st_cb_drawpixels.c revision 738d1ae3b59b073605c7fb58ec6f0a85a185ffd3
1/**************************************************************************
2 *
3 * Copyright 2007 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  * Authors:
30  *   Brian Paul
31  */
32
33#include "main/imports.h"
34#include "main/image.h"
35#include "main/bufferobj.h"
36#include "main/format_pack.h"
37#include "main/macros.h"
38#include "main/mfeatures.h"
39#include "main/mtypes.h"
40#include "main/pack.h"
41#include "main/pbo.h"
42#include "main/readpix.h"
43#include "main/texformat.h"
44#include "main/teximage.h"
45#include "main/texstore.h"
46#include "main/glformats.h"
47#include "program/program.h"
48#include "program/prog_print.h"
49#include "program/prog_instruction.h"
50
51#include "st_atom.h"
52#include "st_atom_constbuf.h"
53#include "st_cb_drawpixels.h"
54#include "st_cb_readpixels.h"
55#include "st_cb_fbo.h"
56#include "st_context.h"
57#include "st_debug.h"
58#include "st_format.h"
59#include "st_program.h"
60#include "st_texture.h"
61
62#include "pipe/p_context.h"
63#include "pipe/p_defines.h"
64#include "tgsi/tgsi_ureg.h"
65#include "util/u_draw_quad.h"
66#include "util/u_format.h"
67#include "util/u_inlines.h"
68#include "util/u_math.h"
69#include "util/u_tile.h"
70#include "util/u_upload_mgr.h"
71#include "cso_cache/cso_context.h"
72
73
74#if FEATURE_drawpix
75
76/**
77 * Check if the given program is:
78 * 0: MOVE result.color, fragment.color;
79 * 1: END;
80 */
81static GLboolean
82is_passthrough_program(const struct gl_fragment_program *prog)
83{
84   if (prog->Base.NumInstructions == 2) {
85      const struct prog_instruction *inst = prog->Base.Instructions;
86      if (inst[0].Opcode == OPCODE_MOV &&
87          inst[1].Opcode == OPCODE_END &&
88          inst[0].DstReg.File == PROGRAM_OUTPUT &&
89          inst[0].DstReg.Index == FRAG_RESULT_COLOR &&
90          inst[0].DstReg.WriteMask == WRITEMASK_XYZW &&
91          inst[0].SrcReg[0].File == PROGRAM_INPUT &&
92          inst[0].SrcReg[0].Index == FRAG_ATTRIB_COL0 &&
93          inst[0].SrcReg[0].Swizzle == SWIZZLE_XYZW) {
94         return GL_TRUE;
95      }
96   }
97   return GL_FALSE;
98}
99
100
101/**
102 * Returns a fragment program which implements the current pixel transfer ops.
103 */
104static struct gl_fragment_program *
105get_glsl_pixel_transfer_program(struct st_context *st,
106                                struct st_fragment_program *orig)
107{
108   int pixelMaps = 0, scaleAndBias = 0;
109   struct gl_context *ctx = st->ctx;
110   struct st_fragment_program *fp = (struct st_fragment_program *)
111      ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
112
113   if (!fp)
114      return NULL;
115
116   if (ctx->Pixel.RedBias != 0.0 || ctx->Pixel.RedScale != 1.0 ||
117       ctx->Pixel.GreenBias != 0.0 || ctx->Pixel.GreenScale != 1.0 ||
118       ctx->Pixel.BlueBias != 0.0 || ctx->Pixel.BlueScale != 1.0 ||
119       ctx->Pixel.AlphaBias != 0.0 || ctx->Pixel.AlphaScale != 1.0) {
120      scaleAndBias = 1;
121   }
122
123   pixelMaps = ctx->Pixel.MapColorFlag;
124
125   if (pixelMaps) {
126      /* create the colormap/texture now if not already done */
127      if (!st->pixel_xfer.pixelmap_texture) {
128         st->pixel_xfer.pixelmap_texture = st_create_color_map_texture(ctx);
129         st->pixel_xfer.pixelmap_sampler_view =
130            st_create_texture_sampler_view(st->pipe,
131                                           st->pixel_xfer.pixelmap_texture);
132      }
133   }
134
135   get_pixel_transfer_visitor(fp, orig->glsl_to_tgsi,
136                              scaleAndBias, pixelMaps);
137
138   return &fp->Base;
139}
140
141
142/**
143 * Make fragment shader for glDraw/CopyPixels.  This shader is made
144 * by combining the pixel transfer shader with the user-defined shader.
145 * \param fpIn  the current/incoming fragment program
146 * \param fpOut  returns the combined fragment program
147 */
148void
149st_make_drawpix_fragment_program(struct st_context *st,
150                                 struct gl_fragment_program *fpIn,
151                                 struct gl_fragment_program **fpOut)
152{
153   struct gl_program *newProg;
154   struct st_fragment_program *stfp = (struct st_fragment_program *) fpIn;
155
156   if (is_passthrough_program(fpIn)) {
157      newProg = (struct gl_program *) _mesa_clone_fragment_program(st->ctx,
158                                             &st->pixel_xfer.program->Base);
159   }
160   else if (stfp->glsl_to_tgsi != NULL) {
161      newProg = (struct gl_program *) get_glsl_pixel_transfer_program(st, stfp);
162   }
163   else {
164#if 0
165      /* debug */
166      printf("Base program:\n");
167      _mesa_print_program(&fpIn->Base);
168      printf("DrawPix program:\n");
169      _mesa_print_program(&st->pixel_xfer.program->Base.Base);
170#endif
171      newProg = _mesa_combine_programs(st->ctx,
172                                       &st->pixel_xfer.program->Base.Base,
173                                       &fpIn->Base);
174   }
175
176#if 0
177   /* debug */
178   printf("Combined DrawPixels program:\n");
179   _mesa_print_program(newProg);
180   printf("InputsRead: 0x%x\n", newProg->InputsRead);
181   printf("OutputsWritten: 0x%x\n", newProg->OutputsWritten);
182   _mesa_print_parameter_list(newProg->Parameters);
183#endif
184
185   *fpOut = (struct gl_fragment_program *) newProg;
186}
187
188
189/**
190 * Create fragment program that does a TEX() instruction to get a Z and/or
191 * stencil value value, then writes to FRAG_RESULT_DEPTH/FRAG_RESULT_STENCIL.
192 * Used for glDrawPixels(GL_DEPTH_COMPONENT / GL_STENCIL_INDEX).
193 * Pass fragment color through as-is.
194 * \return pointer to the gl_fragment program
195 */
196struct gl_fragment_program *
197st_make_drawpix_z_stencil_program(struct st_context *st,
198                                  GLboolean write_depth,
199                                  GLboolean write_stencil)
200{
201   struct gl_context *ctx = st->ctx;
202   struct gl_program *p;
203   struct gl_fragment_program *fp;
204   GLuint ic = 0;
205   const GLuint shaderIndex = write_depth * 2 + write_stencil;
206
207   assert(shaderIndex < Elements(st->drawpix.shaders));
208
209   if (st->drawpix.shaders[shaderIndex]) {
210      /* already have the proper shader */
211      return st->drawpix.shaders[shaderIndex];
212   }
213
214   /*
215    * Create shader now
216    */
217   p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
218   if (!p)
219      return NULL;
220
221   p->NumInstructions = write_depth ? 3 : 1;
222   p->NumInstructions += write_stencil ? 1 : 0;
223
224   p->Instructions = _mesa_alloc_instructions(p->NumInstructions);
225   if (!p->Instructions) {
226      ctx->Driver.DeleteProgram(ctx, p);
227      return NULL;
228   }
229   _mesa_init_instructions(p->Instructions, p->NumInstructions);
230
231   if (write_depth) {
232      /* TEX result.depth, fragment.texcoord[0], texture[0], 2D; */
233      p->Instructions[ic].Opcode = OPCODE_TEX;
234      p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
235      p->Instructions[ic].DstReg.Index = FRAG_RESULT_DEPTH;
236      p->Instructions[ic].DstReg.WriteMask = WRITEMASK_Z;
237      p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
238      p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
239      p->Instructions[ic].TexSrcUnit = 0;
240      p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX;
241      ic++;
242      /* MOV result.color, fragment.color; */
243      p->Instructions[ic].Opcode = OPCODE_MOV;
244      p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
245      p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLOR;
246      p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
247      p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_COL0;
248      ic++;
249   }
250
251   if (write_stencil) {
252      /* TEX result.stencil, fragment.texcoord[0], texture[0], 2D; */
253      p->Instructions[ic].Opcode = OPCODE_TEX;
254      p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
255      p->Instructions[ic].DstReg.Index = FRAG_RESULT_STENCIL;
256      p->Instructions[ic].DstReg.WriteMask = WRITEMASK_Y;
257      p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
258      p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
259      p->Instructions[ic].TexSrcUnit = 1;
260      p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX;
261      ic++;
262   }
263
264   /* END; */
265   p->Instructions[ic++].Opcode = OPCODE_END;
266
267   assert(ic == p->NumInstructions);
268
269   p->InputsRead = FRAG_BIT_TEX0 | FRAG_BIT_COL0;
270   p->OutputsWritten = 0;
271   if (write_depth) {
272      p->OutputsWritten |= BITFIELD64_BIT(FRAG_RESULT_DEPTH);
273      p->OutputsWritten |= BITFIELD64_BIT(FRAG_RESULT_COLOR);
274   }
275   if (write_stencil)
276      p->OutputsWritten |= BITFIELD64_BIT(FRAG_RESULT_STENCIL);
277
278   p->SamplersUsed =  0x1;  /* sampler 0 (bit 0) is used */
279   if (write_stencil)
280      p->SamplersUsed |= 1 << 1;
281
282   fp = (struct gl_fragment_program *) p;
283
284   /* save the new shader */
285   st->drawpix.shaders[shaderIndex] = fp;
286
287   return fp;
288}
289
290
291/**
292 * Create a simple vertex shader that just passes through the
293 * vertex position and texcoord (and optionally, color).
294 */
295static void *
296make_passthrough_vertex_shader(struct st_context *st,
297                               GLboolean passColor)
298{
299   if (!st->drawpix.vert_shaders[passColor]) {
300      struct ureg_program *ureg = ureg_create( TGSI_PROCESSOR_VERTEX );
301
302      if (ureg == NULL)
303         return NULL;
304
305      /* MOV result.pos, vertex.pos; */
306      ureg_MOV(ureg,
307               ureg_DECL_output( ureg, TGSI_SEMANTIC_POSITION, 0 ),
308               ureg_DECL_vs_input( ureg, 0 ));
309
310      /* MOV result.texcoord0, vertex.attr[1]; */
311      ureg_MOV(ureg,
312               ureg_DECL_output( ureg, TGSI_SEMANTIC_GENERIC, 0 ),
313               ureg_DECL_vs_input( ureg, 1 ));
314
315      if (passColor) {
316         /* MOV result.color0, vertex.attr[2]; */
317         ureg_MOV(ureg,
318                  ureg_DECL_output( ureg, TGSI_SEMANTIC_COLOR, 0 ),
319                  ureg_DECL_vs_input( ureg, 2 ));
320      }
321
322      ureg_END( ureg );
323
324      st->drawpix.vert_shaders[passColor] =
325         ureg_create_shader_and_destroy( ureg, st->pipe );
326   }
327
328   return st->drawpix.vert_shaders[passColor];
329}
330
331
332/**
333 * Return a texture internalFormat for drawing/copying an image
334 * of the given format and type.
335 */
336static GLenum
337internal_format(struct gl_context *ctx, GLenum format, GLenum type)
338{
339   switch (format) {
340   case GL_DEPTH_COMPONENT:
341      switch (type) {
342      case GL_UNSIGNED_SHORT:
343         return GL_DEPTH_COMPONENT16;
344
345      case GL_UNSIGNED_INT:
346         return GL_DEPTH_COMPONENT32;
347
348      case GL_FLOAT:
349         if (ctx->Extensions.ARB_depth_buffer_float)
350            return GL_DEPTH_COMPONENT32F;
351         else
352            return GL_DEPTH_COMPONENT;
353
354      default:
355         return GL_DEPTH_COMPONENT;
356      }
357
358   case GL_DEPTH_STENCIL:
359      switch (type) {
360      case GL_FLOAT_32_UNSIGNED_INT_24_8_REV:
361         return GL_DEPTH32F_STENCIL8;
362
363      case GL_UNSIGNED_INT_24_8:
364      default:
365         return GL_DEPTH24_STENCIL8;
366      }
367
368   case GL_STENCIL_INDEX:
369      return GL_STENCIL_INDEX;
370
371   default:
372      if (_mesa_is_enum_format_integer(format)) {
373         switch (type) {
374         case GL_BYTE:
375            return GL_RGBA8I;
376         case GL_UNSIGNED_BYTE:
377            return GL_RGBA8UI;
378         case GL_SHORT:
379            return GL_RGBA16I;
380         case GL_UNSIGNED_SHORT:
381            return GL_RGBA16UI;
382         case GL_INT:
383            return GL_RGBA32I;
384         case GL_UNSIGNED_INT:
385            return GL_RGBA32UI;
386         default:
387            assert(0 && "Unexpected type in internal_format()");
388            return GL_RGBA_INTEGER;
389         }
390      }
391      else {
392         switch (type) {
393         case GL_UNSIGNED_BYTE:
394         case GL_UNSIGNED_INT_8_8_8_8:
395         case GL_UNSIGNED_INT_8_8_8_8_REV:
396         default:
397            return GL_RGBA8;
398
399         case GL_UNSIGNED_BYTE_3_3_2:
400         case GL_UNSIGNED_BYTE_2_3_3_REV:
401            return GL_R3_G3_B2;
402
403         case GL_UNSIGNED_SHORT_4_4_4_4:
404         case GL_UNSIGNED_SHORT_4_4_4_4_REV:
405            return GL_RGBA4;
406
407         case GL_UNSIGNED_SHORT_5_6_5:
408         case GL_UNSIGNED_SHORT_5_6_5_REV:
409            return GL_RGB565;
410
411         case GL_UNSIGNED_SHORT_5_5_5_1:
412         case GL_UNSIGNED_SHORT_1_5_5_5_REV:
413            return GL_RGB5_A1;
414
415         case GL_UNSIGNED_INT_10_10_10_2:
416         case GL_UNSIGNED_INT_2_10_10_10_REV:
417            return GL_RGB10_A2;
418
419         case GL_UNSIGNED_SHORT:
420         case GL_UNSIGNED_INT:
421            return GL_RGBA16;
422
423         case GL_BYTE:
424            return
425               ctx->Extensions.EXT_texture_snorm ? GL_RGBA8_SNORM : GL_RGBA8;
426
427         case GL_SHORT:
428         case GL_INT:
429            return
430               ctx->Extensions.EXT_texture_snorm ? GL_RGBA16_SNORM : GL_RGBA16;
431
432         case GL_HALF_FLOAT_ARB:
433            return
434               ctx->Extensions.ARB_texture_float ? GL_RGBA16F :
435               ctx->Extensions.EXT_texture_snorm ? GL_RGBA16_SNORM : GL_RGBA16;
436
437         case GL_FLOAT:
438         case GL_DOUBLE:
439            return
440               ctx->Extensions.ARB_texture_float ? GL_RGBA32F :
441               ctx->Extensions.EXT_texture_snorm ? GL_RGBA16_SNORM : GL_RGBA16;
442
443         case GL_UNSIGNED_INT_5_9_9_9_REV:
444            assert(ctx->Extensions.EXT_texture_shared_exponent);
445            return GL_RGB9_E5;
446
447         case GL_UNSIGNED_INT_10F_11F_11F_REV:
448            assert(ctx->Extensions.EXT_packed_float);
449            return GL_R11F_G11F_B10F;
450         }
451      }
452   }
453}
454
455
456/**
457 * Create a temporary texture to hold an image of the given size.
458 * If width, height are not POT and the driver only handles POT textures,
459 * allocate the next larger size of texture that is POT.
460 */
461static struct pipe_resource *
462alloc_texture(struct st_context *st, GLsizei width, GLsizei height,
463              enum pipe_format texFormat)
464{
465   struct pipe_resource *pt;
466
467   pt = st_texture_create(st, st->internal_target, texFormat, 0,
468                          width, height, 1, 1, PIPE_BIND_SAMPLER_VIEW);
469
470   return pt;
471}
472
473
474/**
475 * Make texture containing an image for glDrawPixels image.
476 * If 'pixels' is NULL, leave the texture image data undefined.
477 */
478static struct pipe_resource *
479make_texture(struct st_context *st,
480	     GLsizei width, GLsizei height, GLenum format, GLenum type,
481	     const struct gl_pixelstore_attrib *unpack,
482	     const GLvoid *pixels)
483{
484   struct gl_context *ctx = st->ctx;
485   struct pipe_context *pipe = st->pipe;
486   gl_format mformat;
487   struct pipe_resource *pt;
488   enum pipe_format pipeFormat;
489   GLenum baseInternalFormat, intFormat;
490
491   intFormat = internal_format(ctx, format, type);
492   baseInternalFormat = _mesa_base_tex_format(ctx, intFormat);
493
494   mformat = st_ChooseTextureFormat_renderable(ctx, intFormat,
495                                               format, type, GL_FALSE);
496   assert(mformat);
497
498   pipeFormat = st_mesa_format_to_pipe_format(mformat);
499   assert(pipeFormat);
500
501   pixels = _mesa_map_pbo_source(ctx, unpack, pixels);
502   if (!pixels)
503      return NULL;
504
505   /* alloc temporary texture */
506   pt = alloc_texture(st, width, height, pipeFormat);
507   if (!pt) {
508      _mesa_unmap_pbo_source(ctx, unpack);
509      return NULL;
510   }
511
512   {
513      struct pipe_transfer *transfer;
514      GLboolean success;
515      GLubyte *dest;
516      const GLbitfield imageTransferStateSave = ctx->_ImageTransferState;
517
518      /* we'll do pixel transfer in a fragment shader */
519      ctx->_ImageTransferState = 0x0;
520
521      transfer = pipe_get_transfer(st->pipe, pt, 0, 0,
522                                   PIPE_TRANSFER_WRITE, 0, 0,
523                                   width, height);
524
525      /* map texture transfer */
526      dest = pipe_transfer_map(pipe, transfer);
527
528
529      /* Put image into texture transfer.
530       * Note that the image is actually going to be upside down in
531       * the texture.  We deal with that with texcoords.
532       */
533      success = _mesa_texstore(ctx, 2,           /* dims */
534                               baseInternalFormat, /* baseInternalFormat */
535                               mformat,          /* gl_format */
536                               transfer->stride, /* dstRowStride, bytes */
537                               &dest,            /* destSlices */
538                               width, height, 1, /* size */
539                               format, type,     /* src format/type */
540                               pixels,           /* data source */
541                               unpack);
542
543      /* unmap */
544      pipe_transfer_unmap(pipe, transfer);
545      pipe->transfer_destroy(pipe, transfer);
546
547      assert(success);
548
549      /* restore */
550      ctx->_ImageTransferState = imageTransferStateSave;
551   }
552
553   _mesa_unmap_pbo_source(ctx, unpack);
554
555   return pt;
556}
557
558
559/**
560 * Draw quad with texcoords and optional color.
561 * Coords are gallium window coords with y=0=top.
562 * \param color  may be null
563 * \param invertTex  if true, flip texcoords vertically
564 */
565static void
566draw_quad(struct gl_context *ctx, GLfloat x0, GLfloat y0, GLfloat z,
567          GLfloat x1, GLfloat y1, const GLfloat *color,
568          GLboolean invertTex, GLfloat maxXcoord, GLfloat maxYcoord)
569{
570   struct st_context *st = st_context(ctx);
571   struct pipe_context *pipe = st->pipe;
572   GLfloat (*verts)[3][4]; /* four verts, three attribs, XYZW */
573   struct pipe_resource *buf = NULL;
574   unsigned offset;
575
576   u_upload_alloc(st->uploader, 0, 4 * sizeof(verts[0]), &offset, &buf,
577		  (void**)&verts);
578   if (!buf) {
579      return;
580   }
581
582   /* setup vertex data */
583   {
584      const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
585      const GLfloat fb_width = (GLfloat) fb->Width;
586      const GLfloat fb_height = (GLfloat) fb->Height;
587      const GLfloat clip_x0 = x0 / fb_width * 2.0f - 1.0f;
588      const GLfloat clip_y0 = y0 / fb_height * 2.0f - 1.0f;
589      const GLfloat clip_x1 = x1 / fb_width * 2.0f - 1.0f;
590      const GLfloat clip_y1 = y1 / fb_height * 2.0f - 1.0f;
591      const GLfloat sLeft = 0.0f, sRight = maxXcoord;
592      const GLfloat tTop = invertTex ? maxYcoord : 0.0f;
593      const GLfloat tBot = invertTex ? 0.0f : maxYcoord;
594      GLuint i;
595
596      /* upper-left */
597      verts[0][0][0] = clip_x0;    /* v[0].attr[0].x */
598      verts[0][0][1] = clip_y0;    /* v[0].attr[0].y */
599
600      /* upper-right */
601      verts[1][0][0] = clip_x1;
602      verts[1][0][1] = clip_y0;
603
604      /* lower-right */
605      verts[2][0][0] = clip_x1;
606      verts[2][0][1] = clip_y1;
607
608      /* lower-left */
609      verts[3][0][0] = clip_x0;
610      verts[3][0][1] = clip_y1;
611
612      verts[0][1][0] = sLeft; /* v[0].attr[1].S */
613      verts[0][1][1] = tTop;  /* v[0].attr[1].T */
614      verts[1][1][0] = sRight;
615      verts[1][1][1] = tTop;
616      verts[2][1][0] = sRight;
617      verts[2][1][1] = tBot;
618      verts[3][1][0] = sLeft;
619      verts[3][1][1] = tBot;
620
621      /* same for all verts: */
622      if (color) {
623         for (i = 0; i < 4; i++) {
624            verts[i][0][2] = z;         /* v[i].attr[0].z */
625            verts[i][0][3] = 1.0f;      /* v[i].attr[0].w */
626            verts[i][2][0] = color[0];  /* v[i].attr[2].r */
627            verts[i][2][1] = color[1];  /* v[i].attr[2].g */
628            verts[i][2][2] = color[2];  /* v[i].attr[2].b */
629            verts[i][2][3] = color[3];  /* v[i].attr[2].a */
630            verts[i][1][2] = 0.0f;      /* v[i].attr[1].R */
631            verts[i][1][3] = 1.0f;      /* v[i].attr[1].Q */
632         }
633      }
634      else {
635         for (i = 0; i < 4; i++) {
636            verts[i][0][2] = z;    /*Z*/
637            verts[i][0][3] = 1.0f; /*W*/
638            verts[i][1][2] = 0.0f; /*R*/
639            verts[i][1][3] = 1.0f; /*Q*/
640         }
641      }
642   }
643
644   u_upload_unmap(st->uploader);
645   util_draw_vertex_buffer(pipe, st->cso_context, buf, offset,
646			   PIPE_PRIM_QUADS,
647			   4,  /* verts */
648			   3); /* attribs/vert */
649   pipe_resource_reference(&buf, NULL);
650}
651
652
653
654static void
655draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
656                   GLsizei width, GLsizei height,
657                   GLfloat zoomX, GLfloat zoomY,
658                   struct pipe_sampler_view **sv,
659                   int num_sampler_view,
660                   void *driver_vp,
661                   void *driver_fp,
662                   const GLfloat *color,
663                   GLboolean invertTex,
664                   GLboolean write_depth, GLboolean write_stencil)
665{
666   struct st_context *st = st_context(ctx);
667   struct pipe_context *pipe = st->pipe;
668   struct cso_context *cso = st->cso_context;
669   GLfloat x0, y0, x1, y1;
670   GLsizei maxSize;
671   boolean normalized = sv[0]->texture->target != PIPE_TEXTURE_RECT;
672
673   /* limit checks */
674   /* XXX if DrawPixels image is larger than max texture size, break
675    * it up into chunks.
676    */
677   maxSize = 1 << (pipe->screen->get_param(pipe->screen,
678                                        PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
679   assert(width <= maxSize);
680   assert(height <= maxSize);
681
682   cso_save_rasterizer(cso);
683   cso_save_viewport(cso);
684   cso_save_samplers(cso, PIPE_SHADER_FRAGMENT);
685   cso_save_sampler_views(cso, PIPE_SHADER_FRAGMENT);
686   cso_save_fragment_shader(cso);
687   cso_save_stream_outputs(cso);
688   cso_save_vertex_shader(cso);
689   cso_save_geometry_shader(cso);
690   cso_save_vertex_elements(cso);
691   cso_save_vertex_buffers(cso);
692   if (write_stencil) {
693      cso_save_depth_stencil_alpha(cso);
694      cso_save_blend(cso);
695   }
696
697   /* rasterizer state: just scissor */
698   {
699      struct pipe_rasterizer_state rasterizer;
700      memset(&rasterizer, 0, sizeof(rasterizer));
701      rasterizer.clamp_fragment_color = !st->clamp_frag_color_in_shader &&
702                                        ctx->Color._ClampFragmentColor &&
703                                        !ctx->DrawBuffer->_IntegerColor;
704      rasterizer.gl_rasterization_rules = 1;
705      rasterizer.depth_clip = !ctx->Transform.DepthClamp;
706      rasterizer.scissor = ctx->Scissor.Enabled;
707      cso_set_rasterizer(cso, &rasterizer);
708   }
709
710   if (write_stencil) {
711      /* Stencil writing bypasses the normal fragment pipeline to
712       * disable color writing and set stencil test to always pass.
713       */
714      struct pipe_depth_stencil_alpha_state dsa;
715      struct pipe_blend_state blend;
716
717      /* depth/stencil */
718      memset(&dsa, 0, sizeof(dsa));
719      dsa.stencil[0].enabled = 1;
720      dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
721      dsa.stencil[0].writemask = ctx->Stencil.WriteMask[0] & 0xff;
722      dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
723      if (write_depth) {
724         /* writing depth+stencil: depth test always passes */
725         dsa.depth.enabled = 1;
726         dsa.depth.writemask = ctx->Depth.Mask;
727         dsa.depth.func = PIPE_FUNC_ALWAYS;
728      }
729      cso_set_depth_stencil_alpha(cso, &dsa);
730
731      /* blend (colormask) */
732      memset(&blend, 0, sizeof(blend));
733      cso_set_blend(cso, &blend);
734   }
735
736   /* fragment shader state: TEX lookup program */
737   cso_set_fragment_shader_handle(cso, driver_fp);
738
739   /* vertex shader state: position + texcoord pass-through */
740   cso_set_vertex_shader_handle(cso, driver_vp);
741
742   /* geometry shader state: disabled */
743   cso_set_geometry_shader_handle(cso, NULL);
744
745   /* texture sampling state: */
746   {
747      struct pipe_sampler_state sampler;
748      memset(&sampler, 0, sizeof(sampler));
749      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP;
750      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP;
751      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP;
752      sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
753      sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
754      sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
755      sampler.normalized_coords = normalized;
756
757      cso_single_sampler(cso, PIPE_SHADER_FRAGMENT, 0, &sampler);
758      if (num_sampler_view > 1) {
759         cso_single_sampler(cso, PIPE_SHADER_FRAGMENT, 1, &sampler);
760      }
761      cso_single_sampler_done(cso, PIPE_SHADER_FRAGMENT);
762   }
763
764   /* viewport state: viewport matching window dims */
765   {
766      const float w = (float) ctx->DrawBuffer->Width;
767      const float h = (float) ctx->DrawBuffer->Height;
768      struct pipe_viewport_state vp;
769      vp.scale[0] =  0.5f * w;
770      vp.scale[1] = -0.5f * h;
771      vp.scale[2] = 0.5f;
772      vp.scale[3] = 1.0f;
773      vp.translate[0] = 0.5f * w;
774      vp.translate[1] = 0.5f * h;
775      vp.translate[2] = 0.5f;
776      vp.translate[3] = 0.0f;
777      cso_set_viewport(cso, &vp);
778   }
779
780   cso_set_vertex_elements(cso, 3, st->velems_util_draw);
781   cso_set_stream_outputs(st->cso_context, 0, NULL, 0);
782
783   /* texture state: */
784   cso_set_sampler_views(cso, PIPE_SHADER_FRAGMENT, num_sampler_view, sv);
785
786   /* Compute Gallium window coords (y=0=top) with pixel zoom.
787    * Recall that these coords are transformed by the current
788    * vertex shader and viewport transformation.
789    */
790   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_BOTTOM) {
791      y = ctx->DrawBuffer->Height - (int) (y + height * ctx->Pixel.ZoomY);
792      invertTex = !invertTex;
793   }
794
795   x0 = (GLfloat) x;
796   x1 = x + width * ctx->Pixel.ZoomX;
797   y0 = (GLfloat) y;
798   y1 = y + height * ctx->Pixel.ZoomY;
799
800   /* convert Z from [0,1] to [-1,-1] to match viewport Z scale/bias */
801   z = z * 2.0 - 1.0;
802
803   draw_quad(ctx, x0, y0, z, x1, y1, color, invertTex,
804             normalized ? ((GLfloat) width / sv[0]->texture->width0) : (GLfloat)width,
805             normalized ? ((GLfloat) height / sv[0]->texture->height0) : (GLfloat)height);
806
807   /* restore state */
808   cso_restore_rasterizer(cso);
809   cso_restore_viewport(cso);
810   cso_restore_samplers(cso, PIPE_SHADER_FRAGMENT);
811   cso_restore_sampler_views(cso, PIPE_SHADER_FRAGMENT);
812   cso_restore_fragment_shader(cso);
813   cso_restore_vertex_shader(cso);
814   cso_restore_geometry_shader(cso);
815   cso_restore_vertex_elements(cso);
816   cso_restore_vertex_buffers(cso);
817   cso_restore_stream_outputs(cso);
818   if (write_stencil) {
819      cso_restore_depth_stencil_alpha(cso);
820      cso_restore_blend(cso);
821   }
822}
823
824
825/**
826 * Software fallback to do glDrawPixels(GL_STENCIL_INDEX) when we
827 * can't use a fragment shader to write stencil values.
828 */
829static void
830draw_stencil_pixels(struct gl_context *ctx, GLint x, GLint y,
831                    GLsizei width, GLsizei height, GLenum format, GLenum type,
832                    const struct gl_pixelstore_attrib *unpack,
833                    const GLvoid *pixels)
834{
835   struct st_context *st = st_context(ctx);
836   struct pipe_context *pipe = st->pipe;
837   struct st_renderbuffer *strb;
838   enum pipe_transfer_usage usage;
839   struct pipe_transfer *pt;
840   const GLboolean zoom = ctx->Pixel.ZoomX != 1.0 || ctx->Pixel.ZoomY != 1.0;
841   ubyte *stmap;
842   struct gl_pixelstore_attrib clippedUnpack = *unpack;
843   GLubyte *sValues;
844   GLuint *zValues;
845
846   if (!zoom) {
847      if (!_mesa_clip_drawpixels(ctx, &x, &y, &width, &height,
848                                 &clippedUnpack)) {
849         /* totally clipped */
850         return;
851      }
852   }
853
854   strb = st_renderbuffer(ctx->DrawBuffer->
855                          Attachment[BUFFER_STENCIL].Renderbuffer);
856
857   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
858      y = ctx->DrawBuffer->Height - y - height;
859   }
860
861   if (format == GL_STENCIL_INDEX &&
862       _mesa_is_format_packed_depth_stencil(strb->Base.Format)) {
863      /* writing stencil to a combined depth+stencil buffer */
864      usage = PIPE_TRANSFER_READ_WRITE;
865   }
866   else {
867      usage = PIPE_TRANSFER_WRITE;
868   }
869
870   pt = pipe_get_transfer(pipe, strb->texture,
871                          strb->rtt_level, strb->rtt_face + strb->rtt_slice,
872                          usage, x, y,
873                          width, height);
874
875   stmap = pipe_transfer_map(pipe, pt);
876
877   pixels = _mesa_map_pbo_source(ctx, &clippedUnpack, pixels);
878   assert(pixels);
879
880   sValues = (GLubyte *) malloc(width * sizeof(GLubyte));
881   zValues = (GLuint *) malloc(width * sizeof(GLuint));
882
883   if (sValues && zValues) {
884      GLint row;
885      for (row = 0; row < height; row++) {
886         GLfloat *zValuesFloat = (GLfloat*)zValues;
887         GLenum destType = GL_UNSIGNED_BYTE;
888         const GLvoid *source = _mesa_image_address2d(&clippedUnpack, pixels,
889                                                      width, height,
890                                                      format, type,
891                                                      row, 0);
892         _mesa_unpack_stencil_span(ctx, width, destType, sValues,
893                                   type, source, &clippedUnpack,
894                                   ctx->_ImageTransferState);
895
896         if (format == GL_DEPTH_STENCIL) {
897            GLenum ztype =
898               pt->resource->format == PIPE_FORMAT_Z32_FLOAT_S8X24_UINT ?
899               GL_FLOAT : GL_UNSIGNED_INT;
900
901            _mesa_unpack_depth_span(ctx, width, ztype, zValues,
902                                    (1 << 24) - 1, type, source,
903                                    &clippedUnpack);
904         }
905
906         if (zoom) {
907            _mesa_problem(ctx, "Gallium glDrawPixels(GL_STENCIL) with "
908                          "zoom not complete");
909         }
910
911         {
912            GLint spanY;
913
914            if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
915               spanY = height - row - 1;
916            }
917            else {
918               spanY = row;
919            }
920
921            /* now pack the stencil (and Z) values in the dest format */
922            switch (pt->resource->format) {
923            case PIPE_FORMAT_S8_UINT:
924               {
925                  ubyte *dest = stmap + spanY * pt->stride;
926                  assert(usage == PIPE_TRANSFER_WRITE);
927                  memcpy(dest, sValues, width);
928               }
929               break;
930            case PIPE_FORMAT_Z24_UNORM_S8_UINT:
931               if (format == GL_DEPTH_STENCIL) {
932                  uint *dest = (uint *) (stmap + spanY * pt->stride);
933                  GLint k;
934                  assert(usage == PIPE_TRANSFER_WRITE);
935                  for (k = 0; k < width; k++) {
936                     dest[k] = zValues[k] | (sValues[k] << 24);
937                  }
938               }
939               else {
940                  uint *dest = (uint *) (stmap + spanY * pt->stride);
941                  GLint k;
942                  assert(usage == PIPE_TRANSFER_READ_WRITE);
943                  for (k = 0; k < width; k++) {
944                     dest[k] = (dest[k] & 0xffffff) | (sValues[k] << 24);
945                  }
946               }
947               break;
948            case PIPE_FORMAT_S8_UINT_Z24_UNORM:
949               if (format == GL_DEPTH_STENCIL) {
950                  uint *dest = (uint *) (stmap + spanY * pt->stride);
951                  GLint k;
952                  assert(usage == PIPE_TRANSFER_WRITE);
953                  for (k = 0; k < width; k++) {
954                     dest[k] = (zValues[k] << 8) | (sValues[k] & 0xff);
955                  }
956               }
957               else {
958                  uint *dest = (uint *) (stmap + spanY * pt->stride);
959                  GLint k;
960                  assert(usage == PIPE_TRANSFER_READ_WRITE);
961                  for (k = 0; k < width; k++) {
962                     dest[k] = (dest[k] & 0xffffff00) | (sValues[k] & 0xff);
963                  }
964               }
965               break;
966            case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT:
967               if (format == GL_DEPTH_STENCIL) {
968                  uint *dest = (uint *) (stmap + spanY * pt->stride);
969                  GLfloat *destf = (GLfloat*)dest;
970                  GLint k;
971                  assert(usage == PIPE_TRANSFER_WRITE);
972                  for (k = 0; k < width; k++) {
973                     destf[k*2] = zValuesFloat[k];
974                     dest[k*2+1] = sValues[k] & 0xff;
975                  }
976               }
977               else {
978                  uint *dest = (uint *) (stmap + spanY * pt->stride);
979                  GLint k;
980                  assert(usage == PIPE_TRANSFER_READ_WRITE);
981                  for (k = 0; k < width; k++) {
982                     dest[k*2+1] = sValues[k] & 0xff;
983                  }
984               }
985               break;
986            default:
987               assert(0);
988            }
989         }
990      }
991   }
992   else {
993      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glDrawPixels()");
994   }
995
996   free(sValues);
997   free(zValues);
998
999   _mesa_unmap_pbo_source(ctx, &clippedUnpack);
1000
1001   /* unmap the stencil buffer */
1002   pipe_transfer_unmap(pipe, pt);
1003   pipe->transfer_destroy(pipe, pt);
1004}
1005
1006
1007/**
1008 * Get fragment program variant for a glDrawPixels or glCopyPixels
1009 * command for RGBA data.
1010 */
1011static struct st_fp_variant *
1012get_color_fp_variant(struct st_context *st)
1013{
1014   struct gl_context *ctx = st->ctx;
1015   struct st_fp_variant_key key;
1016   struct st_fp_variant *fpv;
1017
1018   memset(&key, 0, sizeof(key));
1019
1020   key.st = st;
1021   key.drawpixels = 1;
1022   key.scaleAndBias = (ctx->Pixel.RedBias != 0.0 ||
1023                       ctx->Pixel.RedScale != 1.0 ||
1024                       ctx->Pixel.GreenBias != 0.0 ||
1025                       ctx->Pixel.GreenScale != 1.0 ||
1026                       ctx->Pixel.BlueBias != 0.0 ||
1027                       ctx->Pixel.BlueScale != 1.0 ||
1028                       ctx->Pixel.AlphaBias != 0.0 ||
1029                       ctx->Pixel.AlphaScale != 1.0);
1030   key.pixelMaps = ctx->Pixel.MapColorFlag;
1031   key.clamp_color = st->clamp_frag_color_in_shader &&
1032                     st->ctx->Color._ClampFragmentColor &&
1033                     !st->ctx->DrawBuffer->_IntegerColor;
1034
1035   fpv = st_get_fp_variant(st, st->fp, &key);
1036
1037   return fpv;
1038}
1039
1040
1041/**
1042 * Get fragment program variant for a glDrawPixels or glCopyPixels
1043 * command for depth/stencil data.
1044 */
1045static struct st_fp_variant *
1046get_depth_stencil_fp_variant(struct st_context *st, GLboolean write_depth,
1047                             GLboolean write_stencil)
1048{
1049   struct st_fp_variant_key key;
1050   struct st_fp_variant *fpv;
1051
1052   memset(&key, 0, sizeof(key));
1053
1054   key.st = st;
1055   key.drawpixels = 1;
1056   key.drawpixels_z = write_depth;
1057   key.drawpixels_stencil = write_stencil;
1058
1059   fpv = st_get_fp_variant(st, st->fp, &key);
1060
1061   return fpv;
1062}
1063
1064
1065/**
1066 * Clamp glDrawPixels width and height to the maximum texture size.
1067 */
1068static void
1069clamp_size(struct pipe_context *pipe, GLsizei *width, GLsizei *height,
1070           struct gl_pixelstore_attrib *unpack)
1071{
1072   const unsigned maxSize =
1073      1 << (pipe->screen->get_param(pipe->screen,
1074                                    PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
1075
1076   if (*width > maxSize) {
1077      if (unpack->RowLength == 0)
1078         unpack->RowLength = *width;
1079      *width = maxSize;
1080   }
1081   if (*height > maxSize) {
1082      *height = maxSize;
1083   }
1084}
1085
1086
1087/**
1088 * Called via ctx->Driver.DrawPixels()
1089 */
1090static void
1091st_DrawPixels(struct gl_context *ctx, GLint x, GLint y,
1092              GLsizei width, GLsizei height,
1093              GLenum format, GLenum type,
1094              const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels)
1095{
1096   void *driver_vp, *driver_fp;
1097   struct st_context *st = st_context(ctx);
1098   const GLfloat *color;
1099   struct pipe_context *pipe = st->pipe;
1100   GLboolean write_stencil = GL_FALSE, write_depth = GL_FALSE;
1101   struct pipe_sampler_view *sv[2];
1102   int num_sampler_view = 1;
1103   struct st_fp_variant *fpv;
1104   struct gl_pixelstore_attrib clippedUnpack;
1105
1106   /* Mesa state should be up to date by now */
1107   assert(ctx->NewState == 0x0);
1108
1109   st_validate_state(st);
1110
1111   /* Limit the size of the glDrawPixels to the max texture size.
1112    * Strictly speaking, that's not correct but since we don't handle
1113    * larger images yet, this is better than crashing.
1114    */
1115   clippedUnpack = *unpack;
1116   unpack = &clippedUnpack;
1117   clamp_size(st->pipe, &width, &height, &clippedUnpack);
1118
1119   if (format == GL_DEPTH_STENCIL)
1120      write_stencil = write_depth = GL_TRUE;
1121   else if (format == GL_STENCIL_INDEX)
1122      write_stencil = GL_TRUE;
1123   else if (format == GL_DEPTH_COMPONENT)
1124      write_depth = GL_TRUE;
1125
1126   if (write_stencil &&
1127       !pipe->screen->get_param(pipe->screen, PIPE_CAP_SHADER_STENCIL_EXPORT)) {
1128      /* software fallback */
1129      draw_stencil_pixels(ctx, x, y, width, height, format, type,
1130                          unpack, pixels);
1131      return;
1132   }
1133
1134   /*
1135    * Get vertex/fragment shaders
1136    */
1137   if (write_depth || write_stencil) {
1138      fpv = get_depth_stencil_fp_variant(st, write_depth, write_stencil);
1139
1140      driver_fp = fpv->driver_shader;
1141
1142      driver_vp = make_passthrough_vertex_shader(st, GL_TRUE);
1143
1144      color = ctx->Current.RasterColor;
1145   }
1146   else {
1147      fpv = get_color_fp_variant(st);
1148
1149      driver_fp = fpv->driver_shader;
1150
1151      driver_vp = make_passthrough_vertex_shader(st, GL_FALSE);
1152
1153      color = NULL;
1154      if (st->pixel_xfer.pixelmap_enabled) {
1155	  sv[1] = st->pixel_xfer.pixelmap_sampler_view;
1156	  num_sampler_view++;
1157      }
1158   }
1159
1160   /* update fragment program constants */
1161   st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
1162
1163   /* draw with textured quad */
1164   {
1165      struct pipe_resource *pt
1166         = make_texture(st, width, height, format, type, unpack, pixels);
1167      if (pt) {
1168         sv[0] = st_create_texture_sampler_view(st->pipe, pt);
1169
1170         if (sv[0]) {
1171            /* Create a second sampler view to read stencil.
1172             * The stencil is written using the shader stencil export
1173             * functionality. */
1174            if (write_stencil) {
1175               enum pipe_format stencil_format =
1176                     util_format_stencil_only(pt->format);
1177
1178               sv[1] = st_create_texture_sampler_view_format(st->pipe, pt,
1179                                                             stencil_format);
1180               num_sampler_view++;
1181            }
1182
1183            draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2],
1184                               width, height,
1185                               ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
1186                               sv,
1187                               num_sampler_view,
1188                               driver_vp,
1189                               driver_fp,
1190                               color, GL_FALSE, write_depth, write_stencil);
1191            pipe_sampler_view_reference(&sv[0], NULL);
1192            if (num_sampler_view > 1)
1193               pipe_sampler_view_reference(&sv[1], NULL);
1194         }
1195         pipe_resource_reference(&pt, NULL);
1196      }
1197   }
1198}
1199
1200
1201
1202/**
1203 * Software fallback for glCopyPixels(GL_STENCIL).
1204 */
1205static void
1206copy_stencil_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
1207                    GLsizei width, GLsizei height,
1208                    GLint dstx, GLint dsty)
1209{
1210   struct st_renderbuffer *rbDraw;
1211   struct pipe_context *pipe = st_context(ctx)->pipe;
1212   enum pipe_transfer_usage usage;
1213   struct pipe_transfer *ptDraw;
1214   ubyte *drawMap;
1215   ubyte *buffer;
1216   int i;
1217
1218   buffer = malloc(width * height * sizeof(ubyte));
1219   if (!buffer) {
1220      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyPixels(stencil)");
1221      return;
1222   }
1223
1224   /* Get the dest renderbuffer */
1225   rbDraw = st_renderbuffer(ctx->DrawBuffer->
1226                            Attachment[BUFFER_STENCIL].Renderbuffer);
1227
1228   /* this will do stencil pixel transfer ops */
1229   _mesa_readpixels(ctx, srcx, srcy, width, height,
1230                    GL_STENCIL_INDEX, GL_UNSIGNED_BYTE,
1231                    &ctx->DefaultPacking, buffer);
1232
1233   if (0) {
1234      /* debug code: dump stencil values */
1235      GLint row, col;
1236      for (row = 0; row < height; row++) {
1237         printf("%3d: ", row);
1238         for (col = 0; col < width; col++) {
1239            printf("%02x ", buffer[col + row * width]);
1240         }
1241         printf("\n");
1242      }
1243   }
1244
1245   if (_mesa_is_format_packed_depth_stencil(rbDraw->Base.Format))
1246      usage = PIPE_TRANSFER_READ_WRITE;
1247   else
1248      usage = PIPE_TRANSFER_WRITE;
1249
1250   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
1251      dsty = rbDraw->Base.Height - dsty - height;
1252   }
1253
1254   ptDraw = pipe_get_transfer(pipe,
1255                              rbDraw->texture,
1256                              rbDraw->rtt_level,
1257                              rbDraw->rtt_face + rbDraw->rtt_slice,
1258                              usage, dstx, dsty,
1259                              width, height);
1260
1261   assert(util_format_get_blockwidth(ptDraw->resource->format) == 1);
1262   assert(util_format_get_blockheight(ptDraw->resource->format) == 1);
1263
1264   /* map the stencil buffer */
1265   drawMap = pipe_transfer_map(pipe, ptDraw);
1266
1267   /* draw */
1268   /* XXX PixelZoom not handled yet */
1269   for (i = 0; i < height; i++) {
1270      ubyte *dst;
1271      const ubyte *src;
1272      int y;
1273
1274      y = i;
1275
1276      if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
1277         y = height - y - 1;
1278      }
1279
1280      dst = drawMap + y * ptDraw->stride;
1281      src = buffer + i * width;
1282
1283      _mesa_pack_ubyte_stencil_row(rbDraw->Base.Format, width, src, dst);
1284   }
1285
1286   free(buffer);
1287
1288   /* unmap the stencil buffer */
1289   pipe_transfer_unmap(pipe, ptDraw);
1290   pipe->transfer_destroy(pipe, ptDraw);
1291}
1292
1293
1294/**
1295 * Return renderbuffer to use for reading color pixels for glCopyPixels
1296 */
1297static struct st_renderbuffer *
1298st_get_color_read_renderbuffer(struct gl_context *ctx)
1299{
1300   struct gl_framebuffer *fb = ctx->ReadBuffer;
1301   struct st_renderbuffer *strb =
1302      st_renderbuffer(fb->_ColorReadBuffer);
1303
1304   return strb;
1305}
1306
1307
1308/** Do the src/dest regions overlap? */
1309static GLboolean
1310regions_overlap(GLint srcX, GLint srcY, GLint dstX, GLint dstY,
1311                GLsizei width, GLsizei height)
1312{
1313   if (srcX + width <= dstX ||
1314       dstX + width <= srcX ||
1315       srcY + height <= dstY ||
1316       dstY + height <= srcY)
1317      return GL_FALSE;
1318   else
1319      return GL_TRUE;
1320}
1321
1322
1323/**
1324 * Try to do a glCopyPixels for simple cases with a blit by calling
1325 * pipe->resource_copy_region().
1326 *
1327 * We can do this when we're copying color pixels (depth/stencil
1328 * eventually) with no pixel zoom, no pixel transfer ops, no
1329 * per-fragment ops, the src/dest regions don't overlap and the
1330 * src/dest pixel formats are the same.
1331 */
1332static GLboolean
1333blit_copy_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
1334                 GLsizei width, GLsizei height,
1335                 GLint dstx, GLint dsty, GLenum type)
1336{
1337   struct st_context *st = st_context(ctx);
1338   struct pipe_context *pipe = st->pipe;
1339   struct gl_pixelstore_attrib pack, unpack;
1340   GLint readX, readY, readW, readH;
1341
1342   if (type == GL_COLOR &&
1343       ctx->Pixel.ZoomX == 1.0 &&
1344       ctx->Pixel.ZoomY == 1.0 &&
1345       ctx->_ImageTransferState == 0x0 &&
1346       !ctx->Color.BlendEnabled &&
1347       !ctx->Color.AlphaEnabled &&
1348       !ctx->Depth.Test &&
1349       !ctx->Fog.Enabled &&
1350       !ctx->Stencil.Enabled &&
1351       !ctx->FragmentProgram.Enabled &&
1352       !ctx->VertexProgram.Enabled &&
1353       !ctx->Shader.CurrentFragmentProgram &&
1354       st_fb_orientation(ctx->ReadBuffer) == st_fb_orientation(ctx->DrawBuffer) &&
1355       ctx->DrawBuffer->_NumColorDrawBuffers == 1 &&
1356       !ctx->Query.CondRenderQuery) {
1357      struct st_renderbuffer *rbRead, *rbDraw;
1358      GLint drawX, drawY;
1359
1360      /*
1361       * Clip the read region against the src buffer bounds.
1362       * We'll still allocate a temporary buffer/texture for the original
1363       * src region size but we'll only read the region which is on-screen.
1364       * This may mean that we draw garbage pixels into the dest region, but
1365       * that's expected.
1366       */
1367      readX = srcx;
1368      readY = srcy;
1369      readW = width;
1370      readH = height;
1371      pack = ctx->DefaultPacking;
1372      if (!_mesa_clip_readpixels(ctx, &readX, &readY, &readW, &readH, &pack))
1373         return GL_TRUE; /* all done */
1374
1375      /* clip against dest buffer bounds and scissor box */
1376      drawX = dstx + pack.SkipPixels;
1377      drawY = dsty + pack.SkipRows;
1378      unpack = pack;
1379      if (!_mesa_clip_drawpixels(ctx, &drawX, &drawY, &readW, &readH, &unpack))
1380         return GL_TRUE; /* all done */
1381
1382      readX = readX - pack.SkipPixels + unpack.SkipPixels;
1383      readY = readY - pack.SkipRows + unpack.SkipRows;
1384
1385      rbRead = st_get_color_read_renderbuffer(ctx);
1386      rbDraw = st_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0]);
1387
1388      if ((rbRead != rbDraw ||
1389           !regions_overlap(readX, readY, drawX, drawY, readW, readH)) &&
1390          rbRead->Base.Format == rbDraw->Base.Format) {
1391         struct pipe_box srcBox;
1392
1393         /* flip src/dst position if needed */
1394         if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
1395            /* both buffers will have the same orientation */
1396            readY = ctx->ReadBuffer->Height - readY - readH;
1397            drawY = ctx->DrawBuffer->Height - drawY - readH;
1398         }
1399
1400         u_box_2d(readX, readY, readW, readH, &srcBox);
1401
1402         pipe->resource_copy_region(pipe,
1403                                    rbDraw->texture,
1404                                    rbDraw->rtt_level, drawX, drawY, 0,
1405                                    rbRead->texture,
1406                                    rbRead->rtt_level, &srcBox);
1407         return GL_TRUE;
1408      }
1409   }
1410
1411   return GL_FALSE;
1412}
1413
1414
1415static void
1416st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
1417              GLsizei width, GLsizei height,
1418              GLint dstx, GLint dsty, GLenum type)
1419{
1420   struct st_context *st = st_context(ctx);
1421   struct pipe_context *pipe = st->pipe;
1422   struct pipe_screen *screen = pipe->screen;
1423   struct st_renderbuffer *rbRead;
1424   void *driver_vp, *driver_fp;
1425   struct pipe_resource *pt;
1426   struct pipe_sampler_view *sv[2];
1427   int num_sampler_view = 1;
1428   GLfloat *color;
1429   enum pipe_format srcFormat, texFormat;
1430   GLboolean invertTex = GL_FALSE;
1431   GLint readX, readY, readW, readH;
1432   GLuint sample_count;
1433   struct gl_pixelstore_attrib pack = ctx->DefaultPacking;
1434   struct st_fp_variant *fpv;
1435
1436   st_validate_state(st);
1437
1438   if (type == GL_DEPTH_STENCIL) {
1439      /* XXX make this more efficient */
1440      st_CopyPixels(ctx, srcx, srcy, width, height, dstx, dsty, GL_STENCIL);
1441      st_CopyPixels(ctx, srcx, srcy, width, height, dstx, dsty, GL_DEPTH);
1442      return;
1443   }
1444
1445   if (type == GL_STENCIL) {
1446      /* can't use texturing to do stencil */
1447      copy_stencil_pixels(ctx, srcx, srcy, width, height, dstx, dsty);
1448      return;
1449   }
1450
1451   if (blit_copy_pixels(ctx, srcx, srcy, width, height, dstx, dsty, type))
1452      return;
1453
1454   /*
1455    * The subsequent code implements glCopyPixels by copying the source
1456    * pixels into a temporary texture that's then applied to a textured quad.
1457    * When we draw the textured quad, all the usual per-fragment operations
1458    * are handled.
1459    */
1460
1461
1462   /*
1463    * Get vertex/fragment shaders
1464    */
1465   if (type == GL_COLOR) {
1466      rbRead = st_get_color_read_renderbuffer(ctx);
1467      color = NULL;
1468
1469      fpv = get_color_fp_variant(st);
1470      driver_fp = fpv->driver_shader;
1471
1472      driver_vp = make_passthrough_vertex_shader(st, GL_FALSE);
1473
1474      if (st->pixel_xfer.pixelmap_enabled) {
1475	  sv[1] = st->pixel_xfer.pixelmap_sampler_view;
1476	  num_sampler_view++;
1477      }
1478   }
1479   else {
1480      assert(type == GL_DEPTH);
1481      rbRead = st_renderbuffer(ctx->ReadBuffer->
1482                               Attachment[BUFFER_DEPTH].Renderbuffer);
1483      color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
1484
1485      fpv = get_depth_stencil_fp_variant(st, GL_TRUE, GL_FALSE);
1486      driver_fp = fpv->driver_shader;
1487
1488      driver_vp = make_passthrough_vertex_shader(st, GL_TRUE);
1489   }
1490
1491   /* update fragment program constants */
1492   st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
1493
1494   sample_count = rbRead->texture->nr_samples;
1495   /* I believe this would be legal, presumably would need to do a resolve
1496      for color, and for depth/stencil spec says to just use one of the
1497      depth/stencil samples per pixel? Need some transfer clarifications. */
1498   assert(sample_count < 2);
1499
1500   srcFormat = rbRead->texture->format;
1501
1502   if (screen->is_format_supported(screen, srcFormat, st->internal_target,
1503                                   sample_count,
1504                                   PIPE_BIND_SAMPLER_VIEW)) {
1505      texFormat = srcFormat;
1506   }
1507   else {
1508      /* srcFormat can't be used as a texture format */
1509      if (type == GL_DEPTH) {
1510         texFormat = st_choose_format(screen, GL_DEPTH_COMPONENT,
1511                                      GL_NONE, GL_NONE, st->internal_target,
1512				      sample_count, PIPE_BIND_DEPTH_STENCIL);
1513         assert(texFormat != PIPE_FORMAT_NONE);
1514      }
1515      else {
1516         /* default color format */
1517         texFormat = st_choose_format(screen, GL_RGBA,
1518                                      GL_NONE, GL_NONE, st->internal_target,
1519                                      sample_count, PIPE_BIND_SAMPLER_VIEW);
1520         assert(texFormat != PIPE_FORMAT_NONE);
1521      }
1522   }
1523
1524   /* Invert src region if needed */
1525   if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
1526      srcy = ctx->ReadBuffer->Height - srcy - height;
1527      invertTex = !invertTex;
1528   }
1529
1530   /* Clip the read region against the src buffer bounds.
1531    * We'll still allocate a temporary buffer/texture for the original
1532    * src region size but we'll only read the region which is on-screen.
1533    * This may mean that we draw garbage pixels into the dest region, but
1534    * that's expected.
1535    */
1536   readX = srcx;
1537   readY = srcy;
1538   readW = width;
1539   readH = height;
1540   if (!_mesa_clip_readpixels(ctx, &readX, &readY, &readW, &readH, &pack)) {
1541      /* The source region is completely out of bounds.  Do nothing.
1542       * The GL spec says "Results of copies from outside the window,
1543       * or from regions of the window that are not exposed, are
1544       * hardware dependent and undefined."
1545       */
1546      return;
1547   }
1548
1549   readW = MAX2(0, readW);
1550   readH = MAX2(0, readH);
1551
1552   /* alloc temporary texture */
1553   pt = alloc_texture(st, width, height, texFormat);
1554   if (!pt)
1555      return;
1556
1557   sv[0] = st_create_texture_sampler_view(st->pipe, pt);
1558   if (!sv[0]) {
1559      pipe_resource_reference(&pt, NULL);
1560      return;
1561   }
1562
1563   /* Make temporary texture which is a copy of the src region.
1564    */
1565   if (srcFormat == texFormat) {
1566      struct pipe_box src_box;
1567      u_box_2d(readX, readY, readW, readH, &src_box);
1568      /* copy source framebuffer surface into mipmap/texture */
1569      pipe->resource_copy_region(pipe,
1570                                 pt,                                /* dest tex */
1571                                 0,                                 /* dest lvl */
1572                                 pack.SkipPixels, pack.SkipRows, 0, /* dest pos */
1573                                 rbRead->texture,                   /* src tex */
1574                                 rbRead->rtt_level,                 /* src lvl */
1575                                 &src_box);
1576
1577   }
1578   else {
1579      /* CPU-based fallback/conversion */
1580      struct pipe_transfer *ptRead =
1581         pipe_get_transfer(st->pipe, rbRead->texture,
1582                           rbRead->rtt_level,
1583                           rbRead->rtt_face + rbRead->rtt_slice,
1584                           PIPE_TRANSFER_READ,
1585                           readX, readY, readW, readH);
1586      struct pipe_transfer *ptTex;
1587      enum pipe_transfer_usage transfer_usage;
1588
1589      if (ST_DEBUG & DEBUG_FALLBACK)
1590         debug_printf("%s: fallback processing\n", __FUNCTION__);
1591
1592      if (type == GL_DEPTH && util_format_is_depth_and_stencil(pt->format))
1593         transfer_usage = PIPE_TRANSFER_READ_WRITE;
1594      else
1595         transfer_usage = PIPE_TRANSFER_WRITE;
1596
1597      ptTex = pipe_get_transfer(st->pipe, pt, 0, 0, transfer_usage,
1598                                0, 0, width, height);
1599
1600      /* copy image from ptRead surface to ptTex surface */
1601      if (type == GL_COLOR) {
1602         /* alternate path using get/put_tile() */
1603         GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
1604         enum pipe_format readFormat, drawFormat;
1605         readFormat = util_format_linear(rbRead->texture->format);
1606         drawFormat = util_format_linear(pt->format);
1607         pipe_get_tile_rgba_format(pipe, ptRead, 0, 0, readW, readH,
1608                                   readFormat, buf);
1609         pipe_put_tile_rgba_format(pipe, ptTex, pack.SkipPixels, pack.SkipRows,
1610                                   readW, readH, drawFormat, buf);
1611         free(buf);
1612      }
1613      else {
1614         /* GL_DEPTH */
1615         GLuint *buf = (GLuint *) malloc(width * height * sizeof(GLuint));
1616         pipe_get_tile_z(pipe, ptRead, 0, 0, readW, readH, buf);
1617         pipe_put_tile_z(pipe, ptTex, pack.SkipPixels, pack.SkipRows,
1618                         readW, readH, buf);
1619         free(buf);
1620      }
1621
1622      pipe->transfer_destroy(pipe, ptRead);
1623      pipe->transfer_destroy(pipe, ptTex);
1624   }
1625
1626   /* OK, the texture 'pt' contains the src image/pixels.  Now draw a
1627    * textured quad with that texture.
1628    */
1629   draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2],
1630                      width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
1631                      sv,
1632                      num_sampler_view,
1633                      driver_vp,
1634                      driver_fp,
1635                      color, invertTex, GL_FALSE, GL_FALSE);
1636
1637   pipe_resource_reference(&pt, NULL);
1638   pipe_sampler_view_reference(&sv[0], NULL);
1639}
1640
1641
1642
1643void st_init_drawpixels_functions(struct dd_function_table *functions)
1644{
1645   functions->DrawPixels = st_DrawPixels;
1646   functions->CopyPixels = st_CopyPixels;
1647}
1648
1649
1650void
1651st_destroy_drawpix(struct st_context *st)
1652{
1653   GLuint i;
1654
1655   for (i = 0; i < Elements(st->drawpix.shaders); i++) {
1656      if (st->drawpix.shaders[i])
1657         _mesa_reference_fragprog(st->ctx, &st->drawpix.shaders[i], NULL);
1658   }
1659
1660   st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL);
1661   if (st->drawpix.vert_shaders[0])
1662      cso_delete_vertex_shader(st->cso_context, st->drawpix.vert_shaders[0]);
1663   if (st->drawpix.vert_shaders[1])
1664      cso_delete_vertex_shader(st->cso_context, st->drawpix.vert_shaders[1]);
1665}
1666
1667#endif /* FEATURE_drawpix */
1668