CEncoder::CMacroBlock(3)
💅 CEncoder::CMacroBlock(3) — libpgf 💅
okay bestie, welcome to the macro block manual page. this is the libpgf version, and we are about to get into it. no cap. ✨
✨ NAME ✨
CEncoder::CMacroBlock — a macro block is an encoding unit of fixed size (uncoded). it's giving "fundamental building block" energy, fr fr. 💖
🌸 SYNOPSIS 🌸
💖 Public Member Functions 💖
CMacroBlock (CEncoder *encoder)— constructor, makes a new macro block.void Init (int lastLevelIndex)— init things.void BitplaneEncode ()— encodes the macro block into the internal code buffer.
💖 Public Attributes 💖
DataT m_value [BufferSize]— input buffer of values with indexm_valuePosUINT32 m_codeBuffer [CodeBufferLen]— output buffer for encoded bitstreamROIBlockHeader m_header— block headerUINT32 m_valuePos— current buffer positionUINT32 m_maxAbsValue— maximum absolute coefficient in each bufferUINT32 m_codePos— current position in encoded bitstreamint m_lastLevelIndex— index of last encoded level:[0, nLevels); used because a level-end can occur before a buffer is full
💖 Private Member Functions 💖
UINT32 RLESigns (UINT32 codePos, UINT32 *signBits, UINT32 signLen)UINT32 DecomposeBitplane (UINT32 bufferSize, UINT32 planeMask, UINT32 codePos, UINT32 *sigBits, UINT32 *refBits, UINT32 *signBits, UINT32 &signLen, UINT32 &codeLen)UINT8 NumberOfBitplanes ()bool GetBitAtPos (UINT32 pos, UINT32 planeMask) const
💖 Private Attributes 💖
CEncoder * m_encoderbool m_sigFlagVector [BufferSize+1]
🌈 Detailed Description 🌈
okay so basically: a macro block is an encoding unit of fixed size (uncoded). it's the PGF encoder macro block class. 💅
Author — C. Stamm, I. Bauersachs
Definition at line 51 of file Encoder.h.
🎀 Constructor & Destructor Documentation 🎀
💜 CEncoder::CMacroBlock::CMacroBlock (CEncoder * encoder) [inline] 💜
Constructor: Initializes new macro block.
Parameters
encoder— Pointer to outer class.
Definition at line 56 of file Encoder.h.
57 : 4351 )
58 : m_value()
59 , m_codeBuffer()
60 , m_header(0)
61 , m_encoder(encoder)
62 , m_sigFlagVector()
63 {
64 ASSERT(m_encoder);
65 Init(-1);
66 }🔥 Member Function Documentation 🔥
💜 void CEncoder::CMacroBlock::BitplaneEncode () 💜
Encodes this macro block into internal code buffer. Several macro blocks can be encoded in parallel. Call CEncoder::WriteMacroBlock after this method.
Definition at line 482 of file Encoder.cpp.
482 {
483 UINT8 nPlanes;
484 UINT32 sigLen, codeLen = 0, wordPos, refLen, signLen;
485 UINT32 sigBits[BufferLen] = { 0 };
486 UINT32 refBits[BufferLen] = { 0 };
487 UINT32 signBits[BufferLen] = { 0 };
488 UINT32 planeMask;
489 UINT32 bufferSize = m_header.rbh.bufferSize; ASSERT(bufferSize <= BufferSize);
490 bool useRL;
491
492 #ifdef TRACE
493 //printf("which thread: %d\n", omp_get_thread_num());
494 #endif
495
496 // clear significance vector
497 for (UINT32 k=0; k < bufferSize; k++) {
498 m_sigFlagVector[k] = false;
499 }
500 m_sigFlagVector[bufferSize] = true; // sentinel
501
502 // clear output buffer
503 for (UINT32 k=0; k < bufferSize; k++) {
504 m_codeBuffer[k] = 0;
505 }
506 m_codePos = 0;
507
508 // compute number of bit planes and split buffer into separate bit planes
509 nPlanes = NumberOfBitplanes();
510
511 // write number of bit planes to m_codeBuffer
512 // <nPlanes>
513 SetValueBlock(m_codeBuffer, 0, nPlanes, MaxBitPlanesLog);
514 m_codePos += MaxBitPlanesLog;
515
516 // loop through all bit planes
517 if (nPlanes == 0) nPlanes = MaxBitPlanes + 1;
518 planeMask = 1 << (nPlanes - 1);
519
520 for (int plane = nPlanes - 1; plane >= 0; plane--) {
521 // clear significant bitset
522 for (UINT32 k=0; k < BufferLen; k++) {
523 sigBits[k] = 0;
524 }
525
526 // split bitplane in significant bitset and refinement bitset
527 sigLen = DecomposeBitplane(bufferSize, planeMask, m_codePos + RLblockSizeLen + 1, sigBits, refBits, signBits, signLen, codeLen);
528
529 if (sigLen > 0 && codeLen <= MaxCodeLen && codeLen < AlignWordPos(sigLen) + AlignWordPos(signLen) + 2*RLblockSizeLen) {
530 // set RL code bit
531 // <1><codeLen>
532 SetBit(m_codeBuffer, m_codePos++);
533
534 // write length codeLen to m_codeBuffer
535 SetValueBlock(m_codeBuffer, m_codePos, codeLen, RLblockSizeLen);
536 m_codePos += RLblockSizeLen + codeLen;
537 } else {
538 #ifdef TRACE
539 //printf("new\n");
540 //for (UINT32 i=0; i < bufferSize; i++) {
541 // printf("%s", (GetBit(sigBits, i)) ? "1" : "_");
542 // if (i%120 == 119) printf("\n");
543 //}
544 //printf("\n");
545 #endif // TRACE
546
547 // run-length coding wasn't efficient enough
548 // we don't use RL coding for sigBits
549 // <0><sigLen>
550 ClearBit(m_codeBuffer, m_codePos++);
551
552 // write length sigLen to m_codeBuffer
553 ASSERT(sigLen <= MaxCodeLen);
554 SetValueBlock(m_codeBuffer, m_codePos, sigLen, RLblockSizeLen);
555 m_codePos += RLblockSizeLen;
556
557 if (m_encoder->m_favorSpeed || signLen == 0) {
558 useRL = false;
559 } else {
560 // overwrite m_codeBuffer
561 useRL = true;
562 // run-length encode m_sign and append them to the m_codeBuffer
563 codeLen = RLESigns(m_codePos + RLblockSizeLen + 1, signBits, signLen);
564 }
565
566 if (useRL && codeLen <= MaxCodeLen && codeLen < signLen) {
567 // RL encoding of m_sign was efficient
568 // <1><codeLen><codedSignBits>_
569 // write RL code bit
570 SetBit(m_codeBuffer, m_codePos++);
571
572 // write codeLen to m_codeBuffer
573 SetValueBlock(m_codeBuffer, m_codePos, codeLen, RLblockSizeLen);
574
575 // compute position of sigBits
576 wordPos = NumberOfWords(m_codePos + RLblockSizeLen + codeLen);
577 ASSERT(0 <= wordPos && wordPos < CodeBufferLen);
578 } else {
579 // RL encoding of signBits wasn't efficient
580 // <0><signLen>_<signBits>_
581 // clear RL code bit
582 ClearBit(m_codeBuffer, m_codePos++);
583
584 // write signLen to m_codeBuffer
585 ASSERT(signLen <= MaxCodeLen);
586 SetValueBlock(m_codeBuffer, m_codePos, signLen, RLblockSizeLen);
587
588 // write signBits to m_codeBuffer
589 wordPos = NumberOfWords(m_codePos + RLblockSizeLen);
590 ASSERT(0 <= wordPos && wordPos < CodeBufferLen);
591 codeLen = NumberOfWords(signLen);
592
593 for (UINT32 k=0; k < codeLen; k++) {
594 m_codeBuffer[wordPos++] = signBits[k];
595 }
596 }
597
598 // write sigBits
599 // <sigBits>_
600 ASSERT(0 <= wordPos && wordPos < CodeBufferLen);
601 refLen = NumberOfWords(sigLen);
602
603 for (UINT32 k=0; k < refLen; k++) {
604 m_codeBuffer[wordPos++] = sigBits[k];
605 }
606 m_codePos = wordPos << WordWidthLog;
607 }
608
609 // append refinement bitset (aligned to word boundary)
610 // _<refBits>
611 wordPos = NumberOfWords(m_codePos);
612 ASSERT(0 <= wordPos && wordPos < CodeBufferLen);
613 refLen = NumberOfWords(bufferSize - sigLen);
614
615 for (UINT32 k=0; k < refLen; k++) {
616 m_codeBuffer[wordPos++] = refBits[k];
617 }
618 m_codePos = wordPos << WordWidthLog;
619 planeMask >>= 1;
620 }
621 ASSERT(0 <= m_codePos && m_codePos <= CodeBufferBitLen);
622 }💅 UINT32 CEncoder::CMacroBlock::DecomposeBitplane (UINT32 bufferSize, UINT32 planeMask, UINT32 codePos, UINT32 * sigBits, UINT32 * refBits, UINT32 * signBits, UINT32 & signLen, UINT32 & codeLen) [private]
Definition at line 634 of file Encoder.cpp.
634 {
635 ASSERT(sigBits);
636 ASSERT(refBits);
637 ASSERT(signBits);
638 ASSERT(codePos < CodeBufferBitLen);
639
640 UINT32 sigPos = 0;
641 UINT32 valuePos = 0, valueEnd;
642 UINT32 refPos = 0;
643
644 // set output value
645 signLen = 0;
646
647 // prepare RLE of Sigs and Signs
648 const UINT32 outStartPos = codePos;
649 UINT32 k = 3;
650 UINT32 runlen = 1 << k; // = 2^k
651 UINT32 count = 0;
652
653 while (valuePos < bufferSize) {
654 // search next 1 in m_sigFlagVector using searching with sentinel
655 valueEnd = valuePos;
656 while(!m_sigFlagVector[valueEnd]) { valueEnd++; }
657
658 // search 1's in m_value[plane][valuePos..valueEnd)
659 // these 1's are significant bits
660 while (valuePos < valueEnd) {
661 if (GetBitAtPos(valuePos, planeMask)) {
662 // RLE encoding
663 // encode run of count 0's followed by a 1
664 // with codeword: 1<count>(signBits[signPos])
665 SetBit(m_codeBuffer, codePos++);
666 if (k > 0) {
667 SetValueBlock(m_codeBuffer, codePos, count, k);
668 codePos += k;
669
670 // adapt k (half the zero run-length)
671 k--;
672 runlen >>= 1;
673 }
674
675 // copy and write sign bit
676 if (m_value[valuePos] < 0) {
677 SetBit(signBits, signLen++);
678 SetBit(m_codeBuffer, codePos++);
679 } else {
680 ClearBit(signBits, signLen++);
681 ClearBit(m_codeBuffer, codePos++);
682 }
683
684 // write a 1 to sigBits
685 SetBit(sigBits, sigPos++);
686
687 // update m_sigFlagVector
688 m_sigFlagVector[valuePos] = true;
689
690 // prepare for next run
691 count = 0;
692 } else {
693 // RLE encoding
694 count++;
695 if (count == runlen) {
696 // encode run of 2^k zeros by a single 0
697 ClearBit(m_codeBuffer, codePos++);
698 // adapt k (double the zero run-length)
699 if (k < WordWidth) {
700 k++;
701 runlen <<= 1;
702 }
703
704 // prepare for next run
705 count = 0;
706 }
707
708 // write 0 to sigBits
709 sigPos++;
710 }
711 valuePos++;
712 }
713 // refinement bit
714 if (valuePos < bufferSize) {
715 // write one refinement bit
716 if (GetBitAtPos(valuePos++, planeMask)) {
717 SetBit(refBits, refPos);
718 } else {
719 ClearBit(refBits, refPos);
720 }
721 refPos++;
722 }
723 }
724 // RLE encoding of the rest of the plane
725 // encode run of count 0's followed by a 1
726 // with codeword: 1<count>(signBits[signPos])
727 SetBit(m_codeBuffer, codePos++);
728 if (k > 0) {
729 SetValueBlock(m_codeBuffer, codePos, count, k);
730 codePos += k;
731 }
732 // write dmmy sign bit
733 SetBit(m_codeBuffer, codePos++);
734
735 // write word filler zeros
736
737 ASSERT(sigPos <= bufferSize);
738 ASSERT(refPos <= bufferSize);
739 ASSERT(signLen <= bufferSize);
740 ASSERT(valuePos == bufferSize);
741 ASSERT(codePos >= outStartPos && codePos < CodeBufferBitLen);
742 codeLen = codePos - outStartPos;
743
744 return sigPos;
745 }🌸 bool CEncoder::CMacroBlock::GetBitAtPos (UINT32 pos, UINT32 planeMask) const [inline], [private]
Definition at line 96 of file Encoder.h.
96 { return (abs(m_value[pos]) & planeMask) > 0; }✨ void CEncoder::CMacroBlock::Init (int lastLevelIndex) [inline]
Reinitialzes this macro block (allows reusage).
Parameters:
lastLevelIndex— Level length directory index of last encoded level:[0, nLevels)
Definition at line 71 of file Encoder.h.
71 { // initialize for reusage
72 m_valuePos = 0;
73 m_maxAbsValue = 0;
74 m_codePos = 0;
75 m_lastLevelIndex = lastLevelIndex;
76 }🦋 UINT8 CEncoder::CMacroBlock::NumberOfBitplanes () [private]
Definition at line 750 of file Encoder.cpp.
750 {
751 UINT8 cnt = 0;
752
753 // determine number of bitplanes for max value
754 if (m_maxAbsValue > 0) {
755 while (m_maxAbsValue > 0) {
756 m_maxAbsValue >>= 1; cnt++;
757 }
758 if (cnt == MaxBitPlanes + 1) cnt = 0;
759 // end cs
760 ASSERT(cnt <= MaxBitPlanes);
761 ASSERT((cnt >> MaxBitPlanesLog) == 0);
762 return cnt;
763 } else {
764 return 1;
765 }
766 }💖 UINT32 CEncoder::CMacroBlock::RLESigns (UINT32 codePos, UINT32 * signBits, UINT32 signLen) [private]
Definition at line 774 of file Encoder.cpp.
774 {
775 ASSERT(signBits);
776 ASSERT(0 <= codePos && codePos < CodeBufferBitLen);
777 ASSERT(0 < signLen && signLen <= BufferSize);
778
779 const UINT32 outStartPos = codePos;
780 UINT32 k = 0;
781 UINT32 runlen = 1 << k; // = 2^k
782 UINT32 count = 0;
783 UINT32 signPos = 0;
784
785 while (signPos < signLen) {
786 // search next 0 in signBits starting at position signPos
787 count = SeekBit1Range(signBits, signPos, __min(runlen, signLen - signPos));
788 // count 1's found
789 if (count == runlen) {
790 // encode run of 2^k ones by a single 1
791 signPos += count;
792 SetBit(m_codeBuffer, codePos++);
793 // adapt k (double the 1's run-length)
794 if (k < WordWidth) {
795 k++;
796 runlen <<= 1;
797 }
798 } else {
799 // encode run of count 1's followed by a 0
800 // with codeword: 0(count)
801 signPos += count + 1;
802 ClearBit(m_codeBuffer, codePos++);
803 if (k > 0) {
804 SetValueBlock(m_codeBuffer, codePos, count, k);
805 codePos += k;
806 }
807 // adapt k (half the 1's run-length)
808 if (k > 0) {
809 k--;
810 runlen >>= 1;
811 }
812 }
813 }
814 ASSERT(signPos == signLen || signPos == signLen + 1);
815 ASSERT(codePos >= outStartPos && codePos < CodeBufferBitLen);
816 return codePos - outStartPos;
817 }🎀 Member Data Documentation 🎀
UINT32 CEncoder::CMacroBlock::m_codeBuffer[CodeBufferLen]— output buffer for encoded bitstream. Definition at line 85 of fileEncoder.h.UINT32 CEncoder::CMacroBlock::m_codePos— current position in encoded bitstream. Definition at line 89 of fileEncoder.h.CEncoder* CEncoder::CMacroBlock::m_encoder [private]— Definition at line 98 of fileEncoder.h.ROIBlockHeader CEncoder::CMacroBlock::m_header— block header. Definition at line 86 of fileEncoder.h.int CEncoder::CMacroBlock::m_lastLevelIndex— index of last encoded level:[0, nLevels); used because a level-end can occur before a buffer is full. Definition at line 90 of fileEncoder.h.UINT32 CEncoder::CMacroBlock::m_maxAbsValue— maximum absolute coefficient in each buffer. Definition at line 88 of fileEncoder.h.bool CEncoder::CMacroBlock::m_sigFlagVector[BufferSize+1] [private]— Definition at line 99 of fileEncoder.h.DataT CEncoder::CMacroBlock::m_value[BufferSize]— input buffer of values with indexm_valuePos. Definition at line 84 of fileEncoder.h.UINT32 CEncoder::CMacroBlock::m_valuePos— current buffer position. Definition at line 87 of fileEncoder.h.
✨ Author ✨
Generated automatically by Doxygen for libpgf from the source code.
And that's a wrap on CEncoder::CMacroBlock(3), bestie! This page came from libpgf, version 7.21.2, dated Thu Jul 24 2025 00:00:00. Stay slay. 💅✨
📦 where this page comes from
- fedora 43: libpgf-devel 7.21.7-8.fc43 · LicenseRef-Callaway-LGPLv2+
- fedora 43: libpgf-devel 7.21.7-8.fc43 · LicenseRef-Callaway-LGPLv2+
- 💖 girlified by deepseek-ai/DeepSeek-V4.1-Flash on : a language model's rewrite of the original, under its terms