Merge branch '7.8' into master
[mesa.git] / src / mesa / drivers / dri / radeon / radeon_blit.c
1 /*
2 * Copyright (C) 2010 Advanced Micro Devices, Inc.
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 "radeon_context.h"
30 #include "radeon_blit.h"
31
32 static inline uint32_t cmdpacket0(struct radeon_screen *rscrn,
33 int reg, int count)
34 {
35 if (count)
36 return CP_PACKET0(reg, count - 1);
37 return CP_PACKET2;
38 }
39
40 /* common formats supported as both textures and render targets */
41 unsigned r100_check_blit(gl_format mesa_format)
42 {
43 /* XXX others? BE/LE? */
44 switch (mesa_format) {
45 case MESA_FORMAT_ARGB8888:
46 case MESA_FORMAT_XRGB8888:
47 case MESA_FORMAT_RGB565:
48 case MESA_FORMAT_ARGB4444:
49 case MESA_FORMAT_ARGB1555:
50 case MESA_FORMAT_A8:
51 case MESA_FORMAT_L8:
52 case MESA_FORMAT_I8:
53 break;
54 default:
55 return 0;
56 }
57
58 /* ??? */
59 if (_mesa_get_format_bits(mesa_format, GL_DEPTH_BITS) > 0)
60 return 0;
61
62 return 1;
63 }
64
65 static inline void emit_vtx_state(struct r100_context *r100)
66 {
67 BATCH_LOCALS(&r100->radeon);
68
69 BEGIN_BATCH(8);
70 if (r100->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
71 OUT_BATCH_REGVAL(RADEON_SE_CNTL_STATUS, 0);
72 } else {
73 OUT_BATCH_REGVAL(RADEON_SE_CNTL_STATUS, RADEON_TCL_BYPASS);
74
75 }
76 OUT_BATCH_REGVAL(RADEON_SE_COORD_FMT, (RADEON_VTX_XY_PRE_MULT_1_OVER_W0 |
77 RADEON_TEX1_W_ROUTING_USE_W0));
78 OUT_BATCH_REGVAL(RADEON_SE_VTX_FMT, RADEON_SE_VTX_FMT_XY | RADEON_SE_VTX_FMT_ST0);
79 OUT_BATCH_REGVAL(RADEON_SE_CNTL, (RADEON_DIFFUSE_SHADE_GOURAUD |
80 RADEON_BFACE_SOLID |
81 RADEON_FFACE_SOLID |
82 RADEON_VTX_PIX_CENTER_OGL |
83 RADEON_ROUND_MODE_ROUND |
84 RADEON_ROUND_PREC_4TH_PIX));
85 END_BATCH();
86 }
87
88 static void inline emit_tx_setup(struct r100_context *r100,
89 gl_format mesa_format,
90 struct radeon_bo *bo,
91 intptr_t offset,
92 unsigned width,
93 unsigned height,
94 unsigned pitch)
95 {
96 uint32_t txformat = RADEON_TXFORMAT_NON_POWER2;
97 BATCH_LOCALS(&r100->radeon);
98
99 assert(width <= 2047);
100 assert(height <= 2047);
101 assert(offset % 32 == 0);
102
103 /* XXX others? BE/LE? */
104 switch (mesa_format) {
105 case MESA_FORMAT_ARGB8888:
106 txformat |= RADEON_TXFORMAT_ARGB8888 | RADEON_TXFORMAT_ALPHA_IN_MAP;
107 break;
108 case MESA_FORMAT_RGBA8888:
109 txformat |= RADEON_TXFORMAT_RGBA8888 | RADEON_TXFORMAT_ALPHA_IN_MAP;
110 break;
111 case MESA_FORMAT_XRGB8888:
112 txformat |= RADEON_TXFORMAT_ARGB8888;
113 break;
114 case MESA_FORMAT_RGB565:
115 txformat |= RADEON_TXFORMAT_RGB565;
116 break;
117 case MESA_FORMAT_ARGB4444:
118 txformat |= RADEON_TXFORMAT_ARGB4444 | RADEON_TXFORMAT_ALPHA_IN_MAP;
119 break;
120 case MESA_FORMAT_ARGB1555:
121 txformat |= RADEON_TXFORMAT_ARGB1555 | RADEON_TXFORMAT_ALPHA_IN_MAP;
122 break;
123 case MESA_FORMAT_A8:
124 case MESA_FORMAT_I8:
125 txformat |= RADEON_TXFORMAT_I8 | RADEON_TXFORMAT_ALPHA_IN_MAP;
126 break;
127 case MESA_FORMAT_L8:
128 txformat |= RADEON_TXFORMAT_I8;
129 break;
130 case MESA_FORMAT_AL88:
131 txformat |= RADEON_TXFORMAT_AI88 | RADEON_TXFORMAT_ALPHA_IN_MAP;
132 break;
133 default:
134 break;
135 }
136
137 BEGIN_BATCH(18);
138 OUT_BATCH_REGVAL(RADEON_PP_CNTL, RADEON_TEX_0_ENABLE | RADEON_TEX_BLEND_0_ENABLE);
139 OUT_BATCH_REGVAL(RADEON_PP_TXCBLEND_0, (RADEON_COLOR_ARG_A_ZERO |
140 RADEON_COLOR_ARG_B_ZERO |
141 RADEON_COLOR_ARG_C_T0_COLOR |
142 RADEON_BLEND_CTL_ADD |
143 RADEON_CLAMP_TX));
144 OUT_BATCH_REGVAL(RADEON_PP_TXABLEND_0, (RADEON_ALPHA_ARG_A_ZERO |
145 RADEON_ALPHA_ARG_B_ZERO |
146 RADEON_ALPHA_ARG_C_T0_ALPHA |
147 RADEON_BLEND_CTL_ADD |
148 RADEON_CLAMP_TX));
149 OUT_BATCH_REGVAL(RADEON_PP_TXFILTER_0, (RADEON_CLAMP_S_CLAMP_LAST |
150 RADEON_CLAMP_T_CLAMP_LAST |
151 RADEON_MAG_FILTER_NEAREST |
152 RADEON_MIN_FILTER_NEAREST));
153 OUT_BATCH_REGVAL(RADEON_PP_TXFORMAT_0, txformat);
154 OUT_BATCH_REGVAL(RADEON_PP_TEX_SIZE_0, ((width - 1) |
155 ((height - 1) << RADEON_TEX_VSIZE_SHIFT)));
156 OUT_BATCH_REGVAL(RADEON_PP_TEX_PITCH_0, pitch * _mesa_get_format_bytes(mesa_format) - 32);
157
158 OUT_BATCH_REGSEQ(RADEON_PP_TXOFFSET_0, 1);
159 OUT_BATCH_RELOC(0, bo, 0, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
160
161 END_BATCH();
162 }
163
164 static inline void emit_cb_setup(struct r100_context *r100,
165 struct radeon_bo *bo,
166 intptr_t offset,
167 gl_format mesa_format,
168 unsigned pitch,
169 unsigned width,
170 unsigned height)
171 {
172 uint32_t dst_pitch = pitch;
173 uint32_t dst_format = 0;
174 BATCH_LOCALS(&r100->radeon);
175
176 /* XXX others? BE/LE? */
177 switch (mesa_format) {
178 case MESA_FORMAT_ARGB8888:
179 case MESA_FORMAT_XRGB8888:
180 dst_format = RADEON_COLOR_FORMAT_ARGB8888;
181 break;
182 case MESA_FORMAT_RGB565:
183 dst_format = RADEON_COLOR_FORMAT_RGB565;
184 break;
185 case MESA_FORMAT_ARGB4444:
186 dst_format = RADEON_COLOR_FORMAT_ARGB4444;
187 break;
188 case MESA_FORMAT_ARGB1555:
189 dst_format = RADEON_COLOR_FORMAT_ARGB1555;
190 break;
191 case MESA_FORMAT_A8:
192 case MESA_FORMAT_L8:
193 case MESA_FORMAT_I8:
194 dst_format = RADEON_COLOR_FORMAT_RGB8;
195 break;
196 default:
197 break;
198 }
199
200 BEGIN_BATCH_NO_AUTOSTATE(18);
201 OUT_BATCH_REGVAL(RADEON_RE_TOP_LEFT, 0);
202 OUT_BATCH_REGVAL(RADEON_RE_WIDTH_HEIGHT, ((width << RADEON_RE_WIDTH_SHIFT) |
203 (height << RADEON_RE_HEIGHT_SHIFT)));
204 OUT_BATCH_REGVAL(RADEON_RB3D_PLANEMASK, 0xffffffff);
205 OUT_BATCH_REGVAL(RADEON_RB3D_BLENDCNTL, RADEON_SRC_BLEND_GL_ONE | RADEON_DST_BLEND_GL_ZERO);
206 OUT_BATCH_REGVAL(RADEON_RB3D_CNTL, dst_format);
207
208 OUT_BATCH_REGSEQ(RADEON_RB3D_COLOROFFSET, 1);
209 OUT_BATCH_RELOC(0, bo, 0, 0, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0);
210 OUT_BATCH_REGSEQ(RADEON_RB3D_COLORPITCH, 1);
211 OUT_BATCH_RELOC(dst_pitch, bo, dst_pitch, 0, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0);
212
213 END_BATCH();
214 }
215
216 static GLboolean validate_buffers(struct r100_context *r100,
217 struct radeon_bo *src_bo,
218 struct radeon_bo *dst_bo)
219 {
220 int ret;
221
222 radeon_cs_space_reset_bos(r100->radeon.cmdbuf.cs);
223
224 ret = radeon_cs_space_check_with_bo(r100->radeon.cmdbuf.cs,
225 src_bo, RADEON_GEM_DOMAIN_VRAM | RADEON_GEM_DOMAIN_GTT, 0);
226 if (ret)
227 return GL_FALSE;
228
229 ret = radeon_cs_space_check_with_bo(r100->radeon.cmdbuf.cs,
230 dst_bo, 0, RADEON_GEM_DOMAIN_VRAM | RADEON_GEM_DOMAIN_GTT);
231 if (ret)
232 return GL_FALSE;
233
234 return GL_TRUE;
235 }
236
237 /**
238 * Calculate texcoords for given image region.
239 * Output values are [minx, maxx, miny, maxy]
240 */
241 static inline void calc_tex_coords(float img_width, float img_height,
242 float x, float y,
243 float reg_width, float reg_height,
244 unsigned flip_y, float *buf)
245 {
246 buf[0] = x / img_width;
247 buf[1] = buf[0] + reg_width / img_width;
248 buf[2] = y / img_height;
249 buf[3] = buf[2] + reg_height / img_height;
250 if (flip_y)
251 {
252 buf[2] = 1.0 - buf[2];
253 buf[3] = 1.0 - buf[3];
254 }
255 }
256
257 static inline void emit_draw_packet(struct r100_context *r100,
258 unsigned src_width, unsigned src_height,
259 unsigned src_x_offset, unsigned src_y_offset,
260 unsigned dst_x_offset, unsigned dst_y_offset,
261 unsigned reg_width, unsigned reg_height,
262 unsigned flip_y)
263 {
264 float texcoords[4];
265 float verts[12];
266 BATCH_LOCALS(&r100->radeon);
267
268 calc_tex_coords(src_width, src_height,
269 src_x_offset, src_y_offset,
270 reg_width, reg_height,
271 flip_y, texcoords);
272
273 verts[0] = dst_x_offset;
274 verts[1] = dst_y_offset + reg_height;
275 verts[2] = texcoords[0];
276 verts[3] = texcoords[3];
277
278 verts[4] = dst_x_offset + reg_width;
279 verts[5] = dst_y_offset + reg_height;
280 verts[6] = texcoords[1];
281 verts[7] = texcoords[3];
282
283 verts[8] = dst_x_offset + reg_width;
284 verts[9] = dst_y_offset;
285 verts[10] = texcoords[1];
286 verts[11] = texcoords[2];
287
288 BEGIN_BATCH(15);
289 OUT_BATCH(RADEON_CP_PACKET3_3D_DRAW_IMMD | (13 << 16));
290 OUT_BATCH(RADEON_CP_VC_FRMT_XY | RADEON_CP_VC_FRMT_ST0);
291 OUT_BATCH(RADEON_CP_VC_CNTL_PRIM_WALK_RING |
292 RADEON_CP_VC_CNTL_PRIM_TYPE_RECT_LIST |
293 RADEON_CP_VC_CNTL_MAOS_ENABLE |
294 RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE |
295 (3 << 16));
296 OUT_BATCH_TABLE(verts, 12);
297 END_BATCH();
298 }
299
300 /**
301 * Copy a region of [@a width x @a height] pixels from source buffer
302 * to destination buffer.
303 * @param[in] r100 r100 context
304 * @param[in] src_bo source radeon buffer object
305 * @param[in] src_offset offset of the source image in the @a src_bo
306 * @param[in] src_mesaformat source image format
307 * @param[in] src_pitch aligned source image width
308 * @param[in] src_width source image width
309 * @param[in] src_height source image height
310 * @param[in] src_x_offset x offset in the source image
311 * @param[in] src_y_offset y offset in the source image
312 * @param[in] dst_bo destination radeon buffer object
313 * @param[in] dst_offset offset of the destination image in the @a dst_bo
314 * @param[in] dst_mesaformat destination image format
315 * @param[in] dst_pitch aligned destination image width
316 * @param[in] dst_width destination image width
317 * @param[in] dst_height destination image height
318 * @param[in] dst_x_offset x offset in the destination image
319 * @param[in] dst_y_offset y offset in the destination image
320 * @param[in] width region width
321 * @param[in] height region height
322 * @param[in] flip_y set if y coords of the source image need to be flipped
323 */
324 unsigned r100_blit(GLcontext *ctx,
325 struct radeon_bo *src_bo,
326 intptr_t src_offset,
327 gl_format src_mesaformat,
328 unsigned src_pitch,
329 unsigned src_width,
330 unsigned src_height,
331 unsigned src_x_offset,
332 unsigned src_y_offset,
333 struct radeon_bo *dst_bo,
334 intptr_t dst_offset,
335 gl_format dst_mesaformat,
336 unsigned dst_pitch,
337 unsigned dst_width,
338 unsigned dst_height,
339 unsigned dst_x_offset,
340 unsigned dst_y_offset,
341 unsigned reg_width,
342 unsigned reg_height,
343 unsigned flip_y)
344 {
345 struct r100_context *r100 = R100_CONTEXT(ctx);
346
347 if (!r100_check_blit(dst_mesaformat))
348 return GL_FALSE;
349
350 /* Make sure that colorbuffer has even width - hw limitation */
351 if (dst_pitch % 2 > 0)
352 ++dst_pitch;
353
354 /* Rendering to small buffer doesn't work.
355 * Looks like a hw limitation.
356 */
357 if (dst_pitch < 32)
358 return GL_FALSE;
359
360 /* Need to clamp the region size to make sure
361 * we don't read outside of the source buffer
362 * or write outside of the destination buffer.
363 */
364 if (reg_width + src_x_offset > src_width)
365 reg_width = src_width - src_x_offset;
366 if (reg_height + src_y_offset > src_height)
367 reg_height = src_height - src_y_offset;
368 if (reg_width + dst_x_offset > dst_width)
369 reg_width = dst_width - dst_x_offset;
370 if (reg_height + dst_y_offset > dst_height)
371 reg_height = dst_height - dst_y_offset;
372
373 if (src_bo == dst_bo) {
374 return GL_FALSE;
375 }
376
377 if (src_offset % 32 || dst_offset % 32) {
378 return GL_FALSE;
379 }
380
381 if (0) {
382 fprintf(stderr, "src: size [%d x %d], pitch %d, "
383 "offset [%d x %d], format %s, bo %p\n",
384 src_width, src_height, src_pitch,
385 src_x_offset, src_y_offset,
386 _mesa_get_format_name(src_mesaformat),
387 src_bo);
388 fprintf(stderr, "dst: pitch %d, offset[%d x %d], format %s, bo %p\n",
389 dst_pitch, dst_x_offset, dst_y_offset,
390 _mesa_get_format_name(dst_mesaformat), dst_bo);
391 fprintf(stderr, "region: %d x %d\n", reg_width, reg_height);
392 }
393
394 /* Flush is needed to make sure that source buffer has correct data */
395 radeonFlush(ctx);
396
397 rcommonEnsureCmdBufSpace(&r100->radeon, 59, __FUNCTION__);
398
399 if (!validate_buffers(r100, src_bo, dst_bo))
400 return GL_FALSE;
401
402 /* 8 */
403 emit_vtx_state(r100);
404 /* 18 */
405 emit_tx_setup(r100, src_mesaformat, src_bo, src_offset, src_width, src_height, src_pitch);
406 /* 18 */
407 emit_cb_setup(r100, dst_bo, dst_offset, dst_mesaformat, dst_pitch, dst_width, dst_height);
408 /* 15 */
409 emit_draw_packet(r100, src_width, src_height,
410 src_x_offset, src_y_offset,
411 dst_x_offset, dst_y_offset,
412 reg_width, reg_height,
413 flip_y);
414
415 radeonFlush(ctx);
416
417 return GL_TRUE;
418 }