_CODE_ACCESS void *memset(void *dst, int fill, size_t len)
{
char *restrict dst1, *restrict dst2;
int pre_bytes, post_bytes, wfill, i;
double dfill1, dfill2;
dst1 = (char *)dst;
/*--------------------------------------------------------------------*/
/* Replicate the 8-bit value in fill into all 4 bytes of wfill */
/*--------------------------------------------------------------------*/
wfill = _pack2 (fill, fill);
wfill = _packl4(wfill, wfill);
dfill1 = _itod (wfill, wfill);
dfill2 = _itod (wfill, wfill);
/*--------------------------------------------------------------------*/
/* Calculate number of bytes to pre-copy to get to an alignment of 8 */
/*--------------------------------------------------------------------*/
pre_bytes = (8 - (int) dst) & 7;
if (len > pre_bytes)
{
len -= pre_bytes;
if (pre_bytes & 1) { *dst1 = fill; 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 = len > 0 ? len : 0;
dst2 = dst1 + 8;
if (len > 15)
for (i = 0; i < len >> 4; i++)
{
_amemd8(dst1) = dfill1; dst1 += 16;
_amemd8(dst2) = dfill2; dst2 += 16;
post_bytes -= 16;
}
/*--------------------------------------------------------------------*/
/* Finish transfer with 8, 4, 2 and/or 1-byte writes */
/*--------------------------------------------------------------------*/
if (post_bytes & 8) { _memd8(dst1) = dfill1; dst1 += 8; }
if (post_bytes & 4) { _mem4 (dst1) = wfill; dst1 += 4; }
if (post_bytes & 2) { dst1[0] = wfill;
dst1[1] = wfill; dst1 += 2; }
if (post_bytes & 1) { *dst1 = fill; dst1 += 1; }
return dst;
}
/**
This function allows to get the luminance prediction of a non IDR picture when xFracl = 0 and yFracl = 1.
@param image Table of current frame.
@param refPicLXl Table of the reference decoded picture buffer.
@param PicWidthSamples Stride of the reference buffer.
@param stride Stride of the current image.
*/
void luma_sample_interp_0_1_TI(unsigned char image [], unsigned char refPicLXl[], const short PicWidthSamples, const short stride){
/* No horizontal interpolation */
unsigned int uiLine1,uiLine2,uiLine3,uiLine4,uiLine5,uiLine6,uiLine7,uiLine8,uiLine9;
unsigned int uiTmpLine12_h,uiTmpLine34_h,uiTmpLine12_l,uiTmpLine34_l,uiTmpLine1234_4,uiTmpLine1234_2,uiTmpLine1234_3,uiTmpLine1234_1;
unsigned int uiTmpLine56_h,uiTmpLine78_h,uiTmpLine56_l,uiTmpLine78_l,uiTmpLine5678_4,uiTmpLine5678_2,uiTmpLine5678_3,uiTmpLine5678_1;
unsigned int tmpc1_1,tmpc1_2,tmpc2_1,tmpc2_2,tmpc1,tmpc2;
unsigned int tmpc3_1,tmpc3_2,tmpc4_1,tmpc4_2,tmpc3,tmpc4;
unsigned int tmp12,tmp34,tmpend1,tmpend2,tmpend3,tmpend4;
unsigned int tmpl9l,tmpl9h;
unsigned int input1,input2,input3,input4;
unsigned char* pRefImgPtr;
unsigned char* pImgPtr;
pRefImgPtr = refPicLXl-(PicWidthSamples<<1);
pImgPtr = image;
uiLine1 = _mem4(pRefImgPtr);
pRefImgPtr += PicWidthSamples;
uiLine2 = _mem4(pRefImgPtr);
pRefImgPtr += PicWidthSamples;
uiLine3 = _mem4(pRefImgPtr);
pRefImgPtr += PicWidthSamples;
uiLine4 = _mem4(pRefImgPtr);
pRefImgPtr += PicWidthSamples;
uiLine5 = _mem4(pRefImgPtr);
pRefImgPtr += PicWidthSamples;
uiLine6 = _mem4(pRefImgPtr);
pRefImgPtr += PicWidthSamples;
uiLine7 = _mem4(pRefImgPtr);
pRefImgPtr += PicWidthSamples;
uiLine8 = _mem4(pRefImgPtr);
pRefImgPtr += PicWidthSamples;
uiLine9 = _mem4(pRefImgPtr);
input1 = uiLine3;
input2 = uiLine4;
input3 = uiLine5;
input4 = uiLine6;
uiTmpLine12_h = _packh4(uiLine1,uiLine2);
uiTmpLine34_h = _packh4(uiLine3,uiLine4);
uiTmpLine12_l = _packl4(uiLine1,uiLine2);
uiTmpLine34_l = _packl4(uiLine3,uiLine4);
uiTmpLine1234_4 = _packh4(uiTmpLine12_h,uiTmpLine34_h);
uiTmpLine1234_2 = _packl4(uiTmpLine12_h,uiTmpLine34_h);
uiTmpLine1234_3 = _packh4(uiTmpLine12_l,uiTmpLine34_l);
uiTmpLine1234_1 = _packl4(uiTmpLine12_l,uiTmpLine34_l);
uiTmpLine56_h = _packh4(uiLine5,uiLine6);
uiTmpLine78_h = _packh4(uiLine7,uiLine8);
uiTmpLine56_l = _packl4(uiLine5,uiLine6);
uiTmpLine78_l = _packl4(uiLine7,uiLine8);
uiTmpLine5678_4 = _packh4(uiTmpLine56_h,uiTmpLine78_h);
uiTmpLine5678_2 = _packl4(uiTmpLine56_h,uiTmpLine78_h);
uiTmpLine5678_3 = _packh4(uiTmpLine56_l,uiTmpLine78_l);
uiTmpLine5678_1 = _packl4(uiTmpLine56_l,uiTmpLine78_l);
tmpc1_1 = _dotpsu4(0x01FB1414,uiTmpLine1234_1);
tmpc1_2 = _dotpsu4(0xFB010000,uiTmpLine5678_1);
tmpc2_1 = _dotpsu4(0x01FB1414,uiTmpLine1234_2);
tmpc2_2 = _dotpsu4(0xFB010000,uiTmpLine5678_2);
tmpc1 = _spack2(tmpc1_1,tmpc2_1);
tmpc2 = _spack2(tmpc1_2,tmpc2_2);
tmp12 = _sadd2(tmpc1,tmpc2);
tmp12 = _shr2(_sadd2(tmp12,0x00100010),5);
tmpc3_1 = _dotpsu4(0x01FB1414,uiTmpLine1234_3);
tmpc3_2 = _dotpsu4(0xFB010000,uiTmpLine5678_3);
tmpc4_1 = _dotpsu4(0x01FB1414,uiTmpLine1234_4);
tmpc4_2 = _dotpsu4(0xFB010000,uiTmpLine5678_4);
tmpc3 = _spack2(tmpc3_1,tmpc4_1);
tmpc4 = _spack2(tmpc3_2,tmpc4_2);
tmp34 = _sadd2(tmpc3,tmpc4);
tmp34 = _shr2(_sadd2(tmp34,0x00100010),5);
tmpend1 = _spacku4(tmp34,tmp12);
tmpend1 = _swap4(tmpend1);
_amem4(pImgPtr) = _avgu4(tmpend1,input1);
pImgPtr += stride;
tmpc1_1 = _dotpsu4(0x0001FB14,uiTmpLine1234_1);
tmpc1_2 = _dotpsu4(0x14FB0100,uiTmpLine5678_1);
tmpc2_1 = _dotpsu4(0x0001FB14,uiTmpLine1234_2);
tmpc2_2 = _dotpsu4(0x14FB0100,uiTmpLine5678_2);
tmpc1 = _spack2(tmpc1_1,tmpc2_1);
tmpc2 = _spack2(tmpc1_2,tmpc2_2);
tmp12 = _sadd2(tmpc1,tmpc2);
tmp12 = _shr2(_sadd2(tmp12,0x00100010),5);
tmpc3_1 = _dotpsu4(0x0001FB14,uiTmpLine1234_3);
tmpc3_2 = _dotpsu4(0x14FB0100,uiTmpLine5678_3);
tmpc4_1 = _dotpsu4(0x0001FB14,uiTmpLine1234_4);
tmpc4_2 = _dotpsu4(0x14FB0100,uiTmpLine5678_4);
tmpc3 = _spack2(tmpc3_1,tmpc4_1);
tmpc4 = _spack2(tmpc3_2,tmpc4_2);
tmp34 = _sadd2(tmpc3,tmpc4);
tmp34 = _shr2(_sadd2(tmp34,0x00100010),5);
tmpend2 = _spacku4(tmp34,tmp12);
tmpend2 = _swap4(tmpend2);
_amem4(pImgPtr) = _avgu4(tmpend2,input2);
pImgPtr += stride;
tmpc1_1 = _dotpsu4(0x000001FB,uiTmpLine1234_1);
tmpc1_2 = _dotpsu4(0x1414FB01,uiTmpLine5678_1);
tmpc2_1 = _dotpsu4(0x000001FB,uiTmpLine1234_2);
tmpc2_2 = _dotpsu4(0x1414FB01,uiTmpLine5678_2);
tmpc1 = _spack2(tmpc1_1,tmpc2_1);
tmpc2 = _spack2(tmpc1_2,tmpc2_2);
tmp12 = _sadd2(tmpc1,tmpc2);
tmp12 = _shr2(_sadd2(tmp12,0x00100010),5);
tmpc3_1 = _dotpsu4(0x000001FB,uiTmpLine1234_3);
tmpc3_2 = _dotpsu4(0x1414FB01,uiTmpLine5678_3);
tmpc4_1 = _dotpsu4(0x000001FB,uiTmpLine1234_4);
tmpc4_2 = _dotpsu4(0x1414FB01,uiTmpLine5678_4);
tmpc3 = _spack2(tmpc3_1,tmpc4_1);
tmpc4 = _spack2(tmpc3_2,tmpc4_2);
tmp34 = _sadd2(tmpc3,tmpc4);
tmp34 = _shr2(_sadd2(tmp34,0x00100010),5);
tmpend3 = _spacku4(tmp34,tmp12);
tmpend3 = _swap4(tmpend3);
_amem4(pImgPtr) = _avgu4(tmpend3,input3);
pImgPtr += stride;
uiLine9 = _swap4(uiLine9);
tmpl9h = _unpkhu4 (uiLine9);
tmpl9l = _unpklu4 (uiLine9);
tmpc1_1 = _extu(uiTmpLine1234_1,24,24);//_dotpsu4(0x00000001,uiTmpLine1234_1);
tmpc1_2 = _dotpsu4(0xFB1414FB,uiTmpLine5678_1);
tmpc2_1 = _extu(uiTmpLine1234_2,24,24);//_dotpsu4(0x00000001,uiTmpLine1234_2);
tmpc2_2 = _dotpsu4(0xFB1414FB,uiTmpLine5678_2);
tmpc1 = _spack2(tmpc1_1,tmpc2_1);
tmpc2 = _spack2(tmpc1_2,tmpc2_2);
tmp12 = _sadd2(tmpc1,tmpc2);
tmp12 = _sadd2(tmp12,tmpl9l);
tmp12 = _shr2(_sadd2(tmp12,0x00100010),5);
tmpc3_1 = _extu(uiTmpLine1234_3,24,24);//_dotpsu4(0x00000001,uiTmpLine1234_3);
tmpc3_2 = _dotpsu4(0xFB1414FB,uiTmpLine5678_3);
tmpc4_1 = _extu(uiTmpLine1234_4,24,24);//_dotpsu4(0x00000001,uiTmpLine1234_4);
tmpc4_2 = _dotpsu4(0xFB1414FB,uiTmpLine5678_4);
tmpc3 = _spack2(tmpc3_1,tmpc4_1);
tmpc4 = _spack2(tmpc3_2,tmpc4_2);
tmp34 = _sadd2(tmpc3,tmpc4);
tmp34 = _sadd2(tmp34,tmpl9h);
tmp34 = _shr2(_sadd2(tmp34,0x00100010),5);
tmpend4 = _spacku4(tmp34,tmp12);
tmpend4 = _swap4(tmpend4);
_amem4(pImgPtr) = _avgu4(tmpend4,input4);
}
void chroma_sample_interpolation_TI(unsigned char image_Cb [RESTRICT], unsigned char image_Cr [RESTRICT]
, unsigned char refPicLXCb[RESTRICT], unsigned char refPicLXCr[RESTRICT]
, const short xFracl, const short yFracl, const short PicWidthSamples,const short stride)
{
unsigned char* pucCbPtrA = refPicLXCb;
unsigned char* pucCbPtrB = refPicLXCb + 1;
unsigned char* pucCbPtrC = refPicLXCb + PicWidthSamples;
unsigned char* pucCbPtrD = refPicLXCb + PicWidthSamples + 1;
unsigned char* pucCrPtrE = refPicLXCr;
unsigned char* pucCrPtrF = refPicLXCr + 1;
unsigned char* pucCrPtrG = refPicLXCr + PicWidthSamples;
unsigned char* pucCrPtrH = refPicLXCr + PicWidthSamples + 1;
unsigned char* pucOutputCbPtr = image_Cb;
unsigned char* pucOutputCrPtr = image_Cr;
unsigned int uiTmp1,uiTmp2;
unsigned int ui1_1,ui1_2,ui2_1,ui2_2,res_1,res_2,res_3,res_4;
unsigned int tmpend1_1,tmpend1_2,tmpend2_1,tmpend2_2;
unsigned int uiA,uiB,uiC,uiD;
unsigned int uiE,uiF,uiG,uiH;
unsigned int uicst = xFracl * yFracl;
uiTmp1 = _pack2(uicst,uicst);
uiTmp2 = (_pack2(xFracl,yFracl)) << 3;
uiTmp2 = _sub2(uiTmp2,uiTmp1);
uiTmp1 = (uicst) + ((uicst - ((xFracl + yFracl) <<3) + 64) << 16);
uicst = _packh2(uiTmp1,uiTmp2); // cst2 cst3
uiTmp1 = _pack2(uiTmp2,uiTmp1); // cst4 cst1
uicst = _spacku4(uicst,uiTmp1);
uiA = _mem2(pucCbPtrA);
uiB = _mem2(pucCbPtrB);
uiC = _mem2(pucCbPtrC);
uiD = _mem2(pucCbPtrD);
uiE = _mem2(pucCrPtrE);
uiF = _mem2(pucCrPtrF);
uiG = _mem2(pucCrPtrG);
uiH = _mem2(pucCrPtrH);
pucCbPtrA += PicWidthSamples;
pucCbPtrB += PicWidthSamples;
pucCbPtrC += PicWidthSamples;
pucCbPtrD += PicWidthSamples;
pucCrPtrE += PicWidthSamples;
pucCrPtrF += PicWidthSamples;
pucCrPtrG += PicWidthSamples;
pucCrPtrH += PicWidthSamples;
uiA += (_mem2(pucCbPtrA) << 16);
uiB += (_mem2(pucCbPtrB) << 16);
uiC += (_mem2(pucCbPtrC) << 16);
uiD += (_mem2(pucCbPtrD) << 16);
uiE += (_mem2(pucCrPtrE) << 16);
uiF += (_mem2(pucCrPtrF) << 16);
uiG += (_mem2(pucCrPtrG) << 16);
uiH += (_mem2(pucCrPtrH) << 16);
uiTmp1 = _packh4(uiA,uiB);
uiTmp2 = _packh4(uiC,uiD);
ui1_1 = _packh4(uiTmp1,uiTmp2);
ui2_1 = _packl4(uiTmp1,uiTmp2);
uiTmp1 = _packl4(uiA,uiB);
uiTmp2 = _packl4(uiC,uiD);
ui1_2 = _packh4(uiTmp1,uiTmp2);
ui2_2 = _packl4(uiTmp1,uiTmp2);
tmpend1_1 = _dotpu4(uicst,ui1_1);
tmpend1_2 = _dotpu4(uicst,ui1_2);
tmpend2_1 = _dotpu4(uicst,ui2_1);
tmpend2_2 = _dotpu4(uicst,ui2_2);
res_1 = _pack2(tmpend1_1,tmpend1_2);
res_2 = _pack2(tmpend2_1,tmpend2_2);
res_1 = _shr2(_sadd2(res_1,0x00200020),6);
res_2 = _shr2(_sadd2(res_2,0x00200020),6);
res_1 = _spacku4(0x00000000,res_1);
res_2 = _spacku4(0x00000000,res_2);
_mem2(pucOutputCbPtr) = res_2;
pucOutputCbPtr += stride;
_mem2(pucOutputCbPtr) = res_1;
uiTmp1 = _packh4(uiE,uiF);
uiTmp2 = _packh4(uiG,uiH);
ui1_1 = _packh4(uiTmp1,uiTmp2);
ui2_1 = _packl4(uiTmp1,uiTmp2);
uiTmp1 = _packl4(uiE,uiF);
uiTmp2 = _packl4(uiG,uiH);
ui1_2 = _packh4(uiTmp1,uiTmp2);
ui2_2 = _packl4(uiTmp1,uiTmp2);
tmpend1_1 = _dotpu4(uicst,ui1_1);
tmpend1_2 = _dotpu4(uicst,ui1_2);
tmpend2_1 = _dotpu4(uicst,ui2_1);
tmpend2_2 = _dotpu4(uicst,ui2_2);
res_3 = _pack2(tmpend1_1,tmpend1_2);
res_4 = _pack2(tmpend2_1,tmpend2_2);
res_3 = _shr2(_sadd2(res_3,0x00200020),6);
res_4 = _shr2(_sadd2(res_4,0x00200020),6);
res_3 = _spacku4(0x00000000,res_3);
res_4 = _spacku4(0x00000000,res_4);
_mem2(pucOutputCrPtr) = res_4;
pucOutputCrPtr += stride;
_mem2(pucOutputCrPtr) = res_3;
}
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;
}
