/**

 * $Id$

 *

 * Texture cache. Responsible for maintaining a working set of OpenGL 

 * textures. 

 *

 *

 * Copyright (c) 2005 Nathan Keynes.

 *

 * This program is free software; you can redistribute it and/or modify

 * it under the terms of the GNU General Public License as published by

 * the Free Software Foundation; either version 2 of the License, or

 * (at your option) any later version.

 *

 * This program is distributed in the hope that it will be useful,

 * but WITHOUT ANY WARRANTY; without even the implied warranty of

 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

 * GNU General Public License for more details.

 */

#include <assert.h>

#include "pvr2/pvr2.h"

/** Specifies the maximum number of OpenGL

 * textures we're willing to have open at a time. If more are

 * needed, textures will be evicted in LRU order.

 */

#define MAX_TEXTURES 256

/**

 * Data structure:

 *

 * Main operations:

 *    find entry by texture_addr

 *    add new entry

 *    move entry to tail of lru list

 *    remove entry

 */

typedef signed short texcache_entry_index;

#define EMPTY_ENTRY -1

static texcache_entry_index texcache_free_ptr = 0;

static GLuint texcache_free_list[MAX_TEXTURES];

typedef struct texcache_entry {

    uint32_t texture_addr;

    int width, height, mode;

    GLuint texture_id;

    texcache_entry_index next;

    uint32_t lru_count;

} *texcache_entry_t;

static texcache_entry_index texcache_page_lookup[PVR2_RAM_PAGES];

static uint32_t texcache_ref_counter;

static struct texcache_entry texcache_active_list[MAX_TEXTURES];

/**

 * Initialize the texture cache.

 */

void texcache_init( )

{

    int i;

    for( i=0; i<PVR2_RAM_PAGES; i++ ) {

	texcache_page_lookup[i] = EMPTY_ENTRY;

    }

    for( i=0; i<MAX_TEXTURES; i++ ) {

	texcache_free_list[i] = i;

	texcache_active_list[i].texture_addr = -1;

    }

    texcache_free_ptr = 0;

    texcache_ref_counter = 0;

}

/**

 * Setup the initial texture ids (must be called after the GL context is

 * prepared)

 */

void texcache_gl_init( )

{

    int i;

    GLuint texids[MAX_TEXTURES];

    glGenTextures( MAX_TEXTURES, texids );

    for( i=0; i<MAX_TEXTURES; i++ ) {

	texcache_active_list[i].texture_id = texids[i];

    }

}

/**

 * Flush all textures from the cache, returning them to the free list.

 */

void texcache_flush( )

{

    int i;

    /* clear structures */

    for( i=0; i<PVR2_RAM_PAGES; i++ ) {

	texcache_page_lookup[i] = EMPTY_ENTRY;

    }

    for( i=0; i<MAX_TEXTURES; i++ ) {

	texcache_free_list[i] = i;

    }

    texcache_free_ptr = 0;

    texcache_ref_counter = 0;

}

/**

 * Flush all textures and delete. The cache will be non-functional until

 * the next call to texcache_init(). This would typically be done if

 * switching GL targets.

 */    

void texcache_shutdown( )

{

    GLuint texids[MAX_TEXTURES];

    int i;

    texcache_flush();

    for( i=0; i<MAX_TEXTURES; i++ ) {

	texids[i] = texcache_active_list[i].texture_id;

    }

    glDeleteTextures( MAX_TEXTURES, texids );

}

static void texcache_evict( int slot )

{

    /* Remove the selected slot from the lookup table */

    assert( texcache_active_list[slot].texture_addr != -1 );

    uint32_t evict_page = texcache_active_list[slot].texture_addr >> 12;

    texcache_entry_index replace_next = texcache_active_list[slot].next;

    texcache_active_list[slot].texture_addr = -1;

    texcache_active_list[slot].next = EMPTY_ENTRY; /* Just for safety */

    if( texcache_page_lookup[evict_page] == slot ) {

	texcache_page_lookup[evict_page] = replace_next;

    } else {

	texcache_entry_index idx = texcache_page_lookup[evict_page];

	texcache_entry_index next;

	do {

	    next = texcache_active_list[idx].next;

	    if( next == slot ) {

		assert( idx != replace_next );

		texcache_active_list[idx].next = replace_next;

		break;

	    }

	    idx = next;

	} while( next != EMPTY_ENTRY );

    }

}

/**

 * Evict a single texture from the cache.

 * @return the slot of the evicted texture.

 */

static texcache_entry_index texcache_evict_lru( void )

{

    /* Full table scan - take over the entry with the lowest lru value */

    texcache_entry_index slot = 0;

    int lru_value = texcache_active_list[0].lru_count;

    int i;

    for( i=1; i<MAX_TEXTURES; i++ ) {

	/* FIXME: account for rollover */

	if( texcache_active_list[i].lru_count < lru_value ) {

	    slot = i;

	    lru_value = texcache_active_list[i].lru_count;

	}

    }

    texcache_evict(slot);

    return slot;

}

/**

 * Evict all textures contained in the page identified by a texture address.

 */

void texcache_invalidate_page( uint32_t texture_addr ) {

    uint32_t texture_page = texture_addr >> 12;

    texcache_entry_index idx = texcache_page_lookup[texture_page];

    if( idx == EMPTY_ENTRY )

	return;

    assert( texcache_free_ptr >= 0 );

    do {

	texcache_entry_t entry = &texcache_active_list[idx];

	entry->texture_addr = -1;

	/* release entry */

	texcache_free_ptr--;

	texcache_free_list[texcache_free_ptr] = idx;

	idx = entry->next;

	entry->next = EMPTY_ENTRY;

    } while( idx != EMPTY_ENTRY );

    texcache_page_lookup[texture_page] = EMPTY_ENTRY;

}

/**

 * Mark all textures that use the palette table as needing a re-read (ie 

 * for when the palette is changed. We could track exactly which ones are 

 * affected, but it's not clear that the extra maintanence overhead is 

 * worthwhile.

 */

void texcache_invalidate_palette( )

{

    int i;

    for( i=0; i<MAX_TEXTURES; i++ ) {

	if( texcache_active_list[i].texture_addr != -1 &&

	    PVR2_TEX_IS_PALETTE(texcache_active_list[i].mode) ) {

	    texcache_evict( i );

	    texcache_free_ptr--;

	    texcache_free_list[texcache_free_ptr] = i;

	}

    }

}

static void decode_pal8_to_32( uint32_t *out, uint8_t *in, int inbytes, uint32_t *pal )

{

    int i;

    for( i=0; i<inbytes; i++ ) {

	*out++ = pal[*in++];

    }

}

static void decode_pal8_to_16( uint16_t *out, uint8_t *in, int inbytes, uint32_t *pal )

{

    int i;

    for( i=0; i<inbytes; i++ ) {

	*out++ = (uint16_t)pal[*in++];

    }

}

static void decode_pal4_to_32( uint32_t *out, uint8_t *in, int inbytes, uint32_t *pal )

{

    int i;

    for( i=0; i<inbytes; i++ ) {

	*out++ = pal[*in & 0x0F];

	*out++ = pal[(*in >> 4)];

	in++;

    }

}

static void decode_pal4_to_16( uint16_t *out, uint8_t *in, int inbytes, uint32_t *pal )

{

    int i;

    for( i=0; i<inbytes; i++ ) {

	*out++ = (uint16_t)pal[*in & 0x0F];

	*out++ = (uint16_t)pal[(*in >> 4)];

	in++;

    }

}

#define VQ_CODEBOOK_SIZE 2048 /* 256 entries * 4 pixels per quad * 2 byte pixels */

struct vq_codebook {

    uint16_t quad[256][4];

};

static void vq_get_codebook( struct vq_codebook *codebook, 

				uint16_t *input )

{

    /* Detwiddle the codebook, for the sake of my own sanity if nothing else */

    uint16_t *p = (uint16_t *)input;

    int i;

    for( i=0; i<256; i++ ) {

	codebook->quad[i][0] = *p++;

	codebook->quad[i][2] = *p++;

	codebook->quad[i][1] = *p++;

	codebook->quad[i][3] = *p++;

    }

}    

static void vq_decode( uint16_t *output, unsigned char *input, int width, int height, 

		       struct vq_codebook *codebook ) {

    int i,j;

    uint8_t *c = (uint8_t *)input;

    for( j=0; j<height; j+=2 ) {

	for( i=0; i<width; i+=2 ) {

	    uint8_t code = *c++;

	    output[i + j*width] = codebook->quad[code][0];

	    output[i + 1 + j*width] = codebook->quad[code][1];

	    output[i + (j+1)*width] = codebook->quad[code][2];

	    output[i + 1 + (j+1)*width] = codebook->quad[code][3];

	}

    }

}

static inline uint32_t yuv_to_rgb32( float y, float u, float v )

{

    u -= 128;

    v -= 128;

    int r = (int)(y + v*1.375);

    int g = (int)(y - u*0.34375 - v*0.6875);

    int b = (int)(y + u*1.71875);

    if( r > 255 ) { r = 255; } else if( r < 0 ) { r = 0; }

    if( g > 255 ) { g = 255; } else if( g < 0 ) { g = 0; }

    if( b > 255 ) { b = 255; } else if( b < 0 ) { b = 0; }

    return 0xFF000000 | (r<<16) | (g<<8) | (b);

}

/**

 * Convert raster YUV texture data into RGB32 data - most GL implementations don't

 * directly support this format unfortunately. The input data is formatted as

 * 32 bits = 2 horizontal pixels, UYVY. This is currently done rather inefficiently

 * in floating point.

 */

static void yuv_decode( uint32_t *output, uint32_t *input, int width, int height )

{

    int x, y;

    uint32_t *p = input;

    for( y=0; y<height; y++ ) {

	for( x=0; x<width; x+=2 ) {

	    float u = (float)(*p & 0xFF);

	    float y0 = (float)( (*p>>8)&0xFF );

	    float v = (float)( (*p>>16)&0xFF );

	    float y1 = (float)( (*p>>24)&0xFF );

	    *output++ = yuv_to_rgb32( y0, u, v ); 

	    *output++ = yuv_to_rgb32( y1, u, v );

	    p++;

	}

    }

}

/**

 * Load texture data from the given address and parameters into the currently

 * bound OpenGL texture.

 */

static void texcache_load_texture( uint32_t texture_addr, int width, int height,

			      int mode ) {

    int bpp_shift = 1; /* bytes per (output) pixel as a power of 2 */

    GLint intFormat = GL_RGBA, format, type;

    int tex_format = mode & PVR2_TEX_FORMAT_MASK;

    struct vq_codebook codebook;

    GLint filter = GL_LINEAR;

    glPixelStorei( GL_UNPACK_ROW_LENGTH, 0 );

    /* Decode the format parameters */

    switch( tex_format ) {

    case PVR2_TEX_FORMAT_IDX4:

    case PVR2_TEX_FORMAT_IDX8:

	/* For indexed-colour modes, we need to lookup the palette control

	 * word to determine the de-indexed texture format.

	 */

	switch( MMIO_READ( PVR2, RENDER_PALETTE ) & 0x03 ) {

	case 0: /* ARGB1555 */

	    format = GL_BGRA;

	    type = GL_UNSIGNED_SHORT_1_5_5_5_REV;

	    break;

	case 1:  /* RGB565 */

	    intFormat = GL_RGB;

	    format = GL_RGB;

	    type = GL_UNSIGNED_SHORT_5_6_5;

	    break;

	case 2: /* ARGB4444 */

	    format = GL_BGRA;

	    type = GL_UNSIGNED_SHORT_4_4_4_4_REV;

	    break;

	case 3: /* ARGB8888 */

	    format = GL_BGRA;

	    type = GL_UNSIGNED_BYTE;

	    bpp_shift = 2;

	    break;

	default:

	    return; /* Can't happen, but it makes gcc stop complaining */

	}

	break;

    case PVR2_TEX_FORMAT_ARGB1555:

	format = GL_BGRA;

	type = GL_UNSIGNED_SHORT_1_5_5_5_REV;

	break;

    case PVR2_TEX_FORMAT_RGB565:

	intFormat = GL_RGB;

	format = GL_RGB;

	type = GL_UNSIGNED_SHORT_5_6_5;

	break;

    case PVR2_TEX_FORMAT_ARGB4444:

	format = GL_BGRA;

	type = GL_UNSIGNED_SHORT_4_4_4_4_REV;

	break;

    case PVR2_TEX_FORMAT_YUV422:

	/* YUV422 isn't directly supported by most implementations, so decode

	 * it to a (reasonably) standard ARGB32.

	 */

	bpp_shift = 2;

	format = GL_BGRA;

	type = GL_UNSIGNED_BYTE;

	break;

    case PVR2_TEX_FORMAT_BUMPMAP:

	ERROR( "Bumpmap not supported" );

	return;

    default:

	ERROR( "Undefined texture format" );

	return;

    }

    if( PVR2_TEX_IS_STRIDE(mode) && tex_format != PVR2_TEX_FORMAT_IDX4 &&

	tex_format != PVR2_TEX_FORMAT_IDX8 ) {

	/* Stride textures cannot be mip-mapped, compressed, indexed or twiddled */

	uint32_t stride = (MMIO_READ( PVR2, RENDER_TEXSIZE ) & 0x003F) << 5;

	unsigned char data[(width*height) << bpp_shift];

	if( tex_format == PVR2_TEX_FORMAT_YUV422 ) {

	    unsigned char tmp[(width*height)<<1];

	    pvr2_vram64_read_stride( tmp, width<<1, texture_addr, stride<<1, height );

	    yuv_decode( (uint32_t *)data, (uint32_t *)tmp, width, height );

	} else {

	    pvr2_vram64_read_stride( data, width<<bpp_shift, texture_addr, stride<<bpp_shift, height );

	}

	glTexImage2D( GL_TEXTURE_2D, 0, intFormat, width, height, 0, format, type, data );

	glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, filter);

	glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);

	return;

    } 

    int level=0, last_level = 0, mip_width = width, mip_height = height, src_bytes, dest_bytes;

    if( PVR2_TEX_IS_MIPMAPPED(mode) ) {

	int i;

	for( i=0; 1<<i < width; i++ );

	last_level = i;

	mip_width = 2;

	mip_height= 2;

	filter = GL_LINEAR_MIPMAP_LINEAR;

    }

    dest_bytes = (mip_width * mip_height) << bpp_shift;

    src_bytes = dest_bytes; // Modes will change this (below)

    if( PVR2_TEX_IS_COMPRESSED(mode) ) {

	uint16_t tmp[VQ_CODEBOOK_SIZE];

	pvr2_vram64_read( (unsigned char *)tmp, texture_addr, VQ_CODEBOOK_SIZE );

	texture_addr += VQ_CODEBOOK_SIZE;

	vq_get_codebook( &codebook, tmp );

    }

    for( level=last_level; level>= 0; level-- ) {

	unsigned char data[dest_bytes];

	/* load data from image, detwiddling/uncompressing as required */

	if( tex_format == PVR2_TEX_FORMAT_IDX8 ) {

	    src_bytes = (mip_width * mip_height);

	    int bank = (mode >> 25) &0x03;

	    uint32_t *palette = ((uint32_t *)mmio_region_PVR2PAL.mem) + (bank<<8);

	    unsigned char tmp[src_bytes];

	    pvr2_vram64_read_twiddled_8( tmp, texture_addr, mip_width, mip_height );

	    if( bpp_shift == 2 ) {

		decode_pal8_to_32( (uint32_t *)data, tmp, src_bytes, palette );

	    } else {

		decode_pal8_to_16( (uint16_t *)data, tmp, src_bytes, palette );

	    }

	} else if( tex_format == PVR2_TEX_FORMAT_IDX4 ) {

	    src_bytes = (mip_width * mip_height) >> 1;

	    int bank = (mode >>21 ) & 0x3F;

	    uint32_t *palette = ((uint32_t *)mmio_region_PVR2PAL.mem) + (bank<<4);

	    unsigned char tmp[src_bytes];

	    pvr2_vram64_read_twiddled_4( tmp, texture_addr, mip_width, mip_height );

	    if( bpp_shift == 2 ) {

		decode_pal4_to_32( (uint32_t *)data, tmp, src_bytes, palette );

	    } else {

		decode_pal4_to_16( (uint16_t *)data, tmp, src_bytes, palette );

	    }

	} else if( tex_format == PVR2_TEX_FORMAT_YUV422 ) {

	    src_bytes = ((mip_width*mip_height)<<1);

	    unsigned char tmp[src_bytes];

	    if( PVR2_TEX_IS_TWIDDLED(mode) ) {

		pvr2_vram64_read_twiddled_16( tmp, texture_addr, mip_width, mip_height );

	    } else {

		pvr2_vram64_read( tmp, texture_addr, src_bytes );

	    }

	    yuv_decode( (uint32_t *)data, (uint32_t *)tmp, mip_width, mip_height );

	} else if( PVR2_TEX_IS_COMPRESSED(mode) ) {

	    src_bytes = ((mip_width*mip_height) >> 2);

	    unsigned char tmp[src_bytes];

	    if( PVR2_TEX_IS_TWIDDLED(mode) ) {

		pvr2_vram64_read_twiddled_8( tmp, texture_addr, mip_width>>1, mip_height>>1 );

	    } else {

		pvr2_vram64_read( tmp, texture_addr, src_bytes );

	    }

	    vq_decode( (uint16_t *)data, tmp, mip_width, mip_height, &codebook );

	} else if( PVR2_TEX_IS_TWIDDLED(mode) ) {

	    pvr2_vram64_read_twiddled_16( data, texture_addr, mip_width, mip_height );

	} else {

	    pvr2_vram64_read( data, texture_addr, src_bytes );

	}

	/* Pass to GL */

	if( level == last_level && level != 0 ) { /* 1x1 stored within a 2x2 */

	    glTexImage2D( GL_TEXTURE_2D, level, intFormat, 1, 1, 0, format, type,

			  data + (3 << bpp_shift) );

	    texture_addr += src_bytes;

	} else {

	    glTexImage2D( GL_TEXTURE_2D, level, intFormat, mip_width, mip_height, 0, format, type,

			  data );

	    texture_addr += src_bytes;

	    mip_width <<= 1;

	    mip_height <<= 1;

	    dest_bytes <<= 2;

	    src_bytes <<= 2;

	}

    }

    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, filter);

    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);

}

/**

 * Return a texture ID for the texture specified at the supplied address

 * and given parameters (the same sequence of bytes could in theory have

 * multiple interpretations). We use the texture address as the primary

 * index, but allow for multiple instances at each address. The texture

 * will be bound to the GL_TEXTURE_2D target before being returned.

 * 

 * If the texture has already been bound, return the ID to which it was

 * bound. Otherwise obtain an unused texture ID and set it up appropriately.

 */

GLuint texcache_get_texture( uint32_t texture_addr, int width, int height,

			     int mode )

{

    uint32_t texture_page = texture_addr >> 12;

    texcache_entry_index next;

    texcache_entry_index idx = texcache_page_lookup[texture_page];

    while( idx != EMPTY_ENTRY ) {

	texcache_entry_t entry = &texcache_active_list[idx];

	if( entry->texture_addr == texture_addr &&

	    entry->mode == mode &&

	    entry->width == width &&

	    entry->height == height ) {

	    entry->lru_count = texcache_ref_counter++;

	    glBindTexture( GL_TEXTURE_2D, entry->texture_id );

	    return entry->texture_id;

	}

        idx = entry->next;

    }

    /* Not found - check the free list */

    texcache_entry_index slot = 0;

    if( texcache_free_ptr < MAX_TEXTURES ) {

	slot = texcache_free_list[texcache_free_ptr++];

    } else {

	slot = texcache_evict_lru();

    }

    /* Construct new entry */

    texcache_active_list[slot].texture_addr = texture_addr;

    texcache_active_list[slot].width = width;

    texcache_active_list[slot].height = height;

    texcache_active_list[slot].mode = mode;

    texcache_active_list[slot].lru_count = texcache_ref_counter++;

    /* Add entry to the lookup table */

    next = texcache_page_lookup[texture_page];

    if( next == slot ) {

	int i;

	fprintf( stderr, "Active list: " );

	for( i=0; i<MAX_TEXTURES; i++ ) {

	    fprintf( stderr, "%d, ", texcache_active_list[i].next );

	}

	fprintf( stderr, "\n" );

	assert( next != slot );

    }

    assert( next != slot );

    texcache_active_list[slot].next = next;

    texcache_page_lookup[texture_page] = slot;

    /* Construct the GL texture */

    glBindTexture( GL_TEXTURE_2D, texcache_active_list[slot].texture_id );

    texcache_load_texture( texture_addr, width, height, mode );

    return texcache_active_list[slot].texture_id;

}