void write_back_motion_TI( const int ai_iB_stride, const int ai_iB8_stride , short MvdL0[ ],short ai_tiMv_cache[ ][2],
short ao_tiRef[ ], short ai_tiRef_cache[])
{
//1 Iteration
_mem8(MvdL0) =_mem8(ai_tiMv_cache+12);
_mem8(MvdL0+4)=_mem8(ai_tiMv_cache+14);
MvdL0 += ai_iB_stride;
//2 iteration
_mem8(MvdL0) =_mem8(ai_tiMv_cache+20);
_mem8(MvdL0+4)=_mem8(ai_tiMv_cache+22);
MvdL0 += ai_iB_stride;
//3 iteration
_mem8(MvdL0) =_mem8(ai_tiMv_cache+28);
_mem8(MvdL0+4)=_mem8(ai_tiMv_cache+30);
MvdL0 += ai_iB_stride;
//4 iteration
_mem8(MvdL0) =_mem8(ai_tiMv_cache+36);
_mem8(MvdL0+4)=_mem8(ai_tiMv_cache+38);
MvdL0 += ai_iB_stride;
_mem4(ao_tiRef)=(ai_tiRef_cache[14]<<16|ai_tiRef_cache[12]);
_mem4(ao_tiRef+ai_iB8_stride)=(ai_tiRef_cache[30]<<16|ai_tiRef_cache[28]);
}
static __inline void *optimized_mem_set(void *mem, int ch, size_t n)
{
char * restrict dst1, * restrict dst2;
int pre_bytes, post_bytes, wfill, i;
unsigned char *outbuf = mem;
unsigned int count = n;
dst1 = (char *)outbuf;
#if defined(_TMS320C6400) || defined(_TMS320C6740) || defined(_TMS320C6600) || \
defined(_TI_C6X_TESLA)
/*---------------------------------------------------------------------*/
/* We do not use 'dwfill' on other variations of the C6x architecture, */
/* so limit 'dwfill' references to the architectures that use it. */
/*---------------------------------------------------------------------*/
{
long long dwfill;
/*------------------------------------------------------------------*/
/* Set up 64-bit and 32-bit fill values. */
/*------------------------------------------------------------------*/
wfill = _pack2 (ch, ch);
wfill = _packl4(wfill, wfill);
dwfill = _itoll (wfill, wfill);
/*------------------------------------------------------------------*/
/* Calculate # of bytes to pre-copy to get to an alignment of 8 */
/*------------------------------------------------------------------*/
pre_bytes = (8 - (int) dst1) & 7;
if (count > pre_bytes)
{
count -= pre_bytes;
if (pre_bytes & 1) { *dst1 = ch; dst1 += 1; }
if (pre_bytes & 2) { _amem2(dst1) = wfill; dst1 += 2; }
if (pre_bytes & 4) { _amem4(dst1) = wfill; dst1 += 4; }
}
/*------------------------------------------------------------------*/
/* Double word fills */
/*------------------------------------------------------------------*/
post_bytes = count > 0 ? count : 0;
dst2 = dst1 + 8;
if (count > 15)
for (i = 0; i < count >> 4; i++)
{
_amem8(dst1) = dwfill; dst1 += 16;
_amem8(dst2) = dwfill; dst2 += 16;
post_bytes -= 16;
}
/*------------------------------------------------------------------*/
/* Finish transfer with 8, 4, 2 and/or 1-byte writes */
/*------------------------------------------------------------------*/
if (post_bytes & 8) { _mem8(dst1) = dwfill; dst1 += 8; }
if (post_bytes & 4) { _mem4(dst1) = wfill; dst1 += 4; }
if (post_bytes & 2) { dst1[0] = ch;
dst1[1] = ch; dst1 += 2; }
if (post_bytes & 1) { *dst1 = ch; dst1 += 1; }
}
#else
/*--------------------------------------------------------------------*/
/* Set up 32-bit fill value. */
/*--------------------------------------------------------------------*/
wfill = _mpy(0x101, (int)ch);
wfill += (wfill << 16);
/*--------------------------------------------------------------------*/
/* Calculate number of bytes to pre-copy to get to an alignment of 4 */
/*--------------------------------------------------------------------*/
pre_bytes = (4 - (int) dst1) & 3;
if (count > pre_bytes)
{
count -= pre_bytes;
if (pre_bytes & 1) { *dst1 = ch; dst1 += 1; }
if (pre_bytes & 2) { _amem2(dst1) = wfill; dst1 += 2; }
}
/*--------------------------------------------------------------------*/
/* Double word fills */
/*--------------------------------------------------------------------*/
post_bytes = count > 0 ? count : 0;
dst2 = dst1 + 4;
if (count > 7)
for (i = 0; i < count >> 3; i++)
{
_amem4(dst1) = wfill; dst1 += 8;
_amem4(dst2) = wfill; dst2 += 8;
post_bytes -= 8;
}
/*--------------------------------------------------------------------*/
/* Finish transfer with up to 7 single-byte writes. */
/*--------------------------------------------------------------------*/
if (post_bytes) { *dst1++ = ch; post_bytes--; }
if (post_bytes) { *dst1++ = ch; post_bytes--; }
if (post_bytes) { *dst1++ = ch; post_bytes--; }
if (post_bytes) { *dst1++ = ch; post_bytes--; }
if (post_bytes) { *dst1++ = ch; post_bytes--; }
if (post_bytes) { *dst1++ = ch; post_bytes--; }
if (post_bytes) { *dst1++ = ch; post_bytes--; }
#endif
return dst1;
}
