home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
CU Amiga Super CD-ROM 15
/
CU Amiga Magazine's Super CD-ROM 15 (1997)(EMAP Images)(GB)[!][issue 1997-10].iso
/
CUCD
/
Graphics
/
Ghostscript
/
source
/
gsbitops.h
< prev
next >
Wrap
C/C++ Source or Header
|
1995-12-05
|
7KB
|
174 lines
/* Copyright (C) 1991, 1995 Aladdin Enterprises. All rights reserved.
This file is part of Aladdin Ghostscript.
Aladdin Ghostscript is distributed with NO WARRANTY OF ANY KIND. No author
or distributor accepts any responsibility for the consequences of using it,
or for whether it serves any particular purpose or works at all, unless he
or she says so in writing. Refer to the Aladdin Ghostscript Free Public
License (the "License") for full details.
Every copy of Aladdin Ghostscript must include a copy of the License,
normally in a plain ASCII text file named PUBLIC. The License grants you
the right to copy, modify and redistribute Aladdin Ghostscript, but only
under certain conditions described in the License. Among other things, the
License requires that the copyright notice and this notice be preserved on
all copies.
*/
/* gsbitops.h */
/* Definitions for bitmap operations */
/* ---------------- Definitions ---------------- */
/*
* Macros for processing bitmaps in the largest possible chunks.
* Bits within a byte are always stored big-endian;
* bytes are likewise stored in left-to-right order, i.e., big-endian.
* Note that this is the format used for the source of copy_mono.
* It used to be the case that bytes were stored in the natural
* platform order, and the client had force them into big-endian order
* by calling gdev_mem_ensure_byte_order, but this no longer necessary.
*
* Note that we use type uint for register variables holding a chunk:
* for this reason, the chunk size cannot be larger than uint.
*/
/* Generic macros for chunk accessing. */
#define cbytes(ct) size_of(ct) /* sizeof may be unsigned */
# define chunk_bytes cbytes(chunk)
/* The clog2_bytes macro assumes that ints are 2, 4, or 8 bytes in size. */
#define clog2_bytes(ct) (size_of(ct) == 8 ? 3 : size_of(ct)>>1)
# define chunk_log2_bytes clog2_bytes(chunk)
#define cbits(ct) (size_of(ct)*8) /* sizeof may be unsigned */
# define chunk_bits cbits(chunk)
#define clog2_bits(ct) (clog2_bytes(ct)+3)
# define chunk_log2_bits clog2_bits(chunk)
#define cbit_mask(ct) (cbits(ct)-1)
# define chunk_bit_mask cbit_mask(chunk)
#define calign_bytes(ct)\
(sizeof(ct) == 1 ? 1:\
sizeof(ct) == sizeof(short) ? arch_align_short_mod :\
sizeof(ct) == sizeof(int) ? arch_align_int_mod : arch_align_long_mod)
# define chunk_align_bytes calign_bytes(chunk)
#define calign_bit_mask(ct) (calign_bytes(ct)*8-1)
# define chunk_align_bit_mask calign_bit_mask(chunk)
/*
* The obvious definition for cmask is:
* #define cmask(ct) ((ct)~(ct)0)
* but this doesn't work on the VAX/VMS compiler, which fails to truncate
* the value to 16 bits when ct is ushort.
* Instead, we have to generate the mask with no extra 1-bits.
* We can't do this in the obvious way:
* #define cmask(ct) ((1 << (size_of(ct) * 8)) - 1)
* because some compilers won't allow a shift of the full type size.
* Instead, we have to do something really awkward:
*/
#define cmask(ct) ((ct) (((((ct)1 << (size_of(ct)*8-2)) - 1) << 2) + 3))
# define chunk_all_bits cmask(chunk)
/*
* The obvious definition for chi_bits is:
* #define chi_bits(ct,n) (cmask(ct)-(cmask(ct)>>(n)))
* but this doesn't work on the DEC/MIPS compilers.
* Instead, we have to restrict chi_bits to only working for values of n
* between 0 and cbits(ct)-1, and use
*/
#define chi_bits(ct,n) (ct)(~(ct)1 << (cbits(ct)-1 - (n)))
# define chunk_hi_bits(n) chi_bits(chunk,n)
/* Define whether this is a machine where chunks are long, */
/* but the machine can't shift a long by its full width. */
#define arch_cant_shift_full_chunk\
(arch_is_big_endian && !arch_ints_are_short && !arch_can_shift_full_long)
/* Pointer arithmetic macros. */
#define inc_ptr(ptr,delta)\
ptr = (void *)((byte *)ptr + (delta))
/* Define macros for setting up left- and right-end masks. */
/* These are used for monobit operations, and for filling */
/* with 2- and 4-bit-per-pixel patterns. */
/*
* Define the chunk size for monobit copying operations.
* (The chunk size for monobit filling is always uint.)
*/
#define mono_fill_chunk uint
#define mono_fill_chunk_bytes arch_sizeof_int
#if arch_is_big_endian
# define mono_copy_chunk uint
# define set_mono_right_mask(var, w)\
var = ((w) == chunk_bits ? chunk_all_bits : chunk_hi_bits(w))
/*
* We have to split the following statement because of a bug in the Xenix C
* compiler (it produces a signed rather than an unsigned shift if we don't
* split).
*/
# define set_mono_thin_mask(var, w, bit)\
set_mono_right_mask(var, w), var >>= (bit)
/*
* We have to split the following statement in two because of a bug
* in the DEC VAX/VMS C compiler.
*/
# define set_mono_left_mask(var, bit)\
var = chunk_all_bits, var >>= (bit)
#else
# define mono_copy_chunk bits16
extern const bits16 mono_copy_masks[17];
# if mono_fill_chunk_bytes == 2
# define mono_fill_masks mono_copy_masks
# else
extern const bits32 mono_fill_masks[33];
# endif
/*
* We define mono_masks as either mono_fill_masks or
* mono_copy_masks before using the following macros.
*/
# define set_mono_left_mask(var, bit)\
var = mono_masks[bit]
# define set_mono_thin_mask(var, w, bit)\
var = ~mono_masks[(w) + (bit)] & mono_masks[bit]
# define set_mono_right_mask(var, ebit)\
var = ~mono_masks[ebit]
#endif
/* ---------------- Procedures ---------------- */
/* Fill a rectangle of bits with an 8x1 pattern. */
/* The pattern argument must consist of the pattern in every byte, */
/* e.g., if the desired pattern is 0xaa, the pattern argument must */
/* have the value 0xaaaa (if ints are short) or 0xaaaaaaaa. */
#if mono_fill_chunk_bytes == 2
# define mono_fill_make_pattern(byt) (uint)((uint)(byt) * 0x0101)
#else
# define mono_fill_make_pattern(byt) (uint)((uint)(byt) * 0x01010101)
#endif
void bits_fill_rectangle(P6(byte *dest, int dest_bit, uint raster,
mono_fill_chunk pattern, int width_bits, int height));
/* Replicate a bitmap horizontally in place. */
void bits_replicate_horizontally(P6(byte *data, uint width, uint height,
uint raster, uint replicated_width, uint replicated_raster));
/* Replicate a bitmap vertically in place. */
void bits_replicate_vertically(P4(byte *data, uint height, uint raster,
uint replicated_height));
/* Find the bounding box of a bitmap. */
void bits_bounding_box(P4(const byte *data, uint height, uint raster,
gs_int_rect *pbox));
/* Compress an oversampled image, possibly in place. */
/* The width and height must be multiples of the respective scale factors. */
/* The source must be an aligned bitmap, as usual. */
void bits_compress_scaled(P9(const byte *src, int srcx, uint width,
uint height, uint sraster, byte *dest, uint draster,
const gs_log2_scale_point *plog2_scale, int log2_out_bits));
/* Fill a rectangle of bytes. */
void bytes_fill_rectangle(P5(byte *dest, uint raster,
byte value, int width_bytes, int height));
/* Copy a rectangle of bytes. */
void bytes_copy_rectangle(P6(byte *dest, uint dest_raster,
const byte *src, uint src_raster, int width_bytes, int height));