void lab4check(char* folderPath, char* testName) {
LARGE_INTEGER frequency;
LARGE_INTEGER t1, t2;
QueryPerformanceFrequency(&frequency);
QueryPerformanceCounter(&t1);
std::string path(folderPath);
std::string fileName;
printf(" read %s\n", testName);
CCharBuffer text(fileName.assign(path).append(testName).c_str());
printf(" check %s:\n", testName);
CCharBuffer patterns[4*mLen*3*tests];
GeneratePatterns(text, patterns);
CSuffixArray sArr(text);
const unsigned int cc = tests, bc = 3*tests, ac = mLen*bc, kt[4] = { 1, 1, 3, 1 }, taskMask[4] = { 1, 0, 0, 0 };
for(unsigned int i=0; i<mLen; i++) {
printf(" length:\t%u\n", m[i]);
printf(" suffix tree:");
for(unsigned int k=0; k<4; k++) if((taskMask[k] != 0) &&((k!=3) || (i<3))) {
QueryPerformanceCounter(&t1);
for(unsigned int l=0; l<kt[k]; l++)
for(unsigned int j=0; j<tests; j++)
binarySearch(sArr, patterns[k*ac+i*bc+l*cc+j]);
QueryPerformanceCounter(&t2);
printf("\t%.2lf ms", (t2.QuadPart - t1.QuadPart) * 1000.0 / frequency.QuadPart);
}
printf("\n");
}
printf("\n");
}
void ff(TextureQuantizeRAW,format,A)(RESOURCEINFO &texo, RESOURCEINFO &texd, ULONG *texs, ULONG *texr, int level, int l) {
/* square dimension of this surface-level */
/* square area of this surface-level */
const int lv = (1 << l);
const int av = lv * lv;
/* ------------------------------------------------------------------------------------------------------- */
const int NORMALS_SCALEBYLEVEL = ::NORMALS_SCALEBYLEVEL;
const int ALPHAS_SCALEBYLEVEL = ::ALPHAS_SCALEBYLEVEL;
const float colorgamma = ::colorgamma;
const float alphacontrast = ::alphacontrast;
const float colorgammainv = ::colorgammainv;
const float alphacontrastinv = ::alphacontrastinv;
int iwidth = texo.Width;
int iheight = texo.Height;
int owidth = texd.Width;
int oheight = texd.Height;
int cwidth = owidth;
int cheight = oheight;
/* get the data back to the CPU */
#if (TCOMPRESS_CHANNELS(format) == 4)
/* ABGR -> ARGB */ cwidth = (cwidth + 0) >> 0; /* 1x LONG to 1x LONG */
#elif (TCOMPRESS_CHANNELS(format) == 3)
/* -BGR -> -RGB */ cwidth = (cwidth + 1) >> 1; /* 1x LONG to 1x SHORT */
#elif (TCOMPRESS_CHANNELS(format) == 2)
/* LA-- -> AL-- */ cwidth = (cwidth + 3) >> 2; /* 1x LONG to 1x CHAR */
#elif (TCOMPRESS_CHANNELS(format) == 1)
/* A--- -> A--- */ cwidth = (cwidth + 31) >> 5; /* 8x LONG to 1x CHAR */
#else
#error
#endif
/* ensure tile ability (bit on overhead for non-4 resolutions) */
owidth = (owidth + (TX - 1)) & (~(TX - 1));
oheight = (oheight + (TY - 1)) & (~(TY - 1));
assert((owidth & (TX - 1)) == 0);
assert((oheight & (TY - 1)) == 0);
#if defined(SQUASH_USE_AMP) && !defined(SQUASH_USE_AMP_DEBUG)
/* get a two-dimensional extend over the whole output (without re-cast to LONG),
* then get a tile-extend over that one ()
*/
Concurrency::extent<2> ee(oheight, owidth);
Concurrency::tiled_extent<TY, TX> te(ee);
Concurrency::array_view<const unsigned int, 2> sArr(iheight, iwidth, (const unsigned int *)texs);
Concurrency::array_view< unsigned int, 2> dArr(cheight, cwidth, ( unsigned int *)texr);
Concurrency::parallel_for_each(te /*dArr.extent.tile<TY, TX>(osize)*/, [=](tiled_index<TY, TX> elm) restrict(amp) {
typedef type accu[DIM];
/* tile static memory */
// tile_static UTYPE bTex[2][TY][TX];
tile_static int bTex[2][TY][TX];
tile_static type fTex[2][TY][TX][DIM];
// const int y = elm.global[0] - ly;
// const int x = elm.global[1] - lx;
const int y = elm.tile[0] * TY;
const int x = elm.tile[1] * TX;
const int ly = elm.local[0];
const int lx = elm.local[1];
#else
array_view<const unsigned int, 2> sArr(iheight, iwidth, (const unsigned int *)texs);
array_view< unsigned int, 2> dArr(cheight, cwidth, ( unsigned int *)texr, true);
for (int groupsy = 0; groupsy < (owidth / TY); groupsy++)
for (int groupsx = 0; groupsx < (oheight / TX); groupsx++) {
typedef type accu[DIM];
/* tile static memory */
// UTYPE bTex[2][TY][TX];
int bTex[2][TY][TX];
type fTex[2][TY][TX][DIM];
for (int tiley = 0; tiley < TY; tiley++)
for (int tilex = 0; tilex < TX; tilex++)
{
const int y = groupsy * TY;
const int x = groupsx * TX;
const int ly = tiley;
const int lx = tilex;
#endif
/* generate this level's 4x4-block from the original surface */
{
const int yl = ((y + ly) << l);
const int xl = ((x + lx) << l);
accu tt; tt[0] = tt[1] = tt[2] = tt[3] = tt[4] = tt[5] = tt[6] = tt[7] = 0;
/* access all pixels this level's 4x4-block represents in
* the full dimension original surface (high quality mip-mapping)
*/
for (int oy = 0; oy < lv; oy += 1)
for (int ox = 0; ox < lv; ox += 1) {
/* assume seamless tiling: wrap pixels around */
const int posx = (xl + ox) % iwidth;
const int posy = (yl + oy) % iheight;
const ULONG &t = sArr(posy, posx);
Accu(tt, t); // +=
}
/* build average of each channel */
Norm(fTex[0][ly][lx], tt, av, level, l);
}
#if defined(SQUASH_USE_AMP) && !defined(SQUASH_USE_AMP_DEBUG)
tile_static accu tr; tr[(ly * TX + lx) & 7] = 0;
tile_static_memory_fence(elm.barrier);
// elm.barrier.wait_with_tile_static_memory_fence();
#else
}
accu tr = {0};
#endif
/* runs on only 1 thread per tile (reduction) */
#if defined(SQUASH_USE_AMP) && !defined(SQUASH_USE_AMP_DEBUG)
if (elm.local == index<2>(0, 0))
#endif
{
/* analyze this level's 4x4-block */
for (int ly = 0; ly < TY; ly += 1)
for (int lx = 0; lx < TX; lx += 1) {
Look(fTex[0][ly][lx], tr);
}
}
#if defined(SQUASH_USE_AMP) && !defined(SQUASH_USE_AMP_DEBUG)
tile_static_memory_fence(elm.barrier);
// elm.barrier.wait_with_tile_static_memory_fence();
#else
for (int tiley = 0; tiley < TY; tiley++)
for (int tilex = 0; tilex < TX; tilex++)
{
const int y = groupsy;
const int x = groupsx;
const int ly = tiley;
const int lx = tilex;
#endif
/* generate this level's 4x4-block from the original surface */
{
/* build average of each channel an join */
ULONG t;
Code(fTex[0][ly][lx], tr, (A > 2 ? 4 : (A > 1 ? 10 : (A > 0 ? 5 : 6)))); t =
Qunt(fTex[0][ly][lx], tr, (A > 2 ? 4 : (A > 1 ? 10 : (A > 0 ? 5 : 6))));
/* write the result */
/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
#if (TCOMPRESS_CHANNELS(format) == 4)
/* ABGR -> RGBA */
bTex[0][ly][lx] = t;
/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
#elif (TCOMPRESS_CHANNELS(format) == 3)
/* -BGR -> RGB- */
bTex[0][ly][lx] = t;
/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
#elif (TCOMPRESS_CHANNELS(format) == 2)
/* AL-- -> LA-- */
bTex[0][ly][lx] = t;
#else
#error
#endif
}
/* put this level's 4x4-block into the destination surface */
{
/* assume seamless tiling: wrap pixels around */
const int posx = (x + lx) % owidth;
const int posy = (y + ly) % oheight;
/* convert unaligned output location to "int"-space output location */
const int linear = ((posy * owidth) + posx) * 1;
const int lposx = (linear << 0) % (cwidth << 0);
const int lposy = (linear << 0) / (cwidth << 0);
/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
#if (TCOMPRESS_CHANNELS(format) <= 4)
/* ABGR -> ARGB */
if (sizeof(UTYPE) == 4) {
/* every single thread */
{
int t0 = bTex[0][ly][lx + 0];
/* write combining */
unsigned int val = (ULONG)t0;
/* write out all of an "int" */
dArr(lposy, lposx) = val;
}
}
/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
#elif (TCOMPRESS_CHANNELS(format) <= 3)
/* -BGR -> -RGB */
if (sizeof(UTYPE) == 2) {
/* every second thread */
if (!(elm.local[1] & 1)) {
int t0 = bTex[0][ly][lx + 0];
int t1 = bTex[0][ly][lx + 1];
/* write combining */
unsigned int val = (ULONG)((t1 << 16) + (t0 << 0));
/* write out all of an "int" */
dArr(lposy, lposx) = val;
}
}
/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
#elif (TCOMPRESS_CHANNELS(format) <= 2)
/* --YX -> XY-- */
/* LA-- -> AL-- */
if (sizeof(UTYPE) == 1) {
/* every fourth thread */
if (!(elm.local[1] & 3)) {
int t0 = bTex[0][ly][lx + 0];
int t1 = bTex[0][ly][lx + 1];
int t2 = bTex[0][ly][lx + 2];
int t3 = bTex[0][ly][lx + 3];
/* write combining */
unsigned int val = (ULONG)((t3 << 24) + (t2 << 16) + (t1 << 8) + (t0 << 0));
/* write out all of an "int" */
dArr(lposy, lposx) = val;
}
}
#else
#error
#endif
}
// dTex += 0;
#if defined(SQUASH_USE_AMP) && !defined(SQUASH_USE_AMP_DEBUG)
});
dArr.synchronize();
#else
}}
void ff(TextureCompressDXT,format,coding,fiting)(RESOURCEINFO &texo, RESOURCEINFO &texd, ULONG *texs, ULONG *texr, int level, int l, int blocksize, int flags) {
/* square dimension of this surface-level */
/* square area of this surface-level */
const int lv = (1 << l);
const int av = lv * lv;
/* ------------------------------------------------------------------------------------------------------- */
const int NORMALS_SCALEBYLEVEL = ::NORMALS_SCALEBYLEVEL;
const int ALPHAS_SCALEBYLEVEL = ::ALPHAS_SCALEBYLEVEL;
const float colorgamma = ::colorgamma;
const float alphacontrast = ::alphacontrast;
const float colorgammainv = ::colorgammainv;
const float alphacontrastinv = ::alphacontrastinv;
int iwidth = texo.Width;
int iheight = texo.Height;
int owidth = texd.Width;
int oheight = texd.Height;
int cwidth = owidth;
int cheight = oheight;
/* get the data back to the CPU */
cheight = (cheight + 3) / 4; /* 4x4 LONG ... */
cwidth = (cwidth + 3) / 4; /* 4x4 LONG ... */
cwidth *= 2 * blocksize; /* ... to 2x|4x LONG */
/* ensure tile ability (bit on overhead for non-4 resolutions) */
owidth = (owidth + (TX - 1)) & (~(TX - 1));
oheight = (oheight + (TY - 1)) & (~(TY - 1));
assert((owidth & (TX - 1)) == 0);
assert((oheight & (TY - 1)) == 0);
#if defined(SQUASH_USE_AMP) && !defined(SQUASH_USE_AMP_DEBUG)
/* constant buffer array */
Concurrency::array_view<const SingleColourLookup_CCR, 2> lArr(2, 256, (const SingleColourLookup_CCR *)::lookup_34_56_ccr);
Concurrency::array_view<const IndexBlockLookup_CCR, 2> yArr(4, 8, (const IndexBlockLookup_CCR *)::lookup_c34a57_ccr);
/* get a two-dimensional extend over the whole output (without re-cast to LONG),
* then get a tile-extend over that one ()
*/
Concurrency::extent<2> ee(oheight, owidth);
Concurrency::tiled_extent<TY, TX> te(ee);
Concurrency::array_view<const unsigned int, 2> sArr(iheight, iwidth, (const unsigned int *)texs);
Concurrency::array_view< unsigned int, 2> dArr(cheight, cwidth, ( unsigned int *)texr);
Concurrency::parallel_for_each(te /*dArr.extent.tile<TY, TX>(osize)*/, [=](tiled_index<TY, TX> elm) restrict(amp) {
typedef type accu[DIM];
/* tile static memory */
// tile_static UTYPE bTex[2][TY*TX];
tile_static type fTex[2][TY*TX][DIM];
tile_static int iTex[2][TY*TX][DIM];
/* generate this level's 4x4-block from the original surface */
// const int y = elm.global[0] - ly;
// const int x = elm.global[1] - lx;
const int y = elm.tile[0] * TY;
const int x = elm.tile[1] * TX;
const int ly = elm.local[0];
const int lx = elm.local[1];
const int lxy = ly * TX + lx;
#else
Concurrency::array_view<const SingleColourLookup_CCR, 2> lArr(2, 256, (const SingleColourLookup_CCR *)::lookup_34_56_ccr);
Concurrency::array_view<const unsigned int, 2> sArr(iheight, iwidth, (const unsigned int *)texs);
Concurrency::array_view< unsigned int, 2> dArr(cheight, cwidth, ( unsigned int *)texr, true);
for (int groupsy = 0; groupsy < (owidth / TY); groupsy++)
for (int groupsx = 0; groupsx < (oheight / TX); groupsx++) {
typedef type accu[DIM];
/* tile static memory */
// UTYPE bTex[2][TY*TX];
type fTex[2][TY*TX][DIM];
int iTex[2][TY*TX][DIM];
for (int tiley = 0; tiley < TY; tiley++)
for (int tilex = 0; tilex < TX; tilex++)
{
const int y = groupsy * TY;
const int x = groupsx * TX;
const int ly = tiley;
const int lx = tilex;
const int lxy = ly * TX + lx;
#endif
{
const int yl = ((y + ly) << l);
const int xl = ((x + lx) << l);
accu tt = {0};
/* access all pixels this level's 4x4-block represents in
* the full dimension original surface (high quality mip-mapping)
*/
for (int oy = 0; oy < lv; oy += 1) {
for (int ox = 0; ox < lv; ox += 1) {
/* assume seamless tiling: wrap pixels around */
const int posx = (xl + ox) % iwidth;
const int posy = (yl + oy) % iheight;
const ULONG &t = sArr(posy, posx);
Accu(tt, t); // +=
}
}
/* build average of each channel */
Norm(fTex[0][lxy], tt, av, level, l);
}
#if defined(SQUASH_USE_AMP) && !defined(SQUASH_USE_AMP_DEBUG)
tile_static accu tr; tr[lxy & 7] = 0;
tile_static_memory_fence(elm.barrier);
// elm.barrier.wait_with_tile_static_memory_fence();
#else
}
accu tr = {0};
#endif
/* runs on only 1 thread per tile (reduction) */
#if defined(SQUASH_USE_AMP) && !defined(SQUASH_USE_AMP_DEBUG)
if (elm.local == index<2>(0, 0))
#endif
{
/* analyze this level's 4x4-block */
for (int lxy = 0; lxy < TY*TX; lxy += 1) {
Look(fTex[0][lxy], tr);
}
}
#if defined(SQUASH_USE_AMP) && !defined(SQUASH_USE_AMP_DEBUG)
tile_static_memory_fence(elm.barrier);
// elm.barrier.wait_with_tile_static_memory_fence();
#else
for (int tiley = 0; tiley < TY; tiley++)
for (int tilex = 0; tilex < TX; tilex++)
{
const int y = groupsy;
const int x = groupsx;
const int ly = tiley;
const int lx = tilex;
const int lxy = ly * TX + lx;
#endif
/* generate this level's 4x4-block from the original surface */
{
/* build average of each channel an join */
Code (fTex[0][lxy], tr,
(TCOMPRESS_CHANNELS(format) +
(TCOMPRESS_GREYS (format) ? 2 : 0)) == 2 ? 8 :
(TCOMPRESS_SWIZZL (format) ? 6 : 5));
Range(iTex[0][lxy],
fTex[0][lxy]);
#if (TCOMPRESS_SWIZZL(format))
/* swizzle ABGR -> AGBR */
{
int swap = iTex[0][lxy][1];
iTex[0][lxy][1] = iTex[0][lxy][2];
iTex[0][lxy][2] = swap ;
}
#endif
/* write the result */
/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
#if ((TCOMPRESS_CHANNELS(format) + (TCOMPRESS_GREYS(format) ? 2 : 0)) == 4)
/* ABGR -> RGBA */
/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
#elif ((TCOMPRESS_CHANNELS(format) + (TCOMPRESS_GREYS(format) ? 2 : 0)) == 3)
/* -BGR -> RGB- */
iTex[0][lxy][0] = 0xFF;
/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
#elif ((TCOMPRESS_CHANNELS(format) + (TCOMPRESS_GREYS(format) ? 2 : 0)) == 2)
/* --YX -> XY-- */
/* AL-- -> LA-- */
#if (format == TCOMPRESS_XYz)
iTex[0][lxy][0] = iTex[0][lxy][2], // Y
iTex[1][lxy][0] = iTex[0][lxy][3]; // X
#else
iTex[0][lxy][0] = iTex[0][lxy][0], // A
iTex[1][lxy][0] = iTex[0][lxy][1]; // Z
#endif
/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
#elif ((TCOMPRESS_CHANNELS(format) + (TCOMPRESS_GREYS(format) ? 2 : 0)) == 1)
/* -Z-- -> Z--- */
/* A--- -> A--- */
/* -LLL -> L--- */
#if (format == TCOMPRESS_a )
iTex[0][lxy][0] = iTex[0][lxy][0]; // A
#elif (format == TCOMPRESS_A )
iTex[0][lxy][0] = iTex[0][lxy][0]; // A
#elif (format == TCOMPRESS_xyZ)
iTex[0][lxy][0] = iTex[0][lxy][1]; // Z
#else
iTex[0][lxy][0] = iTex[0][lxy][3]; // X
#endif
#else
#error
#endif
}
#if defined(SQUASH_USE_AMP) && !defined(SQUASH_USE_AMP_DEBUG)
tile_static_memory_fence(elm.barrier);
// elm.barrier.wait_with_tile_static_memory_fence();
#define local_is(a,b) elm.local == index<2>(a, b)
#else
}
for (int tiley = 0; tiley < TY; tiley++)
for (int tilex = 0; tilex < TX; tilex++)
{
const int y = groupsy;
const int x = groupsx;
const int ly = tiley;
const int lx = tilex;
const int lxy = ly * TX + lx;
#define local_is(a,b) ((ly == a) && (lx == b))
#endif
/* put this level's 4x4-block into the destination surface */
{
/* round down */
int posx = (x + lx) >> 2;
int posy = (y + ly) >> 2;
/* first and second block */
unsigned int b[2][2];
/* compress to DXT1/DXT3/DXT5/ATI1/ATI2 */
#define sflgs TCOMPRESS_COLOR(format) ? SQUISH_METRIC_PERCEPTUAL : SQUISH_METRIC_UNIFORM, \
TCOMPRESS_TRANS(format), \
fiting
/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
#if ((TCOMPRESS_CHANNELS(format) + (TCOMPRESS_GREYS(format) ? 2 : 0)) == 4) || \
((TCOMPRESS_CHANNELS(format) + (TCOMPRESS_GREYS(format) ? 2 : 0)) == 3)
/* 1x LONG per block for DXT1, 2x for the others */
#if (coding == 1)
{ posx <<= 0;
squish::CompressColorBtc1(elm.barrier, lxy, iTex[0], 0xFFFF, b[1], sflgs, yArr, lArr);
dArr(posy, posx + 0) = b[1][0];
dArr(posy, posx + 1) = b[1][1];
}
#elif (coding == 2)
{ posx <<= 1;
squish::CompressAlphaBtc2(elm.barrier, lxy, iTex[0], 0xFFFF, b[0] , yArr );
squish::CompressColorBtc2(elm.barrier, lxy, iTex[0], 0xFFFF, b[1], sflgs, yArr, lArr);
dArr(posy, posx + 0) = b[0][0];
dArr(posy, posx + 1) = b[0][1];
dArr(posy, posx + 2) = b[1][0];
dArr(posy, posx + 3) = b[1][1];
}
#elif (coding == 3)
{ posx <<= 1;
squish::CompressAlphaBtc3(elm.barrier, lxy, iTex[0], 0xFFFF, b[0] , yArr );
squish::CompressColorBtc3(elm.barrier, lxy, iTex[0], 0xFFFF, b[1], sflgs, yArr, lArr);
dArr(posy, posx + 0) = b[0][0];
dArr(posy, posx + 1) = b[0][1];
dArr(posy, posx + 2) = b[1][0];
dArr(posy, posx + 3) = b[1][1];
}
#else
#error
#endif
/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
#elif ((TCOMPRESS_CHANNELS(format) + (TCOMPRESS_GREYS(format) ? 2 : 0)) == 1)
/* 1x LONG for ATI1 */
#if (coding == 4)
{ posx <<= 0;
squish::CompressAlphaBtc3(elm.barrier, lxy, iTex[0], 0xFFFF, b[0] , yArr );
dArr(posy, posx + 0) = b[0][0];
dArr(posy, posx + 1) = b[0][1];
}
#else
#error
#endif
/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
#elif ((TCOMPRESS_CHANNELS(format) + (TCOMPRESS_GREYS(format) ? 2 : 0)) == 2)
/* 2x LONG for ATI2 */
#if (coding == 5)
{ posx <<= 1;
squish::CompressAlphaBtc3(elm.barrier, lxy, iTex[0], 0xFFFF, b[0] , yArr );
squish::CompressAlphaBtc3(elm.barrier, lxy, iTex[1], 0xFFFF, b[1] , yArr );
dArr(posy, posx + 0) = b[0][0];
dArr(posy, posx + 1) = b[0][1];
dArr(posy, posx + 2) = b[1][0];
dArr(posy, posx + 3) = b[1][1];
}
#else
#error
#endif
#else
#error
#endif
#undef sflgs
// elm.barrier.wait();
/* advance pointer of compressed blocks */
// wTex += blocksize;
// dTex += blocksize;
}
#if defined(SQUASH_USE_AMP) && !defined(SQUASH_USE_AMP_DEBUG)
// elm.barrier.wait();
// dTex += 0;
});
dArr.synchronize();
#else
}}
