pointcloud/lib/pc_dimensional.c
2013-02-18 20:54:37 -08:00

580 lines
14 KiB
C

/***********************************************************************
* pc_dimensional.c
*
* Support for "dimensional compression", which is a catch-all
* term for applying compression separately on each dimension
* of a PCPATCH collection of PCPOINTS.
*
* Depending on the character of the data, one of these schemes
* will be used:
*
* - run-length encoding
* - significant-bit removal
* - deflate
*
* Portions Copyright (c) 2012, OpenGeo
*
***********************************************************************/
#include <stdarg.h>
#include "pc_api_internal.h"
/**
* How many distinct runs of values are there in this array?
* One? Two? Five? Great news for run-length encoding!
* N? Not so great news.
*/
uint32_t
pc_bytes_run_count(const uint8_t *bytes, uint32_t interpretation, uint32_t nelems)
{
int i;
const uint8_t *ptr0;
const uint8_t *ptr1;
size_t size = INTERPRETATION_SIZES[interpretation];
uint32_t runcount = 1;
for ( i = 1; i < nelems; i++ )
{
ptr0 = bytes + (i-1)*size;
ptr1 = bytes + i*size;
if ( memcmp(ptr0, ptr1, size) != 0 )
{
runcount++;
}
}
return runcount;
}
/**
* Take the uncompressed bytes and run-length encode (RLE) them.
* Structure of RLE array as:
* <uint8> number of elements
* <val> value
* ...
*/
uint8_t *
pc_bytes_run_length_encode(const uint8_t *bytes, uint32_t interpretation, uint32_t nelems, size_t *bytes_rle_size)
{
int i;
uint8_t *buf, *bufptr;
const uint8_t *bytesptr;
const uint8_t *runstart;
uint8_t *bytes_rle;
size_t size = INTERPRETATION_SIZES[interpretation];
uint8_t runlength = 1;
/* Allocate more size than we need (worst case: n elements, n runs) */
buf = pcalloc(nelems*size + sizeof(uint8_t)*size);
bufptr = buf;
/* First run starts at the start! */
runstart = bytes;
for ( i = 1; i <= nelems; i++ )
{
bytesptr = bytes + i*size;
/* Run continues... */
if ( i < nelems && runlength < 255 && memcmp(runstart, bytesptr, size) == 0 )
{
runlength++;
}
else
{
/* Write # elements in the run */
*bufptr = runlength;
bufptr += 1;
/* Write element value */
memcpy(bufptr, runstart, size);
bufptr += size;
/* Advance read head */
runstart = bytesptr;
runlength = 1;
}
}
/* Length of buffer */
if ( bytes_rle_size )
{
*bytes_rle_size = (bufptr - buf);
}
/* Write out shortest buffer possible */
bytes_rle = pcalloc(*bytes_rle_size);
memcpy(bytes_rle, buf, *bytes_rle_size);
pcfree(buf);
return bytes_rle;
}
/**
* Take the compressed bytes and run-length dencode (RLE) them.
* Structure of RLE array is:
* <uint8> number of elements
* <val> value
* ...
*/
uint8_t *
pc_bytes_run_length_decode(const uint8_t *bytes_rle, size_t bytes_rle_size, uint32_t interpretation, uint32_t *bytes_nelems)
{
int i, n;
uint8_t *bytes;
uint8_t *bytes_ptr;
const uint8_t *bytes_rle_ptr = bytes_rle;
const uint8_t *bytes_rle_end = bytes_rle + bytes_rle_size;
size_t size = INTERPRETATION_SIZES[interpretation];
uint8_t runlength;
uint32_t nelems = 0;
/* Count up how big our output is. */
while( bytes_rle_ptr < bytes_rle_end )
{
nelems += *bytes_rle_ptr;
bytes_rle_ptr += 1 + size;
}
*bytes_nelems = nelems;
/* Alocate output and fill it up */
bytes = pcalloc(size * nelems);
bytes_ptr = bytes;
bytes_rle_ptr = bytes_rle;
while ( bytes_rle_ptr < bytes_rle_end )
{
n = *bytes_rle_ptr;
bytes_rle_ptr += 1;
for ( i = 0; i < n; i++ )
{
memcpy(bytes_ptr, bytes_rle_ptr, size);
bytes_ptr += size;
}
bytes_rle_ptr += size;
}
return bytes;
}
uint8_t
pc_sigbits_8(const uint8_t *bytes, uint32_t nelems, uint32_t *nsigbits)
{
static uint8_t nbits = 8;
uint8_t elem_and = bytes[0];
uint8_t elem_or = bytes[0];
uint32_t commonbits = nbits;
int i;
for ( i = 0; i < nelems; i++ )
{
elem_and &= bytes[i];
elem_or |= bytes[i];
}
while ( elem_and != elem_or )
{
elem_and >>= 1;
elem_or >>= 1;
commonbits -= 1;
}
elem_and <<= nbits - commonbits;
if ( nsigbits ) *nsigbits = commonbits;
return elem_and;
}
uint16_t
pc_sigbits_16(const uint8_t *bytes8, uint32_t nelems, uint32_t *nsigbits)
{
static int nbits = 16;
uint16_t *bytes = (uint16_t*)bytes8;
uint16_t elem_and = bytes[0];
uint16_t elem_or = bytes[0];
uint32_t commonbits = nbits;
int i;
for ( i = 0; i < nelems; i++ )
{
elem_and &= bytes[i];
elem_or |= bytes[i];
}
while ( elem_and != elem_or )
{
elem_and >>= 1;
elem_or >>= 1;
commonbits -= 1;
}
elem_and <<= nbits - commonbits;
if ( nsigbits ) *nsigbits = commonbits;
return elem_and;
}
uint32_t
pc_sigbits_32(const uint8_t *bytes8, uint32_t nelems, uint32_t *nsigbits)
{
static int nbits = 32;
uint32_t *bytes = (uint32_t*)bytes8;
uint32_t elem_and = bytes[0];
uint32_t elem_or = bytes[0];
uint32_t commonbits = nbits;
int i;
for ( i = 0; i < nelems; i++ )
{
elem_and &= bytes[i];
elem_or |= bytes[i];
}
while ( elem_and != elem_or )
{
elem_and >>= 1;
elem_or >>= 1;
commonbits -= 1;
}
elem_and <<= nbits - commonbits;
if ( nsigbits ) *nsigbits = commonbits;
return elem_and;
}
uint64_t
pc_sigbits_64(const uint8_t *bytes8, uint32_t nelems, uint32_t *nsigbits)
{
static int nbits = 64;
uint64_t *bytes = (uint64_t*)bytes8;
uint64_t elem_and = bytes[0];
uint64_t elem_or = bytes[0];
uint32_t commonbits = nbits;
int i;
for ( i = 0; i < nelems; i++ )
{
elem_and &= bytes[i];
elem_or |= bytes[i];
}
while ( elem_and != elem_or )
{
elem_and >>= 1;
elem_or >>= 1;
commonbits -= 1;
}
elem_and <<= nbits - commonbits;
if ( nsigbits ) *nsigbits = commonbits;
return elem_and;
}
/**
* How many bits are shared by all elements of this array?
*/
uint32_t
pc_sigbits_count(const uint8_t *bytes, uint32_t interpretation, uint32_t nelems)
{
int i, j, start, end, incr;
const uint8_t *bytes_ptr;
uint8_t bytes_and[8];
uint8_t bytes_or[8];
uint8_t bytes_sig[8];
size_t size = INTERPRETATION_SIZES[interpretation];
uint32_t count = 0;
memset(bytes_sig, 0, 8);
memset(bytes_and, 0, 8);
memset(bytes_or, 0, 8);
memcpy(bytes_and, bytes, size);
memcpy(bytes_or, bytes, size);
/* Figure out the global "and" and "or" of all the values */
for ( i = 1; i < nelems; i++ )
{
bytes_ptr = bytes + i*size;
for ( j = 0; j < size; j++ )
{
bytes_and[j] &= bytes_ptr[j];
bytes_or[j] |= bytes_ptr[j];
}
}
/* Count down the bytes for little-endian, up for big */
if ( machine_endian() == PC_NDR )
{
start = size-1;
end = -1;
incr = -1;
}
else
{
start = 0;
end = size;
incr = 1;
}
for ( i = start; i != end; i += incr )
{
/* If bytes are the same, all 8 bits are shared! */
if ( bytes_and[i] == bytes_or[i] )
{
count += 8;
bytes_sig[i] = bytes_and[i];
}
/* If bytes are different, find if they share any top bits */
else
{
int commonbits = 8;
uint8_t b_and = bytes_and[i];
uint8_t b_or = bytes_or[i];
/* Slide off bottom bits until the values match */
while ( b_and != b_or )
{
b_and >>= 1;
b_or >>= 1;
commonbits -= 1;
}
count += commonbits;
/* Save the common bits to the significant bit value */
b_and <<= 8 - commonbits;
bytes_sig[i] = b_and;
break;
}
}
return count;
}
uint8_t *
pc_bytes_sigbits_encode_8(const uint8_t *bytes, uint32_t nelems, uint8_t commonvalue, uint8_t commonbits, size_t *bytes_size)
{
int i;
int shift;
/* How wide are our words? */
static int bitwidth = 8;
/* How wide are our unique values? */
int nbits = bitwidth - commonbits;
/* Size of output buffer */
size_t size_out = (nbits * nelems / 8) + 2;
uint8_t *bytes_out = pcalloc(size_out);
/* Use this to zero out the parts that are common */
uint8_t mask = (0xFF >> commonbits);
/* Write head */
uint8_t *byte_ptr = bytes_out + 1;
/* What bit are we writing to now? */
int bit = bitwidth;
/* Common value goes up front, unmolested */
bytes_out[0] = commonvalue;
for ( i = 0; i < nelems; i++ )
{
uint8_t val = bytes[i];
/* Clear off common parts */
val &= mask;
/* How far to move unique parts to get to write head? */
shift = bit - nbits;
/* If positive, we can fit this part into the current word */
if ( shift >= 0 )
{
val <<= shift;
*byte_ptr |= val;
bit -= nbits;
if ( bit <= 0 )
{
bit = bitwidth;
byte_ptr++;
}
}
/* If negative, then we need to split this part across words */
else
{
/* First the bit into the current word */
uint8_t v = val;
int s = abs(shift);
v >>= s;
*byte_ptr |= v;
/* The reset to write the next word */
bit = bitwidth;
byte_ptr++;
v = val;
shift = bit - s;
/* But only those parts we didn't already write */
v <<= shift;
*byte_ptr |= v;
bit -= s;
}
}
*bytes_size = size_out;
return bytes_out;
}
uint8_t *
pc_bytes_sigbits_encode_16(const uint8_t *bytes8, uint32_t nelems, uint16_t commonvalue, uint8_t commonbits, size_t *bytes_size)
{
int i;
int shift;
uint16_t *bytes = (uint16_t*)bytes8;
/* How wide are our words? */
static int bitwidth = 16;
/* How wide are our unique values? */
int nbits = bitwidth - commonbits;
/* Size of output buffer */
size_t size_out = (nbits * nelems / 8) + 4;
uint16_t *bytes_out = pcalloc(size_out);
/* Use this to zero out the parts that are common */
uint16_t mask = (0xFFFF >> commonbits);
/* Write head */
uint16_t *byte_ptr = bytes_out + 1;
/* What bit are we writing to now? */
int bit = bitwidth;
/* Common value goes up front, unmolested */
bytes_out[0] = commonvalue;
for ( i = 0; i < nelems; i++ )
{
uint16_t val = bytes[i];
/* Clear off common parts */
val &= mask;
/* How far to move unique parts to get to write head? */
shift = bit - nbits;
/* If positive, we can fit this part into the current word */
if ( shift >= 0 )
{
val <<= shift;
*byte_ptr |= val;
bit -= nbits;
if ( bit <= 0 )
{
bit = bitwidth;
byte_ptr++;
}
}
/* If negative, then we need to split this part across words */
else
{
/* First the bit into the current word */
uint16_t v = val;
int s = abs(shift);
v >>= s;
*byte_ptr |= v;
/* The reset to write the next word */
bit = bitwidth;
byte_ptr++;
v = val;
shift = bit - s;
/* But only those parts we didn't already write */
v <<= shift;
*byte_ptr |= v;
bit -= s;
}
}
*bytes_size = size_out;
return (uint8_t*)bytes_out;
}
uint8_t *
pc_bytes_sigbits_encode_32(const uint8_t *bytes8, uint32_t nelems, uint32_t commonvalue, uint8_t commonbits, size_t *bytes_size)
{
int i;
int shift;
uint32_t *bytes = (uint32_t*)bytes8;
/* How wide are our words? */
static int bitwidth = 32;
/* How wide are our unique values? */
int nbits = bitwidth - commonbits;
/* Size of output buffer */
size_t size_out = (nbits * nelems / 8) + 8;
uint32_t *bytes_out = pcalloc(size_out);
/* Use this to zero out the parts that are common */
uint32_t mask = (0xFFFFFFFF >> commonbits);
/* Write head */
uint32_t *byte_ptr = bytes_out + 1;
/* What bit are we writing to now? */
int bit = bitwidth;
/* Common value goes up front, unmolested */
bytes_out[0] = commonvalue;
for ( i = 0; i < nelems; i++ )
{
uint32_t val = bytes[i];
/* Clear off common parts */
val &= mask;
/* How far to move unique parts to get to write head? */
shift = bit - nbits;
/* If positive, we can fit this part into the current word */
if ( shift >= 0 )
{
val <<= shift;
*byte_ptr |= val;
bit -= nbits;
if ( bit <= 0 )
{
bit = bitwidth;
byte_ptr++;
}
}
/* If negative, then we need to split this part across words */
else
{
/* First the bit into the current word */
uint32_t v = val;
int s = abs(shift);
v >>= s;
*byte_ptr |= v;
/* The reset to write the next word */
bit = bitwidth;
byte_ptr++;
v = val;
shift = bit - s;
/* But only those parts we didn't already write */
v <<= shift;
*byte_ptr |= v;
bit -= s;
}
}
*bytes_size = size_out;
return (uint8_t*)bytes_out;
}
uint8_t *
pc_bytes_sigbits_encode(const uint8_t *bytes, uint32_t interpretation, uint32_t nelems, size_t *ebytes_size)
{
size_t size = INTERPRETATION_SIZES[interpretation];
uint32_t nbits;
switch ( size )
{
case 1:
{
uint8_t commonvalue = pc_sigbits_8(bytes, nelems, &nbits);
return pc_bytes_sigbits_encode_8(bytes, nelems, commonvalue, nbits, ebytes_size);
}
case 2:
{
uint16_t commonvalue = pc_sigbits_16(bytes, nelems, &nbits);
return pc_bytes_sigbits_encode_16(bytes, nelems, commonvalue, nbits, ebytes_size);
}
case 4:
{
uint16_t commonvalue = pc_sigbits_32(bytes, nelems, &nbits);
return pc_bytes_sigbits_encode_32(bytes, nelems, commonvalue, nbits, ebytes_size);
}
default:
{
pcerror("Uh oh");
}
}
pcerror("Uh Oh");
return NULL;
}