st_cb_drawpixels.c revision 03b78ceb50b97611bcaa2d2354ff5b505306b0a1
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         case GL_UNSIGNED_SHORT_4_4_4_4:
402         case GL_UNSIGNED_SHORT_4_4_4_4_REV:
403            return GL_RGBA4;
404
405         case GL_UNSIGNED_SHORT_5_6_5:
406         case GL_UNSIGNED_SHORT_5_6_5_REV:
407            return GL_RGB565;
408
409         case GL_UNSIGNED_SHORT_5_5_5_1:
410         case GL_UNSIGNED_SHORT_1_5_5_5_REV:
411            return GL_RGB5_A1;
412
413         case GL_UNSIGNED_INT_10_10_10_2:
414         case GL_UNSIGNED_INT_2_10_10_10_REV:
415            return GL_RGB10_A2;
416
417         case GL_UNSIGNED_SHORT:
418         case GL_UNSIGNED_INT:
419            return GL_RGBA16;
420
421         case GL_BYTE:
422            return
423               ctx->Extensions.EXT_texture_snorm ? GL_RGBA8_SNORM : GL_RGBA8;
424
425         case GL_SHORT:
426         case GL_INT:
427            return
428               ctx->Extensions.EXT_texture_snorm ? GL_RGBA16_SNORM : GL_RGBA16;
429
430         case GL_HALF_FLOAT_ARB:
431            return
432               ctx->Extensions.ARB_texture_float ? GL_RGBA16F :
433               ctx->Extensions.EXT_texture_snorm ? GL_RGBA16_SNORM : GL_RGBA16;
434
435         case GL_FLOAT:
436         case GL_DOUBLE:
437            return
438               ctx->Extensions.ARB_texture_float ? GL_RGBA32F :
439               ctx->Extensions.EXT_texture_snorm ? GL_RGBA16_SNORM : GL_RGBA16;
440
441         case GL_UNSIGNED_INT_5_9_9_9_REV:
442            assert(ctx->Extensions.EXT_texture_shared_exponent);
443            return GL_RGB9_E5;
444
445         case GL_UNSIGNED_INT_10F_11F_11F_REV:
446            assert(ctx->Extensions.EXT_packed_float);
447            return GL_R11F_G11F_B10F;
448         }
449      }
450   }
451}
452
453
454/**
455 * Create a temporary texture to hold an image of the given size.
456 * If width, height are not POT and the driver only handles POT textures,
457 * allocate the next larger size of texture that is POT.
458 */
459static struct pipe_resource *
460alloc_texture(struct st_context *st, GLsizei width, GLsizei height,
461              enum pipe_format texFormat)
462{
463   struct pipe_resource *pt;
464
465   pt = st_texture_create(st, st->internal_target, texFormat, 0,
466                          width, height, 1, 1, PIPE_BIND_SAMPLER_VIEW);
467
468   return pt;
469}
470
471
472/**
473 * Make texture containing an image for glDrawPixels image.
474 * If 'pixels' is NULL, leave the texture image data undefined.
475 */
476static struct pipe_resource *
477make_texture(struct st_context *st,
478	     GLsizei width, GLsizei height, GLenum format, GLenum type,
479	     const struct gl_pixelstore_attrib *unpack,
480	     const GLvoid *pixels)
481{
482   struct gl_context *ctx = st->ctx;
483   struct pipe_context *pipe = st->pipe;
484   gl_format mformat;
485   struct pipe_resource *pt;
486   enum pipe_format pipeFormat;
487   GLenum baseInternalFormat, intFormat;
488
489   intFormat = internal_format(ctx, format, type);
490   baseInternalFormat = _mesa_base_tex_format(ctx, intFormat);
491
492   mformat = st_ChooseTextureFormat_renderable(ctx, intFormat,
493                                               format, type, GL_FALSE);
494   assert(mformat);
495
496   pipeFormat = st_mesa_format_to_pipe_format(mformat);
497   assert(pipeFormat);
498
499   pixels = _mesa_map_pbo_source(ctx, unpack, pixels);
500   if (!pixels)
501      return NULL;
502
503   /* alloc temporary texture */
504   pt = alloc_texture(st, width, height, pipeFormat);
505   if (!pt) {
506      _mesa_unmap_pbo_source(ctx, unpack);
507      return NULL;
508   }
509
510   {
511      struct pipe_transfer *transfer;
512      GLboolean success;
513      GLubyte *dest;
514      const GLbitfield imageTransferStateSave = ctx->_ImageTransferState;
515
516      /* we'll do pixel transfer in a fragment shader */
517      ctx->_ImageTransferState = 0x0;
518
519      transfer = pipe_get_transfer(st->pipe, pt, 0, 0,
520                                   PIPE_TRANSFER_WRITE, 0, 0,
521                                   width, height);
522
523      /* map texture transfer */
524      dest = pipe_transfer_map(pipe, transfer);
525
526
527      /* Put image into texture transfer.
528       * Note that the image is actually going to be upside down in
529       * the texture.  We deal with that with texcoords.
530       */
531      success = _mesa_texstore(ctx, 2,           /* dims */
532                               baseInternalFormat, /* baseInternalFormat */
533                               mformat,          /* gl_format */
534                               transfer->stride, /* dstRowStride, bytes */
535                               &dest,            /* destSlices */
536                               width, height, 1, /* size */
537                               format, type,     /* src format/type */
538                               pixels,           /* data source */
539                               unpack);
540
541      /* unmap */
542      pipe_transfer_unmap(pipe, transfer);
543      pipe->transfer_destroy(pipe, transfer);
544
545      assert(success);
546
547      /* restore */
548      ctx->_ImageTransferState = imageTransferStateSave;
549   }
550
551   _mesa_unmap_pbo_source(ctx, unpack);
552
553   return pt;
554}
555
556
557/**
558 * Draw quad with texcoords and optional color.
559 * Coords are gallium window coords with y=0=top.
560 * \param color  may be null
561 * \param invertTex  if true, flip texcoords vertically
562 */
563static void
564draw_quad(struct gl_context *ctx, GLfloat x0, GLfloat y0, GLfloat z,
565          GLfloat x1, GLfloat y1, const GLfloat *color,
566          GLboolean invertTex, GLfloat maxXcoord, GLfloat maxYcoord)
567{
568   struct st_context *st = st_context(ctx);
569   struct pipe_context *pipe = st->pipe;
570   GLfloat (*verts)[3][4]; /* four verts, three attribs, XYZW */
571   struct pipe_resource *buf = NULL;
572   unsigned offset;
573
574   u_upload_alloc(st->uploader, 0, 4 * sizeof(verts[0]), &offset, &buf,
575		  (void**)&verts);
576   if (!buf) {
577      return;
578   }
579
580   /* setup vertex data */
581   {
582      const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
583      const GLfloat fb_width = (GLfloat) fb->Width;
584      const GLfloat fb_height = (GLfloat) fb->Height;
585      const GLfloat clip_x0 = x0 / fb_width * 2.0f - 1.0f;
586      const GLfloat clip_y0 = y0 / fb_height * 2.0f - 1.0f;
587      const GLfloat clip_x1 = x1 / fb_width * 2.0f - 1.0f;
588      const GLfloat clip_y1 = y1 / fb_height * 2.0f - 1.0f;
589      const GLfloat sLeft = 0.0f, sRight = maxXcoord;
590      const GLfloat tTop = invertTex ? maxYcoord : 0.0f;
591      const GLfloat tBot = invertTex ? 0.0f : maxYcoord;
592      GLuint i;
593
594      /* upper-left */
595      verts[0][0][0] = clip_x0;    /* v[0].attr[0].x */
596      verts[0][0][1] = clip_y0;    /* v[0].attr[0].y */
597
598      /* upper-right */
599      verts[1][0][0] = clip_x1;
600      verts[1][0][1] = clip_y0;
601
602      /* lower-right */
603      verts[2][0][0] = clip_x1;
604      verts[2][0][1] = clip_y1;
605
606      /* lower-left */
607      verts[3][0][0] = clip_x0;
608      verts[3][0][1] = clip_y1;
609
610      verts[0][1][0] = sLeft; /* v[0].attr[1].S */
611      verts[0][1][1] = tTop;  /* v[0].attr[1].T */
612      verts[1][1][0] = sRight;
613      verts[1][1][1] = tTop;
614      verts[2][1][0] = sRight;
615      verts[2][1][1] = tBot;
616      verts[3][1][0] = sLeft;
617      verts[3][1][1] = tBot;
618
619      /* same for all verts: */
620      if (color) {
621         for (i = 0; i < 4; i++) {
622            verts[i][0][2] = z;         /* v[i].attr[0].z */
623            verts[i][0][3] = 1.0f;      /* v[i].attr[0].w */
624            verts[i][2][0] = color[0];  /* v[i].attr[2].r */
625            verts[i][2][1] = color[1];  /* v[i].attr[2].g */
626            verts[i][2][2] = color[2];  /* v[i].attr[2].b */
627            verts[i][2][3] = color[3];  /* v[i].attr[2].a */
628            verts[i][1][2] = 0.0f;      /* v[i].attr[1].R */
629            verts[i][1][3] = 1.0f;      /* v[i].attr[1].Q */
630         }
631      }
632      else {
633         for (i = 0; i < 4; i++) {
634            verts[i][0][2] = z;    /*Z*/
635            verts[i][0][3] = 1.0f; /*W*/
636            verts[i][1][2] = 0.0f; /*R*/
637            verts[i][1][3] = 1.0f; /*Q*/
638         }
639      }
640   }
641
642   u_upload_unmap(st->uploader);
643   util_draw_vertex_buffer(pipe, st->cso_context, buf, offset,
644			   PIPE_PRIM_QUADS,
645			   4,  /* verts */
646			   3); /* attribs/vert */
647   pipe_resource_reference(&buf, NULL);
648}
649
650
651
652static void
653draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
654                   GLsizei width, GLsizei height,
655                   GLfloat zoomX, GLfloat zoomY,
656                   struct pipe_sampler_view **sv,
657                   int num_sampler_view,
658                   void *driver_vp,
659                   void *driver_fp,
660                   const GLfloat *color,
661                   GLboolean invertTex,
662                   GLboolean write_depth, GLboolean write_stencil)
663{
664   struct st_context *st = st_context(ctx);
665   struct pipe_context *pipe = st->pipe;
666   struct cso_context *cso = st->cso_context;
667   GLfloat x0, y0, x1, y1;
668   GLsizei maxSize;
669   boolean normalized = sv[0]->texture->target != PIPE_TEXTURE_RECT;
670
671   /* limit checks */
672   /* XXX if DrawPixels image is larger than max texture size, break
673    * it up into chunks.
674    */
675   maxSize = 1 << (pipe->screen->get_param(pipe->screen,
676                                        PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
677   assert(width <= maxSize);
678   assert(height <= maxSize);
679
680   cso_save_rasterizer(cso);
681   cso_save_viewport(cso);
682   cso_save_samplers(cso, PIPE_SHADER_FRAGMENT);
683   cso_save_sampler_views(cso, PIPE_SHADER_FRAGMENT);
684   cso_save_fragment_shader(cso);
685   cso_save_stream_outputs(cso);
686   cso_save_vertex_shader(cso);
687   cso_save_geometry_shader(cso);
688   cso_save_vertex_elements(cso);
689   cso_save_vertex_buffers(cso);
690   if (write_stencil) {
691      cso_save_depth_stencil_alpha(cso);
692      cso_save_blend(cso);
693   }
694
695   /* rasterizer state: just scissor */
696   {
697      struct pipe_rasterizer_state rasterizer;
698      memset(&rasterizer, 0, sizeof(rasterizer));
699      rasterizer.clamp_fragment_color = !st->clamp_frag_color_in_shader &&
700                                        ctx->Color._ClampFragmentColor &&
701                                        !ctx->DrawBuffer->_IntegerColor;
702      rasterizer.gl_rasterization_rules = 1;
703      rasterizer.depth_clip = !ctx->Transform.DepthClamp;
704      rasterizer.scissor = ctx->Scissor.Enabled;
705      cso_set_rasterizer(cso, &rasterizer);
706   }
707
708   if (write_stencil) {
709      /* Stencil writing bypasses the normal fragment pipeline to
710       * disable color writing and set stencil test to always pass.
711       */
712      struct pipe_depth_stencil_alpha_state dsa;
713      struct pipe_blend_state blend;
714
715      /* depth/stencil */
716      memset(&dsa, 0, sizeof(dsa));
717      dsa.stencil[0].enabled = 1;
718      dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
719      dsa.stencil[0].writemask = ctx->Stencil.WriteMask[0] & 0xff;
720      dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
721      if (write_depth) {
722         /* writing depth+stencil: depth test always passes */
723         dsa.depth.enabled = 1;
724         dsa.depth.writemask = ctx->Depth.Mask;
725         dsa.depth.func = PIPE_FUNC_ALWAYS;
726      }
727      cso_set_depth_stencil_alpha(cso, &dsa);
728
729      /* blend (colormask) */
730      memset(&blend, 0, sizeof(blend));
731      cso_set_blend(cso, &blend);
732   }
733
734   /* fragment shader state: TEX lookup program */
735   cso_set_fragment_shader_handle(cso, driver_fp);
736
737   /* vertex shader state: position + texcoord pass-through */
738   cso_set_vertex_shader_handle(cso, driver_vp);
739
740   /* geometry shader state: disabled */
741   cso_set_geometry_shader_handle(cso, NULL);
742
743   /* texture sampling state: */
744   {
745      struct pipe_sampler_state sampler;
746      memset(&sampler, 0, sizeof(sampler));
747      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP;
748      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP;
749      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP;
750      sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
751      sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
752      sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
753      sampler.normalized_coords = normalized;
754
755      cso_single_sampler(cso, PIPE_SHADER_FRAGMENT, 0, &sampler);
756      if (num_sampler_view > 1) {
757         cso_single_sampler(cso, PIPE_SHADER_FRAGMENT, 1, &sampler);
758      }
759      cso_single_sampler_done(cso, PIPE_SHADER_FRAGMENT);
760   }
761
762   /* viewport state: viewport matching window dims */
763   {
764      const float w = (float) ctx->DrawBuffer->Width;
765      const float h = (float) ctx->DrawBuffer->Height;
766      struct pipe_viewport_state vp;
767      vp.scale[0] =  0.5f * w;
768      vp.scale[1] = -0.5f * h;
769      vp.scale[2] = 0.5f;
770      vp.scale[3] = 1.0f;
771      vp.translate[0] = 0.5f * w;
772      vp.translate[1] = 0.5f * h;
773      vp.translate[2] = 0.5f;
774      vp.translate[3] = 0.0f;
775      cso_set_viewport(cso, &vp);
776   }
777
778   cso_set_vertex_elements(cso, 3, st->velems_util_draw);
779   cso_set_stream_outputs(st->cso_context, 0, NULL, 0);
780
781   /* texture state: */
782   cso_set_sampler_views(cso, PIPE_SHADER_FRAGMENT, num_sampler_view, sv);
783
784   /* Compute Gallium window coords (y=0=top) with pixel zoom.
785    * Recall that these coords are transformed by the current
786    * vertex shader and viewport transformation.
787    */
788   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_BOTTOM) {
789      y = ctx->DrawBuffer->Height - (int) (y + height * ctx->Pixel.ZoomY);
790      invertTex = !invertTex;
791   }
792
793   x0 = (GLfloat) x;
794   x1 = x + width * ctx->Pixel.ZoomX;
795   y0 = (GLfloat) y;
796   y1 = y + height * ctx->Pixel.ZoomY;
797
798   /* convert Z from [0,1] to [-1,-1] to match viewport Z scale/bias */
799   z = z * 2.0 - 1.0;
800
801   draw_quad(ctx, x0, y0, z, x1, y1, color, invertTex,
802             normalized ? ((GLfloat) width / sv[0]->texture->width0) : (GLfloat)width,
803             normalized ? ((GLfloat) height / sv[0]->texture->height0) : (GLfloat)height);
804
805   /* restore state */
806   cso_restore_rasterizer(cso);
807   cso_restore_viewport(cso);
808   cso_restore_samplers(cso, PIPE_SHADER_FRAGMENT);
809   cso_restore_sampler_views(cso, PIPE_SHADER_FRAGMENT);
810   cso_restore_fragment_shader(cso);
811   cso_restore_vertex_shader(cso);
812   cso_restore_geometry_shader(cso);
813   cso_restore_vertex_elements(cso);
814   cso_restore_vertex_buffers(cso);
815   cso_restore_stream_outputs(cso);
816   if (write_stencil) {
817      cso_restore_depth_stencil_alpha(cso);
818      cso_restore_blend(cso);
819   }
820}
821
822
823/**
824 * Software fallback to do glDrawPixels(GL_STENCIL_INDEX) when we
825 * can't use a fragment shader to write stencil values.
826 */
827static void
828draw_stencil_pixels(struct gl_context *ctx, GLint x, GLint y,
829                    GLsizei width, GLsizei height, GLenum format, GLenum type,
830                    const struct gl_pixelstore_attrib *unpack,
831                    const GLvoid *pixels)
832{
833   struct st_context *st = st_context(ctx);
834   struct pipe_context *pipe = st->pipe;
835   struct st_renderbuffer *strb;
836   enum pipe_transfer_usage usage;
837   struct pipe_transfer *pt;
838   const GLboolean zoom = ctx->Pixel.ZoomX != 1.0 || ctx->Pixel.ZoomY != 1.0;
839   ubyte *stmap;
840   struct gl_pixelstore_attrib clippedUnpack = *unpack;
841   GLubyte *sValues;
842   GLuint *zValues;
843
844   if (!zoom) {
845      if (!_mesa_clip_drawpixels(ctx, &x, &y, &width, &height,
846                                 &clippedUnpack)) {
847         /* totally clipped */
848         return;
849      }
850   }
851
852   strb = st_renderbuffer(ctx->DrawBuffer->
853                          Attachment[BUFFER_STENCIL].Renderbuffer);
854
855   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
856      y = ctx->DrawBuffer->Height - y - height;
857   }
858
859   if (format == GL_STENCIL_INDEX &&
860       _mesa_is_format_packed_depth_stencil(strb->Base.Format)) {
861      /* writing stencil to a combined depth+stencil buffer */
862      usage = PIPE_TRANSFER_READ_WRITE;
863   }
864   else {
865      usage = PIPE_TRANSFER_WRITE;
866   }
867
868   pt = pipe_get_transfer(pipe, strb->texture,
869                          strb->rtt_level, strb->rtt_face + strb->rtt_slice,
870                          usage, x, y,
871                          width, height);
872
873   stmap = pipe_transfer_map(pipe, pt);
874
875   pixels = _mesa_map_pbo_source(ctx, &clippedUnpack, pixels);
876   assert(pixels);
877
878   sValues = (GLubyte *) malloc(width * sizeof(GLubyte));
879   zValues = (GLuint *) malloc(width * sizeof(GLuint));
880
881   if (sValues && zValues) {
882      GLint row;
883      for (row = 0; row < height; row++) {
884         GLfloat *zValuesFloat = (GLfloat*)zValues;
885         GLenum destType = GL_UNSIGNED_BYTE;
886         const GLvoid *source = _mesa_image_address2d(&clippedUnpack, pixels,
887                                                      width, height,
888                                                      format, type,
889                                                      row, 0);
890         _mesa_unpack_stencil_span(ctx, width, destType, sValues,
891                                   type, source, &clippedUnpack,
892                                   ctx->_ImageTransferState);
893
894         if (format == GL_DEPTH_STENCIL) {
895            GLenum ztype =
896               pt->resource->format == PIPE_FORMAT_Z32_FLOAT_S8X24_UINT ?
897               GL_FLOAT : GL_UNSIGNED_INT;
898
899            _mesa_unpack_depth_span(ctx, width, ztype, zValues,
900                                    (1 << 24) - 1, type, source,
901                                    &clippedUnpack);
902         }
903
904         if (zoom) {
905            _mesa_problem(ctx, "Gallium glDrawPixels(GL_STENCIL) with "
906                          "zoom not complete");
907         }
908
909         {
910            GLint spanY;
911
912            if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
913               spanY = height - row - 1;
914            }
915            else {
916               spanY = row;
917            }
918
919            /* now pack the stencil (and Z) values in the dest format */
920            switch (pt->resource->format) {
921            case PIPE_FORMAT_S8_UINT:
922               {
923                  ubyte *dest = stmap + spanY * pt->stride;
924                  assert(usage == PIPE_TRANSFER_WRITE);
925                  memcpy(dest, sValues, width);
926               }
927               break;
928            case PIPE_FORMAT_Z24_UNORM_S8_UINT:
929               if (format == GL_DEPTH_STENCIL) {
930                  uint *dest = (uint *) (stmap + spanY * pt->stride);
931                  GLint k;
932                  assert(usage == PIPE_TRANSFER_WRITE);
933                  for (k = 0; k < width; k++) {
934                     dest[k] = zValues[k] | (sValues[k] << 24);
935                  }
936               }
937               else {
938                  uint *dest = (uint *) (stmap + spanY * pt->stride);
939                  GLint k;
940                  assert(usage == PIPE_TRANSFER_READ_WRITE);
941                  for (k = 0; k < width; k++) {
942                     dest[k] = (dest[k] & 0xffffff) | (sValues[k] << 24);
943                  }
944               }
945               break;
946            case PIPE_FORMAT_S8_UINT_Z24_UNORM:
947               if (format == GL_DEPTH_STENCIL) {
948                  uint *dest = (uint *) (stmap + spanY * pt->stride);
949                  GLint k;
950                  assert(usage == PIPE_TRANSFER_WRITE);
951                  for (k = 0; k < width; k++) {
952                     dest[k] = (zValues[k] << 8) | (sValues[k] & 0xff);
953                  }
954               }
955               else {
956                  uint *dest = (uint *) (stmap + spanY * pt->stride);
957                  GLint k;
958                  assert(usage == PIPE_TRANSFER_READ_WRITE);
959                  for (k = 0; k < width; k++) {
960                     dest[k] = (dest[k] & 0xffffff00) | (sValues[k] & 0xff);
961                  }
962               }
963               break;
964            case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT:
965               if (format == GL_DEPTH_STENCIL) {
966                  uint *dest = (uint *) (stmap + spanY * pt->stride);
967                  GLfloat *destf = (GLfloat*)dest;
968                  GLint k;
969                  assert(usage == PIPE_TRANSFER_WRITE);
970                  for (k = 0; k < width; k++) {
971                     destf[k*2] = zValuesFloat[k];
972                     dest[k*2+1] = sValues[k] & 0xff;
973                  }
974               }
975               else {
976                  uint *dest = (uint *) (stmap + spanY * pt->stride);
977                  GLint k;
978                  assert(usage == PIPE_TRANSFER_READ_WRITE);
979                  for (k = 0; k < width; k++) {
980                     dest[k*2+1] = sValues[k] & 0xff;
981                  }
982               }
983               break;
984            default:
985               assert(0);
986            }
987         }
988      }
989   }
990   else {
991      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glDrawPixels()");
992   }
993
994   free(sValues);
995   free(zValues);
996
997   _mesa_unmap_pbo_source(ctx, &clippedUnpack);
998
999   /* unmap the stencil buffer */
1000   pipe_transfer_unmap(pipe, pt);
1001   pipe->transfer_destroy(pipe, pt);
1002}
1003
1004
1005/**
1006 * Get fragment program variant for a glDrawPixels or glCopyPixels
1007 * command for RGBA data.
1008 */
1009static struct st_fp_variant *
1010get_color_fp_variant(struct st_context *st)
1011{
1012   struct gl_context *ctx = st->ctx;
1013   struct st_fp_variant_key key;
1014   struct st_fp_variant *fpv;
1015
1016   memset(&key, 0, sizeof(key));
1017
1018   key.st = st;
1019   key.drawpixels = 1;
1020   key.scaleAndBias = (ctx->Pixel.RedBias != 0.0 ||
1021                       ctx->Pixel.RedScale != 1.0 ||
1022                       ctx->Pixel.GreenBias != 0.0 ||
1023                       ctx->Pixel.GreenScale != 1.0 ||
1024                       ctx->Pixel.BlueBias != 0.0 ||
1025                       ctx->Pixel.BlueScale != 1.0 ||
1026                       ctx->Pixel.AlphaBias != 0.0 ||
1027                       ctx->Pixel.AlphaScale != 1.0);
1028   key.pixelMaps = ctx->Pixel.MapColorFlag;
1029   key.clamp_color = st->clamp_frag_color_in_shader &&
1030                     st->ctx->Color._ClampFragmentColor &&
1031                     !st->ctx->DrawBuffer->_IntegerColor;
1032
1033   fpv = st_get_fp_variant(st, st->fp, &key);
1034
1035   return fpv;
1036}
1037
1038
1039/**
1040 * Get fragment program variant for a glDrawPixels or glCopyPixels
1041 * command for depth/stencil data.
1042 */
1043static struct st_fp_variant *
1044get_depth_stencil_fp_variant(struct st_context *st, GLboolean write_depth,
1045                             GLboolean write_stencil)
1046{
1047   struct st_fp_variant_key key;
1048   struct st_fp_variant *fpv;
1049
1050   memset(&key, 0, sizeof(key));
1051
1052   key.st = st;
1053   key.drawpixels = 1;
1054   key.drawpixels_z = write_depth;
1055   key.drawpixels_stencil = write_stencil;
1056
1057   fpv = st_get_fp_variant(st, st->fp, &key);
1058
1059   return fpv;
1060}
1061
1062
1063/**
1064 * Clamp glDrawPixels width and height to the maximum texture size.
1065 */
1066static void
1067clamp_size(struct pipe_context *pipe, GLsizei *width, GLsizei *height,
1068           struct gl_pixelstore_attrib *unpack)
1069{
1070   const unsigned maxSize =
1071      1 << (pipe->screen->get_param(pipe->screen,
1072                                    PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
1073
1074   if (*width > maxSize) {
1075      if (unpack->RowLength == 0)
1076         unpack->RowLength = *width;
1077      *width = maxSize;
1078   }
1079   if (*height > maxSize) {
1080      *height = maxSize;
1081   }
1082}
1083
1084
1085/**
1086 * Called via ctx->Driver.DrawPixels()
1087 */
1088static void
1089st_DrawPixels(struct gl_context *ctx, GLint x, GLint y,
1090              GLsizei width, GLsizei height,
1091              GLenum format, GLenum type,
1092              const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels)
1093{
1094   void *driver_vp, *driver_fp;
1095   struct st_context *st = st_context(ctx);
1096   const GLfloat *color;
1097   struct pipe_context *pipe = st->pipe;
1098   GLboolean write_stencil = GL_FALSE, write_depth = GL_FALSE;
1099   struct pipe_sampler_view *sv[2];
1100   int num_sampler_view = 1;
1101   struct st_fp_variant *fpv;
1102   struct gl_pixelstore_attrib clippedUnpack;
1103
1104   /* Mesa state should be up to date by now */
1105   assert(ctx->NewState == 0x0);
1106
1107   st_validate_state(st);
1108
1109   /* Limit the size of the glDrawPixels to the max texture size.
1110    * Strictly speaking, that's not correct but since we don't handle
1111    * larger images yet, this is better than crashing.
1112    */
1113   clippedUnpack = *unpack;
1114   unpack = &clippedUnpack;
1115   clamp_size(st->pipe, &width, &height, &clippedUnpack);
1116
1117   if (format == GL_DEPTH_STENCIL)
1118      write_stencil = write_depth = GL_TRUE;
1119   else if (format == GL_STENCIL_INDEX)
1120      write_stencil = GL_TRUE;
1121   else if (format == GL_DEPTH_COMPONENT)
1122      write_depth = GL_TRUE;
1123
1124   if (write_stencil &&
1125       !pipe->screen->get_param(pipe->screen, PIPE_CAP_SHADER_STENCIL_EXPORT)) {
1126      /* software fallback */
1127      draw_stencil_pixels(ctx, x, y, width, height, format, type,
1128                          unpack, pixels);
1129      return;
1130   }
1131
1132   /*
1133    * Get vertex/fragment shaders
1134    */
1135   if (write_depth || write_stencil) {
1136      fpv = get_depth_stencil_fp_variant(st, write_depth, write_stencil);
1137
1138      driver_fp = fpv->driver_shader;
1139
1140      driver_vp = make_passthrough_vertex_shader(st, GL_TRUE);
1141
1142      color = ctx->Current.RasterColor;
1143   }
1144   else {
1145      fpv = get_color_fp_variant(st);
1146
1147      driver_fp = fpv->driver_shader;
1148
1149      driver_vp = make_passthrough_vertex_shader(st, GL_FALSE);
1150
1151      color = NULL;
1152      if (st->pixel_xfer.pixelmap_enabled) {
1153	  sv[1] = st->pixel_xfer.pixelmap_sampler_view;
1154	  num_sampler_view++;
1155      }
1156   }
1157
1158   /* update fragment program constants */
1159   st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
1160
1161   /* draw with textured quad */
1162   {
1163      struct pipe_resource *pt
1164         = make_texture(st, width, height, format, type, unpack, pixels);
1165      if (pt) {
1166         sv[0] = st_create_texture_sampler_view(st->pipe, pt);
1167
1168         if (sv[0]) {
1169            /* Create a second sampler view to read stencil.
1170             * The stencil is written using the shader stencil export
1171             * functionality. */
1172            if (write_stencil) {
1173               enum pipe_format stencil_format =
1174                     util_format_stencil_only(pt->format);
1175
1176               sv[1] = st_create_texture_sampler_view_format(st->pipe, pt,
1177                                                             stencil_format);
1178               num_sampler_view++;
1179            }
1180
1181            draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2],
1182                               width, height,
1183                               ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
1184                               sv,
1185                               num_sampler_view,
1186                               driver_vp,
1187                               driver_fp,
1188                               color, GL_FALSE, write_depth, write_stencil);
1189            pipe_sampler_view_reference(&sv[0], NULL);
1190            if (num_sampler_view > 1)
1191               pipe_sampler_view_reference(&sv[1], NULL);
1192         }
1193         pipe_resource_reference(&pt, NULL);
1194      }
1195   }
1196}
1197
1198
1199
1200/**
1201 * Software fallback for glCopyPixels(GL_STENCIL).
1202 */
1203static void
1204copy_stencil_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
1205                    GLsizei width, GLsizei height,
1206                    GLint dstx, GLint dsty)
1207{
1208   struct st_renderbuffer *rbDraw;
1209   struct pipe_context *pipe = st_context(ctx)->pipe;
1210   enum pipe_transfer_usage usage;
1211   struct pipe_transfer *ptDraw;
1212   ubyte *drawMap;
1213   ubyte *buffer;
1214   int i;
1215
1216   buffer = malloc(width * height * sizeof(ubyte));
1217   if (!buffer) {
1218      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyPixels(stencil)");
1219      return;
1220   }
1221
1222   /* Get the dest renderbuffer */
1223   rbDraw = st_renderbuffer(ctx->DrawBuffer->
1224                            Attachment[BUFFER_STENCIL].Renderbuffer);
1225
1226   /* this will do stencil pixel transfer ops */
1227   _mesa_readpixels(ctx, srcx, srcy, width, height,
1228                    GL_STENCIL_INDEX, GL_UNSIGNED_BYTE,
1229                    &ctx->DefaultPacking, buffer);
1230
1231   if (0) {
1232      /* debug code: dump stencil values */
1233      GLint row, col;
1234      for (row = 0; row < height; row++) {
1235         printf("%3d: ", row);
1236         for (col = 0; col < width; col++) {
1237            printf("%02x ", buffer[col + row * width]);
1238         }
1239         printf("\n");
1240      }
1241   }
1242
1243   if (_mesa_is_format_packed_depth_stencil(rbDraw->Base.Format))
1244      usage = PIPE_TRANSFER_READ_WRITE;
1245   else
1246      usage = PIPE_TRANSFER_WRITE;
1247
1248   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
1249      dsty = rbDraw->Base.Height - dsty - height;
1250   }
1251
1252   ptDraw = pipe_get_transfer(pipe,
1253                              rbDraw->texture,
1254                              rbDraw->rtt_level,
1255                              rbDraw->rtt_face + rbDraw->rtt_slice,
1256                              usage, dstx, dsty,
1257                              width, height);
1258
1259   assert(util_format_get_blockwidth(ptDraw->resource->format) == 1);
1260   assert(util_format_get_blockheight(ptDraw->resource->format) == 1);
1261
1262   /* map the stencil buffer */
1263   drawMap = pipe_transfer_map(pipe, ptDraw);
1264
1265   /* draw */
1266   /* XXX PixelZoom not handled yet */
1267   for (i = 0; i < height; i++) {
1268      ubyte *dst;
1269      const ubyte *src;
1270      int y;
1271
1272      y = i;
1273
1274      if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
1275         y = height - y - 1;
1276      }
1277
1278      dst = drawMap + y * ptDraw->stride;
1279      src = buffer + i * width;
1280
1281      _mesa_pack_ubyte_stencil_row(rbDraw->Base.Format, width, src, dst);
1282   }
1283
1284   free(buffer);
1285
1286   /* unmap the stencil buffer */
1287   pipe_transfer_unmap(pipe, ptDraw);
1288   pipe->transfer_destroy(pipe, ptDraw);
1289}
1290
1291
1292/**
1293 * Return renderbuffer to use for reading color pixels for glCopyPixels
1294 */
1295static struct st_renderbuffer *
1296st_get_color_read_renderbuffer(struct gl_context *ctx)
1297{
1298   struct gl_framebuffer *fb = ctx->ReadBuffer;
1299   struct st_renderbuffer *strb =
1300      st_renderbuffer(fb->_ColorReadBuffer);
1301
1302   return strb;
1303}
1304
1305
1306/** Do the src/dest regions overlap? */
1307static GLboolean
1308regions_overlap(GLint srcX, GLint srcY, GLint dstX, GLint dstY,
1309                GLsizei width, GLsizei height)
1310{
1311   if (srcX + width <= dstX ||
1312       dstX + width <= srcX ||
1313       srcY + height <= dstY ||
1314       dstY + height <= srcY)
1315      return GL_FALSE;
1316   else
1317      return GL_TRUE;
1318}
1319
1320
1321/**
1322 * Try to do a glCopyPixels for simple cases with a blit by calling
1323 * pipe->resource_copy_region().
1324 *
1325 * We can do this when we're copying color pixels (depth/stencil
1326 * eventually) with no pixel zoom, no pixel transfer ops, no
1327 * per-fragment ops, the src/dest regions don't overlap and the
1328 * src/dest pixel formats are the same.
1329 */
1330static GLboolean
1331blit_copy_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
1332                 GLsizei width, GLsizei height,
1333                 GLint dstx, GLint dsty, GLenum type)
1334{
1335   struct st_context *st = st_context(ctx);
1336   struct pipe_context *pipe = st->pipe;
1337   struct gl_pixelstore_attrib pack, unpack;
1338   GLint readX, readY, readW, readH;
1339
1340   if (type == GL_COLOR &&
1341       ctx->Pixel.ZoomX == 1.0 &&
1342       ctx->Pixel.ZoomY == 1.0 &&
1343       ctx->_ImageTransferState == 0x0 &&
1344       !ctx->Color.BlendEnabled &&
1345       !ctx->Color.AlphaEnabled &&
1346       !ctx->Depth.Test &&
1347       !ctx->Fog.Enabled &&
1348       !ctx->Stencil.Enabled &&
1349       !ctx->FragmentProgram.Enabled &&
1350       !ctx->VertexProgram.Enabled &&
1351       !ctx->Shader.CurrentFragmentProgram &&
1352       st_fb_orientation(ctx->ReadBuffer) == st_fb_orientation(ctx->DrawBuffer) &&
1353       ctx->DrawBuffer->_NumColorDrawBuffers == 1 &&
1354       !ctx->Query.CondRenderQuery) {
1355      struct st_renderbuffer *rbRead, *rbDraw;
1356      GLint drawX, drawY;
1357
1358      /*
1359       * Clip the read region against the src buffer bounds.
1360       * We'll still allocate a temporary buffer/texture for the original
1361       * src region size but we'll only read the region which is on-screen.
1362       * This may mean that we draw garbage pixels into the dest region, but
1363       * that's expected.
1364       */
1365      readX = srcx;
1366      readY = srcy;
1367      readW = width;
1368      readH = height;
1369      pack = ctx->DefaultPacking;
1370      if (!_mesa_clip_readpixels(ctx, &readX, &readY, &readW, &readH, &pack))
1371         return GL_TRUE; /* all done */
1372
1373      /* clip against dest buffer bounds and scissor box */
1374      drawX = dstx + pack.SkipPixels;
1375      drawY = dsty + pack.SkipRows;
1376      unpack = pack;
1377      if (!_mesa_clip_drawpixels(ctx, &drawX, &drawY, &readW, &readH, &unpack))
1378         return GL_TRUE; /* all done */
1379
1380      readX = readX - pack.SkipPixels + unpack.SkipPixels;
1381      readY = readY - pack.SkipRows + unpack.SkipRows;
1382
1383      rbRead = st_get_color_read_renderbuffer(ctx);
1384      rbDraw = st_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0]);
1385
1386      if ((rbRead != rbDraw ||
1387           !regions_overlap(readX, readY, drawX, drawY, readW, readH)) &&
1388          rbRead->Base.Format == rbDraw->Base.Format) {
1389         struct pipe_box srcBox;
1390
1391         /* flip src/dst position if needed */
1392         if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
1393            /* both buffers will have the same orientation */
1394            readY = ctx->ReadBuffer->Height - readY - readH;
1395            drawY = ctx->DrawBuffer->Height - drawY - readH;
1396         }
1397
1398         u_box_2d(readX, readY, readW, readH, &srcBox);
1399
1400         pipe->resource_copy_region(pipe,
1401                                    rbDraw->texture,
1402                                    rbDraw->rtt_level, drawX, drawY, 0,
1403                                    rbRead->texture,
1404                                    rbRead->rtt_level, &srcBox);
1405         return GL_TRUE;
1406      }
1407   }
1408
1409   return GL_FALSE;
1410}
1411
1412
1413static void
1414st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
1415              GLsizei width, GLsizei height,
1416              GLint dstx, GLint dsty, GLenum type)
1417{
1418   struct st_context *st = st_context(ctx);
1419   struct pipe_context *pipe = st->pipe;
1420   struct pipe_screen *screen = pipe->screen;
1421   struct st_renderbuffer *rbRead;
1422   void *driver_vp, *driver_fp;
1423   struct pipe_resource *pt;
1424   struct pipe_sampler_view *sv[2];
1425   int num_sampler_view = 1;
1426   GLfloat *color;
1427   enum pipe_format srcFormat, texFormat;
1428   GLboolean invertTex = GL_FALSE;
1429   GLint readX, readY, readW, readH;
1430   GLuint sample_count;
1431   struct gl_pixelstore_attrib pack = ctx->DefaultPacking;
1432   struct st_fp_variant *fpv;
1433
1434   st_validate_state(st);
1435
1436   if (type == GL_DEPTH_STENCIL) {
1437      /* XXX make this more efficient */
1438      st_CopyPixels(ctx, srcx, srcy, width, height, dstx, dsty, GL_STENCIL);
1439      st_CopyPixels(ctx, srcx, srcy, width, height, dstx, dsty, GL_DEPTH);
1440      return;
1441   }
1442
1443   if (type == GL_STENCIL) {
1444      /* can't use texturing to do stencil */
1445      copy_stencil_pixels(ctx, srcx, srcy, width, height, dstx, dsty);
1446      return;
1447   }
1448
1449   if (blit_copy_pixels(ctx, srcx, srcy, width, height, dstx, dsty, type))
1450      return;
1451
1452   /*
1453    * The subsequent code implements glCopyPixels by copying the source
1454    * pixels into a temporary texture that's then applied to a textured quad.
1455    * When we draw the textured quad, all the usual per-fragment operations
1456    * are handled.
1457    */
1458
1459
1460   /*
1461    * Get vertex/fragment shaders
1462    */
1463   if (type == GL_COLOR) {
1464      rbRead = st_get_color_read_renderbuffer(ctx);
1465      color = NULL;
1466
1467      fpv = get_color_fp_variant(st);
1468      driver_fp = fpv->driver_shader;
1469
1470      driver_vp = make_passthrough_vertex_shader(st, GL_FALSE);
1471
1472      if (st->pixel_xfer.pixelmap_enabled) {
1473	  sv[1] = st->pixel_xfer.pixelmap_sampler_view;
1474	  num_sampler_view++;
1475      }
1476   }
1477   else {
1478      assert(type == GL_DEPTH);
1479      rbRead = st_renderbuffer(ctx->ReadBuffer->
1480                               Attachment[BUFFER_DEPTH].Renderbuffer);
1481      color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
1482
1483      fpv = get_depth_stencil_fp_variant(st, GL_TRUE, GL_FALSE);
1484      driver_fp = fpv->driver_shader;
1485
1486      driver_vp = make_passthrough_vertex_shader(st, GL_TRUE);
1487   }
1488
1489   /* update fragment program constants */
1490   st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
1491
1492   sample_count = rbRead->texture->nr_samples;
1493   /* I believe this would be legal, presumably would need to do a resolve
1494      for color, and for depth/stencil spec says to just use one of the
1495      depth/stencil samples per pixel? Need some transfer clarifications. */
1496   assert(sample_count < 2);
1497
1498   srcFormat = rbRead->texture->format;
1499
1500   if (screen->is_format_supported(screen, srcFormat, st->internal_target,
1501                                   sample_count,
1502                                   PIPE_BIND_SAMPLER_VIEW)) {
1503      texFormat = srcFormat;
1504   }
1505   else {
1506      /* srcFormat can't be used as a texture format */
1507      if (type == GL_DEPTH) {
1508         texFormat = st_choose_format(screen, GL_DEPTH_COMPONENT,
1509                                      GL_NONE, GL_NONE, st->internal_target,
1510				      sample_count, PIPE_BIND_DEPTH_STENCIL);
1511         assert(texFormat != PIPE_FORMAT_NONE);
1512      }
1513      else {
1514         /* default color format */
1515         texFormat = st_choose_format(screen, GL_RGBA,
1516                                      GL_NONE, GL_NONE, st->internal_target,
1517                                      sample_count, PIPE_BIND_SAMPLER_VIEW);
1518         assert(texFormat != PIPE_FORMAT_NONE);
1519      }
1520   }
1521
1522   /* Invert src region if needed */
1523   if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
1524      srcy = ctx->ReadBuffer->Height - srcy - height;
1525      invertTex = !invertTex;
1526   }
1527
1528   /* Clip the read region against the src buffer bounds.
1529    * We'll still allocate a temporary buffer/texture for the original
1530    * src region size but we'll only read the region which is on-screen.
1531    * This may mean that we draw garbage pixels into the dest region, but
1532    * that's expected.
1533    */
1534   readX = srcx;
1535   readY = srcy;
1536   readW = width;
1537   readH = height;
1538   if (!_mesa_clip_readpixels(ctx, &readX, &readY, &readW, &readH, &pack)) {
1539      /* The source region is completely out of bounds.  Do nothing.
1540       * The GL spec says "Results of copies from outside the window,
1541       * or from regions of the window that are not exposed, are
1542       * hardware dependent and undefined."
1543       */
1544      return;
1545   }
1546
1547   readW = MAX2(0, readW);
1548   readH = MAX2(0, readH);
1549
1550   /* alloc temporary texture */
1551   pt = alloc_texture(st, width, height, texFormat);
1552   if (!pt)
1553      return;
1554
1555   sv[0] = st_create_texture_sampler_view(st->pipe, pt);
1556   if (!sv[0]) {
1557      pipe_resource_reference(&pt, NULL);
1558      return;
1559   }
1560
1561   /* Make temporary texture which is a copy of the src region.
1562    */
1563   if (srcFormat == texFormat) {
1564      struct pipe_box src_box;
1565      u_box_2d(readX, readY, readW, readH, &src_box);
1566      /* copy source framebuffer surface into mipmap/texture */
1567      pipe->resource_copy_region(pipe,
1568                                 pt,                                /* dest tex */
1569                                 0,                                 /* dest lvl */
1570                                 pack.SkipPixels, pack.SkipRows, 0, /* dest pos */
1571                                 rbRead->texture,                   /* src tex */
1572                                 rbRead->rtt_level,                 /* src lvl */
1573                                 &src_box);
1574
1575   }
1576   else {
1577      /* CPU-based fallback/conversion */
1578      struct pipe_transfer *ptRead =
1579         pipe_get_transfer(st->pipe, rbRead->texture,
1580                           rbRead->rtt_level,
1581                           rbRead->rtt_face + rbRead->rtt_slice,
1582                           PIPE_TRANSFER_READ,
1583                           readX, readY, readW, readH);
1584      struct pipe_transfer *ptTex;
1585      enum pipe_transfer_usage transfer_usage;
1586
1587      if (ST_DEBUG & DEBUG_FALLBACK)
1588         debug_printf("%s: fallback processing\n", __FUNCTION__);
1589
1590      if (type == GL_DEPTH && util_format_is_depth_and_stencil(pt->format))
1591         transfer_usage = PIPE_TRANSFER_READ_WRITE;
1592      else
1593         transfer_usage = PIPE_TRANSFER_WRITE;
1594
1595      ptTex = pipe_get_transfer(st->pipe, pt, 0, 0, transfer_usage,
1596                                0, 0, width, height);
1597
1598      /* copy image from ptRead surface to ptTex surface */
1599      if (type == GL_COLOR) {
1600         /* alternate path using get/put_tile() */
1601         GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
1602         enum pipe_format readFormat, drawFormat;
1603         readFormat = util_format_linear(rbRead->texture->format);
1604         drawFormat = util_format_linear(pt->format);
1605         pipe_get_tile_rgba_format(pipe, ptRead, 0, 0, readW, readH,
1606                                   readFormat, buf);
1607         pipe_put_tile_rgba_format(pipe, ptTex, pack.SkipPixels, pack.SkipRows,
1608                                   readW, readH, drawFormat, buf);
1609         free(buf);
1610      }
1611      else {
1612         /* GL_DEPTH */
1613         GLuint *buf = (GLuint *) malloc(width * height * sizeof(GLuint));
1614         pipe_get_tile_z(pipe, ptRead, 0, 0, readW, readH, buf);
1615         pipe_put_tile_z(pipe, ptTex, pack.SkipPixels, pack.SkipRows,
1616                         readW, readH, buf);
1617         free(buf);
1618      }
1619
1620      pipe->transfer_destroy(pipe, ptRead);
1621      pipe->transfer_destroy(pipe, ptTex);
1622   }
1623
1624   /* OK, the texture 'pt' contains the src image/pixels.  Now draw a
1625    * textured quad with that texture.
1626    */
1627   draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2],
1628                      width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
1629                      sv,
1630                      num_sampler_view,
1631                      driver_vp,
1632                      driver_fp,
1633                      color, invertTex, GL_FALSE, GL_FALSE);
1634
1635   pipe_resource_reference(&pt, NULL);
1636   pipe_sampler_view_reference(&sv[0], NULL);
1637}
1638
1639
1640
1641void st_init_drawpixels_functions(struct dd_function_table *functions)
1642{
1643   functions->DrawPixels = st_DrawPixels;
1644   functions->CopyPixels = st_CopyPixels;
1645}
1646
1647
1648void
1649st_destroy_drawpix(struct st_context *st)
1650{
1651   GLuint i;
1652
1653   for (i = 0; i < Elements(st->drawpix.shaders); i++) {
1654      if (st->drawpix.shaders[i])
1655         _mesa_reference_fragprog(st->ctx, &st->drawpix.shaders[i], NULL);
1656   }
1657
1658   st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL);
1659   if (st->drawpix.vert_shaders[0])
1660      cso_delete_vertex_shader(st->cso_context, st->drawpix.vert_shaders[0]);
1661   if (st->drawpix.vert_shaders[1])
1662      cso_delete_vertex_shader(st->cso_context, st->drawpix.vert_shaders[1]);
1663}
1664
1665#endif /* FEATURE_drawpix */
1666