comparison libswscale/swscale_altivec_template.c @ 25750:06bf0eb70a55

Cosmetics: whitespaces

25749:8d082a234a6d

static inline void
altivec_packIntArrayToCharArray(int *val, uint8_t* dest, int dstW) {
    register int i;
    vector unsigned int altivec_vectorShiftInt19 =
        vec_add(vec_splat_u32(10),vec_splat_u32(9));
    if ((unsigned long)dest % 16) {
        /* badly aligned store, we force store alignment */
        /* and will handle load misalignment on val w/ vec_perm */
        vector unsigned char perm1;
        vector signed int v1;

            int offset = i << 2;
            vector signed int v2 = vec_ld(offset + 16, val);
            vector signed int v3 = vec_ld(offset + 32, val);
            vector signed int v4 = vec_ld(offset + 48, val);
            vector signed int v5 = vec_ld(offset + 64, val);
            vector signed int v12 = vec_perm(v1,v2,perm1);
            vector signed int v23 = vec_perm(v2,v3,perm1);
            vector signed int v34 = vec_perm(v3,v4,perm1);
            vector signed int v45 = vec_perm(v4,v5,perm1);

            vector signed int vA = vec_sra(v12, altivec_vectorShiftInt19);
            vector signed int vB = vec_sra(v23, altivec_vectorShiftInt19);
            vector signed int vC = vec_sra(v34, altivec_vectorShiftInt19);
            vector signed int vD = vec_sra(v45, altivec_vectorShiftInt19);

            }
            for ( ; i < dstW; i++) {
                val[i] += lumSrc[j][i] * lumFilter[j];
            }
        }
        altivec_packIntArrayToCharArray(val,dest,dstW);
    }
    if (uDest != 0) {
        int  __attribute__ ((aligned (16))) u[chrDstW];
        int  __attribute__ ((aligned (16))) v[chrDstW];

            for ( ; i < chrDstW; i++) {
                u[i] += chrSrc[j][i] * chrFilter[j];
                v[i] += chrSrc[j][i + 2048] * chrFilter[j];
            }
        }
        altivec_packIntArrayToCharArray(u,uDest,chrDstW);
        altivec_packIntArrayToCharArray(v,vDest,chrDstW);
    }
}

static inline void hScale_altivec_real(int16_t *dst, int dstW, uint8_t *src, int srcW, int xInc, int16_t *filter, int16_t *filterPos, int filterSize) {
    register int i;

        // the neat trick : we only care for half the elements,
        // high or low depending on (i<<3)%16 (it's 0 or 8 here),
        // and we're going to use vec_mule, so we chose
        // carefully how to "unpack" the elements into the even slots
        if ((i << 3) % 16)
            filter_v = vec_mergel(filter_v,(vector signed short)vzero);
        else
            filter_v = vec_mergeh(filter_v,(vector signed short)vzero);

        val_vEven = vec_mule(src_v, filter_v);
        val_s = vec_sums(val_vEven, vzero);
        vec_st(val_s, 0, tempo);
        dst[i] = av_clip(tempo[3]>>7, 0, (1<<15)-1);

}

static inline int yv12toyuy2_unscaled_altivec(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
                                              int srcSliceH, uint8_t* dstParam[], int dstStride_a[]) {
    uint8_t *dst=dstParam[0] + dstStride_a[0]*srcSliceY;
    // yv12toyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
    uint8_t *ysrc = src[0];
    uint8_t *usrc = src[1];
    uint8_t *vsrc = src[2];
    const int width = c->srcW;
    const int height = srcSliceH;

    const vector unsigned char yperm = vec_lvsl(0, ysrc);
    const int vertLumPerChroma = 2;
    register unsigned int y;

    if (width&15) {
        yv12toyuy2(ysrc, usrc, vsrc, dst,c->srcW,srcSliceH, lumStride, chromStride, dstStride);
        return srcSliceH;
    }

    /* this code assume:

}

static inline int yv12touyvy_unscaled_altivec(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
                                              int srcSliceH, uint8_t* dstParam[], int dstStride_a[]) {
    uint8_t *dst=dstParam[0] + dstStride_a[0]*srcSliceY;
    // yv12toyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
    uint8_t *ysrc = src[0];
    uint8_t *usrc = src[1];
    uint8_t *vsrc = src[2];
    const int width = c->srcW;
    const int height = srcSliceH;

    const int vertLumPerChroma = 2;
    const vector unsigned char yperm = vec_lvsl(0, ysrc);
    register unsigned int y;

    if (width&15) {
        yv12touyvy(ysrc, usrc, vsrc, dst,c->srcW,srcSliceH, lumStride, chromStride, dstStride);
        return srcSliceH;
    }

    /* this code assume:

25750:06bf0eb70a55

static inline void
altivec_packIntArrayToCharArray(int *val, uint8_t* dest, int dstW) {
    register int i;
    vector unsigned int altivec_vectorShiftInt19 =
        vec_add(vec_splat_u32(10), vec_splat_u32(9));
    if ((unsigned long)dest % 16) {
        /* badly aligned store, we force store alignment */
        /* and will handle load misalignment on val w/ vec_perm */
        vector unsigned char perm1;
        vector signed int v1;

            int offset = i << 2;
            vector signed int v2 = vec_ld(offset + 16, val);
            vector signed int v3 = vec_ld(offset + 32, val);
            vector signed int v4 = vec_ld(offset + 48, val);
            vector signed int v5 = vec_ld(offset + 64, val);
            vector signed int v12 = vec_perm(v1, v2, perm1);
            vector signed int v23 = vec_perm(v2, v3, perm1);
            vector signed int v34 = vec_perm(v3, v4, perm1);
            vector signed int v45 = vec_perm(v4, v5, perm1);

            vector signed int vA = vec_sra(v12, altivec_vectorShiftInt19);
            vector signed int vB = vec_sra(v23, altivec_vectorShiftInt19);
            vector signed int vC = vec_sra(v34, altivec_vectorShiftInt19);
            vector signed int vD = vec_sra(v45, altivec_vectorShiftInt19);

            }
            for ( ; i < dstW; i++) {
                val[i] += lumSrc[j][i] * lumFilter[j];
            }
        }
        altivec_packIntArrayToCharArray(val, dest, dstW);
    }
    if (uDest != 0) {
        int  __attribute__ ((aligned (16))) u[chrDstW];
        int  __attribute__ ((aligned (16))) v[chrDstW];

            for ( ; i < chrDstW; i++) {
                u[i] += chrSrc[j][i] * chrFilter[j];
                v[i] += chrSrc[j][i + 2048] * chrFilter[j];
            }
        }
        altivec_packIntArrayToCharArray(u, uDest, chrDstW);
        altivec_packIntArrayToCharArray(v, vDest, chrDstW);
    }
}

static inline void hScale_altivec_real(int16_t *dst, int dstW, uint8_t *src, int srcW, int xInc, int16_t *filter, int16_t *filterPos, int filterSize) {
    register int i;

        // the neat trick : we only care for half the elements,
        // high or low depending on (i<<3)%16 (it's 0 or 8 here),
        // and we're going to use vec_mule, so we chose
        // carefully how to "unpack" the elements into the even slots
        if ((i << 3) % 16)
            filter_v = vec_mergel(filter_v, (vector signed short)vzero);
        else
            filter_v = vec_mergeh(filter_v, (vector signed short)vzero);

        val_vEven = vec_mule(src_v, filter_v);
        val_s = vec_sums(val_vEven, vzero);
        vec_st(val_s, 0, tempo);
        dst[i] = av_clip(tempo[3]>>7, 0, (1<<15)-1);

}

static inline int yv12toyuy2_unscaled_altivec(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
                                              int srcSliceH, uint8_t* dstParam[], int dstStride_a[]) {
    uint8_t *dst=dstParam[0] + dstStride_a[0]*srcSliceY;
    // yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
    uint8_t *ysrc = src[0];
    uint8_t *usrc = src[1];
    uint8_t *vsrc = src[2];
    const int width = c->srcW;
    const int height = srcSliceH;

    const vector unsigned char yperm = vec_lvsl(0, ysrc);
    const int vertLumPerChroma = 2;
    register unsigned int y;

    if (width&15) {
        yv12toyuy2(ysrc, usrc, vsrc, dst, c->srcW, srcSliceH, lumStride, chromStride, dstStride);
        return srcSliceH;
    }

    /* this code assume:

}

static inline int yv12touyvy_unscaled_altivec(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
                                              int srcSliceH, uint8_t* dstParam[], int dstStride_a[]) {
    uint8_t *dst=dstParam[0] + dstStride_a[0]*srcSliceY;
    // yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
    uint8_t *ysrc = src[0];
    uint8_t *usrc = src[1];
    uint8_t *vsrc = src[2];
    const int width = c->srcW;
    const int height = srcSliceH;

    const int vertLumPerChroma = 2;
    const vector unsigned char yperm = vec_lvsl(0, ysrc);
    register unsigned int y;

    if (width&15) {
        yv12touyvy(ysrc, usrc, vsrc, dst, c->srcW, srcSliceH, lumStride, chromStride, dstStride);
        return srcSliceH;
    }

    /* this code assume: