static av_always_inline void color_correlation(uint8_t *dst, int dst_linesize,
float **src, int src_linesize,
int w, int h,
int r, int g, int b)
{
int x, y;
const float *src_r = src[0];
const float *src_g = src[1];
const float *src_b = src[2];
for (y = 0; y < h; y++) {
uint8_t *dstp = dst;
for (x = 0; x < w; x++) {
dstp[r] = av_clip_uint8(src_r[x] * DCT3X3_0_0 + src_g[x] * DCT3X3_1_0 + src_b[x] * DCT3X3_2_0);
dstp[g] = av_clip_uint8(src_r[x] * DCT3X3_0_1 + src_b[x] * DCT3X3_2_1);
dstp[b] = av_clip_uint8(src_r[x] * DCT3X3_0_2 + src_g[x] * DCT3X3_1_2 + src_b[x] * DCT3X3_2_2);
dstp += 3;
}
dst += dst_linesize;
src_r += src_linesize;
src_g += src_linesize;
src_b += src_linesize;
}
}
static void render_line(int x0, int y0, int x1, int y1, float *buf)
{
int dy = y1 - y0;
int adx = x1 - x0;
int ady = FFABS(dy);
int sy = dy < 0 ? -1 : 1;
buf[x0] = ff_vorbis_floor1_inverse_db_table[av_clip_uint8(y0)];
if (ady*2 <= adx) { // optimized common case
render_line_unrolled(x0, y0, x1, sy, ady, adx, buf);
} else {
int base = dy / adx;
int x = x0;
int y = y0;
int err = -adx;
ady -= FFABS(base) * adx;
while (++x < x1) {
y += base;
err += ady;
if (err >= 0) {
err -= adx;
y += sy;
}
buf[x] = ff_vorbis_floor1_inverse_db_table[av_clip_uint8(y)];
}
}
}
/** Function used to do motion compensation with bicubic interpolation
*/
static void vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride, int hmode, int vmode, int rnd)
{
int i, j;
uint8_t tmp[8*11], *tptr;
int r;
r = rnd;
src -= stride;
tptr = tmp;
for(j = 0; j < 11; j++) {
for(i = 0; i < 8; i++)
tptr[i] = av_clip_uint8(vc1_mspel_filter(src + i, 1, hmode, r));
src += stride;
tptr += 8;
}
r = 1 - rnd;
tptr = tmp + 8;
for(j = 0; j < 8; j++) {
for(i = 0; i < 8; i++)
dst[i] = av_clip_uint8(vc1_mspel_filter(tptr + i, 8, vmode, r));
dst += stride;
tptr += 8;
}
}
/**
* VC-1 in-loop deblocking filter for one line
* @param src source block type
* @param stride block stride
* @param pq block quantizer
* @return whether other 3 pairs should be filtered or not
* @see 8.6
*/
static av_always_inline int vc1_filter_line(uint8_t* src, int stride, int pq) {
int a0 = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3;
int a0_sign = a0 >> 31; /* Store sign */
a0 = (a0 ^ a0_sign) - a0_sign; /* a0 = FFABS(a0); */
if(a0 < pq) {
int a1 = FFABS((2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3);
int a2 = FFABS((2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3);
if(a1 < a0 || a2 < a0) {
int clip = src[-1*stride] - src[ 0*stride];
int clip_sign = clip >> 31;
clip = ((clip ^ clip_sign) - clip_sign)>>1;
if(clip) {
int a3 = FFMIN(a1, a2);
int d = 5 * (a3 - a0);
int d_sign = (d >> 31);
d = ((d ^ d_sign) - d_sign) >> 3;
d_sign ^= a0_sign;
if( d_sign ^ clip_sign )
d = 0;
else {
d = FFMIN(d, clip);
d = (d ^ d_sign) - d_sign; /* Restore sign */
src[-1*stride] = av_clip_uint8(src[-1*stride] - d);
src[ 0*stride] = av_clip_uint8(src[ 0*stride] + d);
}
return 1;
}
}
int av_audio_convert(AVAudioConvert *ctx,
void * const out[6], const int out_stride[6],
const void * const in[6], const int in_stride[6], int len)
{
int ch;
//FIXME optimize common cases
for(ch=0; ch<ctx->out_channels; ch++){
const int is= in_stride[ch];
const int os= out_stride[ch];
const uint8_t *pi= in[ch];
uint8_t *po= out[ch];
uint8_t *end= po + os*len;
if(!out[ch])
continue;
#define CONV(ofmt, otype, ifmt, expr)\
if(ctx->fmt_pair == ofmt + AV_SAMPLE_FMT_NB*ifmt){\
do{\
*(otype*)po = expr; pi += is; po += os;\
}while(po < end);\
}
//FIXME put things below under ifdefs so we do not waste space for cases no codec will need
//FIXME rounding ?
CONV(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_U8 , *(const uint8_t*)pi)
else CONV(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)<<8)
else CONV(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)<<24)
else CONV(AV_SAMPLE_FMT_FLT, float , AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)*(1.0 / (1<<7)))
else CONV(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)*(1.0 / (1<<7)))
else CONV(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_S16, (*(const int16_t*)pi>>8) + 0x80)
else CONV(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_S16, *(const int16_t*)pi)
else CONV(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_S16, *(const int16_t*)pi<<16)
else CONV(AV_SAMPLE_FMT_FLT, float , AV_SAMPLE_FMT_S16, *(const int16_t*)pi*(1.0 / (1<<15)))
else CONV(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_S16, *(const int16_t*)pi*(1.0 / (1<<15)))
else CONV(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_S32, (*(const int32_t*)pi>>24) + 0x80)
else CONV(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_S32, *(const int32_t*)pi>>16)
else CONV(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_S32, *(const int32_t*)pi)
else CONV(AV_SAMPLE_FMT_FLT, float , AV_SAMPLE_FMT_S32, *(const int32_t*)pi*(1.0 / (1U<<31)))
else CONV(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_S32, *(const int32_t*)pi*(1.0 / (1U<<31)))
else CONV(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_FLT, av_clip_uint8( lrintf(*(const float*)pi * (1<<7)) + 0x80))
else CONV(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, av_clip_int16( lrintf(*(const float*)pi * (1<<15))))
else CONV(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, av_clipl_int32(llrintf(*(const float*)pi * (1U<<31))))
else CONV(AV_SAMPLE_FMT_FLT, float , AV_SAMPLE_FMT_FLT, *(const float*)pi)
else CONV(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_FLT, *(const float*)pi)
else CONV(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_DBL, av_clip_uint8( lrint(*(const double*)pi * (1<<7)) + 0x80))
else CONV(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, av_clip_int16( lrint(*(const double*)pi * (1<<15))))
else CONV(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, av_clipl_int32(llrint(*(const double*)pi * (1U<<31))))
else CONV(AV_SAMPLE_FMT_FLT, float , AV_SAMPLE_FMT_DBL, *(const double*)pi)
else CONV(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_DBL, *(const double*)pi)
else return -1;
}
return 0;
}
static int vibrance_slice8(AVFilterContext *avctx, void *arg, int jobnr, int nb_jobs)
{
VibranceContext *s = avctx->priv;
AVFrame *frame = arg;
const int width = frame->width;
const int height = frame->height;
const float scale = 1.f / 255.f;
const float gc = s->lcoeffs[0];
const float bc = s->lcoeffs[1];
const float rc = s->lcoeffs[2];
const float intensity = s->intensity;
const float alternate = s->alternate ? 1.f : -1.f;
const float gintensity = intensity * s->balance[0];
const float bintensity = intensity * s->balance[1];
const float rintensity = intensity * s->balance[2];
const float sgintensity = alternate * FFSIGN(gintensity);
const float sbintensity = alternate * FFSIGN(bintensity);
const float srintensity = alternate * FFSIGN(rintensity);
const int slice_start = (height * jobnr) / nb_jobs;
const int slice_end = (height * (jobnr + 1)) / nb_jobs;
const int glinesize = frame->linesize[0];
const int blinesize = frame->linesize[1];
const int rlinesize = frame->linesize[2];
uint8_t *gptr = frame->data[0] + slice_start * glinesize;
uint8_t *bptr = frame->data[1] + slice_start * blinesize;
uint8_t *rptr = frame->data[2] + slice_start * rlinesize;
for (int y = slice_start; y < slice_end; y++) {
for (int x = 0; x < width; x++) {
float g = gptr[x] * scale;
float b = bptr[x] * scale;
float r = rptr[x] * scale;
float max_color = FFMAX3(r, g, b);
float min_color = FFMIN3(r, g, b);
float color_saturation = max_color - min_color;
float luma = g * gc + r * rc + b * bc;
const float cg = 1.f + gintensity * (1.f - sgintensity * color_saturation);
const float cb = 1.f + bintensity * (1.f - sbintensity * color_saturation);
const float cr = 1.f + rintensity * (1.f - srintensity * color_saturation);
g = lerpf(luma, g, cg);
b = lerpf(luma, b, cb);
r = lerpf(luma, r, cr);
gptr[x] = av_clip_uint8(g * 255.f);
bptr[x] = av_clip_uint8(b * 255.f);
rptr[x] = av_clip_uint8(r * 255.f);
}
gptr += glinesize;
bptr += blinesize;
rptr += rlinesize;
}
return 0;
}
static void vp56_edge_filter(VP56Context *s, uint8_t *yuv,
int pix_inc, int line_inc, int t)
{
int pix2_inc = 2 * pix_inc;
int i, v;
for (i=0; i<12; i++) {
v = (yuv[-pix2_inc] + 3*(yuv[0]-yuv[-pix_inc]) - yuv[pix_inc] + 4) >>3;
v = s->adjust(v, t);
yuv[-pix_inc] = av_clip_uint8(yuv[-pix_inc] + v);
yuv[0] = av_clip_uint8(yuv[0] - v);
yuv += line_inc;
}
}
static void put_signed_rect_clamped_8bit_c(uint8_t *dst, int dst_stride, const uint8_t *_src, int src_stride, int width, int height)
{
int x, y;
int16_t *src = (int16_t *)_src;
for (y = 0; y < height; y++) {
for (x = 0; x < width; x+=4) {
dst[x ] = av_clip_uint8(src[x ] + 128);
dst[x+1] = av_clip_uint8(src[x+1] + 128);
dst[x+2] = av_clip_uint8(src[x+2] + 128);
dst[x+3] = av_clip_uint8(src[x+3] + 128);
}
dst += dst_stride;
src += src_stride >> 1;
}
}
void ff_mss34_dct_put(uint8_t *dst, int stride, int *block)
{
int i, j;
int *ptr;
ptr = block;
for (i = 0; i < 8; i++) {
DCT_TEMPLATE(ptr, 1, SOP_ROW, 13);
ptr += 8;
}
ptr = block;
for (i = 0; i < 8; i++) {
DCT_TEMPLATE(ptr, 8, SOP_COL, 22);
ptr++;
}
ptr = block;
for (j = 0; j < 8; j++) {
for (i = 0; i < 8; i++)
dst[i] = av_clip_uint8(ptr[i] + 128);
dst += stride;
ptr += 8;
}
}
static void midequalizer8(const uint8_t *in0, const uint8_t *in1,
uint8_t *dst,
ptrdiff_t linesize1, ptrdiff_t linesize2,
ptrdiff_t dlinesize,
int w0, int h0,
int w1, int h1,
float *histogram1, float *histogram2,
unsigned *cchange,
size_t hsize)
{
int x, y;
compute_histogram8(in0, linesize1, w0, h0, histogram1, hsize);
compute_histogram8(in1, linesize2, w1, h1, histogram2, hsize);
compute_contrast_change(histogram1, histogram2, cchange, hsize);
for (y = 0; y < h0; y++) {
for (x = 0; x < w0; x++) {
dst[x] = av_clip_uint8(cchange[in0[x]]);
}
dst += dlinesize;
in0 += linesize1;
}
}
static void idct(uint8_t *dst, int dst_linesize, int src[64])
{
int i, j, k;
double tmp[64];
for (i = 0; i < 8; i++) {
for (j = 0; j < 8; j++) {
double sum = 0.0;
for (k = 0; k < 8; k++)
sum += c[k*8+j] * src[8*i+k];
tmp[8*i+j] = sum;
}
}
for (j = 0; j < 8; j++) {
for (i = 0; i < 8; i++) {
double sum = 0.0;
for (k = 0; k < 8; k++)
sum += c[k*8+i]*tmp[8*k+j];
dst[dst_linesize*i + j] = av_clip_uint8((int)floor(sum+0.5));
}
}
}
static inline void comp(unsigned char *dst, int dst_stride,
unsigned char *src, int src_stride, int add)
{
int j, i;
for (j=0; j<8; j++)
for (i=0; i<8; i++)
dst[j*dst_stride + i] = av_clip_uint8(src[j*src_stride + i] + add);
}
static void filter_mb_mbaff_edgev( H264Context *h, uint8_t *pix, int stride, int16_t bS[4], int bsi, int qp ) {
int i;
const int bit_depth = h->sps.bit_depth_luma;
const int qp_bd_offset = 6*(bit_depth-8);
int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;
int alpha = alpha_table[index_a] << (bit_depth-8);
int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset] << (bit_depth-8);
for( i = 0; i < 8; i++, pix += stride) {
const int bS_index = (i >> 1) * bsi;
if( bS[bS_index] == 0 ) {
continue;
}
if( bS[bS_index] < 4 ) {
const int tc0 = tc0_table[index_a][bS[bS_index]] << (bit_depth-8);
const int p0 = pix[-1];
const int p1 = pix[-2];
const int p2 = pix[-3];
const int q0 = pix[0];
const int q1 = pix[1];
const int q2 = pix[2];
if( FFABS( p0 - q0 ) < alpha &&
FFABS( p1 - p0 ) < beta &&
FFABS( q1 - q0 ) < beta ) {
int tc = tc0;
int i_delta;
if( FFABS( p2 - p0 ) < beta ) {
if(tc0)
pix[-2] = p1 + av_clip( ( p2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( p1 << 1 ) ) >> 1, -tc0, tc0 );
tc++;
}
if( FFABS( q2 - q0 ) < beta ) {
if(tc0)
pix[1] = q1 + av_clip( ( q2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( q1 << 1 ) ) >> 1, -tc0, tc0 );
tc++;
}
i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
pix[-1] = av_clip_uint8( p0 + i_delta ); /* p0' */
pix[0] = av_clip_uint8( q0 - i_delta ); /* q0' */
tprintf(h->s.avctx, "filter_mb_mbaff_edgev i:%d, qp:%d, indexA:%d, alpha:%d, beta:%d, tc:%d\n# bS:%d -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x]\n", i, qp[qp_index], index_a, alpha, beta, tc, bS[bS_index], pix[-3], p1, p0, q0, q1, pix[2], p1, pix[-1], pix[0], q1);
}
}else{
void ff_ivi_recompose_haar(const IVIPlaneDesc *plane, uint8_t *dst,
const int dst_pitch)
{
int x, y, indx, b0, b1, b2, b3, p0, p1, p2, p3;
const short *b0_ptr, *b1_ptr, *b2_ptr, *b3_ptr;
int32_t pitch;
/* all bands should have the same pitch */
pitch = plane->bands[0].pitch;
/* get pointers to the wavelet bands */
b0_ptr = plane->bands[0].buf;
b1_ptr = plane->bands[1].buf;
b2_ptr = plane->bands[2].buf;
b3_ptr = plane->bands[3].buf;
for (y = 0; y < plane->height; y += 2) {
for (x = 0, indx = 0; x < plane->width; x += 2, indx++) {
/* load coefficients */
b0 = b0_ptr[indx]; //should be: b0 = (num_bands > 0) ? b0_ptr[indx] : 0;
b1 = b1_ptr[indx]; //should be: b1 = (num_bands > 1) ? b1_ptr[indx] : 0;
b2 = b2_ptr[indx]; //should be: b2 = (num_bands > 2) ? b2_ptr[indx] : 0;
b3 = b3_ptr[indx]; //should be: b3 = (num_bands > 3) ? b3_ptr[indx] : 0;
/* haar wavelet recomposition */
p0 = (b0 + b1 + b2 + b3 + 2) >> 2;
p1 = (b0 + b1 - b2 - b3 + 2) >> 2;
p2 = (b0 - b1 + b2 - b3 + 2) >> 2;
p3 = (b0 - b1 - b2 + b3 + 2) >> 2;
/* bias, convert and output four pixels */
dst[x] = av_clip_uint8(p0 + 128);
dst[x + 1] = av_clip_uint8(p1 + 128);
dst[dst_pitch + x] = av_clip_uint8(p2 + 128);
dst[dst_pitch + x + 1] = av_clip_uint8(p3 + 128);
}// for x
dst += dst_pitch << 1;
b0_ptr += pitch;
b1_ptr += pitch;
b2_ptr += pitch;
b3_ptr += pitch;
}// for y
}
static void yuv2plane1_8_u(const int16_t *src, uint8_t *dest, int dstW,
const uint8_t *dither, int offset, int start)
{
int i;
for (i = start; i < dstW; i++) {
int val = (src[i] + dither[(i + offset) & 7]) >> 7;
dest[i] = av_clip_uint8(val);
}
}
static void filter_5x5(ConvolutionContext *s, AVFrame *in, AVFrame *out, int plane)
{
const uint8_t *src = in->data[plane];
uint8_t *dst = out->data[plane];
const int stride = in->linesize[plane];
const int bstride = s->bstride;
const int height = s->planeheight[plane];
const int width = s->planewidth[plane];
uint8_t *p0 = s->buffer + 16;
uint8_t *p1 = p0 + bstride;
uint8_t *p2 = p1 + bstride;
uint8_t *p3 = p2 + bstride;
uint8_t *p4 = p3 + bstride;
uint8_t *orig = p0, *end = p4;
const int *matrix = s->matrix[plane];
float rdiv = s->rdiv[plane];
float bias = s->bias[plane];
int y, x, i;
line_copy8(p0, src + 2 * stride, width, 2);
line_copy8(p1, src + stride, width, 2);
line_copy8(p2, src, width, 2);
src += stride;
line_copy8(p3, src, width, 2);
for (y = 0; y < height; y++) {
uint8_t *array[] = {
p0 - 2, p0 - 1, p0, p0 + 1, p0 + 2,
p1 - 2, p1 - 1, p1, p1 + 1, p1 + 2,
p2 - 2, p2 - 1, p2, p2 + 1, p2 + 2,
p3 - 2, p3 - 1, p3, p3 + 1, p3 + 2,
p4 - 2, p4 - 1, p4, p4 + 1, p4 + 2
};
src += stride * (y < height - 2 ? 1 : -1);
line_copy8(p4, src, width, 2);
for (x = 0; x < width; x++) {
int sum = 0;
for (i = 0; i < 25; i++) {
sum += *(array[i] + x) * matrix[i];
}
sum = (int)(sum * rdiv + bias + 0.5f);
dst[x] = av_clip_uint8(sum);
}
p0 = p1;
p1 = p2;
p2 = p3;
p3 = p4;
p4 = (p4 == end) ? orig: p4 + bstride;
dst += out->linesize[plane];
}
}
AVWRAP_DECL int AVWrapper_WriteFrame(uint8_t *buf)
{
int x, y, stride = g_Width * 4;
uint8_t *data[3];
// copy pointers, prepare source
memcpy(data, g_pVFrame->data, sizeof(data));
buf += (g_Height - 1) * stride;
// convert to YUV 4:2:0
for (y = 0; y < g_Height; y++) {
for (x = 0; x < g_Width; x++) {
int r = buf[x * 4 + 0];
int g = buf[x * 4 + 1];
int b = buf[x * 4 + 2];
int luma = (int)(0.299f * r + 0.587f * g + 0.114f * b);
data[0][x] = av_clip_uint8(luma);
if (!(x & 1) && !(y & 1)) {
int r = (buf[x * 4 + 0] + buf[(x + 1) * 4 + 0] +
buf[x * 4 + 0 + stride] + buf[(x + 1) * 4 + 0 + stride]) / 4;
int g = (buf[x * 4 + 1] + buf[(x + 1) * 4 + 1] +
buf[x * 4 + 1 + stride] + buf[(x + 1) * 4 + 1 + stride]) / 4;
int b = (buf[x * 4 + 2] + buf[(x + 1) * 4 + 2] +
buf[x * 4 + 2 + stride] + buf[(x + 1) * 4 + 2 + stride]) / 4;
int cr = (int)(-0.14713f * r - 0.28886f * g + 0.436f * b);
int cb = (int)( 0.615f * r - 0.51499f * g - 0.10001f * b);
data[1][x / 2] = av_clip_uint8(128 + cr);
data[2][x / 2] = av_clip_uint8(128 + cb);
}
}
buf += -stride;
data[0] += g_pVFrame->linesize[0];
if (y & 1) {
data[1] += g_pVFrame->linesize[1];
data[2] += g_pVFrame->linesize[2];
}
}
return WriteFrame(g_pVFrame);
}
static int filter_sobel(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
{
ConvolutionContext *s = ctx->priv;
ThreadData *td = arg;
AVFrame *in = td->in;
AVFrame *out = td->out;
const int plane = td->plane;
const int stride = in->linesize[plane];
const int bstride = s->bstride;
const int height = s->planeheight[plane];
const int width = s->planewidth[plane];
const int slice_start = (height * jobnr) / nb_jobs;
const int slice_end = (height * (jobnr+1)) / nb_jobs;
const uint8_t *src = in->data[plane] + slice_start * stride;
uint8_t *dst = out->data[plane] + slice_start * out->linesize[plane];
const float scale = s->scale;
const float delta = s->delta;
uint8_t *p0 = s->bptrs[jobnr] + 16;
uint8_t *p1 = p0 + bstride;
uint8_t *p2 = p1 + bstride;
uint8_t *orig = p0, *end = p2;
int y, x;
line_copy8(p0, src + stride * (slice_start == 0 ? 1 : -1), width, 1);
line_copy8(p1, src, width, 1);
for (y = slice_start; y < slice_end; y++) {
src += stride * (y < height - 1 ? 1 : -1);
line_copy8(p2, src, width, 1);
for (x = 0; x < width; x++) {
int suma = p0[x - 1] * -1 +
p0[x] * -2 +
p0[x + 1] * -1 +
p2[x - 1] * 1 +
p2[x] * 2 +
p2[x + 1] * 1;
int sumb = p0[x - 1] * -1 +
p0[x + 1] * 1 +
p1[x - 1] * -2 +
p1[x + 1] * 2 +
p2[x - 1] * -1 +
p2[x + 1] * 1;
dst[x] = av_clip_uint8(sqrt(suma*suma + sumb*sumb) * scale + delta);
}
p0 = p1;
p1 = p2;
p2 = (p2 == end) ? orig: p2 + bstride;
dst += out->linesize[plane];
}
return 0;
}
static void dirac_hpel_filter(uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, const uint8_t *src,
int stride, int width, int height)
{
int x, y;
for (y = 0; y < height; y++) {
for (x = -3; x < width+5; x++)
dstv[x] = av_clip_uint8(FILTER(src+x, stride));
for (x = 0; x < width; x++)
dstc[x] = av_clip_uint8(FILTER(dstv+x, 1));
for (x = 0; x < width; x++)
dsth[x] = av_clip_uint8(FILTER(src+x, 1));
src += stride;
dsth += stride;
dstv += stride;
dstc += stride;
}
}
static int filter_3x3(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
{
ConvolutionContext *s = ctx->priv;
ThreadData *td = arg;
AVFrame *in = td->in;
AVFrame *out = td->out;
const int plane = td->plane;
const int stride = in->linesize[plane];
const int bstride = s->bstride;
const int height = s->planeheight[plane];
const int width = s->planewidth[plane];
const int slice_start = (height * jobnr) / nb_jobs;
const int slice_end = (height * (jobnr+1)) / nb_jobs;
const uint8_t *src = in->data[plane] + slice_start * stride;
uint8_t *dst = out->data[plane] + slice_start * out->linesize[plane];
uint8_t *p0 = s->bptrs[jobnr] + 16;
uint8_t *p1 = p0 + bstride;
uint8_t *p2 = p1 + bstride;
uint8_t *orig = p0, *end = p2;
const int *matrix = s->matrix[plane];
const float rdiv = s->rdiv[plane];
const float bias = s->bias[plane];
int y, x;
line_copy8(p0, src + stride * (slice_start == 0 ? 1 : -1), width, 1);
line_copy8(p1, src, width, 1);
for (y = slice_start; y < slice_end; y++) {
src += stride * (y < height - 1 ? 1 : -1);
line_copy8(p2, src, width, 1);
for (x = 0; x < width; x++) {
int sum = p0[x - 1] * matrix[0] +
p0[x] * matrix[1] +
p0[x + 1] * matrix[2] +
p1[x - 1] * matrix[3] +
p1[x] * matrix[4] +
p1[x + 1] * matrix[5] +
p2[x - 1] * matrix[6] +
p2[x] * matrix[7] +
p2[x + 1] * matrix[8];
sum = (int)(sum * rdiv + bias + 0.5f);
dst[x] = av_clip_uint8(sum);
}
p0 = p1;
p1 = p2;
p2 = (p2 == end) ? orig: p2 + bstride;
dst += out->linesize[plane];
}
return 0;
}
/** Apply overlap transform to vertical edge
*/
static void vc1_h_overlap_c(uint8_t* src, int stride)
{
int i;
int a, b, c, d;
int d1, d2;
int rnd = 1;
for(i = 0; i < 8; i++) {
a = src[-2];
b = src[-1];
c = src[0];
d = src[1];
d1 = (a - d + 3 + rnd) >> 3;
d2 = (a - d + b - c + 4 - rnd) >> 3;
src[-2] = a - d1;
src[-1] = av_clip_uint8(b - d2);
src[0] = av_clip_uint8(c + d2);
src[1] = d + d1;
src += stride;
rnd = !rnd;
}
}
static inline void render_line_unrolled(intptr_t x, int y, int x1,
intptr_t sy, int ady, int adx,
float *buf)
{
int err = -adx;
x -= x1 - 1;
buf += x1 - 1;
while (++x < 0) {
err += ady;
if (err >= 0) {
err += ady - adx;
y += sy;
buf[x++] = ff_vorbis_floor1_inverse_db_table[av_clip_uint8(y)];
}
buf[x] = ff_vorbis_floor1_inverse_db_table[av_clip_uint8(y)];
}
if (x <= 0) {
if (err + ady >= 0)
y += sy;
buf[x] = ff_vorbis_floor1_inverse_db_table[av_clip_uint8(y)];
}
}
static void set(uint8_t *a[], int ch, int index, int ch_count, enum AVSampleFormat f, double v){
uint8_t *p;
if(av_sample_fmt_is_planar(f)){
f= av_get_alt_sample_fmt(f, 0);
p= a[ch];
}else{
p= a[0];
index= ch + index*ch_count;
}
switch(f){
case AV_SAMPLE_FMT_U8 : ((uint8_t*)p)[index]= av_clip_uint8 (lrint((v+1.0)*127)); break;
case AV_SAMPLE_FMT_S16: ((int16_t*)p)[index]= av_clip_int16 (lrint(v*32767)); break;
case AV_SAMPLE_FMT_S32: ((int32_t*)p)[index]= av_clipl_int32(llrint(v*2147483647)); break;
case AV_SAMPLE_FMT_FLT: ((float *)p)[index]= v; break;
case AV_SAMPLE_FMT_DBL: ((double *)p)[index]= v; break;
default: av_assert2(0);
}
}
static void filter_3x3(ConvolutionContext *s, AVFrame *in, AVFrame *out, int plane)
{
const uint8_t *src = in->data[plane];
uint8_t *dst = out->data[plane];
const int stride = in->linesize[plane];
const int bstride = s->bstride;
const int height = s->planeheight[plane];
const int width = s->planewidth[plane];
uint8_t *p0 = s->buffer + 16;
uint8_t *p1 = p0 + bstride;
uint8_t *p2 = p1 + bstride;
uint8_t *orig = p0, *end = p2;
const int *matrix = s->matrix[plane];
const float rdiv = s->rdiv[plane];
const float bias = s->bias[plane];
int y, x;
line_copy8(p0, src + stride, width, 1);
line_copy8(p1, src, width, 1);
for (y = 0; y < height; y++) {
src += stride * (y < height - 1 ? 1 : -1);
line_copy8(p2, src, width, 1);
for (x = 0; x < width; x++) {
int sum = p0[x - 1] * matrix[0] +
p0[x] * matrix[1] +
p0[x + 1] * matrix[2] +
p1[x - 1] * matrix[3] +
p1[x] * matrix[4] +
p1[x + 1] * matrix[5] +
p2[x - 1] * matrix[6] +
p2[x] * matrix[7] +
p2[x + 1] * matrix[8];
sum = (int)(sum * rdiv + bias + 0.5f);
dst[x] = av_clip_uint8(sum);
}
p0 = p1;
p1 = p2;
p2 = (p2 == end) ? orig: p2 + bstride;
dst += out->linesize[plane];
}
}
static void wmv2_idct_add_c(uint8_t *dest, int line_size, int16_t *block)
{
int i;
for (i = 0; i < 64; i += 8)
wmv2_idct_row(block + i);
for (i = 0; i < 8; i++)
wmv2_idct_col(block + i);
for (i = 0; i < 8; i++) {
dest[0] = av_clip_uint8(dest[0] + block[0]);
dest[1] = av_clip_uint8(dest[1] + block[1]);
dest[2] = av_clip_uint8(dest[2] + block[2]);
dest[3] = av_clip_uint8(dest[3] + block[3]);
dest[4] = av_clip_uint8(dest[4] + block[4]);
dest[5] = av_clip_uint8(dest[5] + block[5]);
dest[6] = av_clip_uint8(dest[6] + block[6]);
dest[7] = av_clip_uint8(dest[7] + block[7]);
dest += line_size;
block += 8;
}
}
static void run_postproc(AVCodecContext *avctx, AVFrame *frame)
{
DDSContext *ctx = avctx->priv_data;
int i, x_off;
switch (ctx->postproc) {
case DDS_ALPHA_EXP:
/* Alpha-exponential mode divides each channel by the maximum
* R, G or B value, and stores the multiplying factor in the
* alpha channel. */
av_log(avctx, AV_LOG_DEBUG, "Post-processing alpha exponent.\n");
for (i = 0; i < frame->linesize[0] * frame->height; i += 4) {
uint8_t *src = frame->data[0] + i;
int r = src[0];
int g = src[1];
int b = src[2];
int a = src[3];
src[0] = r * a / 255;
src[1] = g * a / 255;
src[2] = b * a / 255;
src[3] = 255;
}
break;
case DDS_NORMAL_MAP:
/* Normal maps work in the XYZ color space and they encode
* X in R or in A, depending on the texture type, Y in G and
* derive Z with a square root of the distance.
*
* http://www.realtimecollisiondetection.net/blog/?p=28 */
av_log(avctx, AV_LOG_DEBUG, "Post-processing normal map.\n");
x_off = ctx->tex_ratio == 8 ? 0 : 3;
for (i = 0; i < frame->linesize[0] * frame->height; i += 4) {
uint8_t *src = frame->data[0] + i;
int x = src[x_off];
int y = src[1];
int z = 127;
int d = (255 * 255 - x * x - y * y) / 2;
if (d > 0)
z = lrint(sqrtf(d));
src[0] = x;
src[1] = y;
src[2] = z;
src[3] = 255;
}
break;
case DDS_RAW_YCOCG:
/* Data is Y-Co-Cg-A and not RGBA, but they are represented
* with the same masks in the DDPF header. */
av_log(avctx, AV_LOG_DEBUG, "Post-processing raw YCoCg.\n");
for (i = 0; i < frame->linesize[0] * frame->height; i += 4) {
uint8_t *src = frame->data[0] + i;
int a = src[0];
int cg = src[1] - 128;
int co = src[2] - 128;
int y = src[3];
src[0] = av_clip_uint8(y + co - cg);
src[1] = av_clip_uint8(y + cg);
src[2] = av_clip_uint8(y - co - cg);
src[3] = a;
}
break;
case DDS_SWAP_ALPHA:
/* Alpha and Luma are stored swapped. */
av_log(avctx, AV_LOG_DEBUG, "Post-processing swapped Luma/Alpha.\n");
for (i = 0; i < frame->linesize[0] * frame->height; i += 2) {
uint8_t *src = frame->data[0] + i;
FFSWAP(uint8_t, src[0], src[1]);
}
break;
case DDS_SWIZZLE_A2XY:
/* Swap R and G, often used to restore a standard RGTC2. */
av_log(avctx, AV_LOG_DEBUG, "Post-processing A2XY swizzle.\n");
do_swizzle(frame, 0, 1);
break;
case DDS_SWIZZLE_RBXG:
/* Swap G and A, then B and new A (G). */
av_log(avctx, AV_LOG_DEBUG, "Post-processing RBXG swizzle.\n");
do_swizzle(frame, 1, 3);
do_swizzle(frame, 2, 3);
break;
case DDS_SWIZZLE_RGXB:
/* Swap B and A. */
av_log(avctx, AV_LOG_DEBUG, "Post-processing RGXB swizzle.\n");
do_swizzle(frame, 2, 3);
break;
case DDS_SWIZZLE_RXBG:
/* Swap G and A. */
av_log(avctx, AV_LOG_DEBUG, "Post-processing RXBG swizzle.\n");
do_swizzle(frame, 1, 3);
break;
case DDS_SWIZZLE_RXGB:
/* Swap R and A (misleading name). */
av_log(avctx, AV_LOG_DEBUG, "Post-processing RXGB swizzle.\n");
do_swizzle(frame, 0, 3);
break;
case DDS_SWIZZLE_XGBR:
/* Swap B and A, then R and new A (B). */
av_log(avctx, AV_LOG_DEBUG, "Post-processing XGBR swizzle.\n");
do_swizzle(frame, 2, 3);
do_swizzle(frame, 0, 3);
break;
case DDS_SWIZZLE_XGXR:
/* Swap G and A, then R and new A (G), then new R (G) and new G (A).
* This variant does not store any B component. */
av_log(avctx, AV_LOG_DEBUG, "Post-processing XGXR swizzle.\n");
do_swizzle(frame, 1, 3);
do_swizzle(frame, 0, 3);
do_swizzle(frame, 0, 1);
break;
case DDS_SWIZZLE_XRBG:
/* Swap G and A, then R and new A (G). */
av_log(avctx, AV_LOG_DEBUG, "Post-processing XRBG swizzle.\n");
do_swizzle(frame, 1, 3);
do_swizzle(frame, 0, 3);
break;
}
}
static void search_for_quantizers_faac(AVCodecContext *avctx, AACEncContext *s,
SingleChannelElement *sce,
const float lambda)
{
int start = 0, i, w, w2, g;
float uplim[128], maxq[128];
int minq, maxsf;
float distfact = ((sce->ics.num_windows > 1) ? 85.80 : 147.84) / lambda;
int last = 0, lastband = 0, curband = 0;
float avg_energy = 0.0;
if (sce->ics.num_windows == 1) {
start = 0;
for (i = 0; i < 1024; i++) {
if (i - start >= sce->ics.swb_sizes[curband]) {
start += sce->ics.swb_sizes[curband];
curband++;
}
if (sce->coeffs[i]) {
avg_energy += sce->coeffs[i] * sce->coeffs[i];
last = i;
lastband = curband;
}
}
} else {
for (w = 0; w < 8; w++) {
const float *coeffs = sce->coeffs + w*128;
start = 0;
for (i = 0; i < 128; i++) {
if (i - start >= sce->ics.swb_sizes[curband]) {
start += sce->ics.swb_sizes[curband];
curband++;
}
if (coeffs[i]) {
avg_energy += coeffs[i] * coeffs[i];
last = FFMAX(last, i);
lastband = FFMAX(lastband, curband);
}
}
}
}
last++;
avg_energy /= last;
if (avg_energy == 0.0f) {
for (i = 0; i < FF_ARRAY_ELEMS(sce->sf_idx); i++)
sce->sf_idx[i] = SCALE_ONE_POS;
return;
}
for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
start = w*128;
for (g = 0; g < sce->ics.num_swb; g++) {
float *coefs = sce->coeffs + start;
const int size = sce->ics.swb_sizes[g];
int start2 = start, end2 = start + size, peakpos = start;
float maxval = -1, thr = 0.0f, t;
maxq[w*16+g] = 0.0f;
if (g > lastband) {
maxq[w*16+g] = 0.0f;
start += size;
for (w2 = 0; w2 < sce->ics.group_len[w]; w2++)
memset(coefs + w2*128, 0, sizeof(coefs[0])*size);
continue;
}
for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
for (i = 0; i < size; i++) {
float t = coefs[w2*128+i]*coefs[w2*128+i];
maxq[w*16+g] = FFMAX(maxq[w*16+g], fabsf(coefs[w2*128 + i]));
thr += t;
if (sce->ics.num_windows == 1 && maxval < t) {
maxval = t;
peakpos = start+i;
}
}
}
if (sce->ics.num_windows == 1) {
start2 = FFMAX(peakpos - 2, start2);
end2 = FFMIN(peakpos + 3, end2);
} else {
start2 -= start;
end2 -= start;
}
start += size;
thr = pow(thr / (avg_energy * (end2 - start2)), 0.3 + 0.1*(lastband - g) / lastband);
t = 1.0 - (1.0 * start2 / last);
uplim[w*16+g] = distfact / (1.4 * thr + t*t*t + 0.075);
}
}
memset(sce->sf_idx, 0, sizeof(sce->sf_idx));
abs_pow34_v(s->scoefs, sce->coeffs, 1024);
for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
start = w*128;
for (g = 0; g < sce->ics.num_swb; g++) {
const float *coefs = sce->coeffs + start;
const float *scaled = s->scoefs + start;
const int size = sce->ics.swb_sizes[g];
int scf, prev_scf, step;
int min_scf = -1, max_scf = 256;
float curdiff;
if (maxq[w*16+g] < 21.544) {
sce->zeroes[w*16+g] = 1;
start += size;
continue;
}
sce->zeroes[w*16+g] = 0;
scf = prev_scf = av_clip(SCALE_ONE_POS - SCALE_DIV_512 - log2f(1/maxq[w*16+g])*16/3, 60, 218);
step = 16;
for (;;) {
float dist = 0.0f;
int quant_max;
for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
int b;
dist += quantize_band_cost(s, coefs + w2*128,
scaled + w2*128,
sce->ics.swb_sizes[g],
scf,
ESC_BT,
lambda,
INFINITY,
&b);
dist -= b;
}
dist *= 1.0f / 512.0f / lambda;
quant_max = quant(maxq[w*16+g], ff_aac_pow2sf_tab[POW_SF2_ZERO - scf + SCALE_ONE_POS - SCALE_DIV_512]);
if (quant_max >= 8191) { // too much, return to the previous quantizer
sce->sf_idx[w*16+g] = prev_scf;
break;
}
prev_scf = scf;
curdiff = fabsf(dist - uplim[w*16+g]);
if (curdiff <= 1.0f)
step = 0;
else
step = log2f(curdiff);
if (dist > uplim[w*16+g])
step = -step;
scf += step;
scf = av_clip_uint8(scf);
step = scf - prev_scf;
if (FFABS(step) <= 1 || (step > 0 && scf >= max_scf) || (step < 0 && scf <= min_scf)) {
sce->sf_idx[w*16+g] = av_clip(scf, min_scf, max_scf);
break;
}
if (step > 0)
min_scf = prev_scf;
else
max_scf = prev_scf;
}
start += size;
}
}
minq = sce->sf_idx[0] ? sce->sf_idx[0] : INT_MAX;
for (i = 1; i < 128; i++) {
if (!sce->sf_idx[i])
sce->sf_idx[i] = sce->sf_idx[i-1];
else
minq = FFMIN(minq, sce->sf_idx[i]);
}
if (minq == INT_MAX)
minq = 0;
minq = FFMIN(minq, SCALE_MAX_POS);
maxsf = FFMIN(minq + SCALE_MAX_DIFF, SCALE_MAX_POS);
for (i = 126; i >= 0; i--) {
if (!sce->sf_idx[i])
sce->sf_idx[i] = sce->sf_idx[i+1];
sce->sf_idx[i] = av_clip(sce->sf_idx[i], minq, maxsf);
}
}
static int dpcm_decode_frame(AVCodecContext *avctx, void *data, int *data_size,
AVPacket *avpkt)
{
const uint8_t *buf = avpkt->data;
int buf_size = avpkt->size;
const uint8_t *buf_end = buf + buf_size;
DPCMContext *s = avctx->priv_data;
int out = 0;
int predictor[2];
int ch = 0;
int stereo = s->channels - 1;
int16_t *output_samples = data;
/* calculate output size */
switch(avctx->codec->id) {
case CODEC_ID_ROQ_DPCM:
out = buf_size - 8;
break;
case CODEC_ID_INTERPLAY_DPCM:
out = buf_size - 6 - s->channels;
break;
case CODEC_ID_XAN_DPCM:
out = buf_size - 2 * s->channels;
break;
case CODEC_ID_SOL_DPCM:
if (avctx->codec_tag != 3)
out = buf_size * 2;
else
out = buf_size;
break;
}
out *= av_get_bytes_per_sample(avctx->sample_fmt);
if (out <= 0) {
av_log(avctx, AV_LOG_ERROR, "packet is too small\n");
return AVERROR(EINVAL);
}
if (*data_size < out) {
av_log(avctx, AV_LOG_ERROR, "output buffer is too small\n");
return AVERROR(EINVAL);
}
switch(avctx->codec->id) {
case CODEC_ID_ROQ_DPCM:
buf += 6;
if (stereo) {
predictor[1] = (int16_t)(bytestream_get_byte(&buf) << 8);
predictor[0] = (int16_t)(bytestream_get_byte(&buf) << 8);
} else {
predictor[0] = (int16_t)bytestream_get_le16(&buf);
}
/* decode the samples */
while (buf < buf_end) {
predictor[ch] += s->roq_square_array[*buf++];
predictor[ch] = av_clip_int16(predictor[ch]);
*output_samples++ = predictor[ch];
/* toggle channel */
ch ^= stereo;
}
break;
case CODEC_ID_INTERPLAY_DPCM:
buf += 6; /* skip over the stream mask and stream length */
for (ch = 0; ch < s->channels; ch++) {
predictor[ch] = (int16_t)bytestream_get_le16(&buf);
*output_samples++ = predictor[ch];
}
ch = 0;
while (buf < buf_end) {
predictor[ch] += interplay_delta_table[*buf++];
predictor[ch] = av_clip_int16(predictor[ch]);
*output_samples++ = predictor[ch];
/* toggle channel */
ch ^= stereo;
}
break;
case CODEC_ID_XAN_DPCM:
{
int shift[2] = { 4, 4 };
for (ch = 0; ch < s->channels; ch++)
predictor[ch] = (int16_t)bytestream_get_le16(&buf);
ch = 0;
while (buf < buf_end) {
uint8_t n = *buf++;
int16_t diff = (n & 0xFC) << 8;
if ((n & 0x03) == 3)
shift[ch]++;
else
shift[ch] -= (2 * (n & 3));
/* saturate the shifter to a lower limit of 0 */
if (shift[ch] < 0)
shift[ch] = 0;
diff >>= shift[ch];
predictor[ch] += diff;
predictor[ch] = av_clip_int16(predictor[ch]);
*output_samples++ = predictor[ch];
/* toggle channel */
ch ^= stereo;
}
break;
}
case CODEC_ID_SOL_DPCM:
if (avctx->codec_tag != 3) {
uint8_t *output_samples_u8 = data;
while (buf < buf_end) {
uint8_t n = *buf++;
s->sample[0] += s->sol_table[n >> 4];
s->sample[0] = av_clip_uint8(s->sample[0]);
*output_samples_u8++ = s->sample[0];
s->sample[stereo] += s->sol_table[n & 0x0F];
s->sample[stereo] = av_clip_uint8(s->sample[stereo]);
*output_samples_u8++ = s->sample[stereo];
}
} else {
while (buf < buf_end) {
static inline void p8idct(DCTELEM data[64], FLOAT temp[64], uint8_t *dest, int stride, int x, int y, int type){
int i;
FLOAT av_unused tmp0;
FLOAT s04, d04, s17, d17, s26, d26, s53, d53;
FLOAT os07, os16, os25, os34;
FLOAT od07, od16, od25, od34;
for(i=0; i<y*8; i+=y){
s17= temp[1*x + i] + temp[7*x + i];
d17= temp[1*x + i] - temp[7*x + i];
s53= temp[5*x + i] + temp[3*x + i];
d53= temp[5*x + i] - temp[3*x + i];
od07= s17 + s53;
od25= (s17 - s53)*(2*A4);
#if 0 //these 2 are equivalent
tmp0= (d17 + d53)*(2*A2);
od34= d17*( 2*B6) - tmp0;
od16= d53*(-2*B2) + tmp0;
#else
od34= d17*(2*(B6-A2)) - d53*(2*A2);
od16= d53*(2*(A2-B2)) + d17*(2*A2);
#endif
od16 -= od07;
od25 -= od16;
od34 += od25;
s26 = temp[2*x + i] + temp[6*x + i];
d26 = temp[2*x + i] - temp[6*x + i];
d26*= 2*A4;
d26-= s26;
s04= temp[0*x + i] + temp[4*x + i];
d04= temp[0*x + i] - temp[4*x + i];
os07= s04 + s26;
os34= s04 - s26;
os16= d04 + d26;
os25= d04 - d26;
if(type==0){
temp[0*x + i]= os07 + od07;
temp[7*x + i]= os07 - od07;
temp[1*x + i]= os16 + od16;
temp[6*x + i]= os16 - od16;
temp[2*x + i]= os25 + od25;
temp[5*x + i]= os25 - od25;
temp[3*x + i]= os34 - od34;
temp[4*x + i]= os34 + od34;
}else if(type==1){
data[0*x + i]= lrintf(os07 + od07);
data[7*x + i]= lrintf(os07 - od07);
data[1*x + i]= lrintf(os16 + od16);
data[6*x + i]= lrintf(os16 - od16);
data[2*x + i]= lrintf(os25 + od25);
data[5*x + i]= lrintf(os25 - od25);
data[3*x + i]= lrintf(os34 - od34);
data[4*x + i]= lrintf(os34 + od34);
}else if(type==2){
dest[0*stride + i]= av_clip_uint8(((int)dest[0*stride + i]) + lrintf(os07 + od07));
dest[7*stride + i]= av_clip_uint8(((int)dest[7*stride + i]) + lrintf(os07 - od07));
dest[1*stride + i]= av_clip_uint8(((int)dest[1*stride + i]) + lrintf(os16 + od16));
dest[6*stride + i]= av_clip_uint8(((int)dest[6*stride + i]) + lrintf(os16 - od16));
dest[2*stride + i]= av_clip_uint8(((int)dest[2*stride + i]) + lrintf(os25 + od25));
dest[5*stride + i]= av_clip_uint8(((int)dest[5*stride + i]) + lrintf(os25 - od25));
dest[3*stride + i]= av_clip_uint8(((int)dest[3*stride + i]) + lrintf(os34 - od34));
dest[4*stride + i]= av_clip_uint8(((int)dest[4*stride + i]) + lrintf(os34 + od34));
}else{
dest[0*stride + i]= av_clip_uint8(lrintf(os07 + od07));
dest[7*stride + i]= av_clip_uint8(lrintf(os07 - od07));
dest[1*stride + i]= av_clip_uint8(lrintf(os16 + od16));
dest[6*stride + i]= av_clip_uint8(lrintf(os16 - od16));
dest[2*stride + i]= av_clip_uint8(lrintf(os25 + od25));
dest[5*stride + i]= av_clip_uint8(lrintf(os25 - od25));
dest[3*stride + i]= av_clip_uint8(lrintf(os34 - od34));
dest[4*stride + i]= av_clip_uint8(lrintf(os34 + od34));
}
}
}
CONV_FUNC(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_U8 , *(const uint8_t*)pi)
CONV_FUNC(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80U)<<8)
CONV_FUNC(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80U)<<24)
CONV_FUNC(AV_SAMPLE_FMT_FLT, float , AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)*(1.0f/ (1<<7)))
CONV_FUNC(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)*(1.0 / (1<<7)))
CONV_FUNC(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_S16, (*(const int16_t*)pi>>8) + 0x80)
CONV_FUNC(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_S16, *(const int16_t*)pi)
CONV_FUNC(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_S16, *(const int16_t*)pi<<16)
CONV_FUNC(AV_SAMPLE_FMT_FLT, float , AV_SAMPLE_FMT_S16, *(const int16_t*)pi*(1.0f/ (1<<15)))
CONV_FUNC(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_S16, *(const int16_t*)pi*(1.0 / (1<<15)))
CONV_FUNC(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_S32, (*(const int32_t*)pi>>24) + 0x80)
CONV_FUNC(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_S32, *(const int32_t*)pi>>16)
CONV_FUNC(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_S32, *(const int32_t*)pi)
CONV_FUNC(AV_SAMPLE_FMT_FLT, float , AV_SAMPLE_FMT_S32, *(const int32_t*)pi*(1.0f/ (1U<<31)))
CONV_FUNC(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_S32, *(const int32_t*)pi*(1.0 / (1U<<31)))
CONV_FUNC(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_FLT, av_clip_uint8( lrintf(*(const float*)pi * (1<<7)) + 0x80))
CONV_FUNC(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, av_clip_int16( lrintf(*(const float*)pi * (1<<15))))
CONV_FUNC(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, av_clipl_int32(llrintf(*(const float*)pi * (1U<<31))))
CONV_FUNC(AV_SAMPLE_FMT_FLT, float , AV_SAMPLE_FMT_FLT, *(const float*)pi)
CONV_FUNC(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_FLT, *(const float*)pi)
CONV_FUNC(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_DBL, av_clip_uint8( lrint(*(const double*)pi * (1<<7)) + 0x80))
CONV_FUNC(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, av_clip_int16( lrint(*(const double*)pi * (1<<15))))
CONV_FUNC(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, av_clipl_int32(llrint(*(const double*)pi * (1U<<31))))
CONV_FUNC(AV_SAMPLE_FMT_FLT, float , AV_SAMPLE_FMT_DBL, *(const double*)pi)
CONV_FUNC(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_DBL, *(const double*)pi)
#define FMT_PAIR_FUNC(out, in) [(out) + AV_SAMPLE_FMT_NB*(in)] = CONV_FUNC_NAME(out, in)
static conv_func_type * const fmt_pair_to_conv_functions[AV_SAMPLE_FMT_NB*AV_SAMPLE_FMT_NB] = {
FMT_PAIR_FUNC(AV_SAMPLE_FMT_U8 , AV_SAMPLE_FMT_U8 ),
FMT_PAIR_FUNC(AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_U8 ),
