Merge "Code cleanup." into experimental

This commit is contained in:
Dmitry Kovalev 2013-02-27 16:43:04 -08:00 committed by Gerrit Code Review
commit c6421433c8
4 changed files with 114 additions and 190 deletions

View File

@ -71,8 +71,7 @@ typedef enum {
INTER_FRAME = 1 INTER_FRAME = 1
} FRAME_TYPE; } FRAME_TYPE;
typedef enum typedef enum {
{
#if CONFIG_ENABLE_6TAP #if CONFIG_ENABLE_6TAP
SIXTAP, SIXTAP,
#endif #endif
@ -83,8 +82,7 @@ typedef enum
SWITCHABLE /* should be the last one */ SWITCHABLE /* should be the last one */
} INTERPOLATIONFILTERTYPE; } INTERPOLATIONFILTERTYPE;
typedef enum typedef enum {
{
DC_PRED, /* average of above and left pixels */ DC_PRED, /* average of above and left pixels */
V_PRED, /* vertical prediction */ V_PRED, /* vertical prediction */
H_PRED, /* horizontal prediction */ H_PRED, /* horizontal prediction */
@ -411,66 +409,38 @@ typedef struct macroblockd {
// convert MB_PREDICTION_MODE to B_PREDICTION_MODE // convert MB_PREDICTION_MODE to B_PREDICTION_MODE
static B_PREDICTION_MODE pred_mode_conv(MB_PREDICTION_MODE mode) { static B_PREDICTION_MODE pred_mode_conv(MB_PREDICTION_MODE mode) {
B_PREDICTION_MODE b_mode;
switch (mode) { switch (mode) {
case DC_PRED: case DC_PRED: return B_DC_PRED;
b_mode = B_DC_PRED; case V_PRED: return B_VE_PRED;
break; case H_PRED: return B_HE_PRED;
case V_PRED: case TM_PRED: return B_TM_PRED;
b_mode = B_VE_PRED; case D45_PRED: return B_LD_PRED;
break; case D135_PRED: return B_RD_PRED;
case H_PRED: case D117_PRED: return B_VR_PRED;
b_mode = B_HE_PRED; case D153_PRED: return B_HD_PRED;
break; case D27_PRED: return B_HU_PRED;
case TM_PRED: case D63_PRED: return B_VL_PRED;
b_mode = B_TM_PRED; default:
break;
case D45_PRED:
b_mode = B_LD_PRED;
break;
case D135_PRED:
b_mode = B_RD_PRED;
break;
case D117_PRED:
b_mode = B_VR_PRED;
break;
case D153_PRED:
b_mode = B_HD_PRED;
break;
case D27_PRED:
b_mode = B_HU_PRED;
break;
case D63_PRED:
b_mode = B_VL_PRED;
break;
default :
// for debug purpose, to be removed after full testing
assert(0); assert(0);
break; return B_MODE_COUNT; // Dummy value
} }
return b_mode;
} }
// transform mapping // transform mapping
static TX_TYPE txfm_map(B_PREDICTION_MODE bmode) { static TX_TYPE txfm_map(B_PREDICTION_MODE bmode) {
// map transform type
TX_TYPE tx_type;
switch (bmode) { switch (bmode) {
case B_TM_PRED : case B_TM_PRED :
case B_RD_PRED : case B_RD_PRED :
tx_type = ADST_ADST; return ADST_ADST;
break;
case B_VE_PRED : case B_VE_PRED :
case B_VR_PRED : case B_VR_PRED :
tx_type = ADST_DCT; return ADST_DCT;
break;
case B_HE_PRED : case B_HE_PRED :
case B_HD_PRED : case B_HD_PRED :
case B_HU_PRED : case B_HU_PRED :
tx_type = DCT_ADST; return DCT_ADST;
break;
#if CONFIG_NEWBINTRAMODES #if CONFIG_NEWBINTRAMODES
case B_CONTEXT_PRED: case B_CONTEXT_PRED:
@ -478,11 +448,9 @@ static TX_TYPE txfm_map(B_PREDICTION_MODE bmode) {
break; break;
#endif #endif
default : default:
tx_type = DCT_DCT; return DCT_DCT;
break;
} }
return tx_type;
} }
extern const uint8_t vp9_block2left[TX_SIZE_MAX_SB][24]; extern const uint8_t vp9_block2left[TX_SIZE_MAX_SB][24];
@ -622,16 +590,13 @@ void vp9_build_block_doffsets(MACROBLOCKD *xd);
void vp9_setup_block_dptrs(MACROBLOCKD *xd); void vp9_setup_block_dptrs(MACROBLOCKD *xd);
static void update_blockd_bmi(MACROBLOCKD *xd) { static void update_blockd_bmi(MACROBLOCKD *xd) {
int i; const MB_PREDICTION_MODE mode = xd->mode_info_context->mbmi.mode;
int is_4x4;
is_4x4 = (xd->mode_info_context->mbmi.mode == SPLITMV) ||
(xd->mode_info_context->mbmi.mode == I8X8_PRED) ||
(xd->mode_info_context->mbmi.mode == B_PRED);
if (is_4x4) { if (mode == SPLITMV || mode == I8X8_PRED || mode == B_PRED) {
for (i = 0; i < 16; i++) { int i;
for (i = 0; i < 16; i++)
xd->block[i].bmi = xd->mode_info_context->bmi[i]; xd->block[i].bmi = xd->mode_info_context->bmi[i];
} }
}
} }
#endif // VP9_COMMON_VP9_BLOCKD_H_ #endif // VP9_COMMON_VP9_BLOCKD_H_

View File

@ -8,8 +8,8 @@
* be found in the AUTHORS file in the root of the source tree. * be found in the AUTHORS file in the root of the source tree.
*/ */
#ifndef VP9_COMMON_VP9_IDCT_H #ifndef VP9_COMMON_VP9_IDCT_H_
#define VP9_COMMON_VP9_IDCT_H #define VP9_COMMON_VP9_IDCT_H_
#include "./vpx_config.h" #include "./vpx_config.h"
@ -75,4 +75,17 @@ static INLINE int dct_32_round(int input) {
return rv; return rv;
} }
#endif typedef void (*transform_1d)(int16_t*, int16_t*);
typedef struct {
transform_1d cols, rows; // vertical and horizontal
} transform_2d;
#define ROUND_POWER_OF_TWO(value, n) (((value) + (1 << ((n) - 1))) >> (n))
/* If we don't want to use ROUND_POWER_OF_TWO macro
static INLINE int16_t round_power_of_two(int16_t value, int n) {
return (value + (1 << (n - 1))) >> n;
}*/
#endif // VP9_COMMON_VP9_IDCT_H_

View File

@ -31,12 +31,6 @@
#include "vp9/common/vp9_common.h" #include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_idct.h" #include "vp9/common/vp9_idct.h"
typedef void (*transform_1d)(int16_t*, int16_t*);
typedef struct {
transform_1d cols, rows; // vertical and horizontal
} transform_2d;
void vp9_short_inv_walsh4x4_x8_c(int16_t *input, int16_t *output, int pitch) { void vp9_short_inv_walsh4x4_x8_c(int16_t *input, int16_t *output, int pitch) {
int i; int i;
int a1, b1, c1, d1; int a1, b1, c1, d1;

View File

@ -36,13 +36,14 @@ static void fdct4_1d(int16_t *input, int16_t *output) {
output[3] = dct_const_round_shift(temp2); output[3] = dct_const_round_shift(temp2);
} }
void vp9_short_fdct4x4_c(short *input, short *output, int pitch) { void vp9_short_fdct4x4_c(int16_t *input, int16_t *output, int pitch) {
int16_t out[4 * 4]; int16_t out[4 * 4];
int16_t *outptr = &out[0]; int16_t *outptr = &out[0];
const int short_pitch = pitch >> 1; const int short_pitch = pitch >> 1;
int i, j; int i, j;
int16_t temp_in[4], temp_out[4]; int16_t temp_in[4], temp_out[4];
// First transform cols
// Columns
for (i = 0; i < 4; ++i) { for (i = 0; i < 4; ++i) {
for (j = 0; j < 4; ++j) for (j = 0; j < 4; ++j)
temp_in[j] = input[j * short_pitch + i] << 4; temp_in[j] = input[j * short_pitch + i] << 4;
@ -52,6 +53,8 @@ void vp9_short_fdct4x4_c(short *input, short *output, int pitch) {
for (j = 0; j < 4; ++j) for (j = 0; j < 4; ++j)
outptr[j * 4 + i] = temp_out[j]; outptr[j * 4 + i] = temp_out[j];
} }
// Rows
for (i = 0; i < 4; ++i) { for (i = 0; i < 4; ++i) {
for (j = 0; j < 4; ++j) for (j = 0; j < 4; ++j)
temp_in[j] = out[j + i * 4]; temp_in[j] = out[j + i * 4];
@ -101,61 +104,43 @@ static void fadst4_1d(int16_t *input, int16_t *output) {
output[3] = dct_const_round_shift(s3); output[3] = dct_const_round_shift(s3);
} }
static const transform_2d FHT_4[] = {
{ fdct4_1d, fdct4_1d }, // DCT_DCT = 0
{ fadst4_1d, fdct4_1d }, // ADST_DCT = 1
{ fdct4_1d, fadst4_1d }, // DCT_ADST = 2
{ fadst4_1d, fadst4_1d } // ADST_ADST = 3
};
void vp9_short_fht4x4_c(int16_t *input, int16_t *output, void vp9_short_fht4x4_c(int16_t *input, int16_t *output,
int pitch, TX_TYPE tx_type) { int pitch, TX_TYPE tx_type) {
int16_t out[4 * 4]; int16_t out[4 * 4];
int16_t *outptr = &out[0]; int16_t *outptr = &out[0];
int i, j; int i, j;
int16_t temp_in[4], temp_out[4]; int16_t temp_in[4], temp_out[4];
const transform_2d ht = FHT_4[tx_type];
void (*fwdr)(int16_t*, int16_t*); // Columns
void (*fwdc)(int16_t*, int16_t*);
switch (tx_type) {
case ADST_ADST:
fwdc = &fadst4_1d;
fwdr = &fadst4_1d;
break;
case ADST_DCT:
fwdc = &fadst4_1d;
fwdr = &fdct4_1d;
break;
case DCT_ADST:
fwdc = &fdct4_1d;
fwdr = &fadst4_1d;
break;
case DCT_DCT:
fwdc = &fdct4_1d;
fwdr = &fdct4_1d;
break;
default:
assert(0);
}
// column transform
for (i = 0; i < 4; ++i) { for (i = 0; i < 4; ++i) {
for (j = 0; j < 4; ++j) for (j = 0; j < 4; ++j)
temp_in[j] = input[j * pitch + i] << 4; temp_in[j] = input[j * pitch + i] << 4;
if (i == 0 && temp_in[0]) if (i == 0 && temp_in[0])
temp_in[0] += 1; temp_in[0] += 1;
fwdc(temp_in, temp_out); ht.cols(temp_in, temp_out);
for (j = 0; j < 4; ++j) for (j = 0; j < 4; ++j)
outptr[j * 4 + i] = temp_out[j]; outptr[j * 4 + i] = temp_out[j];
} }
// row transform // Rows
for (i = 0; i < 4; ++i) { for (i = 0; i < 4; ++i) {
for (j = 0; j < 4; ++j) for (j = 0; j < 4; ++j)
temp_in[j] = out[j + i * 4]; temp_in[j] = out[j + i * 4];
fwdr(temp_in, temp_out); ht.rows(temp_in, temp_out);
for (j = 0; j < 4; ++j) for (j = 0; j < 4; ++j)
output[j + i * 4] = (temp_out[j] + 1) >> 2; output[j + i * 4] = (temp_out[j] + 1) >> 2;
} }
} }
void vp9_short_fdct8x4_c(short *input, short *output, int pitch) void vp9_short_fdct8x4_c(int16_t *input, int16_t *output, int pitch) {
{
vp9_short_fdct4x4_c(input, output, pitch); vp9_short_fdct4x4_c(input, output, pitch);
vp9_short_fdct4x4_c(input + 4, output + 16, pitch); vp9_short_fdct4x4_c(input + 4, output + 16, pitch);
} }
@ -212,7 +197,7 @@ void vp9_short_fdct8x8_c(int16_t *input, int16_t *output, int pitch) {
int16_t out[64]; int16_t out[64];
int16_t temp_in[8], temp_out[8]; int16_t temp_in[8], temp_out[8];
// First transform columns // Columns
for (i = 0; i < 8; i++) { for (i = 0; i < 8; i++) {
for (j = 0; j < 8; j++) for (j = 0; j < 8; j++)
temp_in[j] = input[j * shortpitch + i] << 2; temp_in[j] = input[j * shortpitch + i] << 2;
@ -221,7 +206,7 @@ void vp9_short_fdct8x8_c(int16_t *input, int16_t *output, int pitch) {
out[j * 8 + i] = temp_out[j]; out[j * 8 + i] = temp_out[j];
} }
// Then transform rows // Rows
for (i = 0; i < 8; ++i) { for (i = 0; i < 8; ++i) {
for (j = 0; j < 8; ++j) for (j = 0; j < 8; ++j)
temp_in[j] = out[j + i * 8]; temp_in[j] = out[j + i * 8];
@ -232,17 +217,16 @@ void vp9_short_fdct8x8_c(int16_t *input, int16_t *output, int pitch) {
} }
static void fadst8_1d(int16_t *input, int16_t *output) { static void fadst8_1d(int16_t *input, int16_t *output) {
int x0, x1, x2, x3, x4, x5, x6, x7;
int s0, s1, s2, s3, s4, s5, s6, s7; int s0, s1, s2, s3, s4, s5, s6, s7;
x0 = input[7]; int x0 = input[7];
x1 = input[0]; int x1 = input[0];
x2 = input[5]; int x2 = input[5];
x3 = input[2]; int x3 = input[2];
x4 = input[3]; int x4 = input[3];
x5 = input[4]; int x5 = input[4];
x6 = input[1]; int x6 = input[1];
x7 = input[6]; int x7 = input[6];
// stage 1 // stage 1
s0 = cospi_2_64 * x0 + cospi_30_64 * x1; s0 = cospi_2_64 * x0 + cospi_30_64 * x1;
@ -303,51 +287,35 @@ static void fadst8_1d(int16_t *input, int16_t *output) {
output[7] = - x1; output[7] = - x1;
} }
static const transform_2d FHT_8[] = {
{ fdct8_1d, fdct8_1d }, // DCT_DCT = 0
{ fadst8_1d, fdct8_1d }, // ADST_DCT = 1
{ fdct8_1d, fadst8_1d }, // DCT_ADST = 2
{ fadst8_1d, fadst8_1d } // ADST_ADST = 3
};
void vp9_short_fht8x8_c(int16_t *input, int16_t *output, void vp9_short_fht8x8_c(int16_t *input, int16_t *output,
int pitch, TX_TYPE tx_type) { int pitch, TX_TYPE tx_type) {
int16_t out[64]; int16_t out[64];
int16_t *outptr = &out[0]; int16_t *outptr = &out[0];
int i, j; int i, j;
int16_t temp_in[8], temp_out[8]; int16_t temp_in[8], temp_out[8];
const transform_2d ht = FHT_8[tx_type];
void (*fwdr)(int16_t*, int16_t*); // Columns
void (*fwdc)(int16_t*, int16_t*);
switch (tx_type) {
case ADST_ADST:
fwdc = &fadst8_1d;
fwdr = &fadst8_1d;
break;
case ADST_DCT:
fwdc = &fadst8_1d;
fwdr = &fdct8_1d;
break;
case DCT_ADST:
fwdc = &fdct8_1d;
fwdr = &fadst8_1d;
break;
case DCT_DCT:
fwdc = &fdct8_1d;
fwdr = &fdct8_1d;
break;
default:
assert(0);
}
// column transform
for (i = 0; i < 8; ++i) { for (i = 0; i < 8; ++i) {
for (j = 0; j < 8; ++j) for (j = 0; j < 8; ++j)
temp_in[j] = input[j * pitch + i] << 2; temp_in[j] = input[j * pitch + i] << 2;
fwdc(temp_in, temp_out); ht.cols(temp_in, temp_out);
for (j = 0; j < 8; ++j) for (j = 0; j < 8; ++j)
outptr[j * 8 + i] = temp_out[j]; outptr[j * 8 + i] = temp_out[j];
} }
// row transform // Rows
for (i = 0; i < 8; ++i) { for (i = 0; i < 8; ++i) {
for (j = 0; j < 8; ++j) for (j = 0; j < 8; ++j)
temp_in[j] = out[j + i * 8]; temp_in[j] = out[j + i * 8];
fwdr(temp_in, temp_out); ht.rows(temp_in, temp_out);
for (j = 0; j < 8; ++j) for (j = 0; j < 8; ++j)
output[j + i * 8] = temp_out[j] >> 1; output[j + i * 8] = temp_out[j] >> 1;
} }
@ -529,25 +497,24 @@ void vp9_short_fdct16x16_c(int16_t *input, int16_t *out, int pitch) {
} }
void fadst16_1d(int16_t *input, int16_t *output) { void fadst16_1d(int16_t *input, int16_t *output) {
int x0, x1, x2, x3, x4, x5, x6, x7, x8, x9, x10, x11, x12, x13, x14, x15;
int s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, s14, s15; int s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, s14, s15;
x0 = input[15]; int x0 = input[15];
x1 = input[0]; int x1 = input[0];
x2 = input[13]; int x2 = input[13];
x3 = input[2]; int x3 = input[2];
x4 = input[11]; int x4 = input[11];
x5 = input[4]; int x5 = input[4];
x6 = input[9]; int x6 = input[9];
x7 = input[6]; int x7 = input[6];
x8 = input[7]; int x8 = input[7];
x9 = input[8]; int x9 = input[8];
x10 = input[5]; int x10 = input[5];
x11 = input[10]; int x11 = input[10];
x12 = input[3]; int x12 = input[3];
x13 = input[12]; int x13 = input[12];
x14 = input[1]; int x14 = input[1];
x15 = input[14]; int x15 = input[14];
// stage 1 // stage 1
s0 = x0 * cospi_1_64 + x1 * cospi_31_64; s0 = x0 * cospi_1_64 + x1 * cospi_31_64;
@ -691,51 +658,35 @@ void fadst16_1d(int16_t *input, int16_t *output) {
output[15] = - x1; output[15] = - x1;
} }
static const transform_2d FHT_16[] = {
{ fdct16_1d, fdct16_1d }, // DCT_DCT = 0
{ fadst16_1d, fdct16_1d }, // ADST_DCT = 1
{ fdct16_1d, fadst16_1d }, // DCT_ADST = 2
{ fadst16_1d, fadst16_1d } // ADST_ADST = 3
};
void vp9_short_fht16x16_c(int16_t *input, int16_t *output, void vp9_short_fht16x16_c(int16_t *input, int16_t *output,
int pitch, TX_TYPE tx_type) { int pitch, TX_TYPE tx_type) {
int16_t out[256]; int16_t out[256];
int16_t *outptr = &out[0]; int16_t *outptr = &out[0];
int i, j; int i, j;
int16_t temp_in[16], temp_out[16]; int16_t temp_in[16], temp_out[16];
const transform_2d ht = FHT_16[tx_type];
void (*fwdr)(int16_t*, int16_t*); // Columns
void (*fwdc)(int16_t*, int16_t*);
switch (tx_type) {
case ADST_ADST:
fwdc = &fadst16_1d;
fwdr = &fadst16_1d;
break;
case ADST_DCT:
fwdc = &fadst16_1d;
fwdr = &fdct16_1d;
break;
case DCT_ADST:
fwdc = &fdct16_1d;
fwdr = &fadst16_1d;
break;
case DCT_DCT:
fwdc = &fdct16_1d;
fwdr = &fdct16_1d;
break;
default:
assert(0);
}
// column transform
for (i = 0; i < 16; ++i) { for (i = 0; i < 16; ++i) {
for (j = 0; j < 16; ++j) for (j = 0; j < 16; ++j)
temp_in[j] = input[j * pitch + i] << 2; temp_in[j] = input[j * pitch + i] << 2;
fwdc(temp_in, temp_out); ht.cols(temp_in, temp_out);
for (j = 0; j < 16; ++j) for (j = 0; j < 16; ++j)
outptr[j * 16 + i] = (temp_out[j] + 1 + (temp_out[j] > 0)) >> 2; outptr[j * 16 + i] = (temp_out[j] + 1 + (temp_out[j] > 0)) >> 2;
} }
// row transform // Rows
for (i = 0; i < 16; ++i) { for (i = 0; i < 16; ++i) {
for (j = 0; j < 16; ++j) for (j = 0; j < 16; ++j)
temp_in[j] = out[j + i * 16]; temp_in[j] = out[j + i * 16];
fwdr(temp_in, temp_out); ht.rows(temp_in, temp_out);
for (j = 0; j < 16; ++j) for (j = 0; j < 16; ++j)
output[j + i * 16] = temp_out[j]; output[j + i * 16] = temp_out[j];
} }
@ -1030,8 +981,9 @@ static void dct32_1d(int *input, int *output) {
void vp9_short_fdct32x32_c(int16_t *input, int16_t *out, int pitch) { void vp9_short_fdct32x32_c(int16_t *input, int16_t *out, int pitch) {
int shortpitch = pitch >> 1; int shortpitch = pitch >> 1;
int i, j; int i, j;
int output[1024]; int output[32 * 32];
// First transform columns
// Columns
for (i = 0; i < 32; i++) { for (i = 0; i < 32; i++) {
int temp_in[32], temp_out[32]; int temp_in[32], temp_out[32];
for (j = 0; j < 32; j++) for (j = 0; j < 32; j++)
@ -1041,7 +993,7 @@ void vp9_short_fdct32x32_c(int16_t *input, int16_t *out, int pitch) {
output[j * 32 + i] = (temp_out[j] + 1 + (temp_out[j] > 0)) >> 2; output[j * 32 + i] = (temp_out[j] + 1 + (temp_out[j] > 0)) >> 2;
} }
// Then transform rows // Rows
for (i = 0; i < 32; ++i) { for (i = 0; i < 32; ++i) {
int temp_in[32], temp_out[32]; int temp_in[32], temp_out[32];
for (j = 0; j < 32; ++j) for (j = 0; j < 32; ++j)