rockbox/apps/codecs/libtremor/codebook.c

588 lines
17 KiB
C
Raw Normal View History

/********************************************************************
* *
* THIS FILE IS PART OF THE OggVorbis 'TREMOR' CODEC SOURCE CODE. *
* *
* USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS *
* GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
* IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING. *
* *
* THE OggVorbis 'TREMOR' SOURCE CODE IS (C) COPYRIGHT 1994-2002 *
* BY THE Xiph.Org FOUNDATION http://www.xiph.org/ *
* *
********************************************************************
function: basic codebook pack/unpack/code/decode operations
********************************************************************/
#include "config-tremor.h"
#include <string.h>
#include <math.h>
#include "ogg.h"
#include "ivorbiscodec.h"
#include "codebook.h"
#include "misc.h"
#include "os.h"
/* unpacks a codebook from the packet buffer into the codebook struct,
readies the codebook auxiliary structures for decode *************/
static_codebook *vorbis_staticbook_unpack(oggpack_buffer *opb){
long i,j;
static_codebook *s=_ogg_calloc(1,sizeof(*s));
/* make sure alignment is correct */
if(oggpack_read(opb,24)!=0x564342)goto _eofout;
/* first the basic parameters */
s->dim=oggpack_read(opb,16);
s->entries=oggpack_read(opb,24);
if(s->entries==-1)goto _eofout;
if(_ilog(s->dim)+_ilog(s->entries)>24)goto _eofout;
/* codeword ordering.... length ordered or unordered? */
switch((int)oggpack_read(opb,1)){
case 0:{
long unused;
/* allocated but unused entries? */
unused=oggpack_read(opb,1);
if((s->entries*(unused?1:5)+7)>>3>opb->storage-oggpack_bytes(opb))
goto _eofout;
/* unordered */
s->lengthlist=(long *)_ogg_malloc(sizeof(*s->lengthlist)*s->entries);
/* allocated but unused entries? */
if(unused){
/* yes, unused entries */
for(i=0;i<s->entries;i++){
if(oggpack_read(opb,1)){
long num=oggpack_read(opb,5);
if(num==-1)goto _eofout;
s->lengthlist[i]=num+1;
}else
s->lengthlist[i]=0;
}
}else{
/* all entries used; no tagging */
for(i=0;i<s->entries;i++){
long num=oggpack_read(opb,5);
if(num==-1)goto _eofout;
s->lengthlist[i]=num+1;
}
}
break;
}
case 1:
/* ordered */
{
long length=oggpack_read(opb,5)+1;
if(length==0)goto _eofout;
s->lengthlist=(long *)_ogg_malloc(sizeof(*s->lengthlist)*s->entries);
for(i=0;i<s->entries;){
long num=oggpack_read(opb,_ilog(s->entries-i));
if(num==-1)goto _eofout;
if(length>32 || num>s->entries-i ||
(num>0 && (num-1)>>(length>>1)>>((length+1)>>1))>0){
goto _errout;
}
for(j=0;j<num;j++,i++)
s->lengthlist[i]=length;
length++;
}
}
break;
default:
/* EOF */
goto _eofout;
}
/* Do we have a mapping to unpack? */
switch((s->maptype=oggpack_read(opb,4))){
case 0:
/* no mapping */
break;
case 1: case 2:
/* implicitly populated value mapping */
/* explicitly populated value mapping */
s->q_min=oggpack_read(opb,32);
s->q_delta=oggpack_read(opb,32);
s->q_quant=oggpack_read(opb,4)+1;
s->q_sequencep=oggpack_read(opb,1);
if(s->q_sequencep==-1)goto _eofout;
{
int quantvals=0;
switch(s->maptype){
case 1:
quantvals=(s->dim==0?0:_book_maptype1_quantvals(s));
break;
case 2:
quantvals=s->entries*s->dim;
break;
}
/* quantized values */
if((quantvals*s->q_quant+7)>>3>opb->storage-oggpack_bytes(opb))
goto _eofout;
s->quantlist=(long *)_ogg_malloc(sizeof(*s->quantlist)*quantvals);
for(i=0;i<quantvals;i++)
s->quantlist[i]=oggpack_read(opb,s->q_quant);
if(quantvals&&s->quantlist[quantvals-1]==-1)goto _eofout;
}
break;
default:
goto _errout;
}
/* all set */
return(s);
_errout:
_eofout:
vorbis_staticbook_destroy(s);
return(NULL);
}
/* the 'eliminate the decode tree' optimization actually requires the
codewords to be MSb first, not LSb. This is an annoying inelegancy
(and one of the first places where carefully thought out design
turned out to be wrong; Vorbis II and future Ogg codecs should go
to an MSb bitpacker), but not actually the huge hit it appears to
be. The first-stage decode table catches most words so that
bitreverse is not in the main execution path. */
static inline ogg_uint32_t bitreverse(register ogg_uint32_t x)
{
unsigned tmp, ret;
#ifdef _ARM_ASSEM_
#if ARM_ARCH >= 6
unsigned mask = 0x0f0f0f0f;
#else
unsigned mask = 0x00ff00ff;
#endif
asm (
#if ARM_ARCH >= 6
"rev %[r], %[x] \n" /* swap halfwords and bytes */
"and %[t], %[m], %[r] \n" /* Sequence is one instruction */
"eor %[r], %[t], %[r] \n" /* longer than on <= ARMv5, but */
"mov %[t], %[t], lsl #4 \n" /* interlock free */
"orr %[r], %[t], %[r], lsr #4\n" /* nibbles swapped */
"eor %[m], %[m], %[m], lsl #2\n" /* mask = 0x33333333 */
"and %[t], %[m], %[r] \n"
"eor %[r], %[t], %[r] \n"
"mov %[t], %[t], lsl #2 \n"
"orr %[r], %[t], %[r], lsr #2\n" /* dibits swapped */
"eor %[m], %[m], %[m], lsl #1\n" /* mask = 0x55555555 */
"and %[t], %[m], %[r] \n"
"eor %[r], %[t], %[r] \n"
"mov %[t], %[t], lsl #1 \n"
"orr %[r], %[t], %[r], lsr #1\n" /* bits swapped */
#else /* ARM_ARCH <= 5 */
"mov %[r], %[x], ror #16 \n" /* swap halfwords */
"and %[t], %[m], %[r], lsr #8\n"
"eor %[r], %[r], %[t], lsl #8\n"
"orr %[r], %[t], %[r], lsl #8\n" /* bytes swapped */
"eor %[m], %[m], %[m], lsl #4\n" /* mask = 0x0f0f0f0f */
"and %[t], %[m], %[r], lsr #4\n"
"eor %[r], %[r], %[t], lsl #4\n"
"orr %[r], %[t], %[r], lsl #4\n" /* nibbles swapped */
"eor %[m], %[m], %[m], lsl #2\n" /* mask = 0x33333333 */
"and %[t], %[m], %[r], lsr #2\n"
"eor %[r], %[r], %[t], lsl #2\n"
"orr %[r], %[t], %[r], lsl #2\n" /* dibits swapped */
"eor %[m], %[m], %[m], lsl #1\n" /* mask = 0x55555555 */
"and %[t], %[m], %[r], lsr #1\n"
"eor %[r], %[r], %[t], lsl #1\n"
"orr %[r], %[t], %[r], lsl #1\n" /* bits swapped */
#endif /* ARM_ARCH */
: /* outputs */
[m]"+r"(mask),
[r]"=r"(ret),
[t]"=r"(tmp)
: /* inputs */
[x]"r"(x)
);
#else /* !_ARM_ASSEM_ */
ret = (x>>16) | (x<<16);
tmp = ret & 0x00ff00ff;
ret ^= tmp;
ret = (ret >> 8) | (tmp << 8); /* bytes swapped */
tmp = ret & 0x0f0f0f0f;
ret ^= tmp;
ret = (ret >> 4) | (tmp << 4); /* 4-bit units swapped */
tmp = ret & 0x33333333;
ret ^= tmp;
ret = (ret >> 2) | (tmp << 2); /* 2-bit units swapped */
tmp = ret & 0x55555555;
ret ^= tmp;
ret = (ret >> 1) | (tmp << 1); /* done */
#endif /* !_ARM_ASSEM_ */
return ret;
}
static inline long bisect_codelist(long lo, long hi, ogg_uint32_t cache,
const ogg_uint32_t *codelist)
{
ogg_uint32_t testword=bitreverse(cache);
long p;
while(LIKELY(p = (hi-lo) >> 1) > 0){
if(codelist[lo+p] > testword)
hi -= p;
else
lo += p;
}
return lo;
}
STIN long decode_packed_entry_number(codebook *book,
oggpack_buffer *b){
int read=book->dec_maxlength;
long lo,hi;
long lok = oggpack_look(b,book->dec_firsttablen);
if (LIKELY(lok >= 0)) {
ogg_int32_t entry = book->dec_firsttable[lok];
if(UNLIKELY(entry < 0)){
lo=(entry>>15)&0x7fff;
hi=book->used_entries-(entry&0x7fff);
}else{
oggpack_adv(b, book->dec_codelengths[entry-1]);
return(entry-1);
}
}else{
lo=0;
hi=book->used_entries;
}
lok = oggpack_look(b, read);
while(lok<0 && read>1)
lok = oggpack_look(b, --read);
if(lok<0){
oggpack_adv(b,1); /* force eop */
return -1;
}
/* bisect search for the codeword in the ordered list */
{
lo = bisect_codelist(lo, hi, lok, book->codelist);
if(book->dec_codelengths[lo]<=read){
oggpack_adv(b, book->dec_codelengths[lo]);
return(lo);
}
}
oggpack_adv(b, read+1);
return(-1);
}
static long decode_packed_block(codebook *book, oggpack_buffer *b,
long *buf, int n){
long *bufptr = buf;
long *bufend = buf + n;
while (bufptr<bufend) {
if(b->endbyte < b->storage - 8) {
ogg_uint32_t *ptr;
unsigned long bit, bitend;
unsigned long adr;
ogg_uint32_t cache = 0;
int cachesize = 0;
const unsigned int cachemask = (1<<book->dec_firsttablen)-1;
const int book_dec_maxlength = book->dec_maxlength;
const ogg_uint32_t *book_dec_firsttable = book->dec_firsttable;
const long book_used_entries = book->used_entries;
const ogg_uint32_t *book_codelist = book->codelist;
const char *book_dec_codelengths = book->dec_codelengths;
adr = (unsigned long)b->ptr;
bit = (adr&3)*8+b->endbit;
ptr = (ogg_uint32_t*)(adr&~3);
bitend = ((adr&3)+(b->storage-b->endbyte))*8;
while (bufptr<bufend){
if (UNLIKELY(cachesize<book_dec_maxlength)) {
if (bit-cachesize+32>=bitend)
break;
bit-=cachesize;
cache = letoh32(ptr[bit>>5]);
if (bit&31) {
cache >>= (bit&31);
cache |= letoh32(ptr[(bit>>5)+1]) << (32-(bit&31));
}
cachesize=32;
bit+=32;
}
ogg_int32_t entry = book_dec_firsttable[cache&cachemask];
if(UNLIKELY(entry < 0)){
const long lo = (entry>>15)&0x7fff, hi = book_used_entries-(entry&0x7fff);
entry = bisect_codelist(lo, hi, cache, book_codelist);
}else
entry--;
*bufptr++ = entry;
int l = book_dec_codelengths[entry];
cachesize -= l;
cache >>= l;
}
adr=(unsigned long)b->ptr;
bit-=(adr&3)*8+cachesize;
b->endbyte+=bit/8;
b->ptr+=bit/8;
b->endbit=bit&7;
} else {
long r = decode_packed_entry_number(book, b);
if (r == -1) return bufptr-buf;
*bufptr++ = r;
}
}
return n;
}
/* Decode side is specced and easier, because we don't need to find
matches using different criteria; we simply read and map. There are
two things we need to do 'depending':
We may need to support interleave. We don't really, but it's
convenient to do it here rather than rebuild the vector later.
Cascades may be additive or multiplicitive; this is not inherent in
the codebook, but set in the code using the codebook. Like
interleaving, it's easiest to do it here.
addmul==0 -> declarative (set the value)
addmul==1 -> additive
addmul==2 -> multiplicitive */
/* returns the [original, not compacted] entry number or -1 on eof *********/
long vorbis_book_decode(codebook *book, oggpack_buffer *b){
if(book->used_entries>0){
long packed_entry=decode_packed_entry_number(book,b);
if(packed_entry>=0)
return(book->dec_index[packed_entry]);
}
/* if there's no dec_index, the codebook unpacking isn't collapsed */
return(-1);
}
/* returns 0 on OK or -1 on eof *************************************/
/* decode vector / dim granularity gaurding is done in the upper layer */
long vorbis_book_decodevs_add(codebook *book,ogg_int32_t *a,
oggpack_buffer *b,int n,int point){
if(book->used_entries>0){
int step=n/book->dim;
long *entry = (long *)alloca(sizeof(*entry)*step);
ogg_int32_t **t = (ogg_int32_t **)alloca(sizeof(*t)*step);
int i,j,o;
int shift=point-book->binarypoint;
if(shift>=0){
for (i = 0; i < step; i++) {
entry[i]=decode_packed_entry_number(book,b);
if(entry[i]==-1)return(-1);
t[i] = book->valuelist+entry[i]*book->dim;
}
for(i=0,o=0;i<book->dim;i++,o+=step)
for (j=0;j<step;j++)
a[o+j]+=t[j][i]>>shift;
}else{
for (i = 0; i < step; i++) {
entry[i]=decode_packed_entry_number(book,b);
if(entry[i]==-1)return(-1);
t[i] = book->valuelist+entry[i]*book->dim;
}
for(i=0,o=0;i<book->dim;i++,o+=step)
for (j=0;j<step;j++)
a[o+j]+=t[j][i]<<-shift;
}
}
return(0);
}
/* decode vector / dim granularity gaurding is done in the upper layer */
long vorbis_book_decodev_add(codebook *book,ogg_int32_t *a,
oggpack_buffer *b,int n,int point){
if(book->used_entries>0){
int i,j,entry;
ogg_int32_t *t;
int shift=point-book->binarypoint;
if(shift>=0){
for(i=0;i<n;){
entry = decode_packed_entry_number(book,b);
if(entry==-1)return(-1);
t = book->valuelist+entry*book->dim;
for (j=0;j<book->dim;)
a[i++]+=t[j++]>>shift;
}
}else{
shift = -shift;
for(i=0;i<n;){
entry = decode_packed_entry_number(book,b);
if(entry==-1)return(-1);
t = book->valuelist+entry*book->dim;
for (j=0;j<book->dim;)
a[i++]+=t[j++]<<shift;
}
}
}
return(0);
}
/* unlike the others, we guard against n not being an integer number
of <dim> internally rather than in the upper layer (called only by
floor0) */
long vorbis_book_decodev_set(codebook *book,ogg_int32_t *a,
oggpack_buffer *b,int n,int point){
if(book->used_entries>0){
int i,j,entry;
ogg_int32_t *t;
int shift=point-book->binarypoint;
if(shift>=0){
for(i=0;i<n;){
entry = decode_packed_entry_number(book,b);
if(entry==-1)return(-1);
t = book->valuelist+entry*book->dim;
for (j=0;i<n && j<book->dim;){
a[i++]=t[j++]>>shift;
}
}
}else{
shift = -shift;
for(i=0;i<n;){
entry = decode_packed_entry_number(book,b);
if(entry==-1)return(-1);
t = book->valuelist+entry*book->dim;
for (j=0;i<n && j<book->dim;){
a[i++]=t[j++]<<shift;
}
}
}
}else{
int i;
for(i=0;i<n;){
a[i++]=0;
}
}
return(0);
}
/* decode vector / dim granularity gaurding is done in the upper layer */
static long vorbis_book_decodevv_add_2ch_even(codebook *book,ogg_int32_t **a,
long offset,oggpack_buffer *b,
unsigned int n,int point){
long k,chunk,read;
int shift=point-book->binarypoint;
long entries[32];
ogg_int32_t *p0 = &(a[0][offset]);
ogg_int32_t *p1 = &(a[1][offset]);
const unsigned long dim = book->dim;
const ogg_int32_t * const vlist = book->valuelist;
if(shift>=0){
while(n>0){
chunk=32;
if (16*dim>n)
chunk=(n*2-1)/dim + 1;
read = decode_packed_block(book,b,entries,chunk);
for(k=0;k<read;k++){
const ogg_int32_t *t = vlist+entries[k]*dim;
const ogg_int32_t *u = t+dim;
do{
*p0++ += *t++>>shift;
*p1++ += *t++>>shift;
}while(t<u);
}
if (read<chunk)return-1;
n -= read*dim/2;
}
}else{
shift = -shift;
while(n>0){
chunk=32;
if (16*dim>n)
chunk=(n*2-1)/dim + 1;
read = decode_packed_block(book,b,entries,chunk);
for(k=0;k<read;k++){
const ogg_int32_t *t = vlist+entries[k]*dim;
const ogg_int32_t *u = t+dim;
do{
*p0++ += *t++<<shift;
*p1++ += *t++<<shift;
}while(t<u);
}
if (read<chunk)return-1;
n -= read*dim/2;
}
}
return(0);
}
long vorbis_book_decodevv_add(codebook *book,ogg_int32_t **a,
long offset,int ch,
oggpack_buffer *b,int n,int point){
if(LIKELY(book->used_entries>0)){
long i,j,k,chunk,read;
int chptr=0;
int shift=point-book->binarypoint;
long entries[32];
if (!(book->dim&1) && ch==2)
return vorbis_book_decodevv_add_2ch_even(book,a,offset,b,n,point);
if(shift>=0){
for(i=offset;i<offset+n;){
chunk=32;
if (chunk*book->dim>(offset+n-i)*ch)
chunk=((offset+n-i)*ch+book->dim-1)/book->dim;
read = decode_packed_block(book,b,entries,chunk);
for(k=0;k<read;k++){
const ogg_int32_t *t = book->valuelist+entries[k]*book->dim;
for (j=0;j<book->dim;j++){
a[chptr++][i]+=t[j]>>shift;
if(chptr==ch){
chptr=0;
i++;
}
}
}
if (read<chunk)return-1;
}
}else{
shift = -shift;
for(i=offset;i<offset+n;){
chunk=32;
if (chunk*book->dim>(offset+n-i)*ch)
chunk=((offset+n-i)*ch+book->dim-1)/book->dim;
read = decode_packed_block(book,b,entries,chunk);
for(k=0;k<read;k++){
const ogg_int32_t *t = book->valuelist+entries[k]*book->dim;
for (j=0;j<book->dim;j++){
a[chptr++][i]+=t[j]<<shift;
if(chptr==ch){
chptr=0;
i++;
}
}
}
if (read<chunk)return-1;
}
}
}
return(0);
}