libpostproc/postprocess_altivec_template.c File Reference

#include "libavutil/avutil.h"

Go to the source code of this file.

Defines

#define ALTIVEC_TRANSPOSE_8x8_SHORT(src_a, src_b, src_c, src_d, src_e, src_f, src_g, src_h)
#define LOAD_LINE(i)
#define LOAD_LINE_ALIGNED(i)
#define ITER(i, j)
#define LOAD_LINE(i)
#define LOAD_LINE_ALIGNED(i)
#define COMPUTE_VR(i, j, k)
#define PACK_AND_STORE(i)
#define PACK_AND_STORE_ALIGNED(i)
#define LOAD_LINE(i)
#define STORE(i)
#define LOAD_LINE(i)
#define EXTRACT(op)
#define COMPARE(i)
#define F_INIT()
#define F2(i, j, k, l)
#define ITER(i, j, k)
#define STORE_LINE(i)
#define doHorizLowPass_altivec(a...)   doHorizLowPass_C(a)
#define doHorizDefFilter_altivec(a...)   doHorizDefFilter_C(a)
#define do_a_deblock_altivec(a...)   do_a_deblock_C(a)
#define LOAD_LINE(src, i)
#define ACCUMULATE_DIFFS(i)
#define OP(i)   v_tempBlurredAss##i = vec_avg(v_tempBlurredAss##i, v_srcAss##i);
#define OP(i)   v_tempBlurredAss##i = v_srcAss##i;
#define OP(i)
#define OP(i)
#define PACK_AND_STORE(src, i)
#define LOAD_DOUBLE_LINE(i, j)
#define LOAD_DOUBLE_LINE(i, j)
#define STORE_DOUBLE_LINE(i, j)

Functions

static int vertClassify_altivec (uint8_t src[], int stride, PPContext *c)
static void doVertLowPass_altivec (uint8_t *src, int stride, PPContext *c)
static void doVertDefFilter_altivec (uint8_t src[], int stride, PPContext *c)
static void dering_altivec (uint8_t src[], int stride, PPContext *c)
static void tempNoiseReducer_TMPL (uint8_t *src, int stride, uint8_t *tempBlurred, uint32_t *tempBlurredPast, int *maxNoise)
static void transpose_16x8_char_toPackedAlign_altivec (unsigned char *dst, unsigned char *src, int stride)
static void transpose_8x16_char_fromPackedAlign_altivec (unsigned char *dst, unsigned char *src, int stride)

Define Documentation

#define ACCUMULATE_DIFFS ( i   ) 

Value:

do {                                \
        vector signed short v_d = vec_sub(v_tempBlurredAss##i,  \
                                          v_srcAss##i);         \
        v_dp = vec_msums(v_d, v_d, v_dp);                       \
        v_sysdp = vec_msums(v_d, vsint16_1, v_sysdp);           \
    } while (0)

Referenced by tempNoiseReducer_TMPL().

#define ALTIVEC_TRANSPOSE_8x8_SHORT ( src_a,
src_b,
src_c,
src_d,
src_e,
src_f,
src_g,
src_h   ) 

Value:

do {                                                          \
        __typeof__(src_a) tempA1, tempB1, tempC1, tempD1;         \
        __typeof__(src_a) tempE1, tempF1, tempG1, tempH1;         \
        __typeof__(src_a) tempA2, tempB2, tempC2, tempD2;         \
        __typeof__(src_a) tempE2, tempF2, tempG2, tempH2;         \
        tempA1 = vec_mergeh (src_a, src_e);                       \
        tempB1 = vec_mergel (src_a, src_e);                       \
        tempC1 = vec_mergeh (src_b, src_f);                       \
        tempD1 = vec_mergel (src_b, src_f);                       \
        tempE1 = vec_mergeh (src_c, src_g);                       \
        tempF1 = vec_mergel (src_c, src_g);                       \
        tempG1 = vec_mergeh (src_d, src_h);                       \
        tempH1 = vec_mergel (src_d, src_h);                       \
        tempA2 = vec_mergeh (tempA1, tempE1);                     \
        tempB2 = vec_mergel (tempA1, tempE1);                     \
        tempC2 = vec_mergeh (tempB1, tempF1);                     \
        tempD2 = vec_mergel (tempB1, tempF1);                     \
        tempE2 = vec_mergeh (tempC1, tempG1);                     \
        tempF2 = vec_mergel (tempC1, tempG1);                     \
        tempG2 = vec_mergeh (tempD1, tempH1);                     \
        tempH2 = vec_mergel (tempD1, tempH1);                     \
        src_a = vec_mergeh (tempA2, tempE2);                      \
        src_b = vec_mergel (tempA2, tempE2);                      \
        src_c = vec_mergeh (tempB2, tempF2);                      \
        src_d = vec_mergel (tempB2, tempF2);                      \
        src_e = vec_mergeh (tempC2, tempG2);                      \
        src_f = vec_mergel (tempC2, tempG2);                      \
        src_g = vec_mergeh (tempD2, tempH2);                      \
        src_h = vec_mergel (tempD2, tempH2);                      \
    } while (0)

Definition at line 25 of file postprocess_altivec_template.c.

#define COMPARE ( i   ) 

Value:

do {                                                                \
        const vector unsigned char cmp =                                \
            (vector unsigned char)vec_cmpgt(src##i, v_avg);             \
        const vector unsigned short cmpHi =                             \
            (vector unsigned short)vec_mergeh(cmp, cmp);                \
        const vector unsigned short cmpLi =                             \
            (vector unsigned short)vec_mergel(cmp, cmp);                \
        const vector signed short cmpHf =                               \
            (vector signed short)vec_and(cmpHi, mask1);                 \
        const vector signed short cmpLf =                               \
            (vector signed short)vec_and(cmpLi, mask2);                 \
        const vector signed int sump = vec_sum4s(cmpHf, zero);          \
        const vector signed int sumq = vec_sum4s(cmpLf, sump);          \
        sum##i  = vec_sums(sumq, zero);                                 \
    } while (0)

Referenced by dering_altivec().

#define COMPUTE_VR ( i,
j,
k   ) 

Value:

const vector signed short temps1##i =                               \
            vec_add(v_sumsB##i, v_sumsB##k);                                \
        const vector signed short temps2##i =                               \
            vec_mladd(vb##j, (vector signed short)v_2, temps1##i);          \
        const vector signed short  vr##j = vec_sra(temps2##i, v_4)

Referenced by doVertLowPass_altivec().

#define do_a_deblock_altivec ( a...   )     do_a_deblock_C(a)

Definition at line 826 of file postprocess_altivec_template.c.

#define doHorizDefFilter_altivec ( a...   )     doHorizDefFilter_C(a)

Definition at line 825 of file postprocess_altivec_template.c.

#define doHorizLowPass_altivec ( a...   )     doHorizLowPass_C(a)

Definition at line 824 of file postprocess_altivec_template.c.

#define EXTRACT ( op   ) 

Value:

do {                                                \
    const vector unsigned char s_1   = vec_##op(trunc_src12, trunc_src34); \
    const vector unsigned char s_2   = vec_##op(trunc_src56, trunc_src78); \
    const vector unsigned char s_6   = vec_##op(s_1, s_2);     \
    const vector unsigned char s_8h  = vec_mergeh(s_6, s_6);   \
    const vector unsigned char s_8l  = vec_mergel(s_6, s_6);   \
    const vector unsigned char s_9   = vec_##op(s_8h, s_8l);   \
    const vector unsigned char s_9h  = vec_mergeh(s_9, s_9);   \
    const vector unsigned char s_9l  = vec_mergel(s_9, s_9);   \
    const vector unsigned char s_10  = vec_##op(s_9h, s_9l);   \
    const vector unsigned char s_10h = vec_mergeh(s_10, s_10); \
    const vector unsigned char s_10l = vec_mergel(s_10, s_10); \
    const vector unsigned char s_11  = vec_##op(s_10h, s_10l); \
    const vector unsigned char s_11h = vec_mergeh(s_11, s_11); \
    const vector unsigned char s_11l = vec_mergel(s_11, s_11); \
    v_##op = vec_##op(s_11h, s_11l);                           \
} while (0)

Referenced by dering_altivec().

#define F2 ( i,
j,
k,
l   ) 

Value:

if (S[i] & (1 << (l+1))) {                                          \
        const vector unsigned char a_A = vec_perm(src##i, src##j, permA1M); \
        const vector unsigned char a_B = vec_perm(a_A, src##k, permA2M); \
        const vector signed int a_sump =                                \
            (vector signed int)vec_msum(a_B, magic, (vector unsigned int)zero);\
        vector signed int F = vec_sr(vec_sums(a_sump, vsint32_8), vuint32_4); \
        const vector signed int p =                                     \
            (vector signed int)vec_perm(src##j, (vector unsigned char)zero, \
                                        extractPermM);                  \
        const vector signed int sum  = vec_add(p, vQP2);                \
        const vector signed int diff = vec_sub(p, vQP2);                \
        vector signed int newpm;                                        \
        vector unsigned char newpm2, mask;                              \
        F = vec_splat(F, 3);                                            \
        if (vec_all_lt(sum, F))                                         \
            newpm = sum;                                                \
        else if (vec_all_gt(diff, F))                                   \
            newpm = diff;                                               \
        else newpm = F;                                                 \
        newpm2 = vec_splat((vector unsigned char)newpm, 15);            \
        mask = vec_add(identity, tenRightM);                            \
        src##j = vec_perm(src##j, newpm2, mask);                        \
    }                                                                   \
    permA1M = vec_add(permA1M, permA1inc);                              \
    permA2M = vec_add(permA2M, permA2inc);                              \
    tenRightM = vec_sro(tenRightM, eightLeft);                          \
    extractPermM = vec_add(extractPermM, extractPermInc)

 
#define F_INIT (  ) 

Value:

vector unsigned char tenRightM = tenRight;         \
    vector unsigned char permA1M = permA1;             \
    vector unsigned char permA2M = permA2;             \
    vector unsigned char extractPermM = extractPerm

#define ITER ( i,
j,
k   ) 

Value:

do {                      \
    F_INIT();                                   \
    F2(i, j, k, 0);                             \
    F2(i, j, k, 1);                             \
    F2(i, j, k, 2);                             \
    F2(i, j, k, 3);                             \
    F2(i, j, k, 4);                             \
    F2(i, j, k, 5);                             \
    F2(i, j, k, 6);                             \
    F2(i, j, k, 7);                             \
} while (0)

#define ITER ( i,
j   ) 

Value:

const vector signed short v_diff##i =                               \
        vec_sub(v_srcAss##i, v_srcAss##j);                              \
    const vector signed short v_sum##i =                                \
        vec_add(v_diff##i, v_dcOffset);                                 \
    const vector signed short v_comp##i =                               \
        (vector signed short)vec_cmplt((vector unsigned short)v_sum##i, \
                                       v_dcThreshold);                  \
    const vector signed short v_part##i = vec_and(mask, v_comp##i);

Referenced by dering_altivec(), and vertClassify_altivec().

#define LOAD_DOUBLE_LINE ( i,
j   ) 

Value:

vector unsigned char src##i = vec_ld(i * 16, src);            \
    vector unsigned char src##j = vec_ld(j * 16, src)

#define LOAD_DOUBLE_LINE ( i,
j   ) 

Value:

vector unsigned char perm1##i = vec_lvsl(i * stride, src);          \
    vector unsigned char perm2##i = vec_lvsl(j * stride, src);          \
    vector unsigned char srcA##i = vec_ld(i * stride, src);             \
    vector unsigned char srcB##i = vec_ld(i * stride + 16, src);        \
    vector unsigned char srcC##i = vec_ld(j * stride, src);             \
    vector unsigned char srcD##i = vec_ld(j * stride+ 16, src);         \
    vector unsigned char src##i = vec_perm(srcA##i, srcB##i, perm1##i); \
    vector unsigned char src##j = vec_perm(srcC##i, srcD##i, perm2##i)

Referenced by transpose_16x8_char_toPackedAlign_altivec(), and transpose_8x16_char_fromPackedAlign_altivec().

#define LOAD_LINE ( src,
i   ) 

Value:

register int j##src##i = i * stride;                                \
    vector unsigned char perm##src##i = vec_lvsl(j##src##i, src);       \
    const vector unsigned char v_##src##A1##i = vec_ld(j##src##i, src); \
    const vector unsigned char v_##src##A2##i = vec_ld(j##src##i + 16, src); \
    const vector unsigned char v_##src##A##i =                          \
        vec_perm(v_##src##A1##i, v_##src##A2##i, perm##src##i);         \
    vector signed short v_##src##Ass##i =                               \
        (vector signed short)vec_mergeh((vector signed char)zero,       \
                                        (vector signed char)v_##src##A##i)

#define LOAD_LINE ( i   ) 

Value:

const vector unsigned char perm##i =                              \
        vec_lvsl(i * stride, srcCopy);                                \
    vector unsigned char sA##i = vec_ld(i * stride, srcCopy);         \
    vector unsigned char sB##i = vec_ld(i * stride + 16, srcCopy);    \
    vector unsigned char src##i = vec_perm(sA##i, sB##i, perm##i)

#define LOAD_LINE ( i   ) 

Value:

const vector unsigned char perm##i =                                \
        vec_lvsl(i * stride, src2);                                     \
    const vector unsigned char vbA##i =                                 \
        vec_ld(i * stride, src2);                                       \
    const vector unsigned char vbB##i =                                 \
        vec_ld(i * stride + 16, src2);                                  \
    const vector unsigned char vbT##i =                                 \
        vec_perm(vbA##i, vbB##i, perm##i);                              \
    const vector signed short vb##i =                                   \
        (vector signed short)vec_mergeh((vector unsigned char)zero,     \
                                        (vector unsigned char)vbT##i)

#define LOAD_LINE ( i   ) 

Value:

perml##i = vec_lvsl(i * stride, src2);                              \
    vbA##i = vec_ld(i * stride, src2);                                  \
    vbB##i = vec_ld(i * stride + 16, src2);                             \
    vbT##i = vec_perm(vbA##i, vbB##i, perml##i);                        \
    vb##i =                                                             \
        (vector signed short)vec_mergeh((vector unsigned char)zero,     \
                                        (vector unsigned char)vbT##i)

#define LOAD_LINE ( i   ) 

Value:

{                                                                   \
    vector unsigned char perm##i = vec_lvsl(j##i, src2);                \
    vector unsigned char v_srcA2##i;                                    \
    vector unsigned char v_srcA1##i = vec_ld(j##i, src2);               \
    if (two_vectors)                                                    \
        v_srcA2##i = vec_ld(j##i + 16, src2);                           \
    v_srcA##i =                                                         \
        vec_perm(v_srcA1##i, v_srcA2##i, perm##i);                      \
    v_srcAss##i =                                                       \
        (vector signed short)vec_mergeh((vector signed char)zero,       \
                                        (vector signed char)v_srcA##i); }

Referenced by dering_altivec(), doVertDefFilter_altivec(), doVertLowPass_altivec(), tempNoiseReducer_TMPL(), and vertClassify_altivec().

#define LOAD_LINE_ALIGNED ( i   ) 

Value:

vbT##i = vec_ld(j##i, src2);                                        \
    vb##i =                                                             \
        (vector signed short)vec_mergeh((vector signed char)zero,       \
                                        (vector signed char)vbT##i)

#define LOAD_LINE_ALIGNED ( i   ) 

Value:

v_srcA##i = vec_ld(j##i, src2);                                     \
    v_srcAss##i =                                                       \
        (vector signed short)vec_mergeh((vector signed char)zero,       \
                                        (vector signed char)v_srcA##i)

Referenced by doVertLowPass_altivec(), and vertClassify_altivec().

#define OP ( i   ) 

Value:

do {                                              \
            const vector signed short v_temp =                  \
                vec_mladd(v_tempBlurredAss##i, vsint16_3, v_srcAss##i); \
            const vector signed short v_temp2 = vec_add(v_temp, vsint16_2); \
            v_tempBlurredAss##i =                                       \
                vec_sr(v_temp2, (vector unsigned short)vsint16_2);      \
        } while (0)

#define OP ( i   ) 

Value:

do {                                                      \
            const vector signed short v_temp =                          \
                vec_mladd(v_tempBlurredAss##i, vsint16_7, v_srcAss##i); \
            const vector signed short v_temp2 = vec_add(v_temp, vsint16_4); \
            v_tempBlurredAss##i = vec_sr(v_temp2, vuint16_3);           \
        } while (0)

#define OP ( i   )     v_tempBlurredAss##i = v_srcAss##i;

#define OP ( i   )     v_tempBlurredAss##i = vec_avg(v_tempBlurredAss##i, v_srcAss##i);

#define PACK_AND_STORE ( src,
i   ) 

Value:

do {                                      \
    const vector unsigned char perms = vec_lvsr(i * stride, src);        \
    const vector unsigned char vf =                                      \
        vec_packsu(v_tempBlurredAss##1, (vector signed short)zero);     \
    const vector unsigned char vg = vec_perm(vf, v_##src##A##i, permHH); \
    const vector unsigned char mask =                                    \
        vec_perm((vector unsigned char)zero, (vector unsigned char)neg1, perms); \
    const vector unsigned char vg2 = vec_perm(vg, vg, perms);            \
    const vector unsigned char svA = vec_sel(v_##src##A1##i, vg2, mask); \
    const vector unsigned char svB = vec_sel(vg2, v_##src##A2##i, mask); \
    vec_st(svA, i * stride, src);                                        \
    vec_st(svB, i * stride + 16, src);                                   \
} while (0)

#define PACK_AND_STORE ( i   ) 

Value:

{   const vector unsigned char perms##i =                       \
        vec_lvsr(i * stride, src2);                             \
    const vector unsigned char vf##i =                          \
        vec_packsu(vr##i, (vector signed short)zero);           \
    const vector unsigned char vg##i =                          \
        vec_perm(vf##i, vbT##i, permHH);                        \
    const vector unsigned char mask##i =                        \
        vec_perm((vector unsigned char)zero, (vector unsigned char)neg1, perms##i); \
    const vector unsigned char vg2##i =                         \
        vec_perm(vg##i, vg##i, perms##i);                       \
    const vector unsigned char svA##i =                         \
        vec_sel(vbA##i, vg2##i, mask##i);                       \
    const vector unsigned char svB##i =                         \
        vec_sel(vg2##i, vbB##i, mask##i);                       \
    vec_st(svA##i, i * stride, src2);                           \
    vec_st(svB##i, i * stride + 16, src2);}

Referenced by doVertLowPass_altivec(), and tempNoiseReducer_TMPL().

#define PACK_AND_STORE_ALIGNED ( i   ) 

Value:

{   const vector unsigned char vf##i =                          \
        vec_packsu(vr##i, (vector signed short)zero);           \
    const vector unsigned char vg##i =                          \
        vec_perm(vf##i, vbT##i, permHH);                        \
    vec_st(vg##i, i * stride, src2);}

Referenced by doVertLowPass_altivec().

#define STORE ( i   ) 

Value:

{    const vector unsigned char perms##i =                      \
         vec_lvsr(i * stride, src2);                            \
     const vector unsigned char vg##i =                         \
         vec_perm(st##i, vbT##i, permHH);                       \
     const vector unsigned char mask##i =                       \
         vec_perm((vector unsigned char)zero, (vector unsigned char)neg1, perms##i); \
     const vector unsigned char vg2##i =                        \
         vec_perm(vg##i, vg##i, perms##i);                      \
     const vector unsigned char svA##i =                        \
         vec_sel(vbA##i, vg2##i, mask##i);                      \
     const vector unsigned char svB##i =                        \
         vec_sel(vg2##i, vbB##i, mask##i);                      \
     vec_st(svA##i, i * stride, src2);                          \
     vec_st(svB##i, i * stride + 16, src2);}

#define STORE_DOUBLE_LINE ( i,
j   ) 

Value:

do {                                    \
    vector unsigned char dstAi = vec_ld(i * stride, dst);               \
    vector unsigned char dstBi = vec_ld(i * stride + 16, dst);          \
    vector unsigned char dstAj = vec_ld(j * stride, dst);               \
    vector unsigned char dstBj = vec_ld(j * stride+ 16, dst);           \
    vector unsigned char aligni = vec_lvsr(i * stride, dst);            \
    vector unsigned char alignj = vec_lvsr(j * stride, dst);            \
    vector unsigned char maski =                                        \
        vec_perm(zero, (vector unsigned char)neg1, aligni);             \
    vector unsigned char maskj =                                        \
        vec_perm(zero, (vector unsigned char)neg1, alignj);             \
    vector unsigned char dstRi = vec_perm(temp##i, temp##i, aligni);    \
    vector unsigned char dstRj = vec_perm(temp##j, temp##j, alignj);    \
    vector unsigned char dstAFi = vec_sel(dstAi, dstRi, maski);         \
    vector unsigned char dstBFi = vec_sel(dstRi, dstBi, maski);         \
    vector unsigned char dstAFj = vec_sel(dstAj, dstRj, maskj);         \
    vector unsigned char dstBFj = vec_sel(dstRj, dstBj, maskj);         \
    vec_st(dstAFi, i * stride, dst);                                    \
    vec_st(dstBFi, i * stride + 16, dst);                               \
    vec_st(dstAFj, j * stride, dst);                                    \
    vec_st(dstBFj, j * stride + 16, dst);                               \
} while (0)

Referenced by transpose_8x16_char_fromPackedAlign_altivec().

#define STORE_LINE ( i   ) 

Value:

do {                              \
    const vector unsigned char permST =                 \
        vec_lvsr(i * stride, srcCopy);                  \
    const vector unsigned char maskST =                 \
        vec_perm((vector unsigned char)zero,            \
                 (vector unsigned char)neg1, permST);   \
    src##i = vec_perm(src##i ,src##i, permST);          \
    sA##i= vec_sel(sA##i, src##i, maskST);              \
    sB##i= vec_sel(src##i, sB##i, maskST);              \
    vec_st(sA##i, i * stride, srcCopy);                 \
    vec_st(sB##i, i * stride + 16, srcCopy);            \
} while (0)

Referenced by dering_altivec().

Function Documentation

static void dering_altivec ( uint8_t  src[],
int  stride,
PPContext *  c 
) [inline, static]

Definition at line 532 of file postprocess_altivec_template.c.

static void doVertDefFilter_altivec ( uint8_t  src[],
int  stride,
PPContext *  c 
) [inline, static]

Definition at line 411 of file postprocess_altivec_template.c.

Referenced by postProcess_TMPL().

static void doVertLowPass_altivec ( uint8_t *  src,
int  stride,
PPContext *  c 
) [inline, static]

Definition at line 213 of file postprocess_altivec_template.c.

Referenced by postProcess_TMPL().

static void tempNoiseReducer_TMPL ( uint8_t *  src,
int  stride,
uint8_t *  tempBlurred,
uint32_t *  tempBlurredPast,
int *  maxNoise 
) [inline, static]

Definition at line 828 of file postprocess_altivec_template.c.

static void transpose_16x8_char_toPackedAlign_altivec ( unsigned char *  dst,
unsigned char *  src,
int  stride 
) [inline, static]

Definition at line 1015 of file postprocess_altivec_template.c.

Referenced by postProcess_TMPL().

static void transpose_8x16_char_fromPackedAlign_altivec ( unsigned char *  dst,
unsigned char *  src,
int  stride 
) [inline, static]

Definition at line 1120 of file postprocess_altivec_template.c.

Referenced by postProcess_TMPL().

static int vertClassify_altivec ( uint8_t  src[],
int  stride,
PPContext *  c 
) [inline, static]

Definition at line 58 of file postprocess_altivec_template.c.

Referenced by postProcess_TMPL().

Generated on Fri Oct 26 02:48:03 2012 for FFmpeg by  doxygen 1.5.8