• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (C) 2009 Maciej Cencora <m.cencora@gmail.com>
3  *
4  * All Rights Reserved.
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining
7  * a copy of this software and associated documentation files (the
8  * "Software"), to deal in the Software without restriction, including
9  * without limitation the rights to use, copy, modify, merge, publish,
10  * distribute, sublicense, and/or sell copies of the Software, and to
11  * permit persons to whom the Software is furnished to do so, subject to
12  * the following conditions:
13  *
14  * The above copyright notice and this permission notice (including the
15  * next paragraph) shall be included in all copies or substantial
16  * portions of the Software.
17  *
18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
19  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
21  * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
22  * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
23  * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
24  * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25  *
26  */
27 
28 #include "radeon_common.h"
29 #include "r200_context.h"
30 #include "r200_blit.h"
31 #include "r200_tex.h"
32 
cmdpacket0(struct radeon_screen * rscrn,int reg,int count)33 static inline uint32_t cmdpacket0(struct radeon_screen *rscrn,
34                                   int reg, int count)
35 {
36     if (count)
37 	    return CP_PACKET0(reg, count - 1);
38     return CP_PACKET2;
39 }
40 
41 /* common formats supported as both textures and render targets */
r200_check_blit(mesa_format mesa_format,uint32_t dst_pitch)42 unsigned r200_check_blit(mesa_format mesa_format, uint32_t dst_pitch)
43 {
44     /* XXX others? */
45     if (_mesa_little_endian()) {
46 	switch (mesa_format) {
47 	case MESA_FORMAT_B8G8R8A8_UNORM:
48 	case MESA_FORMAT_B8G8R8X8_UNORM:
49 	case MESA_FORMAT_B5G6R5_UNORM:
50 	case MESA_FORMAT_B4G4R4A4_UNORM:
51 	case MESA_FORMAT_B5G5R5A1_UNORM:
52 	case MESA_FORMAT_A_UNORM8:
53 	case MESA_FORMAT_L_UNORM8:
54 	case MESA_FORMAT_I_UNORM8:
55 	/* swizzled - probably can't happen with the disabled Choose8888TexFormat code */
56 	case MESA_FORMAT_A8B8G8R8_UNORM:
57 	case MESA_FORMAT_R8G8B8A8_UNORM:
58 	    break;
59 	default:
60 	    return 0;
61 	}
62     }
63     else {
64 	switch (mesa_format) {
65 	case MESA_FORMAT_A8R8G8B8_UNORM:
66 	case MESA_FORMAT_X8R8G8B8_UNORM:
67 	case MESA_FORMAT_R5G6B5_UNORM:
68 	case MESA_FORMAT_A4R4G4B4_UNORM:
69 	case MESA_FORMAT_A1R5G5B5_UNORM:
70 	case MESA_FORMAT_A_UNORM8:
71 	case MESA_FORMAT_L_UNORM8:
72 	case MESA_FORMAT_I_UNORM8:
73 	/* swizzled  - probably can't happen with the disabled Choose8888TexFormat code */
74 	case MESA_FORMAT_R8G8B8A8_UNORM:
75 	case MESA_FORMAT_A8B8G8R8_UNORM:
76 	   break;
77 	default:
78 	   return 0;
79 	}
80     }
81 
82     /* Rendering to small buffer doesn't work.
83      * Looks like a hw limitation.
84      */
85     if (dst_pitch < 32)
86             return 0;
87 
88     /* ??? */
89     if (_mesa_get_format_bits(mesa_format, GL_DEPTH_BITS) > 0)
90 	    return 0;
91 
92     return 1;
93 }
94 
emit_vtx_state(struct r200_context * r200)95 static inline void emit_vtx_state(struct r200_context *r200)
96 {
97     BATCH_LOCALS(&r200->radeon);
98 
99     BEGIN_BATCH(14);
100     if (r200->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
101 	    OUT_BATCH_REGVAL(R200_SE_VAP_CNTL_STATUS, 0);
102     } else {
103 	    OUT_BATCH_REGVAL(R200_SE_VAP_CNTL_STATUS, RADEON_TCL_BYPASS);
104     }
105     OUT_BATCH_REGVAL(R200_SE_VAP_CNTL, (R200_VAP_FORCE_W_TO_ONE |
106 					(9 << R200_VAP_VF_MAX_VTX_NUM__SHIFT)));
107     OUT_BATCH_REGVAL(R200_SE_VTX_STATE_CNTL, 0);
108     OUT_BATCH_REGVAL(R200_SE_VTE_CNTL, 0);
109     OUT_BATCH_REGVAL(R200_SE_VTX_FMT_0, R200_VTX_XY);
110     OUT_BATCH_REGVAL(R200_SE_VTX_FMT_1, (2 << R200_VTX_TEX0_COMP_CNT_SHIFT));
111     OUT_BATCH_REGVAL(RADEON_SE_CNTL, (RADEON_DIFFUSE_SHADE_GOURAUD |
112 				      RADEON_BFACE_SOLID |
113 				      RADEON_FFACE_SOLID |
114 				      RADEON_VTX_PIX_CENTER_OGL |
115 				      RADEON_ROUND_MODE_ROUND |
116 				      RADEON_ROUND_PREC_4TH_PIX));
117     END_BATCH();
118 }
119 
emit_tx_setup(struct r200_context * r200,mesa_format src_mesa_format,mesa_format dst_mesa_format,struct radeon_bo * bo,intptr_t offset,unsigned width,unsigned height,unsigned pitch)120 static void inline emit_tx_setup(struct r200_context *r200,
121 				 mesa_format src_mesa_format,
122 				 mesa_format dst_mesa_format,
123 				 struct radeon_bo *bo,
124 				 intptr_t offset,
125 				 unsigned width,
126 				 unsigned height,
127 				 unsigned pitch)
128 {
129     uint32_t txformat = R200_TXFORMAT_NON_POWER2;
130     BATCH_LOCALS(&r200->radeon);
131 
132     assert(width <= 2048);
133     assert(height <= 2048);
134     assert(offset % 32 == 0);
135 
136     if (_mesa_little_endian()) {
137 	txformat |= tx_table_le[src_mesa_format].format;
138     }
139     else {
140 	txformat |= tx_table_be[src_mesa_format].format;
141     }
142 
143     if (bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
144 	offset |= R200_TXO_MACRO_TILE;
145     if (bo->flags & RADEON_BO_FLAGS_MICRO_TILE)
146 	offset |= R200_TXO_MICRO_TILE;
147 
148     switch (dst_mesa_format) {
149     /* le */
150     case MESA_FORMAT_B8G8R8A8_UNORM:
151     case MESA_FORMAT_B8G8R8X8_UNORM:
152     case MESA_FORMAT_B5G6R5_UNORM:
153     case MESA_FORMAT_B4G4R4A4_UNORM:
154     case MESA_FORMAT_B5G5R5A1_UNORM:
155     /* be */
156     case MESA_FORMAT_A8R8G8B8_UNORM:
157     case MESA_FORMAT_X8R8G8B8_UNORM:
158     case MESA_FORMAT_R5G6B5_UNORM:
159     case MESA_FORMAT_A4R4G4B4_UNORM:
160     case MESA_FORMAT_A1R5G5B5_UNORM:
161     /* little and big */
162     case MESA_FORMAT_A_UNORM8:
163     case MESA_FORMAT_L_UNORM8:
164     case MESA_FORMAT_I_UNORM8:
165     default:
166 	    /* no swizzle required */
167 	    BEGIN_BATCH(10);
168 	    OUT_BATCH_REGVAL(RADEON_PP_CNTL, (RADEON_TEX_0_ENABLE |
169 					      RADEON_TEX_BLEND_0_ENABLE));
170 	    OUT_BATCH_REGVAL(R200_PP_TXCBLEND_0, (R200_TXC_ARG_A_ZERO |
171 						  R200_TXC_ARG_B_ZERO |
172 						  R200_TXC_ARG_C_R0_COLOR |
173 						  R200_TXC_OP_MADD));
174 	    OUT_BATCH_REGVAL(R200_PP_TXCBLEND2_0, (R200_TXC_CLAMP_0_1 |
175 						   R200_TXC_OUTPUT_REG_R0));
176 	    OUT_BATCH_REGVAL(R200_PP_TXABLEND_0, (R200_TXA_ARG_A_ZERO |
177 						  R200_TXA_ARG_B_ZERO |
178 						  R200_TXA_ARG_C_R0_ALPHA |
179 						  R200_TXA_OP_MADD));
180 	    OUT_BATCH_REGVAL(R200_PP_TXABLEND2_0, (R200_TXA_CLAMP_0_1 |
181 						   R200_TXA_OUTPUT_REG_R0));
182 	    END_BATCH();
183 	    break;
184     case MESA_FORMAT_A8B8G8R8_UNORM:
185     case MESA_FORMAT_R8G8B8A8_UNORM:
186        if ((dst_mesa_format == MESA_FORMAT_A8B8G8R8_UNORM && _mesa_little_endian()) ||
187 	   (dst_mesa_format == MESA_FORMAT_R8G8B8A8_UNORM && !_mesa_little_endian())) {
188 	    BEGIN_BATCH(10);
189 	    OUT_BATCH_REGVAL(RADEON_PP_CNTL, (RADEON_TEX_0_ENABLE |
190 					      RADEON_TEX_BLEND_0_ENABLE));
191 	    OUT_BATCH_REGVAL(R200_PP_TXCBLEND_0, (R200_TXC_ARG_A_ZERO |
192 						  R200_TXC_ARG_B_ZERO |
193 						  R200_TXC_ARG_C_R0_COLOR |
194 						  R200_TXC_OP_MADD));
195 	    /* XXX I don't think this can work. This is output rotation, and alpha contains
196 	     * red, not alpha (we'd write gbrr). */
197 	    OUT_BATCH_REGVAL(R200_PP_TXCBLEND2_0, (R200_TXC_CLAMP_0_1 |
198 						   R200_TXC_OUTPUT_ROTATE_GBA |
199 						   R200_TXC_OUTPUT_REG_R0));
200 	    OUT_BATCH_REGVAL(R200_PP_TXABLEND_0, (R200_TXA_ARG_A_ZERO |
201 						  R200_TXA_ARG_B_ZERO |
202 						  R200_TXA_ARG_C_R0_ALPHA |
203 						  R200_TXA_OP_MADD));
204 	    OUT_BATCH_REGVAL(R200_PP_TXABLEND2_0, (R200_TXA_CLAMP_0_1 |
205 						   (R200_TXA_REPL_RED << R200_TXA_REPL_ARG_C_SHIFT) |
206 						   R200_TXA_OUTPUT_REG_R0));
207 	    END_BATCH();
208        }
209        else {
210 	    /* XXX pretty sure could do this with just 2 instead of 4 instructions.
211 	     * Like so:
212 	     * 1st: use RGA output rotation, rgb arg replicate b, a arg r, write mask rb.
213 	     * That's just one instruction in fact but I'm not entirely sure it works
214 	     * if some of those incoming r0 components are never written (due to mask)
215 	     * in the shader itself to r0.
216 	     * In any case this case (and the one above) may not be reachable with
217 	     * disabled Choose8888TexFormat code. */
218 	    BEGIN_BATCH(34);
219 	    OUT_BATCH_REGVAL(RADEON_PP_CNTL, (RADEON_TEX_0_ENABLE |
220 					      RADEON_TEX_BLEND_0_ENABLE |
221 					      RADEON_TEX_BLEND_1_ENABLE |
222 					      RADEON_TEX_BLEND_2_ENABLE |
223 					      RADEON_TEX_BLEND_3_ENABLE));
224 	    /* r1.r = r0.b */
225 	    OUT_BATCH_REGVAL(R200_PP_TXCBLEND_0, (R200_TXC_ARG_A_ZERO |
226 						  R200_TXC_ARG_B_ZERO |
227 						  R200_TXC_ARG_C_R0_COLOR |
228 						  R200_TXC_OP_MADD));
229 	    OUT_BATCH_REGVAL(R200_PP_TXCBLEND2_0, (R200_TXC_CLAMP_0_1 |
230 						   R200_TXC_OUTPUT_MASK_R |
231 						   (R200_TXC_REPL_BLUE << R200_TXC_REPL_ARG_C_SHIFT) |
232 						   R200_TXC_OUTPUT_REG_R1));
233 	    /* r1.a = r0.a */
234 	    OUT_BATCH_REGVAL(R200_PP_TXABLEND_0, (R200_TXA_ARG_A_ZERO |
235 						  R200_TXA_ARG_B_ZERO |
236 						  R200_TXA_ARG_C_R0_ALPHA |
237 						  R200_TXA_OP_MADD));
238 	    OUT_BATCH_REGVAL(R200_PP_TXABLEND2_0, (R200_TXA_CLAMP_0_1 |
239 						   R200_TXA_OUTPUT_REG_R1));
240 	    /* r1.g = r0.g */
241 	    OUT_BATCH_REGVAL(R200_PP_TXCBLEND_1, (R200_TXC_ARG_A_ZERO |
242 						  R200_TXC_ARG_B_ZERO |
243 						  R200_TXC_ARG_C_R0_COLOR |
244 						  R200_TXC_OP_MADD));
245 	    OUT_BATCH_REGVAL(R200_PP_TXCBLEND2_1, (R200_TXC_CLAMP_0_1 |
246 						   R200_TXC_OUTPUT_MASK_G |
247 						   (R200_TXC_REPL_GREEN << R200_TXC_REPL_ARG_C_SHIFT) |
248 						   R200_TXC_OUTPUT_REG_R1));
249 	    /* r1.a = r0.a */
250 	    OUT_BATCH_REGVAL(R200_PP_TXABLEND_1, (R200_TXA_ARG_A_ZERO |
251 						  R200_TXA_ARG_B_ZERO |
252 						  R200_TXA_ARG_C_R0_ALPHA |
253 						  R200_TXA_OP_MADD));
254 	    OUT_BATCH_REGVAL(R200_PP_TXABLEND2_1, (R200_TXA_CLAMP_0_1 |
255 						   R200_TXA_OUTPUT_REG_R1));
256 	    /* r1.b = r0.r */
257 	    OUT_BATCH_REGVAL(R200_PP_TXCBLEND_2, (R200_TXC_ARG_A_ZERO |
258 						  R200_TXC_ARG_B_ZERO |
259 						  R200_TXC_ARG_C_R0_COLOR |
260 						  R200_TXC_OP_MADD));
261 	    OUT_BATCH_REGVAL(R200_PP_TXCBLEND2_2, (R200_TXC_CLAMP_0_1 |
262 						   R200_TXC_OUTPUT_MASK_B |
263 						   (R200_TXC_REPL_RED << R200_TXC_REPL_ARG_C_SHIFT) |
264 						   R200_TXC_OUTPUT_REG_R1));
265 	    /* r1.a = r0.a */
266 	    OUT_BATCH_REGVAL(R200_PP_TXABLEND_2, (R200_TXA_ARG_A_ZERO |
267 						  R200_TXA_ARG_B_ZERO |
268 						  R200_TXA_ARG_C_R0_ALPHA |
269 						  R200_TXA_OP_MADD));
270 	    OUT_BATCH_REGVAL(R200_PP_TXABLEND2_2, (R200_TXA_CLAMP_0_1 |
271 						   R200_TXA_OUTPUT_REG_R1));
272 	    /* r0.rgb = r1.rgb */
273 	    OUT_BATCH_REGVAL(R200_PP_TXCBLEND_3, (R200_TXC_ARG_A_ZERO |
274 						  R200_TXC_ARG_B_ZERO |
275 						  R200_TXC_ARG_C_R1_COLOR |
276 						  R200_TXC_OP_MADD));
277 	    OUT_BATCH_REGVAL(R200_PP_TXCBLEND2_3, (R200_TXC_CLAMP_0_1 |
278 						   R200_TXC_OUTPUT_REG_R0));
279 	    /* r0.a = r1.a */
280 	    OUT_BATCH_REGVAL(R200_PP_TXABLEND_3, (R200_TXA_ARG_A_ZERO |
281 						  R200_TXA_ARG_B_ZERO |
282 						  R200_TXA_ARG_C_R1_ALPHA |
283 						  R200_TXA_OP_MADD));
284 	    OUT_BATCH_REGVAL(R200_PP_TXABLEND2_3, (R200_TXA_CLAMP_0_1 |
285 						   R200_TXA_OUTPUT_REG_R0));
286 	    END_BATCH();
287 	}
288 	break;
289     }
290 
291     BEGIN_BATCH(18);
292     OUT_BATCH_REGVAL(R200_PP_CNTL_X, 0);
293     OUT_BATCH_REGVAL(R200_PP_TXMULTI_CTL_0, 0);
294     OUT_BATCH_REGVAL(R200_PP_TXFILTER_0, (R200_CLAMP_S_CLAMP_LAST |
295 					  R200_CLAMP_T_CLAMP_LAST |
296 					  R200_MAG_FILTER_NEAREST |
297 					  R200_MIN_FILTER_NEAREST));
298     OUT_BATCH_REGVAL(R200_PP_TXFORMAT_0, txformat);
299     OUT_BATCH_REGVAL(R200_PP_TXFORMAT_X_0, 0);
300     OUT_BATCH_REGVAL(R200_PP_TXSIZE_0, ((width - 1) |
301 					((height - 1) << RADEON_TEX_VSIZE_SHIFT)));
302     OUT_BATCH_REGVAL(R200_PP_TXPITCH_0, pitch * _mesa_get_format_bytes(src_mesa_format) - 32);
303 
304     OUT_BATCH_REGSEQ(R200_PP_TXOFFSET_0, 1);
305     OUT_BATCH_RELOC(offset, bo, offset, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
306 
307     END_BATCH();
308 }
309 
emit_cb_setup(struct r200_context * r200,struct radeon_bo * bo,intptr_t offset,mesa_format mesa_format,unsigned pitch,unsigned width,unsigned height)310 static inline void emit_cb_setup(struct r200_context *r200,
311 				 struct radeon_bo *bo,
312 				 intptr_t offset,
313 				 mesa_format mesa_format,
314 				 unsigned pitch,
315 				 unsigned width,
316 				 unsigned height)
317 {
318     uint32_t dst_pitch = pitch;
319     uint32_t dst_format = 0;
320     BATCH_LOCALS(&r200->radeon);
321 
322     switch (mesa_format) {
323     /* The first of each pair is for little, the second for big endian */
324     case MESA_FORMAT_B8G8R8A8_UNORM:
325     case MESA_FORMAT_A8R8G8B8_UNORM:
326     case MESA_FORMAT_B8G8R8X8_UNORM:
327     case MESA_FORMAT_X8R8G8B8_UNORM:
328     /* These two are valid both for little and big endian (swizzled) */
329     case MESA_FORMAT_A8B8G8R8_UNORM:
330     case MESA_FORMAT_R8G8B8A8_UNORM:
331 	    dst_format = RADEON_COLOR_FORMAT_ARGB8888;
332 	    break;
333     case MESA_FORMAT_B5G6R5_UNORM:
334     case MESA_FORMAT_R5G6B5_UNORM:
335 	    dst_format = RADEON_COLOR_FORMAT_RGB565;
336 	    break;
337     case MESA_FORMAT_B4G4R4A4_UNORM:
338     case MESA_FORMAT_A4R4G4B4_UNORM:
339 	    dst_format = RADEON_COLOR_FORMAT_ARGB4444;
340 	    break;
341     case MESA_FORMAT_B5G5R5A1_UNORM:
342     case MESA_FORMAT_A1R5G5B5_UNORM:
343 	    dst_format = RADEON_COLOR_FORMAT_ARGB1555;
344 	    break;
345     case MESA_FORMAT_A_UNORM8:
346     case MESA_FORMAT_L_UNORM8:
347     case MESA_FORMAT_I_UNORM8:
348 	    dst_format = RADEON_COLOR_FORMAT_RGB8;
349 	    break;
350     default:
351 	    break;
352     }
353 
354     if (bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
355 	dst_pitch |= R200_COLOR_TILE_ENABLE;
356     if (bo->flags & RADEON_BO_FLAGS_MICRO_TILE)
357 	dst_pitch |= R200_COLOR_MICROTILE_ENABLE;
358 
359     BEGIN_BATCH(22);
360     OUT_BATCH_REGVAL(R200_RE_AUX_SCISSOR_CNTL, 0);
361     OUT_BATCH_REGVAL(R200_RE_CNTL, 0);
362     OUT_BATCH_REGVAL(RADEON_RE_TOP_LEFT, 0);
363     OUT_BATCH_REGVAL(RADEON_RE_WIDTH_HEIGHT, (((width - 1) << RADEON_RE_WIDTH_SHIFT) |
364 					      ((height - 1) << RADEON_RE_HEIGHT_SHIFT)));
365     OUT_BATCH_REGVAL(RADEON_RB3D_PLANEMASK, 0xffffffff);
366     OUT_BATCH_REGVAL(RADEON_RB3D_BLENDCNTL, RADEON_SRC_BLEND_GL_ONE | RADEON_DST_BLEND_GL_ZERO);
367     OUT_BATCH_REGVAL(RADEON_RB3D_CNTL, dst_format);
368 
369     OUT_BATCH_REGSEQ(RADEON_RB3D_COLOROFFSET, 1);
370     OUT_BATCH_RELOC(offset, bo, offset, 0, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0);
371     OUT_BATCH_REGSEQ(RADEON_RB3D_COLORPITCH, 1);
372     OUT_BATCH_RELOC(dst_pitch, bo, dst_pitch, 0, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0);
373 
374     END_BATCH();
375 }
376 
validate_buffers(struct r200_context * r200,struct radeon_bo * src_bo,struct radeon_bo * dst_bo)377 static GLboolean validate_buffers(struct r200_context *r200,
378                                   struct radeon_bo *src_bo,
379                                   struct radeon_bo *dst_bo)
380 {
381     int ret;
382 
383     radeon_cs_space_reset_bos(r200->radeon.cmdbuf.cs);
384 
385     ret = radeon_cs_space_check_with_bo(r200->radeon.cmdbuf.cs,
386                                         src_bo, RADEON_GEM_DOMAIN_VRAM | RADEON_GEM_DOMAIN_GTT, 0);
387     if (ret)
388         return GL_FALSE;
389 
390     ret = radeon_cs_space_check_with_bo(r200->radeon.cmdbuf.cs,
391                                         dst_bo, 0, RADEON_GEM_DOMAIN_VRAM | RADEON_GEM_DOMAIN_GTT);
392     if (ret)
393         return GL_FALSE;
394 
395     return GL_TRUE;
396 }
397 
398 /**
399  * Calculate texcoords for given image region.
400  * Output values are [minx, maxx, miny, maxy]
401  */
calc_tex_coords(float img_width,float img_height,float x,float y,float reg_width,float reg_height,unsigned flip_y,float * buf)402 static inline void calc_tex_coords(float img_width, float img_height,
403 				   float x, float y,
404 				   float reg_width, float reg_height,
405 				   unsigned flip_y, float *buf)
406 {
407     buf[0] = x / img_width;
408     buf[1] = buf[0] + reg_width / img_width;
409     buf[2] = y / img_height;
410     buf[3] = buf[2] + reg_height / img_height;
411     if (flip_y)
412     {
413         buf[2] = 1.0 - buf[2];
414         buf[3] = 1.0 - buf[3];
415     }
416 }
417 
emit_draw_packet(struct r200_context * r200,unsigned src_width,unsigned src_height,unsigned src_x_offset,unsigned src_y_offset,unsigned dst_x_offset,unsigned dst_y_offset,unsigned reg_width,unsigned reg_height,unsigned flip_y)418 static inline void emit_draw_packet(struct r200_context *r200,
419 				    unsigned src_width, unsigned src_height,
420 				    unsigned src_x_offset, unsigned src_y_offset,
421 				    unsigned dst_x_offset, unsigned dst_y_offset,
422 				    unsigned reg_width, unsigned reg_height,
423 				    unsigned flip_y)
424 {
425     float texcoords[4];
426     float verts[12];
427     BATCH_LOCALS(&r200->radeon);
428 
429     calc_tex_coords(src_width, src_height,
430                     src_x_offset, src_y_offset,
431                     reg_width, reg_height,
432                     flip_y, texcoords);
433 
434     verts[0] = dst_x_offset;
435     verts[1] = dst_y_offset + reg_height;
436     verts[2] = texcoords[0];
437     verts[3] = texcoords[3];
438 
439     verts[4] = dst_x_offset + reg_width;
440     verts[5] = dst_y_offset + reg_height;
441     verts[6] = texcoords[1];
442     verts[7] = texcoords[3];
443 
444     verts[8] = dst_x_offset + reg_width;
445     verts[9] = dst_y_offset;
446     verts[10] = texcoords[1];
447     verts[11] = texcoords[2];
448 
449     BEGIN_BATCH(14);
450     OUT_BATCH(R200_CP_CMD_3D_DRAW_IMMD_2 | (12 << 16));
451     OUT_BATCH(RADEON_CP_VC_CNTL_PRIM_WALK_RING |
452 	      RADEON_CP_VC_CNTL_PRIM_TYPE_RECT_LIST |
453               (3 << 16));
454     OUT_BATCH_TABLE(verts, 12);
455     END_BATCH();
456 }
457 
458 /**
459  * Copy a region of [@a width x @a height] pixels from source buffer
460  * to destination buffer.
461  * @param[in] r200 r200 context
462  * @param[in] src_bo source radeon buffer object
463  * @param[in] src_offset offset of the source image in the @a src_bo
464  * @param[in] src_mesaformat source image format
465  * @param[in] src_pitch aligned source image width
466  * @param[in] src_width source image width
467  * @param[in] src_height source image height
468  * @param[in] src_x_offset x offset in the source image
469  * @param[in] src_y_offset y offset in the source image
470  * @param[in] dst_bo destination radeon buffer object
471  * @param[in] dst_offset offset of the destination image in the @a dst_bo
472  * @param[in] dst_mesaformat destination image format
473  * @param[in] dst_pitch aligned destination image width
474  * @param[in] dst_width destination image width
475  * @param[in] dst_height destination image height
476  * @param[in] dst_x_offset x offset in the destination image
477  * @param[in] dst_y_offset y offset in the destination image
478  * @param[in] width region width
479  * @param[in] height region height
480  * @param[in] flip_y set if y coords of the source image need to be flipped
481  */
r200_blit(struct gl_context * ctx,struct radeon_bo * src_bo,intptr_t src_offset,mesa_format src_mesaformat,unsigned src_pitch,unsigned src_width,unsigned src_height,unsigned src_x_offset,unsigned src_y_offset,struct radeon_bo * dst_bo,intptr_t dst_offset,mesa_format dst_mesaformat,unsigned dst_pitch,unsigned dst_width,unsigned dst_height,unsigned dst_x_offset,unsigned dst_y_offset,unsigned reg_width,unsigned reg_height,unsigned flip_y)482 unsigned r200_blit(struct gl_context *ctx,
483                    struct radeon_bo *src_bo,
484                    intptr_t src_offset,
485                    mesa_format src_mesaformat,
486                    unsigned src_pitch,
487                    unsigned src_width,
488                    unsigned src_height,
489                    unsigned src_x_offset,
490                    unsigned src_y_offset,
491                    struct radeon_bo *dst_bo,
492                    intptr_t dst_offset,
493                    mesa_format dst_mesaformat,
494                    unsigned dst_pitch,
495                    unsigned dst_width,
496                    unsigned dst_height,
497                    unsigned dst_x_offset,
498                    unsigned dst_y_offset,
499                    unsigned reg_width,
500                    unsigned reg_height,
501                    unsigned flip_y)
502 {
503     struct r200_context *r200 = R200_CONTEXT(ctx);
504 
505     if (!r200_check_blit(dst_mesaformat, dst_pitch))
506         return GL_FALSE;
507 
508     /* Make sure that colorbuffer has even width - hw limitation */
509     if (dst_pitch % 2 > 0)
510         ++dst_pitch;
511 
512     /* Need to clamp the region size to make sure
513      * we don't read outside of the source buffer
514      * or write outside of the destination buffer.
515      */
516     if (reg_width + src_x_offset > src_width)
517         reg_width = src_width - src_x_offset;
518     if (reg_height + src_y_offset > src_height)
519         reg_height = src_height - src_y_offset;
520     if (reg_width + dst_x_offset > dst_width)
521         reg_width = dst_width - dst_x_offset;
522     if (reg_height + dst_y_offset > dst_height)
523         reg_height = dst_height - dst_y_offset;
524 
525     if (src_bo == dst_bo) {
526         return GL_FALSE;
527     }
528 
529     if (src_offset % 32 || dst_offset % 32) {
530         return GL_FALSE;
531     }
532 
533     if (0) {
534         fprintf(stderr, "src: size [%d x %d], pitch %d, "
535                 "offset [%d x %d], format %s, bo %p\n",
536                 src_width, src_height, src_pitch,
537                 src_x_offset, src_y_offset,
538                 _mesa_get_format_name(src_mesaformat),
539                 src_bo);
540         fprintf(stderr, "dst: pitch %d, offset[%d x %d], format %s, bo %p\n",
541                 dst_pitch, dst_x_offset, dst_y_offset,
542                 _mesa_get_format_name(dst_mesaformat), dst_bo);
543         fprintf(stderr, "region: %d x %d\n", reg_width, reg_height);
544     }
545 
546     /* Flush is needed to make sure that source buffer has correct data */
547     radeonFlush(&r200->radeon.glCtx);
548 
549     rcommonEnsureCmdBufSpace(&r200->radeon, 102, __func__);
550 
551     if (!validate_buffers(r200, src_bo, dst_bo))
552         return GL_FALSE;
553 
554     /* 14 */
555     emit_vtx_state(r200);
556     /* 52 */
557     emit_tx_setup(r200, src_mesaformat, dst_mesaformat, src_bo, src_offset, src_width, src_height, src_pitch);
558     /* 22 */
559     emit_cb_setup(r200, dst_bo, dst_offset, dst_mesaformat, dst_pitch, dst_width, dst_height);
560     /* 14 */
561     emit_draw_packet(r200, src_width, src_height,
562                      src_x_offset, src_y_offset,
563                      dst_x_offset, dst_y_offset,
564                      reg_width, reg_height,
565                      flip_y);
566 
567     radeonFlush(ctx);
568 
569     /* We submitted those packets outside our state atom mechanism. Thus
570      * make sure the atoms are resubmitted the next time. */
571     r200->hw.cst.dirty = GL_TRUE;
572     r200->hw.ctx.dirty = GL_TRUE;
573     r200->hw.vap.dirty = GL_TRUE;
574     r200->hw.msk.dirty = GL_TRUE;
575     r200->hw.pix[0].dirty = GL_TRUE;
576     r200->hw.pix[1].dirty = GL_TRUE;
577     r200->hw.pix[2].dirty = GL_TRUE;
578     r200->hw.pix[3].dirty = GL_TRUE;
579     r200->hw.sci.dirty = GL_TRUE;
580     r200->hw.set.dirty = GL_TRUE;
581     r200->hw.tex[0].dirty = GL_TRUE;
582     r200->hw.vte.dirty = GL_TRUE;
583     r200->hw.vtx.dirty = GL_TRUE;
584 
585     return GL_TRUE;
586 }
587