2 * GLSL pixel and vertex shader implementation
4 * Copyright 2006 Jason Green
5 * Copyright 2006-2007 Henri Verbeet
6 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
7 * Copyright 2009-2011 Henri Verbeet for CodeWeavers
9 * This library is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
14 * This library is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with this library; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
25 * D3D shader asm has swizzles on source parameters, and write masks for
26 * destination parameters. GLSL uses swizzles for both. The result of this is
27 * that for example "mov dst.xw, src.zyxw" becomes "dst.xw = src.zw" in GLSL.
28 * Ie, to generate a proper GLSL source swizzle, we need to take the D3D write
29 * mask for the destination parameter into account.
33 #include "wine/port.h"
38 #include "wined3d_private.h"
40 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
41 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants);
42 WINE_DECLARE_DEBUG_CHANNEL(d3d);
43 WINE_DECLARE_DEBUG_CHANNEL(winediag);
45 #define WINED3D_GLSL_SAMPLE_PROJECTED 0x1
46 #define WINED3D_GLSL_SAMPLE_RECT 0x2
47 #define WINED3D_GLSL_SAMPLE_LOD 0x4
48 #define WINED3D_GLSL_SAMPLE_GRAD 0x8
62 struct glsl_sample_function
70 HEAP_NODE_TRAVERSE_LEFT,
71 HEAP_NODE_TRAVERSE_RIGHT,
83 struct constant_entry *entries;
84 unsigned int *positions;
88 /* GLSL shader private data */
89 struct shader_glsl_priv {
90 struct wined3d_shader_buffer shader_buffer;
91 struct wine_rb_tree program_lookup;
92 struct glsl_shader_prog_link *glsl_program;
93 struct constant_heap vconst_heap;
94 struct constant_heap pconst_heap;
96 GLhandleARB depth_blt_program_full[tex_type_count];
97 GLhandleARB depth_blt_program_masked[tex_type_count];
98 UINT next_constant_version;
100 const struct fragment_pipeline *fragment_pipe;
101 struct wine_rb_tree ffp_fragment_shaders;
102 BOOL ffp_proj_control;
105 struct glsl_vs_program
107 struct list shader_entry;
109 GLenum vertex_color_clamp;
110 GLint *uniform_f_locations;
111 GLint uniform_i_locations[MAX_CONST_I];
112 GLint pos_fixup_location;
115 struct glsl_gs_program
117 struct list shader_entry;
121 struct glsl_ps_program
123 struct list shader_entry;
125 GLint *uniform_f_locations;
126 GLint uniform_i_locations[MAX_CONST_I];
127 GLint bumpenv_mat_location[MAX_TEXTURES];
128 GLint bumpenv_lum_scale_location[MAX_TEXTURES];
129 GLint bumpenv_lum_offset_location[MAX_TEXTURES];
130 GLint tex_factor_location;
131 GLint specular_enable_location;
132 GLint ycorrection_location;
133 GLint np2_fixup_location;
134 const struct ps_np2fixup_info *np2_fixup_info;
137 /* Struct to maintain data about a linked GLSL program */
138 struct glsl_shader_prog_link
140 struct wine_rb_entry program_lookup_entry;
141 struct glsl_vs_program vs;
142 struct glsl_gs_program gs;
143 struct glsl_ps_program ps;
144 GLhandleARB programId;
145 UINT constant_version;
148 struct glsl_program_key
155 struct shader_glsl_ctx_priv {
156 const struct vs_compile_args *cur_vs_args;
157 const struct ps_compile_args *cur_ps_args;
158 struct ps_np2fixup_info *cur_np2fixup_info;
161 struct glsl_ps_compiled_shader
163 struct ps_compile_args args;
164 struct ps_np2fixup_info np2fixup;
168 struct glsl_vs_compiled_shader
170 struct vs_compile_args args;
174 struct glsl_gs_compiled_shader
179 struct glsl_shader_private
183 struct glsl_vs_compiled_shader *vs;
184 struct glsl_gs_compiled_shader *gs;
185 struct glsl_ps_compiled_shader *ps;
187 UINT num_gl_shaders, shader_array_size;
190 struct glsl_ffp_fragment_shader
192 struct ffp_frag_desc entry;
194 struct list linked_programs;
197 static const char *debug_gl_shader_type(GLenum type)
201 #define WINED3D_TO_STR(u) case u: return #u
202 WINED3D_TO_STR(GL_VERTEX_SHADER_ARB);
203 WINED3D_TO_STR(GL_GEOMETRY_SHADER_ARB);
204 WINED3D_TO_STR(GL_FRAGMENT_SHADER_ARB);
205 #undef WINED3D_TO_STR
207 return wine_dbg_sprintf("UNKNOWN(%#x)", type);
211 static const char *shader_glsl_get_prefix(enum wined3d_shader_type type)
215 case WINED3D_SHADER_TYPE_VERTEX:
218 case WINED3D_SHADER_TYPE_GEOMETRY:
221 case WINED3D_SHADER_TYPE_PIXEL:
225 FIXME("Unhandled shader type %#x.\n", type);
230 /* Extract a line from the info log.
231 * Note that this modifies the source string. */
232 static char *get_info_log_line(char **ptr)
237 if (!(q = strstr(p, "\n")))
239 if (!*p) return NULL;
249 /* Context activation is done by the caller. */
250 static void print_glsl_info_log(const struct wined3d_gl_info *gl_info, GLhandleARB obj)
252 int infologLength = 0;
255 if (!WARN_ON(d3d_shader) && !FIXME_ON(d3d_shader))
258 GL_EXTCALL(glGetObjectParameterivARB(obj,
259 GL_OBJECT_INFO_LOG_LENGTH_ARB,
262 /* A size of 1 is just a null-terminated string, so the log should be bigger than
263 * that if there are errors. */
264 if (infologLength > 1)
268 infoLog = HeapAlloc(GetProcessHeap(), 0, infologLength);
269 /* The info log is supposed to be zero-terminated, but at least some
270 * versions of fglrx don't terminate the string properly. The reported
271 * length does include the terminator, so explicitly set it to zero
273 infoLog[infologLength - 1] = 0;
274 GL_EXTCALL(glGetInfoLogARB(obj, infologLength, NULL, infoLog));
277 if (gl_info->quirks & WINED3D_QUIRK_INFO_LOG_SPAM)
279 WARN("Info log received from GLSL shader #%u:\n", obj);
280 while ((line = get_info_log_line(&ptr))) WARN(" %s\n", line);
284 FIXME("Info log received from GLSL shader #%u:\n", obj);
285 while ((line = get_info_log_line(&ptr))) FIXME(" %s\n", line);
287 HeapFree(GetProcessHeap(), 0, infoLog);
291 /* Context activation is done by the caller. */
292 static void shader_glsl_compile(const struct wined3d_gl_info *gl_info, GLhandleARB shader, const char *src)
294 TRACE("Compiling shader object %u.\n", shader);
295 GL_EXTCALL(glShaderSourceARB(shader, 1, &src, NULL));
296 checkGLcall("glShaderSourceARB");
297 GL_EXTCALL(glCompileShaderARB(shader));
298 checkGLcall("glCompileShaderARB");
299 print_glsl_info_log(gl_info, shader);
302 /* Context activation is done by the caller. */
303 static void shader_glsl_dump_program_source(const struct wined3d_gl_info *gl_info, GLhandleARB program)
305 GLint i, object_count, source_size = -1;
306 GLhandleARB *objects;
309 GL_EXTCALL(glGetObjectParameterivARB(program, GL_OBJECT_ATTACHED_OBJECTS_ARB, &object_count));
310 objects = HeapAlloc(GetProcessHeap(), 0, object_count * sizeof(*objects));
313 ERR("Failed to allocate object array memory.\n");
317 GL_EXTCALL(glGetAttachedObjectsARB(program, object_count, NULL, objects));
318 for (i = 0; i < object_count; ++i)
323 GL_EXTCALL(glGetObjectParameterivARB(objects[i], GL_OBJECT_SHADER_SOURCE_LENGTH_ARB, &tmp));
325 if (source_size < tmp)
327 HeapFree(GetProcessHeap(), 0, source);
329 source = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, tmp);
332 ERR("Failed to allocate %d bytes for shader source.\n", tmp);
333 HeapFree(GetProcessHeap(), 0, objects);
339 FIXME("Object %u:\n", objects[i]);
340 GL_EXTCALL(glGetObjectParameterivARB(objects[i], GL_OBJECT_SUBTYPE_ARB, &tmp));
341 FIXME(" GL_OBJECT_SUBTYPE_ARB: %s.\n", debug_gl_shader_type(tmp));
342 GL_EXTCALL(glGetObjectParameterivARB(objects[i], GL_OBJECT_COMPILE_STATUS_ARB, &tmp));
343 FIXME(" GL_OBJECT_COMPILE_STATUS_ARB: %d.\n", tmp);
347 GL_EXTCALL(glGetShaderSourceARB(objects[i], source_size, NULL, source));
348 while ((line = get_info_log_line(&ptr))) FIXME(" %s\n", line);
352 HeapFree(GetProcessHeap(), 0, source);
353 HeapFree(GetProcessHeap(), 0, objects);
356 /* Context activation is done by the caller. */
357 static void shader_glsl_validate_link(const struct wined3d_gl_info *gl_info, GLhandleARB program)
361 if (!TRACE_ON(d3d_shader) && !FIXME_ON(d3d_shader)) return;
363 GL_EXTCALL(glGetObjectParameterivARB(program, GL_OBJECT_TYPE_ARB, &tmp));
364 if (tmp == GL_PROGRAM_OBJECT_ARB)
366 GL_EXTCALL(glGetObjectParameterivARB(program, GL_OBJECT_LINK_STATUS_ARB, &tmp));
369 FIXME("Program %u link status invalid.\n", program);
370 shader_glsl_dump_program_source(gl_info, program);
374 print_glsl_info_log(gl_info, program);
377 /* Context activation is done by the caller. */
378 static void shader_glsl_load_psamplers(const struct wined3d_gl_info *gl_info,
379 const DWORD *tex_unit_map, GLhandleARB programId)
382 char sampler_name[20];
385 for (i = 0; i < MAX_FRAGMENT_SAMPLERS; ++i)
387 snprintf(sampler_name, sizeof(sampler_name), "ps_sampler%u", i);
388 name_loc = GL_EXTCALL(glGetUniformLocationARB(programId, sampler_name));
389 if (name_loc != -1) {
390 DWORD mapped_unit = tex_unit_map[i];
391 if (mapped_unit != WINED3D_UNMAPPED_STAGE && mapped_unit < gl_info->limits.fragment_samplers)
393 TRACE("Loading %s for texture %d\n", sampler_name, mapped_unit);
394 GL_EXTCALL(glUniform1iARB(name_loc, mapped_unit));
395 checkGLcall("glUniform1iARB");
397 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name, mapped_unit);
403 /* Context activation is done by the caller. */
404 static void shader_glsl_load_vsamplers(const struct wined3d_gl_info *gl_info,
405 const DWORD *tex_unit_map, GLhandleARB programId)
408 char sampler_name[20];
411 for (i = 0; i < MAX_VERTEX_SAMPLERS; ++i)
413 snprintf(sampler_name, sizeof(sampler_name), "vs_sampler%u", i);
414 name_loc = GL_EXTCALL(glGetUniformLocationARB(programId, sampler_name));
415 if (name_loc != -1) {
416 DWORD mapped_unit = tex_unit_map[MAX_FRAGMENT_SAMPLERS + i];
417 if (mapped_unit != WINED3D_UNMAPPED_STAGE && mapped_unit < gl_info->limits.combined_samplers)
419 TRACE("Loading %s for texture %d\n", sampler_name, mapped_unit);
420 GL_EXTCALL(glUniform1iARB(name_loc, mapped_unit));
421 checkGLcall("glUniform1iARB");
423 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name, mapped_unit);
429 /* Context activation is done by the caller. */
430 static inline void walk_constant_heap(const struct wined3d_gl_info *gl_info, const float *constants,
431 const GLint *constant_locations, const struct constant_heap *heap, unsigned char *stack, DWORD version)
434 unsigned int heap_idx = 1;
437 if (heap->entries[heap_idx].version <= version) return;
439 idx = heap->entries[heap_idx].idx;
440 if (constant_locations[idx] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations[idx], 1, &constants[idx * 4]));
441 stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
443 while (stack_idx >= 0)
445 /* Note that we fall through to the next case statement. */
446 switch(stack[stack_idx])
448 case HEAP_NODE_TRAVERSE_LEFT:
450 unsigned int left_idx = heap_idx << 1;
451 if (left_idx < heap->size && heap->entries[left_idx].version > version)
454 idx = heap->entries[heap_idx].idx;
455 if (constant_locations[idx] != -1)
456 GL_EXTCALL(glUniform4fvARB(constant_locations[idx], 1, &constants[idx * 4]));
458 stack[stack_idx++] = HEAP_NODE_TRAVERSE_RIGHT;
459 stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
464 case HEAP_NODE_TRAVERSE_RIGHT:
466 unsigned int right_idx = (heap_idx << 1) + 1;
467 if (right_idx < heap->size && heap->entries[right_idx].version > version)
469 heap_idx = right_idx;
470 idx = heap->entries[heap_idx].idx;
471 if (constant_locations[idx] != -1)
472 GL_EXTCALL(glUniform4fvARB(constant_locations[idx], 1, &constants[idx * 4]));
474 stack[stack_idx++] = HEAP_NODE_POP;
475 stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
486 checkGLcall("walk_constant_heap()");
489 /* Context activation is done by the caller. */
490 static inline void apply_clamped_constant(const struct wined3d_gl_info *gl_info, GLint location, const GLfloat *data)
492 GLfloat clamped_constant[4];
494 if (location == -1) return;
496 clamped_constant[0] = data[0] < -1.0f ? -1.0f : data[0] > 1.0f ? 1.0f : data[0];
497 clamped_constant[1] = data[1] < -1.0f ? -1.0f : data[1] > 1.0f ? 1.0f : data[1];
498 clamped_constant[2] = data[2] < -1.0f ? -1.0f : data[2] > 1.0f ? 1.0f : data[2];
499 clamped_constant[3] = data[3] < -1.0f ? -1.0f : data[3] > 1.0f ? 1.0f : data[3];
501 GL_EXTCALL(glUniform4fvARB(location, 1, clamped_constant));
504 /* Context activation is done by the caller. */
505 static inline void walk_constant_heap_clamped(const struct wined3d_gl_info *gl_info, const float *constants,
506 const GLint *constant_locations, const struct constant_heap *heap, unsigned char *stack, DWORD version)
509 unsigned int heap_idx = 1;
512 if (heap->entries[heap_idx].version <= version) return;
514 idx = heap->entries[heap_idx].idx;
515 apply_clamped_constant(gl_info, constant_locations[idx], &constants[idx * 4]);
516 stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
518 while (stack_idx >= 0)
520 /* Note that we fall through to the next case statement. */
521 switch(stack[stack_idx])
523 case HEAP_NODE_TRAVERSE_LEFT:
525 unsigned int left_idx = heap_idx << 1;
526 if (left_idx < heap->size && heap->entries[left_idx].version > version)
529 idx = heap->entries[heap_idx].idx;
530 apply_clamped_constant(gl_info, constant_locations[idx], &constants[idx * 4]);
532 stack[stack_idx++] = HEAP_NODE_TRAVERSE_RIGHT;
533 stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
538 case HEAP_NODE_TRAVERSE_RIGHT:
540 unsigned int right_idx = (heap_idx << 1) + 1;
541 if (right_idx < heap->size && heap->entries[right_idx].version > version)
543 heap_idx = right_idx;
544 idx = heap->entries[heap_idx].idx;
545 apply_clamped_constant(gl_info, constant_locations[idx], &constants[idx * 4]);
547 stack[stack_idx++] = HEAP_NODE_POP;
548 stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
559 checkGLcall("walk_constant_heap_clamped()");
562 /* Context activation is done by the caller. */
563 static void shader_glsl_load_constantsF(const struct wined3d_shader *shader, const struct wined3d_gl_info *gl_info,
564 const float *constants, const GLint *constant_locations, const struct constant_heap *heap,
565 unsigned char *stack, UINT version)
567 const struct wined3d_shader_lconst *lconst;
569 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
570 if (shader->reg_maps.shader_version.major == 1
571 && shader->reg_maps.shader_version.type == WINED3D_SHADER_TYPE_PIXEL)
572 walk_constant_heap_clamped(gl_info, constants, constant_locations, heap, stack, version);
574 walk_constant_heap(gl_info, constants, constant_locations, heap, stack, version);
576 if (!shader->load_local_constsF)
578 TRACE("No need to load local float constants for this shader\n");
582 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
583 LIST_FOR_EACH_ENTRY(lconst, &shader->constantsF, struct wined3d_shader_lconst, entry)
585 GLint location = constant_locations[lconst->idx];
586 /* We found this uniform name in the program - go ahead and send the data */
587 if (location != -1) GL_EXTCALL(glUniform4fvARB(location, 1, (const GLfloat *)lconst->value));
589 checkGLcall("glUniform4fvARB()");
592 /* Context activation is done by the caller. */
593 static void shader_glsl_load_constantsI(const struct wined3d_shader *shader, const struct wined3d_gl_info *gl_info,
594 const GLint locations[MAX_CONST_I], const int *constants, WORD constants_set)
599 for (i = 0; constants_set; constants_set >>= 1, ++i)
601 if (!(constants_set & 1)) continue;
603 TRACE_(d3d_constants)("Loading constants %u: %i, %i, %i, %i\n",
604 i, constants[i*4], constants[i*4+1], constants[i*4+2], constants[i*4+3]);
606 /* We found this uniform name in the program - go ahead and send the data */
607 GL_EXTCALL(glUniform4ivARB(locations[i], 1, &constants[i*4]));
608 checkGLcall("glUniform4ivARB");
611 /* Load immediate constants */
612 ptr = list_head(&shader->constantsI);
615 const struct wined3d_shader_lconst *lconst = LIST_ENTRY(ptr, const struct wined3d_shader_lconst, entry);
616 unsigned int idx = lconst->idx;
617 const GLint *values = (const GLint *)lconst->value;
619 TRACE_(d3d_constants)("Loading local constants %i: %i, %i, %i, %i\n", idx,
620 values[0], values[1], values[2], values[3]);
622 /* We found this uniform name in the program - go ahead and send the data */
623 GL_EXTCALL(glUniform4ivARB(locations[idx], 1, values));
624 checkGLcall("glUniform4ivARB");
625 ptr = list_next(&shader->constantsI, ptr);
629 /* Context activation is done by the caller. */
630 static void shader_glsl_load_constantsB(const struct wined3d_shader *shader, const struct wined3d_gl_info *gl_info,
631 GLhandleARB programId, const BOOL *constants, WORD constants_set)
639 prefix = shader_glsl_get_prefix(shader->reg_maps.shader_version.type);
641 /* TODO: Benchmark and see if it would be beneficial to store the
642 * locations of the constants to avoid looking up each time */
643 for (i = 0; constants_set; constants_set >>= 1, ++i)
645 if (!(constants_set & 1)) continue;
647 TRACE_(d3d_constants)("Loading constants %i: %i;\n", i, constants[i]);
649 /* TODO: Benchmark and see if it would be beneficial to store the
650 * locations of the constants to avoid looking up each time */
651 snprintf(tmp_name, sizeof(tmp_name), "%s_b[%i]", prefix, i);
652 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
655 /* We found this uniform name in the program - go ahead and send the data */
656 GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, &constants[i]));
657 checkGLcall("glUniform1ivARB");
661 /* Load immediate constants */
662 ptr = list_head(&shader->constantsB);
665 const struct wined3d_shader_lconst *lconst = LIST_ENTRY(ptr, const struct wined3d_shader_lconst, entry);
666 unsigned int idx = lconst->idx;
667 const GLint *values = (const GLint *)lconst->value;
669 TRACE_(d3d_constants)("Loading local constants %i: %i\n", idx, values[0]);
671 snprintf(tmp_name, sizeof(tmp_name), "%s_b[%i]", prefix, idx);
672 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
674 /* We found this uniform name in the program - go ahead and send the data */
675 GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, values));
676 checkGLcall("glUniform1ivARB");
678 ptr = list_next(&shader->constantsB, ptr);
682 static void reset_program_constant_version(struct wine_rb_entry *entry, void *context)
684 WINE_RB_ENTRY_VALUE(entry, struct glsl_shader_prog_link, program_lookup_entry)->constant_version = 0;
687 /* Context activation is done by the caller (state handler). */
688 static void shader_glsl_load_np2fixup_constants(void *shader_priv,
689 const struct wined3d_gl_info *gl_info, const struct wined3d_state *state)
691 struct shader_glsl_priv *glsl_priv = shader_priv;
692 const struct glsl_shader_prog_link *prog = glsl_priv->glsl_program;
694 /* No GLSL program set - nothing to do. */
697 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
698 if (!use_ps(state)) return;
700 if (prog->ps.np2_fixup_info && prog->ps.np2_fixup_location != -1)
703 UINT fixup = prog->ps.np2_fixup_info->active;
704 GLfloat np2fixup_constants[4 * MAX_FRAGMENT_SAMPLERS];
706 for (i = 0; fixup; fixup >>= 1, ++i)
708 const struct wined3d_texture *tex = state->textures[i];
709 const unsigned char idx = prog->ps.np2_fixup_info->idx[i];
710 GLfloat *tex_dim = &np2fixup_constants[(idx >> 1) * 4];
714 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
720 tex_dim[2] = tex->pow2_matrix[0];
721 tex_dim[3] = tex->pow2_matrix[5];
725 tex_dim[0] = tex->pow2_matrix[0];
726 tex_dim[1] = tex->pow2_matrix[5];
730 GL_EXTCALL(glUniform4fvARB(prog->ps.np2_fixup_location,
731 prog->ps.np2_fixup_info->num_consts, np2fixup_constants));
735 /* Context activation is done by the caller (state handler). */
736 static void shader_glsl_load_constants(const struct wined3d_context *context,
737 BOOL usePixelShader, BOOL useVertexShader)
739 const struct wined3d_gl_info *gl_info = context->gl_info;
740 struct wined3d_device *device = context->swapchain->device;
741 struct wined3d_stateblock *stateBlock = device->stateBlock;
742 const struct wined3d_state *state = &stateBlock->state;
743 struct shader_glsl_priv *priv = device->shader_priv;
744 float position_fixup[4];
746 GLhandleARB programId;
747 struct glsl_shader_prog_link *prog = priv->glsl_program;
748 UINT constant_version;
752 /* No GLSL program set - nothing to do. */
755 programId = prog->programId;
756 constant_version = prog->constant_version;
760 const struct wined3d_shader *vshader = state->vertex_shader;
762 /* Load DirectX 9 float constants/uniforms for vertex shader */
763 shader_glsl_load_constantsF(vshader, gl_info, state->vs_consts_f,
764 prog->vs.uniform_f_locations, &priv->vconst_heap, priv->stack, constant_version);
766 /* Load DirectX 9 integer constants/uniforms for vertex shader */
767 shader_glsl_load_constantsI(vshader, gl_info, prog->vs.uniform_i_locations, state->vs_consts_i,
768 stateBlock->changed.vertexShaderConstantsI & vshader->reg_maps.integer_constants);
770 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
771 shader_glsl_load_constantsB(vshader, gl_info, programId, state->vs_consts_b,
772 stateBlock->changed.vertexShaderConstantsB & vshader->reg_maps.boolean_constants);
774 /* Upload the position fixup params */
775 shader_get_position_fixup(context, state, position_fixup);
776 GL_EXTCALL(glUniform4fvARB(prog->vs.pos_fixup_location, 1, position_fixup));
777 checkGLcall("glUniform4fvARB");
782 const struct wined3d_shader *pshader = state->pixel_shader;
784 /* Load DirectX 9 float constants/uniforms for pixel shader */
785 shader_glsl_load_constantsF(pshader, gl_info, state->ps_consts_f,
786 prog->ps.uniform_f_locations, &priv->pconst_heap, priv->stack, constant_version);
788 /* Load DirectX 9 integer constants/uniforms for pixel shader */
789 shader_glsl_load_constantsI(pshader, gl_info, prog->ps.uniform_i_locations, state->ps_consts_i,
790 stateBlock->changed.pixelShaderConstantsI & pshader->reg_maps.integer_constants);
792 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
793 shader_glsl_load_constantsB(pshader, gl_info, programId, state->ps_consts_b,
794 stateBlock->changed.pixelShaderConstantsB & pshader->reg_maps.boolean_constants);
796 /* Upload the environment bump map matrix if needed. The needsbumpmat
797 * member specifies the texture stage to load the matrix from. It
798 * can't be 0 for a valid texbem instruction. */
799 for (i = 0; i < MAX_TEXTURES; ++i)
803 if (prog->ps.bumpenv_mat_location[i] == -1)
806 data = (const float *)&state->texture_states[i][WINED3D_TSS_BUMPENV_MAT00];
807 GL_EXTCALL(glUniformMatrix2fvARB(prog->ps.bumpenv_mat_location[i], 1, 0, data));
808 checkGLcall("glUniformMatrix2fvARB");
810 /* texbeml needs the luminance scale and offset too. If texbeml
811 * is used, needsbumpmat is set too, so we can check that in the
812 * needsbumpmat check. */
813 if (prog->ps.bumpenv_lum_scale_location[i] != -1)
815 const GLfloat *scale = (const GLfloat *)&state->texture_states[i][WINED3D_TSS_BUMPENV_LSCALE];
816 const GLfloat *offset = (const GLfloat *)&state->texture_states[i][WINED3D_TSS_BUMPENV_LOFFSET];
818 GL_EXTCALL(glUniform1fvARB(prog->ps.bumpenv_lum_scale_location[i], 1, scale));
819 checkGLcall("glUniform1fvARB");
820 GL_EXTCALL(glUniform1fvARB(prog->ps.bumpenv_lum_offset_location[i], 1, offset));
821 checkGLcall("glUniform1fvARB");
825 if (prog->ps.ycorrection_location != -1)
827 float correction_params[4];
829 if (context->render_offscreen)
831 correction_params[0] = 0.0f;
832 correction_params[1] = 1.0f;
834 /* position is window relative, not viewport relative */
835 correction_params[0] = (float) context->current_rt->resource.height;
836 correction_params[1] = -1.0f;
838 GL_EXTCALL(glUniform4fvARB(prog->ps.ycorrection_location, 1, correction_params));
841 else if (priv->fragment_pipe == &glsl_fragment_pipe)
845 for (i = 0; i < MAX_TEXTURES; ++i)
847 GL_EXTCALL(glUniformMatrix2fvARB(prog->ps.bumpenv_mat_location[i], 1, 0,
848 (const float *)&state->texture_states[i][WINED3D_TSS_BUMPENV_MAT00]));
849 GL_EXTCALL(glUniform1fARB(prog->ps.bumpenv_lum_scale_location[i],
850 *(const float *)&state->texture_states[i][WINED3D_TSS_BUMPENV_LSCALE]));
851 GL_EXTCALL(glUniform1fARB(prog->ps.bumpenv_lum_offset_location[i],
852 *(const float *)&state->texture_states[i][WINED3D_TSS_BUMPENV_LOFFSET]));
855 D3DCOLORTOGLFLOAT4(state->render_states[WINED3D_RS_TEXTUREFACTOR], col);
856 GL_EXTCALL(glUniform4fARB(prog->ps.tex_factor_location, col[0], col[1], col[2], col[3]));
858 if (state->render_states[WINED3D_RS_SPECULARENABLE])
859 GL_EXTCALL(glUniform4fARB(prog->ps.specular_enable_location, 1.0f, 1.0f, 1.0f, 0.0f));
861 GL_EXTCALL(glUniform4fARB(prog->ps.specular_enable_location, 0.0f, 0.0f, 0.0f, 0.0f));
863 checkGLcall("fixed function uniforms");
866 if (priv->next_constant_version == UINT_MAX)
868 TRACE("Max constant version reached, resetting to 0.\n");
869 wine_rb_for_each_entry(&priv->program_lookup, reset_program_constant_version, NULL);
870 priv->next_constant_version = 1;
874 prog->constant_version = priv->next_constant_version++;
878 static void update_heap_entry(const struct constant_heap *heap, unsigned int idx,
879 unsigned int heap_idx, DWORD new_version)
881 struct constant_entry *entries = heap->entries;
882 unsigned int *positions = heap->positions;
883 unsigned int parent_idx;
887 parent_idx = heap_idx >> 1;
889 if (new_version <= entries[parent_idx].version) break;
891 entries[heap_idx] = entries[parent_idx];
892 positions[entries[parent_idx].idx] = heap_idx;
893 heap_idx = parent_idx;
896 entries[heap_idx].version = new_version;
897 entries[heap_idx].idx = idx;
898 positions[idx] = heap_idx;
901 static void shader_glsl_update_float_vertex_constants(struct wined3d_device *device, UINT start, UINT count)
903 struct shader_glsl_priv *priv = device->shader_priv;
904 struct constant_heap *heap = &priv->vconst_heap;
907 for (i = start; i < count + start; ++i)
909 if (!device->stateBlock->changed.vertexShaderConstantsF[i])
910 update_heap_entry(heap, i, heap->size++, priv->next_constant_version);
912 update_heap_entry(heap, i, heap->positions[i], priv->next_constant_version);
916 static void shader_glsl_update_float_pixel_constants(struct wined3d_device *device, UINT start, UINT count)
918 struct shader_glsl_priv *priv = device->shader_priv;
919 struct constant_heap *heap = &priv->pconst_heap;
922 for (i = start; i < count + start; ++i)
924 if (!device->stateBlock->changed.pixelShaderConstantsF[i])
925 update_heap_entry(heap, i, heap->size++, priv->next_constant_version);
927 update_heap_entry(heap, i, heap->positions[i], priv->next_constant_version);
931 static unsigned int vec4_varyings(DWORD shader_major, const struct wined3d_gl_info *gl_info)
933 unsigned int ret = gl_info->limits.glsl_varyings / 4;
934 /* 4.0 shaders do not write clip coords because d3d10 does not support user clipplanes */
935 if(shader_major > 3) return ret;
937 /* 3.0 shaders may need an extra varying for the clip coord on some cards(mostly dx10 ones) */
938 if (gl_info->quirks & WINED3D_QUIRK_GLSL_CLIP_VARYING) ret -= 1;
942 /** Generate the variable & register declarations for the GLSL output target */
943 static void shader_generate_glsl_declarations(const struct wined3d_context *context,
944 struct wined3d_shader_buffer *buffer, const struct wined3d_shader *shader,
945 const struct wined3d_shader_reg_maps *reg_maps, const struct shader_glsl_ctx_priv *ctx_priv)
947 const struct wined3d_shader_version *version = ®_maps->shader_version;
948 const struct wined3d_state *state = &shader->device->stateBlock->state;
949 const struct ps_compile_args *ps_args = ctx_priv->cur_ps_args;
950 const struct wined3d_gl_info *gl_info = context->gl_info;
951 const struct wined3d_fb_state *fb = &shader->device->fb;
952 unsigned int i, extra_constants_needed = 0;
953 const struct wined3d_shader_lconst *lconst;
957 prefix = shader_glsl_get_prefix(version->type);
959 /* Prototype the subroutines */
960 for (i = 0, map = reg_maps->labels; map; map >>= 1, ++i)
962 if (map & 1) shader_addline(buffer, "void subroutine%u();\n", i);
965 /* Declare the constants (aka uniforms) */
966 if (shader->limits.constant_float > 0)
968 unsigned max_constantsF;
970 /* Unless the shader uses indirect addressing, always declare the
971 * maximum array size and ignore that we need some uniforms privately.
972 * E.g. if GL supports 256 uniforms, and we need 2 for the pos fixup
973 * and immediate values, still declare VC[256]. If the shader needs
974 * more uniforms than we have it won't work in any case. If it uses
975 * less, the compiler will figure out which uniforms are really used
976 * and strip them out. This allows a shader to use c255 on a dx9 card,
977 * as long as it doesn't also use all the other constants.
979 * If the shader uses indirect addressing the compiler must assume
980 * that all declared uniforms are used. In this case, declare only the
981 * amount that we're assured to have.
983 * Thus we run into problems in these two cases:
984 * 1) The shader really uses more uniforms than supported.
985 * 2) The shader uses indirect addressing, less constants than
986 * supported, but uses a constant index > #supported consts. */
987 if (version->type == WINED3D_SHADER_TYPE_PIXEL)
989 /* No indirect addressing here. */
990 max_constantsF = gl_info->limits.glsl_ps_float_constants;
994 if (reg_maps->usesrelconstF)
996 /* Subtract the other potential uniforms from the max
997 * available (bools, ints, and 1 row of projection matrix).
998 * Subtract another uniform for immediate values, which have
999 * to be loaded via uniform by the driver as well. The shader
1000 * code only uses 0.5, 2.0, 1.0, 128 and -128 in vertex
1001 * shader code, so one vec4 should be enough. (Unfortunately
1002 * the Nvidia driver doesn't store 128 and -128 in one float).
1004 * Writing gl_ClipVertex requires one uniform for each
1005 * clipplane as well. */
1006 max_constantsF = gl_info->limits.glsl_vs_float_constants - 3;
1007 if(ctx_priv->cur_vs_args->clip_enabled)
1009 max_constantsF -= gl_info->limits.clipplanes;
1011 max_constantsF -= count_bits(reg_maps->integer_constants);
1012 /* Strictly speaking a bool only uses one scalar, but the nvidia(Linux) compiler doesn't pack them properly,
1013 * so each scalar requires a full vec4. We could work around this by packing the booleans ourselves, but
1014 * for now take this into account when calculating the number of available constants
1016 max_constantsF -= count_bits(reg_maps->boolean_constants);
1017 /* Set by driver quirks in directx.c */
1018 max_constantsF -= gl_info->reserved_glsl_constants;
1020 if (max_constantsF < shader->limits.constant_float)
1022 static unsigned int once;
1025 ERR_(winediag)("The hardware does not support enough uniform components to run this shader,"
1026 " it may not render correctly.\n");
1028 WARN("The hardware does not support enough uniform components to run this shader.\n");
1033 max_constantsF = gl_info->limits.glsl_vs_float_constants;
1036 max_constantsF = min(shader->limits.constant_float, max_constantsF);
1037 shader_addline(buffer, "uniform vec4 %s_c[%u];\n", prefix, max_constantsF);
1040 /* Always declare the full set of constants, the compiler can remove the
1041 * unused ones because d3d doesn't (yet) support indirect int and bool
1042 * constant addressing. This avoids problems if the app uses e.g. i0 and i9. */
1043 if (shader->limits.constant_int > 0 && reg_maps->integer_constants)
1044 shader_addline(buffer, "uniform ivec4 %s_i[%u];\n", prefix, shader->limits.constant_int);
1046 if (shader->limits.constant_bool > 0 && reg_maps->boolean_constants)
1047 shader_addline(buffer, "uniform bool %s_b[%u];\n", prefix, shader->limits.constant_bool);
1049 for (i = 0; i < WINED3D_MAX_CBS; ++i)
1051 if (reg_maps->cb_sizes[i])
1052 shader_addline(buffer, "uniform vec4 %s_cb%u[%u];\n", prefix, i, reg_maps->cb_sizes[i]);
1055 /* Declare texture samplers */
1056 for (i = 0; i < shader->limits.sampler; ++i)
1058 if (reg_maps->sampler_type[i])
1060 BOOL shadow_sampler = version->type == WINED3D_SHADER_TYPE_PIXEL && (ps_args->shadow & (1 << i));
1061 const struct wined3d_texture *texture;
1063 switch (reg_maps->sampler_type[i])
1067 shader_addline(buffer, "uniform sampler1DShadow %s_sampler%u;\n", prefix, i);
1069 shader_addline(buffer, "uniform sampler1D %s_sampler%u;\n", prefix, i);
1072 texture = state->textures[i];
1075 if (texture && texture->target == GL_TEXTURE_RECTANGLE_ARB)
1076 shader_addline(buffer, "uniform sampler2DRectShadow %s_sampler%u;\n", prefix, i);
1078 shader_addline(buffer, "uniform sampler2DShadow %s_sampler%u;\n", prefix, i);
1082 if (texture && texture->target == GL_TEXTURE_RECTANGLE_ARB)
1083 shader_addline(buffer, "uniform sampler2DRect %s_sampler%u;\n", prefix, i);
1085 shader_addline(buffer, "uniform sampler2D %s_sampler%u;\n", prefix, i);
1088 case WINED3DSTT_CUBE:
1090 FIXME("Unsupported Cube shadow sampler.\n");
1091 shader_addline(buffer, "uniform samplerCube %s_sampler%u;\n", prefix, i);
1093 case WINED3DSTT_VOLUME:
1095 FIXME("Unsupported 3D shadow sampler.\n");
1096 shader_addline(buffer, "uniform sampler3D %s_sampler%u;\n", prefix, i);
1099 shader_addline(buffer, "uniform unsupported_sampler %s_sampler%u;\n", prefix, i);
1100 FIXME("Unrecognized sampler type: %#x\n", reg_maps->sampler_type[i]);
1106 /* Declare uniforms for NP2 texcoord fixup:
1107 * This is NOT done inside the loop that declares the texture samplers
1108 * since the NP2 fixup code is currently only used for the GeforceFX
1109 * series and when forcing the ARB_npot extension off. Modern cards just
1110 * skip the code anyway, so put it inside a separate loop. */
1111 if (version->type == WINED3D_SHADER_TYPE_PIXEL && ps_args->np2_fixup)
1113 struct ps_np2fixup_info *fixup = ctx_priv->cur_np2fixup_info;
1116 /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
1117 * while D3D has them in the (normalized) [0,1]x[0,1] range.
1118 * samplerNP2Fixup stores texture dimensions and is updated through
1119 * shader_glsl_load_np2fixup_constants when the sampler changes. */
1121 for (i = 0; i < shader->limits.sampler; ++i)
1123 if (reg_maps->sampler_type[i])
1125 if (!(ps_args->np2_fixup & (1 << i))) continue;
1127 if (WINED3DSTT_2D != reg_maps->sampler_type[i]) {
1128 FIXME("Non-2D texture is flagged for NP2 texcoord fixup.\n");
1132 fixup->idx[i] = cur++;
1136 fixup->num_consts = (cur + 1) >> 1;
1137 fixup->active = ps_args->np2_fixup;
1138 shader_addline(buffer, "uniform vec4 %s_samplerNP2Fixup[%u];\n", prefix, fixup->num_consts);
1141 /* Declare address variables */
1142 for (i = 0, map = reg_maps->address; map; map >>= 1, ++i)
1144 if (map & 1) shader_addline(buffer, "ivec4 A%u;\n", i);
1147 /* Declare texture coordinate temporaries and initialize them */
1148 for (i = 0, map = reg_maps->texcoord; map; map >>= 1, ++i)
1150 if (map & 1) shader_addline(buffer, "vec4 T%u = gl_TexCoord[%u];\n", i, i);
1153 if (version->type == WINED3D_SHADER_TYPE_VERTEX)
1155 /* Declare attributes. */
1156 for (i = 0, map = reg_maps->input_registers; map; map >>= 1, ++i)
1159 shader_addline(buffer, "attribute vec4 %s_in%u;\n", prefix, i);
1162 shader_addline(buffer, "uniform vec4 posFixup;\n");
1163 shader_addline(buffer, "void order_ps_input(in vec4[%u]);\n", shader->limits.packed_output);
1165 else if (version->type == WINED3D_SHADER_TYPE_GEOMETRY)
1167 shader_addline(buffer, "varying in vec4 gs_in[][%u];\n", shader->limits.packed_input);
1169 else if (version->type == WINED3D_SHADER_TYPE_PIXEL)
1171 if (version->major >= 3)
1173 UINT in_count = min(vec4_varyings(version->major, gl_info), shader->limits.packed_input);
1176 shader_addline(buffer, "varying vec4 %s_in[%u];\n", prefix, in_count);
1178 /* TODO: Write a replacement shader for the fixed function
1179 * vertex pipeline, so this isn't needed. For fixed function
1180 * vertex processing + 3.0 pixel shader we need a separate
1181 * function in the pixel shader that reads the fixed function
1182 * color into the packed input registers. */
1183 shader_addline(buffer, "vec4 %s_in[%u];\n", prefix, in_count);
1186 for (i = 0, map = reg_maps->bumpmat; map; map >>= 1, ++i)
1191 shader_addline(buffer, "uniform mat2 bumpenv_mat%u;\n", i);
1193 if (reg_maps->luminanceparams & (1 << i))
1195 shader_addline(buffer, "uniform float bumpenv_lum_scale%u;\n", i);
1196 shader_addline(buffer, "uniform float bumpenv_lum_offset%u;\n", i);
1197 extra_constants_needed++;
1200 extra_constants_needed++;
1203 if (ps_args->srgb_correction)
1205 shader_addline(buffer, "const vec4 srgb_const0 = vec4(%.8e, %.8e, %.8e, %.8e);\n",
1206 srgb_pow, srgb_mul_high, srgb_sub_high, srgb_mul_low);
1207 shader_addline(buffer, "const vec4 srgb_const1 = vec4(%.8e, 0.0, 0.0, 0.0);\n",
1210 if (reg_maps->vpos || reg_maps->usesdsy)
1212 if (shader->limits.constant_float + extra_constants_needed
1213 + 1 < gl_info->limits.glsl_ps_float_constants)
1215 shader_addline(buffer, "uniform vec4 ycorrection;\n");
1216 extra_constants_needed++;
1220 /* This happens because we do not have proper tracking of the constant registers that are
1221 * actually used, only the max limit of the shader version
1223 FIXME("Cannot find a free uniform for vpos correction params\n");
1224 shader_addline(buffer, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
1225 context->render_offscreen ? 0.0f : fb->render_targets[0]->resource.height,
1226 context->render_offscreen ? 1.0f : -1.0f);
1228 shader_addline(buffer, "vec4 vpos;\n");
1232 /* Declare output register temporaries */
1233 if (shader->limits.packed_output)
1234 shader_addline(buffer, "vec4 %s_out[%u];\n", prefix, shader->limits.packed_output);
1236 /* Declare temporary variables */
1237 for (i = 0, map = reg_maps->temporary; map; map >>= 1, ++i)
1239 if (map & 1) shader_addline(buffer, "vec4 R%u;\n", i);
1242 /* Declare loop registers aLx */
1243 if (version->major < 4)
1245 for (i = 0; i < reg_maps->loop_depth; ++i)
1247 shader_addline(buffer, "int aL%u;\n", i);
1248 shader_addline(buffer, "int tmpInt%u;\n", i);
1252 /* Temporary variables for matrix operations */
1253 shader_addline(buffer, "vec4 tmp0;\n");
1254 shader_addline(buffer, "vec4 tmp1;\n");
1256 if (!shader->load_local_constsF)
1258 LIST_FOR_EACH_ENTRY(lconst, &shader->constantsF, struct wined3d_shader_lconst, entry)
1261 value = (const float *)lconst->value;
1262 shader_addline(buffer, "const vec4 %s_lc%u = vec4(%.8e, %.8e, %.8e, %.8e);\n",
1263 prefix, lconst->idx, value[0], value[1], value[2], value[3]);
1267 /* Start the main program. */
1268 shader_addline(buffer, "void main()\n{\n");
1270 /* Direct3D applications expect integer vPos values, while OpenGL drivers
1271 * add approximately 0.5. This causes off-by-one problems as spotted by
1272 * the vPos d3d9 visual test. Unfortunately ATI cards do not add exactly
1273 * 0.5, but rather something like 0.49999999 or 0.50000001, which still
1274 * causes precision troubles when we just subtract 0.5.
1276 * To deal with that, just floor() the position. This will eliminate the
1277 * fraction on all cards.
1279 * TODO: Test how this behaves with multisampling.
1281 * An advantage of floor is that it works even if the driver doesn't add
1282 * 0.5. It is somewhat questionable if 1.5, 2.5, ... are the proper values
1283 * to return in gl_FragCoord, even though coordinates specify the pixel
1284 * centers instead of the pixel corners. This code will behave correctly
1285 * on drivers that returns integer values. */
1286 if (version->type == WINED3D_SHADER_TYPE_PIXEL && reg_maps->vpos)
1287 shader_addline(buffer,
1288 "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
1291 /*****************************************************************************
1292 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
1294 * For more information, see http://wiki.winehq.org/DirectX-Shaders
1295 ****************************************************************************/
1298 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction *ins,
1299 const struct wined3d_shader_src_param *wined3d_src, DWORD mask, struct glsl_src_param *glsl_src);
1301 /** Used for opcode modifiers - They multiply the result by the specified amount */
1302 static const char * const shift_glsl_tab[] = {
1304 "2.0 * ", /* 1 (x2) */
1305 "4.0 * ", /* 2 (x4) */
1306 "8.0 * ", /* 3 (x8) */
1307 "16.0 * ", /* 4 (x16) */
1308 "32.0 * ", /* 5 (x32) */
1315 "0.0625 * ", /* 12 (d16) */
1316 "0.125 * ", /* 13 (d8) */
1317 "0.25 * ", /* 14 (d4) */
1318 "0.5 * " /* 15 (d2) */
1321 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
1322 static void shader_glsl_gen_modifier(enum wined3d_shader_src_modifier src_modifier,
1323 const char *in_reg, const char *in_regswizzle, char *out_str)
1327 switch (src_modifier)
1329 case WINED3DSPSM_DZ: /* Need to handle this in the instructions itself (texld & texcrd). */
1330 case WINED3DSPSM_DW:
1331 case WINED3DSPSM_NONE:
1332 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
1334 case WINED3DSPSM_NEG:
1335 sprintf(out_str, "-%s%s", in_reg, in_regswizzle);
1337 case WINED3DSPSM_NOT:
1338 sprintf(out_str, "!%s%s", in_reg, in_regswizzle);
1340 case WINED3DSPSM_BIAS:
1341 sprintf(out_str, "(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
1343 case WINED3DSPSM_BIASNEG:
1344 sprintf(out_str, "-(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
1346 case WINED3DSPSM_SIGN:
1347 sprintf(out_str, "(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
1349 case WINED3DSPSM_SIGNNEG:
1350 sprintf(out_str, "-(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
1352 case WINED3DSPSM_COMP:
1353 sprintf(out_str, "(1.0 - %s%s)", in_reg, in_regswizzle);
1355 case WINED3DSPSM_X2:
1356 sprintf(out_str, "(2.0 * %s%s)", in_reg, in_regswizzle);
1358 case WINED3DSPSM_X2NEG:
1359 sprintf(out_str, "-(2.0 * %s%s)", in_reg, in_regswizzle);
1361 case WINED3DSPSM_ABS:
1362 sprintf(out_str, "abs(%s%s)", in_reg, in_regswizzle);
1364 case WINED3DSPSM_ABSNEG:
1365 sprintf(out_str, "-abs(%s%s)", in_reg, in_regswizzle);
1368 FIXME("Unhandled modifier %u\n", src_modifier);
1369 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
1373 /** Writes the GLSL variable name that corresponds to the register that the
1374 * DX opcode parameter is trying to access */
1375 static void shader_glsl_get_register_name(const struct wined3d_shader_register *reg,
1376 char *register_name, BOOL *is_color, const struct wined3d_shader_instruction *ins)
1378 /* oPos, oFog and oPts in D3D */
1379 static const char * const hwrastout_reg_names[] = {"vs_out[10]", "vs_out[11].x", "vs_out[11].y"};
1381 const struct wined3d_shader *shader = ins->ctx->shader;
1382 const struct wined3d_shader_reg_maps *reg_maps = ins->ctx->reg_maps;
1383 const struct wined3d_shader_version *version = ®_maps->shader_version;
1384 const struct wined3d_gl_info *gl_info = ins->ctx->gl_info;
1385 const char *prefix = shader_glsl_get_prefix(version->type);
1386 struct glsl_src_param rel_param0, rel_param1;
1388 if (reg->idx[0].offset != ~0U && reg->idx[0].rel_addr)
1389 shader_glsl_add_src_param(ins, reg->idx[0].rel_addr, WINED3DSP_WRITEMASK_0, &rel_param0);
1390 if (reg->idx[1].offset != ~0U && reg->idx[1].rel_addr)
1391 shader_glsl_add_src_param(ins, reg->idx[1].rel_addr, WINED3DSP_WRITEMASK_0, &rel_param1);
1396 case WINED3DSPR_TEMP:
1397 sprintf(register_name, "R%u", reg->idx[0].offset);
1400 case WINED3DSPR_INPUT:
1401 /* vertex shaders */
1402 if (version->type == WINED3D_SHADER_TYPE_VERTEX)
1404 struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
1405 if (priv->cur_vs_args->swizzle_map & (1 << reg->idx[0].offset))
1407 sprintf(register_name, "%s_in%u", prefix, reg->idx[0].offset);
1411 if (version->type == WINED3D_SHADER_TYPE_GEOMETRY)
1413 if (reg->idx[0].rel_addr)
1415 if (reg->idx[1].rel_addr)
1416 sprintf(register_name, "gs_in[%s + %u][%s + %u]",
1417 rel_param0.param_str, reg->idx[0].offset, rel_param1.param_str, reg->idx[1].offset);
1419 sprintf(register_name, "gs_in[%s + %u][%u]",
1420 rel_param0.param_str, reg->idx[0].offset, reg->idx[1].offset);
1422 else if (reg->idx[1].rel_addr)
1423 sprintf(register_name, "gs_in[%u][%s + %u]",
1424 reg->idx[0].offset, rel_param1.param_str, reg->idx[1].offset);
1426 sprintf(register_name, "gs_in[%u][%u]", reg->idx[0].offset, reg->idx[1].offset);
1430 /* pixel shaders >= 3.0 */
1431 if (version->major >= 3)
1433 DWORD idx = shader->u.ps.input_reg_map[reg->idx[0].offset];
1434 unsigned int in_count = vec4_varyings(version->major, gl_info);
1436 if (reg->idx[0].rel_addr)
1438 /* Removing a + 0 would be an obvious optimization, but
1439 * OS X doesn't see the NOP operation there. */
1442 if (shader->u.ps.declared_in_count > in_count)
1444 sprintf(register_name,
1445 "((%s + %u) > %u ? (%s + %u) > %u ? gl_SecondaryColor : gl_Color : %s_in[%s + %u])",
1446 rel_param0.param_str, idx, in_count - 1, rel_param0.param_str, idx, in_count,
1447 prefix, rel_param0.param_str, idx);
1451 sprintf(register_name, "%s_in[%s + %u]", prefix, rel_param0.param_str, idx);
1456 if (shader->u.ps.declared_in_count > in_count)
1458 sprintf(register_name, "((%s) > %u ? (%s) > %u ? gl_SecondaryColor : gl_Color : %s_in[%s])",
1459 rel_param0.param_str, in_count - 1, rel_param0.param_str, in_count,
1460 prefix, rel_param0.param_str);
1464 sprintf(register_name, "%s_in[%s]", prefix, rel_param0.param_str);
1470 if (idx == in_count) sprintf(register_name, "gl_Color");
1471 else if (idx == in_count + 1) sprintf(register_name, "gl_SecondaryColor");
1472 else sprintf(register_name, "%s_in[%u]", prefix, idx);
1477 if (!reg->idx[0].offset)
1478 strcpy(register_name, "gl_Color");
1480 strcpy(register_name, "gl_SecondaryColor");
1485 case WINED3DSPR_CONST:
1487 /* Relative addressing */
1488 if (reg->idx[0].rel_addr)
1490 if (reg->idx[0].offset)
1491 sprintf(register_name, "%s_c[%s + %u]", prefix, rel_param0.param_str, reg->idx[0].offset);
1493 sprintf(register_name, "%s_c[%s]", prefix, rel_param0.param_str);
1497 if (shader_constant_is_local(shader, reg->idx[0].offset))
1498 sprintf(register_name, "%s_lc%u", prefix, reg->idx[0].offset);
1500 sprintf(register_name, "%s_c[%u]", prefix, reg->idx[0].offset);
1505 case WINED3DSPR_CONSTINT:
1506 sprintf(register_name, "%s_i[%u]", prefix, reg->idx[0].offset);
1509 case WINED3DSPR_CONSTBOOL:
1510 sprintf(register_name, "%s_b[%u]", prefix, reg->idx[0].offset);
1513 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
1514 if (version->type == WINED3D_SHADER_TYPE_PIXEL)
1515 sprintf(register_name, "T%u", reg->idx[0].offset);
1517 sprintf(register_name, "A%u", reg->idx[0].offset);
1520 case WINED3DSPR_LOOP:
1521 sprintf(register_name, "aL%u", ins->ctx->loop_state->current_reg - 1);
1524 case WINED3DSPR_SAMPLER:
1525 sprintf(register_name, "%s_sampler%u", prefix, reg->idx[0].offset);
1528 case WINED3DSPR_COLOROUT:
1529 if (reg->idx[0].offset >= gl_info->limits.buffers)
1530 WARN("Write to render target %u, only %d supported.\n",
1531 reg->idx[0].offset, gl_info->limits.buffers);
1533 sprintf(register_name, "gl_FragData[%u]", reg->idx[0].offset);
1536 case WINED3DSPR_RASTOUT:
1537 sprintf(register_name, "%s", hwrastout_reg_names[reg->idx[0].offset]);
1540 case WINED3DSPR_DEPTHOUT:
1541 sprintf(register_name, "gl_FragDepth");
1544 case WINED3DSPR_ATTROUT:
1545 if (!reg->idx[0].offset)
1546 sprintf(register_name, "%s_out[8]", prefix);
1548 sprintf(register_name, "%s_out[9]", prefix);
1551 case WINED3DSPR_TEXCRDOUT:
1552 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1553 sprintf(register_name, "%s_out[%u]", prefix, reg->idx[0].offset);
1556 case WINED3DSPR_MISCTYPE:
1557 if (!reg->idx[0].offset)
1560 sprintf(register_name, "vpos");
1562 else if (reg->idx[0].offset == 1)
1564 /* Note that gl_FrontFacing is a bool, while vFace is
1565 * a float for which the sign determines front/back */
1566 sprintf(register_name, "(gl_FrontFacing ? 1.0 : -1.0)");
1570 FIXME("Unhandled misctype register %u.\n", reg->idx[0].offset);
1571 sprintf(register_name, "unrecognized_register");
1575 case WINED3DSPR_IMMCONST:
1576 switch (reg->immconst_type)
1578 case WINED3D_IMMCONST_SCALAR:
1579 switch (reg->data_type)
1581 case WINED3D_DATA_FLOAT:
1582 sprintf(register_name, "%.8e", *(const float *)reg->immconst_data);
1584 case WINED3D_DATA_INT:
1585 sprintf(register_name, "%#x", reg->immconst_data[0]);
1587 case WINED3D_DATA_RESOURCE:
1588 case WINED3D_DATA_SAMPLER:
1589 case WINED3D_DATA_UINT:
1590 sprintf(register_name, "%#xu", reg->immconst_data[0]);
1593 sprintf(register_name, "<unhandled data type %#x>", reg->data_type);
1598 case WINED3D_IMMCONST_VEC4:
1599 switch (reg->data_type)
1601 case WINED3D_DATA_FLOAT:
1602 sprintf(register_name, "vec4(%.8e, %.8e, %.8e, %.8e)",
1603 *(const float *)®->immconst_data[0], *(const float *)®->immconst_data[1],
1604 *(const float *)®->immconst_data[2], *(const float *)®->immconst_data[3]);
1606 case WINED3D_DATA_INT:
1607 sprintf(register_name, "ivec4(%#x, %#x, %#x, %#x)",
1608 reg->immconst_data[0], reg->immconst_data[1],
1609 reg->immconst_data[2], reg->immconst_data[3]);
1611 case WINED3D_DATA_RESOURCE:
1612 case WINED3D_DATA_SAMPLER:
1613 case WINED3D_DATA_UINT:
1614 sprintf(register_name, "uvec4(%#xu, %#xu, %#xu, %#xu)",
1615 reg->immconst_data[0], reg->immconst_data[1],
1616 reg->immconst_data[2], reg->immconst_data[3]);
1619 sprintf(register_name, "<unhandled data type %#x>", reg->data_type);
1625 FIXME("Unhandled immconst type %#x\n", reg->immconst_type);
1626 sprintf(register_name, "<unhandled_immconst_type %#x>", reg->immconst_type);
1630 case WINED3DSPR_CONSTBUFFER:
1631 if (reg->idx[1].rel_addr)
1632 sprintf(register_name, "%s_cb%u[%s + %u]",
1633 prefix, reg->idx[0].offset, rel_param1.param_str, reg->idx[1].offset);
1635 sprintf(register_name, "%s_cb%u[%u]", prefix, reg->idx[0].offset, reg->idx[1].offset);
1638 case WINED3DSPR_PRIMID:
1639 sprintf(register_name, "uint(gl_PrimitiveIDIn)");
1643 FIXME("Unhandled register type %#x.\n", reg->type);
1644 sprintf(register_name, "unrecognized_register");
1649 static void shader_glsl_write_mask_to_str(DWORD write_mask, char *str)
1652 if (write_mask & WINED3DSP_WRITEMASK_0) *str++ = 'x';
1653 if (write_mask & WINED3DSP_WRITEMASK_1) *str++ = 'y';
1654 if (write_mask & WINED3DSP_WRITEMASK_2) *str++ = 'z';
1655 if (write_mask & WINED3DSP_WRITEMASK_3) *str++ = 'w';
1659 /* Get the GLSL write mask for the destination register */
1660 static DWORD shader_glsl_get_write_mask(const struct wined3d_shader_dst_param *param, char *write_mask)
1662 DWORD mask = param->write_mask;
1664 if (shader_is_scalar(¶m->reg))
1666 mask = WINED3DSP_WRITEMASK_0;
1671 shader_glsl_write_mask_to_str(mask, write_mask);
1677 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask) {
1678 unsigned int size = 0;
1680 if (write_mask & WINED3DSP_WRITEMASK_0) ++size;
1681 if (write_mask & WINED3DSP_WRITEMASK_1) ++size;
1682 if (write_mask & WINED3DSP_WRITEMASK_2) ++size;
1683 if (write_mask & WINED3DSP_WRITEMASK_3) ++size;
1688 static void shader_glsl_swizzle_to_str(const DWORD swizzle, BOOL fixup, DWORD mask, char *str)
1690 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1691 * but addressed as "rgba". To fix this we need to swap the register's x
1692 * and z components. */
1693 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
1696 /* swizzle bits fields: wwzzyyxx */
1697 if (mask & WINED3DSP_WRITEMASK_0) *str++ = swizzle_chars[swizzle & 0x03];
1698 if (mask & WINED3DSP_WRITEMASK_1) *str++ = swizzle_chars[(swizzle >> 2) & 0x03];
1699 if (mask & WINED3DSP_WRITEMASK_2) *str++ = swizzle_chars[(swizzle >> 4) & 0x03];
1700 if (mask & WINED3DSP_WRITEMASK_3) *str++ = swizzle_chars[(swizzle >> 6) & 0x03];
1704 static void shader_glsl_get_swizzle(const struct wined3d_shader_src_param *param,
1705 BOOL fixup, DWORD mask, char *swizzle_str)
1707 if (shader_is_scalar(¶m->reg))
1708 *swizzle_str = '\0';
1710 shader_glsl_swizzle_to_str(param->swizzle, fixup, mask, swizzle_str);
1713 /* From a given parameter token, generate the corresponding GLSL string.
1714 * Also, return the actual register name and swizzle in case the
1715 * caller needs this information as well. */
1716 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction *ins,
1717 const struct wined3d_shader_src_param *wined3d_src, DWORD mask, struct glsl_src_param *glsl_src)
1719 BOOL is_color = FALSE;
1720 char swizzle_str[6];
1722 glsl_src->reg_name[0] = '\0';
1723 glsl_src->param_str[0] = '\0';
1724 swizzle_str[0] = '\0';
1726 shader_glsl_get_register_name(&wined3d_src->reg, glsl_src->reg_name, &is_color, ins);
1727 shader_glsl_get_swizzle(wined3d_src, is_color, mask, swizzle_str);
1729 if (wined3d_src->reg.type == WINED3DSPR_IMMCONST || wined3d_src->reg.type == WINED3DSPR_PRIMID)
1731 shader_glsl_gen_modifier(wined3d_src->modifiers, glsl_src->reg_name, swizzle_str, glsl_src->param_str);
1735 char param_str[200];
1737 shader_glsl_gen_modifier(wined3d_src->modifiers, glsl_src->reg_name, swizzle_str, param_str);
1739 switch (wined3d_src->reg.data_type)
1741 case WINED3D_DATA_FLOAT:
1742 sprintf(glsl_src->param_str, "%s", param_str);
1744 case WINED3D_DATA_INT:
1745 sprintf(glsl_src->param_str, "floatBitsToInt(%s)", param_str);
1747 case WINED3D_DATA_RESOURCE:
1748 case WINED3D_DATA_SAMPLER:
1749 case WINED3D_DATA_UINT:
1750 sprintf(glsl_src->param_str, "floatBitsToUint(%s)", param_str);
1753 FIXME("Unhandled data type %#x.\n", wined3d_src->reg.data_type);
1754 sprintf(glsl_src->param_str, "%s", param_str);
1760 /* From a given parameter token, generate the corresponding GLSL string.
1761 * Also, return the actual register name and swizzle in case the
1762 * caller needs this information as well. */
1763 static DWORD shader_glsl_add_dst_param(const struct wined3d_shader_instruction *ins,
1764 const struct wined3d_shader_dst_param *wined3d_dst, struct glsl_dst_param *glsl_dst)
1766 BOOL is_color = FALSE;
1768 glsl_dst->mask_str[0] = '\0';
1769 glsl_dst->reg_name[0] = '\0';
1771 shader_glsl_get_register_name(&wined3d_dst->reg, glsl_dst->reg_name, &is_color, ins);
1772 return shader_glsl_get_write_mask(wined3d_dst, glsl_dst->mask_str);
1775 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1776 static DWORD shader_glsl_append_dst_ext(struct wined3d_shader_buffer *buffer,
1777 const struct wined3d_shader_instruction *ins, const struct wined3d_shader_dst_param *dst)
1779 struct glsl_dst_param glsl_dst;
1782 if ((mask = shader_glsl_add_dst_param(ins, dst, &glsl_dst)))
1784 switch (dst->reg.data_type)
1786 case WINED3D_DATA_FLOAT:
1787 shader_addline(buffer, "%s%s = %s(",
1788 glsl_dst.reg_name, glsl_dst.mask_str, shift_glsl_tab[dst->shift]);
1790 case WINED3D_DATA_INT:
1791 shader_addline(buffer, "%s%s = %sintBitsToFloat(",
1792 glsl_dst.reg_name, glsl_dst.mask_str, shift_glsl_tab[dst->shift]);
1794 case WINED3D_DATA_RESOURCE:
1795 case WINED3D_DATA_SAMPLER:
1796 case WINED3D_DATA_UINT:
1797 shader_addline(buffer, "%s%s = %suintBitsToFloat(",
1798 glsl_dst.reg_name, glsl_dst.mask_str, shift_glsl_tab[dst->shift]);
1801 FIXME("Unhandled data type %#x.\n", dst->reg.data_type);
1802 shader_addline(buffer, "%s%s = %s(",
1803 glsl_dst.reg_name, glsl_dst.mask_str, shift_glsl_tab[dst->shift]);
1811 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1812 static DWORD shader_glsl_append_dst(struct wined3d_shader_buffer *buffer, const struct wined3d_shader_instruction *ins)
1814 return shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0]);
1817 /** Process GLSL instruction modifiers */
1818 static void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction *ins)
1820 struct glsl_dst_param dst_param;
1823 if (!ins->dst_count) return;
1825 modifiers = ins->dst[0].modifiers;
1826 if (!modifiers) return;
1828 shader_glsl_add_dst_param(ins, &ins->dst[0], &dst_param);
1830 if (modifiers & WINED3DSPDM_SATURATE)
1832 /* _SAT means to clamp the value of the register to between 0 and 1 */
1833 shader_addline(ins->ctx->buffer, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param.reg_name,
1834 dst_param.mask_str, dst_param.reg_name, dst_param.mask_str);
1837 if (modifiers & WINED3DSPDM_MSAMPCENTROID)
1839 FIXME("_centroid modifier not handled\n");
1842 if (modifiers & WINED3DSPDM_PARTIALPRECISION)
1844 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1848 static const char *shader_glsl_get_rel_op(enum wined3d_shader_rel_op op)
1852 case WINED3D_SHADER_REL_OP_GT: return ">";
1853 case WINED3D_SHADER_REL_OP_EQ: return "==";
1854 case WINED3D_SHADER_REL_OP_GE: return ">=";
1855 case WINED3D_SHADER_REL_OP_LT: return "<";
1856 case WINED3D_SHADER_REL_OP_NE: return "!=";
1857 case WINED3D_SHADER_REL_OP_LE: return "<=";
1859 FIXME("Unrecognized operator %#x.\n", op);
1864 static void shader_glsl_get_sample_function(const struct wined3d_shader_context *ctx,
1865 DWORD sampler_idx, DWORD flags, struct glsl_sample_function *sample_function)
1867 enum wined3d_sampler_texture_type sampler_type = ctx->reg_maps->sampler_type[sampler_idx];
1868 const struct wined3d_gl_info *gl_info = ctx->gl_info;
1869 BOOL shadow = ctx->reg_maps->shader_version.type == WINED3D_SHADER_TYPE_PIXEL
1870 && (((const struct shader_glsl_ctx_priv *)ctx->backend_data)->cur_ps_args->shadow & (1 << sampler_idx));
1871 BOOL projected = flags & WINED3D_GLSL_SAMPLE_PROJECTED;
1872 BOOL texrect = flags & WINED3D_GLSL_SAMPLE_RECT;
1873 BOOL lod = flags & WINED3D_GLSL_SAMPLE_LOD;
1874 BOOL grad = flags & WINED3D_GLSL_SAMPLE_GRAD;
1876 /* Note that there's no such thing as a projected cube texture. */
1877 switch(sampler_type) {
1883 sample_function->name = projected ? "shadow1DProjLod" : "shadow1DLod";
1887 if (gl_info->supported[EXT_GPU_SHADER4])
1888 sample_function->name = projected ? "shadow1DProjGrad" : "shadow1DGrad";
1889 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
1890 sample_function->name = projected ? "shadow1DProjGradARB" : "shadow1DGradARB";
1893 FIXME("Unsupported 1D shadow grad function.\n");
1894 sample_function->name = "unsupported1DGrad";
1899 sample_function->name = projected ? "shadow1DProj" : "shadow1D";
1901 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1;
1907 sample_function->name = projected ? "texture1DProjLod" : "texture1DLod";
1911 if (gl_info->supported[EXT_GPU_SHADER4])
1912 sample_function->name = projected ? "texture1DProjGrad" : "texture1DGrad";
1913 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
1914 sample_function->name = projected ? "texture1DProjGradARB" : "texture1DGradARB";
1917 FIXME("Unsupported 1D grad function.\n");
1918 sample_function->name = "unsupported1DGrad";
1923 sample_function->name = projected ? "texture1DProj" : "texture1D";
1925 sample_function->coord_mask = WINED3DSP_WRITEMASK_0;
1936 sample_function->name = projected ? "shadow2DRectProjLod" : "shadow2DRectLod";
1940 if (gl_info->supported[EXT_GPU_SHADER4])
1941 sample_function->name = projected ? "shadow2DRectProjGrad" : "shadow2DRectGrad";
1942 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
1943 sample_function->name = projected ? "shadow2DRectProjGradARB" : "shadow2DRectGradARB";
1946 FIXME("Unsupported RECT shadow grad function.\n");
1947 sample_function->name = "unsupported2DRectGrad";
1952 sample_function->name = projected ? "shadow2DRectProj" : "shadow2DRect";
1959 sample_function->name = projected ? "shadow2DProjLod" : "shadow2DLod";
1963 if (gl_info->supported[EXT_GPU_SHADER4])
1964 sample_function->name = projected ? "shadow2DProjGrad" : "shadow2DGrad";
1965 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
1966 sample_function->name = projected ? "shadow2DProjGradARB" : "shadow2DGradARB";
1969 FIXME("Unsupported 2D shadow grad function.\n");
1970 sample_function->name = "unsupported2DGrad";
1975 sample_function->name = projected ? "shadow2DProj" : "shadow2D";
1978 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1986 sample_function->name = projected ? "texture2DRectProjLod" : "texture2DRectLod";
1990 if (gl_info->supported[EXT_GPU_SHADER4])
1991 sample_function->name = projected ? "texture2DRectProjGrad" : "texture2DRectGrad";
1992 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
1993 sample_function->name = projected ? "texture2DRectProjGradARB" : "texture2DRectGradARB";
1996 FIXME("Unsupported RECT grad function.\n");
1997 sample_function->name = "unsupported2DRectGrad";
2002 sample_function->name = projected ? "texture2DRectProj" : "texture2DRect";
2009 sample_function->name = projected ? "texture2DProjLod" : "texture2DLod";
2013 if (gl_info->supported[EXT_GPU_SHADER4])
2014 sample_function->name = projected ? "texture2DProjGrad" : "texture2DGrad";
2015 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
2016 sample_function->name = projected ? "texture2DProjGradARB" : "texture2DGradARB";
2019 FIXME("Unsupported 2D grad function.\n");
2020 sample_function->name = "unsupported2DGrad";
2025 sample_function->name = projected ? "texture2DProj" : "texture2D";
2028 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1;
2032 case WINED3DSTT_CUBE:
2035 FIXME("Unsupported Cube shadow function.\n");
2036 sample_function->name = "unsupportedCubeShadow";
2037 sample_function->coord_mask = 0;
2043 sample_function->name = "textureCubeLod";
2047 if (gl_info->supported[EXT_GPU_SHADER4])
2048 sample_function->name = "textureCubeGrad";
2049 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
2050 sample_function->name = "textureCubeGradARB";
2053 FIXME("Unsupported Cube grad function.\n");
2054 sample_function->name = "unsupportedCubeGrad";
2059 sample_function->name = "textureCube";
2061 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2065 case WINED3DSTT_VOLUME:
2068 FIXME("Unsupported 3D shadow function.\n");
2069 sample_function->name = "unsupported3DShadow";
2070 sample_function->coord_mask = 0;
2076 sample_function->name = projected ? "texture3DProjLod" : "texture3DLod";
2080 if (gl_info->supported[EXT_GPU_SHADER4])
2081 sample_function->name = projected ? "texture3DProjGrad" : "texture3DGrad";
2082 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
2083 sample_function->name = projected ? "texture3DProjGradARB" : "texture3DGradARB";
2086 FIXME("Unsupported 3D grad function.\n");
2087 sample_function->name = "unsupported3DGrad";
2092 sample_function->name = projected ? "texture3DProj" : "texture3D";
2094 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2099 sample_function->name = "";
2100 sample_function->coord_mask = 0;
2101 FIXME("Unrecognized sampler type: %#x;\n", sampler_type);
2106 static void shader_glsl_append_fixup_arg(char *arguments, const char *reg_name,
2107 BOOL sign_fixup, enum fixup_channel_source channel_source)
2109 switch(channel_source)
2111 case CHANNEL_SOURCE_ZERO:
2112 strcat(arguments, "0.0");
2115 case CHANNEL_SOURCE_ONE:
2116 strcat(arguments, "1.0");
2119 case CHANNEL_SOURCE_X:
2120 strcat(arguments, reg_name);
2121 strcat(arguments, ".x");
2124 case CHANNEL_SOURCE_Y:
2125 strcat(arguments, reg_name);
2126 strcat(arguments, ".y");
2129 case CHANNEL_SOURCE_Z:
2130 strcat(arguments, reg_name);
2131 strcat(arguments, ".z");
2134 case CHANNEL_SOURCE_W:
2135 strcat(arguments, reg_name);
2136 strcat(arguments, ".w");
2140 FIXME("Unhandled channel source %#x\n", channel_source);
2141 strcat(arguments, "undefined");
2145 if (sign_fixup) strcat(arguments, " * 2.0 - 1.0");
2148 static void shader_glsl_color_correction_ext(struct wined3d_shader_buffer *buffer,
2149 const char *reg_name, DWORD mask, struct color_fixup_desc fixup)
2151 unsigned int mask_size, remaining;
2152 DWORD fixup_mask = 0;
2153 char arguments[256];
2156 if (fixup.x_sign_fixup || fixup.x_source != CHANNEL_SOURCE_X) fixup_mask |= WINED3DSP_WRITEMASK_0;
2157 if (fixup.y_sign_fixup || fixup.y_source != CHANNEL_SOURCE_Y) fixup_mask |= WINED3DSP_WRITEMASK_1;
2158 if (fixup.z_sign_fixup || fixup.z_source != CHANNEL_SOURCE_Z) fixup_mask |= WINED3DSP_WRITEMASK_2;
2159 if (fixup.w_sign_fixup || fixup.w_source != CHANNEL_SOURCE_W) fixup_mask |= WINED3DSP_WRITEMASK_3;
2160 if (!(mask &= fixup_mask))
2163 if (is_complex_fixup(fixup))
2165 enum complex_fixup complex_fixup = get_complex_fixup(fixup);
2166 FIXME("Complex fixup (%#x) not supported\n",complex_fixup);
2170 shader_glsl_write_mask_to_str(mask, mask_str);
2171 mask_size = shader_glsl_get_write_mask_size(mask);
2173 arguments[0] = '\0';
2174 remaining = mask_size;
2175 if (mask & WINED3DSP_WRITEMASK_0)
2177 shader_glsl_append_fixup_arg(arguments, reg_name, fixup.x_sign_fixup, fixup.x_source);
2178 if (--remaining) strcat(arguments, ", ");
2180 if (mask & WINED3DSP_WRITEMASK_1)
2182 shader_glsl_append_fixup_arg(arguments, reg_name, fixup.y_sign_fixup, fixup.y_source);
2183 if (--remaining) strcat(arguments, ", ");
2185 if (mask & WINED3DSP_WRITEMASK_2)
2187 shader_glsl_append_fixup_arg(arguments, reg_name, fixup.z_sign_fixup, fixup.z_source);
2188 if (--remaining) strcat(arguments, ", ");
2190 if (mask & WINED3DSP_WRITEMASK_3)
2192 shader_glsl_append_fixup_arg(arguments, reg_name, fixup.w_sign_fixup, fixup.w_source);
2193 if (--remaining) strcat(arguments, ", ");
2197 shader_addline(buffer, "%s%s = vec%u(%s);\n", reg_name, mask_str, mask_size, arguments);
2199 shader_addline(buffer, "%s%s = %s;\n", reg_name, mask_str, arguments);
2202 static void shader_glsl_color_correction(const struct wined3d_shader_instruction *ins, struct color_fixup_desc fixup)
2207 shader_glsl_get_register_name(&ins->dst[0].reg, reg_name, &is_color, ins);
2208 shader_glsl_color_correction_ext(ins->ctx->buffer, reg_name, ins->dst[0].write_mask, fixup);
2211 static void PRINTF_ATTR(8, 9) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction *ins,
2212 DWORD sampler, const struct glsl_sample_function *sample_function, DWORD swizzle,
2213 const char *dx, const char *dy, const char *bias, const char *coord_reg_fmt, ...)
2215 const struct wined3d_shader_version *version = &ins->ctx->reg_maps->shader_version;
2216 char dst_swizzle[6];
2217 struct color_fixup_desc fixup;
2218 BOOL np2_fixup = FALSE;
2221 shader_glsl_swizzle_to_str(swizzle, FALSE, ins->dst[0].write_mask, dst_swizzle);
2223 if (version->type == WINED3D_SHADER_TYPE_PIXEL)
2225 const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
2226 fixup = priv->cur_ps_args->color_fixup[sampler];
2228 if(priv->cur_ps_args->np2_fixup & (1 << sampler)) {
2230 FIXME("Biased sampling from NP2 textures is unsupported\n");
2238 fixup = COLOR_FIXUP_IDENTITY; /* FIXME: Vshader color fixup */
2241 shader_glsl_append_dst(ins->ctx->buffer, ins);
2243 shader_addline(ins->ctx->buffer, "%s(%s_sampler%u, ",
2244 sample_function->name, shader_glsl_get_prefix(version->type), sampler);
2246 va_start(args, coord_reg_fmt);
2247 shader_vaddline(ins->ctx->buffer, coord_reg_fmt, args);
2251 shader_addline(ins->ctx->buffer, ", %s)%s);\n", bias, dst_swizzle);
2254 const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
2255 const unsigned char idx = priv->cur_np2fixup_info->idx[sampler];
2257 shader_addline(ins->ctx->buffer, " * ps_samplerNP2Fixup[%u].%s)%s);\n", idx >> 1,
2258 (idx % 2) ? "zw" : "xy", dst_swizzle);
2259 } else if(dx && dy) {
2260 shader_addline(ins->ctx->buffer, ", %s, %s)%s);\n", dx, dy, dst_swizzle);
2262 shader_addline(ins->ctx->buffer, ")%s);\n", dst_swizzle);
2266 if(!is_identity_fixup(fixup)) {
2267 shader_glsl_color_correction(ins, fixup);
2271 /*****************************************************************************
2272 * Begin processing individual instruction opcodes
2273 ****************************************************************************/
2275 static void shader_glsl_binop(const struct wined3d_shader_instruction *ins)
2277 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2278 struct glsl_src_param src0_param;
2279 struct glsl_src_param src1_param;
2283 /* Determine the GLSL operator to use based on the opcode */
2284 switch (ins->handler_idx)
2286 case WINED3DSIH_ADD: op = "+"; break;
2287 case WINED3DSIH_AND: op = "&"; break;
2288 case WINED3DSIH_DIV: op = "/"; break;
2289 case WINED3DSIH_IADD: op = "+"; break;
2290 case WINED3DSIH_MUL: op = "*"; break;
2291 case WINED3DSIH_SUB: op = "-"; break;
2292 case WINED3DSIH_USHR: op = ">>"; break;
2293 case WINED3DSIH_XOR: op = "^"; break;
2295 op = "<unhandled operator>";
2296 FIXME("Opcode %#x not yet handled in GLSL\n", ins->handler_idx);
2300 write_mask = shader_glsl_append_dst(buffer, ins);
2301 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2302 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2303 shader_addline(buffer, "%s %s %s);\n", src0_param.param_str, op, src1_param.param_str);
2306 static void shader_glsl_relop(const struct wined3d_shader_instruction *ins)
2308 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2309 struct glsl_src_param src0_param;
2310 struct glsl_src_param src1_param;
2311 unsigned int mask_size;
2315 write_mask = shader_glsl_append_dst(buffer, ins);
2316 mask_size = shader_glsl_get_write_mask_size(write_mask);
2317 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2318 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2322 switch (ins->handler_idx)
2324 case WINED3DSIH_EQ: op = "equal"; break;
2325 case WINED3DSIH_GE: op = "greaterThanEqual"; break;
2326 case WINED3DSIH_IGE: op = "greaterThanEqual"; break;
2327 case WINED3DSIH_LT: op = "lessThan"; break;
2329 op = "<unhandled operator>";
2330 ERR("Unhandled opcode %#x.\n", ins->handler_idx);
2334 shader_addline(buffer, "uvec%u(%s(%s, %s)) * 0xffffffffu);\n",
2335 mask_size, op, src0_param.param_str, src1_param.param_str);
2339 switch (ins->handler_idx)
2341 case WINED3DSIH_EQ: op = "=="; break;
2342 case WINED3DSIH_GE: op = ">="; break;
2343 case WINED3DSIH_IGE: op = ">="; break;
2344 case WINED3DSIH_LT: op = "<"; break;
2346 op = "<unhandled operator>";
2347 ERR("Unhandled opcode %#x.\n", ins->handler_idx);
2351 shader_addline(buffer, "%s %s %s ? 0xffffffffu : 0u);\n",
2352 src0_param.param_str, op, src1_param.param_str);
2356 static void shader_glsl_imul(const struct wined3d_shader_instruction *ins)
2358 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2359 struct glsl_src_param src0_param;
2360 struct glsl_src_param src1_param;
2363 /* If we have ARB_gpu_shader5 or GLSL 4.0, we can use imulExtended(). If
2364 * not, we can emulate it. */
2365 if (ins->dst[0].reg.type != WINED3DSPR_NULL)
2366 FIXME("64-bit integer multiplies not implemented.\n");
2368 if (ins->dst[1].reg.type != WINED3DSPR_NULL)
2370 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[1]);
2371 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2372 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2374 shader_addline(ins->ctx->buffer, "%s * %s);\n",
2375 src0_param.param_str, src1_param.param_str);
2379 static void shader_glsl_udiv(const struct wined3d_shader_instruction *ins)
2381 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2382 struct glsl_src_param src0_param, src1_param;
2385 if (ins->dst[0].reg.type != WINED3DSPR_NULL)
2388 if (ins->dst[1].reg.type != WINED3DSPR_NULL)
2392 write_mask = shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
2393 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2394 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2395 shader_addline(buffer, "tmp0%s = %s / %s;\n",
2396 dst_mask, src0_param.param_str, src1_param.param_str);
2398 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[1]);
2399 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2400 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2401 shader_addline(buffer, "%s %% %s));\n", src0_param.param_str, src1_param.param_str);
2403 shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0]);
2404 shader_addline(buffer, "tmp0%s);\n", dst_mask);
2408 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0]);
2409 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2410 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2411 shader_addline(buffer, "%s / %s);\n", src0_param.param_str, src1_param.param_str);
2414 else if (ins->dst[1].reg.type != WINED3DSPR_NULL)
2416 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[1]);
2417 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2418 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2419 shader_addline(buffer, "%s %% %s);\n", src0_param.param_str, src1_param.param_str);
2423 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
2424 static void shader_glsl_mov(const struct wined3d_shader_instruction *ins)
2426 const struct wined3d_gl_info *gl_info = ins->ctx->gl_info;
2427 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2428 struct glsl_src_param src0_param;
2431 write_mask = shader_glsl_append_dst(buffer, ins);
2432 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2434 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
2435 * shader versions WINED3DSIO_MOVA is used for this. */
2436 if (ins->ctx->reg_maps->shader_version.major == 1
2437 && ins->ctx->reg_maps->shader_version.type == WINED3D_SHADER_TYPE_VERTEX
2438 && ins->dst[0].reg.type == WINED3DSPR_ADDR)
2440 /* This is a simple floor() */
2441 unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
2442 if (mask_size > 1) {
2443 shader_addline(buffer, "ivec%d(floor(%s)));\n", mask_size, src0_param.param_str);
2445 shader_addline(buffer, "int(floor(%s)));\n", src0_param.param_str);
2448 else if(ins->handler_idx == WINED3DSIH_MOVA)
2450 /* We need to *round* to the nearest int here. */
2451 unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
2453 if (gl_info->supported[EXT_GPU_SHADER4])
2456 shader_addline(buffer, "ivec%d(round(%s)));\n", mask_size, src0_param.param_str);
2458 shader_addline(buffer, "int(round(%s)));\n", src0_param.param_str);
2463 shader_addline(buffer, "ivec%d(floor(abs(%s) + vec%d(0.5)) * sign(%s)));\n",
2464 mask_size, src0_param.param_str, mask_size, src0_param.param_str);
2466 shader_addline(buffer, "int(floor(abs(%s) + 0.5) * sign(%s)));\n",
2467 src0_param.param_str, src0_param.param_str);
2472 shader_addline(buffer, "%s);\n", src0_param.param_str);
2476 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
2477 static void shader_glsl_dot(const struct wined3d_shader_instruction *ins)
2479 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2480 struct glsl_src_param src0_param;
2481 struct glsl_src_param src1_param;
2482 DWORD dst_write_mask, src_write_mask;
2483 unsigned int dst_size = 0;
2485 dst_write_mask = shader_glsl_append_dst(buffer, ins);
2486 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
2488 /* dp3 works on vec3, dp4 on vec4 */
2489 if (ins->handler_idx == WINED3DSIH_DP4)
2491 src_write_mask = WINED3DSP_WRITEMASK_ALL;
2493 src_write_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2496 shader_glsl_add_src_param(ins, &ins->src[0], src_write_mask, &src0_param);
2497 shader_glsl_add_src_param(ins, &ins->src[1], src_write_mask, &src1_param);
2500 shader_addline(buffer, "vec%d(dot(%s, %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
2502 shader_addline(buffer, "dot(%s, %s));\n", src0_param.param_str, src1_param.param_str);
2506 /* Note that this instruction has some restrictions. The destination write mask
2507 * can't contain the w component, and the source swizzles have to be .xyzw */
2508 static void shader_glsl_cross(const struct wined3d_shader_instruction *ins)
2510 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2511 struct glsl_src_param src0_param;
2512 struct glsl_src_param src1_param;
2515 shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
2516 shader_glsl_append_dst(ins->ctx->buffer, ins);
2517 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
2518 shader_glsl_add_src_param(ins, &ins->src[1], src_mask, &src1_param);
2519 shader_addline(ins->ctx->buffer, "cross(%s, %s)%s);\n", src0_param.param_str, src1_param.param_str, dst_mask);
2522 static void shader_glsl_cut(const struct wined3d_shader_instruction *ins)
2524 shader_addline(ins->ctx->buffer, "EndPrimitive();\n");
2527 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
2528 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
2529 * GLSL uses the value as-is. */
2530 static void shader_glsl_pow(const struct wined3d_shader_instruction *ins)
2532 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2533 struct glsl_src_param src0_param;
2534 struct glsl_src_param src1_param;
2535 DWORD dst_write_mask;
2536 unsigned int dst_size;
2538 dst_write_mask = shader_glsl_append_dst(buffer, ins);
2539 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
2541 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
2542 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0, &src1_param);
2546 shader_addline(buffer, "vec%u(%s == 0.0 ? 1.0 : pow(abs(%s), %s)));\n",
2547 dst_size, src1_param.param_str, src0_param.param_str, src1_param.param_str);
2551 shader_addline(buffer, "%s == 0.0 ? 1.0 : pow(abs(%s), %s));\n",
2552 src1_param.param_str, src0_param.param_str, src1_param.param_str);
2556 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
2557 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
2558 * GLSL uses the value as-is. */
2559 static void shader_glsl_log(const struct wined3d_shader_instruction *ins)
2561 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2562 struct glsl_src_param src0_param;
2563 DWORD dst_write_mask;
2564 unsigned int dst_size;
2566 dst_write_mask = shader_glsl_append_dst(buffer, ins);
2567 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
2569 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
2573 shader_addline(buffer, "vec%u(log2(abs(%s))));\n",
2574 dst_size, src0_param.param_str);
2578 shader_addline(buffer, "log2(abs(%s)));\n",
2579 src0_param.param_str);
2583 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
2584 static void shader_glsl_map2gl(const struct wined3d_shader_instruction *ins)
2586 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2587 struct glsl_src_param src_param;
2588 const char *instruction;
2592 /* Determine the GLSL function to use based on the opcode */
2593 /* TODO: Possibly make this a table for faster lookups */
2594 switch (ins->handler_idx)
2596 case WINED3DSIH_MIN: instruction = "min"; break;
2597 case WINED3DSIH_MAX: instruction = "max"; break;
2598 case WINED3DSIH_ABS: instruction = "abs"; break;
2599 case WINED3DSIH_FRC: instruction = "fract"; break;
2600 case WINED3DSIH_EXP: instruction = "exp2"; break;
2601 case WINED3DSIH_DSX: instruction = "dFdx"; break;
2602 case WINED3DSIH_DSY: instruction = "ycorrection.y * dFdy"; break;
2603 case WINED3DSIH_ROUND_NI: instruction = "floor"; break;
2604 default: instruction = "";
2605 FIXME("Opcode %#x not yet handled in GLSL\n", ins->handler_idx);
2609 write_mask = shader_glsl_append_dst(buffer, ins);
2611 shader_addline(buffer, "%s(", instruction);
2615 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src_param);
2616 shader_addline(buffer, "%s", src_param.param_str);
2617 for (i = 1; i < ins->src_count; ++i)
2619 shader_glsl_add_src_param(ins, &ins->src[i], write_mask, &src_param);
2620 shader_addline(buffer, ", %s", src_param.param_str);
2624 shader_addline(buffer, "));\n");
2627 static void shader_glsl_nop(const struct wined3d_shader_instruction *ins) {}
2629 static void shader_glsl_nrm(const struct wined3d_shader_instruction *ins)
2631 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2632 struct glsl_src_param src_param;
2633 unsigned int mask_size;
2637 write_mask = shader_glsl_get_write_mask(ins->dst, dst_mask);
2638 mask_size = shader_glsl_get_write_mask_size(write_mask);
2639 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src_param);
2641 shader_addline(buffer, "tmp0.x = dot(%s, %s);\n",
2642 src_param.param_str, src_param.param_str);
2643 shader_glsl_append_dst(buffer, ins);
2647 shader_addline(buffer, "tmp0.x == 0.0 ? vec%u(0.0) : (%s * inversesqrt(tmp0.x)));\n",
2648 mask_size, src_param.param_str);
2652 shader_addline(buffer, "tmp0.x == 0.0 ? 0.0 : (%s * inversesqrt(tmp0.x)));\n",
2653 src_param.param_str);
2657 /** Process the WINED3DSIO_EXPP instruction in GLSL:
2658 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
2659 * dst.x = 2^(floor(src))
2660 * dst.y = src - floor(src)
2661 * dst.z = 2^src (partial precision is allowed, but optional)
2663 * For 2.0 shaders, just do this (honoring writemask and swizzle):
2664 * dst = 2^src; (partial precision is allowed, but optional)
2666 static void shader_glsl_expp(const struct wined3d_shader_instruction *ins)
2668 struct glsl_src_param src_param;
2670 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src_param);
2672 if (ins->ctx->reg_maps->shader_version.major < 2)
2676 shader_addline(ins->ctx->buffer, "tmp0.x = exp2(floor(%s));\n", src_param.param_str);
2677 shader_addline(ins->ctx->buffer, "tmp0.y = %s - floor(%s);\n", src_param.param_str, src_param.param_str);
2678 shader_addline(ins->ctx->buffer, "tmp0.z = exp2(%s);\n", src_param.param_str);
2679 shader_addline(ins->ctx->buffer, "tmp0.w = 1.0;\n");
2681 shader_glsl_append_dst(ins->ctx->buffer, ins);
2682 shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
2683 shader_addline(ins->ctx->buffer, "tmp0%s);\n", dst_mask);
2686 unsigned int mask_size;
2688 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
2689 mask_size = shader_glsl_get_write_mask_size(write_mask);
2691 if (mask_size > 1) {
2692 shader_addline(ins->ctx->buffer, "vec%d(exp2(%s)));\n", mask_size, src_param.param_str);
2694 shader_addline(ins->ctx->buffer, "exp2(%s));\n", src_param.param_str);
2699 static void shader_glsl_to_int(const struct wined3d_shader_instruction *ins)
2701 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2702 struct glsl_src_param src_param;
2703 unsigned int mask_size;
2706 write_mask = shader_glsl_append_dst(buffer, ins);
2707 mask_size = shader_glsl_get_write_mask_size(write_mask);
2708 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src_param);
2711 shader_addline(buffer, "ivec%u(%s));\n", mask_size, src_param.param_str);
2713 shader_addline(buffer, "int(%s));\n", src_param.param_str);
2716 static void shader_glsl_to_float(const struct wined3d_shader_instruction *ins)
2718 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2719 struct glsl_src_param src_param;
2720 unsigned int mask_size;
2723 write_mask = shader_glsl_append_dst(buffer, ins);
2724 mask_size = shader_glsl_get_write_mask_size(write_mask);
2725 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src_param);
2728 shader_addline(buffer, "vec%u(%s));\n", mask_size, src_param.param_str);
2730 shader_addline(buffer, "float(%s));\n", src_param.param_str);
2733 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
2734 static void shader_glsl_rcp(const struct wined3d_shader_instruction *ins)
2736 struct glsl_src_param src_param;
2738 unsigned int mask_size;
2740 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
2741 mask_size = shader_glsl_get_write_mask_size(write_mask);
2742 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &src_param);
2746 shader_addline(ins->ctx->buffer, "vec%u(1.0 / %s));\n",
2747 mask_size, src_param.param_str);
2751 shader_addline(ins->ctx->buffer, "1.0 / %s);\n",
2752 src_param.param_str);
2756 static void shader_glsl_rsq(const struct wined3d_shader_instruction *ins)
2758 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2759 struct glsl_src_param src_param;
2761 unsigned int mask_size;
2763 write_mask = shader_glsl_append_dst(buffer, ins);
2764 mask_size = shader_glsl_get_write_mask_size(write_mask);
2766 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &src_param);
2770 shader_addline(buffer, "vec%u(inversesqrt(abs(%s))));\n",
2771 mask_size, src_param.param_str);
2775 shader_addline(buffer, "inversesqrt(abs(%s)));\n",
2776 src_param.param_str);
2780 /** Process signed comparison opcodes in GLSL. */
2781 static void shader_glsl_compare(const struct wined3d_shader_instruction *ins)
2783 struct glsl_src_param src0_param;
2784 struct glsl_src_param src1_param;
2786 unsigned int mask_size;
2788 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
2789 mask_size = shader_glsl_get_write_mask_size(write_mask);
2790 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2791 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2793 if (mask_size > 1) {
2794 const char *compare;
2796 switch(ins->handler_idx)
2798 case WINED3DSIH_SLT: compare = "lessThan"; break;
2799 case WINED3DSIH_SGE: compare = "greaterThanEqual"; break;
2800 default: compare = "";
2801 FIXME("Can't handle opcode %#x\n", ins->handler_idx);
2804 shader_addline(ins->ctx->buffer, "vec%d(%s(%s, %s)));\n", mask_size, compare,
2805 src0_param.param_str, src1_param.param_str);
2807 switch(ins->handler_idx)
2809 case WINED3DSIH_SLT:
2810 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
2811 * to return 0.0 but step returns 1.0 because step is not < x
2812 * An alternative is a bvec compare padded with an unused second component.
2813 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
2814 * issue. Playing with not() is not possible either because not() does not accept
2817 shader_addline(ins->ctx->buffer, "(%s < %s) ? 1.0 : 0.0);\n",
2818 src0_param.param_str, src1_param.param_str);
2820 case WINED3DSIH_SGE:
2821 /* Here we can use the step() function and safe a conditional */
2822 shader_addline(ins->ctx->buffer, "step(%s, %s));\n", src1_param.param_str, src0_param.param_str);
2825 FIXME("Can't handle opcode %#x\n", ins->handler_idx);
2831 static void shader_glsl_conditional_move(const struct wined3d_shader_instruction *ins)
2833 const char *condition_prefix, *condition_suffix;
2834 struct wined3d_shader_dst_param dst;
2835 struct glsl_src_param src0_param;
2836 struct glsl_src_param src1_param;
2837 struct glsl_src_param src2_param;
2838 BOOL temp_destination = FALSE;
2839 DWORD cmp_channel = 0;
2844 switch (ins->handler_idx)
2846 case WINED3DSIH_CMP:
2847 condition_prefix = "";
2848 condition_suffix = " >= 0.0";
2851 case WINED3DSIH_CND:
2852 condition_prefix = "";
2853 condition_suffix = " > 0.5";
2856 case WINED3DSIH_MOVC:
2857 condition_prefix = "bool(";
2858 condition_suffix = ")";
2862 FIXME("Unhandled instruction %#x.\n", ins->handler_idx);
2863 condition_prefix = "<unhandled prefix>";
2864 condition_suffix = "<unhandled suffix>";
2868 if (shader_is_scalar(&ins->dst[0].reg) || shader_is_scalar(&ins->src[0].reg))
2870 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
2871 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2872 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2873 shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
2875 shader_addline(ins->ctx->buffer, "%s%s%s ? %s : %s);\n",
2876 condition_prefix, src0_param.param_str, condition_suffix,
2877 src1_param.param_str, src2_param.param_str);
2883 /* Splitting the instruction up in multiple lines imposes a problem:
2884 * The first lines may overwrite source parameters of the following lines.
2885 * Deal with that by using a temporary destination register if needed. */
2886 if ((ins->src[0].reg.idx[0].offset == dst.reg.idx[0].offset
2887 && ins->src[0].reg.type == dst.reg.type)
2888 || (ins->src[1].reg.idx[0].offset == dst.reg.idx[0].offset
2889 && ins->src[1].reg.type == dst.reg.type)
2890 || (ins->src[2].reg.idx[0].offset == dst.reg.idx[0].offset
2891 && ins->src[2].reg.type == dst.reg.type))
2892 temp_destination = TRUE;
2894 /* Cycle through all source0 channels. */
2895 for (i = 0; i < 4; ++i)
2898 /* Find the destination channels which use the current source0 channel. */
2899 for (j = 0; j < 4; ++j)
2901 if (((ins->src[0].swizzle >> (2 * j)) & 0x3) == i)
2903 write_mask |= WINED3DSP_WRITEMASK_0 << j;
2904 cmp_channel = WINED3DSP_WRITEMASK_0 << j;
2907 dst.write_mask = ins->dst[0].write_mask & write_mask;
2909 if (temp_destination)
2911 if (!(write_mask = shader_glsl_get_write_mask(&dst, mask_char)))
2913 shader_addline(ins->ctx->buffer, "tmp0%s = (", mask_char);
2915 else if (!(write_mask = shader_glsl_append_dst_ext(ins->ctx->buffer, ins, &dst)))
2918 shader_glsl_add_src_param(ins, &ins->src[0], cmp_channel, &src0_param);
2919 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2920 shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
2922 shader_addline(ins->ctx->buffer, "%s%s%s ? %s : %s);\n",
2923 condition_prefix, src0_param.param_str, condition_suffix,
2924 src1_param.param_str, src2_param.param_str);
2927 if (temp_destination)
2929 shader_glsl_get_write_mask(&ins->dst[0], mask_char);
2930 shader_glsl_append_dst(ins->ctx->buffer, ins);
2931 shader_addline(ins->ctx->buffer, "tmp0%s);\n", mask_char);
2935 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
2936 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
2937 * the compare is done per component of src0. */
2938 static void shader_glsl_cnd(const struct wined3d_shader_instruction *ins)
2940 struct glsl_src_param src0_param;
2941 struct glsl_src_param src1_param;
2942 struct glsl_src_param src2_param;
2944 DWORD shader_version = WINED3D_SHADER_VERSION(ins->ctx->reg_maps->shader_version.major,
2945 ins->ctx->reg_maps->shader_version.minor);
2947 if (shader_version < WINED3D_SHADER_VERSION(1, 4))
2949 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
2950 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
2951 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2952 shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
2954 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2957 shader_addline(ins->ctx->buffer, "%s /* COISSUE! */);\n", src1_param.param_str);
2959 shader_addline(ins->ctx->buffer, "%s > 0.5 ? %s : %s);\n",
2960 src0_param.param_str, src1_param.param_str, src2_param.param_str);
2965 shader_glsl_conditional_move(ins);
2968 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2969 static void shader_glsl_mad(const struct wined3d_shader_instruction *ins)
2971 struct glsl_src_param src0_param;
2972 struct glsl_src_param src1_param;
2973 struct glsl_src_param src2_param;
2976 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
2977 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2978 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2979 shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
2980 shader_addline(ins->ctx->buffer, "(%s * %s) + %s);\n",
2981 src0_param.param_str, src1_param.param_str, src2_param.param_str);
2984 /* Handles transforming all WINED3DSIO_M?x? opcodes for
2985 Vertex shaders to GLSL codes */
2986 static void shader_glsl_mnxn(const struct wined3d_shader_instruction *ins)
2989 int nComponents = 0;
2990 struct wined3d_shader_dst_param tmp_dst = {{0}};
2991 struct wined3d_shader_src_param tmp_src[2] = {{{0}}};
2992 struct wined3d_shader_instruction tmp_ins;
2994 memset(&tmp_ins, 0, sizeof(tmp_ins));
2996 /* Set constants for the temporary argument */
2997 tmp_ins.ctx = ins->ctx;
2998 tmp_ins.dst_count = 1;
2999 tmp_ins.dst = &tmp_dst;
3000 tmp_ins.src_count = 2;
3001 tmp_ins.src = tmp_src;
3003 switch(ins->handler_idx)
3005 case WINED3DSIH_M4x4:
3007 tmp_ins.handler_idx = WINED3DSIH_DP4;
3009 case WINED3DSIH_M4x3:
3011 tmp_ins.handler_idx = WINED3DSIH_DP4;
3013 case WINED3DSIH_M3x4:
3015 tmp_ins.handler_idx = WINED3DSIH_DP3;
3017 case WINED3DSIH_M3x3:
3019 tmp_ins.handler_idx = WINED3DSIH_DP3;
3021 case WINED3DSIH_M3x2:
3023 tmp_ins.handler_idx = WINED3DSIH_DP3;
3029 tmp_dst = ins->dst[0];
3030 tmp_src[0] = ins->src[0];
3031 tmp_src[1] = ins->src[1];
3032 for (i = 0; i < nComponents; ++i)
3034 tmp_dst.write_mask = WINED3DSP_WRITEMASK_0 << i;
3035 shader_glsl_dot(&tmp_ins);
3036 ++tmp_src[1].reg.idx[0].offset;
3041 The LRP instruction performs a component-wise linear interpolation
3042 between the second and third operands using the first operand as the
3043 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
3044 This is equivalent to mix(src2, src1, src0);
3046 static void shader_glsl_lrp(const struct wined3d_shader_instruction *ins)
3048 struct glsl_src_param src0_param;
3049 struct glsl_src_param src1_param;
3050 struct glsl_src_param src2_param;
3053 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
3055 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
3056 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
3057 shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
3059 shader_addline(ins->ctx->buffer, "mix(%s, %s, %s));\n",
3060 src2_param.param_str, src1_param.param_str, src0_param.param_str);
3063 /** Process the WINED3DSIO_LIT instruction in GLSL:
3064 * dst.x = dst.w = 1.0
3065 * dst.y = (src0.x > 0) ? src0.x
3066 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
3067 * where src.w is clamped at +- 128
3069 static void shader_glsl_lit(const struct wined3d_shader_instruction *ins)
3071 struct glsl_src_param src0_param;
3072 struct glsl_src_param src1_param;
3073 struct glsl_src_param src3_param;
3076 shader_glsl_append_dst(ins->ctx->buffer, ins);
3077 shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
3079 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
3080 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_1, &src1_param);
3081 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &src3_param);
3083 /* The sdk specifies the instruction like this
3085 * if(src.x > 0.0) dst.y = src.x
3087 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
3090 * (where power = src.w clamped between -128 and 128)
3092 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
3093 * dst.x = 1.0 ... No further explanation needed
3094 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
3095 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
3096 * dst.w = 1.0. ... Nothing fancy.
3098 * So we still have one conditional in there. So do this:
3099 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
3101 * step(0.0, x) will return 1 if src.x > 0.0, and 0 otherwise. So if y is 0 we get pow(0.0 * 1.0, power),
3102 * which sets dst.z to 0. If y > 0, but x = 0.0, we get pow(y * 0.0, power), which results in 0 too.
3103 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to.
3105 * Unfortunately pow(0.0 ^ 0.0) returns NaN on most GPUs, but lit with src.y = 0 and src.w = 0 returns
3106 * a non-NaN value in dst.z. What we return doesn't matter, as long as it is not NaN. Return 0, which is
3107 * what all Windows HW drivers and GL_ARB_vertex_program's LIT do.
3109 shader_addline(ins->ctx->buffer,
3110 "vec4(1.0, max(%s, 0.0), %s == 0.0 ? 0.0 : "
3111 "pow(max(0.0, %s) * step(0.0, %s), clamp(%s, -128.0, 128.0)), 1.0)%s);\n",
3112 src0_param.param_str, src3_param.param_str, src1_param.param_str,
3113 src0_param.param_str, src3_param.param_str, dst_mask);
3116 /** Process the WINED3DSIO_DST instruction in GLSL:
3118 * dst.y = src0.x * src0.y
3122 static void shader_glsl_dst(const struct wined3d_shader_instruction *ins)
3124 struct glsl_src_param src0y_param;
3125 struct glsl_src_param src0z_param;
3126 struct glsl_src_param src1y_param;
3127 struct glsl_src_param src1w_param;
3130 shader_glsl_append_dst(ins->ctx->buffer, ins);
3131 shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
3133 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_1, &src0y_param);
3134 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_2, &src0z_param);
3135 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_1, &src1y_param);
3136 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_3, &src1w_param);
3138 shader_addline(ins->ctx->buffer, "vec4(1.0, %s * %s, %s, %s))%s;\n",
3139 src0y_param.param_str, src1y_param.param_str, src0z_param.param_str, src1w_param.param_str, dst_mask);
3142 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
3143 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
3144 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
3146 * dst.x = cos(src0.?)
3147 * dst.y = sin(src0.?)
3151 static void shader_glsl_sincos(const struct wined3d_shader_instruction *ins)
3153 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3154 struct glsl_src_param src0_param;
3157 if (ins->ctx->reg_maps->shader_version.major < 4)
3159 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
3161 write_mask = shader_glsl_append_dst(buffer, ins);
3164 case WINED3DSP_WRITEMASK_0:
3165 shader_addline(buffer, "cos(%s));\n", src0_param.param_str);
3168 case WINED3DSP_WRITEMASK_1:
3169 shader_addline(buffer, "sin(%s));\n", src0_param.param_str);
3172 case (WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1):
3173 shader_addline(buffer, "vec2(cos(%s), sin(%s)));\n",
3174 src0_param.param_str, src0_param.param_str);
3178 ERR("Write mask should be .x, .y or .xy\n");
3185 if (ins->dst[0].reg.type != WINED3DSPR_NULL)
3188 if (ins->dst[1].reg.type != WINED3DSPR_NULL)
3192 write_mask = shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
3193 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
3194 shader_addline(buffer, "tmp0%s = sin(%s);\n", dst_mask, src0_param.param_str);
3196 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[1]);
3197 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
3198 shader_addline(buffer, "cos(%s));\n", src0_param.param_str);
3200 shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0]);
3201 shader_addline(buffer, "tmp0%s);\n", dst_mask);
3205 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0]);
3206 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
3207 shader_addline(buffer, "sin(%s));\n", src0_param.param_str);
3210 else if (ins->dst[1].reg.type != WINED3DSPR_NULL)
3212 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[1]);
3213 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
3214 shader_addline(buffer, "cos(%s));\n", src0_param.param_str);
3218 /* sgn in vs_2_0 has 2 extra parameters(registers for temporary storage) which we don't use
3219 * here. But those extra parameters require a dedicated function for sgn, since map2gl would
3220 * generate invalid code
3222 static void shader_glsl_sgn(const struct wined3d_shader_instruction *ins)
3224 struct glsl_src_param src0_param;
3227 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
3228 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
3230 shader_addline(ins->ctx->buffer, "sign(%s));\n", src0_param.param_str);
3233 /** Process the WINED3DSIO_LOOP instruction in GLSL:
3234 * Start a for() loop where src1.y is the initial value of aL,
3235 * increment aL by src1.z for a total of src1.x iterations.
3236 * Need to use a temporary variable for this operation.
3238 /* FIXME: I don't think nested loops will work correctly this way. */
3239 static void shader_glsl_loop(const struct wined3d_shader_instruction *ins)
3241 struct wined3d_shader_loop_state *loop_state = ins->ctx->loop_state;
3242 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3243 const struct wined3d_shader *shader = ins->ctx->shader;
3244 const struct wined3d_shader_lconst *constant;
3245 struct glsl_src_param src1_param;
3246 const DWORD *control_values = NULL;
3248 if (ins->ctx->reg_maps->shader_version.major < 4)
3250 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_ALL, &src1_param);
3252 /* Try to hardcode the loop control parameters if possible. Direct3D 9
3253 * class hardware doesn't support real varying indexing, but Microsoft
3254 * designed this feature for Shader model 2.x+. If the loop control is
3255 * known at compile time, the GLSL compiler can unroll the loop, and
3256 * replace indirect addressing with direct addressing. */
3257 if (ins->src[1].reg.type == WINED3DSPR_CONSTINT)
3259 LIST_FOR_EACH_ENTRY(constant, &shader->constantsI, struct wined3d_shader_lconst, entry)
3261 if (constant->idx == ins->src[1].reg.idx[0].offset)
3263 control_values = constant->value;
3271 struct wined3d_shader_loop_control loop_control;
3272 loop_control.count = control_values[0];
3273 loop_control.start = control_values[1];
3274 loop_control.step = (int)control_values[2];
3276 if (loop_control.step > 0)
3278 shader_addline(buffer, "for (aL%u = %u; aL%u < (%u * %d + %u); aL%u += %d)\n{\n",
3279 loop_state->current_depth, loop_control.start,
3280 loop_state->current_depth, loop_control.count, loop_control.step, loop_control.start,
3281 loop_state->current_depth, loop_control.step);
3283 else if (loop_control.step < 0)
3285 shader_addline(buffer, "for (aL%u = %u; aL%u > (%u * %d + %u); aL%u += %d)\n{\n",
3286 loop_state->current_depth, loop_control.start,
3287 loop_state->current_depth, loop_control.count, loop_control.step, loop_control.start,
3288 loop_state->current_depth, loop_control.step);
3292 shader_addline(buffer, "for (aL%u = %u, tmpInt%u = 0; tmpInt%u < %u; tmpInt%u++)\n{\n",
3293 loop_state->current_depth, loop_control.start, loop_state->current_depth,
3294 loop_state->current_depth, loop_control.count,
3295 loop_state->current_depth);
3300 shader_addline(buffer, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z)\n{\n",
3301 loop_state->current_depth, loop_state->current_reg,
3302 src1_param.reg_name, loop_state->current_depth, src1_param.reg_name,
3303 loop_state->current_depth, loop_state->current_reg, src1_param.reg_name);
3306 ++loop_state->current_reg;
3310 shader_addline(buffer, "for (;;)\n{\n");
3313 ++loop_state->current_depth;
3316 static void shader_glsl_end(const struct wined3d_shader_instruction *ins)
3318 struct wined3d_shader_loop_state *loop_state = ins->ctx->loop_state;
3320 shader_addline(ins->ctx->buffer, "}\n");
3322 if (ins->handler_idx == WINED3DSIH_ENDLOOP)
3324 --loop_state->current_depth;
3325 --loop_state->current_reg;
3328 if (ins->handler_idx == WINED3DSIH_ENDREP)
3330 --loop_state->current_depth;
3334 static void shader_glsl_rep(const struct wined3d_shader_instruction *ins)
3336 const struct wined3d_shader *shader = ins->ctx->shader;
3337 struct wined3d_shader_loop_state *loop_state = ins->ctx->loop_state;
3338 const struct wined3d_shader_lconst *constant;
3339 struct glsl_src_param src0_param;
3340 const DWORD *control_values = NULL;
3342 /* Try to hardcode local values to help the GLSL compiler to unroll and optimize the loop */
3343 if (ins->src[0].reg.type == WINED3DSPR_CONSTINT)
3345 LIST_FOR_EACH_ENTRY(constant, &shader->constantsI, struct wined3d_shader_lconst, entry)
3347 if (constant->idx == ins->src[0].reg.idx[0].offset)
3349 control_values = constant->value;
3357 shader_addline(ins->ctx->buffer, "for (tmpInt%d = 0; tmpInt%d < %d; tmpInt%d++) {\n",
3358 loop_state->current_depth, loop_state->current_depth,
3359 control_values[0], loop_state->current_depth);
3363 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
3364 shader_addline(ins->ctx->buffer, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
3365 loop_state->current_depth, loop_state->current_depth,
3366 src0_param.param_str, loop_state->current_depth);
3369 ++loop_state->current_depth;
3372 static void shader_glsl_if(const struct wined3d_shader_instruction *ins)
3374 struct glsl_src_param src0_param;
3376 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
3377 shader_addline(ins->ctx->buffer, "if (%s) {\n", src0_param.param_str);
3380 static void shader_glsl_ifc(const struct wined3d_shader_instruction *ins)
3382 struct glsl_src_param src0_param;
3383 struct glsl_src_param src1_param;
3385 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
3386 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0, &src1_param);
3388 shader_addline(ins->ctx->buffer, "if (%s %s %s) {\n",
3389 src0_param.param_str, shader_glsl_get_rel_op(ins->flags), src1_param.param_str);
3392 static void shader_glsl_else(const struct wined3d_shader_instruction *ins)
3394 shader_addline(ins->ctx->buffer, "} else {\n");
3397 static void shader_glsl_emit(const struct wined3d_shader_instruction *ins)
3399 shader_addline(ins->ctx->buffer, "EmitVertex();\n");
3402 static void shader_glsl_break(const struct wined3d_shader_instruction *ins)
3404 shader_addline(ins->ctx->buffer, "break;\n");
3407 /* FIXME: According to MSDN the compare is done per component. */
3408 static void shader_glsl_breakc(const struct wined3d_shader_instruction *ins)
3410 struct glsl_src_param src0_param;
3411 struct glsl_src_param src1_param;
3413 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
3414 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0, &src1_param);
3416 shader_addline(ins->ctx->buffer, "if (%s %s %s) break;\n",
3417 src0_param.param_str, shader_glsl_get_rel_op(ins->flags), src1_param.param_str);
3420 static void shader_glsl_breakp(const struct wined3d_shader_instruction *ins)
3422 struct glsl_src_param src_param;
3424 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src_param);
3425 shader_addline(ins->ctx->buffer, "if (bool(%s)) break;\n", src_param.param_str);
3428 static void shader_glsl_label(const struct wined3d_shader_instruction *ins)
3430 shader_addline(ins->ctx->buffer, "}\n");
3431 shader_addline(ins->ctx->buffer, "void subroutine%u()\n{\n", ins->src[0].reg.idx[0].offset);
3434 static void shader_glsl_call(const struct wined3d_shader_instruction *ins)
3436 shader_addline(ins->ctx->buffer, "subroutine%u();\n", ins->src[0].reg.idx[0].offset);
3439 static void shader_glsl_callnz(const struct wined3d_shader_instruction *ins)
3441 struct glsl_src_param src1_param;
3443 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0, &src1_param);
3444 shader_addline(ins->ctx->buffer, "if (%s) subroutine%u();\n",
3445 src1_param.param_str, ins->src[0].reg.idx[0].offset);
3448 static void shader_glsl_ret(const struct wined3d_shader_instruction *ins)
3450 /* No-op. The closing } is written when a new function is started, and at the end of the shader. This
3451 * function only suppresses the unhandled instruction warning
3455 /*********************************************
3456 * Pixel Shader Specific Code begins here
3457 ********************************************/
3458 static void shader_glsl_tex(const struct wined3d_shader_instruction *ins)
3460 const struct wined3d_shader *shader = ins->ctx->shader;
3461 struct wined3d_device *device = shader->device;
3462 DWORD shader_version = WINED3D_SHADER_VERSION(ins->ctx->reg_maps->shader_version.major,
3463 ins->ctx->reg_maps->shader_version.minor);
3464 struct glsl_sample_function sample_function;
3465 const struct wined3d_texture *texture;
3466 DWORD sample_flags = 0;
3468 DWORD mask = 0, swizzle;
3470 /* 1.0-1.4: Use destination register as sampler source.
3471 * 2.0+: Use provided sampler source. */
3472 if (shader_version < WINED3D_SHADER_VERSION(2,0))
3473 sampler_idx = ins->dst[0].reg.idx[0].offset;
3475 sampler_idx = ins->src[1].reg.idx[0].offset;
3476 texture = device->stateBlock->state.textures[sampler_idx];
3478 if (shader_version < WINED3D_SHADER_VERSION(1,4))
3480 const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
3481 DWORD flags = (priv->cur_ps_args->tex_transform >> sampler_idx * WINED3D_PSARGS_TEXTRANSFORM_SHIFT)
3482 & WINED3D_PSARGS_TEXTRANSFORM_MASK;
3483 enum wined3d_sampler_texture_type sampler_type = ins->ctx->reg_maps->sampler_type[sampler_idx];
3485 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
3486 if (flags & WINED3D_PSARGS_PROJECTED && sampler_type != WINED3DSTT_CUBE)
3488 sample_flags |= WINED3D_GLSL_SAMPLE_PROJECTED;
3489 switch (flags & ~WINED3D_PSARGS_PROJECTED)
3491 case WINED3D_TTFF_COUNT1:
3492 FIXME("WINED3D_TTFF_PROJECTED with WINED3D_TTFF_COUNT1?\n");
3494 case WINED3D_TTFF_COUNT2:
3495 mask = WINED3DSP_WRITEMASK_1;
3497 case WINED3D_TTFF_COUNT3:
3498 mask = WINED3DSP_WRITEMASK_2;
3500 case WINED3D_TTFF_COUNT4:
3501 case WINED3D_TTFF_DISABLE:
3502 mask = WINED3DSP_WRITEMASK_3;
3507 else if (shader_version < WINED3D_SHADER_VERSION(2,0))
3509 enum wined3d_shader_src_modifier src_mod = ins->src[0].modifiers;
3511 if (src_mod == WINED3DSPSM_DZ) {
3512 sample_flags |= WINED3D_GLSL_SAMPLE_PROJECTED;
3513 mask = WINED3DSP_WRITEMASK_2;
3514 } else if (src_mod == WINED3DSPSM_DW) {
3515 sample_flags |= WINED3D_GLSL_SAMPLE_PROJECTED;
3516 mask = WINED3DSP_WRITEMASK_3;
3519 if (ins->flags & WINED3DSI_TEXLD_PROJECT)
3521 /* ps 2.0 texldp instruction always divides by the fourth component. */
3522 sample_flags |= WINED3D_GLSL_SAMPLE_PROJECTED;
3523 mask = WINED3DSP_WRITEMASK_3;
3527 if (texture && texture->target == GL_TEXTURE_RECTANGLE_ARB)
3528 sample_flags |= WINED3D_GLSL_SAMPLE_RECT;
3530 shader_glsl_get_sample_function(ins->ctx, sampler_idx, sample_flags, &sample_function);
3531 mask |= sample_function.coord_mask;
3533 if (shader_version < WINED3D_SHADER_VERSION(2,0)) swizzle = WINED3DSP_NOSWIZZLE;
3534 else swizzle = ins->src[1].swizzle;
3536 /* 1.0-1.3: Use destination register as coordinate source.
3537 1.4+: Use provided coordinate source register. */
3538 if (shader_version < WINED3D_SHADER_VERSION(1,4))
3541 shader_glsl_write_mask_to_str(mask, coord_mask);
3542 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, NULL, NULL, NULL,
3543 "T%u%s", sampler_idx, coord_mask);
3547 struct glsl_src_param coord_param;
3548 shader_glsl_add_src_param(ins, &ins->src[0], mask, &coord_param);
3549 if (ins->flags & WINED3DSI_TEXLD_BIAS)
3551 struct glsl_src_param bias;
3552 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &bias);
3553 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, NULL, NULL, bias.param_str,
3554 "%s", coord_param.param_str);
3556 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, NULL, NULL, NULL,
3557 "%s", coord_param.param_str);
3562 static void shader_glsl_texldd(const struct wined3d_shader_instruction *ins)
3564 const struct wined3d_shader *shader = ins->ctx->shader;
3565 struct wined3d_device *device = shader->device;
3566 const struct wined3d_gl_info *gl_info = ins->ctx->gl_info;
3567 struct glsl_src_param coord_param, dx_param, dy_param;
3568 DWORD sample_flags = WINED3D_GLSL_SAMPLE_GRAD;
3569 struct glsl_sample_function sample_function;
3571 DWORD swizzle = ins->src[1].swizzle;
3572 const struct wined3d_texture *texture;
3574 if (!gl_info->supported[ARB_SHADER_TEXTURE_LOD] && !gl_info->supported[EXT_GPU_SHADER4])
3576 FIXME("texldd used, but not supported by hardware. Falling back to regular tex\n");
3577 shader_glsl_tex(ins);
3581 sampler_idx = ins->src[1].reg.idx[0].offset;
3582 texture = device->stateBlock->state.textures[sampler_idx];
3583 if (texture && texture->target == GL_TEXTURE_RECTANGLE_ARB)
3584 sample_flags |= WINED3D_GLSL_SAMPLE_RECT;
3586 shader_glsl_get_sample_function(ins->ctx, sampler_idx, sample_flags, &sample_function);
3587 shader_glsl_add_src_param(ins, &ins->src[0], sample_function.coord_mask, &coord_param);
3588 shader_glsl_add_src_param(ins, &ins->src[2], sample_function.coord_mask, &dx_param);
3589 shader_glsl_add_src_param(ins, &ins->src[3], sample_function.coord_mask, &dy_param);
3591 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, dx_param.param_str, dy_param.param_str, NULL,
3592 "%s", coord_param.param_str);
3595 static void shader_glsl_texldl(const struct wined3d_shader_instruction *ins)
3597 const struct wined3d_shader *shader = ins->ctx->shader;
3598 struct wined3d_device *device = shader->device;
3599 const struct wined3d_gl_info *gl_info = ins->ctx->gl_info;
3600 struct glsl_src_param coord_param, lod_param;
3601 DWORD sample_flags = WINED3D_GLSL_SAMPLE_LOD;
3602 struct glsl_sample_function sample_function;
3604 DWORD swizzle = ins->src[1].swizzle;
3605 const struct wined3d_texture *texture;
3607 sampler_idx = ins->src[1].reg.idx[0].offset;
3608 texture = device->stateBlock->state.textures[sampler_idx];
3609 if (texture && texture->target == GL_TEXTURE_RECTANGLE_ARB)
3610 sample_flags |= WINED3D_GLSL_SAMPLE_RECT;
3612 shader_glsl_get_sample_function(ins->ctx, sampler_idx, sample_flags, &sample_function);
3613 shader_glsl_add_src_param(ins, &ins->src[0], sample_function.coord_mask, &coord_param);
3615 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &lod_param);
3617 if (!gl_info->supported[ARB_SHADER_TEXTURE_LOD] && !gl_info->supported[EXT_GPU_SHADER4]
3618 && ins->ctx->reg_maps->shader_version.type == WINED3D_SHADER_TYPE_PIXEL)
3620 /* Plain GLSL only supports Lod sampling functions in vertex shaders.
3621 * However, the NVIDIA drivers allow them in fragment shaders as well,
3622 * even without the appropriate extension. */
3623 WARN("Using %s in fragment shader.\n", sample_function.name);
3625 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, NULL, NULL, lod_param.param_str,
3626 "%s", coord_param.param_str);
3629 static void shader_glsl_texcoord(const struct wined3d_shader_instruction *ins)
3631 /* FIXME: Make this work for more than just 2D textures */
3632 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3633 DWORD write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
3635 if (!(ins->ctx->reg_maps->shader_version.major == 1 && ins->ctx->reg_maps->shader_version.minor == 4))
3639 shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
3640 shader_addline(buffer, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
3641 ins->dst[0].reg.idx[0].offset, dst_mask);
3645 enum wined3d_shader_src_modifier src_mod = ins->src[0].modifiers;
3646 DWORD reg = ins->src[0].reg.idx[0].offset;
3647 char dst_swizzle[6];
3649 shader_glsl_get_swizzle(&ins->src[0], FALSE, write_mask, dst_swizzle);
3651 if (src_mod == WINED3DSPSM_DZ)
3653 unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
3654 struct glsl_src_param div_param;
3656 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_2, &div_param);
3658 if (mask_size > 1) {
3659 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
3661 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
3664 else if (src_mod == WINED3DSPSM_DW)
3666 unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
3667 struct glsl_src_param div_param;
3669 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &div_param);
3671 if (mask_size > 1) {
3672 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
3674 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
3677 shader_addline(buffer, "gl_TexCoord[%u]%s);\n", reg, dst_swizzle);
3682 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
3683 * Take a 3-component dot product of the TexCoord[dstreg] and src,
3684 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
3685 static void shader_glsl_texdp3tex(const struct wined3d_shader_instruction *ins)
3687 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3688 DWORD sampler_idx = ins->dst[0].reg.idx[0].offset;
3689 struct glsl_sample_function sample_function;
3690 struct glsl_src_param src0_param;
3693 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3695 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
3696 * scalar, and projected sampling would require 4.
3698 * It is a dependent read - not valid with conditional NP2 textures
3700 shader_glsl_get_sample_function(ins->ctx, sampler_idx, 0, &sample_function);
3701 mask_size = shader_glsl_get_write_mask_size(sample_function.coord_mask);
3706 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
3707 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx, src0_param.param_str);
3711 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
3712 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx, src0_param.param_str);
3716 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
3717 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx, src0_param.param_str);
3721 FIXME("Unexpected mask size %u\n", mask_size);
3726 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
3727 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
3728 static void shader_glsl_texdp3(const struct wined3d_shader_instruction *ins)
3730 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3731 DWORD dstreg = ins->dst[0].reg.idx[0].offset;
3732 struct glsl_src_param src0_param;
3734 unsigned int mask_size;
3736 dst_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
3737 mask_size = shader_glsl_get_write_mask_size(dst_mask);
3738 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3740 if (mask_size > 1) {
3741 shader_addline(ins->ctx->buffer, "vec%d(dot(T%u.xyz, %s)));\n", mask_size, dstreg, src0_param.param_str);
3743 shader_addline(ins->ctx->buffer, "dot(T%u.xyz, %s));\n", dstreg, src0_param.param_str);
3747 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
3748 * Calculate the depth as dst.x / dst.y */
3749 static void shader_glsl_texdepth(const struct wined3d_shader_instruction *ins)
3751 struct glsl_dst_param dst_param;
3753 shader_glsl_add_dst_param(ins, &ins->dst[0], &dst_param);
3755 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
3756 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
3757 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
3758 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
3761 shader_addline(ins->ctx->buffer, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
3762 dst_param.reg_name, dst_param.reg_name);
3765 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
3766 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
3767 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
3768 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
3770 static void shader_glsl_texm3x2depth(const struct wined3d_shader_instruction *ins)
3772 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3773 DWORD dstreg = ins->dst[0].reg.idx[0].offset;
3774 struct glsl_src_param src0_param;
3776 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3778 shader_addline(ins->ctx->buffer, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg, src0_param.param_str);
3779 shader_addline(ins->ctx->buffer, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
3782 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
3783 * Calculate the 1st of a 2-row matrix multiplication. */
3784 static void shader_glsl_texm3x2pad(const struct wined3d_shader_instruction *ins)
3786 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3787 DWORD reg = ins->dst[0].reg.idx[0].offset;
3788 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3789 struct glsl_src_param src0_param;
3791 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3792 shader_addline(buffer, "tmp0.x = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
3795 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
3796 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
3797 static void shader_glsl_texm3x3pad(const struct wined3d_shader_instruction *ins)
3799 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3800 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3801 struct wined3d_shader_tex_mx *tex_mx = ins->ctx->tex_mx;
3802 DWORD reg = ins->dst[0].reg.idx[0].offset;
3803 struct glsl_src_param src0_param;
3805 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3806 shader_addline(buffer, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + tex_mx->current_row, reg, src0_param.param_str);
3807 tex_mx->texcoord_w[tex_mx->current_row++] = reg;
3810 static void shader_glsl_texm3x2tex(const struct wined3d_shader_instruction *ins)
3812 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3813 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3814 struct glsl_sample_function sample_function;
3815 DWORD reg = ins->dst[0].reg.idx[0].offset;
3816 struct glsl_src_param src0_param;
3818 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3819 shader_addline(buffer, "tmp0.y = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
3821 shader_glsl_get_sample_function(ins->ctx, reg, 0, &sample_function);
3823 /* Sample the texture using the calculated coordinates */
3824 shader_glsl_gen_sample_code(ins, reg, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL, "tmp0.xy");
3827 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
3828 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
3829 static void shader_glsl_texm3x3tex(const struct wined3d_shader_instruction *ins)
3831 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3832 struct wined3d_shader_tex_mx *tex_mx = ins->ctx->tex_mx;
3833 struct glsl_sample_function sample_function;
3834 DWORD reg = ins->dst[0].reg.idx[0].offset;
3835 struct glsl_src_param src0_param;
3837 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3838 shader_addline(ins->ctx->buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
3840 /* Dependent read, not valid with conditional NP2 */
3841 shader_glsl_get_sample_function(ins->ctx, reg, 0, &sample_function);
3843 /* Sample the texture using the calculated coordinates */
3844 shader_glsl_gen_sample_code(ins, reg, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL, "tmp0.xyz");
3846 tex_mx->current_row = 0;
3849 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
3850 * Perform the 3rd row of a 3x3 matrix multiply */
3851 static void shader_glsl_texm3x3(const struct wined3d_shader_instruction *ins)
3853 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3854 struct wined3d_shader_tex_mx *tex_mx = ins->ctx->tex_mx;
3855 DWORD reg = ins->dst[0].reg.idx[0].offset;
3856 struct glsl_src_param src0_param;
3859 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3861 shader_glsl_append_dst(ins->ctx->buffer, ins);
3862 shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
3863 shader_addline(ins->ctx->buffer, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg, src0_param.param_str, dst_mask);
3865 tex_mx->current_row = 0;
3868 /* Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
3869 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3870 static void shader_glsl_texm3x3spec(const struct wined3d_shader_instruction *ins)
3872 struct glsl_src_param src0_param;
3873 struct glsl_src_param src1_param;
3874 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3875 struct wined3d_shader_tex_mx *tex_mx = ins->ctx->tex_mx;
3876 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3877 struct glsl_sample_function sample_function;
3878 DWORD reg = ins->dst[0].reg.idx[0].offset;
3881 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3882 shader_glsl_add_src_param(ins, &ins->src[1], src_mask, &src1_param);
3884 /* Perform the last matrix multiply operation */
3885 shader_addline(buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
3886 /* Reflection calculation */
3887 shader_addline(buffer, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param.param_str);
3889 /* Dependent read, not valid with conditional NP2 */
3890 shader_glsl_get_sample_function(ins->ctx, reg, 0, &sample_function);
3891 shader_glsl_write_mask_to_str(sample_function.coord_mask, coord_mask);
3893 /* Sample the texture */
3894 shader_glsl_gen_sample_code(ins, reg, &sample_function, WINED3DSP_NOSWIZZLE,
3895 NULL, NULL, NULL, "tmp0%s", coord_mask);
3897 tex_mx->current_row = 0;
3900 /* Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
3901 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3902 static void shader_glsl_texm3x3vspec(const struct wined3d_shader_instruction *ins)
3904 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3905 struct wined3d_shader_tex_mx *tex_mx = ins->ctx->tex_mx;
3906 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3907 struct glsl_sample_function sample_function;
3908 DWORD reg = ins->dst[0].reg.idx[0].offset;
3909 struct glsl_src_param src0_param;
3912 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3914 /* Perform the last matrix multiply operation */
3915 shader_addline(buffer, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg, src0_param.param_str);
3917 /* Construct the eye-ray vector from w coordinates */
3918 shader_addline(buffer, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
3919 tex_mx->texcoord_w[0], tex_mx->texcoord_w[1], reg);
3920 shader_addline(buffer, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
3922 /* Dependent read, not valid with conditional NP2 */
3923 shader_glsl_get_sample_function(ins->ctx, reg, 0, &sample_function);
3924 shader_glsl_write_mask_to_str(sample_function.coord_mask, coord_mask);
3926 /* Sample the texture using the calculated coordinates */
3927 shader_glsl_gen_sample_code(ins, reg, &sample_function, WINED3DSP_NOSWIZZLE,
3928 NULL, NULL, NULL, "tmp0%s", coord_mask);
3930 tex_mx->current_row = 0;
3933 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
3934 * Apply a fake bump map transform.
3935 * texbem is pshader <= 1.3 only, this saves a few version checks
3937 static void shader_glsl_texbem(const struct wined3d_shader_instruction *ins)
3939 const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
3940 struct glsl_sample_function sample_function;
3941 struct glsl_src_param coord_param;
3947 sampler_idx = ins->dst[0].reg.idx[0].offset;
3948 flags = (priv->cur_ps_args->tex_transform >> sampler_idx * WINED3D_PSARGS_TEXTRANSFORM_SHIFT)
3949 & WINED3D_PSARGS_TEXTRANSFORM_MASK;
3951 /* Dependent read, not valid with conditional NP2 */
3952 shader_glsl_get_sample_function(ins->ctx, sampler_idx, 0, &sample_function);
3953 mask = sample_function.coord_mask;
3955 shader_glsl_write_mask_to_str(mask, coord_mask);
3957 /* With projected textures, texbem only divides the static texture coord,
3958 * not the displacement, so we can't let GL handle this. */
3959 if (flags & WINED3D_PSARGS_PROJECTED)
3962 char coord_div_mask[3];
3963 switch (flags & ~WINED3D_PSARGS_PROJECTED)
3965 case WINED3D_TTFF_COUNT1:
3966 FIXME("WINED3D_TTFF_PROJECTED with WINED3D_TTFF_COUNT1?\n");
3968 case WINED3D_TTFF_COUNT2:
3969 div_mask = WINED3DSP_WRITEMASK_1;
3971 case WINED3D_TTFF_COUNT3:
3972 div_mask = WINED3DSP_WRITEMASK_2;
3974 case WINED3D_TTFF_COUNT4:
3975 case WINED3D_TTFF_DISABLE:
3976 div_mask = WINED3DSP_WRITEMASK_3;
3979 shader_glsl_write_mask_to_str(div_mask, coord_div_mask);
3980 shader_addline(ins->ctx->buffer, "T%u%s /= T%u%s;\n", sampler_idx, coord_mask, sampler_idx, coord_div_mask);
3983 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &coord_param);
3985 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
3986 "T%u%s + vec4(bumpenv_mat%u * %s, 0.0, 0.0)%s", sampler_idx, coord_mask, sampler_idx,
3987 coord_param.param_str, coord_mask);
3989 if (ins->handler_idx == WINED3DSIH_TEXBEML)
3991 struct glsl_src_param luminance_param;
3992 struct glsl_dst_param dst_param;
3994 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_2, &luminance_param);
3995 shader_glsl_add_dst_param(ins, &ins->dst[0], &dst_param);
3997 shader_addline(ins->ctx->buffer, "%s%s *= (%s * bumpenv_lum_scale%u + bumpenv_lum_offset%u);\n",
3998 dst_param.reg_name, dst_param.mask_str,
3999 luminance_param.param_str, sampler_idx, sampler_idx);
4003 static void shader_glsl_bem(const struct wined3d_shader_instruction *ins)
4005 DWORD sampler_idx = ins->dst[0].reg.idx[0].offset;
4006 struct glsl_src_param src0_param, src1_param;
4008 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src0_param);
4009 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src1_param);
4011 shader_glsl_append_dst(ins->ctx->buffer, ins);
4012 shader_addline(ins->ctx->buffer, "%s + bumpenv_mat%u * %s);\n",
4013 src0_param.param_str, sampler_idx, src1_param.param_str);
4016 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
4017 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
4018 static void shader_glsl_texreg2ar(const struct wined3d_shader_instruction *ins)
4020 DWORD sampler_idx = ins->dst[0].reg.idx[0].offset;
4021 struct glsl_sample_function sample_function;
4022 struct glsl_src_param src0_param;
4024 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
4026 shader_glsl_get_sample_function(ins->ctx, sampler_idx, 0, &sample_function);
4027 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
4028 "%s.wx", src0_param.reg_name);
4031 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
4032 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
4033 static void shader_glsl_texreg2gb(const struct wined3d_shader_instruction *ins)
4035 DWORD sampler_idx = ins->dst[0].reg.idx[0].offset;
4036 struct glsl_sample_function sample_function;
4037 struct glsl_src_param src0_param;
4039 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
4041 shader_glsl_get_sample_function(ins->ctx, sampler_idx, 0, &sample_function);
4042 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
4043 "%s.yz", src0_param.reg_name);
4046 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
4047 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
4048 static void shader_glsl_texreg2rgb(const struct wined3d_shader_instruction *ins)
4050 DWORD sampler_idx = ins->dst[0].reg.idx[0].offset;
4051 struct glsl_sample_function sample_function;
4052 struct glsl_src_param src0_param;
4054 /* Dependent read, not valid with conditional NP2 */
4055 shader_glsl_get_sample_function(ins->ctx, sampler_idx, 0, &sample_function);
4056 shader_glsl_add_src_param(ins, &ins->src[0], sample_function.coord_mask, &src0_param);
4058 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
4059 "%s", src0_param.param_str);
4062 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
4063 * If any of the first 3 components are < 0, discard this pixel */
4064 static void shader_glsl_texkill(const struct wined3d_shader_instruction *ins)
4066 struct glsl_dst_param dst_param;
4068 /* The argument is a destination parameter, and no writemasks are allowed */
4069 shader_glsl_add_dst_param(ins, &ins->dst[0], &dst_param);
4070 if (ins->ctx->reg_maps->shader_version.major >= 2)
4072 /* 2.0 shaders compare all 4 components in texkill */
4073 shader_addline(ins->ctx->buffer, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param.reg_name);
4075 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
4076 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
4077 * 4 components are defined, only the first 3 are used
4079 shader_addline(ins->ctx->buffer, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param.reg_name);
4083 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
4084 * dst = dot2(src0, src1) + src2 */
4085 static void shader_glsl_dp2add(const struct wined3d_shader_instruction *ins)
4087 struct glsl_src_param src0_param;
4088 struct glsl_src_param src1_param;
4089 struct glsl_src_param src2_param;
4091 unsigned int mask_size;
4093 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
4094 mask_size = shader_glsl_get_write_mask_size(write_mask);
4096 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src0_param);
4097 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src1_param);
4098 shader_glsl_add_src_param(ins, &ins->src[2], WINED3DSP_WRITEMASK_0, &src2_param);
4100 if (mask_size > 1) {
4101 shader_addline(ins->ctx->buffer, "vec%d(dot(%s, %s) + %s));\n",
4102 mask_size, src0_param.param_str, src1_param.param_str, src2_param.param_str);
4104 shader_addline(ins->ctx->buffer, "dot(%s, %s) + %s);\n",
4105 src0_param.param_str, src1_param.param_str, src2_param.param_str);
4109 static void shader_glsl_input_pack(const struct wined3d_shader *shader, struct wined3d_shader_buffer *buffer,
4110 const struct wined3d_shader_signature_element *input_signature,
4111 const struct wined3d_shader_reg_maps *reg_maps,
4112 enum vertexprocessing_mode vertexprocessing)
4114 WORD map = reg_maps->input_registers;
4117 for (i = 0; map; map >>= 1, ++i)
4119 const char *semantic_name;
4124 if (!(map & 1)) continue;
4126 semantic_name = input_signature[i].semantic_name;
4127 semantic_idx = input_signature[i].semantic_idx;
4128 shader_glsl_write_mask_to_str(input_signature[i].mask, reg_mask);
4130 if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_TEXCOORD))
4132 if (semantic_idx < 8 && vertexprocessing == pretransformed)
4133 shader_addline(buffer, "ps_in[%u]%s = gl_TexCoord[%u]%s;\n",
4134 shader->u.ps.input_reg_map[i], reg_mask, semantic_idx, reg_mask);
4136 shader_addline(buffer, "ps_in[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
4137 shader->u.ps.input_reg_map[i], reg_mask, reg_mask);
4139 else if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_COLOR))
4142 shader_addline(buffer, "ps_in[%u]%s = vec4(gl_Color)%s;\n",
4143 shader->u.ps.input_reg_map[i], reg_mask, reg_mask);
4144 else if (semantic_idx == 1)
4145 shader_addline(buffer, "ps_in[%u]%s = vec4(gl_SecondaryColor)%s;\n",
4146 shader->u.ps.input_reg_map[i], reg_mask, reg_mask);
4148 shader_addline(buffer, "ps_in[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
4149 shader->u.ps.input_reg_map[i], reg_mask, reg_mask);
4153 shader_addline(buffer, "ps_in[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
4154 shader->u.ps.input_reg_map[i], reg_mask, reg_mask);
4159 /*********************************************
4160 * Vertex Shader Specific Code begins here
4161 ********************************************/
4163 static void add_glsl_program_entry(struct shader_glsl_priv *priv, struct glsl_shader_prog_link *entry)
4165 struct glsl_program_key key;
4167 key.vs_id = entry->vs.id;
4168 key.gs_id = entry->gs.id;
4169 key.ps_id = entry->ps.id;
4171 if (wine_rb_put(&priv->program_lookup, &key, &entry->program_lookup_entry) == -1)
4173 ERR("Failed to insert program entry.\n");
4177 static struct glsl_shader_prog_link *get_glsl_program_entry(const struct shader_glsl_priv *priv,
4178 GLhandleARB vs_id, GLhandleARB gs_id, GLhandleARB ps_id)
4180 struct wine_rb_entry *entry;
4181 struct glsl_program_key key;
4187 entry = wine_rb_get(&priv->program_lookup, &key);
4188 return entry ? WINE_RB_ENTRY_VALUE(entry, struct glsl_shader_prog_link, program_lookup_entry) : NULL;
4191 /* Context activation is done by the caller. */
4192 static void delete_glsl_program_entry(struct shader_glsl_priv *priv, const struct wined3d_gl_info *gl_info,
4193 struct glsl_shader_prog_link *entry)
4195 struct glsl_program_key key;
4197 key.vs_id = entry->vs.id;
4198 key.gs_id = entry->gs.id;
4199 key.ps_id = entry->ps.id;
4200 wine_rb_remove(&priv->program_lookup, &key);
4202 GL_EXTCALL(glDeleteObjectARB(entry->programId));
4204 list_remove(&entry->vs.shader_entry);
4206 list_remove(&entry->gs.shader_entry);
4208 list_remove(&entry->ps.shader_entry);
4209 HeapFree(GetProcessHeap(), 0, entry->vs.uniform_f_locations);
4210 HeapFree(GetProcessHeap(), 0, entry->ps.uniform_f_locations);
4211 HeapFree(GetProcessHeap(), 0, entry);
4214 static void handle_ps3_input(struct wined3d_shader_buffer *buffer,
4215 const struct wined3d_gl_info *gl_info, const DWORD *map,
4216 const struct wined3d_shader_signature_element *input_signature,
4217 const struct wined3d_shader_reg_maps *reg_maps_in,
4218 const struct wined3d_shader_signature_element *output_signature,
4219 const struct wined3d_shader_reg_maps *reg_maps_out)
4222 const char *semantic_name_in;
4223 UINT semantic_idx_in;
4226 unsigned int in_count = vec4_varyings(3, gl_info);
4228 char destination[50];
4229 WORD input_map, output_map;
4231 set = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*set) * (in_count + 2));
4233 input_map = reg_maps_in->input_registers;
4234 for (i = 0; input_map; input_map >>= 1, ++i)
4236 if (!(input_map & 1)) continue;
4239 /* Declared, but not read register */
4240 if (in_idx == ~0U) continue;
4241 if (in_idx >= (in_count + 2))
4243 FIXME("More input varyings declared than supported, expect issues.\n");
4247 if (in_idx == in_count)
4248 sprintf(destination, "gl_FrontColor");
4249 else if (in_idx == in_count + 1)
4250 sprintf(destination, "gl_FrontSecondaryColor");
4252 sprintf(destination, "ps_in[%u]", in_idx);
4254 semantic_name_in = input_signature[i].semantic_name;
4255 semantic_idx_in = input_signature[i].semantic_idx;
4258 output_map = reg_maps_out->output_registers;
4259 for (j = 0; output_map; output_map >>= 1, ++j)
4263 if (!(output_map & 1)
4264 || semantic_idx_in != output_signature[j].semantic_idx
4265 || strcmp(semantic_name_in, output_signature[j].semantic_name)
4266 || !(mask = input_signature[i].mask & output_signature[j].mask))
4270 shader_glsl_write_mask_to_str(mask, reg_mask);
4272 shader_addline(buffer, "%s%s = vs_out[%u]%s;\n",
4273 destination, reg_mask, j, reg_mask);
4277 for (i = 0; i < in_count + 2; ++i)
4281 if (!set[i] || set[i] == WINED3DSP_WRITEMASK_ALL)
4284 if (set[i] == ~0U) set[i] = 0;
4287 if (!(set[i] & WINED3DSP_WRITEMASK_0)) reg_mask[size++] = 'x';
4288 if (!(set[i] & WINED3DSP_WRITEMASK_1)) reg_mask[size++] = 'y';
4289 if (!(set[i] & WINED3DSP_WRITEMASK_2)) reg_mask[size++] = 'z';
4290 if (!(set[i] & WINED3DSP_WRITEMASK_3)) reg_mask[size++] = 'w';
4291 reg_mask[size] = '\0';
4294 sprintf(destination, "gl_FrontColor");
4295 else if (i == in_count + 1)
4296 sprintf(destination, "gl_FrontSecondaryColor");
4298 sprintf(destination, "ps_in[%u]", i);
4300 if (size == 1) shader_addline(buffer, "%s.%s = 0.0;\n", destination, reg_mask);
4301 else shader_addline(buffer, "%s.%s = vec%u(0.0);\n", destination, reg_mask, size);
4304 HeapFree(GetProcessHeap(), 0, set);
4307 /* Context activation is done by the caller. */
4308 static GLhandleARB generate_param_reorder_function(struct wined3d_shader_buffer *buffer,
4309 const struct wined3d_shader *vs, const struct wined3d_shader *ps,
4310 const struct wined3d_gl_info *gl_info)
4312 GLhandleARB ret = 0;
4313 DWORD ps_major = ps ? ps->reg_maps.shader_version.major : 0;
4315 const char *semantic_name;
4318 const struct wined3d_shader_signature_element *output_signature = vs->output_signature;
4319 WORD map = vs->reg_maps.output_registers;
4321 shader_buffer_clear(buffer);
4323 shader_addline(buffer, "#version 120\n");
4327 shader_addline(buffer, "void order_ps_input(in vec4 vs_out[%u])\n{\n", vs->limits.packed_output);
4329 for (i = 0; map; map >>= 1, ++i)
4333 if (!(map & 1)) continue;
4335 semantic_name = output_signature[i].semantic_name;
4336 semantic_idx = output_signature[i].semantic_idx;
4337 write_mask = output_signature[i].mask;
4338 shader_glsl_write_mask_to_str(write_mask, reg_mask);
4340 if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_COLOR))
4343 shader_addline(buffer, "gl_FrontColor%s = vs_out[%u]%s;\n",
4344 reg_mask, i, reg_mask);
4345 else if (semantic_idx == 1)
4346 shader_addline(buffer, "gl_FrontSecondaryColor%s = vs_out[%u]%s;\n",
4347 reg_mask, i, reg_mask);
4349 else if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_POSITION))
4351 shader_addline(buffer, "gl_Position%s = vs_out[%u]%s;\n",
4352 reg_mask, i, reg_mask);
4354 else if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_TEXCOORD))
4356 if (semantic_idx < 8)
4358 if (!(gl_info->quirks & WINED3D_QUIRK_SET_TEXCOORD_W) || ps_major > 0)
4359 write_mask |= WINED3DSP_WRITEMASK_3;
4361 shader_addline(buffer, "gl_TexCoord[%u]%s = vs_out[%u]%s;\n",
4362 semantic_idx, reg_mask, i, reg_mask);
4363 if (!(write_mask & WINED3DSP_WRITEMASK_3))
4364 shader_addline(buffer, "gl_TexCoord[%u].w = 1.0;\n", semantic_idx);
4367 else if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_PSIZE))
4369 shader_addline(buffer, "gl_PointSize = vs_out[%u].%c;\n", i, reg_mask[1]);
4371 else if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_FOG))
4373 shader_addline(buffer, "gl_FogFragCoord = clamp(vs_out[%u].%c, 0.0, 1.0);\n", i, reg_mask[1]);
4376 shader_addline(buffer, "}\n");
4380 UINT in_count = min(vec4_varyings(ps_major, gl_info), ps->limits.packed_input);
4381 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
4382 shader_addline(buffer, "varying vec4 ps_in[%u];\n", in_count);
4383 shader_addline(buffer, "void order_ps_input(in vec4 vs_out[%u])\n{\n", vs->limits.packed_output);
4385 /* First, sort out position and point size. Those are not passed to the pixel shader */
4386 for (i = 0; map; map >>= 1, ++i)
4388 if (!(map & 1)) continue;
4390 semantic_name = output_signature[i].semantic_name;
4391 shader_glsl_write_mask_to_str(output_signature[i].mask, reg_mask);
4393 if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_POSITION))
4395 shader_addline(buffer, "gl_Position%s = vs_out[%u]%s;\n",
4396 reg_mask, i, reg_mask);
4398 else if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_PSIZE))
4400 shader_addline(buffer, "gl_PointSize = vs_out[%u].%c;\n", i, reg_mask[1]);
4404 /* Then, fix the pixel shader input */
4405 handle_ps3_input(buffer, gl_info, ps->u.ps.input_reg_map, ps->input_signature,
4406 &ps->reg_maps, output_signature, &vs->reg_maps);
4408 shader_addline(buffer, "}\n");
4411 ret = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
4412 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
4413 shader_glsl_compile(gl_info, ret, buffer->buffer);
4418 static void shader_glsl_generate_srgb_write_correction(struct wined3d_shader_buffer *buffer)
4420 shader_addline(buffer, "tmp0.xyz = pow(gl_FragData[0].xyz, vec3(srgb_const0.x));\n");
4421 shader_addline(buffer, "tmp0.xyz = tmp0.xyz * vec3(srgb_const0.y) - vec3(srgb_const0.z);\n");
4422 shader_addline(buffer, "tmp1.xyz = gl_FragData[0].xyz * vec3(srgb_const0.w);\n");
4423 shader_addline(buffer, "bvec3 srgb_compare = lessThan(gl_FragData[0].xyz, vec3(srgb_const1.x));\n");
4424 shader_addline(buffer, "gl_FragData[0].xyz = mix(tmp0.xyz, tmp1.xyz, vec3(srgb_compare));\n");
4425 shader_addline(buffer, "gl_FragData[0] = clamp(gl_FragData[0], 0.0, 1.0);\n");
4428 static void shader_glsl_generate_fog_code(struct wined3d_shader_buffer *buffer, enum fogmode mode)
4436 /* Fog = (gl_Fog.end - gl_FogFragCoord) / (gl_Fog.end - gl_Fog.start) */
4437 shader_addline(buffer, "float Fog = (gl_Fog.end - gl_FogFragCoord) / (gl_Fog.end - gl_Fog.start);\n");
4441 /* Fog = e^-(gl_Fog.density * gl_FogFragCoord) */
4442 shader_addline(buffer, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4446 /* Fog = e^-((gl_Fog.density * gl_FogFragCoord)^2) */
4447 shader_addline(buffer, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4451 ERR("Invalid fog mode %#x.\n", mode);
4455 shader_addline(buffer, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, clamp(Fog, 0.0, 1.0));\n");
4458 /* Context activation is done by the caller. */
4459 static GLuint shader_glsl_generate_pshader(const struct wined3d_context *context,
4460 struct wined3d_shader_buffer *buffer, const struct wined3d_shader *shader,
4461 const struct ps_compile_args *args, struct ps_np2fixup_info *np2fixup_info)
4463 const struct wined3d_shader_reg_maps *reg_maps = &shader->reg_maps;
4464 const struct wined3d_gl_info *gl_info = context->gl_info;
4465 const DWORD *function = shader->function;
4466 struct shader_glsl_ctx_priv priv_ctx;
4468 /* Create the hw GLSL shader object and assign it as the shader->prgId */
4469 GLhandleARB shader_obj = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
4471 memset(&priv_ctx, 0, sizeof(priv_ctx));
4472 priv_ctx.cur_ps_args = args;
4473 priv_ctx.cur_np2fixup_info = np2fixup_info;
4475 shader_addline(buffer, "#version 120\n");
4477 if (gl_info->supported[ARB_SHADER_BIT_ENCODING])
4478 shader_addline(buffer, "#extension GL_ARB_shader_bit_encoding : enable\n");
4479 if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
4480 shader_addline(buffer, "#extension GL_ARB_shader_texture_lod : enable\n");
4481 /* The spec says that it doesn't have to be explicitly enabled, but the
4482 * nvidia drivers write a warning if we don't do so. */
4483 if (gl_info->supported[ARB_TEXTURE_RECTANGLE])
4484 shader_addline(buffer, "#extension GL_ARB_texture_rectangle : enable\n");
4485 if (gl_info->supported[EXT_GPU_SHADER4])
4486 shader_addline(buffer, "#extension GL_EXT_gpu_shader4 : enable\n");
4488 /* Base Declarations */
4489 shader_generate_glsl_declarations(context, buffer, shader, reg_maps, &priv_ctx);
4491 /* Pack 3.0 inputs */
4492 if (reg_maps->shader_version.major >= 3 && args->vp_mode != vertexshader)
4493 shader_glsl_input_pack(shader, buffer, shader->input_signature, reg_maps, args->vp_mode);
4495 /* Base Shader Body */
4496 shader_generate_main(shader, buffer, reg_maps, function, &priv_ctx);
4498 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
4499 if (reg_maps->shader_version.major < 2)
4501 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
4502 shader_addline(buffer, "gl_FragData[0] = R0;\n");
4505 if (args->srgb_correction)
4506 shader_glsl_generate_srgb_write_correction(buffer);
4508 /* SM < 3 does not replace the fog stage. */
4509 if (reg_maps->shader_version.major < 3)
4510 shader_glsl_generate_fog_code(buffer, args->fog);
4512 shader_addline(buffer, "}\n");
4514 TRACE("Compiling shader object %u\n", shader_obj);
4515 shader_glsl_compile(gl_info, shader_obj, buffer->buffer);
4517 /* Store the shader object */
4521 /* Context activation is done by the caller. */
4522 static GLuint shader_glsl_generate_vshader(const struct wined3d_context *context,
4523 struct wined3d_shader_buffer *buffer, const struct wined3d_shader *shader,
4524 const struct vs_compile_args *args)
4526 const struct wined3d_shader_reg_maps *reg_maps = &shader->reg_maps;
4527 const struct wined3d_gl_info *gl_info = context->gl_info;
4528 const DWORD *function = shader->function;
4529 struct shader_glsl_ctx_priv priv_ctx;
4531 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4532 GLhandleARB shader_obj = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
4534 shader_addline(buffer, "#version 120\n");
4536 if (gl_info->supported[ARB_SHADER_BIT_ENCODING])
4537 shader_addline(buffer, "#extension GL_ARB_shader_bit_encoding : enable\n");
4538 if (gl_info->supported[EXT_GPU_SHADER4])
4539 shader_addline(buffer, "#extension GL_EXT_gpu_shader4 : enable\n");
4541 memset(&priv_ctx, 0, sizeof(priv_ctx));
4542 priv_ctx.cur_vs_args = args;
4544 /* Base Declarations */
4545 shader_generate_glsl_declarations(context, buffer, shader, reg_maps, &priv_ctx);
4547 /* Base Shader Body */
4548 shader_generate_main(shader, buffer, reg_maps, function, &priv_ctx);
4550 /* Unpack outputs */
4551 shader_addline(buffer, "order_ps_input(vs_out);\n");
4553 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4554 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4555 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4556 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4558 if (args->fog_src == VS_FOG_Z)
4559 shader_addline(buffer, "gl_FogFragCoord = gl_Position.z;\n");
4560 else if (!reg_maps->fog)
4561 shader_addline(buffer, "gl_FogFragCoord = 0.0;\n");
4563 /* We always store the clipplanes without y inversion */
4564 if (args->clip_enabled)
4565 shader_addline(buffer, "gl_ClipVertex = gl_Position;\n");
4567 /* Write the final position.
4569 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4570 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4571 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4572 * contains 1.0 to allow a mad.
4574 shader_addline(buffer, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4575 shader_addline(buffer, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4577 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4579 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4580 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4581 * which is the same as z = z * 2 - w.
4583 shader_addline(buffer, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4585 shader_addline(buffer, "}\n");
4587 TRACE("Compiling shader object %u\n", shader_obj);
4588 shader_glsl_compile(gl_info, shader_obj, buffer->buffer);
4593 /* Context activation is done by the caller. */
4594 static GLhandleARB shader_glsl_generate_geometry_shader(const struct wined3d_context *context,
4595 struct wined3d_shader_buffer *buffer, const struct wined3d_shader *shader)
4597 const struct wined3d_shader_reg_maps *reg_maps = &shader->reg_maps;
4598 const struct wined3d_gl_info *gl_info = context->gl_info;
4599 const DWORD *function = shader->function;
4600 struct shader_glsl_ctx_priv priv_ctx;
4601 GLhandleARB shader_id;
4603 shader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_GEOMETRY_SHADER_ARB));
4605 shader_addline(buffer, "#version 120\n");
4607 if (gl_info->supported[ARB_GEOMETRY_SHADER4])
4608 shader_addline(buffer, "#extension GL_ARB_geometry_shader4 : enable\n");
4609 if (gl_info->supported[ARB_SHADER_BIT_ENCODING])
4610 shader_addline(buffer, "#extension GL_ARB_shader_bit_encoding : enable\n");
4611 if (gl_info->supported[EXT_GPU_SHADER4])
4612 shader_addline(buffer, "#extension GL_EXT_gpu_shader4 : enable\n");
4614 memset(&priv_ctx, 0, sizeof(priv_ctx));
4615 shader_generate_glsl_declarations(context, buffer, shader, reg_maps, &priv_ctx);
4616 shader_generate_main(shader, buffer, reg_maps, function, &priv_ctx);
4617 shader_addline(buffer, "}\n");
4619 TRACE("Compiling shader object %u.\n", shader_id);
4620 shader_glsl_compile(gl_info, shader_id, buffer->buffer);
4625 static GLhandleARB find_glsl_pshader(const struct wined3d_context *context,
4626 struct wined3d_shader_buffer *buffer, struct wined3d_shader *shader,
4627 const struct ps_compile_args *args, const struct ps_np2fixup_info **np2fixup_info)
4629 struct wined3d_state *state = &shader->device->stateBlock->state;
4630 struct glsl_ps_compiled_shader *gl_shaders, *new_array;
4631 struct glsl_shader_private *shader_data;
4632 struct ps_np2fixup_info *np2fixup;
4637 if (!shader->backend_data)
4639 shader->backend_data = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*shader_data));
4640 if (!shader->backend_data)
4642 ERR("Failed to allocate backend data.\n");
4646 shader_data = shader->backend_data;
4647 gl_shaders = shader_data->gl_shaders.ps;
4649 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4650 * so a linear search is more performant than a hashmap or a binary search
4651 * (cache coherency etc)
4653 for (i = 0; i < shader_data->num_gl_shaders; ++i)
4655 if (!memcmp(&gl_shaders[i].args, args, sizeof(*args)))
4657 if (args->np2_fixup)
4658 *np2fixup_info = &gl_shaders[i].np2fixup;
4659 return gl_shaders[i].prgId;
4663 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader);
4664 if(shader_data->shader_array_size == shader_data->num_gl_shaders) {
4665 if (shader_data->num_gl_shaders)
4667 new_size = shader_data->shader_array_size + max(1, shader_data->shader_array_size / 2);
4668 new_array = HeapReAlloc(GetProcessHeap(), 0, shader_data->gl_shaders.ps,
4669 new_size * sizeof(*gl_shaders));
4673 new_array = HeapAlloc(GetProcessHeap(), 0, sizeof(*gl_shaders));
4678 ERR("Out of memory\n");
4681 shader_data->gl_shaders.ps = new_array;
4682 shader_data->shader_array_size = new_size;
4683 gl_shaders = new_array;
4686 gl_shaders[shader_data->num_gl_shaders].args = *args;
4688 np2fixup = &gl_shaders[shader_data->num_gl_shaders].np2fixup;
4689 memset(np2fixup, 0, sizeof(*np2fixup));
4690 *np2fixup_info = args->np2_fixup ? np2fixup : NULL;
4692 pixelshader_update_samplers(&shader->reg_maps, state->textures);
4694 shader_buffer_clear(buffer);
4695 ret = shader_glsl_generate_pshader(context, buffer, shader, args, np2fixup);
4696 gl_shaders[shader_data->num_gl_shaders++].prgId = ret;
4701 static inline BOOL vs_args_equal(const struct vs_compile_args *stored, const struct vs_compile_args *new,
4702 const DWORD use_map) {
4703 if((stored->swizzle_map & use_map) != new->swizzle_map) return FALSE;
4704 if((stored->clip_enabled) != new->clip_enabled) return FALSE;
4705 return stored->fog_src == new->fog_src;
4708 static GLhandleARB find_glsl_vshader(const struct wined3d_context *context,
4709 struct wined3d_shader_buffer *buffer, struct wined3d_shader *shader,
4710 const struct vs_compile_args *args)
4714 DWORD use_map = shader->device->strided_streams.use_map;
4715 struct glsl_vs_compiled_shader *gl_shaders, *new_array;
4716 struct glsl_shader_private *shader_data;
4719 if (!shader->backend_data)
4721 shader->backend_data = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*shader_data));
4722 if (!shader->backend_data)
4724 ERR("Failed to allocate backend data.\n");
4728 shader_data = shader->backend_data;
4729 gl_shaders = shader_data->gl_shaders.vs;
4731 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4732 * so a linear search is more performant than a hashmap or a binary search
4733 * (cache coherency etc)
4735 for (i = 0; i < shader_data->num_gl_shaders; ++i)
4737 if (vs_args_equal(&gl_shaders[i].args, args, use_map))
4738 return gl_shaders[i].prgId;
4741 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader);
4743 if(shader_data->shader_array_size == shader_data->num_gl_shaders) {
4744 if (shader_data->num_gl_shaders)
4746 new_size = shader_data->shader_array_size + max(1, shader_data->shader_array_size / 2);
4747 new_array = HeapReAlloc(GetProcessHeap(), 0, shader_data->gl_shaders.vs,
4748 new_size * sizeof(*gl_shaders));
4752 new_array = HeapAlloc(GetProcessHeap(), 0, sizeof(*gl_shaders));
4757 ERR("Out of memory\n");
4760 shader_data->gl_shaders.vs = new_array;
4761 shader_data->shader_array_size = new_size;
4762 gl_shaders = new_array;
4765 gl_shaders[shader_data->num_gl_shaders].args = *args;
4767 shader_buffer_clear(buffer);
4768 ret = shader_glsl_generate_vshader(context, buffer, shader, args);
4769 gl_shaders[shader_data->num_gl_shaders++].prgId = ret;
4774 static GLhandleARB find_glsl_geometry_shader(const struct wined3d_context *context,
4775 struct wined3d_shader_buffer *buffer, struct wined3d_shader *shader)
4777 struct glsl_gs_compiled_shader *gl_shaders;
4778 struct glsl_shader_private *shader_data;
4781 if (!shader->backend_data)
4783 if (!(shader->backend_data = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*shader_data))))
4785 ERR("Failed to allocate backend data.\n");
4789 shader_data = shader->backend_data;
4790 gl_shaders = shader_data->gl_shaders.gs;
4792 if (shader_data->num_gl_shaders)
4793 return gl_shaders[0].id;
4795 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader);
4797 if (!(shader_data->gl_shaders.gs = HeapAlloc(GetProcessHeap(), 0, sizeof(*gl_shaders))))
4799 ERR("Failed to allocate GL shader array.\n");
4802 shader_data->shader_array_size = 1;
4803 gl_shaders = shader_data->gl_shaders.gs;
4805 shader_buffer_clear(buffer);
4806 ret = shader_glsl_generate_geometry_shader(context, buffer, shader);
4807 gl_shaders[shader_data->num_gl_shaders++].id = ret;
4812 static const char *shader_glsl_get_ffp_fragment_op_arg(struct wined3d_shader_buffer *buffer,
4813 DWORD argnum, unsigned int stage, DWORD arg)
4817 if (arg == ARG_UNUSED)
4818 return "<unused arg>";
4820 switch (arg & WINED3DTA_SELECTMASK)
4822 case WINED3DTA_DIFFUSE:
4826 case WINED3DTA_CURRENT:
4833 case WINED3DTA_TEXTURE:
4836 case 0: ret = "tex0"; break;
4837 case 1: ret = "tex1"; break;
4838 case 2: ret = "tex2"; break;
4839 case 3: ret = "tex3"; break;
4840 case 4: ret = "tex4"; break;
4841 case 5: ret = "tex5"; break;
4842 case 6: ret = "tex6"; break;
4843 case 7: ret = "tex7"; break;
4845 ret = "<invalid texture>";
4850 case WINED3DTA_TFACTOR:
4854 case WINED3DTA_SPECULAR:
4855 ret = "gl_SecondaryColor";
4858 case WINED3DTA_TEMP:
4862 case WINED3DTA_CONSTANT:
4863 FIXME("Per-stage constants not implemented.\n");
4866 case 0: ret = "const0"; break;
4867 case 1: ret = "const1"; break;
4868 case 2: ret = "const2"; break;
4869 case 3: ret = "const3"; break;
4870 case 4: ret = "const4"; break;
4871 case 5: ret = "const5"; break;
4872 case 6: ret = "const6"; break;
4873 case 7: ret = "const7"; break;
4875 ret = "<invalid constant>";
4881 return "<unhandled arg>";
4884 if (arg & WINED3DTA_COMPLEMENT)
4886 shader_addline(buffer, "arg%u = vec4(1.0) - %s;\n", argnum, ret);
4889 else if (argnum == 1)
4891 else if (argnum == 2)
4895 if (arg & WINED3DTA_ALPHAREPLICATE)
4897 shader_addline(buffer, "arg%u = vec4(%s.w);\n", argnum, ret);
4900 else if (argnum == 1)
4902 else if (argnum == 2)
4909 static void shader_glsl_ffp_fragment_op(struct wined3d_shader_buffer *buffer, unsigned int stage, BOOL color,
4910 BOOL alpha, DWORD dst, DWORD op, DWORD dw_arg0, DWORD dw_arg1, DWORD dw_arg2)
4912 const char *dstmask, *dstreg, *arg0, *arg1, *arg2;
4922 dstreg = "temp_reg";
4926 arg0 = shader_glsl_get_ffp_fragment_op_arg(buffer, 0, stage, dw_arg0);
4927 arg1 = shader_glsl_get_ffp_fragment_op_arg(buffer, 1, stage, dw_arg1);
4928 arg2 = shader_glsl_get_ffp_fragment_op_arg(buffer, 2, stage, dw_arg2);
4932 case WINED3D_TOP_DISABLE:
4934 shader_addline(buffer, "%s%s = gl_Color%s;\n", dstreg, dstmask, dstmask);
4937 case WINED3D_TOP_SELECT_ARG1:
4938 shader_addline(buffer, "%s%s = %s%s;\n", dstreg, dstmask, arg1, dstmask);
4941 case WINED3D_TOP_SELECT_ARG2:
4942 shader_addline(buffer, "%s%s = %s%s;\n", dstreg, dstmask, arg2, dstmask);
4945 case WINED3D_TOP_MODULATE:
4946 shader_addline(buffer, "%s%s = %s%s * %s%s;\n", dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4949 case WINED3D_TOP_MODULATE_4X:
4950 shader_addline(buffer, "%s%s = clamp(%s%s * %s%s * 4.0, 0.0, 1.0);\n",
4951 dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4954 case WINED3D_TOP_MODULATE_2X:
4955 shader_addline(buffer, "%s%s = clamp(%s%s * %s%s * 2.0, 0.0, 1.0);\n",
4956 dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4959 case WINED3D_TOP_ADD:
4960 shader_addline(buffer, "%s%s = clamp(%s%s + %s%s, 0.0, 1.0);\n",
4961 dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4964 case WINED3D_TOP_ADD_SIGNED:
4965 shader_addline(buffer, "%s%s = clamp(%s%s + (%s - vec4(0.5))%s, 0.0, 1.0);\n",
4966 dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4969 case WINED3D_TOP_ADD_SIGNED_2X:
4970 shader_addline(buffer, "%s%s = clamp((%s%s + (%s - vec4(0.5))%s) * 2.0, 0.0, 1.0);\n",
4971 dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4974 case WINED3D_TOP_SUBTRACT:
4975 shader_addline(buffer, "%s%s = clamp(%s%s - %s%s, 0.0, 1.0);\n",
4976 dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4979 case WINED3D_TOP_ADD_SMOOTH:
4980 shader_addline(buffer, "%s%s = clamp((vec4(1.0) - %s)%s * %s%s + %s%s, 0.0, 1.0);\n",
4981 dstreg, dstmask, arg1, dstmask, arg2, dstmask, arg1, dstmask);
4984 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA:
4985 arg0 = shader_glsl_get_ffp_fragment_op_arg(buffer, 0, stage, WINED3DTA_DIFFUSE);
4986 shader_addline(buffer, "%s%s = mix(%s%s, %s%s, %s.w);\n",
4987 dstreg, dstmask, arg2, dstmask, arg1, dstmask, arg0);
4990 case WINED3D_TOP_BLEND_TEXTURE_ALPHA:
4991 arg0 = shader_glsl_get_ffp_fragment_op_arg(buffer, 0, stage, WINED3DTA_TEXTURE);
4992 shader_addline(buffer, "%s%s = mix(%s%s, %s%s, %s.w);\n",
4993 dstreg, dstmask, arg2, dstmask, arg1, dstmask, arg0);
4996 case WINED3D_TOP_BLEND_FACTOR_ALPHA:
4997 arg0 = shader_glsl_get_ffp_fragment_op_arg(buffer, 0, stage, WINED3DTA_TFACTOR);
4998 shader_addline(buffer, "%s%s = mix(%s%s, %s%s, %s.w);\n",
4999 dstreg, dstmask, arg2, dstmask, arg1, dstmask, arg0);
5002 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM:
5003 arg0 = shader_glsl_get_ffp_fragment_op_arg(buffer, 0, stage, WINED3DTA_TEXTURE);
5004 shader_addline(buffer, "%s%s = clamp(%s%s * (1.0 - %s.w) + %s%s, 0.0, 1.0);\n",
5005 dstreg, dstmask, arg2, dstmask, arg0, arg1, dstmask);
5008 case WINED3D_TOP_BLEND_CURRENT_ALPHA:
5009 arg0 = shader_glsl_get_ffp_fragment_op_arg(buffer, 0, stage, WINED3DTA_CURRENT);
5010 shader_addline(buffer, "%s%s = mix(%s%s, %s%s, %s.w);\n",
5011 dstreg, dstmask, arg2, dstmask, arg1, dstmask, arg0);
5014 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR:
5015 shader_addline(buffer, "%s%s = clamp(%s%s * %s.w + %s%s, 0.0, 1.0);\n",
5016 dstreg, dstmask, arg2, dstmask, arg1, arg1, dstmask);
5019 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA:
5020 shader_addline(buffer, "%s%s = clamp(%s%s * %s%s + %s.w, 0.0, 1.0);\n",
5021 dstreg, dstmask, arg1, dstmask, arg2, dstmask, arg1);
5024 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR:
5025 shader_addline(buffer, "%s%s = clamp(%s%s * (1.0 - %s.w) + %s%s, 0.0, 1.0);\n",
5026 dstreg, dstmask, arg2, dstmask, arg1, arg1, dstmask);
5028 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA:
5029 shader_addline(buffer, "%s%s = clamp((vec4(1.0) - %s)%s * %s%s + %s.w, 0.0, 1.0);\n",
5030 dstreg, dstmask, arg1, dstmask, arg2, dstmask, arg1);
5033 case WINED3D_TOP_BUMPENVMAP:
5034 case WINED3D_TOP_BUMPENVMAP_LUMINANCE:
5035 /* These are handled in the first pass, nothing to do. */
5038 case WINED3D_TOP_DOTPRODUCT3:
5039 shader_addline(buffer, "%s%s = vec4(clamp(dot(%s.xyz - 0.5, %s.xyz - 0.5) * 4.0, 0.0, 1.0))%s;\n",
5040 dstreg, dstmask, arg1, arg2, dstmask);
5043 case WINED3D_TOP_MULTIPLY_ADD:
5044 shader_addline(buffer, "%s%s = clamp(%s%s * %s%s + %s%s, 0.0, 1.0);\n",
5045 dstreg, dstmask, arg1, dstmask, arg2, dstmask, arg0, dstmask);
5048 case WINED3D_TOP_LERP:
5049 /* MSDN isn't quite right here. */
5050 shader_addline(buffer, "%s%s = mix(%s%s, %s%s, %s%s);\n",
5051 dstreg, dstmask, arg2, dstmask, arg1, dstmask, arg0, dstmask);
5055 FIXME("Unhandled operation %#x.\n", op);
5060 /* Context activation is done by the caller. */
5061 static GLuint shader_glsl_generate_ffp_fragment_shader(struct wined3d_shader_buffer *buffer,
5062 const struct ffp_frag_settings *settings, const struct wined3d_gl_info *gl_info)
5064 BOOL tempreg_used = FALSE, tfactor_used = FALSE;
5065 BYTE lum_map = 0, bump_map = 0, tex_map = 0;
5066 const char *final_combiner_src = "ret";
5067 UINT lowest_disabled_stage;
5068 GLhandleARB shader_obj;
5069 DWORD arg0, arg1, arg2;
5072 shader_buffer_clear(buffer);
5074 /* Find out which textures are read */
5075 for (stage = 0; stage < MAX_TEXTURES; ++stage)
5077 if (settings->op[stage].cop == WINED3D_TOP_DISABLE)
5080 arg0 = settings->op[stage].carg0 & WINED3DTA_SELECTMASK;
5081 arg1 = settings->op[stage].carg1 & WINED3DTA_SELECTMASK;
5082 arg2 = settings->op[stage].carg2 & WINED3DTA_SELECTMASK;
5084 if (arg0 == WINED3DTA_TEXTURE || arg1 == WINED3DTA_TEXTURE || arg2 == WINED3DTA_TEXTURE)
5085 tex_map |= 1 << stage;
5086 if (arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR)
5087 tfactor_used = TRUE;
5088 if (arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP)
5089 tempreg_used = TRUE;
5090 if (settings->op[stage].dst == tempreg)
5091 tempreg_used = TRUE;
5093 switch (settings->op[stage].cop)
5095 case WINED3D_TOP_BUMPENVMAP_LUMINANCE:
5096 lum_map |= 1 << stage;
5098 case WINED3D_TOP_BUMPENVMAP:
5099 bump_map |= 1 << stage;
5101 case WINED3D_TOP_BLEND_TEXTURE_ALPHA:
5102 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM:
5103 tex_map |= 1 << stage;
5106 case WINED3D_TOP_BLEND_FACTOR_ALPHA:
5107 tfactor_used = TRUE;
5114 if (settings->op[stage].aop == WINED3D_TOP_DISABLE)
5117 arg0 = settings->op[stage].aarg0 & WINED3DTA_SELECTMASK;
5118 arg1 = settings->op[stage].aarg1 & WINED3DTA_SELECTMASK;
5119 arg2 = settings->op[stage].aarg2 & WINED3DTA_SELECTMASK;
5121 if (arg0 == WINED3DTA_TEXTURE || arg1 == WINED3DTA_TEXTURE || arg2 == WINED3DTA_TEXTURE)
5122 tex_map |= 1 << stage;
5123 if (arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR)
5124 tfactor_used = TRUE;
5125 if (arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP)
5126 tempreg_used = TRUE;
5128 lowest_disabled_stage = stage;
5130 shader_addline(buffer, "#version 120\n");
5132 shader_addline(buffer, "vec4 tmp0, tmp1;\n");
5133 shader_addline(buffer, "vec4 ret;\n");
5134 if (tempreg_used || settings->sRGB_write)
5135 shader_addline(buffer, "vec4 temp_reg;\n");
5136 shader_addline(buffer, "vec4 arg0, arg1, arg2;\n");
5138 for (stage = 0; stage < MAX_TEXTURES; ++stage)
5140 if (!(tex_map & (1 << stage)))
5143 switch (settings->op[stage].tex_type)
5146 shader_addline(buffer, "uniform sampler1D ps_sampler%u;\n", stage);
5149 shader_addline(buffer, "uniform sampler2D ps_sampler%u;\n", stage);
5152 shader_addline(buffer, "uniform sampler3D ps_sampler%u;\n", stage);
5155 shader_addline(buffer, "uniform samplerCube ps_sampler%u;\n", stage);
5158 shader_addline(buffer, "uniform sampler2DRect ps_sampler%u;\n", stage);
5161 FIXME("Unhandled sampler type %#x.\n", settings->op[stage].tex_type);
5165 shader_addline(buffer, "vec4 tex%u;\n", stage);
5167 if (!(bump_map & (1 << stage)))
5169 shader_addline(buffer, "uniform mat2 bumpenv_mat%u;\n", stage);
5171 if (!(lum_map & (1 << stage)))
5173 shader_addline(buffer, "uniform float bumpenv_lum_scale%u;\n", stage);
5174 shader_addline(buffer, "uniform float bumpenv_lum_offset%u;\n", stage);
5177 shader_addline(buffer, "uniform vec4 tex_factor;\n");
5178 shader_addline(buffer, "uniform vec4 specular_enable;\n");
5180 if (settings->sRGB_write)
5182 shader_addline(buffer, "const vec4 srgb_const0 = vec4(%.8e, %.8e, %.8e, %.8e);\n",
5183 srgb_pow, srgb_mul_high, srgb_sub_high, srgb_mul_low);
5184 shader_addline(buffer, "const vec4 srgb_const1 = vec4(%.8e, 0.0, 0.0, 0.0);\n",
5188 shader_addline(buffer, "void main()\n{\n");
5190 if (lowest_disabled_stage < 7 && settings->emul_clipplanes)
5191 shader_addline(buffer, "if (any(lessThan(gl_texCoord[7], vec4(0.0)))) discard;\n");
5193 /* Generate texture sampling instructions) */
5194 for (stage = 0; stage < MAX_TEXTURES && settings->op[stage].cop != WINED3D_TOP_DISABLE; ++stage)
5196 const char *texture_function, *coord_mask;
5197 char tex_reg_name[8];
5200 if (!(tex_map & (1 << stage)))
5203 if (settings->op[stage].projected == proj_none)
5207 else if (settings->op[stage].projected == proj_count4
5208 || settings->op[stage].projected == proj_count3)
5214 FIXME("Unexpected projection mode %d\n", settings->op[stage].projected);
5218 if (settings->op[stage].cop == WINED3D_TOP_BUMPENVMAP
5219 || settings->op[stage].cop == WINED3D_TOP_BUMPENVMAP_LUMINANCE)
5224 switch (settings->op[stage].tex_type)
5229 texture_function = "texture1DProj";
5234 texture_function = "texture1D";
5241 texture_function = "texture2DProj";
5246 texture_function = "texture2D";
5253 texture_function = "texture3DProj";
5254 coord_mask = "xyzw";
5258 texture_function = "texture3D";
5263 texture_function = "textureCube";
5269 texture_function = "texture2DRectProj";
5274 texture_function = "texture2DRect";
5279 FIXME("Unhandled texture type %#x.\n", settings->op[stage].tex_type);
5280 texture_function = "";
5281 coord_mask = "xyzw";
5286 && (settings->op[stage - 1].cop == WINED3D_TOP_BUMPENVMAP
5287 || settings->op[stage - 1].cop == WINED3D_TOP_BUMPENVMAP_LUMINANCE))
5289 shader_addline(buffer, "ret.xy = bumpenv_mat%u * tex%u.xy;\n", stage - 1, stage - 1);
5291 /* With projective textures, texbem only divides the static
5292 * texture coord, not the displacement, so multiply the
5293 * displacement with the dividing parameter before passing it to
5295 if (settings->op[stage].projected != proj_none)
5297 if (settings->op[stage].projected == proj_count4)
5299 shader_addline(buffer, "ret.xy = (ret.xy * gl_TexCoord[%u].w) + gl_TexCoord[%u].xy;\n",
5301 shader_addline(buffer, "ret.zw = gl_TexCoord[%u].ww;\n", stage);
5305 shader_addline(buffer, "ret.xy = (ret.xy * gl_TexCoord[%u].z) + gl_TexCoord[%u].xy;\n",
5307 shader_addline(buffer, "ret.zw = gl_TexCoord[%u].zz;\n", stage);
5312 shader_addline(buffer, "ret = gl_TexCoord[%u] + ret.xyxy;\n", stage);
5316 shader_addline(buffer, "tex%u = clamp(%s(ps_sampler%u, ret.%s), 0.0, 1.0);\n",
5317 stage, texture_function, stage, coord_mask);
5319 shader_addline(buffer, "tex%u = %s(ps_sampler%u, ret.%s);\n",
5320 stage, texture_function, stage, coord_mask);
5322 if (settings->op[stage - 1].cop == WINED3D_TOP_BUMPENVMAP_LUMINANCE)
5323 shader_addline(buffer, "tex%u *= clamp(tex%u.z * bumpenv_lum_scale%u + bumpenv_lum_offset%u, 0.0, 1.0);\n",
5324 stage, stage - 1, stage - 1, stage - 1);
5326 else if (settings->op[stage].projected == proj_count3)
5329 shader_addline(buffer, "tex%u = clamp(%s(ps_sampler%u, gl_TexCoord[%u].xyz), 0.0, 1.0);\n",
5330 stage, texture_function, stage, stage);
5332 shader_addline(buffer, "tex%u = %s(ps_sampler%u, gl_TexCoord[%u].xyz);\n",
5333 stage, texture_function, stage, stage);
5338 shader_addline(buffer, "tex%u = clamp(%s(ps_sampler%u, gl_TexCoord[%u].%s), 0.0, 1.0);\n",
5339 stage, texture_function, stage, stage, coord_mask);
5341 shader_addline(buffer, "tex%u = %s(ps_sampler%u, gl_TexCoord[%u].%s);\n",
5342 stage, texture_function, stage, stage, coord_mask);
5345 sprintf(tex_reg_name, "tex%u", stage);
5346 shader_glsl_color_correction_ext(buffer, tex_reg_name, WINED3DSP_WRITEMASK_ALL,
5347 settings->op[stage].color_fixup);
5350 /* Generate the main shader */
5351 for (stage = 0; stage < MAX_TEXTURES; ++stage)
5355 if (settings->op[stage].cop == WINED3D_TOP_DISABLE)
5358 final_combiner_src = "gl_Color";
5362 if (settings->op[stage].cop == WINED3D_TOP_SELECT_ARG1
5363 && settings->op[stage].aop == WINED3D_TOP_SELECT_ARG1)
5364 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg1;
5365 else if (settings->op[stage].cop == WINED3D_TOP_SELECT_ARG1
5366 && settings->op[stage].aop == WINED3D_TOP_SELECT_ARG2)
5367 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg2;
5368 else if (settings->op[stage].cop == WINED3D_TOP_SELECT_ARG2
5369 && settings->op[stage].aop == WINED3D_TOP_SELECT_ARG1)
5370 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg1;
5371 else if (settings->op[stage].cop == WINED3D_TOP_SELECT_ARG2
5372 && settings->op[stage].aop == WINED3D_TOP_SELECT_ARG2)
5373 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg2;
5375 op_equal = settings->op[stage].aop == settings->op[stage].cop
5376 && settings->op[stage].carg0 == settings->op[stage].aarg0
5377 && settings->op[stage].carg1 == settings->op[stage].aarg1
5378 && settings->op[stage].carg2 == settings->op[stage].aarg2;
5380 if (settings->op[stage].aop == WINED3D_TOP_DISABLE)
5382 shader_glsl_ffp_fragment_op(buffer, stage, TRUE, FALSE, settings->op[stage].dst,
5383 settings->op[stage].cop, settings->op[stage].carg0,
5384 settings->op[stage].carg1, settings->op[stage].carg2);
5386 shader_addline(buffer, "ret.w = gl_Color.w;\n");
5390 shader_glsl_ffp_fragment_op(buffer, stage, TRUE, TRUE, settings->op[stage].dst,
5391 settings->op[stage].cop, settings->op[stage].carg0,
5392 settings->op[stage].carg1, settings->op[stage].carg2);
5396 shader_glsl_ffp_fragment_op(buffer, stage, TRUE, FALSE, settings->op[stage].dst,
5397 settings->op[stage].cop, settings->op[stage].carg0,
5398 settings->op[stage].carg1, settings->op[stage].carg2);
5399 shader_glsl_ffp_fragment_op(buffer, stage, FALSE, TRUE, settings->op[stage].dst,
5400 settings->op[stage].aop, settings->op[stage].aarg0,
5401 settings->op[stage].aarg1, settings->op[stage].aarg2);
5405 shader_addline(buffer, "gl_FragData[0] = gl_SecondaryColor * specular_enable + %s;\n", final_combiner_src);
5407 if (settings->sRGB_write)
5408 shader_glsl_generate_srgb_write_correction(buffer);
5410 shader_glsl_generate_fog_code(buffer, settings->fog);
5412 shader_addline(buffer, "}\n");
5414 shader_obj = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
5415 shader_glsl_compile(gl_info, shader_obj, buffer->buffer);
5419 static struct glsl_ffp_fragment_shader *shader_glsl_find_ffp_fragment_shader(struct shader_glsl_priv *priv,
5420 const struct wined3d_gl_info *gl_info, const struct ffp_frag_settings *args)
5422 struct glsl_ffp_fragment_shader *glsl_desc;
5423 const struct ffp_frag_desc *desc;
5425 if ((desc = find_ffp_frag_shader(&priv->ffp_fragment_shaders, args)))
5426 return CONTAINING_RECORD(desc, struct glsl_ffp_fragment_shader, entry);
5428 if (!(glsl_desc = HeapAlloc(GetProcessHeap(), 0, sizeof(*glsl_desc))))
5431 glsl_desc->entry.settings = *args;
5432 glsl_desc->id = shader_glsl_generate_ffp_fragment_shader(&priv->shader_buffer, args, gl_info);
5433 list_init(&glsl_desc->linked_programs);
5434 add_ffp_frag_shader(&priv->ffp_fragment_shaders, &glsl_desc->entry);
5440 static void shader_glsl_init_vs_uniform_locations(const struct wined3d_gl_info *gl_info,
5441 GLhandleARB program_id, struct glsl_vs_program *vs)
5446 vs->uniform_f_locations = HeapAlloc(GetProcessHeap(), 0,
5447 sizeof(GLhandleARB) * gl_info->limits.glsl_vs_float_constants);
5448 for (i = 0; i < gl_info->limits.glsl_vs_float_constants; ++i)
5450 snprintf(name, sizeof(name), "vs_c[%u]", i);
5451 vs->uniform_f_locations[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5454 for (i = 0; i < MAX_CONST_I; ++i)
5456 snprintf(name, sizeof(name), "vs_i[%u]", i);
5457 vs->uniform_i_locations[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5460 vs->pos_fixup_location = GL_EXTCALL(glGetUniformLocationARB(program_id, "posFixup"));
5463 static void shader_glsl_init_ps_uniform_locations(const struct wined3d_gl_info *gl_info,
5464 GLhandleARB program_id, struct glsl_ps_program *ps)
5469 ps->uniform_f_locations = HeapAlloc(GetProcessHeap(), 0,
5470 sizeof(GLhandleARB) * gl_info->limits.glsl_ps_float_constants);
5471 for (i = 0; i < gl_info->limits.glsl_ps_float_constants; ++i)
5473 snprintf(name, sizeof(name), "ps_c[%u]", i);
5474 ps->uniform_f_locations[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5477 for (i = 0; i < MAX_CONST_I; ++i)
5479 snprintf(name, sizeof(name), "ps_i[%u]", i);
5480 ps->uniform_i_locations[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5483 for (i = 0; i < MAX_TEXTURES; ++i)
5485 snprintf(name, sizeof(name), "bumpenv_mat%u", i);
5486 ps->bumpenv_mat_location[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5487 snprintf(name, sizeof(name), "bumpenv_lum_scale%u", i);
5488 ps->bumpenv_lum_scale_location[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5489 snprintf(name, sizeof(name), "bumpenv_lum_offset%u", i);
5490 ps->bumpenv_lum_offset_location[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5493 ps->tex_factor_location = GL_EXTCALL(glGetUniformLocationARB(program_id, "tex_factor"));
5494 ps->specular_enable_location = GL_EXTCALL(glGetUniformLocationARB(program_id, "specular_enable"));
5495 ps->np2_fixup_location = GL_EXTCALL(glGetUniformLocationARB(program_id, "ps_samplerNP2Fixup"));
5496 ps->ycorrection_location = GL_EXTCALL(glGetUniformLocationARB(program_id, "ycorrection"));
5499 /* Context activation is done by the caller. */
5500 static void set_glsl_shader_program(const struct wined3d_context *context, struct wined3d_device *device,
5501 enum wined3d_shader_mode vertex_mode, enum wined3d_shader_mode fragment_mode)
5503 const struct wined3d_state *state = &device->stateBlock->state;
5504 const struct wined3d_gl_info *gl_info = context->gl_info;
5505 const struct ps_np2fixup_info *np2fixup_info = NULL;
5506 struct shader_glsl_priv *priv = device->shader_priv;
5507 struct glsl_shader_prog_link *entry = NULL;
5508 struct wined3d_shader *vshader = NULL;
5509 struct wined3d_shader *gshader = NULL;
5510 struct wined3d_shader *pshader = NULL;
5511 GLhandleARB programId = 0;
5512 GLhandleARB reorder_shader_id = 0;
5514 struct ps_compile_args ps_compile_args;
5515 struct vs_compile_args vs_compile_args;
5516 GLhandleARB vs_id, gs_id, ps_id;
5517 struct list *ps_list;
5519 if (vertex_mode == WINED3D_SHADER_MODE_SHADER)
5521 vshader = state->vertex_shader;
5522 find_vs_compile_args(state, vshader, &vs_compile_args);
5523 vs_id = find_glsl_vshader(context, &priv->shader_buffer, vshader, &vs_compile_args);
5525 if ((gshader = state->geometry_shader))
5526 gs_id = find_glsl_geometry_shader(context, &priv->shader_buffer, gshader);
5536 if (fragment_mode == WINED3D_SHADER_MODE_SHADER)
5538 pshader = state->pixel_shader;
5539 find_ps_compile_args(state, pshader, &ps_compile_args);
5540 ps_id = find_glsl_pshader(context, &priv->shader_buffer,
5541 pshader, &ps_compile_args, &np2fixup_info);
5542 ps_list = &pshader->linked_programs;
5544 else if (fragment_mode == WINED3D_SHADER_MODE_FFP && priv->fragment_pipe == &glsl_fragment_pipe)
5546 struct glsl_ffp_fragment_shader *ffp_shader;
5547 struct ffp_frag_settings settings;
5549 gen_ffp_frag_op(device, state, &settings, FALSE);
5550 ffp_shader = shader_glsl_find_ffp_fragment_shader(priv, gl_info, &settings);
5551 ps_id = ffp_shader->id;
5552 ps_list = &ffp_shader->linked_programs;
5559 if ((!vs_id && !gs_id && !ps_id) || (entry = get_glsl_program_entry(priv, vs_id, gs_id, ps_id)))
5561 priv->glsl_program = entry;
5565 /* If we get to this point, then no matching program exists, so we create one */
5566 programId = GL_EXTCALL(glCreateProgramObjectARB());
5567 TRACE("Created new GLSL shader program %u\n", programId);
5569 /* Create the entry */
5570 entry = HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link));
5571 entry->programId = programId;
5572 entry->vs.id = vs_id;
5573 entry->gs.id = gs_id;
5574 entry->ps.id = ps_id;
5575 entry->constant_version = 0;
5576 entry->ps.np2_fixup_info = np2fixup_info;
5577 /* Add the hash table entry */
5578 add_glsl_program_entry(priv, entry);
5580 /* Set the current program */
5581 priv->glsl_program = entry;
5583 /* Attach GLSL vshader */
5586 WORD map = vshader->reg_maps.input_registers;
5589 reorder_shader_id = generate_param_reorder_function(&priv->shader_buffer, vshader, pshader, gl_info);
5590 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id, programId);
5591 GL_EXTCALL(glAttachObjectARB(programId, reorder_shader_id));
5592 checkGLcall("glAttachObjectARB");
5593 /* Flag the reorder function for deletion, then it will be freed automatically when the program
5596 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id));
5598 TRACE("Attaching GLSL shader object %u to program %u.\n", vs_id, programId);
5599 GL_EXTCALL(glAttachObjectARB(programId, vs_id));
5600 checkGLcall("glAttachObjectARB");
5602 /* Bind vertex attributes to a corresponding index number to match
5603 * the same index numbers as ARB_vertex_programs (makes loading
5604 * vertex attributes simpler). With this method, we can use the
5605 * exact same code to load the attributes later for both ARB and
5608 * We have to do this here because we need to know the Program ID
5609 * in order to make the bindings work, and it has to be done prior
5610 * to linking the GLSL program. */
5611 for (i = 0; map; map >>= 1, ++i)
5613 if (!(map & 1)) continue;
5615 snprintf(tmp_name, sizeof(tmp_name), "vs_in%u", i);
5616 GL_EXTCALL(glBindAttribLocationARB(programId, i, tmp_name));
5618 checkGLcall("glBindAttribLocationARB");
5620 list_add_head(&vshader->linked_programs, &entry->vs.shader_entry);
5625 TRACE("Attaching GLSL geometry shader object %u to program %u.\n", gs_id, programId);
5626 GL_EXTCALL(glAttachObjectARB(programId, gs_id));
5627 checkGLcall("glAttachObjectARB");
5629 TRACE("input type %s, output type %s, vertices out %u.\n",
5630 debug_d3dprimitivetype(gshader->u.gs.input_type),
5631 debug_d3dprimitivetype(gshader->u.gs.output_type),
5632 gshader->u.gs.vertices_out);
5633 GL_EXTCALL(glProgramParameteriARB(programId, GL_GEOMETRY_INPUT_TYPE_ARB,
5634 gl_primitive_type_from_d3d(gshader->u.gs.input_type)));
5635 GL_EXTCALL(glProgramParameteriARB(programId, GL_GEOMETRY_OUTPUT_TYPE_ARB,
5636 gl_primitive_type_from_d3d(gshader->u.gs.output_type)));
5637 GL_EXTCALL(glProgramParameteriARB(programId, GL_GEOMETRY_VERTICES_OUT_ARB,
5638 gshader->u.gs.vertices_out));
5639 checkGLcall("glProgramParameteriARB");
5641 list_add_head(&gshader->linked_programs, &entry->gs.shader_entry);
5644 /* Attach GLSL pshader */
5647 TRACE("Attaching GLSL shader object %u to program %u.\n", ps_id, programId);
5648 GL_EXTCALL(glAttachObjectARB(programId, ps_id));
5649 checkGLcall("glAttachObjectARB");
5651 list_add_head(ps_list, &entry->ps.shader_entry);
5654 /* Link the program */
5655 TRACE("Linking GLSL shader program %u\n", programId);
5656 GL_EXTCALL(glLinkProgramARB(programId));
5657 shader_glsl_validate_link(gl_info, programId);
5659 shader_glsl_init_vs_uniform_locations(gl_info, programId, &entry->vs);
5660 shader_glsl_init_ps_uniform_locations(gl_info, programId, &entry->ps);
5661 checkGLcall("Find glsl program uniform locations");
5663 if (pshader && pshader->reg_maps.shader_version.major >= 3
5664 && pshader->u.ps.declared_in_count > vec4_varyings(3, gl_info))
5666 TRACE("Shader %d needs vertex color clamping disabled\n", programId);
5667 entry->vs.vertex_color_clamp = GL_FALSE;
5671 entry->vs.vertex_color_clamp = GL_FIXED_ONLY_ARB;
5674 /* Set the shader to allow uniform loading on it */
5675 GL_EXTCALL(glUseProgramObjectARB(programId));
5676 checkGLcall("glUseProgramObjectARB(programId)");
5678 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
5679 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
5680 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
5681 * vertex shader with fixed function pixel processing is used we make sure that the card
5682 * supports enough samplers to allow the max number of vertex samplers with all possible
5683 * fixed function fragment processing setups. So once the program is linked these samplers
5686 shader_glsl_load_vsamplers(gl_info, device->texUnitMap, programId);
5687 shader_glsl_load_psamplers(gl_info, device->texUnitMap, programId);
5690 /* Context activation is done by the caller. */
5691 static GLhandleARB create_glsl_blt_shader(const struct wined3d_gl_info *gl_info, enum tex_types tex_type, BOOL masked)
5693 GLhandleARB program_id;
5694 GLhandleARB vshader_id, pshader_id;
5695 const char *blt_pshader;
5697 static const char *blt_vshader =
5701 " gl_Position = gl_Vertex;\n"
5702 " gl_FrontColor = vec4(1.0);\n"
5703 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
5706 static const char * const blt_pshaders_full[tex_type_count] =
5712 "uniform sampler2D sampler;\n"
5715 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
5721 "uniform samplerCube sampler;\n"
5724 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
5728 "#extension GL_ARB_texture_rectangle : enable\n"
5729 "uniform sampler2DRect sampler;\n"
5732 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
5736 static const char * const blt_pshaders_masked[tex_type_count] =
5742 "uniform sampler2D sampler;\n"
5743 "uniform vec4 mask;\n"
5746 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
5747 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
5753 "uniform samplerCube sampler;\n"
5754 "uniform vec4 mask;\n"
5757 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
5758 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
5762 "#extension GL_ARB_texture_rectangle : enable\n"
5763 "uniform sampler2DRect sampler;\n"
5764 "uniform vec4 mask;\n"
5767 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
5768 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
5772 blt_pshader = masked ? blt_pshaders_masked[tex_type] : blt_pshaders_full[tex_type];
5775 FIXME("tex_type %#x not supported\n", tex_type);
5779 vshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
5780 shader_glsl_compile(gl_info, vshader_id, blt_vshader);
5782 pshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
5783 shader_glsl_compile(gl_info, pshader_id, blt_pshader);
5785 program_id = GL_EXTCALL(glCreateProgramObjectARB());
5786 GL_EXTCALL(glAttachObjectARB(program_id, vshader_id));
5787 GL_EXTCALL(glAttachObjectARB(program_id, pshader_id));
5788 GL_EXTCALL(glLinkProgramARB(program_id));
5790 shader_glsl_validate_link(gl_info, program_id);
5792 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
5795 GL_EXTCALL(glDeleteObjectARB(vshader_id));
5796 GL_EXTCALL(glDeleteObjectARB(pshader_id));
5800 /* Context activation is done by the caller. */
5801 static void shader_glsl_select(const struct wined3d_context *context, enum wined3d_shader_mode vertex_mode,
5802 enum wined3d_shader_mode fragment_mode)
5804 const struct wined3d_gl_info *gl_info = context->gl_info;
5805 struct wined3d_device *device = context->swapchain->device;
5806 struct shader_glsl_priv *priv = device->shader_priv;
5807 GLhandleARB program_id = 0;
5808 GLenum old_vertex_color_clamp, current_vertex_color_clamp;
5810 old_vertex_color_clamp = priv->glsl_program ? priv->glsl_program->vs.vertex_color_clamp : GL_FIXED_ONLY_ARB;
5811 set_glsl_shader_program(context, device, vertex_mode, fragment_mode);
5812 current_vertex_color_clamp = priv->glsl_program ? priv->glsl_program->vs.vertex_color_clamp : GL_FIXED_ONLY_ARB;
5813 if (old_vertex_color_clamp != current_vertex_color_clamp)
5815 if (gl_info->supported[ARB_COLOR_BUFFER_FLOAT])
5817 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB, current_vertex_color_clamp));
5818 checkGLcall("glClampColorARB");
5822 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
5826 program_id = priv->glsl_program ? priv->glsl_program->programId : 0;
5827 if (program_id) TRACE("Using GLSL program %u\n", program_id);
5828 GL_EXTCALL(glUseProgramObjectARB(program_id));
5829 checkGLcall("glUseProgramObjectARB");
5831 /* In case that NP2 texcoord fixup data is found for the selected program, trigger a reload of the
5832 * constants. This has to be done because it can't be guaranteed that sampler() (from state.c) is
5833 * called between selecting the shader and using it, which results in wrong fixup for some frames. */
5834 if (priv->glsl_program && priv->glsl_program->ps.np2_fixup_info)
5836 shader_glsl_load_np2fixup_constants(priv, gl_info, &device->stateBlock->state);
5840 /* Context activation is done by the caller. */
5841 static void shader_glsl_select_depth_blt(void *shader_priv, const struct wined3d_gl_info *gl_info,
5842 enum tex_types tex_type, const SIZE *ds_mask_size)
5844 BOOL masked = ds_mask_size->cx && ds_mask_size->cy;
5845 struct shader_glsl_priv *priv = shader_priv;
5846 GLhandleARB *blt_program;
5849 blt_program = masked ? &priv->depth_blt_program_masked[tex_type] : &priv->depth_blt_program_full[tex_type];
5852 *blt_program = create_glsl_blt_shader(gl_info, tex_type, masked);
5853 loc = GL_EXTCALL(glGetUniformLocationARB(*blt_program, "sampler"));
5854 GL_EXTCALL(glUseProgramObjectARB(*blt_program));
5855 GL_EXTCALL(glUniform1iARB(loc, 0));
5859 GL_EXTCALL(glUseProgramObjectARB(*blt_program));
5864 loc = GL_EXTCALL(glGetUniformLocationARB(*blt_program, "mask"));
5865 GL_EXTCALL(glUniform4fARB(loc, 0.0f, 0.0f, (float)ds_mask_size->cx, (float)ds_mask_size->cy));
5869 /* Context activation is done by the caller. */
5870 static void shader_glsl_deselect_depth_blt(void *shader_priv, const struct wined3d_gl_info *gl_info)
5872 struct shader_glsl_priv *priv = shader_priv;
5873 GLhandleARB program_id;
5875 program_id = priv->glsl_program ? priv->glsl_program->programId : 0;
5876 if (program_id) TRACE("Using GLSL program %u\n", program_id);
5878 GL_EXTCALL(glUseProgramObjectARB(program_id));
5879 checkGLcall("glUseProgramObjectARB");
5882 static void shader_glsl_destroy(struct wined3d_shader *shader)
5884 struct glsl_shader_private *shader_data = shader->backend_data;
5885 struct wined3d_device *device = shader->device;
5886 struct shader_glsl_priv *priv = device->shader_priv;
5887 const struct wined3d_gl_info *gl_info;
5888 const struct list *linked_programs;
5889 struct wined3d_context *context;
5891 if (!shader_data || !shader_data->num_gl_shaders)
5893 HeapFree(GetProcessHeap(), 0, shader_data);
5894 shader->backend_data = NULL;
5898 context = context_acquire(device, NULL);
5899 gl_info = context->gl_info;
5901 TRACE("Deleting linked programs.\n");
5902 linked_programs = &shader->linked_programs;
5903 if (linked_programs->next)
5905 struct glsl_shader_prog_link *entry, *entry2;
5908 switch (shader->reg_maps.shader_version.type)
5910 case WINED3D_SHADER_TYPE_PIXEL:
5912 struct glsl_ps_compiled_shader *gl_shaders = shader_data->gl_shaders.ps;
5914 LIST_FOR_EACH_ENTRY_SAFE(entry, entry2, linked_programs,
5915 struct glsl_shader_prog_link, ps.shader_entry)
5917 delete_glsl_program_entry(priv, gl_info, entry);
5920 for (i = 0; i < shader_data->num_gl_shaders; ++i)
5922 TRACE("Deleting pixel shader %u.\n", gl_shaders[i].prgId);
5923 if (priv->glsl_program && priv->glsl_program->ps.id == gl_shaders[i].prgId)
5924 shader_glsl_select(context, WINED3D_SHADER_MODE_NONE, WINED3D_SHADER_MODE_NONE);
5925 GL_EXTCALL(glDeleteObjectARB(gl_shaders[i].prgId));
5926 checkGLcall("glDeleteObjectARB");
5928 HeapFree(GetProcessHeap(), 0, shader_data->gl_shaders.ps);
5933 case WINED3D_SHADER_TYPE_VERTEX:
5935 struct glsl_vs_compiled_shader *gl_shaders = shader_data->gl_shaders.vs;
5937 LIST_FOR_EACH_ENTRY_SAFE(entry, entry2, linked_programs,
5938 struct glsl_shader_prog_link, vs.shader_entry)
5940 delete_glsl_program_entry(priv, gl_info, entry);
5943 for (i = 0; i < shader_data->num_gl_shaders; ++i)
5945 TRACE("Deleting vertex shader %u.\n", gl_shaders[i].prgId);
5946 if (priv->glsl_program && priv->glsl_program->vs.id == gl_shaders[i].prgId)
5947 shader_glsl_select(context, WINED3D_SHADER_MODE_NONE, WINED3D_SHADER_MODE_NONE);
5948 GL_EXTCALL(glDeleteObjectARB(gl_shaders[i].prgId));
5949 checkGLcall("glDeleteObjectARB");
5951 HeapFree(GetProcessHeap(), 0, shader_data->gl_shaders.vs);
5956 case WINED3D_SHADER_TYPE_GEOMETRY:
5958 struct glsl_gs_compiled_shader *gl_shaders = shader_data->gl_shaders.gs;
5960 LIST_FOR_EACH_ENTRY_SAFE(entry, entry2, linked_programs,
5961 struct glsl_shader_prog_link, gs.shader_entry)
5963 delete_glsl_program_entry(priv, gl_info, entry);
5966 for (i = 0; i < shader_data->num_gl_shaders; ++i)
5968 TRACE("Deleting geometry shader %u.\n", gl_shaders[i].id);
5969 if (priv->glsl_program && priv->glsl_program->gs.id == gl_shaders[i].id)
5970 shader_glsl_select(context, WINED3D_SHADER_MODE_NONE, WINED3D_SHADER_MODE_NONE);
5971 GL_EXTCALL(glDeleteObjectARB(gl_shaders[i].id));
5972 checkGLcall("glDeleteObjectARB");
5974 HeapFree(GetProcessHeap(), 0, shader_data->gl_shaders.gs);
5980 ERR("Unhandled shader type %#x.\n", shader->reg_maps.shader_version.type);
5985 HeapFree(GetProcessHeap(), 0, shader->backend_data);
5986 shader->backend_data = NULL;
5988 context_release(context);
5991 static int glsl_program_key_compare(const void *key, const struct wine_rb_entry *entry)
5993 const struct glsl_program_key *k = key;
5994 const struct glsl_shader_prog_link *prog = WINE_RB_ENTRY_VALUE(entry,
5995 const struct glsl_shader_prog_link, program_lookup_entry);
5997 if (k->vs_id > prog->vs.id) return 1;
5998 else if (k->vs_id < prog->vs.id) return -1;
6000 if (k->gs_id > prog->gs.id) return 1;
6001 else if (k->gs_id < prog->gs.id) return -1;
6003 if (k->ps_id > prog->ps.id) return 1;
6004 else if (k->ps_id < prog->ps.id) return -1;
6009 static BOOL constant_heap_init(struct constant_heap *heap, unsigned int constant_count)
6011 SIZE_T size = (constant_count + 1) * sizeof(*heap->entries) + constant_count * sizeof(*heap->positions);
6012 void *mem = HeapAlloc(GetProcessHeap(), 0, size);
6016 ERR("Failed to allocate memory\n");
6020 heap->entries = mem;
6021 heap->entries[1].version = 0;
6022 heap->positions = (unsigned int *)(heap->entries + constant_count + 1);
6028 static void constant_heap_free(struct constant_heap *heap)
6030 HeapFree(GetProcessHeap(), 0, heap->entries);
6033 static const struct wine_rb_functions wined3d_glsl_program_rb_functions =
6038 glsl_program_key_compare,
6041 static HRESULT shader_glsl_alloc(struct wined3d_device *device, const struct fragment_pipeline *fragment_pipe)
6043 const struct wined3d_gl_info *gl_info = &device->adapter->gl_info;
6044 struct shader_glsl_priv *priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_glsl_priv));
6045 SIZE_T stack_size = wined3d_log2i(max(gl_info->limits.glsl_vs_float_constants,
6046 gl_info->limits.glsl_ps_float_constants)) + 1;
6047 struct fragment_caps fragment_caps;
6048 void *fragment_priv;
6050 if (!(fragment_priv = fragment_pipe->alloc_private(&glsl_shader_backend, priv)))
6052 ERR("Failed to initialize fragment pipe.\n");
6053 HeapFree(GetProcessHeap(), 0, priv);
6057 if (!shader_buffer_init(&priv->shader_buffer))
6059 ERR("Failed to initialize shader buffer.\n");
6063 priv->stack = HeapAlloc(GetProcessHeap(), 0, stack_size * sizeof(*priv->stack));
6066 ERR("Failed to allocate memory.\n");
6070 if (!constant_heap_init(&priv->vconst_heap, gl_info->limits.glsl_vs_float_constants))
6072 ERR("Failed to initialize vertex shader constant heap\n");
6076 if (!constant_heap_init(&priv->pconst_heap, gl_info->limits.glsl_ps_float_constants))
6078 ERR("Failed to initialize pixel shader constant heap\n");
6082 if (wine_rb_init(&priv->program_lookup, &wined3d_glsl_program_rb_functions) == -1)
6084 ERR("Failed to initialize rbtree.\n");
6088 priv->next_constant_version = 1;
6089 fragment_pipe->get_caps(gl_info, &fragment_caps);
6090 priv->ffp_proj_control = fragment_caps.wined3d_caps & WINED3D_FRAGMENT_CAP_PROJ_CONTROL;
6091 device->fragment_priv = fragment_priv;
6092 priv->fragment_pipe = fragment_pipe;
6094 device->shader_priv = priv;
6098 constant_heap_free(&priv->pconst_heap);
6099 constant_heap_free(&priv->vconst_heap);
6100 HeapFree(GetProcessHeap(), 0, priv->stack);
6101 shader_buffer_free(&priv->shader_buffer);
6102 fragment_pipe->free_private(device);
6103 HeapFree(GetProcessHeap(), 0, priv);
6104 return E_OUTOFMEMORY;
6107 /* Context activation is done by the caller. */
6108 static void shader_glsl_free(struct wined3d_device *device)
6110 const struct wined3d_gl_info *gl_info = &device->adapter->gl_info;
6111 struct shader_glsl_priv *priv = device->shader_priv;
6114 for (i = 0; i < tex_type_count; ++i)
6116 if (priv->depth_blt_program_full[i])
6118 GL_EXTCALL(glDeleteObjectARB(priv->depth_blt_program_full[i]));
6120 if (priv->depth_blt_program_masked[i])
6122 GL_EXTCALL(glDeleteObjectARB(priv->depth_blt_program_masked[i]));
6126 wine_rb_destroy(&priv->program_lookup, NULL, NULL);
6127 constant_heap_free(&priv->pconst_heap);
6128 constant_heap_free(&priv->vconst_heap);
6129 HeapFree(GetProcessHeap(), 0, priv->stack);
6130 shader_buffer_free(&priv->shader_buffer);
6131 priv->fragment_pipe->free_private(device);
6133 HeapFree(GetProcessHeap(), 0, device->shader_priv);
6134 device->shader_priv = NULL;
6137 static void shader_glsl_context_destroyed(void *shader_priv, const struct wined3d_context *context) {}
6139 static void shader_glsl_get_caps(const struct wined3d_gl_info *gl_info, struct shader_caps *caps)
6143 if (gl_info->supported[EXT_GPU_SHADER4] && gl_info->supported[ARB_SHADER_BIT_ENCODING]
6144 && gl_info->supported[ARB_GEOMETRY_SHADER4] && gl_info->glsl_version >= MAKEDWORD_VERSION(1, 50)
6145 && gl_info->supported[ARB_DRAW_ELEMENTS_BASE_VERTEX] && gl_info->supported[ARB_DRAW_INSTANCED])
6147 /* ARB_shader_texture_lod or EXT_gpu_shader4 is required for the SM3
6148 * texldd and texldl instructions. */
6149 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD] || gl_info->supported[EXT_GPU_SHADER4])
6153 TRACE("Shader model %u.\n", shader_model);
6155 caps->vs_version = min(wined3d_settings.max_sm_vs, shader_model);
6156 caps->gs_version = min(wined3d_settings.max_sm_gs, shader_model);
6157 caps->ps_version = min(wined3d_settings.max_sm_ps, shader_model);
6159 caps->vs_uniform_count = gl_info->limits.glsl_vs_float_constants;
6160 caps->ps_uniform_count = gl_info->limits.glsl_ps_float_constants;
6162 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
6163 * Direct3D minimum requirement.
6165 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
6166 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
6168 * The problem is that the refrast clamps temporary results in the shader to
6169 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
6170 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
6171 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
6172 * offer a way to query this.
6174 caps->ps_1x_max_value = 8.0;
6176 /* Ideally we'd only set caps like sRGB writes here if supported by both
6177 * the shader backend and the fragment pipe, but we can get called before
6178 * shader_glsl_alloc(). */
6179 caps->wined3d_caps = WINED3D_SHADER_CAP_VS_CLIPPING
6180 | WINED3D_SHADER_CAP_SRGB_WRITE;
6183 static BOOL shader_glsl_color_fixup_supported(struct color_fixup_desc fixup)
6185 if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
6187 TRACE("Checking support for fixup:\n");
6188 dump_color_fixup_desc(fixup);
6191 /* We support everything except YUV conversions. */
6192 if (!is_complex_fixup(fixup))
6198 TRACE("[FAILED]\n");
6202 static const SHADER_HANDLER shader_glsl_instruction_handler_table[WINED3DSIH_TABLE_SIZE] =
6204 /* WINED3DSIH_ABS */ shader_glsl_map2gl,
6205 /* WINED3DSIH_ADD */ shader_glsl_binop,
6206 /* WINED3DSIH_AND */ shader_glsl_binop,
6207 /* WINED3DSIH_BEM */ shader_glsl_bem,
6208 /* WINED3DSIH_BREAK */ shader_glsl_break,
6209 /* WINED3DSIH_BREAKC */ shader_glsl_breakc,
6210 /* WINED3DSIH_BREAKP */ shader_glsl_breakp,
6211 /* WINED3DSIH_CALL */ shader_glsl_call,
6212 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz,
6213 /* WINED3DSIH_CMP */ shader_glsl_conditional_move,
6214 /* WINED3DSIH_CND */ shader_glsl_cnd,
6215 /* WINED3DSIH_CRS */ shader_glsl_cross,
6216 /* WINED3DSIH_CUT */ shader_glsl_cut,
6217 /* WINED3DSIH_DCL */ shader_glsl_nop,
6218 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_glsl_nop,
6219 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_glsl_nop,
6220 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_glsl_nop,
6221 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_glsl_nop,
6222 /* WINED3DSIH_DEF */ shader_glsl_nop,
6223 /* WINED3DSIH_DEFB */ shader_glsl_nop,
6224 /* WINED3DSIH_DEFI */ shader_glsl_nop,
6225 /* WINED3DSIH_DIV */ shader_glsl_binop,
6226 /* WINED3DSIH_DP2ADD */ shader_glsl_dp2add,
6227 /* WINED3DSIH_DP3 */ shader_glsl_dot,
6228 /* WINED3DSIH_DP4 */ shader_glsl_dot,
6229 /* WINED3DSIH_DST */ shader_glsl_dst,
6230 /* WINED3DSIH_DSX */ shader_glsl_map2gl,
6231 /* WINED3DSIH_DSY */ shader_glsl_map2gl,
6232 /* WINED3DSIH_ELSE */ shader_glsl_else,
6233 /* WINED3DSIH_EMIT */ shader_glsl_emit,
6234 /* WINED3DSIH_ENDIF */ shader_glsl_end,
6235 /* WINED3DSIH_ENDLOOP */ shader_glsl_end,
6236 /* WINED3DSIH_ENDREP */ shader_glsl_end,
6237 /* WINED3DSIH_EQ */ shader_glsl_relop,
6238 /* WINED3DSIH_EXP */ shader_glsl_map2gl,
6239 /* WINED3DSIH_EXPP */ shader_glsl_expp,
6240 /* WINED3DSIH_FRC */ shader_glsl_map2gl,
6241 /* WINED3DSIH_FTOI */ shader_glsl_to_int,
6242 /* WINED3DSIH_GE */ shader_glsl_relop,
6243 /* WINED3DSIH_IADD */ shader_glsl_binop,
6244 /* WINED3DSIH_IEQ */ NULL,
6245 /* WINED3DSIH_IF */ shader_glsl_if,
6246 /* WINED3DSIH_IFC */ shader_glsl_ifc,
6247 /* WINED3DSIH_IGE */ shader_glsl_relop,
6248 /* WINED3DSIH_IMUL */ shader_glsl_imul,
6249 /* WINED3DSIH_ITOF */ shader_glsl_to_float,
6250 /* WINED3DSIH_LABEL */ shader_glsl_label,
6251 /* WINED3DSIH_LD */ NULL,
6252 /* WINED3DSIH_LIT */ shader_glsl_lit,
6253 /* WINED3DSIH_LOG */ shader_glsl_log,
6254 /* WINED3DSIH_LOGP */ shader_glsl_log,
6255 /* WINED3DSIH_LOOP */ shader_glsl_loop,
6256 /* WINED3DSIH_LRP */ shader_glsl_lrp,
6257 /* WINED3DSIH_LT */ shader_glsl_relop,
6258 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn,
6259 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn,
6260 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn,
6261 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn,
6262 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn,
6263 /* WINED3DSIH_MAD */ shader_glsl_mad,
6264 /* WINED3DSIH_MAX */ shader_glsl_map2gl,
6265 /* WINED3DSIH_MIN */ shader_glsl_map2gl,
6266 /* WINED3DSIH_MOV */ shader_glsl_mov,
6267 /* WINED3DSIH_MOVA */ shader_glsl_mov,
6268 /* WINED3DSIH_MOVC */ shader_glsl_conditional_move,
6269 /* WINED3DSIH_MUL */ shader_glsl_binop,
6270 /* WINED3DSIH_NOP */ shader_glsl_nop,
6271 /* WINED3DSIH_NRM */ shader_glsl_nrm,
6272 /* WINED3DSIH_PHASE */ shader_glsl_nop,
6273 /* WINED3DSIH_POW */ shader_glsl_pow,
6274 /* WINED3DSIH_RCP */ shader_glsl_rcp,
6275 /* WINED3DSIH_REP */ shader_glsl_rep,
6276 /* WINED3DSIH_RET */ shader_glsl_ret,
6277 /* WINED3DSIH_ROUND_NI */ shader_glsl_map2gl,
6278 /* WINED3DSIH_RSQ */ shader_glsl_rsq,
6279 /* WINED3DSIH_SAMPLE */ NULL,
6280 /* WINED3DSIH_SAMPLE_GRAD */ NULL,
6281 /* WINED3DSIH_SAMPLE_LOD */ NULL,
6282 /* WINED3DSIH_SETP */ NULL,
6283 /* WINED3DSIH_SGE */ shader_glsl_compare,
6284 /* WINED3DSIH_SGN */ shader_glsl_sgn,
6285 /* WINED3DSIH_SINCOS */ shader_glsl_sincos,
6286 /* WINED3DSIH_SLT */ shader_glsl_compare,
6287 /* WINED3DSIH_SQRT */ NULL,
6288 /* WINED3DSIH_SUB */ shader_glsl_binop,
6289 /* WINED3DSIH_TEX */ shader_glsl_tex,
6290 /* WINED3DSIH_TEXBEM */ shader_glsl_texbem,
6291 /* WINED3DSIH_TEXBEML */ shader_glsl_texbem,
6292 /* WINED3DSIH_TEXCOORD */ shader_glsl_texcoord,
6293 /* WINED3DSIH_TEXDEPTH */ shader_glsl_texdepth,
6294 /* WINED3DSIH_TEXDP3 */ shader_glsl_texdp3,
6295 /* WINED3DSIH_TEXDP3TEX */ shader_glsl_texdp3tex,
6296 /* WINED3DSIH_TEXKILL */ shader_glsl_texkill,
6297 /* WINED3DSIH_TEXLDD */ shader_glsl_texldd,
6298 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl,
6299 /* WINED3DSIH_TEXM3x2DEPTH */ shader_glsl_texm3x2depth,
6300 /* WINED3DSIH_TEXM3x2PAD */ shader_glsl_texm3x2pad,
6301 /* WINED3DSIH_TEXM3x2TEX */ shader_glsl_texm3x2tex,
6302 /* WINED3DSIH_TEXM3x3 */ shader_glsl_texm3x3,
6303 /* WINED3DSIH_TEXM3x3DIFF */ NULL,
6304 /* WINED3DSIH_TEXM3x3PAD */ shader_glsl_texm3x3pad,
6305 /* WINED3DSIH_TEXM3x3SPEC */ shader_glsl_texm3x3spec,
6306 /* WINED3DSIH_TEXM3x3TEX */ shader_glsl_texm3x3tex,
6307 /* WINED3DSIH_TEXM3x3VSPEC */ shader_glsl_texm3x3vspec,
6308 /* WINED3DSIH_TEXREG2AR */ shader_glsl_texreg2ar,
6309 /* WINED3DSIH_TEXREG2GB */ shader_glsl_texreg2gb,
6310 /* WINED3DSIH_TEXREG2RGB */ shader_glsl_texreg2rgb,
6311 /* WINED3DSIH_UDIV */ shader_glsl_udiv,
6312 /* WINED3DSIH_USHR */ shader_glsl_binop,
6313 /* WINED3DSIH_UTOF */ shader_glsl_to_float,
6314 /* WINED3DSIH_XOR */ shader_glsl_binop,
6317 static void shader_glsl_handle_instruction(const struct wined3d_shader_instruction *ins) {
6318 SHADER_HANDLER hw_fct;
6320 /* Select handler */
6321 hw_fct = shader_glsl_instruction_handler_table[ins->handler_idx];
6323 /* Unhandled opcode */
6326 FIXME("Backend can't handle opcode %#x\n", ins->handler_idx);
6331 shader_glsl_add_instruction_modifiers(ins);
6334 static BOOL shader_glsl_has_ffp_proj_control(void *shader_priv)
6336 struct shader_glsl_priv *priv = shader_priv;
6338 return priv->ffp_proj_control;
6341 const struct wined3d_shader_backend_ops glsl_shader_backend =
6343 shader_glsl_handle_instruction,
6345 shader_glsl_select_depth_blt,
6346 shader_glsl_deselect_depth_blt,
6347 shader_glsl_update_float_vertex_constants,
6348 shader_glsl_update_float_pixel_constants,
6349 shader_glsl_load_constants,
6350 shader_glsl_load_np2fixup_constants,
6351 shader_glsl_destroy,
6354 shader_glsl_context_destroyed,
6355 shader_glsl_get_caps,
6356 shader_glsl_color_fixup_supported,
6357 shader_glsl_has_ffp_proj_control,
6360 static void glsl_fragment_pipe_enable(const struct wined3d_gl_info *gl_info, BOOL enable)
6362 /* Nothing to do. */
6365 static void glsl_fragment_pipe_get_caps(const struct wined3d_gl_info *gl_info, struct fragment_caps *caps)
6367 caps->wined3d_caps = WINED3D_FRAGMENT_CAP_PROJ_CONTROL
6368 | WINED3D_FRAGMENT_CAP_SRGB_WRITE;
6369 caps->PrimitiveMiscCaps = WINED3DPMISCCAPS_TSSARGTEMP;
6370 caps->TextureOpCaps = WINED3DTEXOPCAPS_DISABLE
6371 | WINED3DTEXOPCAPS_SELECTARG1
6372 | WINED3DTEXOPCAPS_SELECTARG2
6373 | WINED3DTEXOPCAPS_MODULATE4X
6374 | WINED3DTEXOPCAPS_MODULATE2X
6375 | WINED3DTEXOPCAPS_MODULATE
6376 | WINED3DTEXOPCAPS_ADDSIGNED2X
6377 | WINED3DTEXOPCAPS_ADDSIGNED
6378 | WINED3DTEXOPCAPS_ADD
6379 | WINED3DTEXOPCAPS_SUBTRACT
6380 | WINED3DTEXOPCAPS_ADDSMOOTH
6381 | WINED3DTEXOPCAPS_BLENDCURRENTALPHA
6382 | WINED3DTEXOPCAPS_BLENDFACTORALPHA
6383 | WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
6384 | WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
6385 | WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
6386 | WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
6387 | WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
6388 | WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
6389 | WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
6390 | WINED3DTEXOPCAPS_DOTPRODUCT3
6391 | WINED3DTEXOPCAPS_MULTIPLYADD
6392 | WINED3DTEXOPCAPS_LERP
6393 | WINED3DTEXOPCAPS_BUMPENVMAP
6394 | WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE;
6395 caps->MaxTextureBlendStages = 8;
6396 caps->MaxSimultaneousTextures = min(gl_info->limits.fragment_samplers, 8);
6399 static void *glsl_fragment_pipe_alloc(const struct wined3d_shader_backend_ops *shader_backend, void *shader_priv)
6401 struct shader_glsl_priv *priv;
6403 if (shader_backend == &glsl_shader_backend)
6407 if (wine_rb_init(&priv->ffp_fragment_shaders, &wined3d_ffp_frag_program_rb_functions) == -1)
6409 ERR("Failed to initialize rbtree.\n");
6416 FIXME("GLSL fragment pipe without GLSL shader backend not implemented.\n");
6421 struct glsl_ffp_destroy_ctx
6423 struct shader_glsl_priv *priv;
6424 const struct wined3d_gl_info *gl_info;
6427 static void shader_glsl_free_ffp_fragment_shader(struct wine_rb_entry *entry, void *context)
6429 struct glsl_ffp_fragment_shader *shader = WINE_RB_ENTRY_VALUE(entry,
6430 struct glsl_ffp_fragment_shader, entry.entry);
6431 struct glsl_shader_prog_link *program, *program2;
6432 struct glsl_ffp_destroy_ctx *ctx = context;
6434 LIST_FOR_EACH_ENTRY_SAFE(program, program2, &shader->linked_programs,
6435 struct glsl_shader_prog_link, ps.shader_entry)
6437 delete_glsl_program_entry(ctx->priv, ctx->gl_info, program);
6439 ctx->gl_info->gl_ops.ext.p_glDeleteObjectARB(shader->id);
6440 HeapFree(GetProcessHeap(), 0, shader);
6443 /* Context activation is done by the caller. */
6444 static void glsl_fragment_pipe_free(struct wined3d_device *device)
6446 struct shader_glsl_priv *priv = device->fragment_priv;
6447 struct glsl_ffp_destroy_ctx ctx;
6450 ctx.gl_info = &device->adapter->gl_info;
6451 wine_rb_destroy(&priv->ffp_fragment_shaders, shader_glsl_free_ffp_fragment_shader, &ctx);
6454 static void glsl_fragment_pipe_shader(struct wined3d_context *context,
6455 const struct wined3d_state *state, DWORD state_id)
6457 context->last_was_pshader = use_ps(state);
6459 context->select_shader = 1;
6460 context->load_constants = 1;
6463 static void glsl_fragment_pipe_fog(struct wined3d_context *context,
6464 const struct wined3d_state *state, DWORD state_id)
6466 BOOL use_vshader = use_vs(state);
6467 enum fogsource new_source;
6469 context->select_shader = 1;
6470 context->load_constants = 1;
6472 if (!state->render_states[WINED3D_RS_FOGENABLE])
6475 if (state->render_states[WINED3D_RS_FOGTABLEMODE] == WINED3D_FOG_NONE)
6478 new_source = FOGSOURCE_VS;
6479 else if (state->render_states[WINED3D_RS_FOGVERTEXMODE] == WINED3D_FOG_NONE || context->last_was_rhw)
6480 new_source = FOGSOURCE_COORD;
6482 new_source = FOGSOURCE_FFP;
6486 new_source = FOGSOURCE_FFP;
6489 if (new_source != context->fog_source)
6491 context->fog_source = new_source;
6492 state_fogstartend(context, state, STATE_RENDER(WINED3D_RS_FOGSTART));
6496 static void glsl_fragment_pipe_tex_transform(struct wined3d_context *context,
6497 const struct wined3d_state *state, DWORD state_id)
6499 context->select_shader = 1;
6500 context->load_constants = 1;
6503 static void glsl_fragment_pipe_invalidate_constants(struct wined3d_context *context,
6504 const struct wined3d_state *state, DWORD state_id)
6506 context->load_constants = 1;
6509 static const struct StateEntryTemplate glsl_fragment_pipe_state_template[] =
6511 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR), {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6512 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6513 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6514 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6515 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6516 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6517 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6518 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6519 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6520 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6521 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6522 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6523 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6524 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6525 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6526 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6527 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6528 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6529 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6530 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6531 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6532 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6533 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6534 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6535 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6536 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6537 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6538 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6539 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6540 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6541 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6542 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6543 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6544 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6545 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6546 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6547 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6548 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6549 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6550 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6551 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6552 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6553 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6554 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6555 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6556 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6557 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6558 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6559 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6560 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6561 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6562 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6563 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6564 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6565 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6566 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6567 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6568 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6569 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6570 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6571 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6572 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6573 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6574 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6575 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6576 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6577 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6578 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6579 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6580 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6581 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6582 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6583 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6584 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6585 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6586 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6587 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6588 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6589 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6590 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6591 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6592 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6593 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6594 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6595 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6596 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6597 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6598 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6599 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6600 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6601 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6602 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6603 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6604 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6605 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6606 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6607 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6608 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6609 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6610 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6611 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6612 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6613 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6614 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6615 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6616 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6617 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6618 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6619 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6620 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6621 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6622 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6623 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6624 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6625 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6626 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6627 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6628 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6629 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6630 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6631 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6632 {STATE_PIXELSHADER, {STATE_PIXELSHADER, glsl_fragment_pipe_shader }, WINED3D_GL_EXT_NONE },
6633 {STATE_RENDER(WINED3D_RS_FOGENABLE), {STATE_RENDER(WINED3D_RS_FOGENABLE), glsl_fragment_pipe_fog }, WINED3D_GL_EXT_NONE },
6634 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE), {STATE_RENDER(WINED3D_RS_FOGENABLE), NULL }, WINED3D_GL_EXT_NONE },
6635 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE), {STATE_RENDER(WINED3D_RS_FOGENABLE), NULL }, WINED3D_GL_EXT_NONE },
6636 {STATE_RENDER(WINED3D_RS_FOGSTART), {STATE_RENDER(WINED3D_RS_FOGSTART), state_fogstartend }, WINED3D_GL_EXT_NONE },
6637 {STATE_RENDER(WINED3D_RS_FOGEND), {STATE_RENDER(WINED3D_RS_FOGSTART), NULL }, WINED3D_GL_EXT_NONE },
6638 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE), {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE), state_srgbwrite }, ARB_FRAMEBUFFER_SRGB},
6639 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6640 {STATE_RENDER(WINED3D_RS_FOGCOLOR), {STATE_RENDER(WINED3D_RS_FOGCOLOR), state_fogcolor }, WINED3D_GL_EXT_NONE },
6641 {STATE_RENDER(WINED3D_RS_FOGDENSITY), {STATE_RENDER(WINED3D_RS_FOGDENSITY), state_fogdensity }, WINED3D_GL_EXT_NONE },
6642 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), {STATE_TEXTURESTAGE(0, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), glsl_fragment_pipe_tex_transform }, WINED3D_GL_EXT_NONE },
6643 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), {STATE_TEXTURESTAGE(1, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), glsl_fragment_pipe_tex_transform }, WINED3D_GL_EXT_NONE },
6644 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), {STATE_TEXTURESTAGE(2, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), glsl_fragment_pipe_tex_transform }, WINED3D_GL_EXT_NONE },
6645 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), {STATE_TEXTURESTAGE(3, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), glsl_fragment_pipe_tex_transform }, WINED3D_GL_EXT_NONE },
6646 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), {STATE_TEXTURESTAGE(4, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), glsl_fragment_pipe_tex_transform }, WINED3D_GL_EXT_NONE },
6647 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), {STATE_TEXTURESTAGE(5, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), glsl_fragment_pipe_tex_transform }, WINED3D_GL_EXT_NONE },
6648 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), {STATE_TEXTURESTAGE(6, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), glsl_fragment_pipe_tex_transform }, WINED3D_GL_EXT_NONE },
6649 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), {STATE_TEXTURESTAGE(7, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS), glsl_fragment_pipe_tex_transform }, WINED3D_GL_EXT_NONE },
6650 {STATE_RENDER(WINED3D_RS_SPECULARENABLE), {STATE_RENDER(WINED3D_RS_SPECULARENABLE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6651 {0 /* Terminate */, {0, 0 }, WINED3D_GL_EXT_NONE },
6654 const struct fragment_pipeline glsl_fragment_pipe =
6656 glsl_fragment_pipe_enable,
6657 glsl_fragment_pipe_get_caps,
6658 glsl_fragment_pipe_alloc,
6659 glsl_fragment_pipe_free,
6660 shader_glsl_color_fixup_supported,
6661 glsl_fragment_pipe_state_template,