Merge branch '7.8'
[mesa.git] / src / mesa / drivers / dri / intel / intel_blit.c
1 /**************************************************************************
2 *
3 * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
4 * All Rights Reserved.
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * 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, sub license, 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 portions
16 * of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25 *
26 **************************************************************************/
27
28
29 #include "main/mtypes.h"
30 #include "main/context.h"
31 #include "main/enums.h"
32 #include "main/colormac.h"
33
34 #include "intel_blit.h"
35 #include "intel_buffers.h"
36 #include "intel_context.h"
37 #include "intel_fbo.h"
38 #include "intel_reg.h"
39 #include "intel_regions.h"
40 #include "intel_batchbuffer.h"
41
42 #define FILE_DEBUG_FLAG DEBUG_BLIT
43
44 static GLuint translate_raster_op(GLenum logicop)
45 {
46 switch(logicop) {
47 case GL_CLEAR: return 0x00;
48 case GL_AND: return 0x88;
49 case GL_AND_REVERSE: return 0x44;
50 case GL_COPY: return 0xCC;
51 case GL_AND_INVERTED: return 0x22;
52 case GL_NOOP: return 0xAA;
53 case GL_XOR: return 0x66;
54 case GL_OR: return 0xEE;
55 case GL_NOR: return 0x11;
56 case GL_EQUIV: return 0x99;
57 case GL_INVERT: return 0x55;
58 case GL_OR_REVERSE: return 0xDD;
59 case GL_COPY_INVERTED: return 0x33;
60 case GL_OR_INVERTED: return 0xBB;
61 case GL_NAND: return 0x77;
62 case GL_SET: return 0xFF;
63 default: return 0;
64 }
65 }
66
67
68 /* Copy BitBlt
69 */
70 GLboolean
71 intelEmitCopyBlit(struct intel_context *intel,
72 GLuint cpp,
73 GLshort src_pitch,
74 dri_bo *src_buffer,
75 GLuint src_offset,
76 uint32_t src_tiling,
77 GLshort dst_pitch,
78 dri_bo *dst_buffer,
79 GLuint dst_offset,
80 uint32_t dst_tiling,
81 GLshort src_x, GLshort src_y,
82 GLshort dst_x, GLshort dst_y,
83 GLshort w, GLshort h,
84 GLenum logic_op)
85 {
86 GLuint CMD, BR13, pass = 0;
87 int dst_y2 = dst_y + h;
88 int dst_x2 = dst_x + w;
89 dri_bo *aper_array[3];
90 BATCH_LOCALS;
91
92 /* Blits are in a different ringbuffer so we don't use them. */
93 if (intel->gen >= 6)
94 return GL_FALSE;
95
96 if (dst_tiling != I915_TILING_NONE) {
97 if (dst_offset & 4095)
98 return GL_FALSE;
99 if (dst_tiling == I915_TILING_Y)
100 return GL_FALSE;
101 }
102 if (src_tiling != I915_TILING_NONE) {
103 if (src_offset & 4095)
104 return GL_FALSE;
105 if (src_tiling == I915_TILING_Y)
106 return GL_FALSE;
107 }
108
109 /* do space check before going any further */
110 do {
111 aper_array[0] = intel->batch->buf;
112 aper_array[1] = dst_buffer;
113 aper_array[2] = src_buffer;
114
115 if (dri_bufmgr_check_aperture_space(aper_array, 3) != 0) {
116 intel_batchbuffer_flush(intel->batch);
117 pass++;
118 } else
119 break;
120 } while (pass < 2);
121
122 if (pass >= 2)
123 return GL_FALSE;
124
125 intel_batchbuffer_require_space(intel->batch, 8 * 4);
126 DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
127 __FUNCTION__,
128 src_buffer, src_pitch, src_offset, src_x, src_y,
129 dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h);
130
131 src_pitch *= cpp;
132 dst_pitch *= cpp;
133
134 BR13 = translate_raster_op(logic_op) << 16;
135
136 switch (cpp) {
137 case 1:
138 CMD = XY_SRC_COPY_BLT_CMD;
139 break;
140 case 2:
141 BR13 |= BR13_565;
142 CMD = XY_SRC_COPY_BLT_CMD;
143 break;
144 case 4:
145 BR13 |= BR13_8888;
146 CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
147 break;
148 default:
149 return GL_FALSE;
150 }
151
152 #ifndef I915
153 if (dst_tiling != I915_TILING_NONE) {
154 CMD |= XY_DST_TILED;
155 dst_pitch /= 4;
156 }
157 if (src_tiling != I915_TILING_NONE) {
158 CMD |= XY_SRC_TILED;
159 src_pitch /= 4;
160 }
161 #endif
162
163 if (dst_y2 <= dst_y || dst_x2 <= dst_x) {
164 return GL_TRUE;
165 }
166
167 assert(dst_x < dst_x2);
168 assert(dst_y < dst_y2);
169
170 BEGIN_BATCH(8);
171 OUT_BATCH(CMD);
172 OUT_BATCH(BR13 | (uint16_t)dst_pitch);
173 OUT_BATCH((dst_y << 16) | dst_x);
174 OUT_BATCH((dst_y2 << 16) | dst_x2);
175 OUT_RELOC_FENCED(dst_buffer,
176 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
177 dst_offset);
178 OUT_BATCH((src_y << 16) | src_x);
179 OUT_BATCH((uint16_t)src_pitch);
180 OUT_RELOC_FENCED(src_buffer,
181 I915_GEM_DOMAIN_RENDER, 0,
182 src_offset);
183 ADVANCE_BATCH();
184
185 intel_batchbuffer_emit_mi_flush(intel->batch);
186
187 return GL_TRUE;
188 }
189
190
191 /**
192 * Use blitting to clear the renderbuffers named by 'flags'.
193 * Note: we can't use the ctx->DrawBuffer->_ColorDrawBufferIndexes field
194 * since that might include software renderbuffers or renderbuffers
195 * which we're clearing with triangles.
196 * \param mask bitmask of BUFFER_BIT_* values indicating buffers to clear
197 */
198 void
199 intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
200 {
201 struct intel_context *intel = intel_context(ctx);
202 struct gl_framebuffer *fb = ctx->DrawBuffer;
203 GLuint clear_depth;
204 GLboolean all;
205 GLint cx, cy, cw, ch;
206 BATCH_LOCALS;
207
208 /* Blits are in a different ringbuffer so we don't use them. */
209 assert(intel->gen < 6);
210
211 /*
212 * Compute values for clearing the buffers.
213 */
214 clear_depth = 0;
215 if (mask & BUFFER_BIT_DEPTH) {
216 clear_depth = (GLuint) (fb->_DepthMax * ctx->Depth.Clear);
217 }
218 if (mask & BUFFER_BIT_STENCIL) {
219 clear_depth |= (ctx->Stencil.Clear & 0xff) << 24;
220 }
221
222 cx = fb->_Xmin;
223 if (fb->Name == 0)
224 cy = ctx->DrawBuffer->Height - fb->_Ymax;
225 else
226 cy = fb->_Ymin;
227 cw = fb->_Xmax - fb->_Xmin;
228 ch = fb->_Ymax - fb->_Ymin;
229
230 if (cw == 0 || ch == 0)
231 return;
232
233 GLuint buf;
234 all = (cw == fb->Width && ch == fb->Height);
235
236 /* Loop over all renderbuffers */
237 for (buf = 0; buf < BUFFER_COUNT && mask; buf++) {
238 const GLbitfield bufBit = 1 << buf;
239 struct intel_renderbuffer *irb;
240 drm_intel_bo *write_buffer;
241 int x1, y1, x2, y2;
242 uint32_t clear_val;
243 uint32_t BR13, CMD;
244 int pitch, cpp;
245 drm_intel_bo *aper_array[2];
246
247 if (!(mask & bufBit))
248 continue;
249
250 /* OK, clear this renderbuffer */
251 irb = intel_get_renderbuffer(fb, buf);
252 write_buffer = intel_region_buffer(intel, irb->region,
253 all ? INTEL_WRITE_FULL :
254 INTEL_WRITE_PART);
255 x1 = cx + irb->region->draw_x;
256 y1 = cy + irb->region->draw_y;
257 x2 = cx + cw + irb->region->draw_x;
258 y2 = cy + ch + irb->region->draw_y;
259
260 pitch = irb->region->pitch;
261 cpp = irb->region->cpp;
262
263 DBG("%s dst:buf(%p)/%d %d,%d sz:%dx%d\n",
264 __FUNCTION__,
265 irb->region->buffer, (pitch * cpp),
266 x1, y1, x2 - x1, y2 - y1);
267
268 BR13 = 0xf0 << 16;
269 CMD = XY_COLOR_BLT_CMD;
270
271 /* Setup the blit command */
272 if (cpp == 4) {
273 BR13 |= BR13_8888;
274 if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) {
275 if (mask & BUFFER_BIT_DEPTH)
276 CMD |= XY_BLT_WRITE_RGB;
277 if (mask & BUFFER_BIT_STENCIL)
278 CMD |= XY_BLT_WRITE_ALPHA;
279 } else {
280 /* clearing RGBA */
281 CMD |= XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
282 }
283 } else {
284 ASSERT(cpp == 2);
285 BR13 |= BR13_565;
286 }
287
288 assert(irb->region->tiling != I915_TILING_Y);
289
290 #ifndef I915
291 if (irb->region->tiling != I915_TILING_NONE) {
292 CMD |= XY_DST_TILED;
293 pitch /= 4;
294 }
295 #endif
296 BR13 |= (pitch * cpp);
297
298 if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) {
299 clear_val = clear_depth;
300 } else {
301 uint8_t clear[4];
302 GLclampf *color = ctx->Color.ClearColor;
303
304 CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]);
305 CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]);
306 CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]);
307 CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]);
308
309 switch (irb->Base.Format) {
310 case MESA_FORMAT_ARGB8888:
311 case MESA_FORMAT_XRGB8888:
312 clear_val = PACK_COLOR_8888(clear[3], clear[0],
313 clear[1], clear[2]);
314 break;
315 case MESA_FORMAT_RGB565:
316 clear_val = PACK_COLOR_565(clear[0], clear[1], clear[2]);
317 break;
318 case MESA_FORMAT_ARGB4444:
319 clear_val = PACK_COLOR_4444(clear[3], clear[0],
320 clear[1], clear[2]);
321 break;
322 case MESA_FORMAT_ARGB1555:
323 clear_val = PACK_COLOR_1555(clear[3], clear[0],
324 clear[1], clear[2]);
325 break;
326 default:
327 _mesa_problem(ctx, "Unexpected renderbuffer format: %d\n",
328 irb->Base.Format);
329 clear_val = 0;
330 }
331 }
332
333 assert(x1 < x2);
334 assert(y1 < y2);
335
336 /* do space check before going any further */
337 aper_array[0] = intel->batch->buf;
338 aper_array[1] = write_buffer;
339
340 if (drm_intel_bufmgr_check_aperture_space(aper_array,
341 ARRAY_SIZE(aper_array)) != 0) {
342 intel_batchbuffer_flush(intel->batch);
343 }
344
345 BEGIN_BATCH(6);
346 OUT_BATCH(CMD);
347 OUT_BATCH(BR13);
348 OUT_BATCH((y1 << 16) | x1);
349 OUT_BATCH((y2 << 16) | x2);
350 OUT_RELOC_FENCED(write_buffer,
351 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
352 0);
353 OUT_BATCH(clear_val);
354 ADVANCE_BATCH();
355
356 if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL)
357 mask &= ~(BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL);
358 else
359 mask &= ~bufBit; /* turn off bit, for faster loop exit */
360 }
361 }
362
363 GLboolean
364 intelEmitImmediateColorExpandBlit(struct intel_context *intel,
365 GLuint cpp,
366 GLubyte *src_bits, GLuint src_size,
367 GLuint fg_color,
368 GLshort dst_pitch,
369 dri_bo *dst_buffer,
370 GLuint dst_offset,
371 uint32_t dst_tiling,
372 GLshort x, GLshort y,
373 GLshort w, GLshort h,
374 GLenum logic_op)
375 {
376 int dwords = ALIGN(src_size, 8) / 4;
377 uint32_t opcode, br13, blit_cmd;
378
379 /* Blits are in a different ringbuffer so we don't use them. */
380 if (intel->gen >= 6)
381 return GL_FALSE;
382
383 if (dst_tiling != I915_TILING_NONE) {
384 if (dst_offset & 4095)
385 return GL_FALSE;
386 if (dst_tiling == I915_TILING_Y)
387 return GL_FALSE;
388 }
389
390 assert( logic_op - GL_CLEAR >= 0 );
391 assert( logic_op - GL_CLEAR < 0x10 );
392 assert(dst_pitch > 0);
393
394 if (w < 0 || h < 0)
395 return GL_TRUE;
396
397 dst_pitch *= cpp;
398
399 DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d, %d bytes %d dwords\n",
400 __FUNCTION__,
401 dst_buffer, dst_pitch, dst_offset, x, y, w, h, src_size, dwords);
402
403 intel_batchbuffer_require_space( intel->batch,
404 (8 * 4) +
405 (3 * 4) +
406 dwords * 4 );
407
408 opcode = XY_SETUP_BLT_CMD;
409 if (cpp == 4)
410 opcode |= XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
411 #ifndef I915
412 if (dst_tiling != I915_TILING_NONE) {
413 opcode |= XY_DST_TILED;
414 dst_pitch /= 4;
415 }
416 #endif
417
418 br13 = dst_pitch | (translate_raster_op(logic_op) << 16) | (1 << 29);
419 if (cpp == 2)
420 br13 |= BR13_565;
421 else
422 br13 |= BR13_8888;
423
424 blit_cmd = XY_TEXT_IMMEDIATE_BLIT_CMD | XY_TEXT_BYTE_PACKED; /* packing? */
425 if (dst_tiling != I915_TILING_NONE)
426 blit_cmd |= XY_DST_TILED;
427
428 BEGIN_BATCH(8 + 3);
429 OUT_BATCH(opcode);
430 OUT_BATCH(br13);
431 OUT_BATCH((0 << 16) | 0); /* clip x1, y1 */
432 OUT_BATCH((100 << 16) | 100); /* clip x2, y2 */
433 OUT_RELOC_FENCED(dst_buffer,
434 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
435 dst_offset);
436 OUT_BATCH(0); /* bg */
437 OUT_BATCH(fg_color); /* fg */
438 OUT_BATCH(0); /* pattern base addr */
439
440 OUT_BATCH(blit_cmd | ((3 - 2) + dwords));
441 OUT_BATCH((y << 16) | x);
442 OUT_BATCH(((y + h) << 16) | (x + w));
443 ADVANCE_BATCH();
444
445 intel_batchbuffer_data( intel->batch,
446 src_bits,
447 dwords * 4 );
448
449 intel_batchbuffer_emit_mi_flush(intel->batch);
450
451 return GL_TRUE;
452 }
453
454 /* We don't have a memmove-type blit like some other hardware, so we'll do a
455 * rectangular blit covering a large space, then emit 1-scanline blit at the
456 * end to cover the last if we need.
457 */
458 void
459 intel_emit_linear_blit(struct intel_context *intel,
460 drm_intel_bo *dst_bo,
461 unsigned int dst_offset,
462 drm_intel_bo *src_bo,
463 unsigned int src_offset,
464 unsigned int size)
465 {
466 GLuint pitch, height;
467 GLboolean ok;
468
469 /* Blits are in a different ringbuffer so we don't use them. */
470 assert(intel->gen < 6);
471
472 /* The pitch is a signed value. */
473 pitch = MIN2(size, (1 << 15) - 1);
474 height = size / pitch;
475 ok = intelEmitCopyBlit(intel, 1,
476 pitch, src_bo, src_offset, I915_TILING_NONE,
477 pitch, dst_bo, dst_offset, I915_TILING_NONE,
478 0, 0, /* src x/y */
479 0, 0, /* dst x/y */
480 pitch, height, /* w, h */
481 GL_COPY);
482 assert(ok);
483
484 src_offset += pitch * height;
485 dst_offset += pitch * height;
486 size -= pitch * height;
487 assert (size < (1 << 15));
488 if (size != 0) {
489 ok = intelEmitCopyBlit(intel, 1,
490 size, src_bo, src_offset, I915_TILING_NONE,
491 size, dst_bo, dst_offset, I915_TILING_NONE,
492 0, 0, /* src x/y */
493 0, 0, /* dst x/y */
494 size, 1, /* w, h */
495 GL_COPY);
496 assert(ok);
497 }
498 }