc->y2ac_delta_q); + put_delta_q(&header_bc, pc->y2dc_delta_q_unused); + put_delta_q(&header_bc, pc->y2ac_delta_q_unused); put_delta_q(&header_bc, pc->uvdc_delta_q); put_delta_q(&header_bc, pc->uvac_delta_q); diff --git a/vp9/encoder/vp9_block.h b/vp9/encoder/vp9_block.h index d5110c810..555a5ae99 100644 --- a/vp9/encoder/vp9_block.h +++ b/vp9/encoder/vp9_block.h @@ -89,11 +89,10 @@ typedef struct superblock { } SUPERBLOCK; typedef struct macroblock { - DECLARE_ALIGNED(16, int16_t, src_diff[400]); // 16x16 Y 8x8 U 8x8 V 4x4 2nd Y - DECLARE_ALIGNED(16, int16_t, coeff[400]); // 16x16 Y 8x8 U 8x8 V 4x4 2nd Y + DECLARE_ALIGNED(16, int16_t, src_diff[384]); // 16x16 Y 8x8 U 8x8 V + DECLARE_ALIGNED(16, int16_t, coeff[384]); // 16x16 Y 8x8 U 8x8 V // 16 Y blocks, 4 U blocks, 4 V blocks, - // 1 DC 2nd order block each with 16 entries - BLOCK block[25]; + BLOCK block[24]; SUPERBLOCK sb_coeff_data; @@ -171,15 +170,12 @@ typedef struct macroblock { void (*fwd_txm4x4)(int16_t *input, int16_t *output, int pitch); void (*fwd_txm8x4)(int16_t *input, int16_t *output, int pitch); - void (*fwd_2ndtxm4x4)(int16_t *input, int16_t *output, int pitch); void (*fwd_txm8x8)(int16_t *input, int16_t *output, int pitch); void (*fwd_txm16x16)(int16_t *input, int16_t *output, int pitch); - void (*fwd_2ndtxm2x2)(int16_t *input, int16_t *output, int pitch); void (*quantize_b_4x4)(BLOCK *b, BLOCKD *d); void (*quantize_b_4x4_pair)(BLOCK *b1, BLOCK *b2, BLOCKD *d0, BLOCKD *d1); void (*quantize_b_16x16)(BLOCK *b, BLOCKD *d); void (*quantize_b_8x8)(BLOCK *b, BLOCKD *d); - void (*quantize_b_2x2)(BLOCK *b, BLOCKD *d); } MACROBLOCK; #endif // VP9_ENCODER_VP9_BLOCK_H_ diff --git a/vp9/encoder/vp9_dct.c b/vp9/encoder/vp9_dct.c index 230c7e986..97794e3f6 100644 --- a/vp9/encoder/vp9_dct.c +++ b/vp9/encoder/vp9_dct.c @@ -564,22 +564,6 @@ void vp9_short_fdct8x8_c(short *InputData, short *OutputData, int pitch) { } #endif -void vp9_short_fhaar2x2_c(short *input, short *output, int pitch) { - /* [1 1; 1 -1] orthogonal transform */ - /* use position: 0,1, 4, 8 */ - int i; - short *ip1 = input; - short *op1 = output; - for (i = 0; i < 16; i++) { - op1[i] = 0; - } - - op1[0] = (ip1[0] + ip1[1] + ip1[4] + ip1[8] + 1) >> 1; - op1[1] = (ip1[0] - ip1[1] + ip1[4] - ip1[8]) >> 1; - op1[4] = (ip1[0] + ip1[1] - ip1[4] - ip1[8]) >> 1; - op1[8] = (ip1[0] - ip1[1] - ip1[4] + ip1[8]) >> 1; -} - /* For test */ #define TEST_INT 1 #if TEST_INT @@ -912,87 +896,7 @@ void vp9_short_fdct8x8_c(int16_t *input, int16_t *output, int pitch) { } #endif -void vp9_short_walsh4x4_c(short *input, short *output, int pitch) { - int i; - int a1, b1, c1, d1; - short *ip = input; - short *op = output; - int pitch_short = pitch >> 1; - - for (i = 0; i < 4; i++) { - a1 = ip[0 * pitch_short] + ip[3 * pitch_short]; - b1 = ip[1 * pitch_short] + ip[2 * pitch_short]; - c1 = ip[1 * pitch_short] - ip[2 * pitch_short]; - d1 = ip[0 * pitch_short] - ip[3 * pitch_short]; - - op[0] = (a1 + b1 + 1) >> 1; - op[4] = (c1 + d1) >> 1; - op[8] = (a1 - b1) >> 1; - op[12] = (d1 - c1) >> 1; - - ip++; - op++; - } - ip = output; - op = output; - - for (i = 0; i < 4; i++) { - a1 = ip[0] + ip[3]; - b1 = ip[1] + ip[2]; - c1 = ip[1] - ip[2]; - d1 = ip[0] - ip[3]; - - op[0] = (a1 + b1 + 1) >> 1; - op[1] = (c1 + d1) >> 1; - op[2] = (a1 - b1) >> 1; - op[3] = (d1 - c1) >> 1; - - ip += 4; - op += 4; - } -} - #if CONFIG_LOSSLESS -void vp9_short_walsh4x4_lossless_c(short *input, short *output, int pitch) { - int i; - int a1, b1, c1, d1; - short *ip = input; - short *op = output; - int pitch_short = pitch >> 1; - - for (i = 0; i < 4; i++) { - a1 = (ip[0 * pitch_short] + ip[3 * pitch_short]) >> Y2_WHT_UPSCALE_FACTOR; - b1 = (ip[1 * pitch_short] + ip[2 * pitch_short]) >> Y2_WHT_UPSCALE_FACTOR; - c1 = (ip[1 * pitch_short] - ip[2 * pitch_short]) >> Y2_WHT_UPSCALE_FACTOR; - d1 = (ip[0 * pitch_short] - ip[3 * pitch_short]) >> Y2_WHT_UPSCALE_FACTOR; - - op[0] = (a1 + b1 + 1) >> 1; - op[4] = (c1 + d1) >> 1; - op[8] = (a1 - b1) >> 1; - op[12] = (d1 - c1) >> 1; - - ip++; - op++; - } - ip = output; - op = output; - - for (i = 0; i < 4; i++) { - a1 = ip[0] + ip[3]; - b1 = ip[1] + ip[2]; - c1 = ip[1] - ip[2]; - d1 = ip[0] - ip[3]; - - op[0] = ((a1 + b1 + 1) >> 1) << Y2_WHT_UPSCALE_FACTOR; - op[1] = ((c1 + d1) >> 1) << Y2_WHT_UPSCALE_FACTOR; - op[2] = ((a1 - b1) >> 1) << Y2_WHT_UPSCALE_FACTOR; - op[3] = ((d1 - c1) >> 1) << Y2_WHT_UPSCALE_FACTOR; - - ip += 4; - op += 4; - } -} - void vp9_short_walsh4x4_x8_c(short *input, short *output, int pitch) { int i; int a1, b1, c1, d1; diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c index fe33f2ebf..6708f62e9 100644 --- a/vp9/encoder/vp9_encodeframe.c +++ b/vp9/encoder/vp9_encodeframe.c @@ -1688,10 +1688,7 @@ void vp9_setup_block_ptrs(MACROBLOCK *x) { } } - x->block[24].src_diff = x->src_diff + 384; - - - for (i = 0; i < 25; i++) { + for (i = 0; i < 24; i++) { x->block[i].coeff = x->coeff + i * 16; } } diff --git a/vp9/encoder/vp9_encodemb.c b/vp9/encoder/vp9_encodemb.c index c5185dcf6..11d599e37 100644 --- a/vp9/encoder/vp9_encodemb.c +++ b/vp9/encoder/vp9_encodemb.c @@ -166,26 +166,14 @@ static void subtract_mb(MACROBLOCK *x) { x->e_mbd.predictor, x->src.uv_stride); } -static void build_dcblock_4x4(MACROBLOCK *x) { - int16_t *src_diff_ptr = &x->src_diff[384]; - int i; - - for (i = 0; i < 16; i++) { - src_diff_ptr[i] = x->coeff[i * 16]; - x->coeff[i * 16] = 0; - } -} - void vp9_transform_mby_4x4(MACROBLOCK *x) { int i; MACROBLOCKD *xd = &x->e_mbd; - int has_2nd_order = get_2nd_order_usage(xd); for (i = 0; i < 16; i++) { BLOCK *b = &x->block[i]; TX_TYPE tx_type = get_tx_type_4x4(xd, &xd->block[i]); if (tx_type != DCT_DCT) { - assert(has_2nd_order == 0); vp9_fht_c(b->src_diff, 32, b->coeff, tx_type, 4); } else if (!(i & 1) && get_tx_type_4x4(xd, &xd->block[i + 1]) == DCT_DCT) { x->fwd_txm8x4(&x->block[i].src_diff[0], @@ -196,17 +184,6 @@ void vp9_transform_mby_4x4(MACROBLOCK *x) { &x->block[i].coeff[0], 32); } } - - if (has_2nd_order) { - // build dc block from 16 y dc values - build_dcblock_4x4(x); - - // do 2nd order transform on the dc block - x->fwd_2ndtxm4x4(&x->block[24].src_diff[0], - &x->block[24].coeff[0], 8); - } else { - vpx_memset(x->block[24].coeff, 0, 16 * sizeof(x->block[24].coeff[0])); - } } void vp9_transform_mbuv_4x4(MACROBLOCK *x) { @@ -223,34 +200,15 @@ static void transform_mb_4x4(MACROBLOCK *x) { vp9_transform_mbuv_4x4(x); } -static void build_dcblock_8x8(MACROBLOCK *x) { - int16_t *src_diff_ptr = x->block[24].src_diff; - int i; - - for (i = 0; i < 16; i++) { - src_diff_ptr[i] = 0; - } - src_diff_ptr[0] = x->coeff[0 * 16]; - src_diff_ptr[1] = x->coeff[4 * 16]; - src_diff_ptr[4] = x->coeff[8 * 16]; - src_diff_ptr[8] = x->coeff[12 * 16]; - x->coeff[0 * 16] = 0; - x->coeff[4 * 16] = 0; - x->coeff[8 * 16] = 0; - x->coeff[12 * 16] = 0; -} - void vp9_transform_mby_8x8(MACROBLOCK *x) { int i; MACROBLOCKD *xd = &x->e_mbd; TX_TYPE tx_type; - int has_2nd_order = get_2nd_order_usage(xd); for (i = 0; i < 9; i += 8) { BLOCK *b = &x->block[i]; tx_type = get_tx_type_8x8(xd, &xd->block[i]); if (tx_type != DCT_DCT) { - assert(has_2nd_order == 0); vp9_fht_c(b->src_diff, 32, b->coeff, tx_type, 8); } else { x->fwd_txm8x8(&x->block[i].src_diff[0], @@ -261,24 +219,12 @@ void vp9_transform_mby_8x8(MACROBLOCK *x) { BLOCK *b = &x->block[i]; tx_type = get_tx_type_8x8(xd, &xd->block[i]); if (tx_type != DCT_DCT) { - assert(has_2nd_order == 0); vp9_fht_c(b->src_diff, 32, (b + 2)->coeff, tx_type, 8); } else { x->fwd_txm8x8(&x->block[i].src_diff[0], &x->block[i + 2].coeff[0], 32); } } - - if (has_2nd_order) { - // build dc block from 2x2 y dc values - build_dcblock_8x8(x); - - // do 2nd order transform on the dc block - x->fwd_2ndtxm2x2(&x->block[24].src_diff[0], - &x->block[24].coeff[0], 8); - } else { - vpx_memset(x->block[24].coeff, 0, 16 * sizeof(x->block[24].coeff[0])); - } } void vp9_transform_mbuv_8x8(MACROBLOCK *x) { @@ -379,7 +325,7 @@ static void optimize_b(MACROBLOCK *mb, int i, PLANE_TYPE type, int16_t *qcoeff_ptr = d->qcoeff; int16_t *dqcoeff_ptr = d->dqcoeff; int eob = d->eob, final_eob, sz = 0; - int i0 = (type == PLANE_TYPE_Y_NO_DC); + const int i0 = 0; int rc, x, next; int64_t rdmult, rddiv, rd_cost0, rd_cost1; int rate0, rate1, error0, error1, t0, t1; @@ -582,80 +528,11 @@ static void optimize_b(MACROBLOCK *mb, int i, PLANE_TYPE type, final_eob++; d->eob = final_eob; - *a = *l = (d->eob > !type); -} - -/************************************************************************** -our inverse hadamard transform effectively is weighted sum of all 16 inputs -with weight either 1 or -1. It has a last stage scaling of (sum+1)>>2. And -dc only idct is (dc+16)>>5. So if all the sums are between -65 and 63 the -output after inverse wht and idct will be all zero. A sum of absolute value -smaller than 65 guarantees all 16 different (+1/-1) weighted sums in wht -fall between -65 and +65. -**************************************************************************/ -#define SUM_2ND_COEFF_THRESH 65 - -static void check_reset_2nd_coeffs(MACROBLOCKD *xd, - ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l) { - int sum = 0; - int i; - BLOCKD *bd = &xd->block[24]; - if (bd->dequant[0] >= SUM_2ND_COEFF_THRESH - && bd->dequant[1] >= SUM_2ND_COEFF_THRESH) - return; - - for (i = 0; i < bd->eob; i++) { - int coef = bd->dqcoeff[vp9_default_zig_zag1d_4x4[i]]; - sum += (coef >= 0) ? coef : -coef; - if (sum >= SUM_2ND_COEFF_THRESH) - return; - } - - if (sum < SUM_2ND_COEFF_THRESH) { - for (i = 0; i < bd->eob; i++) { - int rc = vp9_default_zig_zag1d_4x4[i]; - bd->qcoeff[rc] = 0; - bd->dqcoeff[rc] = 0; - } - bd->eob = 0; - *a = *l = (bd->eob != 0); - } -} - -#define SUM_2ND_COEFF_THRESH_8X8 32 -static void check_reset_8x8_2nd_coeffs(MACROBLOCKD *xd, - ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l) { - int sum = 0; - BLOCKD *bd = &xd->block[24]; - int coef; - - coef = bd->dqcoeff[0]; - sum += (coef >= 0) ? coef : -coef; - coef = bd->dqcoeff[1]; - sum += (coef >= 0) ? coef : -coef; - coef = bd->dqcoeff[4]; - sum += (coef >= 0) ? coef : -coef; - coef = bd->dqcoeff[8]; - sum += (coef >= 0) ? coef : -coef; - - if (sum < SUM_2ND_COEFF_THRESH_8X8) { - bd->qcoeff[0] = 0; - bd->dqcoeff[0] = 0; - bd->qcoeff[1] = 0; - bd->dqcoeff[1] = 0; - bd->qcoeff[4] = 0; - bd->dqcoeff[4] = 0; - bd->qcoeff[8] = 0; - bd->dqcoeff[8] = 0; - bd->eob = 0; - *a = *l = (bd->eob != 0); - } + *a = *l = (d->eob > 0); } void vp9_optimize_mby_4x4(MACROBLOCK *x) { int b; - PLANE_TYPE type; - int has_2nd_order; ENTROPY_CONTEXT_PLANES t_above, t_left; ENTROPY_CONTEXT *ta; ENTROPY_CONTEXT *tl; @@ -669,25 +546,11 @@ void vp9_optimize_mby_4x4(MACROBLOCK *x) { ta = (ENTROPY_CONTEXT *)&t_above; tl = (ENTROPY_CONTEXT *)&t_left; - has_2nd_order = get_2nd_order_usage(&x->e_mbd); - - type = has_2nd_order ? PLANE_TYPE_Y_NO_DC : PLANE_TYPE_Y_WITH_DC; - for (b = 0; b < 16; b++) { - optimize_b(x, b, type, + optimize_b(x, b, PLANE_TYPE_Y_WITH_DC, ta + vp9_block2above[TX_4X4][b], tl + vp9_block2left[TX_4X4][b], TX_4X4); } - - if (has_2nd_order) { - b = 24; - optimize_b(x, b, PLANE_TYPE_Y2, - ta + vp9_block2above[TX_4X4][b], - tl + vp9_block2left[TX_4X4][b], TX_4X4); - check_reset_2nd_coeffs(&x->e_mbd, - ta + vp9_block2above[TX_4X4][b], - tl + vp9_block2left[TX_4X4][b]); - } } void vp9_optimize_mbuv_4x4(MACROBLOCK *x) { @@ -719,11 +582,9 @@ static void optimize_mb_4x4(MACROBLOCK *x) { void vp9_optimize_mby_8x8(MACROBLOCK *x) { int b; - PLANE_TYPE type; ENTROPY_CONTEXT_PLANES t_above, t_left; ENTROPY_CONTEXT *ta; ENTROPY_CONTEXT *tl; - int has_2nd_order = get_2nd_order_usage(&x->e_mbd); if (!x->e_mbd.above_context || !x->e_mbd.left_context) return; @@ -733,7 +594,6 @@ void vp9_optimize_mby_8x8(MACROBLOCK *x) { ta = (ENTROPY_CONTEXT *)&t_above; tl = (ENTROPY_CONTEXT *)&t_left; - type = has_2nd_order ? PLANE_TYPE_Y_NO_DC : PLANE_TYPE_Y_WITH_DC; for (b = 0; b < 16; b += 4) { ENTROPY_CONTEXT *const a = ta + vp9_block2above[TX_8X8][b]; ENTROPY_CONTEXT *const l = tl + vp9_block2left[TX_8X8][b]; @@ -744,17 +604,10 @@ void vp9_optimize_mby_8x8(MACROBLOCK *x) { ENTROPY_CONTEXT above_ec = a[0]; ENTROPY_CONTEXT left_ec = l[0]; #endif - optimize_b(x, b, type, &above_ec, &left_ec, TX_8X8); + optimize_b(x, b, PLANE_TYPE_Y_WITH_DC, &above_ec, &left_ec, TX_8X8); a[1] = a[0] = above_ec; l[1] = l[0] = left_ec; } - - // 8x8 always have 2nd order block - if (has_2nd_order) { - check_reset_8x8_2nd_coeffs(&x->e_mbd, - ta + vp9_block2above[TX_8X8][24], - tl + vp9_block2left[TX_8X8][24]); - } } void vp9_optimize_mbuv_8x8(MACROBLOCK *x) { diff --git a/vp9/encoder/vp9_onyx_if.c b/vp9/encoder/vp9_onyx_if.c index 3e5940f55..b54f4adc7 100644 --- a/vp9/encoder/vp9_onyx_if.c +++ b/vp9/encoder/vp9_onyx_if.c @@ -845,14 +845,11 @@ void vp9_set_speed_features(VP9_COMP *cpi) { cpi->mb.fwd_txm8x8 = vp9_short_fdct8x8; cpi->mb.fwd_txm8x4 = vp9_short_fdct8x4; cpi->mb.fwd_txm4x4 = vp9_short_fdct4x4; - cpi->mb.fwd_2ndtxm4x4 = vp9_short_walsh4x4; - cpi->mb.fwd_2ndtxm2x2 = vp9_short_fhaar2x2; #if CONFIG_LOSSLESS if (cpi->oxcf.lossless) { cpi->mb.fwd_txm8x4 = vp9_short_walsh8x4_x8; cpi->mb.fwd_txm4x4 = vp9_short_walsh4x4_x8; - cpi->mb.fwd_2ndtxm4x4 = vp9_short_walsh4x4_lossless; } #endif @@ -860,7 +857,6 @@ void vp9_set_speed_features(VP9_COMP *cpi) { cpi->mb.quantize_b_4x4_pair = vp9_regular_quantize_b_4x4_pair; cpi->mb.quantize_b_8x8 = vp9_regular_quantize_b_8x8; cpi->mb.quantize_b_16x16 = vp9_regular_quantize_b_16x16; - cpi->mb.quantize_b_2x2 = vp9_regular_quantize_b_2x2; vp9_init_quantizer(cpi); @@ -1222,16 +1218,12 @@ void vp9_change_config(VP9_PTR ptr, VP9_CONFIG *oxcf) { cpi->mb.e_mbd.inv_txm4x4_1 = vp9_short_idct4x4llm_1; cpi->mb.e_mbd.inv_txm4x4 = vp9_short_idct4x4llm; - cpi->mb.e_mbd.inv_2ndtxm4x4_1 = vp9_short_inv_walsh4x4_1; - cpi->mb.e_mbd.inv_2ndtxm4x4 = vp9_short_inv_walsh4x4; #if CONFIG_LOSSLESS cpi->oxcf.lossless = oxcf->lossless; if (cpi->oxcf.lossless) { cpi->mb.e_mbd.inv_txm4x4_1 = vp9_short_inv_walsh4x4_1_x8; cpi->mb.e_mbd.inv_txm4x4 = vp9_short_inv_walsh4x4_x8; - cpi->mb.e_mbd.inv_2ndtxm4x4_1 = vp9_short_inv_walsh4x4_1_lossless; - cpi->mb.e_mbd.inv_2ndtxm4x4 = vp9_short_inv_walsh4x4_lossless; } #endif diff --git a/vp9/encoder/vp9_onyx_int.h b/vp9/encoder/vp9_onyx_int.h index 1476de4da..2a24112b3 100644 --- a/vp9/encoder/vp9_onyx_int.h +++ b/vp9/encoder/vp9_onyx_int.h @@ -303,18 +303,12 @@ typedef struct VP9_COMP { DECLARE_ALIGNED(16, short, Y1zbin[QINDEX_RANGE][16]); DECLARE_ALIGNED(16, short, Y1round[QINDEX_RANGE][16]); - DECLARE_ALIGNED(16, short, Y2quant[QINDEX_RANGE][16]); - DECLARE_ALIGNED(16, unsigned char, Y2quant_shift[QINDEX_RANGE][16]); - DECLARE_ALIGNED(16, short, Y2zbin[QINDEX_RANGE][16]); - DECLARE_ALIGNED(16, short, Y2round[QINDEX_RANGE][16]); - DECLARE_ALIGNED(16, short, UVquant[QINDEX_RANGE][16]); DECLARE_ALIGNED(16, unsigned char, UVquant_shift[QINDEX_RANGE][16]); DECLARE_ALIGNED(16, short, UVzbin[QINDEX_RANGE][16]); DECLARE_ALIGNED(16, short, UVround[QINDEX_RANGE][16]); DECLARE_ALIGNED(16, short, zrun_zbin_boost_y1[QINDEX_RANGE][16]); - DECLARE_ALIGNED(16, short, zrun_zbin_boost_y2[QINDEX_RANGE][16]); DECLARE_ALIGNED(16, short, zrun_zbin_boost_uv[QINDEX_RANGE][16]); MACROBLOCK mb; diff --git a/vp9/encoder/vp9_quantize.c b/vp9/encoder/vp9_quantize.c index e66db7499..6ff4f6ea6 100644 --- a/vp9/encoder/vp9_quantize.c +++ b/vp9/encoder/vp9_quantize.c @@ -140,26 +140,15 @@ void vp9_regular_quantize_b_4x4(BLOCK *b, BLOCKD *d) { void vp9_quantize_mby_4x4_c(MACROBLOCK *x) { int i; - int has_2nd_order = get_2nd_order_usage(&x->e_mbd); for (i = 0; i < 16; i++) { TX_TYPE tx_type = get_tx_type_4x4(&x->e_mbd, &x->e_mbd.block[i]); if (tx_type != DCT_DCT) { - assert(has_2nd_order == 0); vp9_ht_quantize_b_4x4(&x->block[i], &x->e_mbd.block[i], tx_type); } else { x->quantize_b_4x4(&x->block[i], &x->e_mbd.block[i]); } } - if (has_2nd_order) { - x->quantize_b_4x4(&x->block[24], &x->e_mbd.block[24]); - } else { - vpx_memset(x->e_mbd.block[24].qcoeff, 0, - 16 * sizeof(x->e_mbd.block[24].qcoeff[0])); - vpx_memset(x->e_mbd.block[24].dqcoeff, 0, - 16 * sizeof(x->e_mbd.block[24].dqcoeff[0])); - x->e_mbd.block[24].eob = 0; - } } void vp9_quantize_mbuv_4x4_c(MACROBLOCK *x) { @@ -174,59 +163,6 @@ void vp9_quantize_mb_4x4_c(MACROBLOCK *x) { vp9_quantize_mbuv_4x4_c(x); } -void vp9_regular_quantize_b_2x2(BLOCK *b, BLOCKD *d) { - int i, rc, eob; - int zbin; - int x, y, z, sz; - int16_t *zbin_boost_ptr = b->zrun_zbin_boost; - int zbin_zrun_index = 0; - int16_t *coeff_ptr = b->coeff; - int16_t *zbin_ptr = b->zbin; - int16_t *round_ptr = b->round; - int16_t *quant_ptr = b->quant; - uint8_t *quant_shift_ptr = b->quant_shift; - int16_t *qcoeff_ptr = d->qcoeff; - int16_t *dqcoeff_ptr = d->dqcoeff; - int16_t *dequant_ptr = d->dequant; - int zbin_oq_value = b->zbin_extra; - - // double q2nd = 4; - vpx_memset(qcoeff_ptr, 0, 32); - vpx_memset(dqcoeff_ptr, 0, 32); - - eob = -1; - - if (!b->skip_block) { - for (i = 0; i < 4; i++) { - rc = vp9_default_zig_zag1d_4x4[i]; - z = coeff_ptr[rc]; - - zbin_boost_ptr = &b->zrun_zbin_boost[zbin_zrun_index]; - zbin_zrun_index += 4; - zbin = (zbin_ptr[rc] + *zbin_boost_ptr + zbin_oq_value); - - sz = (z >> 31); // sign of z - x = (z ^ sz) - sz; // x = abs(z) - - if (x >= zbin) { - x += (round_ptr[rc]); - y = ((int)((int)(x * quant_ptr[rc]) >> 16) + x) - >> quant_shift_ptr[rc]; // quantize (x) - x = (y ^ sz) - sz; // get the sign back - qcoeff_ptr[rc] = x; // write to destination - dqcoeff_ptr[rc] = x * dequant_ptr[rc]; // dequantized value - - if (y) { - eob = i; // last nonzero coeffs - zbin_zrun_index = 0; - } - } - } - } - - d->eob = eob + 1; -} - void vp9_regular_quantize_b_8x8(BLOCK *b, BLOCKD *d) { int16_t *qcoeff_ptr = d->qcoeff; int16_t *dqcoeff_ptr = d->dqcoeff; @@ -308,29 +244,13 @@ void vp9_regular_quantize_b_8x8(BLOCK *b, BLOCKD *d) { void vp9_quantize_mby_8x8(MACROBLOCK *x) { int i; - int has_2nd_order = get_2nd_order_usage(&x->e_mbd); for (i = 0; i < 16; i ++) { x->e_mbd.block[i].eob = 0; } - x->e_mbd.block[24].eob = 0; for (i = 0; i < 16; i += 4) { - int ib = (i & 8) + ((i & 4) >> 1); - TX_TYPE tx_type = get_tx_type_8x8(&x->e_mbd, &x->e_mbd.block[ib]); - if (tx_type != DCT_DCT) - assert(has_2nd_order == 0); x->quantize_b_8x8(&x->block[i], &x->e_mbd.block[i]); } - - if (has_2nd_order) { - x->quantize_b_2x2(&x->block[24], &x->e_mbd.block[24]); - } else { - vpx_memset(x->e_mbd.block[24].qcoeff, 0, - 16 * sizeof(x->e_mbd.block[24].qcoeff[0])); - vpx_memset(x->e_mbd.block[24].dqcoeff, 0, - 16 * sizeof(x->e_mbd.block[24].dqcoeff[0])); - x->e_mbd.block[24].eob = 0; - } } void vp9_quantize_mbuv_8x8(MACROBLOCK *x) { @@ -352,7 +272,6 @@ void vp9_quantize_mby_16x16(MACROBLOCK *x) { for (i = 0; i < 16; i++) x->e_mbd.block[i].eob = 0; - x->e_mbd.block[24].eob = 0; x->quantize_b_16x16(&x->block[0], &x->e_mbd.block[0]); } @@ -510,13 +429,6 @@ void vp9_init_quantizer(VP9_COMP *cpi) { cpi->common.Y1dequant[Q][0] = quant_val; cpi->zrun_zbin_boost_y1[Q][0] = (quant_val * zbin_boost[0]) >> 7; - quant_val = vp9_dc2quant(Q, cpi->common.y2dc_delta_q); - invert_quant(cpi->Y2quant[Q] + 0, - cpi->Y2quant_shift[Q] + 0, quant_val); - cpi->Y2zbin[Q][0] = ((qzbin_factor * quant_val) + 64) >> 7; - cpi->Y2round[Q][0] = (qrounding_factor * quant_val) >> 7; - cpi->common.Y2dequant[Q][0] = quant_val; - cpi->zrun_zbin_boost_y2[Q][0] = (quant_val * zbin_boost[0]) >> 7; quant_val = vp9_dc_uv_quant(Q, cpi->common.uvdc_delta_q); invert_quant(cpi->UVquant[Q] + 0, cpi->UVquant_shift[Q] + 0, quant_val); @@ -538,15 +450,6 @@ void vp9_init_quantizer(VP9_COMP *cpi) { cpi->zrun_zbin_boost_y1[Q][i] = ((quant_val * zbin_boost[i]) + 64) >> 7; - quant_val = vp9_ac2quant(Q, cpi->common.y2ac_delta_q); - invert_quant(cpi->Y2quant[Q] + rc, - cpi->Y2quant_shift[Q] + rc, quant_val); - cpi->Y2zbin[Q][rc] = ((qzbin_factor * quant_val) + 64) >> 7; - cpi->Y2round[Q][rc] = (qrounding_factor * quant_val) >> 7; - cpi->common.Y2dequant[Q][rc] = quant_val; - cpi->zrun_zbin_boost_y2[Q][i] = - ((quant_val * zbin_boost[i]) + 64) >> 7; - quant_val = vp9_ac_uv_quant(Q, cpi->common.uvac_delta_q); invert_quant(cpi->UVquant[Q] + rc, cpi->UVquant_shift[Q] + rc, quant_val); @@ -623,25 +526,6 @@ void vp9_mb_init_quantizer(VP9_COMP *cpi, MACROBLOCK *x) { vp9_segfeature_active(xd, segment_id, SEG_LVL_SKIP); } - // Y2 - zbin_extra = (cpi->common.Y2dequant[QIndex][1] * - ((cpi->zbin_over_quant / 2) + - cpi->zbin_mode_boost + - x->act_zbin_adj)) >> 7; - - x->block[24].quant = cpi->Y2quant[QIndex]; - x->block[24].quant_shift = cpi->Y2quant_shift[QIndex]; - x->block[24].zbin = cpi->Y2zbin[QIndex]; - x->block[24].round = cpi->Y2round[QIndex]; - x->e_mbd.block[24].dequant = cpi->common.Y2dequant[QIndex]; - x->block[24].zrun_zbin_boost = cpi->zrun_zbin_boost_y2[QIndex]; - x->block[24].zbin_extra = (int16_t)zbin_extra; - - // TBD perhaps not use for Y2 - // Segment skip feature. - x->block[24].skip_block = - vp9_segfeature_active(xd, segment_id, SEG_LVL_SKIP); - /* save this macroblock QIndex for vp9_update_zbin_extra() */ x->e_mbd.q_index = QIndex; } @@ -669,14 +553,6 @@ void vp9_update_zbin_extra(VP9_COMP *cpi, MACROBLOCK *x) { for (i = 16; i < 24; i++) { x->block[i].zbin_extra = (int16_t)zbin_extra; } - - // Y2 - zbin_extra = (cpi->common.Y2dequant[QIndex][1] * - ((cpi->zbin_over_quant / 2) + - cpi->zbin_mode_boost + - x->act_zbin_adj)) >> 7; - - x->block[24].zbin_extra = (int16_t)zbin_extra; } void vp9_frame_init_quantizer(VP9_COMP *cpi) { @@ -695,10 +571,10 @@ void vp9_set_quantizer(struct VP9_COMP *cpi, int Q) { // if any of the delta_q values are changing update flag will // have to be set. cm->y1dc_delta_q = 0; - cm->y2ac_delta_q = 0; + cm->y2ac_delta_q_unused = 0; cm->uvdc_delta_q = 0; cm->uvac_delta_q = 0; - cm->y2dc_delta_q = 0; + cm->y2dc_delta_q_unused = 0; // quantizer has to be reinitialized if any delta_q changes. // As there are not any here for now this is inactive code. diff --git a/vp9/encoder/vp9_quantize.h b/vp9/encoder/vp9_quantize.h index ac44a751c..9a8e35d2c 100644 --- a/vp9/encoder/vp9_quantize.h +++ b/vp9/encoder/vp9_quantize.h @@ -50,11 +50,6 @@ extern prototype_quantize_block(vp9_quantize_quantb_8x8); #endif extern prototype_quantize_block(vp9_quantize_quantb_16x16); -#ifndef vp9_quantize_quantb_2x2 -#define vp9_quantize_quantb_2x2 vp9_regular_quantize_b_2x2 -#endif -extern prototype_quantize_block(vp9_quantize_quantb_2x2); - #ifndef vp9_quantize_mb_4x4 #define vp9_quantize_mb_4x4 vp9_quantize_mb_4x4_c #endif diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c index a05b52f36..4a49728f6 100644 --- a/vp9/encoder/vp9_rdopt.c +++ b/vp9/encoder/vp9_rdopt.c @@ -320,26 +320,7 @@ int vp9_block_error_c(int16_t *coeff, int16_t *dqcoeff, int block_size) { return error; } -int vp9_mbblock_error_8x8_c(MACROBLOCK *mb, int dc) { - BLOCK *be; - BLOCKD *bd; - int i, j; - int berror, error = 0; - - for (i = 0; i < 16; i+=4) { - be = &mb->block[i]; - bd = &mb->e_mbd.block[i]; - berror = 0; - for (j = dc; j < 64; j++) { - int this_diff = be->coeff[j] - bd->dqcoeff[j]; - berror += this_diff * this_diff; - } - error += berror; - } - return error; -} - -int vp9_mbblock_error_c(MACROBLOCK *mb, int dc) { +int vp9_mbblock_error_c(MACROBLOCK *mb) { BLOCK *be; BLOCKD *bd; int i, j; @@ -349,7 +330,7 @@ int vp9_mbblock_error_c(MACROBLOCK *mb, int dc) { be = &mb->block[i]; bd = &mb->e_mbd.block[i]; berror = 0; - for (j = dc; j < 16; j++) { + for (j = 0; j < 16; j++) { int this_diff = be->coeff[j] - bd->dqcoeff[j]; berror += this_diff * this_diff; } @@ -428,7 +409,7 @@ static INLINE int cost_coeffs(MACROBLOCK *mb, const int eob = b->eob; MACROBLOCKD *xd = &mb->e_mbd; const int ib = (int)(b - xd->block); - int c = (type == PLANE_TYPE_Y_NO_DC) ? 1 : 0; + int c = 0; int cost = 0, seg_eob; const int segment_id = xd->mode_info_context->mbmi.segment_id; const int *scan; @@ -453,13 +434,8 @@ static INLINE int cost_coeffs(MACROBLOCK *mb, } break; case TX_8X8: - if (type == PLANE_TYPE_Y2) { - scan = vp9_default_zig_zag1d_4x4; - seg_eob = 4; - } else { - scan = vp9_default_zig_zag1d_8x8; - seg_eob = 64; - } + scan = vp9_default_zig_zag1d_8x8; + seg_eob = 64; break; case TX_16X16: scan = vp9_default_zig_zag1d_16x16; @@ -511,12 +487,12 @@ static INLINE int cost_coeffs(MACROBLOCK *mb, } // is eob first coefficient; - pt = (c > !type); + pt = (c > 0); *a = *l = pt; return cost; } -static int rdcost_mby_4x4(MACROBLOCK *mb, int has_2nd_order, int backup) { +static int rdcost_mby_4x4(MACROBLOCK *mb, int backup) { int cost = 0; int b; MACROBLOCKD *xd = &mb->e_mbd; @@ -536,19 +512,11 @@ static int rdcost_mby_4x4(MACROBLOCK *mb, int has_2nd_order, int backup) { } for (b = 0; b < 16; b++) - cost += cost_coeffs(mb, xd->block + b, - (has_2nd_order ? - PLANE_TYPE_Y_NO_DC : PLANE_TYPE_Y_WITH_DC), + cost += cost_coeffs(mb, xd->block + b, PLANE_TYPE_Y_WITH_DC, ta + vp9_block2above[TX_4X4][b], tl + vp9_block2left[TX_4X4][b], TX_4X4); - if (has_2nd_order) - cost += cost_coeffs(mb, xd->block + 24, PLANE_TYPE_Y2, - ta + vp9_block2above[TX_4X4][24], - tl + vp9_block2left[TX_4X4][24], - TX_4X4); - return cost; } @@ -557,26 +525,17 @@ static void macro_block_yrd_4x4(MACROBLOCK *mb, int *Distortion, int *skippable, int backup) { MACROBLOCKD *const xd = &mb->e_mbd; - BLOCK *const mb_y2 = mb->block + 24; - BLOCKD *const x_y2 = xd->block + 24; - int d, has_2nd_order; xd->mode_info_context->mbmi.txfm_size = TX_4X4; - has_2nd_order = get_2nd_order_usage(xd); - // Fdct and building the 2nd order block vp9_transform_mby_4x4(mb); vp9_quantize_mby_4x4(mb); - d = vp9_mbblock_error(mb, has_2nd_order); - if (has_2nd_order) - d += vp9_block_error(mb_y2->coeff, x_y2->dqcoeff, 16); - - *Distortion = (d >> 2); - // rate - *Rate = rdcost_mby_4x4(mb, has_2nd_order, backup); - *skippable = vp9_mby_is_skippable_4x4(&mb->e_mbd, has_2nd_order); + + *Distortion = vp9_mbblock_error(mb) >> 2; + *Rate = rdcost_mby_4x4(mb, backup); + *skippable = vp9_mby_is_skippable_4x4(xd); } -static int rdcost_mby_8x8(MACROBLOCK *mb, int has_2nd_order, int backup) { +static int rdcost_mby_8x8(MACROBLOCK *mb, int backup) { int cost = 0; int b; MACROBLOCKD *xd = &mb->e_mbd; @@ -596,18 +555,11 @@ static int rdcost_mby_8x8(MACROBLOCK *mb, int has_2nd_order, int backup) { } for (b = 0; b < 16; b += 4) - cost += cost_coeffs(mb, xd->block + b, - (has_2nd_order ? - PLANE_TYPE_Y_NO_DC : PLANE_TYPE_Y_WITH_DC), + cost += cost_coeffs(mb, xd->block + b, PLANE_TYPE_Y_WITH_DC, ta + vp9_block2above[TX_8X8][b], tl + vp9_block2left[TX_8X8][b], TX_8X8); - if (has_2nd_order) - cost += cost_coeffs(mb, xd->block + 24, PLANE_TYPE_Y2, - ta + vp9_block2above[TX_8X8][24], - tl + vp9_block2left[TX_8X8][24], - TX_8X8); return cost; } @@ -616,23 +568,14 @@ static void macro_block_yrd_8x8(MACROBLOCK *mb, int *Distortion, int *skippable, int backup) { MACROBLOCKD *const xd = &mb->e_mbd; - BLOCK *const mb_y2 = mb->block + 24; - BLOCKD *const x_y2 = xd->block + 24; - int d, has_2nd_order; xd->mode_info_context->mbmi.txfm_size = TX_8X8; - vp9_transform_mby_8x8(mb); vp9_quantize_mby_8x8(mb); - has_2nd_order = get_2nd_order_usage(xd); - d = vp9_mbblock_error_8x8_c(mb, has_2nd_order); - if (has_2nd_order) - d += vp9_block_error(mb_y2->coeff, x_y2->dqcoeff, 16); - - *Distortion = (d >> 2); - // rate - *Rate = rdcost_mby_8x8(mb, has_2nd_order, backup); - *skippable = vp9_mby_is_skippable_8x8(&mb->e_mbd, has_2nd_order); + + *Distortion = vp9_mbblock_error(mb) >> 2; + *Rate = rdcost_mby_8x8(mb, backup); + *skippable = vp9_mby_is_skippable_8x8(xd); } static int rdcost_mby_16x16(MACROBLOCK *mb, int backup) { @@ -658,7 +601,6 @@ static int rdcost_mby_16x16(MACROBLOCK *mb, int backup) { static void macro_block_yrd_16x16(MACROBLOCK *mb, int *Rate, int *Distortion, int *skippable, int backup) { - int d; MACROBLOCKD *xd = &mb->e_mbd; xd->mode_info_context->mbmi.txfm_size = TX_16X16; @@ -671,12 +613,9 @@ static void macro_block_yrd_16x16(MACROBLOCK *mb, int *Rate, int *Distortion, xd->mode_info_context->mbmi.mode < I8X8_PRED) vp9_optimize_mby_16x16(mb); - d = vp9_mbblock_error(mb, 0); - - *Distortion = (d >> 2); - // rate + *Distortion = vp9_mbblock_error(mb) >> 2; *Rate = rdcost_mby_16x16(mb, backup); - *skippable = vp9_mby_is_skippable_16x16(&mb->e_mbd); + *skippable = vp9_mby_is_skippable_16x16(xd); } static void choose_txfm_size_from_rd(VP9_COMP *cpi, MACROBLOCK *x, @@ -1401,7 +1340,6 @@ static int64_t rd_pick_intra8x8block(VP9_COMP *cpi, MACROBLOCK *x, int ib, vp9_subtract_4b_c(be, b, 16); - assert(get_2nd_order_usage(xd) == 0); if (xd->mode_info_context->mbmi.txfm_size == TX_8X8) { TX_TYPE tx_type = get_tx_type_8x8(xd, b); if (tx_type != DCT_DCT) @@ -2862,8 +2800,8 @@ static int rd_pick_best_mbsegmentation(VP9_COMP *cpi, MACROBLOCK *x, *returndistortion = bsi.d; *returnyrate = bsi.segment_yrate; *skippable = bsi.txfm_size == TX_4X4 ? - vp9_mby_is_skippable_4x4(&x->e_mbd, 0) : - vp9_mby_is_skippable_8x8(&x->e_mbd, 0); + vp9_mby_is_skippable_4x4(&x->e_mbd) : + vp9_mby_is_skippable_8x8(&x->e_mbd); /* save partitions */ mbmi->txfm_size = bsi.txfm_size; @@ -3376,7 +3314,7 @@ static int64_t handle_inter_mode(VP9_COMP *cpi, MACROBLOCK *x, } if ((int)sse < threshold) { - unsigned int q2dc = xd->block[24].dequant[0]; + unsigned int q2dc = xd->block[0].dequant[0]; /* If there is no codeable 2nd order dc or a very small uniform pixel change change */ if ((sse - var < q2dc * q2dc >> 4) || diff --git a/vp9/encoder/vp9_satd_c.c b/vp9/encoder/vp9_satd_c.c deleted file mode 100644 index 212c2243d..000000000 --- a/vp9/encoder/vp9_satd_c.c +++ /dev/null @@ -1,48 +0,0 @@ -/* - * Copyright (c) 2010 The WebM project authors. All Rights Reserved. - * - * Use of this source code is governed by a BSD-style license - * that can be found in the LICENSE file in the root of the source - * tree. An additional intellectual property rights grant can be found - * in the file PATENTS. All contributing project authors may - * be found in the AUTHORS file in the root of the source tree. - */ - -#include <stdlib.h> -#include "vpx_ports/mem.h" -#include "./vp9_rtcd.h" - -unsigned int vp9_satd16x16_c(const uint8_t *src_ptr, - int src_stride, - const uint8_t *ref_ptr, - int ref_stride, - unsigned int *psatd) { - int r, c, i; - unsigned int satd = 0; - DECLARE_ALIGNED(16, int16_t, diff_in[256]); - DECLARE_ALIGNED(16, int16_t, diff_out[16]); - int16_t *in; - - for (r = 0; r < 16; r++) { - for (c = 0; c < 16; c++) { - diff_in[r * 16 + c] = src_ptr[c] - ref_ptr[c]; - } - src_ptr += src_stride; - ref_ptr += ref_stride; - } - - in = diff_in; - for (r = 0; r < 16; r += 4) { - for (c = 0; c < 16; c += 4) { - vp9_short_walsh4x4_c(in + c, diff_out, 32); - for (i = 0; i < 16; i++) - satd += abs(diff_out[i]); - } - in += 64; - } - - if (psatd) - *psatd = satd; - - return satd; -} diff --git a/vp9/encoder/vp9_tokenize.c b/vp9/encoder/vp9_tokenize.c index 98e04591b..09ee9f923 100644 --- a/vp9/encoder/vp9_tokenize.c +++ b/vp9/encoder/vp9_tokenize.c @@ -108,7 +108,7 @@ static void tokenize_b(VP9_COMP *cpi, TX_SIZE tx_size, int dry_run) { int pt; /* near block/prev token context index */ - int c = (type == PLANE_TYPE_Y_NO_DC) ? 1 : 0; + int c = 0; int recent_energy = 0; const BLOCKD * const b = xd->block + ib; const int eob = b->eob; /* one beyond last nonzero coeff */ @@ -153,17 +153,12 @@ static void tokenize_b(VP9_COMP *cpi, } break; case TX_8X8: - if (type == PLANE_TYPE_Y2) { - seg_eob = 4; - scan = vp9_default_zig_zag1d_4x4; - } else { #if CONFIG_CNVCONTEXT - a_ec = (a[0] + a[1]) != 0; - l_ec = (l[0] + l[1]) != 0; + a_ec = (a[0] + a[1]) != 0; + l_ec = (l[0] + l[1]) != 0; #endif - seg_eob = 64; - scan = vp9_default_zig_zag1d_8x8; - } + seg_eob = 64; + scan = vp9_default_zig_zag1d_8x8; if (tx_type != DCT_DCT) { counts = cpi->hybrid_coef_counts_8x8; probs = cpi->common.fc.hybrid_coef_probs_8x8; @@ -235,8 +230,7 @@ static void tokenize_b(VP9_COMP *cpi, t->Token = token; t->context_tree = probs[type][band][pt]; - t->skip_eob_node = (pt == 0) && ((band > 0 && type != PLANE_TYPE_Y_NO_DC) || - (band > 1 && type == PLANE_TYPE_Y_NO_DC)); + t->skip_eob_node = (pt == 0) && (band > 0); assert(vp9_coef_encodings[t->Token].Len - t->skip_eob_node > 0); if (!dry_run) { ++counts[type][band][pt][token]; @@ -247,11 +241,11 @@ static void tokenize_b(VP9_COMP *cpi, } while (c < eob && ++c < seg_eob); *tp = t; - a_ec = l_ec = (c > !type); /* 0 <-> all coeff data is zero */ + a_ec = l_ec = (c > 0); /* 0 <-> all coeff data is zero */ a[0] = a_ec; l[0] = l_ec; - if (tx_size == TX_8X8 && type != PLANE_TYPE_Y2) { + if (tx_size == TX_8X8) { a[1] = a_ec; l[1] = l_ec; } else if (tx_size == TX_16X16) { @@ -270,18 +264,13 @@ static void tokenize_b(VP9_COMP *cpi, } } -int vp9_mby_is_skippable_4x4(MACROBLOCKD *xd, int has_2nd_order) { +int vp9_mby_is_skippable_4x4(MACROBLOCKD *xd) { int skip = 1; int i = 0; - if (has_2nd_order) { - for (i = 0; i < 16; i++) - skip &= (xd->block[i].eob < 2); - skip &= (!xd->block[24].eob); - } else { - for (i = 0; i < 16; i++) - skip &= (!xd->block[i].eob); - } + for (i = 0; i < 16; i++) + skip &= (!xd->block[i].eob); + return skip; } @@ -294,23 +283,18 @@ int vp9_mbuv_is_skippable_4x4(MACROBLOCKD *xd) { return skip; } -static int mb_is_skippable_4x4(MACROBLOCKD *xd, int has_2nd_order) { - return (vp9_mby_is_skippable_4x4(xd, has_2nd_order) & +static int mb_is_skippable_4x4(MACROBLOCKD *xd) { + return (vp9_mby_is_skippable_4x4(xd) & vp9_mbuv_is_skippable_4x4(xd)); } -int vp9_mby_is_skippable_8x8(MACROBLOCKD *xd, int has_2nd_order) { +int vp9_mby_is_skippable_8x8(MACROBLOCKD *xd) { int skip = 1; int i = 0; - if (has_2nd_order) { - for (i = 0; i < 16; i += 4) - skip &= (xd->block[i].eob < 2); - skip &= (!xd->block[24].eob); - } else { - for (i = 0; i < 16; i += 4) - skip &= (!xd->block[i].eob); - } + for (i = 0; i < 16; i += 4) + skip &= (!xd->block[i].eob); + return skip; } @@ -318,13 +302,13 @@ int vp9_mbuv_is_skippable_8x8(MACROBLOCKD *xd) { return (!xd->block[16].eob) & (!xd->block[20].eob); } -static int mb_is_skippable_8x8(MACROBLOCKD *xd, int has_2nd_order) { - return (vp9_mby_is_skippable_8x8(xd, has_2nd_order) & +static int mb_is_skippable_8x8(MACROBLOCKD *xd) { + return (vp9_mby_is_skippable_8x8(xd) & vp9_mbuv_is_skippable_8x8(xd)); } -static int mb_is_skippable_8x8_4x4uv(MACROBLOCKD *xd, int has_2nd_order) { - return (vp9_mby_is_skippable_8x8(xd, has_2nd_order) & +static int mb_is_skippable_8x8_4x4uv(MACROBLOCKD *xd) { + return (vp9_mby_is_skippable_8x8(xd) & vp9_mbuv_is_skippable_4x4(xd)); } @@ -360,10 +344,6 @@ void vp9_tokenize_sb(VP9_COMP *cpi, VP9_COMMON * const cm = &cpi->common; MB_MODE_INFO * const mbmi = &xd->mode_info_context->mbmi; TOKENEXTRA *t_backup = *t; - ENTROPY_CONTEXT *A[2] = { (ENTROPY_CONTEXT *) (xd->above_context + 0), - (ENTROPY_CONTEXT *) (xd->above_context + 1), }; - ENTROPY_CONTEXT *L[2] = { (ENTROPY_CONTEXT *) (xd->left_context + 0), - (ENTROPY_CONTEXT *) (xd->left_context + 1), }; const int mb_skip_context = vp9_get_pred_context(cm, xd, PRED_MBSKIP); const int segment_id = mbmi->segment_id; const int skip_inc = !vp9_segfeature_active(xd, segment_id, SEG_LVL_SKIP); @@ -394,7 +374,6 @@ void vp9_tokenize_sb(VP9_COMP *cpi, tokenize_b(cpi, xd, b, t, PLANE_TYPE_UV, TX_16X16, dry_run); } - A[0][8] = L[0][8] = A[1][8] = L[1][8] = 0; if (dry_run) *t = t_backup; } @@ -403,8 +382,6 @@ void vp9_tokenize_mb(VP9_COMP *cpi, MACROBLOCKD *xd, TOKENEXTRA **t, int dry_run) { - PLANE_TYPE plane_type; - int has_2nd_order; int b; int tx_size = xd->mode_info_context->mbmi.txfm_size; int mb_skip_context = vp9_get_pred_context(&cpi->common, xd, PRED_MBSKIP); @@ -421,8 +398,6 @@ void vp9_tokenize_mb(VP9_COMP *cpi, } else skip_inc = 0; - has_2nd_order = get_2nd_order_usage(xd); - switch (tx_size) { case TX_16X16: @@ -432,15 +407,15 @@ void vp9_tokenize_mb(VP9_COMP *cpi, if (xd->mode_info_context->mbmi.mode == I8X8_PRED || xd->mode_info_context->mbmi.mode == SPLITMV) xd->mode_info_context->mbmi.mb_skip_coeff = - mb_is_skippable_8x8_4x4uv(xd, 0); + mb_is_skippable_8x8_4x4uv(xd); else xd->mode_info_context->mbmi.mb_skip_coeff = - mb_is_skippable_8x8(xd, has_2nd_order); + mb_is_skippable_8x8(xd); break; default: xd->mode_info_context->mbmi.mb_skip_coeff = - mb_is_skippable_4x4(xd, has_2nd_order); + mb_is_skippable_4x4(xd); break; } @@ -461,15 +436,6 @@ void vp9_tokenize_mb(VP9_COMP *cpi, if (!dry_run) cpi->skip_false_count[mb_skip_context] += skip_inc; - if (has_2nd_order) { - tokenize_b(cpi, xd, 24, t, PLANE_TYPE_Y2, tx_size, dry_run); - plane_type = PLANE_TYPE_Y_NO_DC; - } else { - xd->above_context->y2 = 0; - xd->left_context->y2 = 0; - plane_type = PLANE_TYPE_Y_WITH_DC; - } - if (tx_size == TX_16X16) { tokenize_b(cpi, xd, 0, t, PLANE_TYPE_Y_WITH_DC, TX_16X16, dry_run); for (b = 16; b < 24; b += 4) { @@ -477,7 +443,7 @@ void vp9_tokenize_mb(VP9_COMP *cpi, } } else if (tx_size == TX_8X8) { for (b = 0; b < 16; b += 4) { - tokenize_b(cpi, xd, b, t, plane_type, TX_8X8, dry_run); + tokenize_b(cpi, xd, b, t, PLANE_TYPE_Y_WITH_DC, TX_8X8, dry_run); } if (xd->mode_info_context->mbmi.mode == I8X8_PRED || xd->mode_info_context->mbmi.mode == SPLITMV) { @@ -490,11 +456,10 @@ void vp9_tokenize_mb(VP9_COMP *cpi, } } } else { - for (b = 0; b < 24; b++) { - if (b >= 16) - plane_type = PLANE_TYPE_UV; - tokenize_b(cpi, xd, b, t, plane_type, TX_4X4, dry_run); - } + for (b = 0; b < 16; b++) + tokenize_b(cpi, xd, b, t, PLANE_TYPE_Y_WITH_DC, TX_4X4, dry_run); + for (b = 16; b < 24; b++) + tokenize_b(cpi, xd, b, t, PLANE_TYPE_UV, TX_4X4, dry_run); } if (dry_run) *t = t_backup; @@ -727,10 +692,8 @@ static INLINE void stuff_b(VP9_COMP *cpi, break; case TX_8X8: #if CONFIG_CNVCONTEXT - if (type != PLANE_TYPE_Y2) { - a_ec = (a[0] + a[1]) != 0; - l_ec = (l[0] + l[1]) != 0; - } + a_ec = (a[0] + a[1]) != 0; + l_ec = (l[0] + l[1]) != 0; #endif if (tx_type != DCT_DCT) { counts = cpi->hybrid_coef_counts_8x8; @@ -774,14 +737,14 @@ static INLINE void stuff_b(VP9_COMP *cpi, VP9_COMBINEENTROPYCONTEXTS(pt, a_ec, l_ec); - band = get_coef_band((type == PLANE_TYPE_Y_NO_DC) ? 1 : 0); + band = get_coef_band(0); t->Token = DCT_EOB_TOKEN; t->context_tree = probs[type][band][pt]; t->skip_eob_node = 0; ++t; *tp = t; *a = *l = 0; - if (tx_size == TX_8X8 && type != PLANE_TYPE_Y2) { + if (tx_size == TX_8X8) { a[1] = 0; l[1] = 0; } else if (tx_size == TX_16X16) { @@ -806,26 +769,12 @@ static INLINE void stuff_b(VP9_COMP *cpi, static void stuff_mb_8x8(VP9_COMP *cpi, MACROBLOCKD *xd, TOKENEXTRA **t, int dry_run) { - PLANE_TYPE plane_type; int b; - int has_2nd_order = get_2nd_order_usage(xd); - - if (has_2nd_order) { - stuff_b(cpi, xd, 24, t, PLANE_TYPE_Y2, TX_8X8, dry_run); - plane_type = PLANE_TYPE_Y_NO_DC; - } else { -#if CONFIG_CNVCONTEXT - xd->above_context->y2 = 0; - xd->left_context->y2 = 0; -#endif - plane_type = PLANE_TYPE_Y_WITH_DC; - } - for (b = 0; b < 24; b += 4) { - if (b >= 16) - plane_type = PLANE_TYPE_UV; - stuff_b(cpi, xd, b, t, plane_type, TX_8X8, dry_run); - } + for (b = 0; b < 16; b += 4) + stuff_b(cpi, xd, b, t, PLANE_TYPE_Y_WITH_DC, TX_8X8, dry_run); + for (b = 16; b < 24; b += 4) + stuff_b(cpi, xd, b, t, PLANE_TYPE_UV, TX_8X8, dry_run); } static void stuff_mb_16x16(VP9_COMP *cpi, MACROBLOCKD *xd, @@ -836,56 +785,26 @@ static void stuff_mb_16x16(VP9_COMP *cpi, MACROBLOCKD *xd, for (b = 16; b < 24; b += 4) { stuff_b(cpi, xd, b, t, PLANE_TYPE_UV, TX_8X8, dry_run); } -#if CONFIG_CNVCONTEXT - xd->above_context->y2 = 0; - xd->left_context->y2 = 0; -#endif } static void stuff_mb_4x4(VP9_COMP *cpi, MACROBLOCKD *xd, TOKENEXTRA **t, int dry_run) { int b; - PLANE_TYPE plane_type; - int has_2nd_order = get_2nd_order_usage(xd); - - if (has_2nd_order) { - stuff_b(cpi, xd, 24, t, PLANE_TYPE_Y2, TX_4X4, dry_run); - plane_type = PLANE_TYPE_Y_NO_DC; - } else { - xd->above_context->y2 = 0; - xd->left_context->y2 = 0; - plane_type = PLANE_TYPE_Y_WITH_DC; - } - for (b = 0; b < 24; b++) { - if (b >= 16) - plane_type = PLANE_TYPE_UV; - stuff_b(cpi, xd, b, t, plane_type, TX_4X4, dry_run); - } + for (b = 0; b < 16; b++) + stuff_b(cpi, xd, b, t, PLANE_TYPE_Y_WITH_DC, TX_4X4, dry_run); + for (b = 16; b < 24; b++) + stuff_b(cpi, xd, b, t, PLANE_TYPE_UV, TX_4X4, dry_run); } static void stuff_mb_8x8_4x4uv(VP9_COMP *cpi, MACROBLOCKD *xd, TOKENEXTRA **t, int dry_run) { - PLANE_TYPE plane_type; int b; - int has_2nd_order = get_2nd_order_usage(xd); - if (has_2nd_order) { - stuff_b(cpi, xd, 24, t, PLANE_TYPE_Y2, TX_8X8, dry_run); - plane_type = PLANE_TYPE_Y_NO_DC; - } else { - xd->above_context->y2 = 0; - xd->left_context->y2 = 0; - plane_type = PLANE_TYPE_Y_WITH_DC; - } - - for (b = 0; b < 16; b += 4) { - stuff_b(cpi, xd, b, t, plane_type, TX_8X8, dry_run); - } - - for (b = 16; b < 24; b++) { + for (b = 0; b < 16; b += 4) + stuff_b(cpi, xd, b, t, PLANE_TYPE_Y_WITH_DC, TX_8X8, dry_run); + for (b = 16; b < 24; b++) stuff_b(cpi, xd, b, t, PLANE_TYPE_UV, TX_4X4, dry_run); - } } void vp9_stuff_mb(VP9_COMP *cpi, MACROBLOCKD *xd, TOKENEXTRA **t, int dry_run) { diff --git a/vp9/encoder/vp9_tokenize.h b/vp9/encoder/vp9_tokenize.h index 3eeb8fa5a..c1a4c0520 100644 --- a/vp9/encoder/vp9_tokenize.h +++ b/vp9/encoder/vp9_tokenize.h @@ -31,9 +31,9 @@ typedef struct { typedef int64_t vp9_coeff_accum[COEF_BANDS][PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; -extern int vp9_mby_is_skippable_4x4(MACROBLOCKD *xd, int has_y2_block); +extern int vp9_mby_is_skippable_4x4(MACROBLOCKD *xd); extern int vp9_mbuv_is_skippable_4x4(MACROBLOCKD *xd); -extern int vp9_mby_is_skippable_8x8(MACROBLOCKD *xd, int has_y2_block); +extern int vp9_mby_is_skippable_8x8(MACROBLOCKD *xd); extern int vp9_mbuv_is_skippable_8x8(MACROBLOCKD *xd); extern int vp9_mby_is_skippable_16x16(MACROBLOCKD *xd); extern int vp9_sby_is_skippable_32x32(MACROBLOCKD *xd); diff --git a/vp9/encoder/x86/vp9_encodeopt.asm b/vp9/encoder/x86/vp9_encodeopt.asm index 5d9f7769d..90c793d4f 100644 --- a/vp9/encoder/x86/vp9_encodeopt.asm +++ b/vp9/encoder/x86/vp9_encodeopt.asm @@ -125,7 +125,7 @@ sym(vp9_block_error_mmx): ret -;int vp9_mbblock_error_mmx_impl(short *coeff_ptr, short *dcoef_ptr, int dc); +;int vp9_mbblock_error_mmx_impl(short *coeff_ptr, short *dcoef_ptr); global sym(vp9_mbblock_error_mmx_impl) PRIVATE sym(vp9_mbblock_error_mmx_impl): push rbp @@ -142,10 +142,6 @@ sym(vp9_mbblock_error_mmx_impl): mov rdi, arg(1) ;dcoef_ptr pxor mm2, mm2 - movd mm1, dword ptr arg(2) ;dc - por mm1, mm2 - - pcmpeqw mm1, mm7 mov rcx, 16 .mberror_loop_mmx: @@ -160,7 +156,6 @@ sym(vp9_mbblock_error_mmx_impl): pmaddwd mm5, mm5 psubw mm3, mm4 - pand mm3, mm1 pmaddwd mm3, mm3 paddd mm2, mm5 @@ -202,28 +197,24 @@ sym(vp9_mbblock_error_mmx_impl): ret -;int vp9_mbblock_error_xmm_impl(short *coeff_ptr, short *dcoef_ptr, int dc); +;int vp9_mbblock_error_xmm_impl(short *coeff_ptr, short *dcoef_ptr); global sym(vp9_mbblock_error_xmm_impl) PRIVATE sym(vp9_mbblock_error_xmm_impl): push rbp mov rbp, rsp SHADOW_ARGS_TO_STACK 3 - SAVE_XMM 6 + SAVE_XMM 5 push rsi push rdi ; end prolog mov rsi, arg(0) ;coeff_ptr - pxor xmm6, xmm6 + pxor xmm5, xmm5 mov rdi, arg(1) ;dcoef_ptr pxor xmm4, xmm4 - movd xmm5, dword ptr arg(2) ;dc - por xmm5, xmm4 - - pcmpeqw xmm5, xmm6 mov rcx, 16 .mberror_loop: @@ -238,7 +229,6 @@ sym(vp9_mbblock_error_xmm_impl): pmaddwd xmm2, xmm2 psubw xmm0, xmm1 - pand xmm0, xmm5 pmaddwd xmm0, xmm0 add rsi, 32 @@ -252,9 +242,9 @@ sym(vp9_mbblock_error_xmm_impl): jnz .mberror_loop movdqa xmm0, xmm4 - punpckldq xmm0, xmm6 + punpckldq xmm0, xmm5 - punpckhdq xmm4, xmm6 + punpckhdq xmm4, xmm5 paddd xmm0, xmm4 movdqa xmm1, xmm0 diff --git a/vp9/encoder/x86/vp9_x86_csystemdependent.c b/vp9/encoder/x86/vp9_x86_csystemdependent.c index 3beef53a2..2bf32c569 100644 --- a/vp9/encoder/x86/vp9_x86_csystemdependent.c +++ b/vp9/encoder/x86/vp9_x86_csystemdependent.c @@ -23,11 +23,11 @@ void vp9_short_fdct8x4_mmx(short *input, short *output, int pitch) { vp9_short_fdct4x4_mmx(input + 4, output + 16, pitch); } -int vp9_mbblock_error_mmx_impl(short *coeff_ptr, short *dcoef_ptr, int dc); -int vp9_mbblock_error_mmx(MACROBLOCK *mb, int dc) { +int vp9_mbblock_error_mmx_impl(short *coeff_ptr, short *dcoef_ptr); +int vp9_mbblock_error_mmx(MACROBLOCK *mb) { short *coeff_ptr = mb->block[0].coeff; short *dcoef_ptr = mb->e_mbd.block[0].dqcoeff; - return vp9_mbblock_error_mmx_impl(coeff_ptr, dcoef_ptr, dc); + return vp9_mbblock_error_mmx_impl(coeff_ptr, dcoef_ptr); } int vp9_mbuverror_mmx_impl(short *s_ptr, short *d_ptr); @@ -51,11 +51,11 @@ void vp9_subtract_b_mmx(BLOCK *be, BLOCKD *bd, int pitch) { #endif #if HAVE_SSE2 -int vp9_mbblock_error_xmm_impl(short *coeff_ptr, short *dcoef_ptr, int dc); -int vp9_mbblock_error_xmm(MACROBLOCK *mb, int dc) { +int vp9_mbblock_error_xmm_impl(short *coeff_ptr, short *dcoef_ptr); +int vp9_mbblock_error_xmm(MACROBLOCK *mb) { short *coeff_ptr = mb->block[0].coeff; short *dcoef_ptr = mb->e_mbd.block[0].dqcoeff; - return vp9_mbblock_error_xmm_impl(coeff_ptr, dcoef_ptr, dc); + return vp9_mbblock_error_xmm_impl(coeff_ptr, dcoef_ptr); } int vp9_mbuverror_xmm_impl(short *s_ptr, short *d_ptr); diff --git a/vp9/vp9cx.mk b/vp9/vp9cx.mk index b41873540..6e57e67ef 100644 --- a/vp9/vp9cx.mk +++ b/vp9/vp9cx.mk @@ -65,7 +65,6 @@ VP9_CX_SRCS-yes += encoder/vp9_quantize.c VP9_CX_SRCS-yes += encoder/vp9_ratectrl.c VP9_CX_SRCS-yes += encoder/vp9_rdopt.c VP9_CX_SRCS-yes += encoder/vp9_sad_c.c -VP9_CX_SRCS-yes += encoder/vp9_satd_c.c VP9_CX_SRCS-yes += encoder/vp9_segmentation.c VP9_CX_SRCS-yes += encoder/vp9_segmentation.h VP9_CX_SRCS-$(CONFIG_INTERNAL_STATS) += encoder/vp9_ssim.c |