st_cb_drawpixels.c revision c760677c85a1019eef67b2ce0ce97e8218918f44
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   if (u_upload_alloc(st->uploader, 0, 4 * sizeof(verts[0]), &offset,
577                      &buf, (void **) &verts) != PIPE_OK) {
578      return;
579   }
580
581   /* setup vertex data */
582   {
583      const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
584      const GLfloat fb_width = (GLfloat) fb->Width;
585      const GLfloat fb_height = (GLfloat) fb->Height;
586      const GLfloat clip_x0 = x0 / fb_width * 2.0f - 1.0f;
587      const GLfloat clip_y0 = y0 / fb_height * 2.0f - 1.0f;
588      const GLfloat clip_x1 = x1 / fb_width * 2.0f - 1.0f;
589      const GLfloat clip_y1 = y1 / fb_height * 2.0f - 1.0f;
590      const GLfloat sLeft = 0.0f, sRight = maxXcoord;
591      const GLfloat tTop = invertTex ? maxYcoord : 0.0f;
592      const GLfloat tBot = invertTex ? 0.0f : maxYcoord;
593      GLuint i;
594
595      /* upper-left */
596      verts[0][0][0] = clip_x0;    /* v[0].attr[0].x */
597      verts[0][0][1] = clip_y0;    /* v[0].attr[0].y */
598
599      /* upper-right */
600      verts[1][0][0] = clip_x1;
601      verts[1][0][1] = clip_y0;
602
603      /* lower-right */
604      verts[2][0][0] = clip_x1;
605      verts[2][0][1] = clip_y1;
606
607      /* lower-left */
608      verts[3][0][0] = clip_x0;
609      verts[3][0][1] = clip_y1;
610
611      verts[0][1][0] = sLeft; /* v[0].attr[1].S */
612      verts[0][1][1] = tTop;  /* v[0].attr[1].T */
613      verts[1][1][0] = sRight;
614      verts[1][1][1] = tTop;
615      verts[2][1][0] = sRight;
616      verts[2][1][1] = tBot;
617      verts[3][1][0] = sLeft;
618      verts[3][1][1] = tBot;
619
620      /* same for all verts: */
621      if (color) {
622         for (i = 0; i < 4; i++) {
623            verts[i][0][2] = z;         /* v[i].attr[0].z */
624            verts[i][0][3] = 1.0f;      /* v[i].attr[0].w */
625            verts[i][2][0] = color[0];  /* v[i].attr[2].r */
626            verts[i][2][1] = color[1];  /* v[i].attr[2].g */
627            verts[i][2][2] = color[2];  /* v[i].attr[2].b */
628            verts[i][2][3] = color[3];  /* v[i].attr[2].a */
629            verts[i][1][2] = 0.0f;      /* v[i].attr[1].R */
630            verts[i][1][3] = 1.0f;      /* v[i].attr[1].Q */
631         }
632      }
633      else {
634         for (i = 0; i < 4; i++) {
635            verts[i][0][2] = z;    /*Z*/
636            verts[i][0][3] = 1.0f; /*W*/
637            verts[i][1][2] = 0.0f; /*R*/
638            verts[i][1][3] = 1.0f; /*Q*/
639         }
640      }
641   }
642
643   u_upload_unmap(st->uploader);
644   util_draw_vertex_buffer(pipe, st->cso_context, buf, offset,
645			   PIPE_PRIM_QUADS,
646			   4,  /* verts */
647			   3); /* attribs/vert */
648   pipe_resource_reference(&buf, NULL);
649}
650
651
652
653static void
654draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
655                   GLsizei width, GLsizei height,
656                   GLfloat zoomX, GLfloat zoomY,
657                   struct pipe_sampler_view **sv,
658                   int num_sampler_view,
659                   void *driver_vp,
660                   void *driver_fp,
661                   const GLfloat *color,
662                   GLboolean invertTex,
663                   GLboolean write_depth, GLboolean write_stencil)
664{
665   struct st_context *st = st_context(ctx);
666   struct pipe_context *pipe = st->pipe;
667   struct cso_context *cso = st->cso_context;
668   GLfloat x0, y0, x1, y1;
669   GLsizei maxSize;
670   boolean normalized = sv[0]->texture->target != PIPE_TEXTURE_RECT;
671
672   /* limit checks */
673   /* XXX if DrawPixels image is larger than max texture size, break
674    * it up into chunks.
675    */
676   maxSize = 1 << (pipe->screen->get_param(pipe->screen,
677                                        PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
678   assert(width <= maxSize);
679   assert(height <= maxSize);
680
681   cso_save_rasterizer(cso);
682   cso_save_viewport(cso);
683   cso_save_samplers(cso, PIPE_SHADER_FRAGMENT);
684   cso_save_sampler_views(cso, PIPE_SHADER_FRAGMENT);
685   cso_save_fragment_shader(cso);
686   cso_save_stream_outputs(cso);
687   cso_save_vertex_shader(cso);
688   cso_save_geometry_shader(cso);
689   cso_save_vertex_elements(cso);
690   cso_save_vertex_buffers(cso);
691   if (write_stencil) {
692      cso_save_depth_stencil_alpha(cso);
693      cso_save_blend(cso);
694   }
695
696   /* rasterizer state: just scissor */
697   {
698      struct pipe_rasterizer_state rasterizer;
699      memset(&rasterizer, 0, sizeof(rasterizer));
700      rasterizer.clamp_fragment_color = !st->clamp_frag_color_in_shader &&
701                                        ctx->Color._ClampFragmentColor &&
702                                        !ctx->DrawBuffer->_IntegerColor;
703      rasterizer.gl_rasterization_rules = 1;
704      rasterizer.depth_clip = !ctx->Transform.DepthClamp;
705      rasterizer.scissor = ctx->Scissor.Enabled;
706      cso_set_rasterizer(cso, &rasterizer);
707   }
708
709   if (write_stencil) {
710      /* Stencil writing bypasses the normal fragment pipeline to
711       * disable color writing and set stencil test to always pass.
712       */
713      struct pipe_depth_stencil_alpha_state dsa;
714      struct pipe_blend_state blend;
715
716      /* depth/stencil */
717      memset(&dsa, 0, sizeof(dsa));
718      dsa.stencil[0].enabled = 1;
719      dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
720      dsa.stencil[0].writemask = ctx->Stencil.WriteMask[0] & 0xff;
721      dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
722      if (write_depth) {
723         /* writing depth+stencil: depth test always passes */
724         dsa.depth.enabled = 1;
725         dsa.depth.writemask = ctx->Depth.Mask;
726         dsa.depth.func = PIPE_FUNC_ALWAYS;
727      }
728      cso_set_depth_stencil_alpha(cso, &dsa);
729
730      /* blend (colormask) */
731      memset(&blend, 0, sizeof(blend));
732      cso_set_blend(cso, &blend);
733   }
734
735   /* fragment shader state: TEX lookup program */
736   cso_set_fragment_shader_handle(cso, driver_fp);
737
738   /* vertex shader state: position + texcoord pass-through */
739   cso_set_vertex_shader_handle(cso, driver_vp);
740
741   /* geometry shader state: disabled */
742   cso_set_geometry_shader_handle(cso, NULL);
743
744   /* texture sampling state: */
745   {
746      struct pipe_sampler_state sampler;
747      memset(&sampler, 0, sizeof(sampler));
748      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP;
749      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP;
750      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP;
751      sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
752      sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
753      sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
754      sampler.normalized_coords = normalized;
755
756      cso_single_sampler(cso, PIPE_SHADER_FRAGMENT, 0, &sampler);
757      if (num_sampler_view > 1) {
758         cso_single_sampler(cso, PIPE_SHADER_FRAGMENT, 1, &sampler);
759      }
760      cso_single_sampler_done(cso, PIPE_SHADER_FRAGMENT);
761   }
762
763   /* viewport state: viewport matching window dims */
764   {
765      const float w = (float) ctx->DrawBuffer->Width;
766      const float h = (float) ctx->DrawBuffer->Height;
767      struct pipe_viewport_state vp;
768      vp.scale[0] =  0.5f * w;
769      vp.scale[1] = -0.5f * h;
770      vp.scale[2] = 0.5f;
771      vp.scale[3] = 1.0f;
772      vp.translate[0] = 0.5f * w;
773      vp.translate[1] = 0.5f * h;
774      vp.translate[2] = 0.5f;
775      vp.translate[3] = 0.0f;
776      cso_set_viewport(cso, &vp);
777   }
778
779   cso_set_vertex_elements(cso, 3, st->velems_util_draw);
780   cso_set_stream_outputs(st->cso_context, 0, NULL, 0);
781
782   /* texture state: */
783   cso_set_sampler_views(cso, PIPE_SHADER_FRAGMENT, num_sampler_view, sv);
784
785   /* Compute Gallium window coords (y=0=top) with pixel zoom.
786    * Recall that these coords are transformed by the current
787    * vertex shader and viewport transformation.
788    */
789   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_BOTTOM) {
790      y = ctx->DrawBuffer->Height - (int) (y + height * ctx->Pixel.ZoomY);
791      invertTex = !invertTex;
792   }
793
794   x0 = (GLfloat) x;
795   x1 = x + width * ctx->Pixel.ZoomX;
796   y0 = (GLfloat) y;
797   y1 = y + height * ctx->Pixel.ZoomY;
798
799   /* convert Z from [0,1] to [-1,-1] to match viewport Z scale/bias */
800   z = z * 2.0 - 1.0;
801
802   draw_quad(ctx, x0, y0, z, x1, y1, color, invertTex,
803             normalized ? ((GLfloat) width / sv[0]->texture->width0) : (GLfloat)width,
804             normalized ? ((GLfloat) height / sv[0]->texture->height0) : (GLfloat)height);
805
806   /* restore state */
807   cso_restore_rasterizer(cso);
808   cso_restore_viewport(cso);
809   cso_restore_samplers(cso, PIPE_SHADER_FRAGMENT);
810   cso_restore_sampler_views(cso, PIPE_SHADER_FRAGMENT);
811   cso_restore_fragment_shader(cso);
812   cso_restore_vertex_shader(cso);
813   cso_restore_geometry_shader(cso);
814   cso_restore_vertex_elements(cso);
815   cso_restore_vertex_buffers(cso);
816   cso_restore_stream_outputs(cso);
817   if (write_stencil) {
818      cso_restore_depth_stencil_alpha(cso);
819      cso_restore_blend(cso);
820   }
821}
822
823
824/**
825 * Software fallback to do glDrawPixels(GL_STENCIL_INDEX) when we
826 * can't use a fragment shader to write stencil values.
827 */
828static void
829draw_stencil_pixels(struct gl_context *ctx, GLint x, GLint y,
830                    GLsizei width, GLsizei height, GLenum format, GLenum type,
831                    const struct gl_pixelstore_attrib *unpack,
832                    const GLvoid *pixels)
833{
834   struct st_context *st = st_context(ctx);
835   struct pipe_context *pipe = st->pipe;
836   struct st_renderbuffer *strb;
837   enum pipe_transfer_usage usage;
838   struct pipe_transfer *pt;
839   const GLboolean zoom = ctx->Pixel.ZoomX != 1.0 || ctx->Pixel.ZoomY != 1.0;
840   ubyte *stmap;
841   struct gl_pixelstore_attrib clippedUnpack = *unpack;
842   GLubyte *sValues;
843   GLuint *zValues;
844
845   if (!zoom) {
846      if (!_mesa_clip_drawpixels(ctx, &x, &y, &width, &height,
847                                 &clippedUnpack)) {
848         /* totally clipped */
849         return;
850      }
851   }
852
853   strb = st_renderbuffer(ctx->DrawBuffer->
854                          Attachment[BUFFER_STENCIL].Renderbuffer);
855
856   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
857      y = ctx->DrawBuffer->Height - y - height;
858   }
859
860   if (format == GL_STENCIL_INDEX &&
861       _mesa_is_format_packed_depth_stencil(strb->Base.Format)) {
862      /* writing stencil to a combined depth+stencil buffer */
863      usage = PIPE_TRANSFER_READ_WRITE;
864   }
865   else {
866      usage = PIPE_TRANSFER_WRITE;
867   }
868
869   pt = pipe_get_transfer(pipe, strb->texture,
870                          strb->rtt_level, strb->rtt_face + strb->rtt_slice,
871                          usage, x, y,
872                          width, height);
873
874   stmap = pipe_transfer_map(pipe, pt);
875
876   pixels = _mesa_map_pbo_source(ctx, &clippedUnpack, pixels);
877   assert(pixels);
878
879   sValues = (GLubyte *) malloc(width * sizeof(GLubyte));
880   zValues = (GLuint *) malloc(width * sizeof(GLuint));
881
882   if (sValues && zValues) {
883      GLint row;
884      for (row = 0; row < height; row++) {
885         GLfloat *zValuesFloat = (GLfloat*)zValues;
886         GLenum destType = GL_UNSIGNED_BYTE;
887         const GLvoid *source = _mesa_image_address2d(&clippedUnpack, pixels,
888                                                      width, height,
889                                                      format, type,
890                                                      row, 0);
891         _mesa_unpack_stencil_span(ctx, width, destType, sValues,
892                                   type, source, &clippedUnpack,
893                                   ctx->_ImageTransferState);
894
895         if (format == GL_DEPTH_STENCIL) {
896            GLenum ztype =
897               pt->resource->format == PIPE_FORMAT_Z32_FLOAT_S8X24_UINT ?
898               GL_FLOAT : GL_UNSIGNED_INT;
899
900            _mesa_unpack_depth_span(ctx, width, ztype, zValues,
901                                    (1 << 24) - 1, type, source,
902                                    &clippedUnpack);
903         }
904
905         if (zoom) {
906            _mesa_problem(ctx, "Gallium glDrawPixels(GL_STENCIL) with "
907                          "zoom not complete");
908         }
909
910         {
911            GLint spanY;
912
913            if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
914               spanY = height - row - 1;
915            }
916            else {
917               spanY = row;
918            }
919
920            /* now pack the stencil (and Z) values in the dest format */
921            switch (pt->resource->format) {
922            case PIPE_FORMAT_S8_UINT:
923               {
924                  ubyte *dest = stmap + spanY * pt->stride;
925                  assert(usage == PIPE_TRANSFER_WRITE);
926                  memcpy(dest, sValues, width);
927               }
928               break;
929            case PIPE_FORMAT_Z24_UNORM_S8_UINT:
930               if (format == GL_DEPTH_STENCIL) {
931                  uint *dest = (uint *) (stmap + spanY * pt->stride);
932                  GLint k;
933                  assert(usage == PIPE_TRANSFER_WRITE);
934                  for (k = 0; k < width; k++) {
935                     dest[k] = zValues[k] | (sValues[k] << 24);
936                  }
937               }
938               else {
939                  uint *dest = (uint *) (stmap + spanY * pt->stride);
940                  GLint k;
941                  assert(usage == PIPE_TRANSFER_READ_WRITE);
942                  for (k = 0; k < width; k++) {
943                     dest[k] = (dest[k] & 0xffffff) | (sValues[k] << 24);
944                  }
945               }
946               break;
947            case PIPE_FORMAT_S8_UINT_Z24_UNORM:
948               if (format == GL_DEPTH_STENCIL) {
949                  uint *dest = (uint *) (stmap + spanY * pt->stride);
950                  GLint k;
951                  assert(usage == PIPE_TRANSFER_WRITE);
952                  for (k = 0; k < width; k++) {
953                     dest[k] = (zValues[k] << 8) | (sValues[k] & 0xff);
954                  }
955               }
956               else {
957                  uint *dest = (uint *) (stmap + spanY * pt->stride);
958                  GLint k;
959                  assert(usage == PIPE_TRANSFER_READ_WRITE);
960                  for (k = 0; k < width; k++) {
961                     dest[k] = (dest[k] & 0xffffff00) | (sValues[k] & 0xff);
962                  }
963               }
964               break;
965            case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT:
966               if (format == GL_DEPTH_STENCIL) {
967                  uint *dest = (uint *) (stmap + spanY * pt->stride);
968                  GLfloat *destf = (GLfloat*)dest;
969                  GLint k;
970                  assert(usage == PIPE_TRANSFER_WRITE);
971                  for (k = 0; k < width; k++) {
972                     destf[k*2] = zValuesFloat[k];
973                     dest[k*2+1] = sValues[k] & 0xff;
974                  }
975               }
976               else {
977                  uint *dest = (uint *) (stmap + spanY * pt->stride);
978                  GLint k;
979                  assert(usage == PIPE_TRANSFER_READ_WRITE);
980                  for (k = 0; k < width; k++) {
981                     dest[k*2+1] = sValues[k] & 0xff;
982                  }
983               }
984               break;
985            default:
986               assert(0);
987            }
988         }
989      }
990   }
991   else {
992      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glDrawPixels()");
993   }
994
995   free(sValues);
996   free(zValues);
997
998   _mesa_unmap_pbo_source(ctx, &clippedUnpack);
999
1000   /* unmap the stencil buffer */
1001   pipe_transfer_unmap(pipe, pt);
1002   pipe->transfer_destroy(pipe, pt);
1003}
1004
1005
1006/**
1007 * Get fragment program variant for a glDrawPixels or glCopyPixels
1008 * command for RGBA data.
1009 */
1010static struct st_fp_variant *
1011get_color_fp_variant(struct st_context *st)
1012{
1013   struct gl_context *ctx = st->ctx;
1014   struct st_fp_variant_key key;
1015   struct st_fp_variant *fpv;
1016
1017   memset(&key, 0, sizeof(key));
1018
1019   key.st = st;
1020   key.drawpixels = 1;
1021   key.scaleAndBias = (ctx->Pixel.RedBias != 0.0 ||
1022                       ctx->Pixel.RedScale != 1.0 ||
1023                       ctx->Pixel.GreenBias != 0.0 ||
1024                       ctx->Pixel.GreenScale != 1.0 ||
1025                       ctx->Pixel.BlueBias != 0.0 ||
1026                       ctx->Pixel.BlueScale != 1.0 ||
1027                       ctx->Pixel.AlphaBias != 0.0 ||
1028                       ctx->Pixel.AlphaScale != 1.0);
1029   key.pixelMaps = ctx->Pixel.MapColorFlag;
1030   key.clamp_color = st->clamp_frag_color_in_shader &&
1031                     st->ctx->Color._ClampFragmentColor &&
1032                     !st->ctx->DrawBuffer->_IntegerColor;
1033
1034   fpv = st_get_fp_variant(st, st->fp, &key);
1035
1036   return fpv;
1037}
1038
1039
1040/**
1041 * Get fragment program variant for a glDrawPixels or glCopyPixels
1042 * command for depth/stencil data.
1043 */
1044static struct st_fp_variant *
1045get_depth_stencil_fp_variant(struct st_context *st, GLboolean write_depth,
1046                             GLboolean write_stencil)
1047{
1048   struct st_fp_variant_key key;
1049   struct st_fp_variant *fpv;
1050
1051   memset(&key, 0, sizeof(key));
1052
1053   key.st = st;
1054   key.drawpixels = 1;
1055   key.drawpixels_z = write_depth;
1056   key.drawpixels_stencil = write_stencil;
1057
1058   fpv = st_get_fp_variant(st, st->fp, &key);
1059
1060   return fpv;
1061}
1062
1063
1064/**
1065 * Clamp glDrawPixels width and height to the maximum texture size.
1066 */
1067static void
1068clamp_size(struct pipe_context *pipe, GLsizei *width, GLsizei *height,
1069           struct gl_pixelstore_attrib *unpack)
1070{
1071   const unsigned maxSize =
1072      1 << (pipe->screen->get_param(pipe->screen,
1073                                    PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
1074
1075   if (*width > maxSize) {
1076      if (unpack->RowLength == 0)
1077         unpack->RowLength = *width;
1078      *width = maxSize;
1079   }
1080   if (*height > maxSize) {
1081      *height = maxSize;
1082   }
1083}
1084
1085
1086/**
1087 * Called via ctx->Driver.DrawPixels()
1088 */
1089static void
1090st_DrawPixels(struct gl_context *ctx, GLint x, GLint y,
1091              GLsizei width, GLsizei height,
1092              GLenum format, GLenum type,
1093              const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels)
1094{
1095   void *driver_vp, *driver_fp;
1096   struct st_context *st = st_context(ctx);
1097   const GLfloat *color;
1098   struct pipe_context *pipe = st->pipe;
1099   GLboolean write_stencil = GL_FALSE, write_depth = GL_FALSE;
1100   struct pipe_sampler_view *sv[2];
1101   int num_sampler_view = 1;
1102   struct st_fp_variant *fpv;
1103   struct gl_pixelstore_attrib clippedUnpack;
1104
1105   /* Mesa state should be up to date by now */
1106   assert(ctx->NewState == 0x0);
1107
1108   st_validate_state(st);
1109
1110   /* Limit the size of the glDrawPixels to the max texture size.
1111    * Strictly speaking, that's not correct but since we don't handle
1112    * larger images yet, this is better than crashing.
1113    */
1114   clippedUnpack = *unpack;
1115   unpack = &clippedUnpack;
1116   clamp_size(st->pipe, &width, &height, &clippedUnpack);
1117
1118   if (format == GL_DEPTH_STENCIL)
1119      write_stencil = write_depth = GL_TRUE;
1120   else if (format == GL_STENCIL_INDEX)
1121      write_stencil = GL_TRUE;
1122   else if (format == GL_DEPTH_COMPONENT)
1123      write_depth = GL_TRUE;
1124
1125   if (write_stencil &&
1126       !pipe->screen->get_param(pipe->screen, PIPE_CAP_SHADER_STENCIL_EXPORT)) {
1127      /* software fallback */
1128      draw_stencil_pixels(ctx, x, y, width, height, format, type,
1129                          unpack, pixels);
1130      return;
1131   }
1132
1133   /*
1134    * Get vertex/fragment shaders
1135    */
1136   if (write_depth || write_stencil) {
1137      fpv = get_depth_stencil_fp_variant(st, write_depth, write_stencil);
1138
1139      driver_fp = fpv->driver_shader;
1140
1141      driver_vp = make_passthrough_vertex_shader(st, GL_TRUE);
1142
1143      color = ctx->Current.RasterColor;
1144   }
1145   else {
1146      fpv = get_color_fp_variant(st);
1147
1148      driver_fp = fpv->driver_shader;
1149
1150      driver_vp = make_passthrough_vertex_shader(st, GL_FALSE);
1151
1152      color = NULL;
1153      if (st->pixel_xfer.pixelmap_enabled) {
1154	  sv[1] = st->pixel_xfer.pixelmap_sampler_view;
1155	  num_sampler_view++;
1156      }
1157   }
1158
1159   /* update fragment program constants */
1160   st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
1161
1162   /* draw with textured quad */
1163   {
1164      struct pipe_resource *pt
1165         = make_texture(st, width, height, format, type, unpack, pixels);
1166      if (pt) {
1167         sv[0] = st_create_texture_sampler_view(st->pipe, pt);
1168
1169         if (sv[0]) {
1170            /* Create a second sampler view to read stencil.
1171             * The stencil is written using the shader stencil export
1172             * functionality. */
1173            if (write_stencil) {
1174               enum pipe_format stencil_format =
1175                     util_format_stencil_only(pt->format);
1176
1177               sv[1] = st_create_texture_sampler_view_format(st->pipe, pt,
1178                                                             stencil_format);
1179               num_sampler_view++;
1180            }
1181
1182            draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2],
1183                               width, height,
1184                               ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
1185                               sv,
1186                               num_sampler_view,
1187                               driver_vp,
1188                               driver_fp,
1189                               color, GL_FALSE, write_depth, write_stencil);
1190            pipe_sampler_view_reference(&sv[0], NULL);
1191            if (num_sampler_view > 1)
1192               pipe_sampler_view_reference(&sv[1], NULL);
1193         }
1194         pipe_resource_reference(&pt, NULL);
1195      }
1196   }
1197}
1198
1199
1200
1201/**
1202 * Software fallback for glCopyPixels(GL_STENCIL).
1203 */
1204static void
1205copy_stencil_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
1206                    GLsizei width, GLsizei height,
1207                    GLint dstx, GLint dsty)
1208{
1209   struct st_renderbuffer *rbDraw;
1210   struct pipe_context *pipe = st_context(ctx)->pipe;
1211   enum pipe_transfer_usage usage;
1212   struct pipe_transfer *ptDraw;
1213   ubyte *drawMap;
1214   ubyte *buffer;
1215   int i;
1216
1217   buffer = malloc(width * height * sizeof(ubyte));
1218   if (!buffer) {
1219      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyPixels(stencil)");
1220      return;
1221   }
1222
1223   /* Get the dest renderbuffer */
1224   rbDraw = st_renderbuffer(ctx->DrawBuffer->
1225                            Attachment[BUFFER_STENCIL].Renderbuffer);
1226
1227   /* this will do stencil pixel transfer ops */
1228   _mesa_readpixels(ctx, srcx, srcy, width, height,
1229                    GL_STENCIL_INDEX, GL_UNSIGNED_BYTE,
1230                    &ctx->DefaultPacking, buffer);
1231
1232   if (0) {
1233      /* debug code: dump stencil values */
1234      GLint row, col;
1235      for (row = 0; row < height; row++) {
1236         printf("%3d: ", row);
1237         for (col = 0; col < width; col++) {
1238            printf("%02x ", buffer[col + row * width]);
1239         }
1240         printf("\n");
1241      }
1242   }
1243
1244   if (_mesa_is_format_packed_depth_stencil(rbDraw->Base.Format))
1245      usage = PIPE_TRANSFER_READ_WRITE;
1246   else
1247      usage = PIPE_TRANSFER_WRITE;
1248
1249   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
1250      dsty = rbDraw->Base.Height - dsty - height;
1251   }
1252
1253   ptDraw = pipe_get_transfer(pipe,
1254                              rbDraw->texture,
1255                              rbDraw->rtt_level,
1256                              rbDraw->rtt_face + rbDraw->rtt_slice,
1257                              usage, dstx, dsty,
1258                              width, height);
1259
1260   assert(util_format_get_blockwidth(ptDraw->resource->format) == 1);
1261   assert(util_format_get_blockheight(ptDraw->resource->format) == 1);
1262
1263   /* map the stencil buffer */
1264   drawMap = pipe_transfer_map(pipe, ptDraw);
1265
1266   /* draw */
1267   /* XXX PixelZoom not handled yet */
1268   for (i = 0; i < height; i++) {
1269      ubyte *dst;
1270      const ubyte *src;
1271      int y;
1272
1273      y = i;
1274
1275      if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
1276         y = height - y - 1;
1277      }
1278
1279      dst = drawMap + y * ptDraw->stride;
1280      src = buffer + i * width;
1281
1282      _mesa_pack_ubyte_stencil_row(rbDraw->Base.Format, width, src, dst);
1283   }
1284
1285   free(buffer);
1286
1287   /* unmap the stencil buffer */
1288   pipe_transfer_unmap(pipe, ptDraw);
1289   pipe->transfer_destroy(pipe, ptDraw);
1290}
1291
1292
1293/**
1294 * Return renderbuffer to use for reading color pixels for glCopyPixels
1295 */
1296static struct st_renderbuffer *
1297st_get_color_read_renderbuffer(struct gl_context *ctx)
1298{
1299   struct gl_framebuffer *fb = ctx->ReadBuffer;
1300   struct st_renderbuffer *strb =
1301      st_renderbuffer(fb->_ColorReadBuffer);
1302
1303   return strb;
1304}
1305
1306
1307/** Do the src/dest regions overlap? */
1308static GLboolean
1309regions_overlap(GLint srcX, GLint srcY, GLint dstX, GLint dstY,
1310                GLsizei width, GLsizei height)
1311{
1312   if (srcX + width <= dstX ||
1313       dstX + width <= srcX ||
1314       srcY + height <= dstY ||
1315       dstY + height <= srcY)
1316      return GL_FALSE;
1317   else
1318      return GL_TRUE;
1319}
1320
1321
1322/**
1323 * Try to do a glCopyPixels for simple cases with a blit by calling
1324 * pipe->resource_copy_region().
1325 *
1326 * We can do this when we're copying color pixels (depth/stencil
1327 * eventually) with no pixel zoom, no pixel transfer ops, no
1328 * per-fragment ops, the src/dest regions don't overlap and the
1329 * src/dest pixel formats are the same.
1330 */
1331static GLboolean
1332blit_copy_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
1333                 GLsizei width, GLsizei height,
1334                 GLint dstx, GLint dsty, GLenum type)
1335{
1336   struct st_context *st = st_context(ctx);
1337   struct pipe_context *pipe = st->pipe;
1338   struct gl_pixelstore_attrib pack, unpack;
1339   GLint readX, readY, readW, readH;
1340
1341   if (type == GL_COLOR &&
1342       ctx->Pixel.ZoomX == 1.0 &&
1343       ctx->Pixel.ZoomY == 1.0 &&
1344       ctx->_ImageTransferState == 0x0 &&
1345       !ctx->Color.BlendEnabled &&
1346       !ctx->Color.AlphaEnabled &&
1347       !ctx->Depth.Test &&
1348       !ctx->Fog.Enabled &&
1349       !ctx->Stencil.Enabled &&
1350       !ctx->FragmentProgram.Enabled &&
1351       !ctx->VertexProgram.Enabled &&
1352       !ctx->Shader.CurrentFragmentProgram &&
1353       st_fb_orientation(ctx->ReadBuffer) == st_fb_orientation(ctx->DrawBuffer) &&
1354       ctx->DrawBuffer->_NumColorDrawBuffers == 1 &&
1355       !ctx->Query.CondRenderQuery) {
1356      struct st_renderbuffer *rbRead, *rbDraw;
1357      GLint drawX, drawY;
1358
1359      /*
1360       * Clip the read region against the src buffer bounds.
1361       * We'll still allocate a temporary buffer/texture for the original
1362       * src region size but we'll only read the region which is on-screen.
1363       * This may mean that we draw garbage pixels into the dest region, but
1364       * that's expected.
1365       */
1366      readX = srcx;
1367      readY = srcy;
1368      readW = width;
1369      readH = height;
1370      pack = ctx->DefaultPacking;
1371      if (!_mesa_clip_readpixels(ctx, &readX, &readY, &readW, &readH, &pack))
1372         return GL_TRUE; /* all done */
1373
1374      /* clip against dest buffer bounds and scissor box */
1375      drawX = dstx + pack.SkipPixels;
1376      drawY = dsty + pack.SkipRows;
1377      unpack = pack;
1378      if (!_mesa_clip_drawpixels(ctx, &drawX, &drawY, &readW, &readH, &unpack))
1379         return GL_TRUE; /* all done */
1380
1381      readX = readX - pack.SkipPixels + unpack.SkipPixels;
1382      readY = readY - pack.SkipRows + unpack.SkipRows;
1383
1384      rbRead = st_get_color_read_renderbuffer(ctx);
1385      rbDraw = st_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0]);
1386
1387      if ((rbRead != rbDraw ||
1388           !regions_overlap(readX, readY, drawX, drawY, readW, readH)) &&
1389          rbRead->Base.Format == rbDraw->Base.Format) {
1390         struct pipe_box srcBox;
1391
1392         /* flip src/dst position if needed */
1393         if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
1394            /* both buffers will have the same orientation */
1395            readY = ctx->ReadBuffer->Height - readY - readH;
1396            drawY = ctx->DrawBuffer->Height - drawY - readH;
1397         }
1398
1399         u_box_2d(readX, readY, readW, readH, &srcBox);
1400
1401         pipe->resource_copy_region(pipe,
1402                                    rbDraw->texture,
1403                                    rbDraw->rtt_level, drawX, drawY, 0,
1404                                    rbRead->texture,
1405                                    rbRead->rtt_level, &srcBox);
1406         return GL_TRUE;
1407      }
1408   }
1409
1410   return GL_FALSE;
1411}
1412
1413
1414static void
1415st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
1416              GLsizei width, GLsizei height,
1417              GLint dstx, GLint dsty, GLenum type)
1418{
1419   struct st_context *st = st_context(ctx);
1420   struct pipe_context *pipe = st->pipe;
1421   struct pipe_screen *screen = pipe->screen;
1422   struct st_renderbuffer *rbRead;
1423   void *driver_vp, *driver_fp;
1424   struct pipe_resource *pt;
1425   struct pipe_sampler_view *sv[2];
1426   int num_sampler_view = 1;
1427   GLfloat *color;
1428   enum pipe_format srcFormat, texFormat;
1429   GLboolean invertTex = GL_FALSE;
1430   GLint readX, readY, readW, readH;
1431   GLuint sample_count;
1432   struct gl_pixelstore_attrib pack = ctx->DefaultPacking;
1433   struct st_fp_variant *fpv;
1434
1435   st_validate_state(st);
1436
1437   if (type == GL_DEPTH_STENCIL) {
1438      /* XXX make this more efficient */
1439      st_CopyPixels(ctx, srcx, srcy, width, height, dstx, dsty, GL_STENCIL);
1440      st_CopyPixels(ctx, srcx, srcy, width, height, dstx, dsty, GL_DEPTH);
1441      return;
1442   }
1443
1444   if (type == GL_STENCIL) {
1445      /* can't use texturing to do stencil */
1446      copy_stencil_pixels(ctx, srcx, srcy, width, height, dstx, dsty);
1447      return;
1448   }
1449
1450   if (blit_copy_pixels(ctx, srcx, srcy, width, height, dstx, dsty, type))
1451      return;
1452
1453   /*
1454    * The subsequent code implements glCopyPixels by copying the source
1455    * pixels into a temporary texture that's then applied to a textured quad.
1456    * When we draw the textured quad, all the usual per-fragment operations
1457    * are handled.
1458    */
1459
1460
1461   /*
1462    * Get vertex/fragment shaders
1463    */
1464   if (type == GL_COLOR) {
1465      rbRead = st_get_color_read_renderbuffer(ctx);
1466      color = NULL;
1467
1468      fpv = get_color_fp_variant(st);
1469      driver_fp = fpv->driver_shader;
1470
1471      driver_vp = make_passthrough_vertex_shader(st, GL_FALSE);
1472
1473      if (st->pixel_xfer.pixelmap_enabled) {
1474	  sv[1] = st->pixel_xfer.pixelmap_sampler_view;
1475	  num_sampler_view++;
1476      }
1477   }
1478   else {
1479      assert(type == GL_DEPTH);
1480      rbRead = st_renderbuffer(ctx->ReadBuffer->
1481                               Attachment[BUFFER_DEPTH].Renderbuffer);
1482      color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
1483
1484      fpv = get_depth_stencil_fp_variant(st, GL_TRUE, GL_FALSE);
1485      driver_fp = fpv->driver_shader;
1486
1487      driver_vp = make_passthrough_vertex_shader(st, GL_TRUE);
1488   }
1489
1490   /* update fragment program constants */
1491   st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
1492
1493   sample_count = rbRead->texture->nr_samples;
1494   /* I believe this would be legal, presumably would need to do a resolve
1495      for color, and for depth/stencil spec says to just use one of the
1496      depth/stencil samples per pixel? Need some transfer clarifications. */
1497   assert(sample_count < 2);
1498
1499   srcFormat = rbRead->texture->format;
1500
1501   if (screen->is_format_supported(screen, srcFormat, st->internal_target,
1502                                   sample_count,
1503                                   PIPE_BIND_SAMPLER_VIEW)) {
1504      texFormat = srcFormat;
1505   }
1506   else {
1507      /* srcFormat can't be used as a texture format */
1508      if (type == GL_DEPTH) {
1509         texFormat = st_choose_format(screen, GL_DEPTH_COMPONENT,
1510                                      GL_NONE, GL_NONE, st->internal_target,
1511				      sample_count, PIPE_BIND_DEPTH_STENCIL);
1512         assert(texFormat != PIPE_FORMAT_NONE);
1513      }
1514      else {
1515         /* default color format */
1516         texFormat = st_choose_format(screen, GL_RGBA,
1517                                      GL_NONE, GL_NONE, st->internal_target,
1518                                      sample_count, PIPE_BIND_SAMPLER_VIEW);
1519         assert(texFormat != PIPE_FORMAT_NONE);
1520      }
1521   }
1522
1523   /* Invert src region if needed */
1524   if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
1525      srcy = ctx->ReadBuffer->Height - srcy - height;
1526      invertTex = !invertTex;
1527   }
1528
1529   /* Clip the read region against the src buffer bounds.
1530    * We'll still allocate a temporary buffer/texture for the original
1531    * src region size but we'll only read the region which is on-screen.
1532    * This may mean that we draw garbage pixels into the dest region, but
1533    * that's expected.
1534    */
1535   readX = srcx;
1536   readY = srcy;
1537   readW = width;
1538   readH = height;
1539   if (!_mesa_clip_readpixels(ctx, &readX, &readY, &readW, &readH, &pack)) {
1540      /* The source region is completely out of bounds.  Do nothing.
1541       * The GL spec says "Results of copies from outside the window,
1542       * or from regions of the window that are not exposed, are
1543       * hardware dependent and undefined."
1544       */
1545      return;
1546   }
1547
1548   readW = MAX2(0, readW);
1549   readH = MAX2(0, readH);
1550
1551   /* alloc temporary texture */
1552   pt = alloc_texture(st, width, height, texFormat);
1553   if (!pt)
1554      return;
1555
1556   sv[0] = st_create_texture_sampler_view(st->pipe, pt);
1557   if (!sv[0]) {
1558      pipe_resource_reference(&pt, NULL);
1559      return;
1560   }
1561
1562   /* Make temporary texture which is a copy of the src region.
1563    */
1564   if (srcFormat == texFormat) {
1565      struct pipe_box src_box;
1566      u_box_2d(readX, readY, readW, readH, &src_box);
1567      /* copy source framebuffer surface into mipmap/texture */
1568      pipe->resource_copy_region(pipe,
1569                                 pt,                                /* dest tex */
1570                                 0,                                 /* dest lvl */
1571                                 pack.SkipPixels, pack.SkipRows, 0, /* dest pos */
1572                                 rbRead->texture,                   /* src tex */
1573                                 rbRead->rtt_level,                 /* src lvl */
1574                                 &src_box);
1575
1576   }
1577   else {
1578      /* CPU-based fallback/conversion */
1579      struct pipe_transfer *ptRead =
1580         pipe_get_transfer(st->pipe, rbRead->texture,
1581                           rbRead->rtt_level,
1582                           rbRead->rtt_face + rbRead->rtt_slice,
1583                           PIPE_TRANSFER_READ,
1584                           readX, readY, readW, readH);
1585      struct pipe_transfer *ptTex;
1586      enum pipe_transfer_usage transfer_usage;
1587
1588      if (ST_DEBUG & DEBUG_FALLBACK)
1589         debug_printf("%s: fallback processing\n", __FUNCTION__);
1590
1591      if (type == GL_DEPTH && util_format_is_depth_and_stencil(pt->format))
1592         transfer_usage = PIPE_TRANSFER_READ_WRITE;
1593      else
1594         transfer_usage = PIPE_TRANSFER_WRITE;
1595
1596      ptTex = pipe_get_transfer(st->pipe, pt, 0, 0, transfer_usage,
1597                                0, 0, width, height);
1598
1599      /* copy image from ptRead surface to ptTex surface */
1600      if (type == GL_COLOR) {
1601         /* alternate path using get/put_tile() */
1602         GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
1603         enum pipe_format readFormat, drawFormat;
1604         readFormat = util_format_linear(rbRead->texture->format);
1605         drawFormat = util_format_linear(pt->format);
1606         pipe_get_tile_rgba_format(pipe, ptRead, 0, 0, readW, readH,
1607                                   readFormat, buf);
1608         pipe_put_tile_rgba_format(pipe, ptTex, pack.SkipPixels, pack.SkipRows,
1609                                   readW, readH, drawFormat, buf);
1610         free(buf);
1611      }
1612      else {
1613         /* GL_DEPTH */
1614         GLuint *buf = (GLuint *) malloc(width * height * sizeof(GLuint));
1615         pipe_get_tile_z(pipe, ptRead, 0, 0, readW, readH, buf);
1616         pipe_put_tile_z(pipe, ptTex, pack.SkipPixels, pack.SkipRows,
1617                         readW, readH, buf);
1618         free(buf);
1619      }
1620
1621      pipe->transfer_destroy(pipe, ptRead);
1622      pipe->transfer_destroy(pipe, ptTex);
1623   }
1624
1625   /* OK, the texture 'pt' contains the src image/pixels.  Now draw a
1626    * textured quad with that texture.
1627    */
1628   draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2],
1629                      width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
1630                      sv,
1631                      num_sampler_view,
1632                      driver_vp,
1633                      driver_fp,
1634                      color, invertTex, GL_FALSE, GL_FALSE);
1635
1636   pipe_resource_reference(&pt, NULL);
1637   pipe_sampler_view_reference(&sv[0], NULL);
1638}
1639
1640
1641
1642void st_init_drawpixels_functions(struct dd_function_table *functions)
1643{
1644   functions->DrawPixels = st_DrawPixels;
1645   functions->CopyPixels = st_CopyPixels;
1646}
1647
1648
1649void
1650st_destroy_drawpix(struct st_context *st)
1651{
1652   GLuint i;
1653
1654   for (i = 0; i < Elements(st->drawpix.shaders); i++) {
1655      if (st->drawpix.shaders[i])
1656         _mesa_reference_fragprog(st->ctx, &st->drawpix.shaders[i], NULL);
1657   }
1658
1659   st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL);
1660   if (st->drawpix.vert_shaders[0])
1661      cso_delete_vertex_shader(st->cso_context, st->drawpix.vert_shaders[0]);
1662   if (st->drawpix.vert_shaders[1])
1663      cso_delete_vertex_shader(st->cso_context, st->drawpix.vert_shaders[1]);
1664}
1665
1666#endif /* FEATURE_drawpix */
1667