/*
 * jdhuff.h
 *
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
 * ---------------------------------------------------------------------
 * x86 SIMD extension for IJG JPEG library
 * Copyright (C) 1999-2006, MIYASAKA Masaru.
 * This file has been modified to improve performance.
 * Last Modified : October 31, 2004
 * ---------------------------------------------------------------------
 *
 * This file contains declarations for Huffman entropy decoding routines
 * that are shared between the sequential decoder (jdhuff.c) and the
 * progressive decoder (jdphuff.c).  No other modules need to see these.
 */

/* Short forms of external names for systems with brain-damaged linkers. */

#ifdef NEED_SHORT_EXTERNAL_NAMES
#define jpeg_make_d_derived_tbl	jMkDDerived
#define jpeg_fill_bit_buffer	jFilBitBuf
#define jpeg_huff_decode	jHufDecode
#endif /* NEED_SHORT_EXTERNAL_NAMES */


/* Derived data constructed for each Huffman table */

#define HUFFX_LOOKAHEAD	9	/* # of bits of lookahead */

typedef struct {
  /* Basic tables: (element [0] of each array is unused) */
  INT32 maxcode[18];		/* largest code of length k (-1 if none) */
  /* (maxcode[17] is a sentinel to ensure jpeg_huff_decode terminates) */
  INT32 valoffset[17];		/* huffval[] offset for codes of length k */
  /* valoffset[k] = huffval[] index of 1st symbol of code length k, less
   * the smallest code of length k; so given a code of length k, the
   * corresponding symbol is huffval[code + valoffset[k]]
   */

  /* Link to public Huffman table (needed only in jpeg_huff_decode) */
  JHUFF_TBL *pub;

  /* Lookahead tables: indexed by the next HUFFX_LOOKAHEAD bits of
   * the input data stream.  If the next Huffman code is no more
   * than HUFFX_LOOKAHEAD-1 bits long, we can obtain its length,
   * the corresponding symbol, and the encoded coefficient value
   * directly from these tables.
   */
  UINT8 lookx_nbits[1<<HUFFX_LOOKAHEAD];  /* # bits, or 0 if too long */
  INT16 lookx_val[1<<HUFFX_LOOKAHEAD];  /* coefficient value, or unused */
  UINT8 lookx_sym[1<<HUFFX_LOOKAHEAD];  /* symbol, or unused */
} d_derived_tbl;

/* Expand a Huffman table definition into the derived format */
EXTERN(void) jpeg_make_d_derived_tbl
	JPP((j_decompress_ptr cinfo, boolean isDC, int tblno,
	     d_derived_tbl ** pdtbl));


/*
 * Fetching the next N bits from the input stream is a time-critical operation
 * for the Huffman decoders.  We implement it with a combination of inline
 * macros and out-of-line subroutines.  Note that N (the number of bits
 * demanded at one time) never exceeds 15 for JPEG use.
 *
 * We read source bytes into get_buffer and dole out bits as needed.
 * If get_buffer already contains enough bits, they are fetched in-line
 * by the macros CHECK_BIT_BUFFER and GET_BITS.  When there aren't enough
 * bits, jpeg_fill_bit_buffer is called; it will attempt to fill get_buffer
 * as full as possible (not just to the number of bits needed; this
 * prefetching reduces the overhead cost of calling jpeg_fill_bit_buffer).
 * Note that jpeg_fill_bit_buffer may return FALSE to indicate suspension.
 * On TRUE return, jpeg_fill_bit_buffer guarantees that get_buffer contains
 * at least the requested number of bits --- dummy zeroes are inserted if
 * necessary.
 */

typedef INT32 bit_buf_type;	/* type of bit-extraction buffer */
#define BIT_BUF_SIZE  32	/* size of buffer in bits */

/* If long is > 32 bits on your machine, and shifting/masking longs is
 * reasonably fast, making bit_buf_type be long and setting BIT_BUF_SIZE
 * appropriately should be a win.  Unfortunately we can't define the size
 * with something like  #define BIT_BUF_SIZE (sizeof(bit_buf_type)*8)
 * because not all machines measure sizeof in 8-bit bytes.
 */

#ifdef SLOW_SHIFT_32
#define MIN_GET_BITS  15	/* minimum allowable value */
#else
#define MIN_GET_BITS  (BIT_BUF_SIZE-7)
#endif

/* On most machines MIN_GET_BITS should be 25 to allow the full 32-bit width
 * of get_buffer to be used.  (On machines with wider words, an even larger
 * buffer could be used.)  However, on some machines 32-bit shifts are
 * quite slow and take time proportional to the number of places shifted.
 * (This is true with most PC compilers, for instance.)  In this case it may
 * be a win to set MIN_GET_BITS to the minimum value of 15.  This reduces the
 * average shift distance at the cost of more calls to jpeg_fill_bit_buffer.
 */

typedef struct {		/* Bitreading state saved across MCUs */
  bit_buf_type get_buffer;	/* current bit-extraction buffer */
  int bits_left;		/* # of unused bits in it */
} bitread_perm_state;

typedef struct {		/* Bitreading working state within an MCU */
  /* Current data source location */
  /* We need a copy, rather than munging the original, in case of suspension */
  const JOCTET * next_input_byte; /* => next byte to read from source */
  size_t bytes_in_buffer;	/* # of bytes remaining in source buffer */
  /* Bit input buffer --- note these values are kept in register variables,
   * not in this struct, inside the inner loops.
   */
  bit_buf_type get_buffer;	/* current bit-extraction buffer */
  int bits_left;		/* # of unused bits in it */
  /* Pointer needed by jpeg_fill_bit_buffer. */
  j_decompress_ptr cinfo;	/* back link to decompress master record */
} bitread_working_state;

/* Macros to declare and load/save bitread local variables. */
#define BITREAD_STATE_VARS  \
	register bit_buf_type get_buffer;  \
	register int bits_left;  \
	bitread_working_state br_state

#define BITREAD_LOAD_STATE(cinfop,permstate)  \
	br_state.cinfo = cinfop; \
	br_state.next_input_byte = cinfop->src->next_input_byte; \
	br_state.bytes_in_buffer = cinfop->src->bytes_in_buffer; \
	get_buffer = permstate.get_buffer; \
	bits_left = permstate.bits_left

#define BITREAD_SAVE_STATE(cinfop,permstate)  \
	cinfop->src->next_input_byte = br_state.next_input_byte; \
	cinfop->src->bytes_in_buffer = br_state.bytes_in_buffer; \
	permstate.get_buffer = get_buffer; \
	permstate.bits_left = bits_left

/*
 * These macros provide the in-line portion of bit fetching.
 * Use CHECK_BIT_BUFFER to ensure there are N bits in get_buffer
 * before using GET_BITS, PEEK_BITS, or DROP_BITS.
 * The variables get_buffer and bits_left are assumed to be locals,
 * but the state struct might not be (jpeg_huff_decode needs this).
 *	CHECK_BIT_BUFFER(state,n,action);
 *		Ensure there are N bits in get_buffer; if suspend, take action.
 *      val = GET_BITS(n);
 *		Fetch next N bits.
 *      val = PEEK_BITS(n);
 *		Fetch next N bits without removing them from the buffer.
 *	DROP_BITS(n);
 *		Discard next N bits.
 * The value N should be a simple variable, not an expression, because it
 * is evaluated multiple times.
 */

#define CHECK_BIT_BUFFER(state,nbits,action) \
	{ if (bits_left < (nbits)) {  \
	    if (! jpeg_fill_bit_buffer(&(state),get_buffer,bits_left,nbits))  \
	      { action; }  \
	    get_buffer = (state).get_buffer; bits_left = (state).bits_left; } }

#define GET_BITS(nbits) \
	(((int) (get_buffer >> (bits_left -= (nbits)))) & ((1<<(nbits))-1))

#define PEEK_BITS(nbits) \
	(((int) (get_buffer >> (bits_left -  (nbits)))) & ((1<<(nbits))-1))

#define DROP_BITS(nbits) \
	(bits_left -= (nbits))

/* Load up the bit buffer to a depth of at least nbits */
EXTERN(boolean) jpeg_fill_bit_buffer
	JPP((bitread_working_state * state, register bit_buf_type get_buffer,
	     register int bits_left, int nbits));

/* Out-of-line case for Huffman code fetching */
EXTERN(int) jpeg_huff_decode
	JPP((bitread_working_state * state, register bit_buf_type get_buffer,
	     register int bits_left, d_derived_tbl * htbl, int min_bits));


/*
 * Figure F.12: extend sign bit.
 */

#define HUFF_EXTEND(x,s)  ((x) < (1<<((s)-1)) ? (x) + (((-1)<<(s)) + 1) : (x))