static void dequant_idct_block_8x8(int16_t *in_data, int16_t *out_data, float *quant_tbl)
{
float mb[8 * 8] __attribute((aligned(32)));
float mb2[8 * 8] __attribute((aligned(32)));
int i, v;
for (i = 0; i < 64; ++i)
{
mb[i] = in_data[i];
}
dequantize_block(mb, mb2, quant_tbl);
scale_block(mb2, mb);
/* Two 1D inverse DCT operations with transpose */
for (v = 0; v < 8; ++v)
{
dct_1d_general(mb + v * 8, mb2 + v * 8, dctlookup_trans);
}
transpose_block(mb2, mb);
for (v = 0; v < 8; ++v)
{
dct_1d_general(mb + v * 8, mb2 + v * 8, dctlookup_trans);
}
transpose_block(mb2, mb);
for (i = 0; i < 64; ++i)
{
out_data[i] = mb[i];
}
}
// for debugging and profiling
void rotate_core(int dim, pixel *src, pixel *dst, int D)
{
if (dim < 256)
{
#if DEBUG_PRINT
log("doing improved naive rotate");
#endif
new_rotate(dim, src, dst, D*2, D/2);
}
else
{
#if DEBUG_PRINT
log("doing 2 step rotate on input size %d", dim);
#endif
#if DEBUG_PROFILE_TIME
clock_t test, start, middle, end;
test = clock();
copy_block(dim, src, dst, D);
start = clock();
#elif DEBUG_PROFILE_TIMES
unsigned test, start, middle, end;
test = get_seconds();
copy_block(dim, src, dst, D);
start = get_seconds();
#endif
// transpose the image in src and store
transpose_block(dim, src, dst, D);
#if DEBUG_PROFILE_TIME
middle = clock();
#elif DEBUG_PROFILE_TIMES
middle = get_seconds();
#endif
// perform row exchange on the transposed image
flip_ip(dim, dst);
#if DEBUG_PROFILE_TIME
end = clock();
log("test took %u transpose took %u and flip took %u", (float)start - (float)test, (float)middle - (float)start, (float)end - (float)middle);
#elif DEBUG_PROFILE_TIMES
end = get_seconds();
log("test took %u transpose took %u and flip took %u", start - test, middle - start, end - middle);
#endif
}
}
