FORCE_INLINE
int __ext_viterbi_brick_init_fast(int32 frame_len, int16 code_r, int16 depth) {
ob_count = 0;
m_viterbi.Reset();
frame_length = frame_len;
code_rate = code_r;
if (m_outbuf != NULL) free((void *)m_outbuf);
TRELLIS_DEPTH = (size_t) depth;
m_outbuf = (unsigned char*) malloc((size_t) (TRELLIS_DEPTH + 1));
if (m_outbuf == NULL)
{
printf("Viterbi memory allocation failure!\n");
exit(1);
}
return 0;
}
FORCE_INLINE
int16 __ext_viterbi_brick_decode_fast(num8 * intInput, int len1, unsigned char* bit, int len2)
{
static const int trellis_prefix = 6; // 6 bit zero prefix
static const int trellis_lookahead = 24;
//uchar m_outbuf[TRELLIS_DEPTH / 8 + 1];
unsigned int total_byte_count = 0;
// vector128 constants
#if defined(__GNUC__) && !defined(__ARM_NEON__)
static const __m128i * const pVITMA = VIT_MA; // Branch Metric A
static const __m128i * const pVITMB = VIT_MB; // Branch Metric B
#else
static const vub * const pVITMA = (const vub*)VIT_MA; // Branch Metric A
static const vub * const pVITMB = (const vub*)VIT_MB; // Branch Metric B
#endif
//uchar* input = ipin.peek();
unsigned char* input = (unsigned char*)intInput;
//uchar* input_end = input + NUM_INPUT;
unsigned char* input_end = input + len1;
while (input < input_end) {
// advance branch
if (code_rate == CR_12)
{
m_viterbi.BranchACS(pVITMA, input[0], pVITMB, input[1]);
input += 2; // jump to data
}
else if (code_rate == CR_34)
{
m_viterbi.BranchACS(pVITMA, input[0], pVITMB, input[1]);
m_viterbi.BranchACS(pVITMA, input[2]);
m_viterbi.BranchACS(pVITMB, input[3]);
input += 4;
}
else if (code_rate == CR_23)
{
m_viterbi.BranchACS(pVITMA, input[0], pVITMB, input[1]);
m_viterbi.BranchACS(pVITMA, input[2]);
input += 3;
}
unsigned int tr_index = m_viterbi.trellis_index();
if ((tr_index & 7) == 0) {
m_viterbi.Normalize();
}
// check trace_back
unsigned int output_count = 0;
unsigned int lookahead = 0;
const unsigned int tr_index_end = frame_length * 8 + trellis_prefix;
if (tr_index >= tr_index_end) {
// all bytes have been processed - track back all of them
output_count = tr_index_end - ob_count
- trellis_prefix;
lookahead = tr_index - tr_index_end;
}
else if (tr_index >= ob_count + TRELLIS_DEPTH + trellis_lookahead + trellis_prefix)
{
// trace back partially
// Note: tr_index increase during 'BranchACS', to decode out complete bytes, we may lookahead the
// trellis without output a few more than 'trellis_lookahead'
unsigned int remain = (tr_index - (ob_count + TRELLIS_DEPTH + trellis_lookahead + trellis_prefix)) % 8;
output_count = TRELLIS_DEPTH;
lookahead = trellis_lookahead + remain;
}
if (output_count) {
m_viterbi.Traceback((num8 *)m_outbuf, output_count, lookahead);
ob_count += output_count;
unsigned int last_byte_count = 0;
while (last_byte_count < output_count / 8) {
bit[total_byte_count] = m_outbuf[last_byte_count];
last_byte_count++;
total_byte_count++;
}
}
}
return total_byte_count * 8;
}
int16 __ext_viterbi_brick_decode_fast(char* intInput, int len1, uchar* bit, int len2)
{
static const int trellis_prefix = 6; // 6 bit zero prefix
static const int trellis_lookahead = 24;
//uchar m_outbuf[TRELLIS_DEPTH / 8 + 1];
uint ii = 0;
// vector128 constants
static const vub * const pVITMA = (const vub*)VIT_MA; // Branch Metric A
static const vub * const pVITMB = (const vub*)VIT_MB; // Branch Metric B
//uchar* input = ipin.peek();
uchar* input = (uchar*)intInput;
//uchar* input_end = input + NUM_INPUT;
uchar* input_end = input + len1;
while (input < input_end) {
// advance branch
if (code_rate == CR_12)
{
m_viterbi.BranchACS(pVITMA, input[0], pVITMB, input[1]);
input += 2; // jump to data
}
else if (code_rate == CR_34)
{
m_viterbi.BranchACS(pVITMA, input[0], pVITMB, input[1]);
m_viterbi.BranchACS(pVITMA, input[2]);
m_viterbi.BranchACS(pVITMB, input[3]);
input += 4;
}
else if (code_rate == CR_23)
{
m_viterbi.BranchACS(pVITMA, input[0], pVITMB, input[1]);
m_viterbi.BranchACS(pVITMA, input[2]);
input += 3;
}
uint tr_index = m_viterbi.trellis_index();
if ((tr_index & 7) == 0) {
m_viterbi.Normalize();
}
// check trace_back
uint output_count = 0;
uint lookahead = 0;
const uint tr_index_end = frame_length * 8 + 16 + trellis_prefix;
if (tr_index >= tr_index_end) {
// all bytes have been processed - track back all of them
output_count = tr_index_end - ob_count
- trellis_prefix;
lookahead = tr_index - tr_index_end;
}
else if (tr_index >= ob_count + TRELLIS_DEPTH + trellis_lookahead + trellis_prefix)
{
// trace back partially
// Note: tr_index increase during 'BranchACS', to decode out complete bytes, we may lookahead the
// trellis without output a few more than 'trellis_lookahead'
uint remain = (tr_index - (ob_count + TRELLIS_DEPTH + trellis_lookahead + trellis_prefix)) % 8;
output_count = TRELLIS_DEPTH;
lookahead = trellis_lookahead + remain;
}
if (output_count) {
m_viterbi.Traceback((char*)m_outbuf, output_count, lookahead);
ob_count += output_count;
//output_count >>= 3;
while (ii < output_count) {
//uchar* po = opin().append();
//*po = m_outbuf[ii++];
uchar po = m_outbuf[ii];
bit[ii] = po;
ii++;
//printf ("%d (%x) \n", *po, *po );
//Next()->Process(opin0());
}
//printf ( "\n" );
}
}
//ipin.pop();
return ii;
}