2 * $Id: pvr2mem.c,v 1.5 2007-01-23 11:19:32 nkeynes Exp $
4 * PVR2 (Video) VRAM handling routines (mainly for the 64-bit region)
6 * Copyright (c) 2005 Nathan Keynes.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
22 extern char *video_base;
24 void pvr2_vram64_write( sh4addr_t destaddr, char *src, uint32_t length )
26 int bank_flag = (destaddr & 0x04) >> 2;
31 destaddr = destaddr & 0x7FFFFF;
32 if( destaddr + length > 0x800000 ) {
33 length = 0x800000 - destaddr;
36 for( i=destaddr & 0xFFFFF000; i < destaddr + length; i+= PAGE_SIZE ) {
37 texcache_invalidate_page( i );
40 banks[0] = ((uint32_t *)(video_base + ((destaddr & 0x007FFFF8) >>1)));
41 banks[1] = banks[0] + 0x100000;
45 /* Handle non-aligned start of source */
46 if( destaddr & 0x03 ) {
47 char *dest = ((char *)banks[bank_flag]) + (destaddr & 0x03);
48 for( i= destaddr & 0x03; i < 4 && length > 0; i++, length-- ) {
51 bank_flag = !bank_flag;
54 dwsrc = (uint32_t *)src;
55 while( length >= 4 ) {
56 *banks[bank_flag]++ = *dwsrc++;
57 bank_flag = !bank_flag;
61 /* Handle non-aligned end of source */
64 char *dest = (char *)banks[bank_flag];
65 while( length-- > 0 ) {
72 * Write an image to 64-bit vram, with a line-stride different from the line-size.
73 * The destaddr must be 32-bit aligned, and both line_bytes and line_stride_bytes
74 * must be multiples of 4.
76 void pvr2_vram64_write_stride( sh4addr_t destaddr, char *src, uint32_t line_bytes,
77 uint32_t line_stride_bytes, uint32_t line_count )
79 int bank_flag = (destaddr & 0x04) >> 2;
86 destaddr = destaddr & 0x7FFFF8;
87 i = line_stride_bytes - line_bytes;
88 line_gap_flag = i & 0x04;
92 for( i=destaddr & 0xFFFFF000; i < destaddr + line_stride_bytes*line_count; i+= PAGE_SIZE ) {
93 texcache_invalidate_page( i );
96 banks[0] = (uint32_t *)(video_base + (destaddr >>1));
97 banks[1] = banks[0] + 0x100000;
101 dwsrc = (uint32_t *)src;
102 for( i=0; i<line_count; i++ ) {
103 for( j=0; j<line_bytes; j++ ) {
104 *banks[bank_flag]++ = *dwsrc++;
105 bank_flag = !bank_flag;
107 banks[0] += line_gap;
108 banks[1] += line_gap;
109 if( line_gap_flag ) {
111 bank_flag = !bank_flag;
117 * Read an image from 64-bit vram, with a destination line-stride different from the line-size.
118 * The srcaddr must be 32-bit aligned, and both line_bytes and line_stride_bytes
119 * must be multiples of 4. line_stride_bytes must be >= line_bytes.
120 * This method is used to extract a "stride" texture from vram.
122 void pvr2_vram64_read_stride( char *dest, uint32_t dest_line_bytes, sh4addr_t srcaddr,
123 uint32_t src_line_bytes, uint32_t line_count )
125 int bank_flag = (srcaddr & 0x04) >> 2;
128 uint32_t dest_line_gap;
129 uint32_t src_line_gap;
131 int src_line_gap_flag;
134 srcaddr = srcaddr & 0x7FFFF8;
135 if( src_line_bytes <= dest_line_bytes ) {
136 dest_line_gap = (dest_line_bytes - src_line_bytes) >> 2;
138 src_line_gap_flag = 0;
139 line_bytes = src_line_bytes >> 2;
141 i = (src_line_bytes - dest_line_bytes);
142 src_line_gap_flag = i & 0x04;
143 src_line_gap = i >> 3;
144 line_bytes = dest_line_bytes >> 2;
147 banks[0] = (uint32_t *)(video_base + (srcaddr>>1));
148 banks[1] = banks[0] + 0x100000;
152 dwdest = (uint32_t *)dest;
153 for( i=0; i<line_count; i++ ) {
154 for( j=0; j<line_bytes; j++ ) {
155 *dwdest++ = *banks[bank_flag]++;
156 bank_flag = !bank_flag;
158 dwdest += dest_line_gap;
159 banks[0] += src_line_gap;
160 banks[1] += src_line_gap;
161 if( src_line_gap_flag ) {
163 bank_flag = !bank_flag;
170 * @param dest Destination image buffer
171 * @param banks Source data expressed as two bank pointers
172 * @param offset Offset into banks[0] specifying where the next byte
174 * @param x1,y1 Destination coordinates
175 * @param width Width of current destination block
176 * @param stride Total width of image (ie stride) in bytes
179 static void pvr2_vram64_detwiddle_4( uint8_t *dest, uint8_t *banks[2], int offset,
180 int x1, int y1, int width, int stride )
184 uint8_t t1 = *banks[offset<4?0:1]++;
185 uint8_t t2 = *banks[offset<3?0:1]++;
186 dest[y1*stride + x1] = (t1 & 0x0F) | (t2<<4);
187 dest[(y1+1)*stride + x1] = (t1>>4) | (t2&0xF0);
188 } else if( width == 4 ) {
189 pvr2_vram64_detwiddle_4( dest, banks, offset, x1, y1, 2, stride );
190 pvr2_vram64_detwiddle_4( dest, banks, offset+2, x1, y1+2, 2, stride );
191 pvr2_vram64_detwiddle_4( dest, banks, offset+4, x1+2, y1, 2, stride );
192 pvr2_vram64_detwiddle_4( dest, banks, offset+6, x1+2, y1+2, 2, stride );
195 int subdivide = width >> 1;
196 pvr2_vram64_detwiddle_4( dest, banks, offset, x1, y1, subdivide, stride );
197 pvr2_vram64_detwiddle_4( dest, banks, offset, x1, y1+subdivide, subdivide, stride );
198 pvr2_vram64_detwiddle_4( dest, banks, offset, x1+subdivide, y1, subdivide, stride );
199 pvr2_vram64_detwiddle_4( dest, banks, offset, x1+subdivide, y1+subdivide, subdivide, stride );
204 * @param dest Destination image buffer
205 * @param banks Source data expressed as two bank pointers
206 * @param offset Offset into banks[0] specifying where the next byte
208 * @param x1,y1 Destination coordinates
209 * @param width Width of current destination block
210 * @param stride Total width of image (ie stride)
213 static void pvr2_vram64_detwiddle_8( uint8_t *dest, uint8_t *banks[2], int offset,
214 int x1, int y1, int width, int stride )
217 dest[y1*stride + x1] = *banks[0]++;
218 dest[(y1+1)*stride + x1] = *banks[offset<3?0:1]++;
219 dest[y1*stride + x1 + 1] = *banks[offset<2?0:1]++;
220 dest[(y1+1)*stride + x1 + 1] = *banks[offset==0?0:1]++;
221 uint8_t *tmp = banks[0]; /* swap banks */
225 int subdivide = width >> 1;
226 pvr2_vram64_detwiddle_8( dest, banks, offset, x1, y1, subdivide, stride );
227 pvr2_vram64_detwiddle_8( dest, banks, offset, x1, y1+subdivide, subdivide, stride );
228 pvr2_vram64_detwiddle_8( dest, banks, offset, x1+subdivide, y1, subdivide, stride );
229 pvr2_vram64_detwiddle_8( dest, banks, offset, x1+subdivide, y1+subdivide, subdivide, stride );
234 * @param dest Destination image buffer
235 * @param banks Source data expressed as two bank pointers
236 * @param offset Offset into banks[0] specifying where the next word
237 * to read is (0 or 1)
238 * @param x1,y1 Destination coordinates
239 * @param width Width of current destination block
240 * @param stride Total width of image (ie stride)
243 static void pvr2_vram64_detwiddle_16( uint16_t *dest, uint16_t *banks[2], int offset,
244 int x1, int y1, int width, int stride )
247 dest[y1*stride + x1] = *banks[0]++;
248 dest[(y1+1)*stride + x1] = *banks[offset]++;
249 dest[y1*stride + x1 + 1] = *banks[1]++;
250 dest[(y1+1)*stride + x1 + 1] = *banks[offset^1]++;
252 int subdivide = width >> 1;
253 pvr2_vram64_detwiddle_16( dest, banks, offset, x1, y1, subdivide, stride );
254 pvr2_vram64_detwiddle_16( dest, banks, offset, x1, y1+subdivide, subdivide, stride );
255 pvr2_vram64_detwiddle_16( dest, banks, offset, x1+subdivide, y1, subdivide, stride );
256 pvr2_vram64_detwiddle_16( dest, banks, offset, x1+subdivide, y1+subdivide, subdivide, stride );
261 * Read an image from 64-bit vram stored as twiddled 4-bit pixels. The
262 * image is written out to the destination in detwiddled form.
263 * @param dest destination buffer, which must be at least width*height/2 in length
264 * @param srcaddr source address in vram
265 * @param width image width (must be a power of 2)
266 * @param height image height (must be a power of 2)
268 void pvr2_vram64_read_twiddled_4( char *dest, sh4addr_t srcaddr, uint32_t width, uint32_t height )
270 int offset_flag = (srcaddr & 0x07);
272 uint8_t *wdest = (uint8_t*)dest;
273 uint32_t stride = width >> 1;
276 srcaddr = srcaddr & 0x7FFFF8;
278 banks[0] = (uint8_t *)(video_base + (srcaddr>>1));
279 banks[1] = banks[0] + 0x400000;
280 if( offset_flag & 0x04 ) { // If source is not 64-bit aligned, swap the banks
281 uint8_t *tmp = banks[0];
286 banks[0] += offset_flag;
288 if( width > height ) {
289 for( i=0; i<width; i+=height ) {
290 pvr2_vram64_detwiddle_4( wdest, banks, offset_flag, i, 0, height, stride );
292 } else if( height > width ) {
293 for( i=0; i<height; i+=width ) {
294 pvr2_vram64_detwiddle_4( wdest, banks, offset_flag, 0, i, width, stride );
296 } else if( width == 1 ) {
299 pvr2_vram64_detwiddle_4( wdest, banks, offset_flag, 0, 0, width, stride );
304 * Read an image from 64-bit vram stored as twiddled 8-bit pixels. The
305 * image is written out to the destination in detwiddled form.
306 * @param dest destination buffer, which must be at least width*height in length
307 * @param srcaddr source address in vram
308 * @param width image width (must be a power of 2)
309 * @param height image height (must be a power of 2)
311 void pvr2_vram64_read_twiddled_8( char *dest, sh4addr_t srcaddr, uint32_t width, uint32_t height )
313 int offset_flag = (srcaddr & 0x07);
315 uint8_t *wdest = (uint8_t*)dest;
318 srcaddr = srcaddr & 0x7FFFF8;
320 banks[0] = (uint8_t *)(video_base + (srcaddr>>1));
321 banks[1] = banks[0] + 0x400000;
322 if( offset_flag & 0x04 ) { // If source is not 64-bit aligned, swap the banks
323 uint8_t *tmp = banks[0];
328 banks[0] += offset_flag;
330 if( width > height ) {
331 for( i=0; i<width; i+=height ) {
332 pvr2_vram64_detwiddle_8( wdest, banks, offset_flag, i, 0, height, width );
334 } else if( height > width ) {
335 for( i=0; i<height; i+=width ) {
336 pvr2_vram64_detwiddle_8( wdest, banks, offset_flag, 0, i, width, width );
338 } else if( width == 1 ) {
341 pvr2_vram64_detwiddle_8( wdest, banks, offset_flag, 0, 0, width, width );
346 * Read an image from 64-bit vram stored as twiddled 16-bit pixels. The
347 * image is written out to the destination in detwiddled form.
348 * @param dest destination buffer, which must be at least width*height*2 in length
349 * @param srcaddr source address in vram (must be 16-bit aligned)
350 * @param width image width (must be a power of 2)
351 * @param height image height (must be a power of 2)
353 void pvr2_vram64_read_twiddled_16( char *dest, sh4addr_t srcaddr, uint32_t width, uint32_t height ) {
354 int offset_flag = (srcaddr & 0x06) >> 1;
356 uint16_t *wdest = (uint16_t*)dest;
359 srcaddr = srcaddr & 0x7FFFF8;
361 banks[0] = (uint16_t *)(video_base + (srcaddr>>1));
362 banks[1] = banks[0] + 0x200000;
363 if( offset_flag & 0x02 ) { // If source is not 64-bit aligned, swap the banks
364 uint16_t *tmp = banks[0];
369 banks[0] += offset_flag;
372 if( width > height ) {
373 for( i=0; i<width; i+=height ) {
374 pvr2_vram64_detwiddle_16( wdest, banks, offset_flag, i, 0, height, width );
376 } else if( height > width ) {
377 for( i=0; i<height; i+=width ) {
378 pvr2_vram64_detwiddle_16( wdest, banks, offset_flag, 0, i, width, width );
380 } else if( width == 1 ) {
383 pvr2_vram64_detwiddle_16( wdest, banks, offset_flag, 0, 0, width, width );
387 void pvr2_vram_write_invert( sh4addr_t destaddr, char *src, uint32_t length, uint32_t line_length )
389 char *dest = video_base + (destaddr & 0x007FFFFF);
390 char *p = src + length - line_length;
392 memcpy( dest, p, line_length );
398 void pvr2_vram64_read( char *dest, sh4addr_t srcaddr, uint32_t length )
400 int bank_flag = (srcaddr & 0x04) >> 2;
405 srcaddr = srcaddr & 0x7FFFFF;
406 if( srcaddr + length > 0x800000 )
407 length = 0x800000 - srcaddr;
409 banks[0] = ((uint32_t *)(video_base + ((srcaddr&0x007FFFF8)>>1)));
410 banks[1] = banks[0] + 0x100000;
414 /* Handle non-aligned start of source */
415 if( srcaddr & 0x03 ) {
416 char *src = ((char *)banks[bank_flag]) + (srcaddr & 0x03);
417 for( i= srcaddr & 0x03; i < 4 && length > 0; i++, length-- ) {
420 bank_flag = !bank_flag;
423 dwdest = (uint32_t *)dest;
424 while( length >= 4 ) {
425 *dwdest++ = *banks[bank_flag]++;
426 bank_flag = !bank_flag;
430 /* Handle non-aligned end of source */
432 dest = (char *)dwdest;
433 char *src = (char *)banks[bank_flag];
434 while( length-- > 0 ) {
440 void pvr2_vram64_dump_file( sh4addr_t addr, uint32_t length, gchar *filename )
442 uint32_t tmp[length>>2];
443 FILE *f = fopen(filename, "wo");
447 ERROR( "Unable to write to dump file '%s' (%s)", filename, strerror(errno) );
450 pvr2_vram64_read( tmp, addr, length );
451 fprintf( f, "%08X\n", addr );
452 for( i =0; i<length>>2; i+=8 ) {
453 for( j=i; j<i+8; j++ ) {
455 fprintf( f, " %08X", tmp[j] );
464 void pvr2_vram64_dump( sh4addr_t addr, uint32_t length, FILE *f )
467 pvr2_vram64_read( tmp, addr, length );
468 fwrite_dump( tmp, length, f );
474 * Flush the indicated render buffer back to PVR. Caller is responsible for
475 * tracking whether there is actually anything in the buffer.
477 * @param buffer A render buffer indicating the address to store to, and the
478 * format the data needs to be in.
479 * @param backBuffer TRUE to flush the back buffer, FALSE for
482 void pvr2_render_buffer_copy_to_sh4( pvr2_render_buffer_t buffer,
483 gboolean backBuffer )
485 if( buffer->render_addr == -1 )
487 GLenum type, format = GL_BGRA;
488 int line_size = buffer->width, size;
490 switch( buffer->colour_format ) {
492 type = GL_UNSIGNED_SHORT_5_6_5;
497 type = GL_UNSIGNED_BYTE;
499 line_size = (line_size<<1)+line_size;
501 case COLFMT_ARGB1555:
502 type = GL_UNSIGNED_SHORT_5_5_5_1;
505 case COLFMT_ARGB4444:
506 type = GL_UNSIGNED_SHORT_4_4_4_4;
509 case COLFMT_ARGB8888:
510 type = GL_UNSIGNED_INT_8_8_8_8;
514 size = line_size * buffer->height;
518 glReadBuffer( GL_BACK );
520 glReadBuffer( GL_FRONT );
523 if( buffer->render_addr & 0xFF000000 == 0x04000000 ) {
524 /* Interlaced buffer. Go the double copy... :( */
526 glReadPixels( 0, 0, buffer->width, buffer->height, format, type, target );
527 pvr2_vram64_write( buffer->render_addr, target, size );
531 glReadPixels( 0, 0, buffer->width, buffer->height, format, type, target );
532 pvr2_vram_write_invert( buffer->render_addr, target, size, line_size );
538 * Copy data from PVR ram into the GL render buffer.
540 * @param buffer A render buffer indicating the address to read from, and the
541 * format the data is in.
542 * @param backBuffer TRUE to write the back buffer, FALSE for
545 void pvr2_render_buffer_copy_from_sh4( pvr2_render_buffer_t buffer,
546 gboolean backBuffer )
548 if( buffer->render_addr == -1 )
550 GLenum type, format = GL_RGBA;
551 int size = buffer->width * buffer->height;
553 switch( buffer->colour_format ) {
555 type = GL_UNSIGNED_SHORT_5_6_5;
560 type = GL_UNSIGNED_BYTE;
562 size = (size<<1)+size;
564 case COLFMT_ARGB1555:
565 type = GL_UNSIGNED_SHORT_5_5_5_1;
568 case COLFMT_ARGB4444:
569 type = GL_UNSIGNED_SHORT_4_4_4_4;
572 case COLFMT_ARGB8888:
573 type = GL_UNSIGNED_INT_8_8_8_8;
579 glDrawBuffer( GL_BACK );
581 glDrawBuffer( GL_FRONT );
584 glRasterPos2i( 0, 0 );
585 if( buffer->render_addr & 0xFF000000 == 0x04000000 ) {
586 /* Interlaced buffer. Go the double copy... :( */
588 pvr2_vram64_read( target, buffer->render_addr, size );
589 glDrawPixels( buffer->width, buffer->height,
590 format, type, target );
592 /* Regular buffer - go direct */
593 char *target = mem_get_region( buffer->render_addr );
594 glDrawPixels( buffer->width, buffer->height,
595 format, type, target );
.