diff options
author | James Cook <james@lindenlab.com> | 2007-01-02 08:33:20 +0000 |
---|---|---|
committer | James Cook <james@lindenlab.com> | 2007-01-02 08:33:20 +0000 |
commit | 420b91db29485df39fd6e724e782c449158811cb (patch) | |
tree | b471a94563af914d3ed3edd3e856d21cb1b69945 /indra/llmessage/patch_idct.cpp |
Print done when done.
Diffstat (limited to 'indra/llmessage/patch_idct.cpp')
-rw-r--r-- | indra/llmessage/patch_idct.cpp | 666 |
1 files changed, 666 insertions, 0 deletions
diff --git a/indra/llmessage/patch_idct.cpp b/indra/llmessage/patch_idct.cpp new file mode 100644 index 0000000000..cfc52c551d --- /dev/null +++ b/indra/llmessage/patch_idct.cpp @@ -0,0 +1,666 @@ +/** + * @file patch_idct.cpp + * @brief IDCT patch. + * + * Copyright (c) 2000-$CurrentYear$, Linden Research, Inc. + * $License$ + */ + +#include "linden_common.h" + +#include "llmath.h" +//#include "vmath.h" +#include "v3math.h" +#include "patch_dct.h" + +LLGroupHeader *gGOPP; + +void set_group_of_patch_header(LLGroupHeader *gopp) +{ + gGOPP = gopp; +} + +F32 gPatchDequantizeTable[LARGE_PATCH_SIZE*LARGE_PATCH_SIZE]; +void build_patch_dequantize_table(S32 size) +{ + S32 i, j; + for (j = 0; j < size; j++) + { + for (i = 0; i < size; i++) + { + gPatchDequantizeTable[j*size + i] = (1.f + 2.f*(i+j)); + } + } +} + +S32 gCurrentDeSize = 0; + +F32 gPatchICosines[LARGE_PATCH_SIZE*LARGE_PATCH_SIZE]; + +void setup_patch_icosines(S32 size) +{ + S32 n, u; + F32 oosob = F_PI*0.5f/size; + + for (u = 0; u < size; u++) + { + for (n = 0; n < size; n++) + { + gPatchICosines[u*size+n] = cosf((2.f*n+1.f)*u*oosob); + } + } +} + +S32 gDeCopyMatrix[LARGE_PATCH_SIZE*LARGE_PATCH_SIZE]; + +void build_decopy_matrix(S32 size) +{ + S32 i, j, count; + BOOL b_diag = FALSE; + BOOL b_right = TRUE; + + i = 0; + j = 0; + count = 0; + + while ( (i < size) + &&(j < size)) + { + gDeCopyMatrix[j*size + i] = count; + + count++; + + if (!b_diag) + { + if (b_right) + { + if (i < size - 1) + i++; + else + j++; + b_right = FALSE; + b_diag = TRUE; + } + else + { + if (j < size - 1) + j++; + else + i++; + b_right = TRUE; + b_diag = TRUE; + } + } + else + { + if (b_right) + { + i++; + j--; + if ( (i == size - 1) + ||(j == 0)) + { + b_diag = FALSE; + } + } + else + { + i--; + j++; + if ( (i == 0) + ||(j == size - 1)) + { + b_diag = FALSE; + } + } + } + } +} + +void init_patch_decompressor(S32 size) +{ + if (size != gCurrentDeSize) + { + gCurrentDeSize = size; + build_patch_dequantize_table(size); + setup_patch_icosines(size); + build_decopy_matrix(size); + } +} + +inline void idct_line(F32 *linein, F32 *lineout, S32 line) +{ + S32 n; + F32 total; + F32 *pcp = gPatchICosines; + +#ifdef _PATCH_SIZE_16_AND_32_ONLY + F32 oosob = 2.f/16.f; + S32 line_size = line*NORMAL_PATCH_SIZE; + F32 *tlinein, *tpcp; + + + for (n = 0; n < NORMAL_PATCH_SIZE; n++) + { + tpcp = pcp + n; + tlinein = linein + line_size; + + total = OO_SQRT2*(*(tlinein++)); + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein)*(*(tpcp += NORMAL_PATCH_SIZE)); + + *(lineout + line_size + n) = total*oosob; + } +#else + F32 oosob = 2.f/size; + S32 size = gGOPP->patch_size; + S32 line_size = line*size; + S32 u; + for (n = 0; n < size; n++) + { + total = OO_SQRT2*linein[line_size]; + for (u = 1; u < size; u++) + { + total += linein[line_size + u]*pcp[u*size+n]; + } + lineout[line_size + n] = total*oosob; + } +#endif +} + +inline void idct_line_large_slow(F32 *linein, F32 *lineout, S32 line) +{ + S32 n; + F32 total; + F32 *pcp = gPatchICosines; + + F32 oosob = 2.f/32.f; + S32 line_size = line*LARGE_PATCH_SIZE; + F32 *tlinein, *tpcp; + + + for (n = 0; n < LARGE_PATCH_SIZE; n++) + { + tpcp = pcp + n; + tlinein = linein + line_size; + + total = OO_SQRT2*(*(tlinein++)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein)*(*(tpcp += LARGE_PATCH_SIZE)); + + *(lineout + line_size + n) = total*oosob; + } +} + +// Nota Bene: assumes that coefficients beyond 128 are 0! + +void idct_line_large(F32 *linein, F32 *lineout, S32 line) +{ + S32 n; + F32 total; + F32 *pcp = gPatchICosines; + + F32 oosob = 2.f/32.f; + S32 line_size = line*LARGE_PATCH_SIZE; + F32 *tlinein, *tpcp; + F32 *baselinein = linein + line_size; + F32 *baselineout = lineout + line_size; + + + for (n = 0; n < LARGE_PATCH_SIZE; n++) + { + tpcp = pcp++; + tlinein = baselinein; + + total = OO_SQRT2*(*(tlinein++)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein++)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein)*(*(tpcp)); + + *baselineout++ = total*oosob; + } +} + +inline void idct_column(F32 *linein, F32 *lineout, S32 column) +{ + S32 n; + F32 total; + F32 *pcp = gPatchICosines; + +#ifdef _PATCH_SIZE_16_AND_32_ONLY + F32 *tlinein, *tpcp; + + for (n = 0; n < NORMAL_PATCH_SIZE; n++) + { + tpcp = pcp + n; + tlinein = linein + column; + + total = OO_SQRT2*(*tlinein); + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + total += *(tlinein += NORMAL_PATCH_SIZE)*(*(tpcp += NORMAL_PATCH_SIZE)); + + *(lineout + (n<<4) + column) = total; + } + +#else + S32 size = gGOPP->patch_size; + S32 u; + S32 u_size; + + for (n = 0; n < size; n++) + { + total = OO_SQRT2*linein[column]; + for (u = 1; u < size; u++) + { + u_size = u*size; + total += linein[u_size + column]*pcp[u_size+n]; + } + lineout[size*n + column] = total; + } +#endif +} + +inline void idct_column_large_slow(F32 *linein, F32 *lineout, S32 column) +{ + S32 n; + F32 total; + F32 *pcp = gPatchICosines; + + F32 *tlinein, *tpcp; + + for (n = 0; n < LARGE_PATCH_SIZE; n++) + { + tpcp = pcp + n; + tlinein = linein + column; + + total = OO_SQRT2*(*tlinein); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + + *(lineout + (n<<5) + column) = total; + } +} + +// Nota Bene: assumes that coefficients beyond 128 are 0! + +void idct_column_large(F32 *linein, F32 *lineout, S32 column) +{ + S32 n, m; + F32 total; + F32 *pcp = gPatchICosines; + + F32 *tlinein, *tpcp; + F32 *baselinein = linein + column; + F32 *baselineout = lineout + column; + + for (n = 0; n < LARGE_PATCH_SIZE; n++) + { + tpcp = pcp++; + tlinein = baselinein; + + total = OO_SQRT2*(*tlinein); + for (m = 1; m < NORMAL_PATCH_SIZE; m++) + total += *(tlinein += LARGE_PATCH_SIZE)*(*(tpcp += LARGE_PATCH_SIZE)); + + *(baselineout + (n<<5)) = total; + } +} + +inline void idct_patch(F32 *block) +{ + F32 temp[LARGE_PATCH_SIZE*LARGE_PATCH_SIZE]; + +#ifdef _PATCH_SIZE_16_AND_32_ONLY + idct_column(block, temp, 0); + idct_column(block, temp, 1); + idct_column(block, temp, 2); + idct_column(block, temp, 3); + + idct_column(block, temp, 4); + idct_column(block, temp, 5); + idct_column(block, temp, 6); + idct_column(block, temp, 7); + + idct_column(block, temp, 8); + idct_column(block, temp, 9); + idct_column(block, temp, 10); + idct_column(block, temp, 11); + + idct_column(block, temp, 12); + idct_column(block, temp, 13); + idct_column(block, temp, 14); + idct_column(block, temp, 15); + + idct_line(temp, block, 0); + idct_line(temp, block, 1); + idct_line(temp, block, 2); + idct_line(temp, block, 3); + + idct_line(temp, block, 4); + idct_line(temp, block, 5); + idct_line(temp, block, 6); + idct_line(temp, block, 7); + + idct_line(temp, block, 8); + idct_line(temp, block, 9); + idct_line(temp, block, 10); + idct_line(temp, block, 11); + + idct_line(temp, block, 12); + idct_line(temp, block, 13); + idct_line(temp, block, 14); + idct_line(temp, block, 15); +#else + S32 i; + S32 size = gGOPP->patch_size; + for (i = 0; i < size; i++) + { + idct_column(block, temp, i); + } + for (i = 0; i < size; i++) + { + idct_line(temp, block, i); + } +#endif +} + +inline void idct_patch_large(F32 *block) +{ + F32 temp[LARGE_PATCH_SIZE*LARGE_PATCH_SIZE]; + + idct_column_large_slow(block, temp, 0); + idct_column_large_slow(block, temp, 1); + idct_column_large_slow(block, temp, 2); + idct_column_large_slow(block, temp, 3); + + idct_column_large_slow(block, temp, 4); + idct_column_large_slow(block, temp, 5); + idct_column_large_slow(block, temp, 6); + idct_column_large_slow(block, temp, 7); + + idct_column_large_slow(block, temp, 8); + idct_column_large_slow(block, temp, 9); + idct_column_large_slow(block, temp, 10); + idct_column_large_slow(block, temp, 11); + + idct_column_large_slow(block, temp, 12); + idct_column_large_slow(block, temp, 13); + idct_column_large_slow(block, temp, 14); + idct_column_large_slow(block, temp, 15); + + idct_column_large_slow(block, temp, 16); + idct_column_large_slow(block, temp, 17); + idct_column_large_slow(block, temp, 18); + idct_column_large_slow(block, temp, 19); + + idct_column_large_slow(block, temp, 20); + idct_column_large_slow(block, temp, 21); + idct_column_large_slow(block, temp, 22); + idct_column_large_slow(block, temp, 23); + + idct_column_large_slow(block, temp, 24); + idct_column_large_slow(block, temp, 25); + idct_column_large_slow(block, temp, 26); + idct_column_large_slow(block, temp, 27); + + idct_column_large_slow(block, temp, 28); + idct_column_large_slow(block, temp, 29); + idct_column_large_slow(block, temp, 30); + idct_column_large_slow(block, temp, 31); + + idct_line_large_slow(temp, block, 0); + idct_line_large_slow(temp, block, 1); + idct_line_large_slow(temp, block, 2); + idct_line_large_slow(temp, block, 3); + + idct_line_large_slow(temp, block, 4); + idct_line_large_slow(temp, block, 5); + idct_line_large_slow(temp, block, 6); + idct_line_large_slow(temp, block, 7); + + idct_line_large_slow(temp, block, 8); + idct_line_large_slow(temp, block, 9); + idct_line_large_slow(temp, block, 10); + idct_line_large_slow(temp, block, 11); + + idct_line_large_slow(temp, block, 12); + idct_line_large_slow(temp, block, 13); + idct_line_large_slow(temp, block, 14); + idct_line_large_slow(temp, block, 15); + + idct_line_large_slow(temp, block, 16); + idct_line_large_slow(temp, block, 17); + idct_line_large_slow(temp, block, 18); + idct_line_large_slow(temp, block, 19); + + idct_line_large_slow(temp, block, 20); + idct_line_large_slow(temp, block, 21); + idct_line_large_slow(temp, block, 22); + idct_line_large_slow(temp, block, 23); + + idct_line_large_slow(temp, block, 24); + idct_line_large_slow(temp, block, 25); + idct_line_large_slow(temp, block, 26); + idct_line_large_slow(temp, block, 27); + + idct_line_large_slow(temp, block, 28); + idct_line_large_slow(temp, block, 29); + idct_line_large_slow(temp, block, 30); + idct_line_large_slow(temp, block, 31); +} + +S32 gDitherNoise = 128; + +void decompress_patch(F32 *patch, S32 *cpatch, LLPatchHeader *ph) +{ + S32 i, j; + + F32 block[LARGE_PATCH_SIZE*LARGE_PATCH_SIZE], *tblock = block; + F32 *tpatch; + + LLGroupHeader *gopp = gGOPP; + S32 size = gopp->patch_size; + F32 range = ph->range; + S32 prequant = (ph->quant_wbits >> 4) + 2; + S32 quantize = 1<<prequant; + F32 hmin = ph->dc_offset; + S32 stride = gopp->stride; + + F32 ooq = 1.f/(F32)quantize; + F32 *dq = gPatchDequantizeTable; + S32 *decopy_matrix = gDeCopyMatrix; + + F32 mult = ooq*range; + F32 addval = mult*(F32)(1<<(prequant - 1))+hmin; + + for (i = 0; i < size*size; i++) + { + *(tblock++) = *(cpatch + *(decopy_matrix++))*(*dq++); + } + + if (size == 16) + { + idct_patch(block); + } + else + { + idct_patch_large(block); + } + + for (j = 0; j < size; j++) + { + tpatch = patch + j*stride; + tblock = block + j*size; + for (i = 0; i < size; i++) + { + *(tpatch++) = *(tblock++)*mult+addval; + } + } +} + + +void decompress_patchv(LLVector3 *v, S32 *cpatch, LLPatchHeader *ph) +{ + S32 i, j; + + F32 block[LARGE_PATCH_SIZE*LARGE_PATCH_SIZE], *tblock = block; + LLVector3 *tvec; + + LLGroupHeader *gopp = gGOPP; + S32 size = gopp->patch_size; + F32 range = ph->range; + S32 prequant = (ph->quant_wbits >> 4) + 2; + S32 quantize = 1<<prequant; + F32 hmin = ph->dc_offset; + S32 stride = gopp->stride; + + F32 ooq = 1.f/(F32)quantize; + F32 *dq = gPatchDequantizeTable; + S32 *decopy_matrix = gDeCopyMatrix; + + F32 mult = ooq*range; + F32 addval = mult*(F32)(1<<(prequant - 1))+hmin; + +// BOOL b_diag = FALSE; +// BOOL b_right = TRUE; + + for (i = 0; i < size*size; i++) + { + *(tblock++) = *(cpatch + *(decopy_matrix++))*(*dq++); + } + + if (size == 16) + idct_patch(block); + else + idct_patch_large(block); + + for (j = 0; j < size; j++) + { + tvec = v + j*stride; + tblock = block + j*size; + for (i = 0; i < size; i++) + { + (*tvec++).mV[VZ] = *(tblock++)*mult+addval; + } + } +} + |