nkeynes@284: /**
nkeynes@561:  * $Id$
nkeynes@284:  *
nkeynes@284:  * PVR2 (Video) VRAM handling routines (mainly for the 64-bit region)
nkeynes@284:  *
nkeynes@284:  * Copyright (c) 2005 Nathan Keynes.
nkeynes@284:  *
nkeynes@284:  * This program is free software; you can redistribute it and/or modify
nkeynes@284:  * it under the terms of the GNU General Public License as published by
nkeynes@284:  * the Free Software Foundation; either version 2 of the License, or
nkeynes@284:  * (at your option) any later version.
nkeynes@284:  *
nkeynes@284:  * This program is distributed in the hope that it will be useful,
nkeynes@284:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
nkeynes@284:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
nkeynes@284:  * GNU General Public License for more details.
nkeynes@284:  */
nkeynes@645: #include <string.h>
nkeynes@645: #include <stdio.h>
nkeynes@645: #include <errno.h>
nkeynes@929: #include "sh4/sh4core.h"
nkeynes@284: #include "pvr2.h"
nkeynes@325: #include "asic.h"
nkeynes@669: #include "dream.h"
nkeynes@284: 
nkeynes@429: extern unsigned char *video_base;
nkeynes@284: 
nkeynes@929: /************************* VRAM32 address space ***************************/
nkeynes@929: 
nkeynes@929: static int32_t FASTCALL pvr2_vram32_read_long( sh4addr_t addr )
nkeynes@929: {
nkeynes@929:     pvr2_render_buffer_invalidate(addr, FALSE);
nkeynes@929:     return *((int32_t *)(video_base+(addr&0x007FFFFF)));
nkeynes@929: }
nkeynes@929: static int32_t FASTCALL pvr2_vram32_read_word( sh4addr_t addr )
nkeynes@929: {
nkeynes@929:     pvr2_render_buffer_invalidate(addr, FALSE);
nkeynes@929:     return SIGNEXT16(*((int16_t *)(video_base+(addr&0x007FFFFF))));
nkeynes@929: }
nkeynes@929: static int32_t FASTCALL pvr2_vram32_read_byte( sh4addr_t addr )
nkeynes@929: {
nkeynes@929:     pvr2_render_buffer_invalidate(addr, FALSE);
nkeynes@929:     return SIGNEXT8(*((int8_t *)(video_base+(addr&0x007FFFFF))));
nkeynes@929: }
nkeynes@929: static void FASTCALL pvr2_vram32_write_long( sh4addr_t addr, uint32_t val )
nkeynes@929: {
nkeynes@929:     pvr2_render_buffer_invalidate(addr, TRUE);
nkeynes@929:     *(uint32_t *)(video_base + (addr&0x007FFFFF)) = val;
nkeynes@929: }
nkeynes@929: static void FASTCALL pvr2_vram32_write_word( sh4addr_t addr, uint32_t val )
nkeynes@929: {
nkeynes@929:     pvr2_render_buffer_invalidate(addr, TRUE);
nkeynes@929:     *(uint16_t *)(video_base + (addr&0x007FFFFF)) = (uint16_t)val;
nkeynes@929: }
nkeynes@929: static void FASTCALL pvr2_vram32_write_byte( sh4addr_t addr, uint32_t val )
nkeynes@929: {
nkeynes@929:     pvr2_render_buffer_invalidate(addr, TRUE);
nkeynes@929:     *(uint8_t *)(video_base + (addr&0x007FFFFF)) = (uint8_t)val;
nkeynes@929: }
nkeynes@929: static void FASTCALL pvr2_vram32_read_burst( unsigned char *dest, sh4addr_t addr )
nkeynes@929: {
nkeynes@929:     // Render buffers pretty much have to be (at least) 32-byte aligned
nkeynes@929:     pvr2_render_buffer_invalidate(addr, FALSE);
nkeynes@929:     memcpy( dest, (video_base + (addr&0x007FFFFF)), 32 );
nkeynes@929: }
nkeynes@929: static void FASTCALL pvr2_vram32_write_burst( sh4addr_t addr, unsigned char *src )
nkeynes@929: {
nkeynes@929:     // Render buffers pretty much have to be (at least) 32-byte aligned
nkeynes@929:     pvr2_render_buffer_invalidate(addr, TRUE);
nkeynes@929:     memcpy( (video_base + (addr&0x007FFFFF)), src, 32 );    
nkeynes@929: }
nkeynes@929: 
nkeynes@929: struct mem_region_fn mem_region_vram32 = { pvr2_vram32_read_long, pvr2_vram32_write_long, 
nkeynes@929:         pvr2_vram32_read_word, pvr2_vram32_write_word, 
nkeynes@929:         pvr2_vram32_read_byte, pvr2_vram32_write_byte, 
nkeynes@929:         pvr2_vram32_read_burst, pvr2_vram32_write_burst }; 
nkeynes@929: 
nkeynes@929: /************************* VRAM64 address space ***************************/
nkeynes@929: 
nkeynes@929: #define TRANSLATE_VIDEO_64BIT_ADDRESS(a)  ( (((a)&0x00FFFFF8)>>1)|(((a)&0x00000004)<<20)|((a)&0x03) )
nkeynes@929: 
nkeynes@929: static int32_t FASTCALL pvr2_vram64_read_long( sh4addr_t addr )
nkeynes@929: {
nkeynes@929:     addr = TRANSLATE_VIDEO_64BIT_ADDRESS(addr);
nkeynes@929:     pvr2_render_buffer_invalidate(addr, FALSE);
nkeynes@929:     return *((int32_t *)(video_base+(addr&0x007FFFFF)));
nkeynes@929: }
nkeynes@929: static int32_t FASTCALL pvr2_vram64_read_word( sh4addr_t addr )
nkeynes@929: {
nkeynes@929:     addr = TRANSLATE_VIDEO_64BIT_ADDRESS(addr);
nkeynes@929:     pvr2_render_buffer_invalidate(addr, FALSE);
nkeynes@929:     return SIGNEXT16(*((int16_t *)(video_base+(addr&0x007FFFFF))));
nkeynes@929: }
nkeynes@929: static int32_t FASTCALL pvr2_vram64_read_byte( sh4addr_t addr )
nkeynes@929: {
nkeynes@929:     addr = TRANSLATE_VIDEO_64BIT_ADDRESS(addr);
nkeynes@929:     pvr2_render_buffer_invalidate(addr, FALSE);
nkeynes@929:     return SIGNEXT8(*((int8_t *)(video_base+(addr&0x007FFFFF))));
nkeynes@929: }
nkeynes@929: static void FASTCALL pvr2_vram64_write_long( sh4addr_t addr, uint32_t val )
nkeynes@929: {
nkeynes@929:     texcache_invalidate_page(addr& 0x007FFFFF);
nkeynes@929:     addr = TRANSLATE_VIDEO_64BIT_ADDRESS(addr);
nkeynes@929:     pvr2_render_buffer_invalidate(addr, TRUE);
nkeynes@929:     *(uint32_t *)(video_base + (addr&0x007FFFFF)) = val;
nkeynes@929: }
nkeynes@929: static void FASTCALL pvr2_vram64_write_word( sh4addr_t addr, uint32_t val )
nkeynes@929: {
nkeynes@929:     texcache_invalidate_page(addr& 0x007FFFFF);
nkeynes@929:     addr = TRANSLATE_VIDEO_64BIT_ADDRESS(addr);
nkeynes@929:     pvr2_render_buffer_invalidate(addr, TRUE);
nkeynes@929:     *(uint16_t *)(video_base + (addr&0x007FFFFF)) = (uint16_t)val;
nkeynes@929: }
nkeynes@929: static void FASTCALL pvr2_vram64_write_byte( sh4addr_t addr, uint32_t val )
nkeynes@929: {
nkeynes@929:     texcache_invalidate_page(addr& 0x007FFFFF);
nkeynes@929:     addr = TRANSLATE_VIDEO_64BIT_ADDRESS(addr);
nkeynes@929:     pvr2_render_buffer_invalidate(addr, TRUE);
nkeynes@929:     *(uint8_t *)(video_base + (addr&0x007FFFFF)) = (uint8_t)val;
nkeynes@929: }
nkeynes@929: static void FASTCALL pvr2_vram64_read_burst( unsigned char *dest, sh4addr_t addr )
nkeynes@929: {
nkeynes@929:     pvr2_vram64_read( dest, addr, 32 );
nkeynes@929: }
nkeynes@929: static void FASTCALL pvr2_vram64_write_burst( sh4addr_t addr, unsigned char *src )
nkeynes@929: {
nkeynes@929:     pvr2_vram64_write( addr, src, 32 );
nkeynes@929: }
nkeynes@929: 
nkeynes@929: struct mem_region_fn mem_region_vram64 = { pvr2_vram64_read_long, pvr2_vram64_write_long, 
nkeynes@929:         pvr2_vram64_read_word, pvr2_vram64_write_word, 
nkeynes@929:         pvr2_vram64_read_byte, pvr2_vram64_write_byte, 
nkeynes@929:         pvr2_vram64_read_burst, pvr2_vram64_write_burst }; 
nkeynes@929: 
nkeynes@929: 
nkeynes@929: 
nkeynes@429: void pvr2_dma_write( sh4addr_t destaddr, unsigned char *src, uint32_t count )
nkeynes@325: {
nkeynes@325:     int region;
nkeynes@325: 
nkeynes@325:     switch( destaddr & 0x13800000 ) {
nkeynes@325:     case 0x10000000:
nkeynes@325:     case 0x12000000:
nkeynes@736:         pvr2_ta_write( src, count );
nkeynes@736:         break;
nkeynes@325:     case 0x11000000:
nkeynes@325:     case 0x11800000:
nkeynes@736:         region = MMIO_READ( ASIC, PVRDMARGN1 );
nkeynes@736:         if( region == 0 ) {
nkeynes@736:             pvr2_vram64_write( destaddr, src, count );
nkeynes@736:         } else {
nkeynes@834:             destaddr &= PVR2_RAM_MASK;
nkeynes@834:             unsigned char *dest = video_base + destaddr;
nkeynes@834:             if( PVR2_RAM_SIZE - destaddr < count ) {
nkeynes@834:                 count = PVR2_RAM_SIZE - destaddr;
nkeynes@834:             }
nkeynes@736:             memcpy( dest, src, count );
nkeynes@736:         }
nkeynes@736:         break;
nkeynes@325:     case 0x10800000:
nkeynes@325:     case 0x12800000:
nkeynes@736:         pvr2_yuv_write( src, count );
nkeynes@736:         break;
nkeynes@325:     case 0x13000000:
nkeynes@325:     case 0x13800000:
nkeynes@736:         region = MMIO_READ( ASIC, PVRDMARGN2 );
nkeynes@736:         if( region == 0 ) {
nkeynes@736:             pvr2_vram64_write( destaddr, src, count );
nkeynes@736:         } else {
nkeynes@834:             destaddr &= PVR2_RAM_MASK;
nkeynes@834:             unsigned char *dest = video_base + destaddr;
nkeynes@834:             if( PVR2_RAM_SIZE - destaddr < count ) {
nkeynes@834:                 count = PVR2_RAM_SIZE - destaddr;
nkeynes@834:             }
nkeynes@736:             memcpy( dest, src, count );
nkeynes@736:         }
nkeynes@827:     }
nkeynes@325: }
nkeynes@325: 
nkeynes@429: void pvr2_vram64_write( sh4addr_t destaddr, unsigned char *src, uint32_t length )
nkeynes@284: {
nkeynes@284:     int bank_flag = (destaddr & 0x04) >> 2;
nkeynes@284:     uint32_t *banks[2];
nkeynes@284:     uint32_t *dwsrc;
nkeynes@284:     int i;
nkeynes@284: 
nkeynes@284:     destaddr = destaddr & 0x7FFFFF;
nkeynes@284:     if( destaddr + length > 0x800000 ) {
nkeynes@736:         length = 0x800000 - destaddr;
nkeynes@284:     }
nkeynes@284: 
nkeynes@796:     for( i=destaddr & 0xFFFFF000; i < destaddr + length; i+= LXDREAM_PAGE_SIZE ) {
nkeynes@736:         texcache_invalidate_page( i );
nkeynes@284:     }
nkeynes@284: 
nkeynes@284:     banks[0] = ((uint32_t *)(video_base + ((destaddr & 0x007FFFF8) >>1)));
nkeynes@284:     banks[1] = banks[0] + 0x100000;
nkeynes@827:     if( bank_flag )
nkeynes@736:         banks[0]++;
nkeynes@736: 
nkeynes@284:     /* Handle non-aligned start of source */
nkeynes@284:     if( destaddr & 0x03 ) {
nkeynes@736:         unsigned char *dest = ((unsigned char *)banks[bank_flag]) + (destaddr & 0x03);
nkeynes@736:         for( i= destaddr & 0x03; i < 4 && length > 0; i++, length-- ) {
nkeynes@736:             *dest++ = *src++;
nkeynes@736:         }
nkeynes@736:         bank_flag = !bank_flag;
nkeynes@284:     }
nkeynes@284: 
nkeynes@284:     dwsrc = (uint32_t *)src;
nkeynes@284:     while( length >= 4 ) {
nkeynes@736:         *banks[bank_flag]++ = *dwsrc++;
nkeynes@736:         bank_flag = !bank_flag;
nkeynes@736:         length -= 4;
nkeynes@284:     }
nkeynes@736: 
nkeynes@284:     /* Handle non-aligned end of source */
nkeynes@284:     if( length ) {
nkeynes@736:         src = (unsigned char *)dwsrc;
nkeynes@736:         unsigned char *dest = (unsigned char *)banks[bank_flag];
nkeynes@736:         while( length-- > 0 ) {
nkeynes@736:             *dest++ = *src++;
nkeynes@736:         }
nkeynes@827:     }
nkeynes@284: }
nkeynes@284: 
nkeynes@284: /**
nkeynes@284:  * Write an image to 64-bit vram, with a line-stride different from the line-size.
nkeynes@869:  * The destaddr must be 64-bit aligned, and both line_bytes and line_stride_bytes
nkeynes@869:  * must be multiples of 8.
nkeynes@284:  */
nkeynes@827: void pvr2_vram64_write_stride( sh4addr_t destaddr, unsigned char *src, uint32_t line_bytes,
nkeynes@736:                                uint32_t line_stride_bytes, uint32_t line_count )
nkeynes@284: {
nkeynes@869:     int i,j;
nkeynes@284:     uint32_t *banks[2];
nkeynes@869:     uint32_t *dwsrc = (uint32_t *)src;
nkeynes@869:     uint32_t line_gap = (line_stride_bytes - line_bytes) >> 3;
nkeynes@284: 
nkeynes@284:     destaddr = destaddr & 0x7FFFF8;
nkeynes@869:     line_bytes >>= 3;
nkeynes@284: 
nkeynes@869:     for( i=destaddr; i < destaddr + line_stride_bytes*line_count; i+= LXDREAM_PAGE_SIZE ) {
nkeynes@736:         texcache_invalidate_page( i );
nkeynes@284:     }
nkeynes@284: 
nkeynes@284:     banks[0] = (uint32_t *)(video_base + (destaddr >>1));
nkeynes@284:     banks[1] = banks[0] + 0x100000;
nkeynes@736: 
nkeynes@284:     for( i=0; i<line_count; i++ ) {
nkeynes@736:         for( j=0; j<line_bytes; j++ ) {
nkeynes@869:             *banks[0]++ = *dwsrc++;
nkeynes@869:             *banks[1]++ = *dwsrc++;
nkeynes@736:         }
nkeynes@736:         banks[0] += line_gap;
nkeynes@736:         banks[1] += line_gap;
nkeynes@827:     }
nkeynes@284: }
nkeynes@284: 
nkeynes@284: /**
nkeynes@284:  * Read an image from 64-bit vram, with a destination line-stride different from the line-size.
nkeynes@284:  * The srcaddr must be 32-bit aligned, and both line_bytes and line_stride_bytes
nkeynes@284:  * must be multiples of 4. line_stride_bytes must be >= line_bytes.
nkeynes@284:  * This method is used to extract a "stride" texture from vram.
nkeynes@284:  */
nkeynes@429: void pvr2_vram64_read_stride( unsigned char *dest, uint32_t dest_line_bytes, sh4addr_t srcaddr,
nkeynes@736:                               uint32_t src_line_bytes, uint32_t line_count )
nkeynes@284: {
nkeynes@284:     int bank_flag = (srcaddr & 0x04) >> 2;
nkeynes@284:     uint32_t *banks[2];
nkeynes@284:     uint32_t *dwdest;
nkeynes@429:     uint32_t dest_line_gap = 0;
nkeynes@429:     uint32_t src_line_gap = 0;
nkeynes@284:     uint32_t line_bytes;
nkeynes@284:     int src_line_gap_flag;
nkeynes@284:     int i,j;
nkeynes@284: 
nkeynes@284:     srcaddr = srcaddr & 0x7FFFF8;
nkeynes@284:     if( src_line_bytes <= dest_line_bytes ) {
nkeynes@736:         dest_line_gap = (dest_line_bytes - src_line_bytes) >> 2;
nkeynes@736:         src_line_gap = 0;
nkeynes@736:         src_line_gap_flag = 0;
nkeynes@736:         line_bytes = src_line_bytes >> 2;
nkeynes@284:     } else {
nkeynes@736:         i = (src_line_bytes - dest_line_bytes);
nkeynes@736:         src_line_gap_flag = i & 0x04;
nkeynes@736:         src_line_gap = i >> 3;
nkeynes@736:         line_bytes = dest_line_bytes >> 2;
nkeynes@284:     }
nkeynes@736: 
nkeynes@284:     banks[0] = (uint32_t *)(video_base + (srcaddr>>1));
nkeynes@284:     banks[1] = banks[0] + 0x100000;
nkeynes@284:     if( bank_flag )
nkeynes@736:         banks[0]++;
nkeynes@736: 
nkeynes@284:     dwdest = (uint32_t *)dest;
nkeynes@284:     for( i=0; i<line_count; i++ ) {
nkeynes@736:         for( j=0; j<line_bytes; j++ ) {
nkeynes@736:             *dwdest++ = *banks[bank_flag]++;
nkeynes@736:             bank_flag = !bank_flag;
nkeynes@736:         }
nkeynes@736:         dwdest += dest_line_gap;
nkeynes@736:         banks[0] += src_line_gap;
nkeynes@736:         banks[1] += src_line_gap;
nkeynes@736:         if( src_line_gap_flag ) {
nkeynes@736:             banks[bank_flag]++;
nkeynes@736:             bank_flag = !bank_flag;
nkeynes@736:         }
nkeynes@827:     }
nkeynes@310: }
nkeynes@310: 
nkeynes@315: 
nkeynes@310: /**
nkeynes@310:  * @param dest Destination image buffer
nkeynes@310:  * @param banks Source data expressed as two bank pointers
nkeynes@310:  * @param offset Offset into banks[0] specifying where the next byte
nkeynes@310:  *  to read is (0..3)
nkeynes@310:  * @param x1,y1 Destination coordinates
nkeynes@310:  * @param width Width of current destination block
nkeynes@315:  * @param stride Total width of image (ie stride) in bytes
nkeynes@315:  */
nkeynes@315: 
nkeynes@315: static void pvr2_vram64_detwiddle_4( uint8_t *dest, uint8_t *banks[2], int offset,
nkeynes@736:                                      int x1, int y1, int width, int stride )
nkeynes@315: {
nkeynes@315:     if( width == 2 ) {
nkeynes@736:         x1 = x1 >> 1;
nkeynes@736:         uint8_t t1 = *banks[offset<4?0:1]++;
nkeynes@736:         uint8_t t2 = *banks[offset<3?0:1]++;
nkeynes@736:         dest[y1*stride + x1] = (t1 & 0x0F) | (t2<<4);
nkeynes@736:         dest[(y1+1)*stride + x1] = (t1>>4) | (t2&0xF0);
nkeynes@315:     } else if( width == 4 ) {
nkeynes@736:         pvr2_vram64_detwiddle_4( dest, banks, offset, x1, y1, 2, stride );
nkeynes@736:         pvr2_vram64_detwiddle_4( dest, banks, offset+2, x1, y1+2, 2, stride );
nkeynes@736:         pvr2_vram64_detwiddle_4( dest, banks, offset+4, x1+2, y1, 2, stride );
nkeynes@736:         pvr2_vram64_detwiddle_4( dest, banks, offset+6, x1+2, y1+2, 2, stride );
nkeynes@736: 
nkeynes@315:     } else {
nkeynes@736:         int subdivide = width >> 1;
nkeynes@736:         pvr2_vram64_detwiddle_4( dest, banks, offset, x1, y1, subdivide, stride );
nkeynes@736:         pvr2_vram64_detwiddle_4( dest, banks, offset, x1, y1+subdivide, subdivide, stride );
nkeynes@736:         pvr2_vram64_detwiddle_4( dest, banks, offset, x1+subdivide, y1, subdivide, stride );
nkeynes@736:         pvr2_vram64_detwiddle_4( dest, banks, offset, x1+subdivide, y1+subdivide, subdivide, stride );
nkeynes@315:     }
nkeynes@315: }
nkeynes@315: 
nkeynes@315: /**
nkeynes@315:  * @param dest Destination image buffer
nkeynes@315:  * @param banks Source data expressed as two bank pointers
nkeynes@315:  * @param offset Offset into banks[0] specifying where the next byte
nkeynes@315:  *  to read is (0..3)
nkeynes@315:  * @param x1,y1 Destination coordinates
nkeynes@315:  * @param width Width of current destination block
nkeynes@315:  * @param stride Total width of image (ie stride)
nkeynes@310:  */
nkeynes@310: 
nkeynes@310: static void pvr2_vram64_detwiddle_8( uint8_t *dest, uint8_t *banks[2], int offset,
nkeynes@736:                                      int x1, int y1, int width, int stride )
nkeynes@310: {
nkeynes@310:     if( width == 2 ) {
nkeynes@736:         dest[y1*stride + x1] = *banks[0]++;
nkeynes@736:         dest[(y1+1)*stride + x1] = *banks[offset<3?0:1]++;
nkeynes@736:         dest[y1*stride + x1 + 1] = *banks[offset<2?0:1]++;
nkeynes@736:         dest[(y1+1)*stride + x1 + 1] = *banks[offset==0?0:1]++;
nkeynes@736:         uint8_t *tmp = banks[0]; /* swap banks */
nkeynes@736:         banks[0] = banks[1];
nkeynes@736:         banks[1] = tmp;
nkeynes@310:     } else {
nkeynes@736:         int subdivide = width >> 1;
nkeynes@736:         pvr2_vram64_detwiddle_8( dest, banks, offset, x1, y1, subdivide, stride );
nkeynes@736:         pvr2_vram64_detwiddle_8( dest, banks, offset, x1, y1+subdivide, subdivide, stride );
nkeynes@736:         pvr2_vram64_detwiddle_8( dest, banks, offset, x1+subdivide, y1, subdivide, stride );
nkeynes@736:         pvr2_vram64_detwiddle_8( dest, banks, offset, x1+subdivide, y1+subdivide, subdivide, stride );
nkeynes@284:     }
nkeynes@310: }
nkeynes@310: 
nkeynes@310: /**
nkeynes@310:  * @param dest Destination image buffer
nkeynes@310:  * @param banks Source data expressed as two bank pointers
nkeynes@310:  * @param offset Offset into banks[0] specifying where the next word
nkeynes@310:  *  to read is (0 or 1)
nkeynes@310:  * @param x1,y1 Destination coordinates
nkeynes@310:  * @param width Width of current destination block
nkeynes@315:  * @param stride Total width of image (ie stride)
nkeynes@310:  */
nkeynes@310: 
nkeynes@310: static void pvr2_vram64_detwiddle_16( uint16_t *dest, uint16_t *banks[2], int offset,
nkeynes@736:                                       int x1, int y1, int width, int stride )
nkeynes@310: {
nkeynes@310:     if( width == 2 ) {
nkeynes@736:         dest[y1*stride + x1] = *banks[0]++;
nkeynes@736:         dest[(y1+1)*stride + x1] = *banks[offset]++;
nkeynes@736:         dest[y1*stride + x1 + 1] = *banks[1]++;
nkeynes@736:         dest[(y1+1)*stride + x1 + 1] = *banks[offset^1]++;
nkeynes@310:     } else {
nkeynes@736:         int subdivide = width >> 1;
nkeynes@736:         pvr2_vram64_detwiddle_16( dest, banks, offset, x1, y1, subdivide, stride );
nkeynes@736:         pvr2_vram64_detwiddle_16( dest, banks, offset, x1, y1+subdivide, subdivide, stride );
nkeynes@736:         pvr2_vram64_detwiddle_16( dest, banks, offset, x1+subdivide, y1, subdivide, stride );
nkeynes@736:         pvr2_vram64_detwiddle_16( dest, banks, offset, x1+subdivide, y1+subdivide, subdivide, stride );
nkeynes@310:     }
nkeynes@310: }
nkeynes@310: 
nkeynes@310: /**
nkeynes@827:  * Read an image from 64-bit vram stored as twiddled 4-bit pixels. The
nkeynes@315:  * image is written out to the destination in detwiddled form.
nkeynes@315:  * @param dest destination buffer, which must be at least width*height/2 in length
nkeynes@315:  * @param srcaddr source address in vram
nkeynes@315:  * @param width image width (must be a power of 2)
nkeynes@315:  * @param height image height (must be a power of 2)
nkeynes@315:  */
nkeynes@429: void pvr2_vram64_read_twiddled_4( unsigned char *dest, sh4addr_t srcaddr, uint32_t width, uint32_t height )
nkeynes@315: {
nkeynes@315:     int offset_flag = (srcaddr & 0x07);
nkeynes@315:     uint8_t *banks[2];
nkeynes@315:     uint8_t *wdest = (uint8_t*)dest;
nkeynes@315:     uint32_t stride = width >> 1;
nkeynes@429:     int i;
nkeynes@315: 
nkeynes@315:     srcaddr = srcaddr & 0x7FFFF8;
nkeynes@315: 
nkeynes@315:     banks[0] = (uint8_t *)(video_base + (srcaddr>>1));
nkeynes@315:     banks[1] = banks[0] + 0x400000;
nkeynes@315:     if( offset_flag & 0x04 ) { // If source is not 64-bit aligned, swap the banks
nkeynes@736:         uint8_t *tmp = banks[0];
nkeynes@736:         banks[0] = banks[1];
nkeynes@736:         banks[1] = tmp + 4;
nkeynes@736:         offset_flag &= 0x03;
nkeynes@315:     }
nkeynes@315:     banks[0] += offset_flag;
nkeynes@315: 
nkeynes@315:     if( width > height ) {
nkeynes@736:         for( i=0; i<width; i+=height ) {
nkeynes@736:             pvr2_vram64_detwiddle_4( wdest, banks, offset_flag, i, 0, height, stride );
nkeynes@736:         }
nkeynes@315:     } else if( height > width ) {
nkeynes@736:         for( i=0; i<height; i+=width ) {
nkeynes@736:             pvr2_vram64_detwiddle_4( wdest, banks, offset_flag, 0, i, width, stride );
nkeynes@736:         }
nkeynes@315:     } else if( width == 1 ) {
nkeynes@736:         *wdest = *banks[0];
nkeynes@315:     } else {
nkeynes@736:         pvr2_vram64_detwiddle_4( wdest, banks, offset_flag, 0, 0, width, stride );
nkeynes@827:     }
nkeynes@315: }
nkeynes@315: 
nkeynes@315: /**
nkeynes@827:  * Read an image from 64-bit vram stored as twiddled 8-bit pixels. The
nkeynes@310:  * image is written out to the destination in detwiddled form.
nkeynes@310:  * @param dest destination buffer, which must be at least width*height in length
nkeynes@310:  * @param srcaddr source address in vram
nkeynes@310:  * @param width image width (must be a power of 2)
nkeynes@310:  * @param height image height (must be a power of 2)
nkeynes@310:  */
nkeynes@429: void pvr2_vram64_read_twiddled_8( unsigned char *dest, sh4addr_t srcaddr, uint32_t width, uint32_t height )
nkeynes@310: {
nkeynes@310:     int offset_flag = (srcaddr & 0x07);
nkeynes@310:     uint8_t *banks[2];
nkeynes@310:     uint8_t *wdest = (uint8_t*)dest;
nkeynes@429:     int i;
nkeynes@310: 
nkeynes@310:     srcaddr = srcaddr & 0x7FFFF8;
nkeynes@310: 
nkeynes@310:     banks[0] = (uint8_t *)(video_base + (srcaddr>>1));
nkeynes@310:     banks[1] = banks[0] + 0x400000;
nkeynes@310:     if( offset_flag & 0x04 ) { // If source is not 64-bit aligned, swap the banks
nkeynes@736:         uint8_t *tmp = banks[0];
nkeynes@736:         banks[0] = banks[1];
nkeynes@736:         banks[1] = tmp + 4;
nkeynes@736:         offset_flag &= 0x03;
nkeynes@310:     }
nkeynes@310:     banks[0] += offset_flag;
nkeynes@310: 
nkeynes@310:     if( width > height ) {
nkeynes@736:         for( i=0; i<width; i+=height ) {
nkeynes@736:             pvr2_vram64_detwiddle_8( wdest, banks, offset_flag, i, 0, height, width );
nkeynes@736:         }
nkeynes@310:     } else if( height > width ) {
nkeynes@736:         for( i=0; i<height; i+=width ) {
nkeynes@736:             pvr2_vram64_detwiddle_8( wdest, banks, offset_flag, 0, i, width, width );
nkeynes@736:         }
nkeynes@310:     } else if( width == 1 ) {
nkeynes@736:         *wdest = *banks[0];
nkeynes@310:     } else {
nkeynes@736:         pvr2_vram64_detwiddle_8( wdest, banks, offset_flag, 0, 0, width, width );
nkeynes@827:     }
nkeynes@310: }
nkeynes@310: 
nkeynes@310: /**
nkeynes@827:  * Read an image from 64-bit vram stored as twiddled 16-bit pixels. The
nkeynes@310:  * image is written out to the destination in detwiddled form.
nkeynes@310:  * @param dest destination buffer, which must be at least width*height*2 in length
nkeynes@310:  * @param srcaddr source address in vram (must be 16-bit aligned)
nkeynes@310:  * @param width image width (must be a power of 2)
nkeynes@310:  * @param height image height (must be a power of 2)
nkeynes@310:  */
nkeynes@429: void pvr2_vram64_read_twiddled_16( unsigned char *dest, sh4addr_t srcaddr, uint32_t width, uint32_t height ) {
nkeynes@310:     int offset_flag = (srcaddr & 0x06) >> 1;
nkeynes@310:     uint16_t *banks[2];
nkeynes@310:     uint16_t *wdest = (uint16_t*)dest;
nkeynes@429:     int i;
nkeynes@310: 
nkeynes@310:     srcaddr = srcaddr & 0x7FFFF8;
nkeynes@310: 
nkeynes@310:     banks[0] = (uint16_t *)(video_base + (srcaddr>>1));
nkeynes@310:     banks[1] = banks[0] + 0x200000;
nkeynes@310:     if( offset_flag & 0x02 ) { // If source is not 64-bit aligned, swap the banks
nkeynes@736:         uint16_t *tmp = banks[0];
nkeynes@736:         banks[0] = banks[1];
nkeynes@736:         banks[1] = tmp + 2;
nkeynes@736:         offset_flag &= 0x01;
nkeynes@310:     }
nkeynes@310:     banks[0] += offset_flag;
nkeynes@736: 
nkeynes@310: 
nkeynes@310:     if( width > height ) {
nkeynes@736:         for( i=0; i<width; i+=height ) {
nkeynes@736:             pvr2_vram64_detwiddle_16( wdest, banks, offset_flag, i, 0, height, width );
nkeynes@736:         }
nkeynes@310:     } else if( height > width ) {
nkeynes@736:         for( i=0; i<height; i+=width ) {
nkeynes@736:             pvr2_vram64_detwiddle_16( wdest, banks, offset_flag, 0, i, width, width );
nkeynes@736:         }
nkeynes@310:     } else if( width == 1 ) {
nkeynes@736:         *wdest = *banks[0];
nkeynes@310:     } else {
nkeynes@736:         pvr2_vram64_detwiddle_16( wdest, banks, offset_flag, 0, 0, width, width );
nkeynes@827:     }
nkeynes@284: }
nkeynes@284: 
nkeynes@857: static void pvr2_vram_write_invert( sh4addr_t destaddr, unsigned char *src, uint32_t src_size, 
nkeynes@857:                              uint32_t line_size, uint32_t dest_stride,
nkeynes@736:                              uint32_t src_stride )
nkeynes@284: {
nkeynes@429:     unsigned char *dest = video_base + (destaddr & 0x007FFFFF);
nkeynes@857:     unsigned char *p = src + src_size - src_stride;
nkeynes@284:     while( p >= src ) {
nkeynes@857:         memcpy( dest, p, line_size );
nkeynes@736:         p -= src_stride;
nkeynes@857:         dest += dest_stride;
nkeynes@857:     }
nkeynes@857: }
nkeynes@857: 
nkeynes@869: static void pvr2_vram64_write_invert( sh4addr_t destaddr, unsigned char *src, 
nkeynes@869:                                       uint32_t src_size, uint32_t line_size, 
nkeynes@869:                                       uint32_t dest_stride, uint32_t src_stride )
nkeynes@869: {
nkeynes@869:     int i,j;
nkeynes@869:     uint32_t *banks[2];
nkeynes@869:     uint32_t *dwsrc = (uint32_t *)(src + src_size - src_stride);
nkeynes@869:     int32_t src_line_gap = ((int32_t)src_stride + line_size) >> 2; 
nkeynes@869:     int32_t dest_line_gap = ((int32_t)dest_stride - (int32_t)line_size) >> 3;
nkeynes@869: 
nkeynes@869:     destaddr = destaddr & 0x7FFFF8;
nkeynes@869: 
nkeynes@869:     for( i=destaddr; i < destaddr + dest_stride*(src_size/src_stride); i+= LXDREAM_PAGE_SIZE ) {
nkeynes@869:         texcache_invalidate_page( i );
nkeynes@869:     }
nkeynes@869: 
nkeynes@869:     banks[0] = (uint32_t *)(video_base + (destaddr >>1));
nkeynes@869:     banks[1] = banks[0] + 0x100000;
nkeynes@869: 
nkeynes@869:     while( dwsrc >= (uint32_t *)src ) { 
nkeynes@869:         for( j=0; j<line_size; j+=8 ) {
nkeynes@869:             *banks[0]++ = *dwsrc++;
nkeynes@869:             *banks[1]++ = *dwsrc++;
nkeynes@869:         }
nkeynes@869:         banks[0] += dest_line_gap;
nkeynes@869:         banks[1] += dest_line_gap;
nkeynes@869:         dwsrc -= src_line_gap;
nkeynes@869:     }    
nkeynes@869: }
nkeynes@869: 
nkeynes@857: /**
nkeynes@857:  * Copy a pixel buffer to vram, flipping and scaling at the same time. This
nkeynes@857:  * is not massively efficient, but it's used pretty rarely.
nkeynes@857:  */
nkeynes@857: static void pvr2_vram_write_invert_hscale( sh4addr_t destaddr, unsigned char *src, uint32_t src_size, 
nkeynes@857:                              uint32_t line_size, uint32_t dest_stride,
nkeynes@857:                              uint32_t src_stride, int bpp )
nkeynes@857: {
nkeynes@857:     unsigned char *dest = video_base + (destaddr & 0x007FFFFF);
nkeynes@857:     unsigned char *p = src + src_size - src_stride;
nkeynes@857:     while( p >= src ) {
nkeynes@857:         unsigned char *s = p, *d = dest;
nkeynes@857:         int i;
nkeynes@857:         while( s < p+line_size ) {
nkeynes@857:             for( i=0; i<bpp; i++ ) {
nkeynes@857:                 *d++ = *s++;
nkeynes@857:             }
nkeynes@857:             s+= bpp;
nkeynes@857:         }
nkeynes@857:         p -= src_stride;
nkeynes@857:         dest += dest_stride;
nkeynes@284:     }
nkeynes@284: }
nkeynes@284: 
nkeynes@429: void pvr2_vram64_read( unsigned char *dest, sh4addr_t srcaddr, uint32_t length )
nkeynes@284: {
nkeynes@284:     int bank_flag = (srcaddr & 0x04) >> 2;
nkeynes@284:     uint32_t *banks[2];
nkeynes@284:     uint32_t *dwdest;
nkeynes@284:     int i;
nkeynes@284: 
nkeynes@284:     srcaddr = srcaddr & 0x7FFFFF;
nkeynes@284:     if( srcaddr + length > 0x800000 )
nkeynes@736:         length = 0x800000 - srcaddr;
nkeynes@284: 
nkeynes@284:     banks[0] = ((uint32_t *)(video_base + ((srcaddr&0x007FFFF8)>>1)));
nkeynes@284:     banks[1] = banks[0] + 0x100000;
nkeynes@284:     if( bank_flag )
nkeynes@736:         banks[0]++;
nkeynes@736: 
nkeynes@284:     /* Handle non-aligned start of source */
nkeynes@284:     if( srcaddr & 0x03 ) {
nkeynes@736:         char *src = ((char *)banks[bank_flag]) + (srcaddr & 0x03);
nkeynes@736:         for( i= srcaddr & 0x03; i < 4 && length > 0; i++, length-- ) {
nkeynes@736:             *dest++ = *src++;
nkeynes@736:         }
nkeynes@736:         bank_flag = !bank_flag;
nkeynes@284:     }
nkeynes@284: 
nkeynes@284:     dwdest = (uint32_t *)dest;
nkeynes@284:     while( length >= 4 ) {
nkeynes@736:         *dwdest++ = *banks[bank_flag]++;
nkeynes@736:         bank_flag = !bank_flag;
nkeynes@736:         length -= 4;
nkeynes@284:     }
nkeynes@736: 
nkeynes@284:     /* Handle non-aligned end of source */
nkeynes@284:     if( length ) {
nkeynes@736:         dest = (unsigned char *)dwdest;
nkeynes@736:         unsigned char *src = (unsigned char *)banks[bank_flag];
nkeynes@736:         while( length-- > 0 ) {
nkeynes@736:             *dest++ = *src++;
nkeynes@736:         }
nkeynes@284:     }
nkeynes@284: }
nkeynes@284: 
nkeynes@309: void pvr2_vram64_dump_file( sh4addr_t addr, uint32_t length, gchar *filename )
nkeynes@309: {
nkeynes@309:     uint32_t tmp[length>>2];
nkeynes@309:     FILE *f = fopen(filename, "wo");
nkeynes@309:     unsigned int i, j;
nkeynes@309: 
nkeynes@309:     if( f == NULL ) {
nkeynes@736:         ERROR( "Unable to write to dump file '%s' (%s)", filename, strerror(errno) );
nkeynes@736:         return;
nkeynes@309:     }
nkeynes@429:     pvr2_vram64_read( (unsigned char *)tmp, addr, length );
nkeynes@309:     fprintf( f, "%08X\n", addr );
nkeynes@309:     for( i =0; i<length>>2; i+=8 ) {
nkeynes@736:         for( j=i; j<i+8; j++ ) {
nkeynes@736:             if( j < length )
nkeynes@736:                 fprintf( f, " %08X", tmp[j] );
nkeynes@736:             else
nkeynes@736:                 fprintf( f, "         " );
nkeynes@736:         }
nkeynes@736:         fprintf( f, "\n" );
nkeynes@309:     }
nkeynes@309:     fclose(f);
nkeynes@309: }
nkeynes@309: 
nkeynes@827: void pvr2_vram64_dump( sh4addr_t addr, uint32_t length, FILE *f )
nkeynes@284: {
nkeynes@429:     unsigned char tmp[length];
nkeynes@284:     pvr2_vram64_read( tmp, addr, length );
nkeynes@284:     fwrite_dump( tmp, length, f );
nkeynes@284: }
nkeynes@315: 
nkeynes@315: 
nkeynes@315: 
nkeynes@315: /**
nkeynes@315:  * Flush the indicated render buffer back to PVR. Caller is responsible for
nkeynes@315:  * tracking whether there is actually anything in the buffer.
nkeynes@315:  *
nkeynes@869:  * FIXME: Handle horizontal scaler 
nkeynes@827:  *
nkeynes@315:  * @param buffer A render buffer indicating the address to store to, and the
nkeynes@315:  * format the data needs to be in.
nkeynes@315:  */
nkeynes@352: void pvr2_render_buffer_copy_to_sh4( render_buffer_t buffer )
nkeynes@315: {
nkeynes@869:     int line_size = buffer->width * colour_formats[buffer->colour_format].bpp;
nkeynes@869:     int src_stride = line_size;
nkeynes@869:     unsigned char target[buffer->size];
nkeynes@869: 
nkeynes@869:     display_driver->read_render_buffer( target, buffer, line_size, buffer->colour_format );
nkeynes@869: 
nkeynes@869:     if( (buffer->scale & 0xFFFF) == 0x0800 )
nkeynes@869:         src_stride <<= 1;
nkeynes@869: 
nkeynes@429:     if( (buffer->address & 0xFF000000) == 0x04000000 ) {
nkeynes@869:         pvr2_vram64_write_invert( buffer->address, target, buffer->size, line_size, 
nkeynes@869:                                   buffer->rowstride, src_stride );
nkeynes@315:     } else {
nkeynes@736:         /* Regular buffer */
nkeynes@857:         if( buffer->scale & SCALER_HSCALE ) {
nkeynes@857:             pvr2_vram_write_invert_hscale( buffer->address, target, buffer->size, line_size, buffer->rowstride,
nkeynes@857:                                            src_stride, colour_formats[buffer->colour_format].bpp );
nkeynes@352:         } else {
nkeynes@857:             pvr2_vram_write_invert( buffer->address, target, buffer->size, line_size, buffer->rowstride,
nkeynes@857:                                     src_stride );
nkeynes@352:         }
nkeynes@315:     }
nkeynes@856:     buffer->flushed = TRUE;
nkeynes@315: }
nkeynes@315: