Bump dates to 2011
[x262.git] / common / ppc / quant.c
index 66aaec7..7223dd0 100644 (file)
 /*****************************************************************************
-* quant.c: h264 encoder
-*****************************************************************************
-* Authors: Guillaume Poirier <poirierg@gmail.com>
-*
-* This program is free software; you can redistribute it and/or modify
-* it under the terms of the GNU General Public License as published by
-* the Free Software Foundation; either version 2 of the License, or
-* (at your option) any later version.
-*
-* This program is distributed in the hope that it will be useful,
-* but WITHOUT ANY WARRANTY; without even the implied warranty of
-* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-* GNU General Public License for more details.
-*
-* You should have received a copy of the GNU General Public License
-* along with this program; if not, write to the Free Software
-* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
-*****************************************************************************/
+ * quant.c: ppc quantization
+ *****************************************************************************
+ * Copyright (C) 2007-2011 x264 project
+ *
+ * Authors: Guillaume Poirier <gpoirier@mplayerhq.hu>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
+ *
+ * This program is also available under a commercial proprietary license.
+ * For more information, contact us at licensing@x264.com.
+ *****************************************************************************/
 
 #include "common/common.h"
 #include "ppccommon.h"
-#include "quant.h"            
+#include "quant.h"
 
+#if !HIGH_BIT_DEPTH
 // quant of a whole 4x4 block, unrolled 2x and "pre-scheduled"
-#define QUANT_16_U( dct0, dct1, quant_mf0, quant_mf1, quant_mf2, quant_mf3 ) \
-temp1v = vec_ld((dct0), *dct);                                               \
-temp2v = vec_ld((dct1), *dct);                                               \
-mfvA = (vec_u16_t) vec_packs((vec_u32_t)vec_ld((quant_mf0), *quant_mf), (vec_u32_t)vec_ld((quant_mf1), *quant_mf));    \
-mfvB = (vec_u16_t) vec_packs((vec_u32_t)vec_ld((quant_mf2), *quant_mf), (vec_u32_t)vec_ld((quant_mf3), *quant_mf));    \
-mskA = vec_cmplt(temp1v, zerov);                                             \
-mskB = vec_cmplt(temp2v, zerov);                                             \
-coefvA = (vec_u16_t)vec_max(vec_sub(zerov, temp1v), temp1v);                 \
-coefvB = (vec_u16_t)vec_max(vec_sub(zerov, temp2v), temp2v);                 \
-multEvenvA = vec_mule(coefvA, mfvA);                                         \
-multOddvA = vec_mulo(coefvA, mfvA);                                          \
-multEvenvB = vec_mule(coefvB, mfvB);                                         \
-multOddvB = vec_mulo(coefvB, mfvB);                                          \
-multEvenvA = vec_adds(multEvenvA, fV);                                        \
-multOddvA = vec_adds(multOddvA, fV);                                          \
-multEvenvB = vec_adds(multEvenvB, fV);                                        \
-multOddvB = vec_adds(multOddvB, fV);                                          \
-multEvenvA = vec_sr(multEvenvA, i_qbitsv);                                   \
-multOddvA = vec_sr(multOddvA, i_qbitsv);                                     \
-multEvenvB = vec_sr(multEvenvB, i_qbitsv);                                   \
-multOddvB = vec_sr(multOddvB, i_qbitsv);                                     \
-temp1v = (vec_s16_t) vec_packs(vec_mergeh(multEvenvA, multOddvA), vec_mergel(multEvenvA, multOddvA)); \
-temp2v = (vec_s16_t) vec_packs(vec_mergeh(multEvenvB, multOddvB), vec_mergel(multEvenvB, multOddvB)); \
-temp1v = vec_xor(temp1v, mskA);                                              \
-temp2v = vec_xor(temp2v, mskB);                                              \
-temp1v = vec_adds(temp1v, vec_and(mskA, one));                                \
-vec_st(temp1v, (dct0), (int16_t*)dct);                                        \
-temp2v = vec_adds(temp2v, vec_and(mskB, one));                                \
-vec_st(temp2v, (dct1), (int16_t*)dct);
-                
-void x264_quant_4x4_altivec( int16_t dct[4][4], int quant_mf[4][4], int const i_qbits, int const f ) {
+#define QUANT_16_U( idx0, idx1 )                                    \
+{                                                                   \
+    temp1v = vec_ld((idx0), dct);                                   \
+    temp2v = vec_ld((idx1), dct);                                   \
+    mfvA = vec_ld((idx0), mf);                                      \
+    mfvB = vec_ld((idx1), mf);                                      \
+    biasvA = vec_ld((idx0), bias);                                  \
+    biasvB = vec_ld((idx1), bias);                                  \
+    mskA = vec_cmplt(temp1v, zero_s16v);                            \
+    mskB = vec_cmplt(temp2v, zero_s16v);                            \
+    coefvA = (vec_u16_t)vec_max(vec_sub(zero_s16v, temp1v), temp1v);\
+    coefvB = (vec_u16_t)vec_max(vec_sub(zero_s16v, temp2v), temp2v);\
+    coefvA = vec_adds(coefvA, biasvA);                              \
+    coefvB = vec_adds(coefvB, biasvB);                              \
+    multEvenvA = vec_mule(coefvA, mfvA);                            \
+    multOddvA = vec_mulo(coefvA, mfvA);                             \
+    multEvenvB = vec_mule(coefvB, mfvB);                            \
+    multOddvB = vec_mulo(coefvB, mfvB);                             \
+    multEvenvA = vec_sr(multEvenvA, i_qbitsv);                      \
+    multOddvA = vec_sr(multOddvA, i_qbitsv);                        \
+    multEvenvB = vec_sr(multEvenvB, i_qbitsv);                      \
+    multOddvB = vec_sr(multOddvB, i_qbitsv);                        \
+    temp1v = (vec_s16_t) vec_packs(vec_mergeh(multEvenvA, multOddvA), vec_mergel(multEvenvA, multOddvA)); \
+    temp2v = (vec_s16_t) vec_packs(vec_mergeh(multEvenvB, multOddvB), vec_mergel(multEvenvB, multOddvB)); \
+    temp1v = vec_xor(temp1v, mskA);                                 \
+    temp2v = vec_xor(temp2v, mskB);                                 \
+    temp1v = vec_adds(temp1v, vec_and(mskA, one));                  \
+    vec_st(temp1v, (idx0), dct);                                    \
+    temp2v = vec_adds(temp2v, vec_and(mskB, one));                  \
+    nz = vec_or(nz, vec_or(temp1v, temp2v));                        \
+    vec_st(temp2v, (idx1), dct);                                    \
+}
+
+int x264_quant_4x4_altivec( int16_t dct[16], uint16_t mf[16], uint16_t bias[16] )
+{
+    LOAD_ZERO;
     vector bool short mskA;
     vec_u32_t i_qbitsv;
     vec_u16_t coefvA;
     vec_u32_t multEvenvA, multOddvA;
     vec_u16_t mfvA;
-    vec_s16_t zerov, one;
-    vec_u32_t fV;
+    vec_u16_t biasvA;
+    vec_s16_t one = vec_splat_s16(1);;
+    vec_s16_t nz = zero_s16v;
 
     vector bool short mskB;
     vec_u16_t coefvB;
     vec_u32_t multEvenvB, multOddvB;
     vec_u16_t mfvB;
+    vec_u16_t biasvB;
 
     vec_s16_t temp1v, temp2v;
 
-    vect_int_u qbits_u;
-    qbits_u.s[0]=i_qbits;
+    vec_u32_u qbits_u;
+    qbits_u.s[0]=16;
     i_qbitsv = vec_splat(qbits_u.v, 0);
 
-    vect_int_u f_u;
-    f_u.s[0]=f;
-
-    fV = vec_splat(f_u.v, 0);
-
-    zerov = vec_splat_s16(0);
-    one = vec_splat_s16(1);
-
-    QUANT_16_U( 0, 16, 0, 16, 32, 48 );
+    QUANT_16_U( 0, 16 );
+    return vec_any_ne(nz, zero_s16v);
 }
 
 // DC quant of a whole 4x4 block, unrolled 2x and "pre-scheduled"
-#define QUANT_16_U_DC( dct0, dct1 )                             \
-temp1v = vec_ld((dct0), *dct);                                  \
-temp2v = vec_ld((dct1), *dct);                                  \
-mskA = vec_cmplt(temp1v, zerov);                                \
-mskB = vec_cmplt(temp2v, zerov);                                \
-coefvA = (vec_u16_t) vec_max(vec_sub(zerov, temp1v), temp1v);   \
-coefvB = (vec_u16_t) vec_max(vec_sub(zerov, temp2v), temp2v);   \
-multEvenvA = vec_mule(coefvA, mfv);                             \
-multOddvA = vec_mulo(coefvA, mfv);                              \
-multEvenvB = vec_mule(coefvB, mfv);                             \
-multOddvB = vec_mulo(coefvB, mfv);                              \
-multEvenvA = vec_add(multEvenvA, fV);                           \
-multOddvA = vec_add(multOddvA, fV);                             \
-multEvenvB = vec_add(multEvenvB, fV);                           \
-multOddvB = vec_add(multOddvB, fV);                             \
-multEvenvA = vec_sr(multEvenvA, i_qbitsv);                      \
-multOddvA = vec_sr(multOddvA, i_qbitsv);                        \
-multEvenvB = vec_sr(multEvenvB, i_qbitsv);                      \
-multOddvB = vec_sr(multOddvB, i_qbitsv);                        \
-temp1v = (vec_s16_t) vec_packs(vec_mergeh(multEvenvA, multOddvA), vec_mergel(multEvenvA, multOddvA)); \
-temp2v = (vec_s16_t) vec_packs(vec_mergeh(multEvenvB, multOddvB), vec_mergel(multEvenvB, multOddvB)); \
-temp1v = vec_xor(temp1v, mskA);                                 \
-temp2v = vec_xor(temp2v, mskB);                                 \
-temp1v = vec_add(temp1v, vec_and(mskA, one));                   \
-vec_st(temp1v, (dct0), (int16_t*)dct);                          \
-temp2v = vec_add(temp2v, vec_and(mskB, one));                   \
-vec_st(temp2v, (dct1), (int16_t*)dct);
-
-
-void x264_quant_4x4_dc_altivec( int16_t dct[4][4], int i_quant_mf, int const i_qbits, int const f ) {
+#define QUANT_16_U_DC( idx0, idx1 )                                 \
+{                                                                   \
+    temp1v = vec_ld((idx0), dct);                                   \
+    temp2v = vec_ld((idx1), dct);                                   \
+    mskA = vec_cmplt(temp1v, zero_s16v);                            \
+    mskB = vec_cmplt(temp2v, zero_s16v);                            \
+    coefvA = (vec_u16_t)vec_max(vec_sub(zero_s16v, temp1v), temp1v);\
+    coefvB = (vec_u16_t)vec_max(vec_sub(zero_s16v, temp2v), temp2v);\
+    coefvA = vec_add(coefvA, biasv);                                \
+    coefvB = vec_add(coefvB, biasv);                                \
+    multEvenvA = vec_mule(coefvA, mfv);                             \
+    multOddvA = vec_mulo(coefvA, mfv);                              \
+    multEvenvB = vec_mule(coefvB, mfv);                             \
+    multOddvB = vec_mulo(coefvB, mfv);                              \
+    multEvenvA = vec_sr(multEvenvA, i_qbitsv);                      \
+    multOddvA = vec_sr(multOddvA, i_qbitsv);                        \
+    multEvenvB = vec_sr(multEvenvB, i_qbitsv);                      \
+    multOddvB = vec_sr(multOddvB, i_qbitsv);                        \
+    temp1v = (vec_s16_t) vec_packs(vec_mergeh(multEvenvA, multOddvA), vec_mergel(multEvenvA, multOddvA)); \
+    temp2v = (vec_s16_t) vec_packs(vec_mergeh(multEvenvB, multOddvB), vec_mergel(multEvenvB, multOddvB)); \
+    temp1v = vec_xor(temp1v, mskA);                                 \
+    temp2v = vec_xor(temp2v, mskB);                                 \
+    temp1v = vec_add(temp1v, vec_and(mskA, one));                   \
+    vec_st(temp1v, (idx0), dct);                                    \
+    temp2v = vec_add(temp2v, vec_and(mskB, one));                   \
+    nz = vec_or(nz, vec_or(temp1v, temp2v));                        \
+    vec_st(temp2v, (idx1), dct);                                    \
+}
+
+int x264_quant_4x4_dc_altivec( int16_t dct[16], int mf, int bias )
+{
+    LOAD_ZERO;
     vector bool short mskA;
     vec_u32_t i_qbitsv;
     vec_u16_t coefvA;
     vec_u32_t multEvenvA, multOddvA;
-    vec_s16_t zerov, one;
-    vec_u32_t fV;
+    vec_s16_t one = vec_splat_s16(1);
+    vec_s16_t nz = zero_s16v;
 
     vector bool short mskB;
     vec_u16_t coefvB;
@@ -129,56 +137,234 @@ void x264_quant_4x4_dc_altivec( int16_t dct[4][4], int i_quant_mf, int const i_q
     vec_s16_t temp1v, temp2v;
 
     vec_u16_t mfv;
-    vect_ushort_u mf_u;
-    mf_u.s[0]=i_quant_mf;
+    vec_u16_t biasv;
+
+    vec_u16_u mf_u;
+    mf_u.s[0]=mf;
     mfv = vec_splat( mf_u.v, 0 );
 
-    vect_int_u qbits_u;
-    qbits_u.s[0]=i_qbits;
+    vec_u32_u qbits_u;
+    qbits_u.s[0]=16;
     i_qbitsv = vec_splat(qbits_u.v, 0);
 
-    vect_int_u f_u;
-    f_u.s[0]=f;
-    fV = vec_splat(f_u.v, 0);
-
-    zerov = vec_splat_s16(0);
-    one = vec_splat_s16(1);
+    vec_u16_u bias_u;
+    bias_u.s[0]=bias;
+    biasv = vec_splat(bias_u.v, 0);
 
     QUANT_16_U_DC( 0, 16 );
+    return vec_any_ne(nz, zero_s16v);
+}
+
+// DC quant of a whole 2x2 block
+#define QUANT_4_U_DC( idx0 )                                        \
+{                                                                   \
+    const vec_u16_t sel = (vec_u16_t) CV(-1,-1,-1,-1,0,0,0,0);      \
+    temp1v = vec_ld((idx0), dct);                                   \
+    mskA = vec_cmplt(temp1v, zero_s16v);                            \
+    coefvA = (vec_u16_t)vec_max(vec_sub(zero_s16v, temp1v), temp1v);\
+    coefvA = vec_add(coefvA, biasv);                                \
+    multEvenvA = vec_mule(coefvA, mfv);                             \
+    multOddvA = vec_mulo(coefvA, mfv);                              \
+    multEvenvA = vec_sr(multEvenvA, i_qbitsv);                      \
+    multOddvA = vec_sr(multOddvA, i_qbitsv);                        \
+    temp2v = (vec_s16_t) vec_packs(vec_mergeh(multEvenvA, multOddvA), vec_mergel(multEvenvA, multOddvA)); \
+    temp2v = vec_xor(temp2v, mskA);                                 \
+    temp2v = vec_add(temp2v, vec_and(mskA, one));                   \
+    temp1v = vec_sel(temp1v, temp2v, sel);                          \
+    nz = vec_or(nz, temp1v);                                        \
+    vec_st(temp1v, (idx0), dct);                                    \
 }
 
+int x264_quant_2x2_dc_altivec( int16_t dct[4], int mf, int bias )
+{
+    LOAD_ZERO;
+    vector bool short mskA;
+    vec_u32_t i_qbitsv;
+    vec_u16_t coefvA;
+    vec_u32_t multEvenvA, multOddvA;
+    vec_s16_t one = vec_splat_s16(1);
+    vec_s16_t nz = zero_s16v;
+
+    vec_s16_t temp1v, temp2v;
+
+    vec_u16_t mfv;
+    vec_u16_t biasv;
+
+    vec_u16_u mf_u;
+    mf_u.s[0]=mf;
+    mfv = vec_splat( mf_u.v, 0 );
+
+    vec_u32_u qbits_u;
+    qbits_u.s[0]=16;
+    i_qbitsv = vec_splat(qbits_u.v, 0);
+
+    vec_u16_u bias_u;
+    bias_u.s[0]=bias;
+    biasv = vec_splat(bias_u.v, 0);
+
+    static const vec_s16_t mask2 = CV(-1, -1, -1, -1,  0, 0, 0, 0);
+    QUANT_4_U_DC(0);
+    return vec_any_ne(vec_and(nz, mask2), zero_s16v);
+}
 
-void x264_quant_8x8_altivec( int16_t dct[8][8], int quant_mf[8][8], int const i_qbits, int const f ) {
+int x264_quant_8x8_altivec( int16_t dct[64], uint16_t mf[64], uint16_t bias[64] )
+{
+    LOAD_ZERO;
     vector bool short mskA;
     vec_u32_t i_qbitsv;
     vec_u16_t coefvA;
     vec_u32_t multEvenvA, multOddvA;
     vec_u16_t mfvA;
-    vec_s16_t zerov, one;
-    vec_u32_t fV;
-    
+    vec_u16_t biasvA;
+    vec_s16_t one = vec_splat_s16(1);;
+    vec_s16_t nz = zero_s16v;
+
     vector bool short mskB;
     vec_u16_t coefvB;
     vec_u32_t multEvenvB, multOddvB;
     vec_u16_t mfvB;
-    
+    vec_u16_t biasvB;
+
     vec_s16_t temp1v, temp2v;
-    
-    vect_int_u qbits_u;
-    qbits_u.s[0]=i_qbits;
+
+    vec_u32_u qbits_u;
+    qbits_u.s[0]=16;
     i_qbitsv = vec_splat(qbits_u.v, 0);
 
-    vect_int_u f_u;
-    f_u.s[0]=f;
-    fV = vec_splat(f_u.v, 0);
+    for( int i = 0; i < 4; i++ )
+        QUANT_16_U( i*2*16, i*2*16+16 );
+    return vec_any_ne(nz, zero_s16v);
+}
+
+#define DEQUANT_SHL()                                                \
+{                                                                    \
+    dctv = vec_ld(8*y, dct);                                         \
+    mf1v = vec_ld(16*y, dequant_mf[i_mf]);                           \
+    mf2v = vec_ld(16+16*y, dequant_mf[i_mf]);                        \
+    mfv  = vec_packs(mf1v, mf2v);                                    \
+                                                                     \
+    multEvenvA = vec_mule(dctv, mfv);                                \
+    multOddvA = vec_mulo(dctv, mfv);                                 \
+    dctv = (vec_s16_t) vec_packs(vec_mergeh(multEvenvA, multOddvA),  \
+                                 vec_mergel(multEvenvA, multOddvA)); \
+    dctv = vec_sl(dctv, i_qbitsv);                                   \
+    vec_st(dctv, 8*y, dct);                                          \
+}
+
+#define DEQUANT_SHR()                                          \
+{                                                              \
+    dctv = vec_ld(8*y, dct);                                   \
+    dct1v = vec_mergeh(dctv, dctv);                            \
+    dct2v = vec_mergel(dctv, dctv);                            \
+    mf1v = vec_ld(16*y, dequant_mf[i_mf]);                     \
+    mf2v = vec_ld(16+16*y, dequant_mf[i_mf]);                  \
+                                                               \
+    multEvenvA = vec_mule(dct1v, (vec_s16_t)mf1v);             \
+    multOddvA = vec_mulo(dct1v, (vec_s16_t)mf1v);              \
+    temp1v = vec_add(vec_sl(multEvenvA, sixteenv), multOddvA); \
+    temp1v = vec_add(temp1v, fv);                              \
+    temp1v = vec_sra(temp1v, i_qbitsv);                        \
+                                                               \
+    multEvenvA = vec_mule(dct2v, (vec_s16_t)mf2v);             \
+    multOddvA = vec_mulo(dct2v, (vec_s16_t)mf2v);              \
+    temp2v = vec_add(vec_sl(multEvenvA, sixteenv), multOddvA); \
+    temp2v = vec_add(temp2v, fv);                              \
+    temp2v = vec_sra(temp2v, i_qbitsv);                        \
+                                                               \
+    dctv = (vec_s16_t)vec_packs(temp1v, temp2v);               \
+    vec_st(dctv, y*8, dct);                                    \
+}
+
+void x264_dequant_4x4_altivec( int16_t dct[16], int dequant_mf[6][16], int i_qp )
+{
+    int i_mf = i_qp%6;
+    int i_qbits = i_qp/6 - 4;
+
+    vec_s16_t dctv;
+    vec_s16_t dct1v, dct2v;
+    vec_s32_t mf1v, mf2v;
+    vec_s16_t mfv;
+    vec_s32_t multEvenvA, multOddvA;
+    vec_s32_t temp1v, temp2v;
+
+    if( i_qbits >= 0 )
+    {
+        vec_u16_t i_qbitsv;
+        vec_u16_u qbits_u;
+        qbits_u.s[0]=i_qbits;
+        i_qbitsv = vec_splat(qbits_u.v, 0);
+
+        for( int y = 0; y < 4; y+=2 )
+            DEQUANT_SHL();
+    }
+    else
+    {
+        const int f = 1 << (-i_qbits-1);
+
+        vec_s32_t fv;
+        vec_u32_u f_u;
+        f_u.s[0]=f;
+        fv = (vec_s32_t)vec_splat(f_u.v, 0);
+
+        vec_u32_t i_qbitsv;
+        vec_u32_u qbits_u;
+        qbits_u.s[0]=-i_qbits;
+        i_qbitsv = vec_splat(qbits_u.v, 0);
+
+        vec_u32_t sixteenv;
+        vec_u32_u sixteen_u;
+        sixteen_u.s[0]=16;
+        sixteenv = vec_splat(sixteen_u.v, 0);
+
+        for( int y = 0; y < 4; y+=2 )
+            DEQUANT_SHR();
+    }
+}
+
+void x264_dequant_8x8_altivec( int16_t dct[64], int dequant_mf[6][64], int i_qp )
+{
+    int i_mf = i_qp%6;
+    int i_qbits = i_qp/6 - 6;
+
+    vec_s16_t dctv;
+    vec_s16_t dct1v, dct2v;
+    vec_s32_t mf1v, mf2v;
+    vec_s16_t mfv;
+    vec_s32_t multEvenvA, multOddvA;
+    vec_s32_t temp1v, temp2v;
+
+    if( i_qbits >= 0 )
+    {
+        vec_u16_t i_qbitsv;
+        vec_u16_u qbits_u;
+        qbits_u.s[0]=i_qbits;
+        i_qbitsv = vec_splat(qbits_u.v, 0);
+
+        for( int y = 0; y < 16; y+=2 )
+            DEQUANT_SHL();
+    }
+    else
+    {
+        const int f = 1 << (-i_qbits-1);
+
+        vec_s32_t fv;
+        vec_u32_u f_u;
+        f_u.s[0]=f;
+        fv = (vec_s32_t)vec_splat(f_u.v, 0);
+
+        vec_u32_t i_qbitsv;
+        vec_u32_u qbits_u;
+        qbits_u.s[0]=-i_qbits;
+        i_qbitsv = vec_splat(qbits_u.v, 0);
 
-    zerov = vec_splat_s16(0);
-    one = vec_splat_s16(1);
-    
-    int i;
+        vec_u32_t sixteenv;
+        vec_u32_u sixteen_u;
+        sixteen_u.s[0]=16;
+        sixteenv = vec_splat(sixteen_u.v, 0);
 
-    for ( i=0; i<4; i++ ) {
-      QUANT_16_U( i*2*16, i*2*16+16, i*4*16, i*4*16+16, i*4*16+32, i*4*16+48 );
+        for( int y = 0; y < 16; y+=2 )
+            DEQUANT_SHR();
     }
 }
+#endif // !HIGH_BIT_DEPTH