/* Copyright (C) 1996, 1997, 1998, 1999 Aladdin Enterprises.  All rights reserved.
  
  This software is provided AS-IS with no warranty, either express or
  implied.
  
  This software is distributed under license and may not be copied,
  modified or distributed except as expressly authorized under the terms
  of the license contained in the file LICENSE in this distribution.
  
  For more information about licensing, please refer to
  http://www.ghostscript.com/licensing/. For information on
  commercial licensing, go to http://www.artifex.com/licensing/ or
  contact Artifex Software, Inc., 101 Lucas Valley Road #110,
  San Rafael, CA  94903, U.S.A., +1(415)492-9861.
*/

/* $Id: gsflip.c,v 1.5 2002/06/16 05:48:55 lpd Exp $ */
/* Routines for "flipping" image data */
#include "gx.h"
#include "gserrors.h"		/* for rangecheck in sample macros */
#include "gsbitops.h"
#include "gsbittab.h"
#include "gsflip.h"

#define ARCH_HAS_BYTE_REGS 1

/* Transpose a block of bits between registers. */
#define TRANSPOSE(r,s,mask,shift)\
  r ^= (temp = ((s >> shift) ^ r) & mask);\
  s ^= temp << shift

/* Define the size of byte temporaries.  On Intel CPUs, this should be */
/* byte, but on all other CPUs, it should be uint. */
#if ARCH_HAS_BYTE_REGS
typedef byte byte_var;
#else
typedef uint byte_var;
#endif

#define VTAB(v80,v40,v20,v10,v8,v4,v2,v1)\
  bit_table_8(0,v80,v40,v20,v10,v8,v4,v2,v1)

/* Convert 3Mx1 to 3x1. */
private int
flip3x1(byte * buffer, const byte ** planes, int offset, int nbytes)
{
    byte *out = buffer;
    const byte *in1 = planes[0] + offset;
    const byte *in2 = planes[1] + offset;
    const byte *in3 = planes[2] + offset;
    int n = nbytes;
    static const bits32 tab3x1[256] = {
	VTAB(0x800000, 0x100000, 0x20000, 0x4000, 0x800, 0x100, 0x20, 4)
    };

    for (; n > 0; out += 3, ++in1, ++in2, ++in3, --n) {
	bits32 b24 = tab3x1[*in1] | (tab3x1[*in2] >> 1) | (tab3x1[*in3] >> 2);

	out[0] = (byte) (b24 >> 16);
	out[1] = (byte) (b24 >> 8);
	out[2] = (byte) b24;
    }
    return 0;
}

/* Convert 3Mx2 to 3x2. */
private int
flip3x2(byte * buffer, const byte ** planes, int offset, int nbytes)
{
    byte *out = buffer;
    const byte *in1 = planes[0] + offset;
    const byte *in2 = planes[1] + offset;
    const byte *in3 = planes[2] + offset;
    int n = nbytes;
    static const bits32 tab3x2[256] = {
	VTAB(0x800000, 0x400000, 0x20000, 0x10000, 0x800, 0x400, 0x20, 0x10)
    };

    for (; n > 0; out += 3, ++in1, ++in2, ++in3, --n) {
	bits32 b24 = tab3x2[*in1] | (tab3x2[*in2] >> 2) | (tab3x2[*in3] >> 4);

	out[0] = (byte) (b24 >> 16);
	out[1] = (byte) (b24 >> 8);
	out[2] = (byte) b24;
    }
    return 0;
}

/* Convert 3Mx4 to 3x4. */
private int
flip3x4(byte * buffer, const byte ** planes, int offset, int nbytes)
{
    byte *out = buffer;
    const byte *in1 = planes[0] + offset;
    const byte *in2 = planes[1] + offset;
    const byte *in3 = planes[2] + offset;
    int n = nbytes;

    for (; n > 0; out += 3, ++in1, ++in2, ++in3, --n) {
	byte_var b1 = *in1, b2 = *in2, b3 = *in3;

	out[0] = (b1 & 0xf0) | (b2 >> 4);
	out[1] = (b3 & 0xf0) | (b1 & 0xf);
	out[2] = (byte) (b2 << 4) | (b3 & 0xf);
    }
    return 0;
}

/* Convert 3Mx8 to 3x8. */
private int
flip3x8(byte * buffer, const byte ** planes, int offset, int nbytes)
{
    byte *out = buffer;
    const byte *in1 = planes[0] + offset;
    const byte *in2 = planes[1] + offset;
    const byte *in3 = planes[2] + offset;
    int n = nbytes;

    for (; n > 0; out += 3, ++in1, ++in2, ++in3, --n) {
	out[0] = *in1;
	out[1] = *in2;
	out[2] = *in3;
    }
    return 0;
}

/* Convert 3Mx12 to 3x12. */
private int
flip3x12(byte * buffer, const byte ** planes, int offset, int nbytes)
{
    byte *out = buffer;
    const byte *pa = planes[0] + offset;
    const byte *pb = planes[1] + offset;
    const byte *pc = planes[2] + offset;
    int n = nbytes;

    /*
     * We assume that the input is an integral number of pixels, and
     * round up n to a multiple of 3.
     */
    for (; n > 0; out += 9, pa += 3, pb += 3, pc += 3, n -= 3) {
	byte_var a1 = pa[1], b0 = pb[0], b1 = pb[1], b2 = pb[2], c1 = pc[1];

	out[0] = pa[0];
	out[1] = (a1 & 0xf0) | (b0 >> 4);
	out[2] = (byte) ((b0 << 4) | (b1 >> 4));
	out[3] = pc[0];
	out[4] = (c1 & 0xf0) | (a1 & 0xf);
	out[5] = pa[2];
	out[6] = (byte) ((b1 << 4) | (b2 >> 4));
	out[7] = (byte) ((b2 << 4) | (c1 & 0xf));
	out[8] = pc[2];
    }
    return 0;
}

/* Convert 4Mx1 to 4x1. */
private int
flip4x1(byte * buffer, const byte ** planes, int offset, int nbytes)
{
    byte *out = buffer;
    const byte *in1 = planes[0] + offset;
    const byte *in2 = planes[1] + offset;
    const byte *in3 = planes[2] + offset;
    const byte *in4 = planes[3] + offset;
    int n = nbytes;

    for (; n > 0; out += 4, ++in1, ++in2, ++in3, ++in4, --n) {
	byte_var b1 = *in1, b2 = *in2, b3 = *in3, b4 = *in4;
	byte_var temp;

	/* Transpose blocks of 1 */
	TRANSPOSE(b1, b2, 0x55, 1);
	TRANSPOSE(b3, b4, 0x55, 1);
	/* Transpose blocks of 2 */
	TRANSPOSE(b1, b3, 0x33, 2);
	TRANSPOSE(b2, b4, 0x33, 2);
	/* There's probably a faster way to do this.... */
	out[0] = (b1 & 0xf0) | (b2 >> 4);
	out[1] = (b3 & 0xf0) | (b4 >> 4);
	out[2] = (byte) ((b1 << 4) | (b2 & 0xf));
	out[3] = (byte) ((b3 << 4) | (b4 & 0xf));
    }
    return 0;
}

/* Convert 4Mx2 to 4x2. */
private int
flip4x2(byte * buffer, const byte ** planes, int offset, int nbytes)
{
    byte *out = buffer;
    const byte *in1 = planes[0] + offset;
    const byte *in2 = planes[1] + offset;
    const byte *in3 = planes[2] + offset;
    const byte *in4 = planes[3] + offset;
    int n = nbytes;

    for (; n > 0; out += 4, ++in1, ++in2, ++in3, ++in4, --n) {
	byte_var b1 = *in1, b2 = *in2, b3 = *in3, b4 = *in4;
	byte_var temp;

	/* Transpose blocks of 4x2 */
	TRANSPOSE(b1, b3, 0x0f, 4);
	TRANSPOSE(b2, b4, 0x0f, 4);
	/* Transpose blocks of 2x1 */
	TRANSPOSE(b1, b2, 0x33, 2);
	TRANSPOSE(b3, b4, 0x33, 2);
	out[0] = b1;
	out[1] = b2;
	out[2] = b3;
	out[3] = b4;
    }
    return 0;
}

/* Convert 4Mx4 to 4x4. */
private int
flip4x4(byte * buffer, const byte ** planes, int offset, int nbytes)
{
    byte *out = buffer;
    const byte *in1 = planes[0] + offset;
    const byte *in2 = planes[1] + offset;
    const byte *in3 = planes[2] + offset;
    const byte *in4 = planes[3] + offset;
    int n = nbytes;

    for (; n > 0; out += 4, ++in1, ++in2, ++in3, ++in4, --n) {
	byte_var b1 = *in1, b2 = *in2, b3 = *in3, b4 = *in4;

	out[0] = (b1 & 0xf0) | (b2 >> 4);
	out[1] = (b3 & 0xf0) | (b4 >> 4);
	out[2] = (byte) ((b1 << 4) | (b2 & 0xf));
	out[3] = (byte) ((b3 << 4) | (b4 & 0xf));
    }
    return 0;
}

/* Convert 4Mx8 to 4x8. */
private int
flip4x8(byte * buffer, const byte ** planes, int offset, int nbytes)
{
    byte *out = buffer;
    const byte *in1 = planes[0] + offset;
    const byte *in2 = planes[1] + offset;
    const byte *in3 = planes[2] + offset;
    const byte *in4 = planes[3] + offset;
    int n = nbytes;

    for (; n > 0; out += 4, ++in1, ++in2, ++in3, ++in4, --n) {
	out[0] = *in1;
	out[1] = *in2;
	out[2] = *in3;
	out[3] = *in4;
    }
    return 0;
}

/* Convert 4Mx12 to 4x12. */
private int
flip4x12(byte * buffer, const byte ** planes, int offset, int nbytes)
{
    byte *out = buffer;
    const byte *pa = planes[0] + offset;
    const byte *pb = planes[1] + offset;
    const byte *pc = planes[2] + offset;
    const byte *pd = planes[3] + offset;
    int n = nbytes;

    /*
     * We assume that the input is an integral number of pixels, and
     * round up n to a multiple of 3.
     */
    for (; n > 0; out += 12, pa += 3, pb += 3, pc += 3, pd += 3, n -= 3) {
	byte_var a1 = pa[1], b1 = pb[1], c1 = pc[1], d1 = pd[1];

	{
	    byte_var v0;

	    out[0] = pa[0];
	    v0 = pb[0];
	    out[1] = (a1 & 0xf0) | (v0 >> 4);
	    out[2] = (byte) ((v0 << 4) | (b1 >> 4));
	    out[3] = pc[0];
	    v0 = pd[0];
	    out[4] = (c1 & 0xf0) | (v0 >> 4);
	    out[5] = (byte) ((v0 << 4) | (d1 >> 4));
	}
	{
	    byte_var v2;

	    v2 = pa[2];
	    out[6] = (byte) ((a1 << 4) | (v2 >> 4));
	    out[7] = (byte) ((v2 << 4) | (b1 & 0xf));
	    out[8] = pb[2];
	    v2 = pc[2];
	    out[9] = (byte) ((c1 << 4) | (v2 >> 4));
	    out[10] = (byte) ((v2 << 4) | (d1 & 0xf));
	    out[11] = pd[2];
	}
    }
    return 0;
}

/* Convert NMx{1,2,4,8} to Nx{1,2,4,8}. */
private int
flipNx1to8(byte * buffer, const byte ** planes, int offset, int nbytes,
	   int num_planes, int bits_per_sample)
{
    /* This is only needed for DeviceN colors, so it can be slow. */
    uint mask = (1 << bits_per_sample) - 1;
    int bi, pi;
    sample_store_declare_setup(dptr, dbit, dbbyte, buffer, 0, bits_per_sample);

    for (bi = 0; bi < nbytes * 8; bi += bits_per_sample) {
	for (pi = 0; pi < num_planes; ++pi) {
	    const byte *sptr = planes[pi] + offset + (bi >> 3);
	    uint value = (*sptr >> (8 - (bi & 7) - bits_per_sample)) & mask;

	    sample_store_next8(value, dptr, dbit, bits_per_sample, dbbyte);
	}
    }
    sample_store_flush(dptr, dbit, bits_per_sample, dbbyte);
    return 0;
}

/* Convert NMx12 to Nx12. */
private int
flipNx12(byte * buffer, const byte ** planes, int offset, int nbytes,
	 int num_planes, int ignore_bits_per_sample)
{
    /* This is only needed for DeviceN colors, so it can be slow. */
    int bi, pi;
    sample_store_declare_setup(dptr, dbit, dbbyte, buffer, 0, 12);

    for (bi = 0; bi < nbytes * 8; bi += 12) {
	for (pi = 0; pi < num_planes; ++pi) {
	    const byte *sptr = planes[pi] + offset + (bi >> 3);
	    uint value =
		(bi & 4 ? ((*sptr & 0xf) << 8) | sptr[1] :
		 (*sptr << 4) | (sptr[1] >> 4));

	    sample_store_next_12(value, dptr, dbit, dbbyte);
	}
    }
    sample_store_flush(dptr, dbit, 12, dbbyte);
    return 0;
}

/* Flip data given number of planes and bits per pixel. */
typedef int (*image_flip_proc) (byte *, const byte **, int, int);
private int
flip_fail(byte * buffer, const byte ** planes, int offset, int nbytes)
{
    return -1;
}
private const image_flip_proc image_flip3_procs[13] = {
    flip_fail, flip3x1, flip3x2, flip_fail, flip3x4,
    flip_fail, flip_fail, flip_fail, flip3x8,
    flip_fail, flip_fail, flip_fail, flip3x12
};
private const image_flip_proc image_flip4_procs[13] = {
    flip_fail, flip4x1, flip4x2, flip_fail, flip4x4,
    flip_fail, flip_fail, flip_fail, flip4x8,
    flip_fail, flip_fail, flip_fail, flip4x12
};
typedef int (*image_flipN_proc) (byte *, const byte **, int, int, int, int);
private int
flipN_fail(byte * buffer, const byte ** planes, int offset, int nbytes,
	   int num_planes, int bits_per_sample)
{
    return -1;
}
private const image_flipN_proc image_flipN_procs[13] = {
    flipN_fail, flipNx1to8, flipNx1to8, flipN_fail, flipNx1to8,
    flipN_fail, flipN_fail, flipN_fail, flipNx1to8,
    flipN_fail, flipN_fail, flipN_fail, flipNx12
};

/* Here is the public interface to all of the above. */
int
image_flip_planes(byte * buffer, const byte ** planes, int offset, int nbytes,
		  int num_planes, int bits_per_sample)
{
    if (bits_per_sample < 1 || bits_per_sample > 12)
	return -1;
    switch (num_planes) {

    case 3:
	return image_flip3_procs[bits_per_sample]
	    (buffer, planes, offset, nbytes);
    case 4:
	return image_flip4_procs[bits_per_sample]
	    (buffer, planes, offset, nbytes);
    default:
	if (num_planes < 0)
	    return -1;
	return image_flipN_procs[bits_per_sample]
	    (buffer, planes, offset, nbytes, num_planes, bits_per_sample);
    }
}