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_NPOT 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 GL_EXTCALL(glUniform4fvARB(constant_locations[lconst->idx], 1, (const GLfloat *)lconst->value));
587 checkGLcall("glUniform4fvARB()");
590 /* Context activation is done by the caller. */
591 static void shader_glsl_load_constantsI(const struct wined3d_shader *shader, const struct wined3d_gl_info *gl_info,
592 const GLint locations[MAX_CONST_I], const int *constants, WORD constants_set)
597 for (i = 0; constants_set; constants_set >>= 1, ++i)
599 if (!(constants_set & 1)) continue;
601 TRACE_(d3d_constants)("Loading constants %u: %i, %i, %i, %i\n",
602 i, constants[i*4], constants[i*4+1], constants[i*4+2], constants[i*4+3]);
604 /* We found this uniform name in the program - go ahead and send the data */
605 GL_EXTCALL(glUniform4ivARB(locations[i], 1, &constants[i*4]));
606 checkGLcall("glUniform4ivARB");
609 /* Load immediate constants */
610 ptr = list_head(&shader->constantsI);
613 const struct wined3d_shader_lconst *lconst = LIST_ENTRY(ptr, const struct wined3d_shader_lconst, entry);
614 unsigned int idx = lconst->idx;
615 const GLint *values = (const GLint *)lconst->value;
617 TRACE_(d3d_constants)("Loading local constants %i: %i, %i, %i, %i\n", idx,
618 values[0], values[1], values[2], values[3]);
620 /* We found this uniform name in the program - go ahead and send the data */
621 GL_EXTCALL(glUniform4ivARB(locations[idx], 1, values));
622 checkGLcall("glUniform4ivARB");
623 ptr = list_next(&shader->constantsI, ptr);
627 /* Context activation is done by the caller. */
628 static void shader_glsl_load_constantsB(const struct wined3d_shader *shader, const struct wined3d_gl_info *gl_info,
629 GLhandleARB programId, const BOOL *constants, WORD constants_set)
637 prefix = shader_glsl_get_prefix(shader->reg_maps.shader_version.type);
639 /* TODO: Benchmark and see if it would be beneficial to store the
640 * locations of the constants to avoid looking up each time */
641 for (i = 0; constants_set; constants_set >>= 1, ++i)
643 if (!(constants_set & 1)) continue;
645 TRACE_(d3d_constants)("Loading constants %i: %i;\n", i, constants[i]);
647 /* TODO: Benchmark and see if it would be beneficial to store the
648 * locations of the constants to avoid looking up each time */
649 snprintf(tmp_name, sizeof(tmp_name), "%s_b[%i]", prefix, i);
650 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
651 GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, &constants[i]));
654 /* Load immediate constants */
655 ptr = list_head(&shader->constantsB);
658 const struct wined3d_shader_lconst *lconst = LIST_ENTRY(ptr, const struct wined3d_shader_lconst, entry);
659 unsigned int idx = lconst->idx;
660 const GLint *values = (const GLint *)lconst->value;
662 TRACE_(d3d_constants)("Loading local constants %i: %i\n", idx, values[0]);
664 snprintf(tmp_name, sizeof(tmp_name), "%s_b[%i]", prefix, idx);
665 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
666 GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, values));
667 ptr = list_next(&shader->constantsB, ptr);
670 checkGLcall("shader_glsl_load_constantsB()");
673 static void reset_program_constant_version(struct wine_rb_entry *entry, void *context)
675 WINE_RB_ENTRY_VALUE(entry, struct glsl_shader_prog_link, program_lookup_entry)->constant_version = 0;
678 /* Context activation is done by the caller (state handler). */
679 static void shader_glsl_load_np2fixup_constants(void *shader_priv,
680 const struct wined3d_gl_info *gl_info, const struct wined3d_state *state)
682 struct shader_glsl_priv *glsl_priv = shader_priv;
683 const struct glsl_shader_prog_link *prog = glsl_priv->glsl_program;
685 /* No GLSL program set - nothing to do. */
688 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
689 if (!use_ps(state)) return;
691 if (prog->ps.np2_fixup_info && prog->ps.np2_fixup_location != -1)
694 UINT fixup = prog->ps.np2_fixup_info->active;
695 GLfloat np2fixup_constants[4 * MAX_FRAGMENT_SAMPLERS];
697 for (i = 0; fixup; fixup >>= 1, ++i)
699 const struct wined3d_texture *tex = state->textures[i];
700 const unsigned char idx = prog->ps.np2_fixup_info->idx[i];
701 GLfloat *tex_dim = &np2fixup_constants[(idx >> 1) * 4];
705 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
711 tex_dim[2] = tex->pow2_matrix[0];
712 tex_dim[3] = tex->pow2_matrix[5];
716 tex_dim[0] = tex->pow2_matrix[0];
717 tex_dim[1] = tex->pow2_matrix[5];
721 GL_EXTCALL(glUniform4fvARB(prog->ps.np2_fixup_location,
722 prog->ps.np2_fixup_info->num_consts, np2fixup_constants));
726 /* Context activation is done by the caller (state handler). */
727 static void shader_glsl_load_constants(const struct wined3d_context *context,
728 BOOL usePixelShader, BOOL useVertexShader)
730 const struct wined3d_gl_info *gl_info = context->gl_info;
731 struct wined3d_device *device = context->swapchain->device;
732 struct wined3d_stateblock *stateBlock = device->stateBlock;
733 const struct wined3d_state *state = &stateBlock->state;
734 struct shader_glsl_priv *priv = device->shader_priv;
735 float position_fixup[4];
737 GLhandleARB programId;
738 struct glsl_shader_prog_link *prog = priv->glsl_program;
739 UINT constant_version;
743 /* No GLSL program set - nothing to do. */
746 programId = prog->programId;
747 constant_version = prog->constant_version;
751 const struct wined3d_shader *vshader = state->vertex_shader;
753 /* Load DirectX 9 float constants/uniforms for vertex shader */
754 shader_glsl_load_constantsF(vshader, gl_info, state->vs_consts_f,
755 prog->vs.uniform_f_locations, &priv->vconst_heap, priv->stack, constant_version);
757 /* Load DirectX 9 integer constants/uniforms for vertex shader */
758 shader_glsl_load_constantsI(vshader, gl_info, prog->vs.uniform_i_locations, state->vs_consts_i,
759 stateBlock->changed.vertexShaderConstantsI & vshader->reg_maps.integer_constants);
761 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
762 shader_glsl_load_constantsB(vshader, gl_info, programId, state->vs_consts_b,
763 stateBlock->changed.vertexShaderConstantsB & vshader->reg_maps.boolean_constants);
765 /* Upload the position fixup params */
766 shader_get_position_fixup(context, state, position_fixup);
767 GL_EXTCALL(glUniform4fvARB(prog->vs.pos_fixup_location, 1, position_fixup));
768 checkGLcall("glUniform4fvARB");
773 const struct wined3d_shader *pshader = state->pixel_shader;
775 /* Load DirectX 9 float constants/uniforms for pixel shader */
776 shader_glsl_load_constantsF(pshader, gl_info, state->ps_consts_f,
777 prog->ps.uniform_f_locations, &priv->pconst_heap, priv->stack, constant_version);
779 /* Load DirectX 9 integer constants/uniforms for pixel shader */
780 shader_glsl_load_constantsI(pshader, gl_info, prog->ps.uniform_i_locations, state->ps_consts_i,
781 stateBlock->changed.pixelShaderConstantsI & pshader->reg_maps.integer_constants);
783 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
784 shader_glsl_load_constantsB(pshader, gl_info, programId, state->ps_consts_b,
785 stateBlock->changed.pixelShaderConstantsB & pshader->reg_maps.boolean_constants);
787 /* Upload the environment bump map matrix if needed. The needsbumpmat
788 * member specifies the texture stage to load the matrix from. It
789 * can't be 0 for a valid texbem instruction. */
790 for (i = 0; i < MAX_TEXTURES; ++i)
794 if (prog->ps.bumpenv_mat_location[i] == -1)
797 data = (const float *)&state->texture_states[i][WINED3D_TSS_BUMPENV_MAT00];
798 GL_EXTCALL(glUniformMatrix2fvARB(prog->ps.bumpenv_mat_location[i], 1, 0, data));
799 checkGLcall("glUniformMatrix2fvARB");
801 /* texbeml needs the luminance scale and offset too. If texbeml
802 * is used, needsbumpmat is set too, so we can check that in the
803 * needsbumpmat check. */
804 if (prog->ps.bumpenv_lum_scale_location[i] != -1)
806 const GLfloat *scale = (const GLfloat *)&state->texture_states[i][WINED3D_TSS_BUMPENV_LSCALE];
807 const GLfloat *offset = (const GLfloat *)&state->texture_states[i][WINED3D_TSS_BUMPENV_LOFFSET];
809 GL_EXTCALL(glUniform1fvARB(prog->ps.bumpenv_lum_scale_location[i], 1, scale));
810 checkGLcall("glUniform1fvARB");
811 GL_EXTCALL(glUniform1fvARB(prog->ps.bumpenv_lum_offset_location[i], 1, offset));
812 checkGLcall("glUniform1fvARB");
816 if (prog->ps.ycorrection_location != -1)
818 float correction_params[4];
820 if (context->render_offscreen)
822 correction_params[0] = 0.0f;
823 correction_params[1] = 1.0f;
825 /* position is window relative, not viewport relative */
826 correction_params[0] = (float) context->current_rt->resource.height;
827 correction_params[1] = -1.0f;
829 GL_EXTCALL(glUniform4fvARB(prog->ps.ycorrection_location, 1, correction_params));
832 else if (priv->fragment_pipe == &glsl_fragment_pipe)
836 for (i = 0; i < MAX_TEXTURES; ++i)
838 GL_EXTCALL(glUniformMatrix2fvARB(prog->ps.bumpenv_mat_location[i], 1, 0,
839 (const float *)&state->texture_states[i][WINED3D_TSS_BUMPENV_MAT00]));
840 GL_EXTCALL(glUniform1fARB(prog->ps.bumpenv_lum_scale_location[i],
841 *(const float *)&state->texture_states[i][WINED3D_TSS_BUMPENV_LSCALE]));
842 GL_EXTCALL(glUniform1fARB(prog->ps.bumpenv_lum_offset_location[i],
843 *(const float *)&state->texture_states[i][WINED3D_TSS_BUMPENV_LOFFSET]));
846 D3DCOLORTOGLFLOAT4(state->render_states[WINED3D_RS_TEXTUREFACTOR], col);
847 GL_EXTCALL(glUniform4fARB(prog->ps.tex_factor_location, col[0], col[1], col[2], col[3]));
849 if (state->render_states[WINED3D_RS_SPECULARENABLE])
850 GL_EXTCALL(glUniform4fARB(prog->ps.specular_enable_location, 1.0f, 1.0f, 1.0f, 0.0f));
852 GL_EXTCALL(glUniform4fARB(prog->ps.specular_enable_location, 0.0f, 0.0f, 0.0f, 0.0f));
854 checkGLcall("fixed function uniforms");
857 if (priv->next_constant_version == UINT_MAX)
859 TRACE("Max constant version reached, resetting to 0.\n");
860 wine_rb_for_each_entry(&priv->program_lookup, reset_program_constant_version, NULL);
861 priv->next_constant_version = 1;
865 prog->constant_version = priv->next_constant_version++;
869 static void update_heap_entry(const struct constant_heap *heap, unsigned int idx,
870 unsigned int heap_idx, DWORD new_version)
872 struct constant_entry *entries = heap->entries;
873 unsigned int *positions = heap->positions;
874 unsigned int parent_idx;
878 parent_idx = heap_idx >> 1;
880 if (new_version <= entries[parent_idx].version) break;
882 entries[heap_idx] = entries[parent_idx];
883 positions[entries[parent_idx].idx] = heap_idx;
884 heap_idx = parent_idx;
887 entries[heap_idx].version = new_version;
888 entries[heap_idx].idx = idx;
889 positions[idx] = heap_idx;
892 static void shader_glsl_update_float_vertex_constants(struct wined3d_device *device, UINT start, UINT count)
894 struct shader_glsl_priv *priv = device->shader_priv;
895 struct constant_heap *heap = &priv->vconst_heap;
898 for (i = start; i < count + start; ++i)
900 if (!device->stateBlock->changed.vertexShaderConstantsF[i])
901 update_heap_entry(heap, i, heap->size++, priv->next_constant_version);
903 update_heap_entry(heap, i, heap->positions[i], priv->next_constant_version);
907 static void shader_glsl_update_float_pixel_constants(struct wined3d_device *device, UINT start, UINT count)
909 struct shader_glsl_priv *priv = device->shader_priv;
910 struct constant_heap *heap = &priv->pconst_heap;
913 for (i = start; i < count + start; ++i)
915 if (!device->stateBlock->changed.pixelShaderConstantsF[i])
916 update_heap_entry(heap, i, heap->size++, priv->next_constant_version);
918 update_heap_entry(heap, i, heap->positions[i], priv->next_constant_version);
922 static unsigned int vec4_varyings(DWORD shader_major, const struct wined3d_gl_info *gl_info)
924 unsigned int ret = gl_info->limits.glsl_varyings / 4;
925 /* 4.0 shaders do not write clip coords because d3d10 does not support user clipplanes */
926 if(shader_major > 3) return ret;
928 /* 3.0 shaders may need an extra varying for the clip coord on some cards(mostly dx10 ones) */
929 if (gl_info->quirks & WINED3D_QUIRK_GLSL_CLIP_VARYING) ret -= 1;
933 /** Generate the variable & register declarations for the GLSL output target */
934 static void shader_generate_glsl_declarations(const struct wined3d_context *context,
935 struct wined3d_shader_buffer *buffer, const struct wined3d_shader *shader,
936 const struct wined3d_shader_reg_maps *reg_maps, const struct shader_glsl_ctx_priv *ctx_priv)
938 const struct wined3d_shader_version *version = ®_maps->shader_version;
939 const struct wined3d_state *state = &shader->device->stateBlock->state;
940 const struct ps_compile_args *ps_args = ctx_priv->cur_ps_args;
941 const struct wined3d_gl_info *gl_info = context->gl_info;
942 const struct wined3d_fb_state *fb = &shader->device->fb;
943 unsigned int i, extra_constants_needed = 0;
944 const struct wined3d_shader_lconst *lconst;
948 prefix = shader_glsl_get_prefix(version->type);
950 /* Prototype the subroutines */
951 for (i = 0, map = reg_maps->labels; map; map >>= 1, ++i)
953 if (map & 1) shader_addline(buffer, "void subroutine%u();\n", i);
956 /* Declare the constants (aka uniforms) */
957 if (shader->limits.constant_float > 0)
959 unsigned max_constantsF;
961 /* Unless the shader uses indirect addressing, always declare the
962 * maximum array size and ignore that we need some uniforms privately.
963 * E.g. if GL supports 256 uniforms, and we need 2 for the pos fixup
964 * and immediate values, still declare VC[256]. If the shader needs
965 * more uniforms than we have it won't work in any case. If it uses
966 * less, the compiler will figure out which uniforms are really used
967 * and strip them out. This allows a shader to use c255 on a dx9 card,
968 * as long as it doesn't also use all the other constants.
970 * If the shader uses indirect addressing the compiler must assume
971 * that all declared uniforms are used. In this case, declare only the
972 * amount that we're assured to have.
974 * Thus we run into problems in these two cases:
975 * 1) The shader really uses more uniforms than supported.
976 * 2) The shader uses indirect addressing, less constants than
977 * supported, but uses a constant index > #supported consts. */
978 if (version->type == WINED3D_SHADER_TYPE_PIXEL)
980 /* No indirect addressing here. */
981 max_constantsF = gl_info->limits.glsl_ps_float_constants;
985 if (reg_maps->usesrelconstF)
987 /* Subtract the other potential uniforms from the max
988 * available (bools, ints, and 1 row of projection matrix).
989 * Subtract another uniform for immediate values, which have
990 * to be loaded via uniform by the driver as well. The shader
991 * code only uses 0.5, 2.0, 1.0, 128 and -128 in vertex
992 * shader code, so one vec4 should be enough. (Unfortunately
993 * the Nvidia driver doesn't store 128 and -128 in one float).
995 * Writing gl_ClipVertex requires one uniform for each
996 * clipplane as well. */
997 max_constantsF = gl_info->limits.glsl_vs_float_constants - 3;
998 if(ctx_priv->cur_vs_args->clip_enabled)
1000 max_constantsF -= gl_info->limits.clipplanes;
1002 max_constantsF -= count_bits(reg_maps->integer_constants);
1003 /* Strictly speaking a bool only uses one scalar, but the nvidia(Linux) compiler doesn't pack them properly,
1004 * so each scalar requires a full vec4. We could work around this by packing the booleans ourselves, but
1005 * for now take this into account when calculating the number of available constants
1007 max_constantsF -= count_bits(reg_maps->boolean_constants);
1008 /* Set by driver quirks in directx.c */
1009 max_constantsF -= gl_info->reserved_glsl_constants;
1011 if (max_constantsF < shader->limits.constant_float)
1013 static unsigned int once;
1016 ERR_(winediag)("The hardware does not support enough uniform components to run this shader,"
1017 " it may not render correctly.\n");
1019 WARN("The hardware does not support enough uniform components to run this shader.\n");
1024 max_constantsF = gl_info->limits.glsl_vs_float_constants;
1027 max_constantsF = min(shader->limits.constant_float, max_constantsF);
1028 shader_addline(buffer, "uniform vec4 %s_c[%u];\n", prefix, max_constantsF);
1031 /* Always declare the full set of constants, the compiler can remove the
1032 * unused ones because d3d doesn't (yet) support indirect int and bool
1033 * constant addressing. This avoids problems if the app uses e.g. i0 and i9. */
1034 if (shader->limits.constant_int > 0 && reg_maps->integer_constants)
1035 shader_addline(buffer, "uniform ivec4 %s_i[%u];\n", prefix, shader->limits.constant_int);
1037 if (shader->limits.constant_bool > 0 && reg_maps->boolean_constants)
1038 shader_addline(buffer, "uniform bool %s_b[%u];\n", prefix, shader->limits.constant_bool);
1040 for (i = 0; i < WINED3D_MAX_CBS; ++i)
1042 if (reg_maps->cb_sizes[i])
1043 shader_addline(buffer, "uniform vec4 %s_cb%u[%u];\n", prefix, i, reg_maps->cb_sizes[i]);
1046 /* Declare texture samplers */
1047 for (i = 0; i < shader->limits.sampler; ++i)
1049 if (reg_maps->sampler_type[i])
1051 BOOL shadow_sampler = version->type == WINED3D_SHADER_TYPE_PIXEL && (ps_args->shadow & (1 << i));
1054 switch (reg_maps->sampler_type[i])
1058 shader_addline(buffer, "uniform sampler1DShadow %s_sampler%u;\n", prefix, i);
1060 shader_addline(buffer, "uniform sampler1D %s_sampler%u;\n", prefix, i);
1063 tex_rect = version->type == WINED3D_SHADER_TYPE_PIXEL && (ps_args->np2_fixup & (1 << i));
1064 tex_rect = tex_rect && gl_info->supported[ARB_TEXTURE_RECTANGLE];
1068 shader_addline(buffer, "uniform sampler2DRectShadow %s_sampler%u;\n", prefix, i);
1070 shader_addline(buffer, "uniform sampler2DShadow %s_sampler%u;\n", prefix, i);
1075 shader_addline(buffer, "uniform sampler2DRect %s_sampler%u;\n", prefix, i);
1077 shader_addline(buffer, "uniform sampler2D %s_sampler%u;\n", prefix, i);
1080 case WINED3DSTT_CUBE:
1082 FIXME("Unsupported Cube shadow sampler.\n");
1083 shader_addline(buffer, "uniform samplerCube %s_sampler%u;\n", prefix, i);
1085 case WINED3DSTT_VOLUME:
1087 FIXME("Unsupported 3D shadow sampler.\n");
1088 shader_addline(buffer, "uniform sampler3D %s_sampler%u;\n", prefix, i);
1091 shader_addline(buffer, "uniform unsupported_sampler %s_sampler%u;\n", prefix, i);
1092 FIXME("Unrecognized sampler type: %#x\n", reg_maps->sampler_type[i]);
1098 /* Declare uniforms for NP2 texcoord fixup:
1099 * This is NOT done inside the loop that declares the texture samplers
1100 * since the NP2 fixup code is currently only used for the GeforceFX
1101 * series and when forcing the ARB_npot extension off. Modern cards just
1102 * skip the code anyway, so put it inside a separate loop. */
1103 if (version->type == WINED3D_SHADER_TYPE_PIXEL && ps_args->np2_fixup)
1105 struct ps_np2fixup_info *fixup = ctx_priv->cur_np2fixup_info;
1108 /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
1109 * while D3D has them in the (normalized) [0,1]x[0,1] range.
1110 * samplerNP2Fixup stores texture dimensions and is updated through
1111 * shader_glsl_load_np2fixup_constants when the sampler changes. */
1113 for (i = 0; i < shader->limits.sampler; ++i)
1115 if (reg_maps->sampler_type[i])
1117 if (!(ps_args->np2_fixup & (1 << i))) continue;
1119 if (WINED3DSTT_2D != reg_maps->sampler_type[i]) {
1120 FIXME("Non-2D texture is flagged for NP2 texcoord fixup.\n");
1124 fixup->idx[i] = cur++;
1128 fixup->num_consts = (cur + 1) >> 1;
1129 fixup->active = ps_args->np2_fixup;
1130 shader_addline(buffer, "uniform vec4 %s_samplerNP2Fixup[%u];\n", prefix, fixup->num_consts);
1133 /* Declare address variables */
1134 for (i = 0, map = reg_maps->address; map; map >>= 1, ++i)
1136 if (map & 1) shader_addline(buffer, "ivec4 A%u;\n", i);
1139 /* Declare texture coordinate temporaries and initialize them */
1140 for (i = 0, map = reg_maps->texcoord; map; map >>= 1, ++i)
1142 if (map & 1) shader_addline(buffer, "vec4 T%u = gl_TexCoord[%u];\n", i, i);
1145 if (version->type == WINED3D_SHADER_TYPE_VERTEX)
1147 /* Declare attributes. */
1148 for (i = 0, map = reg_maps->input_registers; map; map >>= 1, ++i)
1151 shader_addline(buffer, "attribute vec4 %s_in%u;\n", prefix, i);
1154 shader_addline(buffer, "uniform vec4 posFixup;\n");
1155 shader_addline(buffer, "void order_ps_input(in vec4[%u]);\n", shader->limits.packed_output);
1157 else if (version->type == WINED3D_SHADER_TYPE_GEOMETRY)
1159 shader_addline(buffer, "varying in vec4 gs_in[][%u];\n", shader->limits.packed_input);
1161 else if (version->type == WINED3D_SHADER_TYPE_PIXEL)
1163 if (version->major >= 3)
1165 UINT in_count = min(vec4_varyings(version->major, gl_info), shader->limits.packed_input);
1168 shader_addline(buffer, "varying vec4 %s_in[%u];\n", prefix, in_count);
1170 /* TODO: Write a replacement shader for the fixed function
1171 * vertex pipeline, so this isn't needed. For fixed function
1172 * vertex processing + 3.0 pixel shader we need a separate
1173 * function in the pixel shader that reads the fixed function
1174 * color into the packed input registers. */
1175 shader_addline(buffer, "vec4 %s_in[%u];\n", prefix, in_count);
1178 for (i = 0, map = reg_maps->bumpmat; map; map >>= 1, ++i)
1183 shader_addline(buffer, "uniform mat2 bumpenv_mat%u;\n", i);
1185 if (reg_maps->luminanceparams & (1 << i))
1187 shader_addline(buffer, "uniform float bumpenv_lum_scale%u;\n", i);
1188 shader_addline(buffer, "uniform float bumpenv_lum_offset%u;\n", i);
1189 extra_constants_needed++;
1192 extra_constants_needed++;
1195 if (ps_args->srgb_correction)
1197 shader_addline(buffer, "const vec4 srgb_const0 = vec4(%.8e, %.8e, %.8e, %.8e);\n",
1198 srgb_pow, srgb_mul_high, srgb_sub_high, srgb_mul_low);
1199 shader_addline(buffer, "const vec4 srgb_const1 = vec4(%.8e, 0.0, 0.0, 0.0);\n",
1202 if (reg_maps->vpos || reg_maps->usesdsy)
1204 if (shader->limits.constant_float + extra_constants_needed
1205 + 1 < gl_info->limits.glsl_ps_float_constants)
1207 shader_addline(buffer, "uniform vec4 ycorrection;\n");
1208 extra_constants_needed++;
1212 /* This happens because we do not have proper tracking of the constant registers that are
1213 * actually used, only the max limit of the shader version
1215 FIXME("Cannot find a free uniform for vpos correction params\n");
1216 shader_addline(buffer, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
1217 context->render_offscreen ? 0.0f : fb->render_targets[0]->resource.height,
1218 context->render_offscreen ? 1.0f : -1.0f);
1220 shader_addline(buffer, "vec4 vpos;\n");
1224 /* Declare output register temporaries */
1225 if (shader->limits.packed_output)
1226 shader_addline(buffer, "vec4 %s_out[%u];\n", prefix, shader->limits.packed_output);
1228 /* Declare temporary variables */
1229 for (i = 0, map = reg_maps->temporary; map; map >>= 1, ++i)
1231 if (map & 1) shader_addline(buffer, "vec4 R%u;\n", i);
1234 /* Declare loop registers aLx */
1235 if (version->major < 4)
1237 for (i = 0; i < reg_maps->loop_depth; ++i)
1239 shader_addline(buffer, "int aL%u;\n", i);
1240 shader_addline(buffer, "int tmpInt%u;\n", i);
1244 /* Temporary variables for matrix operations */
1245 shader_addline(buffer, "vec4 tmp0;\n");
1246 shader_addline(buffer, "vec4 tmp1;\n");
1248 if (!shader->load_local_constsF)
1250 LIST_FOR_EACH_ENTRY(lconst, &shader->constantsF, struct wined3d_shader_lconst, entry)
1253 value = (const float *)lconst->value;
1254 shader_addline(buffer, "const vec4 %s_lc%u = vec4(%.8e, %.8e, %.8e, %.8e);\n",
1255 prefix, lconst->idx, value[0], value[1], value[2], value[3]);
1259 /* Start the main program. */
1260 shader_addline(buffer, "void main()\n{\n");
1262 /* Direct3D applications expect integer vPos values, while OpenGL drivers
1263 * add approximately 0.5. This causes off-by-one problems as spotted by
1264 * the vPos d3d9 visual test. Unfortunately ATI cards do not add exactly
1265 * 0.5, but rather something like 0.49999999 or 0.50000001, which still
1266 * causes precision troubles when we just subtract 0.5.
1268 * To deal with that, just floor() the position. This will eliminate the
1269 * fraction on all cards.
1271 * TODO: Test how this behaves with multisampling.
1273 * An advantage of floor is that it works even if the driver doesn't add
1274 * 0.5. It is somewhat questionable if 1.5, 2.5, ... are the proper values
1275 * to return in gl_FragCoord, even though coordinates specify the pixel
1276 * centers instead of the pixel corners. This code will behave correctly
1277 * on drivers that returns integer values. */
1278 if (version->type == WINED3D_SHADER_TYPE_PIXEL && reg_maps->vpos)
1279 shader_addline(buffer,
1280 "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
1283 /*****************************************************************************
1284 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
1286 * For more information, see http://wiki.winehq.org/DirectX-Shaders
1287 ****************************************************************************/
1290 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction *ins,
1291 const struct wined3d_shader_src_param *wined3d_src, DWORD mask, struct glsl_src_param *glsl_src);
1293 /** Used for opcode modifiers - They multiply the result by the specified amount */
1294 static const char * const shift_glsl_tab[] = {
1296 "2.0 * ", /* 1 (x2) */
1297 "4.0 * ", /* 2 (x4) */
1298 "8.0 * ", /* 3 (x8) */
1299 "16.0 * ", /* 4 (x16) */
1300 "32.0 * ", /* 5 (x32) */
1307 "0.0625 * ", /* 12 (d16) */
1308 "0.125 * ", /* 13 (d8) */
1309 "0.25 * ", /* 14 (d4) */
1310 "0.5 * " /* 15 (d2) */
1313 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
1314 static void shader_glsl_gen_modifier(enum wined3d_shader_src_modifier src_modifier,
1315 const char *in_reg, const char *in_regswizzle, char *out_str)
1319 switch (src_modifier)
1321 case WINED3DSPSM_DZ: /* Need to handle this in the instructions itself (texld & texcrd). */
1322 case WINED3DSPSM_DW:
1323 case WINED3DSPSM_NONE:
1324 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
1326 case WINED3DSPSM_NEG:
1327 sprintf(out_str, "-%s%s", in_reg, in_regswizzle);
1329 case WINED3DSPSM_NOT:
1330 sprintf(out_str, "!%s%s", in_reg, in_regswizzle);
1332 case WINED3DSPSM_BIAS:
1333 sprintf(out_str, "(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
1335 case WINED3DSPSM_BIASNEG:
1336 sprintf(out_str, "-(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
1338 case WINED3DSPSM_SIGN:
1339 sprintf(out_str, "(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
1341 case WINED3DSPSM_SIGNNEG:
1342 sprintf(out_str, "-(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
1344 case WINED3DSPSM_COMP:
1345 sprintf(out_str, "(1.0 - %s%s)", in_reg, in_regswizzle);
1347 case WINED3DSPSM_X2:
1348 sprintf(out_str, "(2.0 * %s%s)", in_reg, in_regswizzle);
1350 case WINED3DSPSM_X2NEG:
1351 sprintf(out_str, "-(2.0 * %s%s)", in_reg, in_regswizzle);
1353 case WINED3DSPSM_ABS:
1354 sprintf(out_str, "abs(%s%s)", in_reg, in_regswizzle);
1356 case WINED3DSPSM_ABSNEG:
1357 sprintf(out_str, "-abs(%s%s)", in_reg, in_regswizzle);
1360 FIXME("Unhandled modifier %u\n", src_modifier);
1361 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
1365 /** Writes the GLSL variable name that corresponds to the register that the
1366 * DX opcode parameter is trying to access */
1367 static void shader_glsl_get_register_name(const struct wined3d_shader_register *reg,
1368 char *register_name, BOOL *is_color, const struct wined3d_shader_instruction *ins)
1370 /* oPos, oFog and oPts in D3D */
1371 static const char * const hwrastout_reg_names[] = {"vs_out[10]", "vs_out[11].x", "vs_out[11].y"};
1373 const struct wined3d_shader *shader = ins->ctx->shader;
1374 const struct wined3d_shader_reg_maps *reg_maps = ins->ctx->reg_maps;
1375 const struct wined3d_shader_version *version = ®_maps->shader_version;
1376 const struct wined3d_gl_info *gl_info = ins->ctx->gl_info;
1377 const char *prefix = shader_glsl_get_prefix(version->type);
1378 struct glsl_src_param rel_param0, rel_param1;
1380 if (reg->idx[0].offset != ~0U && reg->idx[0].rel_addr)
1381 shader_glsl_add_src_param(ins, reg->idx[0].rel_addr, WINED3DSP_WRITEMASK_0, &rel_param0);
1382 if (reg->idx[1].offset != ~0U && reg->idx[1].rel_addr)
1383 shader_glsl_add_src_param(ins, reg->idx[1].rel_addr, WINED3DSP_WRITEMASK_0, &rel_param1);
1388 case WINED3DSPR_TEMP:
1389 sprintf(register_name, "R%u", reg->idx[0].offset);
1392 case WINED3DSPR_INPUT:
1393 /* vertex shaders */
1394 if (version->type == WINED3D_SHADER_TYPE_VERTEX)
1396 struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
1397 if (priv->cur_vs_args->swizzle_map & (1 << reg->idx[0].offset))
1399 sprintf(register_name, "%s_in%u", prefix, reg->idx[0].offset);
1403 if (version->type == WINED3D_SHADER_TYPE_GEOMETRY)
1405 if (reg->idx[0].rel_addr)
1407 if (reg->idx[1].rel_addr)
1408 sprintf(register_name, "gs_in[%s + %u][%s + %u]",
1409 rel_param0.param_str, reg->idx[0].offset, rel_param1.param_str, reg->idx[1].offset);
1411 sprintf(register_name, "gs_in[%s + %u][%u]",
1412 rel_param0.param_str, reg->idx[0].offset, reg->idx[1].offset);
1414 else if (reg->idx[1].rel_addr)
1415 sprintf(register_name, "gs_in[%u][%s + %u]",
1416 reg->idx[0].offset, rel_param1.param_str, reg->idx[1].offset);
1418 sprintf(register_name, "gs_in[%u][%u]", reg->idx[0].offset, reg->idx[1].offset);
1422 /* pixel shaders >= 3.0 */
1423 if (version->major >= 3)
1425 DWORD idx = shader->u.ps.input_reg_map[reg->idx[0].offset];
1426 unsigned int in_count = vec4_varyings(version->major, gl_info);
1428 if (reg->idx[0].rel_addr)
1430 /* Removing a + 0 would be an obvious optimization, but
1431 * OS X doesn't see the NOP operation there. */
1434 if (shader->u.ps.declared_in_count > in_count)
1436 sprintf(register_name,
1437 "((%s + %u) > %u ? (%s + %u) > %u ? gl_SecondaryColor : gl_Color : %s_in[%s + %u])",
1438 rel_param0.param_str, idx, in_count - 1, rel_param0.param_str, idx, in_count,
1439 prefix, rel_param0.param_str, idx);
1443 sprintf(register_name, "%s_in[%s + %u]", prefix, rel_param0.param_str, idx);
1448 if (shader->u.ps.declared_in_count > in_count)
1450 sprintf(register_name, "((%s) > %u ? (%s) > %u ? gl_SecondaryColor : gl_Color : %s_in[%s])",
1451 rel_param0.param_str, in_count - 1, rel_param0.param_str, in_count,
1452 prefix, rel_param0.param_str);
1456 sprintf(register_name, "%s_in[%s]", prefix, rel_param0.param_str);
1462 if (idx == in_count) sprintf(register_name, "gl_Color");
1463 else if (idx == in_count + 1) sprintf(register_name, "gl_SecondaryColor");
1464 else sprintf(register_name, "%s_in[%u]", prefix, idx);
1469 if (!reg->idx[0].offset)
1470 strcpy(register_name, "gl_Color");
1472 strcpy(register_name, "gl_SecondaryColor");
1477 case WINED3DSPR_CONST:
1479 /* Relative addressing */
1480 if (reg->idx[0].rel_addr)
1482 if (reg->idx[0].offset)
1483 sprintf(register_name, "%s_c[%s + %u]", prefix, rel_param0.param_str, reg->idx[0].offset);
1485 sprintf(register_name, "%s_c[%s]", prefix, rel_param0.param_str);
1489 if (shader_constant_is_local(shader, reg->idx[0].offset))
1490 sprintf(register_name, "%s_lc%u", prefix, reg->idx[0].offset);
1492 sprintf(register_name, "%s_c[%u]", prefix, reg->idx[0].offset);
1497 case WINED3DSPR_CONSTINT:
1498 sprintf(register_name, "%s_i[%u]", prefix, reg->idx[0].offset);
1501 case WINED3DSPR_CONSTBOOL:
1502 sprintf(register_name, "%s_b[%u]", prefix, reg->idx[0].offset);
1505 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
1506 if (version->type == WINED3D_SHADER_TYPE_PIXEL)
1507 sprintf(register_name, "T%u", reg->idx[0].offset);
1509 sprintf(register_name, "A%u", reg->idx[0].offset);
1512 case WINED3DSPR_LOOP:
1513 sprintf(register_name, "aL%u", ins->ctx->loop_state->current_reg - 1);
1516 case WINED3DSPR_SAMPLER:
1517 sprintf(register_name, "%s_sampler%u", prefix, reg->idx[0].offset);
1520 case WINED3DSPR_COLOROUT:
1521 if (reg->idx[0].offset >= gl_info->limits.buffers)
1522 WARN("Write to render target %u, only %d supported.\n",
1523 reg->idx[0].offset, gl_info->limits.buffers);
1525 sprintf(register_name, "gl_FragData[%u]", reg->idx[0].offset);
1528 case WINED3DSPR_RASTOUT:
1529 sprintf(register_name, "%s", hwrastout_reg_names[reg->idx[0].offset]);
1532 case WINED3DSPR_DEPTHOUT:
1533 sprintf(register_name, "gl_FragDepth");
1536 case WINED3DSPR_ATTROUT:
1537 if (!reg->idx[0].offset)
1538 sprintf(register_name, "%s_out[8]", prefix);
1540 sprintf(register_name, "%s_out[9]", prefix);
1543 case WINED3DSPR_TEXCRDOUT:
1544 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1545 sprintf(register_name, "%s_out[%u]", prefix, reg->idx[0].offset);
1548 case WINED3DSPR_MISCTYPE:
1549 if (!reg->idx[0].offset)
1552 sprintf(register_name, "vpos");
1554 else if (reg->idx[0].offset == 1)
1556 /* Note that gl_FrontFacing is a bool, while vFace is
1557 * a float for which the sign determines front/back */
1558 sprintf(register_name, "(gl_FrontFacing ? 1.0 : -1.0)");
1562 FIXME("Unhandled misctype register %u.\n", reg->idx[0].offset);
1563 sprintf(register_name, "unrecognized_register");
1567 case WINED3DSPR_IMMCONST:
1568 switch (reg->immconst_type)
1570 case WINED3D_IMMCONST_SCALAR:
1571 switch (reg->data_type)
1573 case WINED3D_DATA_FLOAT:
1574 sprintf(register_name, "%.8e", *(const float *)reg->immconst_data);
1576 case WINED3D_DATA_INT:
1577 sprintf(register_name, "%#x", reg->immconst_data[0]);
1579 case WINED3D_DATA_RESOURCE:
1580 case WINED3D_DATA_SAMPLER:
1581 case WINED3D_DATA_UINT:
1582 sprintf(register_name, "%#xu", reg->immconst_data[0]);
1585 sprintf(register_name, "<unhandled data type %#x>", reg->data_type);
1590 case WINED3D_IMMCONST_VEC4:
1591 switch (reg->data_type)
1593 case WINED3D_DATA_FLOAT:
1594 sprintf(register_name, "vec4(%.8e, %.8e, %.8e, %.8e)",
1595 *(const float *)®->immconst_data[0], *(const float *)®->immconst_data[1],
1596 *(const float *)®->immconst_data[2], *(const float *)®->immconst_data[3]);
1598 case WINED3D_DATA_INT:
1599 sprintf(register_name, "ivec4(%#x, %#x, %#x, %#x)",
1600 reg->immconst_data[0], reg->immconst_data[1],
1601 reg->immconst_data[2], reg->immconst_data[3]);
1603 case WINED3D_DATA_RESOURCE:
1604 case WINED3D_DATA_SAMPLER:
1605 case WINED3D_DATA_UINT:
1606 sprintf(register_name, "uvec4(%#xu, %#xu, %#xu, %#xu)",
1607 reg->immconst_data[0], reg->immconst_data[1],
1608 reg->immconst_data[2], reg->immconst_data[3]);
1611 sprintf(register_name, "<unhandled data type %#x>", reg->data_type);
1617 FIXME("Unhandled immconst type %#x\n", reg->immconst_type);
1618 sprintf(register_name, "<unhandled_immconst_type %#x>", reg->immconst_type);
1622 case WINED3DSPR_CONSTBUFFER:
1623 if (reg->idx[1].rel_addr)
1624 sprintf(register_name, "%s_cb%u[%s + %u]",
1625 prefix, reg->idx[0].offset, rel_param1.param_str, reg->idx[1].offset);
1627 sprintf(register_name, "%s_cb%u[%u]", prefix, reg->idx[0].offset, reg->idx[1].offset);
1630 case WINED3DSPR_PRIMID:
1631 sprintf(register_name, "uint(gl_PrimitiveIDIn)");
1635 FIXME("Unhandled register type %#x.\n", reg->type);
1636 sprintf(register_name, "unrecognized_register");
1641 static void shader_glsl_write_mask_to_str(DWORD write_mask, char *str)
1644 if (write_mask & WINED3DSP_WRITEMASK_0) *str++ = 'x';
1645 if (write_mask & WINED3DSP_WRITEMASK_1) *str++ = 'y';
1646 if (write_mask & WINED3DSP_WRITEMASK_2) *str++ = 'z';
1647 if (write_mask & WINED3DSP_WRITEMASK_3) *str++ = 'w';
1651 /* Get the GLSL write mask for the destination register */
1652 static DWORD shader_glsl_get_write_mask(const struct wined3d_shader_dst_param *param, char *write_mask)
1654 DWORD mask = param->write_mask;
1656 if (shader_is_scalar(¶m->reg))
1658 mask = WINED3DSP_WRITEMASK_0;
1663 shader_glsl_write_mask_to_str(mask, write_mask);
1669 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask) {
1670 unsigned int size = 0;
1672 if (write_mask & WINED3DSP_WRITEMASK_0) ++size;
1673 if (write_mask & WINED3DSP_WRITEMASK_1) ++size;
1674 if (write_mask & WINED3DSP_WRITEMASK_2) ++size;
1675 if (write_mask & WINED3DSP_WRITEMASK_3) ++size;
1680 static void shader_glsl_swizzle_to_str(const DWORD swizzle, BOOL fixup, DWORD mask, char *str)
1682 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1683 * but addressed as "rgba". To fix this we need to swap the register's x
1684 * and z components. */
1685 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
1688 /* swizzle bits fields: wwzzyyxx */
1689 if (mask & WINED3DSP_WRITEMASK_0) *str++ = swizzle_chars[swizzle & 0x03];
1690 if (mask & WINED3DSP_WRITEMASK_1) *str++ = swizzle_chars[(swizzle >> 2) & 0x03];
1691 if (mask & WINED3DSP_WRITEMASK_2) *str++ = swizzle_chars[(swizzle >> 4) & 0x03];
1692 if (mask & WINED3DSP_WRITEMASK_3) *str++ = swizzle_chars[(swizzle >> 6) & 0x03];
1696 static void shader_glsl_get_swizzle(const struct wined3d_shader_src_param *param,
1697 BOOL fixup, DWORD mask, char *swizzle_str)
1699 if (shader_is_scalar(¶m->reg))
1700 *swizzle_str = '\0';
1702 shader_glsl_swizzle_to_str(param->swizzle, fixup, mask, swizzle_str);
1705 /* From a given parameter token, generate the corresponding GLSL string.
1706 * Also, return the actual register name and swizzle in case the
1707 * caller needs this information as well. */
1708 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction *ins,
1709 const struct wined3d_shader_src_param *wined3d_src, DWORD mask, struct glsl_src_param *glsl_src)
1711 BOOL is_color = FALSE;
1712 char swizzle_str[6];
1714 glsl_src->reg_name[0] = '\0';
1715 glsl_src->param_str[0] = '\0';
1716 swizzle_str[0] = '\0';
1718 shader_glsl_get_register_name(&wined3d_src->reg, glsl_src->reg_name, &is_color, ins);
1719 shader_glsl_get_swizzle(wined3d_src, is_color, mask, swizzle_str);
1721 if (wined3d_src->reg.type == WINED3DSPR_IMMCONST || wined3d_src->reg.type == WINED3DSPR_PRIMID)
1723 shader_glsl_gen_modifier(wined3d_src->modifiers, glsl_src->reg_name, swizzle_str, glsl_src->param_str);
1727 char param_str[200];
1729 shader_glsl_gen_modifier(wined3d_src->modifiers, glsl_src->reg_name, swizzle_str, param_str);
1731 switch (wined3d_src->reg.data_type)
1733 case WINED3D_DATA_FLOAT:
1734 sprintf(glsl_src->param_str, "%s", param_str);
1736 case WINED3D_DATA_INT:
1737 sprintf(glsl_src->param_str, "floatBitsToInt(%s)", param_str);
1739 case WINED3D_DATA_RESOURCE:
1740 case WINED3D_DATA_SAMPLER:
1741 case WINED3D_DATA_UINT:
1742 sprintf(glsl_src->param_str, "floatBitsToUint(%s)", param_str);
1745 FIXME("Unhandled data type %#x.\n", wined3d_src->reg.data_type);
1746 sprintf(glsl_src->param_str, "%s", param_str);
1752 /* From a given parameter token, generate the corresponding GLSL string.
1753 * Also, return the actual register name and swizzle in case the
1754 * caller needs this information as well. */
1755 static DWORD shader_glsl_add_dst_param(const struct wined3d_shader_instruction *ins,
1756 const struct wined3d_shader_dst_param *wined3d_dst, struct glsl_dst_param *glsl_dst)
1758 BOOL is_color = FALSE;
1760 glsl_dst->mask_str[0] = '\0';
1761 glsl_dst->reg_name[0] = '\0';
1763 shader_glsl_get_register_name(&wined3d_dst->reg, glsl_dst->reg_name, &is_color, ins);
1764 return shader_glsl_get_write_mask(wined3d_dst, glsl_dst->mask_str);
1767 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1768 static DWORD shader_glsl_append_dst_ext(struct wined3d_shader_buffer *buffer,
1769 const struct wined3d_shader_instruction *ins, const struct wined3d_shader_dst_param *dst)
1771 struct glsl_dst_param glsl_dst;
1774 if ((mask = shader_glsl_add_dst_param(ins, dst, &glsl_dst)))
1776 switch (dst->reg.data_type)
1778 case WINED3D_DATA_FLOAT:
1779 shader_addline(buffer, "%s%s = %s(",
1780 glsl_dst.reg_name, glsl_dst.mask_str, shift_glsl_tab[dst->shift]);
1782 case WINED3D_DATA_INT:
1783 shader_addline(buffer, "%s%s = %sintBitsToFloat(",
1784 glsl_dst.reg_name, glsl_dst.mask_str, shift_glsl_tab[dst->shift]);
1786 case WINED3D_DATA_RESOURCE:
1787 case WINED3D_DATA_SAMPLER:
1788 case WINED3D_DATA_UINT:
1789 shader_addline(buffer, "%s%s = %suintBitsToFloat(",
1790 glsl_dst.reg_name, glsl_dst.mask_str, shift_glsl_tab[dst->shift]);
1793 FIXME("Unhandled data type %#x.\n", dst->reg.data_type);
1794 shader_addline(buffer, "%s%s = %s(",
1795 glsl_dst.reg_name, glsl_dst.mask_str, shift_glsl_tab[dst->shift]);
1803 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1804 static DWORD shader_glsl_append_dst(struct wined3d_shader_buffer *buffer, const struct wined3d_shader_instruction *ins)
1806 return shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0]);
1809 /** Process GLSL instruction modifiers */
1810 static void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction *ins)
1812 struct glsl_dst_param dst_param;
1815 if (!ins->dst_count) return;
1817 modifiers = ins->dst[0].modifiers;
1818 if (!modifiers) return;
1820 shader_glsl_add_dst_param(ins, &ins->dst[0], &dst_param);
1822 if (modifiers & WINED3DSPDM_SATURATE)
1824 /* _SAT means to clamp the value of the register to between 0 and 1 */
1825 shader_addline(ins->ctx->buffer, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param.reg_name,
1826 dst_param.mask_str, dst_param.reg_name, dst_param.mask_str);
1829 if (modifiers & WINED3DSPDM_MSAMPCENTROID)
1831 FIXME("_centroid modifier not handled\n");
1834 if (modifiers & WINED3DSPDM_PARTIALPRECISION)
1836 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1840 static const char *shader_glsl_get_rel_op(enum wined3d_shader_rel_op op)
1844 case WINED3D_SHADER_REL_OP_GT: return ">";
1845 case WINED3D_SHADER_REL_OP_EQ: return "==";
1846 case WINED3D_SHADER_REL_OP_GE: return ">=";
1847 case WINED3D_SHADER_REL_OP_LT: return "<";
1848 case WINED3D_SHADER_REL_OP_NE: return "!=";
1849 case WINED3D_SHADER_REL_OP_LE: return "<=";
1851 FIXME("Unrecognized operator %#x.\n", op);
1856 static void shader_glsl_get_sample_function(const struct wined3d_shader_context *ctx,
1857 DWORD sampler_idx, DWORD flags, struct glsl_sample_function *sample_function)
1859 enum wined3d_sampler_texture_type sampler_type = ctx->reg_maps->sampler_type[sampler_idx];
1860 const struct wined3d_gl_info *gl_info = ctx->gl_info;
1861 BOOL shadow = ctx->reg_maps->shader_version.type == WINED3D_SHADER_TYPE_PIXEL
1862 && (((const struct shader_glsl_ctx_priv *)ctx->backend_data)->cur_ps_args->shadow & (1 << sampler_idx));
1863 BOOL projected = flags & WINED3D_GLSL_SAMPLE_PROJECTED;
1864 BOOL texrect = flags & WINED3D_GLSL_SAMPLE_NPOT && gl_info->supported[ARB_TEXTURE_RECTANGLE];
1865 BOOL lod = flags & WINED3D_GLSL_SAMPLE_LOD;
1866 BOOL grad = flags & WINED3D_GLSL_SAMPLE_GRAD;
1868 /* Note that there's no such thing as a projected cube texture. */
1869 switch(sampler_type) {
1875 sample_function->name = projected ? "shadow1DProjLod" : "shadow1DLod";
1879 if (gl_info->supported[EXT_GPU_SHADER4])
1880 sample_function->name = projected ? "shadow1DProjGrad" : "shadow1DGrad";
1881 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
1882 sample_function->name = projected ? "shadow1DProjGradARB" : "shadow1DGradARB";
1885 FIXME("Unsupported 1D shadow grad function.\n");
1886 sample_function->name = "unsupported1DGrad";
1891 sample_function->name = projected ? "shadow1DProj" : "shadow1D";
1893 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1;
1899 sample_function->name = projected ? "texture1DProjLod" : "texture1DLod";
1903 if (gl_info->supported[EXT_GPU_SHADER4])
1904 sample_function->name = projected ? "texture1DProjGrad" : "texture1DGrad";
1905 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
1906 sample_function->name = projected ? "texture1DProjGradARB" : "texture1DGradARB";
1909 FIXME("Unsupported 1D grad function.\n");
1910 sample_function->name = "unsupported1DGrad";
1915 sample_function->name = projected ? "texture1DProj" : "texture1D";
1917 sample_function->coord_mask = WINED3DSP_WRITEMASK_0;
1928 sample_function->name = projected ? "shadow2DRectProjLod" : "shadow2DRectLod";
1932 if (gl_info->supported[EXT_GPU_SHADER4])
1933 sample_function->name = projected ? "shadow2DRectProjGrad" : "shadow2DRectGrad";
1934 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
1935 sample_function->name = projected ? "shadow2DRectProjGradARB" : "shadow2DRectGradARB";
1938 FIXME("Unsupported RECT shadow grad function.\n");
1939 sample_function->name = "unsupported2DRectGrad";
1944 sample_function->name = projected ? "shadow2DRectProj" : "shadow2DRect";
1951 sample_function->name = projected ? "shadow2DProjLod" : "shadow2DLod";
1955 if (gl_info->supported[EXT_GPU_SHADER4])
1956 sample_function->name = projected ? "shadow2DProjGrad" : "shadow2DGrad";
1957 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
1958 sample_function->name = projected ? "shadow2DProjGradARB" : "shadow2DGradARB";
1961 FIXME("Unsupported 2D shadow grad function.\n");
1962 sample_function->name = "unsupported2DGrad";
1967 sample_function->name = projected ? "shadow2DProj" : "shadow2D";
1970 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1978 sample_function->name = projected ? "texture2DRectProjLod" : "texture2DRectLod";
1982 if (gl_info->supported[EXT_GPU_SHADER4])
1983 sample_function->name = projected ? "texture2DRectProjGrad" : "texture2DRectGrad";
1984 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
1985 sample_function->name = projected ? "texture2DRectProjGradARB" : "texture2DRectGradARB";
1988 FIXME("Unsupported RECT grad function.\n");
1989 sample_function->name = "unsupported2DRectGrad";
1994 sample_function->name = projected ? "texture2DRectProj" : "texture2DRect";
2001 sample_function->name = projected ? "texture2DProjLod" : "texture2DLod";
2005 if (gl_info->supported[EXT_GPU_SHADER4])
2006 sample_function->name = projected ? "texture2DProjGrad" : "texture2DGrad";
2007 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
2008 sample_function->name = projected ? "texture2DProjGradARB" : "texture2DGradARB";
2011 FIXME("Unsupported 2D grad function.\n");
2012 sample_function->name = "unsupported2DGrad";
2017 sample_function->name = projected ? "texture2DProj" : "texture2D";
2020 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1;
2024 case WINED3DSTT_CUBE:
2027 FIXME("Unsupported Cube shadow function.\n");
2028 sample_function->name = "unsupportedCubeShadow";
2029 sample_function->coord_mask = 0;
2035 sample_function->name = "textureCubeLod";
2039 if (gl_info->supported[EXT_GPU_SHADER4])
2040 sample_function->name = "textureCubeGrad";
2041 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
2042 sample_function->name = "textureCubeGradARB";
2045 FIXME("Unsupported Cube grad function.\n");
2046 sample_function->name = "unsupportedCubeGrad";
2051 sample_function->name = "textureCube";
2053 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2057 case WINED3DSTT_VOLUME:
2060 FIXME("Unsupported 3D shadow function.\n");
2061 sample_function->name = "unsupported3DShadow";
2062 sample_function->coord_mask = 0;
2068 sample_function->name = projected ? "texture3DProjLod" : "texture3DLod";
2072 if (gl_info->supported[EXT_GPU_SHADER4])
2073 sample_function->name = projected ? "texture3DProjGrad" : "texture3DGrad";
2074 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
2075 sample_function->name = projected ? "texture3DProjGradARB" : "texture3DGradARB";
2078 FIXME("Unsupported 3D grad function.\n");
2079 sample_function->name = "unsupported3DGrad";
2084 sample_function->name = projected ? "texture3DProj" : "texture3D";
2086 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2091 sample_function->name = "";
2092 sample_function->coord_mask = 0;
2093 FIXME("Unrecognized sampler type: %#x;\n", sampler_type);
2098 static void shader_glsl_append_fixup_arg(char *arguments, const char *reg_name,
2099 BOOL sign_fixup, enum fixup_channel_source channel_source)
2101 switch(channel_source)
2103 case CHANNEL_SOURCE_ZERO:
2104 strcat(arguments, "0.0");
2107 case CHANNEL_SOURCE_ONE:
2108 strcat(arguments, "1.0");
2111 case CHANNEL_SOURCE_X:
2112 strcat(arguments, reg_name);
2113 strcat(arguments, ".x");
2116 case CHANNEL_SOURCE_Y:
2117 strcat(arguments, reg_name);
2118 strcat(arguments, ".y");
2121 case CHANNEL_SOURCE_Z:
2122 strcat(arguments, reg_name);
2123 strcat(arguments, ".z");
2126 case CHANNEL_SOURCE_W:
2127 strcat(arguments, reg_name);
2128 strcat(arguments, ".w");
2132 FIXME("Unhandled channel source %#x\n", channel_source);
2133 strcat(arguments, "undefined");
2137 if (sign_fixup) strcat(arguments, " * 2.0 - 1.0");
2140 static void shader_glsl_color_correction_ext(struct wined3d_shader_buffer *buffer,
2141 const char *reg_name, DWORD mask, struct color_fixup_desc fixup)
2143 unsigned int mask_size, remaining;
2144 DWORD fixup_mask = 0;
2145 char arguments[256];
2148 if (fixup.x_sign_fixup || fixup.x_source != CHANNEL_SOURCE_X) fixup_mask |= WINED3DSP_WRITEMASK_0;
2149 if (fixup.y_sign_fixup || fixup.y_source != CHANNEL_SOURCE_Y) fixup_mask |= WINED3DSP_WRITEMASK_1;
2150 if (fixup.z_sign_fixup || fixup.z_source != CHANNEL_SOURCE_Z) fixup_mask |= WINED3DSP_WRITEMASK_2;
2151 if (fixup.w_sign_fixup || fixup.w_source != CHANNEL_SOURCE_W) fixup_mask |= WINED3DSP_WRITEMASK_3;
2152 if (!(mask &= fixup_mask))
2155 if (is_complex_fixup(fixup))
2157 enum complex_fixup complex_fixup = get_complex_fixup(fixup);
2158 FIXME("Complex fixup (%#x) not supported\n",complex_fixup);
2162 shader_glsl_write_mask_to_str(mask, mask_str);
2163 mask_size = shader_glsl_get_write_mask_size(mask);
2165 arguments[0] = '\0';
2166 remaining = mask_size;
2167 if (mask & WINED3DSP_WRITEMASK_0)
2169 shader_glsl_append_fixup_arg(arguments, reg_name, fixup.x_sign_fixup, fixup.x_source);
2170 if (--remaining) strcat(arguments, ", ");
2172 if (mask & WINED3DSP_WRITEMASK_1)
2174 shader_glsl_append_fixup_arg(arguments, reg_name, fixup.y_sign_fixup, fixup.y_source);
2175 if (--remaining) strcat(arguments, ", ");
2177 if (mask & WINED3DSP_WRITEMASK_2)
2179 shader_glsl_append_fixup_arg(arguments, reg_name, fixup.z_sign_fixup, fixup.z_source);
2180 if (--remaining) strcat(arguments, ", ");
2182 if (mask & WINED3DSP_WRITEMASK_3)
2184 shader_glsl_append_fixup_arg(arguments, reg_name, fixup.w_sign_fixup, fixup.w_source);
2185 if (--remaining) strcat(arguments, ", ");
2189 shader_addline(buffer, "%s%s = vec%u(%s);\n", reg_name, mask_str, mask_size, arguments);
2191 shader_addline(buffer, "%s%s = %s;\n", reg_name, mask_str, arguments);
2194 static void shader_glsl_color_correction(const struct wined3d_shader_instruction *ins, struct color_fixup_desc fixup)
2199 shader_glsl_get_register_name(&ins->dst[0].reg, reg_name, &is_color, ins);
2200 shader_glsl_color_correction_ext(ins->ctx->buffer, reg_name, ins->dst[0].write_mask, fixup);
2203 static void PRINTF_ATTR(8, 9) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction *ins,
2204 DWORD sampler, const struct glsl_sample_function *sample_function, DWORD swizzle,
2205 const char *dx, const char *dy, const char *bias, const char *coord_reg_fmt, ...)
2207 const struct wined3d_shader_version *version = &ins->ctx->reg_maps->shader_version;
2208 char dst_swizzle[6];
2209 struct color_fixup_desc fixup;
2210 BOOL np2_fixup = FALSE;
2213 shader_glsl_swizzle_to_str(swizzle, FALSE, ins->dst[0].write_mask, dst_swizzle);
2215 if (version->type == WINED3D_SHADER_TYPE_PIXEL)
2217 const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
2218 fixup = priv->cur_ps_args->color_fixup[sampler];
2220 if(priv->cur_ps_args->np2_fixup & (1 << sampler)) {
2222 FIXME("Biased sampling from NP2 textures is unsupported\n");
2230 fixup = COLOR_FIXUP_IDENTITY; /* FIXME: Vshader color fixup */
2233 shader_glsl_append_dst(ins->ctx->buffer, ins);
2235 shader_addline(ins->ctx->buffer, "%s(%s_sampler%u, ",
2236 sample_function->name, shader_glsl_get_prefix(version->type), sampler);
2238 va_start(args, coord_reg_fmt);
2239 shader_vaddline(ins->ctx->buffer, coord_reg_fmt, args);
2243 shader_addline(ins->ctx->buffer, ", %s)%s);\n", bias, dst_swizzle);
2246 const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
2247 const unsigned char idx = priv->cur_np2fixup_info->idx[sampler];
2249 shader_addline(ins->ctx->buffer, " * ps_samplerNP2Fixup[%u].%s)%s);\n", idx >> 1,
2250 (idx % 2) ? "zw" : "xy", dst_swizzle);
2251 } else if(dx && dy) {
2252 shader_addline(ins->ctx->buffer, ", %s, %s)%s);\n", dx, dy, dst_swizzle);
2254 shader_addline(ins->ctx->buffer, ")%s);\n", dst_swizzle);
2258 if(!is_identity_fixup(fixup)) {
2259 shader_glsl_color_correction(ins, fixup);
2263 /*****************************************************************************
2264 * Begin processing individual instruction opcodes
2265 ****************************************************************************/
2267 static void shader_glsl_binop(const struct wined3d_shader_instruction *ins)
2269 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2270 struct glsl_src_param src0_param;
2271 struct glsl_src_param src1_param;
2275 /* Determine the GLSL operator to use based on the opcode */
2276 switch (ins->handler_idx)
2278 case WINED3DSIH_ADD: op = "+"; break;
2279 case WINED3DSIH_AND: op = "&"; break;
2280 case WINED3DSIH_DIV: op = "/"; break;
2281 case WINED3DSIH_IADD: op = "+"; break;
2282 case WINED3DSIH_MUL: op = "*"; break;
2283 case WINED3DSIH_SUB: op = "-"; break;
2284 case WINED3DSIH_USHR: op = ">>"; break;
2285 case WINED3DSIH_XOR: op = "^"; break;
2287 op = "<unhandled operator>";
2288 FIXME("Opcode %#x not yet handled in GLSL\n", ins->handler_idx);
2292 write_mask = shader_glsl_append_dst(buffer, ins);
2293 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2294 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2295 shader_addline(buffer, "%s %s %s);\n", src0_param.param_str, op, src1_param.param_str);
2298 static void shader_glsl_relop(const struct wined3d_shader_instruction *ins)
2300 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2301 struct glsl_src_param src0_param;
2302 struct glsl_src_param src1_param;
2303 unsigned int mask_size;
2307 write_mask = shader_glsl_append_dst(buffer, ins);
2308 mask_size = shader_glsl_get_write_mask_size(write_mask);
2309 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2310 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2314 switch (ins->handler_idx)
2316 case WINED3DSIH_EQ: op = "equal"; break;
2317 case WINED3DSIH_GE: op = "greaterThanEqual"; break;
2318 case WINED3DSIH_IGE: op = "greaterThanEqual"; break;
2319 case WINED3DSIH_LT: op = "lessThan"; break;
2321 op = "<unhandled operator>";
2322 ERR("Unhandled opcode %#x.\n", ins->handler_idx);
2326 shader_addline(buffer, "uvec%u(%s(%s, %s)) * 0xffffffffu);\n",
2327 mask_size, op, src0_param.param_str, src1_param.param_str);
2331 switch (ins->handler_idx)
2333 case WINED3DSIH_EQ: op = "=="; break;
2334 case WINED3DSIH_GE: op = ">="; break;
2335 case WINED3DSIH_IGE: op = ">="; break;
2336 case WINED3DSIH_LT: op = "<"; break;
2338 op = "<unhandled operator>";
2339 ERR("Unhandled opcode %#x.\n", ins->handler_idx);
2343 shader_addline(buffer, "%s %s %s ? 0xffffffffu : 0u);\n",
2344 src0_param.param_str, op, src1_param.param_str);
2348 static void shader_glsl_imul(const struct wined3d_shader_instruction *ins)
2350 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2351 struct glsl_src_param src0_param;
2352 struct glsl_src_param src1_param;
2355 /* If we have ARB_gpu_shader5 or GLSL 4.0, we can use imulExtended(). If
2356 * not, we can emulate it. */
2357 if (ins->dst[0].reg.type != WINED3DSPR_NULL)
2358 FIXME("64-bit integer multiplies not implemented.\n");
2360 if (ins->dst[1].reg.type != WINED3DSPR_NULL)
2362 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[1]);
2363 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2364 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2366 shader_addline(ins->ctx->buffer, "%s * %s);\n",
2367 src0_param.param_str, src1_param.param_str);
2371 static void shader_glsl_udiv(const struct wined3d_shader_instruction *ins)
2373 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2374 struct glsl_src_param src0_param, src1_param;
2377 if (ins->dst[0].reg.type != WINED3DSPR_NULL)
2380 if (ins->dst[1].reg.type != WINED3DSPR_NULL)
2384 write_mask = shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
2385 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2386 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2387 shader_addline(buffer, "tmp0%s = %s / %s;\n",
2388 dst_mask, src0_param.param_str, src1_param.param_str);
2390 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[1]);
2391 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2392 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2393 shader_addline(buffer, "%s %% %s));\n", src0_param.param_str, src1_param.param_str);
2395 shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0]);
2396 shader_addline(buffer, "tmp0%s);\n", dst_mask);
2400 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0]);
2401 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2402 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2403 shader_addline(buffer, "%s / %s);\n", src0_param.param_str, src1_param.param_str);
2406 else if (ins->dst[1].reg.type != WINED3DSPR_NULL)
2408 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[1]);
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);
2415 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
2416 static void shader_glsl_mov(const struct wined3d_shader_instruction *ins)
2418 const struct wined3d_gl_info *gl_info = ins->ctx->gl_info;
2419 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2420 struct glsl_src_param src0_param;
2423 write_mask = shader_glsl_append_dst(buffer, ins);
2424 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2426 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
2427 * shader versions WINED3DSIO_MOVA is used for this. */
2428 if (ins->ctx->reg_maps->shader_version.major == 1
2429 && ins->ctx->reg_maps->shader_version.type == WINED3D_SHADER_TYPE_VERTEX
2430 && ins->dst[0].reg.type == WINED3DSPR_ADDR)
2432 /* This is a simple floor() */
2433 unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
2434 if (mask_size > 1) {
2435 shader_addline(buffer, "ivec%d(floor(%s)));\n", mask_size, src0_param.param_str);
2437 shader_addline(buffer, "int(floor(%s)));\n", src0_param.param_str);
2440 else if(ins->handler_idx == WINED3DSIH_MOVA)
2442 /* We need to *round* to the nearest int here. */
2443 unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
2445 if (gl_info->supported[EXT_GPU_SHADER4])
2448 shader_addline(buffer, "ivec%d(round(%s)));\n", mask_size, src0_param.param_str);
2450 shader_addline(buffer, "int(round(%s)));\n", src0_param.param_str);
2455 shader_addline(buffer, "ivec%d(floor(abs(%s) + vec%d(0.5)) * sign(%s)));\n",
2456 mask_size, src0_param.param_str, mask_size, src0_param.param_str);
2458 shader_addline(buffer, "int(floor(abs(%s) + 0.5) * sign(%s)));\n",
2459 src0_param.param_str, src0_param.param_str);
2464 shader_addline(buffer, "%s);\n", src0_param.param_str);
2468 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
2469 static void shader_glsl_dot(const struct wined3d_shader_instruction *ins)
2471 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2472 struct glsl_src_param src0_param;
2473 struct glsl_src_param src1_param;
2474 DWORD dst_write_mask, src_write_mask;
2475 unsigned int dst_size = 0;
2477 dst_write_mask = shader_glsl_append_dst(buffer, ins);
2478 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
2480 /* dp3 works on vec3, dp4 on vec4 */
2481 if (ins->handler_idx == WINED3DSIH_DP4)
2483 src_write_mask = WINED3DSP_WRITEMASK_ALL;
2485 src_write_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2488 shader_glsl_add_src_param(ins, &ins->src[0], src_write_mask, &src0_param);
2489 shader_glsl_add_src_param(ins, &ins->src[1], src_write_mask, &src1_param);
2492 shader_addline(buffer, "vec%d(dot(%s, %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
2494 shader_addline(buffer, "dot(%s, %s));\n", src0_param.param_str, src1_param.param_str);
2498 /* Note that this instruction has some restrictions. The destination write mask
2499 * can't contain the w component, and the source swizzles have to be .xyzw */
2500 static void shader_glsl_cross(const struct wined3d_shader_instruction *ins)
2502 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2503 struct glsl_src_param src0_param;
2504 struct glsl_src_param src1_param;
2507 shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
2508 shader_glsl_append_dst(ins->ctx->buffer, ins);
2509 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
2510 shader_glsl_add_src_param(ins, &ins->src[1], src_mask, &src1_param);
2511 shader_addline(ins->ctx->buffer, "cross(%s, %s)%s);\n", src0_param.param_str, src1_param.param_str, dst_mask);
2514 static void shader_glsl_cut(const struct wined3d_shader_instruction *ins)
2516 shader_addline(ins->ctx->buffer, "EndPrimitive();\n");
2519 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
2520 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
2521 * GLSL uses the value as-is. */
2522 static void shader_glsl_pow(const struct wined3d_shader_instruction *ins)
2524 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2525 struct glsl_src_param src0_param;
2526 struct glsl_src_param src1_param;
2527 DWORD dst_write_mask;
2528 unsigned int dst_size;
2530 dst_write_mask = shader_glsl_append_dst(buffer, ins);
2531 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
2533 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
2534 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0, &src1_param);
2538 shader_addline(buffer, "vec%u(%s == 0.0 ? 1.0 : pow(abs(%s), %s)));\n",
2539 dst_size, src1_param.param_str, src0_param.param_str, src1_param.param_str);
2543 shader_addline(buffer, "%s == 0.0 ? 1.0 : pow(abs(%s), %s));\n",
2544 src1_param.param_str, src0_param.param_str, src1_param.param_str);
2548 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
2549 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
2550 * GLSL uses the value as-is. */
2551 static void shader_glsl_log(const struct wined3d_shader_instruction *ins)
2553 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2554 struct glsl_src_param src0_param;
2555 DWORD dst_write_mask;
2556 unsigned int dst_size;
2558 dst_write_mask = shader_glsl_append_dst(buffer, ins);
2559 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
2561 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
2565 shader_addline(buffer, "vec%u(log2(abs(%s))));\n",
2566 dst_size, src0_param.param_str);
2570 shader_addline(buffer, "log2(abs(%s)));\n",
2571 src0_param.param_str);
2575 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
2576 static void shader_glsl_map2gl(const struct wined3d_shader_instruction *ins)
2578 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2579 struct glsl_src_param src_param;
2580 const char *instruction;
2584 /* Determine the GLSL function to use based on the opcode */
2585 /* TODO: Possibly make this a table for faster lookups */
2586 switch (ins->handler_idx)
2588 case WINED3DSIH_MIN: instruction = "min"; break;
2589 case WINED3DSIH_MAX: instruction = "max"; break;
2590 case WINED3DSIH_ABS: instruction = "abs"; break;
2591 case WINED3DSIH_FRC: instruction = "fract"; break;
2592 case WINED3DSIH_EXP: instruction = "exp2"; break;
2593 case WINED3DSIH_DSX: instruction = "dFdx"; break;
2594 case WINED3DSIH_DSY: instruction = "ycorrection.y * dFdy"; break;
2595 case WINED3DSIH_ROUND_NI: instruction = "floor"; break;
2596 default: instruction = "";
2597 FIXME("Opcode %#x not yet handled in GLSL\n", ins->handler_idx);
2601 write_mask = shader_glsl_append_dst(buffer, ins);
2603 shader_addline(buffer, "%s(", instruction);
2607 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src_param);
2608 shader_addline(buffer, "%s", src_param.param_str);
2609 for (i = 1; i < ins->src_count; ++i)
2611 shader_glsl_add_src_param(ins, &ins->src[i], write_mask, &src_param);
2612 shader_addline(buffer, ", %s", src_param.param_str);
2616 shader_addline(buffer, "));\n");
2619 static void shader_glsl_nop(const struct wined3d_shader_instruction *ins) {}
2621 static void shader_glsl_nrm(const struct wined3d_shader_instruction *ins)
2623 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2624 struct glsl_src_param src_param;
2625 unsigned int mask_size;
2629 write_mask = shader_glsl_get_write_mask(ins->dst, dst_mask);
2630 mask_size = shader_glsl_get_write_mask_size(write_mask);
2631 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src_param);
2633 shader_addline(buffer, "tmp0.x = dot(%s, %s);\n",
2634 src_param.param_str, src_param.param_str);
2635 shader_glsl_append_dst(buffer, ins);
2639 shader_addline(buffer, "tmp0.x == 0.0 ? vec%u(0.0) : (%s * inversesqrt(tmp0.x)));\n",
2640 mask_size, src_param.param_str);
2644 shader_addline(buffer, "tmp0.x == 0.0 ? 0.0 : (%s * inversesqrt(tmp0.x)));\n",
2645 src_param.param_str);
2649 /** Process the WINED3DSIO_EXPP instruction in GLSL:
2650 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
2651 * dst.x = 2^(floor(src))
2652 * dst.y = src - floor(src)
2653 * dst.z = 2^src (partial precision is allowed, but optional)
2655 * For 2.0 shaders, just do this (honoring writemask and swizzle):
2656 * dst = 2^src; (partial precision is allowed, but optional)
2658 static void shader_glsl_expp(const struct wined3d_shader_instruction *ins)
2660 struct glsl_src_param src_param;
2662 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src_param);
2664 if (ins->ctx->reg_maps->shader_version.major < 2)
2668 shader_addline(ins->ctx->buffer, "tmp0.x = exp2(floor(%s));\n", src_param.param_str);
2669 shader_addline(ins->ctx->buffer, "tmp0.y = %s - floor(%s);\n", src_param.param_str, src_param.param_str);
2670 shader_addline(ins->ctx->buffer, "tmp0.z = exp2(%s);\n", src_param.param_str);
2671 shader_addline(ins->ctx->buffer, "tmp0.w = 1.0;\n");
2673 shader_glsl_append_dst(ins->ctx->buffer, ins);
2674 shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
2675 shader_addline(ins->ctx->buffer, "tmp0%s);\n", dst_mask);
2678 unsigned int mask_size;
2680 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
2681 mask_size = shader_glsl_get_write_mask_size(write_mask);
2683 if (mask_size > 1) {
2684 shader_addline(ins->ctx->buffer, "vec%d(exp2(%s)));\n", mask_size, src_param.param_str);
2686 shader_addline(ins->ctx->buffer, "exp2(%s));\n", src_param.param_str);
2691 static void shader_glsl_to_int(const struct wined3d_shader_instruction *ins)
2693 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2694 struct glsl_src_param src_param;
2695 unsigned int mask_size;
2698 write_mask = shader_glsl_append_dst(buffer, ins);
2699 mask_size = shader_glsl_get_write_mask_size(write_mask);
2700 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src_param);
2703 shader_addline(buffer, "ivec%u(%s));\n", mask_size, src_param.param_str);
2705 shader_addline(buffer, "int(%s));\n", src_param.param_str);
2708 static void shader_glsl_to_float(const struct wined3d_shader_instruction *ins)
2710 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2711 struct glsl_src_param src_param;
2712 unsigned int mask_size;
2715 write_mask = shader_glsl_append_dst(buffer, ins);
2716 mask_size = shader_glsl_get_write_mask_size(write_mask);
2717 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src_param);
2720 shader_addline(buffer, "vec%u(%s));\n", mask_size, src_param.param_str);
2722 shader_addline(buffer, "float(%s));\n", src_param.param_str);
2725 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
2726 static void shader_glsl_rcp(const struct wined3d_shader_instruction *ins)
2728 struct glsl_src_param src_param;
2730 unsigned int mask_size;
2732 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
2733 mask_size = shader_glsl_get_write_mask_size(write_mask);
2734 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &src_param);
2738 shader_addline(ins->ctx->buffer, "vec%u(1.0 / %s));\n",
2739 mask_size, src_param.param_str);
2743 shader_addline(ins->ctx->buffer, "1.0 / %s);\n",
2744 src_param.param_str);
2748 static void shader_glsl_rsq(const struct wined3d_shader_instruction *ins)
2750 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
2751 struct glsl_src_param src_param;
2753 unsigned int mask_size;
2755 write_mask = shader_glsl_append_dst(buffer, ins);
2756 mask_size = shader_glsl_get_write_mask_size(write_mask);
2758 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &src_param);
2762 shader_addline(buffer, "vec%u(inversesqrt(abs(%s))));\n",
2763 mask_size, src_param.param_str);
2767 shader_addline(buffer, "inversesqrt(abs(%s)));\n",
2768 src_param.param_str);
2772 /** Process signed comparison opcodes in GLSL. */
2773 static void shader_glsl_compare(const struct wined3d_shader_instruction *ins)
2775 struct glsl_src_param src0_param;
2776 struct glsl_src_param src1_param;
2778 unsigned int mask_size;
2780 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
2781 mask_size = shader_glsl_get_write_mask_size(write_mask);
2782 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2783 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2785 if (mask_size > 1) {
2786 const char *compare;
2788 switch(ins->handler_idx)
2790 case WINED3DSIH_SLT: compare = "lessThan"; break;
2791 case WINED3DSIH_SGE: compare = "greaterThanEqual"; break;
2792 default: compare = "";
2793 FIXME("Can't handle opcode %#x\n", ins->handler_idx);
2796 shader_addline(ins->ctx->buffer, "vec%d(%s(%s, %s)));\n", mask_size, compare,
2797 src0_param.param_str, src1_param.param_str);
2799 switch(ins->handler_idx)
2801 case WINED3DSIH_SLT:
2802 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
2803 * to return 0.0 but step returns 1.0 because step is not < x
2804 * An alternative is a bvec compare padded with an unused second component.
2805 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
2806 * issue. Playing with not() is not possible either because not() does not accept
2809 shader_addline(ins->ctx->buffer, "(%s < %s) ? 1.0 : 0.0);\n",
2810 src0_param.param_str, src1_param.param_str);
2812 case WINED3DSIH_SGE:
2813 /* Here we can use the step() function and safe a conditional */
2814 shader_addline(ins->ctx->buffer, "step(%s, %s));\n", src1_param.param_str, src0_param.param_str);
2817 FIXME("Can't handle opcode %#x\n", ins->handler_idx);
2823 static void shader_glsl_conditional_move(const struct wined3d_shader_instruction *ins)
2825 const char *condition_prefix, *condition_suffix;
2826 struct wined3d_shader_dst_param dst;
2827 struct glsl_src_param src0_param;
2828 struct glsl_src_param src1_param;
2829 struct glsl_src_param src2_param;
2830 BOOL temp_destination = FALSE;
2831 DWORD cmp_channel = 0;
2836 switch (ins->handler_idx)
2838 case WINED3DSIH_CMP:
2839 condition_prefix = "";
2840 condition_suffix = " >= 0.0";
2843 case WINED3DSIH_CND:
2844 condition_prefix = "";
2845 condition_suffix = " > 0.5";
2848 case WINED3DSIH_MOVC:
2849 condition_prefix = "bool(";
2850 condition_suffix = ")";
2854 FIXME("Unhandled instruction %#x.\n", ins->handler_idx);
2855 condition_prefix = "<unhandled prefix>";
2856 condition_suffix = "<unhandled suffix>";
2860 if (shader_is_scalar(&ins->dst[0].reg) || shader_is_scalar(&ins->src[0].reg))
2862 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
2863 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2864 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2865 shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
2867 shader_addline(ins->ctx->buffer, "%s%s%s ? %s : %s);\n",
2868 condition_prefix, src0_param.param_str, condition_suffix,
2869 src1_param.param_str, src2_param.param_str);
2875 /* Splitting the instruction up in multiple lines imposes a problem:
2876 * The first lines may overwrite source parameters of the following lines.
2877 * Deal with that by using a temporary destination register if needed. */
2878 if ((ins->src[0].reg.idx[0].offset == dst.reg.idx[0].offset
2879 && ins->src[0].reg.type == dst.reg.type)
2880 || (ins->src[1].reg.idx[0].offset == dst.reg.idx[0].offset
2881 && ins->src[1].reg.type == dst.reg.type)
2882 || (ins->src[2].reg.idx[0].offset == dst.reg.idx[0].offset
2883 && ins->src[2].reg.type == dst.reg.type))
2884 temp_destination = TRUE;
2886 /* Cycle through all source0 channels. */
2887 for (i = 0; i < 4; ++i)
2890 /* Find the destination channels which use the current source0 channel. */
2891 for (j = 0; j < 4; ++j)
2893 if (((ins->src[0].swizzle >> (2 * j)) & 0x3) == i)
2895 write_mask |= WINED3DSP_WRITEMASK_0 << j;
2896 cmp_channel = WINED3DSP_WRITEMASK_0 << j;
2899 dst.write_mask = ins->dst[0].write_mask & write_mask;
2901 if (temp_destination)
2903 if (!(write_mask = shader_glsl_get_write_mask(&dst, mask_char)))
2905 shader_addline(ins->ctx->buffer, "tmp0%s = (", mask_char);
2907 else if (!(write_mask = shader_glsl_append_dst_ext(ins->ctx->buffer, ins, &dst)))
2910 shader_glsl_add_src_param(ins, &ins->src[0], cmp_channel, &src0_param);
2911 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2912 shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
2914 shader_addline(ins->ctx->buffer, "%s%s%s ? %s : %s);\n",
2915 condition_prefix, src0_param.param_str, condition_suffix,
2916 src1_param.param_str, src2_param.param_str);
2919 if (temp_destination)
2921 shader_glsl_get_write_mask(&ins->dst[0], mask_char);
2922 shader_glsl_append_dst(ins->ctx->buffer, ins);
2923 shader_addline(ins->ctx->buffer, "tmp0%s);\n", mask_char);
2927 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
2928 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
2929 * the compare is done per component of src0. */
2930 static void shader_glsl_cnd(const struct wined3d_shader_instruction *ins)
2932 struct glsl_src_param src0_param;
2933 struct glsl_src_param src1_param;
2934 struct glsl_src_param src2_param;
2936 DWORD shader_version = WINED3D_SHADER_VERSION(ins->ctx->reg_maps->shader_version.major,
2937 ins->ctx->reg_maps->shader_version.minor);
2939 if (shader_version < WINED3D_SHADER_VERSION(1, 4))
2941 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
2942 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
2943 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2944 shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
2946 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2949 shader_addline(ins->ctx->buffer, "%s /* COISSUE! */);\n", src1_param.param_str);
2951 shader_addline(ins->ctx->buffer, "%s > 0.5 ? %s : %s);\n",
2952 src0_param.param_str, src1_param.param_str, src2_param.param_str);
2957 shader_glsl_conditional_move(ins);
2960 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2961 static void shader_glsl_mad(const struct wined3d_shader_instruction *ins)
2963 struct glsl_src_param src0_param;
2964 struct glsl_src_param src1_param;
2965 struct glsl_src_param src2_param;
2968 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
2969 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
2970 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
2971 shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
2972 shader_addline(ins->ctx->buffer, "(%s * %s) + %s);\n",
2973 src0_param.param_str, src1_param.param_str, src2_param.param_str);
2976 /* Handles transforming all WINED3DSIO_M?x? opcodes for
2977 Vertex shaders to GLSL codes */
2978 static void shader_glsl_mnxn(const struct wined3d_shader_instruction *ins)
2981 int nComponents = 0;
2982 struct wined3d_shader_dst_param tmp_dst = {{0}};
2983 struct wined3d_shader_src_param tmp_src[2] = {{{0}}};
2984 struct wined3d_shader_instruction tmp_ins;
2986 memset(&tmp_ins, 0, sizeof(tmp_ins));
2988 /* Set constants for the temporary argument */
2989 tmp_ins.ctx = ins->ctx;
2990 tmp_ins.dst_count = 1;
2991 tmp_ins.dst = &tmp_dst;
2992 tmp_ins.src_count = 2;
2993 tmp_ins.src = tmp_src;
2995 switch(ins->handler_idx)
2997 case WINED3DSIH_M4x4:
2999 tmp_ins.handler_idx = WINED3DSIH_DP4;
3001 case WINED3DSIH_M4x3:
3003 tmp_ins.handler_idx = WINED3DSIH_DP4;
3005 case WINED3DSIH_M3x4:
3007 tmp_ins.handler_idx = WINED3DSIH_DP3;
3009 case WINED3DSIH_M3x3:
3011 tmp_ins.handler_idx = WINED3DSIH_DP3;
3013 case WINED3DSIH_M3x2:
3015 tmp_ins.handler_idx = WINED3DSIH_DP3;
3021 tmp_dst = ins->dst[0];
3022 tmp_src[0] = ins->src[0];
3023 tmp_src[1] = ins->src[1];
3024 for (i = 0; i < nComponents; ++i)
3026 tmp_dst.write_mask = WINED3DSP_WRITEMASK_0 << i;
3027 shader_glsl_dot(&tmp_ins);
3028 ++tmp_src[1].reg.idx[0].offset;
3033 The LRP instruction performs a component-wise linear interpolation
3034 between the second and third operands using the first operand as the
3035 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
3036 This is equivalent to mix(src2, src1, src0);
3038 static void shader_glsl_lrp(const struct wined3d_shader_instruction *ins)
3040 struct glsl_src_param src0_param;
3041 struct glsl_src_param src1_param;
3042 struct glsl_src_param src2_param;
3045 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
3047 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
3048 shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
3049 shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
3051 shader_addline(ins->ctx->buffer, "mix(%s, %s, %s));\n",
3052 src2_param.param_str, src1_param.param_str, src0_param.param_str);
3055 /** Process the WINED3DSIO_LIT instruction in GLSL:
3056 * dst.x = dst.w = 1.0
3057 * dst.y = (src0.x > 0) ? src0.x
3058 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
3059 * where src.w is clamped at +- 128
3061 static void shader_glsl_lit(const struct wined3d_shader_instruction *ins)
3063 struct glsl_src_param src0_param;
3064 struct glsl_src_param src1_param;
3065 struct glsl_src_param src3_param;
3068 shader_glsl_append_dst(ins->ctx->buffer, ins);
3069 shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
3071 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
3072 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_1, &src1_param);
3073 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &src3_param);
3075 /* The sdk specifies the instruction like this
3077 * if(src.x > 0.0) dst.y = src.x
3079 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
3082 * (where power = src.w clamped between -128 and 128)
3084 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
3085 * dst.x = 1.0 ... No further explanation needed
3086 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
3087 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
3088 * dst.w = 1.0. ... Nothing fancy.
3090 * So we still have one conditional in there. So do this:
3091 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
3093 * 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),
3094 * 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.
3095 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to.
3097 * Unfortunately pow(0.0 ^ 0.0) returns NaN on most GPUs, but lit with src.y = 0 and src.w = 0 returns
3098 * a non-NaN value in dst.z. What we return doesn't matter, as long as it is not NaN. Return 0, which is
3099 * what all Windows HW drivers and GL_ARB_vertex_program's LIT do.
3101 shader_addline(ins->ctx->buffer,
3102 "vec4(1.0, max(%s, 0.0), %s == 0.0 ? 0.0 : "
3103 "pow(max(0.0, %s) * step(0.0, %s), clamp(%s, -128.0, 128.0)), 1.0)%s);\n",
3104 src0_param.param_str, src3_param.param_str, src1_param.param_str,
3105 src0_param.param_str, src3_param.param_str, dst_mask);
3108 /** Process the WINED3DSIO_DST instruction in GLSL:
3110 * dst.y = src0.x * src0.y
3114 static void shader_glsl_dst(const struct wined3d_shader_instruction *ins)
3116 struct glsl_src_param src0y_param;
3117 struct glsl_src_param src0z_param;
3118 struct glsl_src_param src1y_param;
3119 struct glsl_src_param src1w_param;
3122 shader_glsl_append_dst(ins->ctx->buffer, ins);
3123 shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
3125 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_1, &src0y_param);
3126 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_2, &src0z_param);
3127 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_1, &src1y_param);
3128 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_3, &src1w_param);
3130 shader_addline(ins->ctx->buffer, "vec4(1.0, %s * %s, %s, %s))%s;\n",
3131 src0y_param.param_str, src1y_param.param_str, src0z_param.param_str, src1w_param.param_str, dst_mask);
3134 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
3135 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
3136 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
3138 * dst.x = cos(src0.?)
3139 * dst.y = sin(src0.?)
3143 static void shader_glsl_sincos(const struct wined3d_shader_instruction *ins)
3145 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3146 struct glsl_src_param src0_param;
3149 if (ins->ctx->reg_maps->shader_version.major < 4)
3151 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
3153 write_mask = shader_glsl_append_dst(buffer, ins);
3156 case WINED3DSP_WRITEMASK_0:
3157 shader_addline(buffer, "cos(%s));\n", src0_param.param_str);
3160 case WINED3DSP_WRITEMASK_1:
3161 shader_addline(buffer, "sin(%s));\n", src0_param.param_str);
3164 case (WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1):
3165 shader_addline(buffer, "vec2(cos(%s), sin(%s)));\n",
3166 src0_param.param_str, src0_param.param_str);
3170 ERR("Write mask should be .x, .y or .xy\n");
3177 if (ins->dst[0].reg.type != WINED3DSPR_NULL)
3180 if (ins->dst[1].reg.type != WINED3DSPR_NULL)
3184 write_mask = shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
3185 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
3186 shader_addline(buffer, "tmp0%s = sin(%s);\n", dst_mask, src0_param.param_str);
3188 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[1]);
3189 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
3190 shader_addline(buffer, "cos(%s));\n", src0_param.param_str);
3192 shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0]);
3193 shader_addline(buffer, "tmp0%s);\n", dst_mask);
3197 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0]);
3198 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
3199 shader_addline(buffer, "sin(%s));\n", src0_param.param_str);
3202 else if (ins->dst[1].reg.type != WINED3DSPR_NULL)
3204 write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[1]);
3205 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
3206 shader_addline(buffer, "cos(%s));\n", src0_param.param_str);
3210 /* sgn in vs_2_0 has 2 extra parameters(registers for temporary storage) which we don't use
3211 * here. But those extra parameters require a dedicated function for sgn, since map2gl would
3212 * generate invalid code
3214 static void shader_glsl_sgn(const struct wined3d_shader_instruction *ins)
3216 struct glsl_src_param src0_param;
3219 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
3220 shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
3222 shader_addline(ins->ctx->buffer, "sign(%s));\n", src0_param.param_str);
3225 /** Process the WINED3DSIO_LOOP instruction in GLSL:
3226 * Start a for() loop where src1.y is the initial value of aL,
3227 * increment aL by src1.z for a total of src1.x iterations.
3228 * Need to use a temporary variable for this operation.
3230 /* FIXME: I don't think nested loops will work correctly this way. */
3231 static void shader_glsl_loop(const struct wined3d_shader_instruction *ins)
3233 struct wined3d_shader_loop_state *loop_state = ins->ctx->loop_state;
3234 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3235 const struct wined3d_shader *shader = ins->ctx->shader;
3236 const struct wined3d_shader_lconst *constant;
3237 struct glsl_src_param src1_param;
3238 const DWORD *control_values = NULL;
3240 if (ins->ctx->reg_maps->shader_version.major < 4)
3242 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_ALL, &src1_param);
3244 /* Try to hardcode the loop control parameters if possible. Direct3D 9
3245 * class hardware doesn't support real varying indexing, but Microsoft
3246 * designed this feature for Shader model 2.x+. If the loop control is
3247 * known at compile time, the GLSL compiler can unroll the loop, and
3248 * replace indirect addressing with direct addressing. */
3249 if (ins->src[1].reg.type == WINED3DSPR_CONSTINT)
3251 LIST_FOR_EACH_ENTRY(constant, &shader->constantsI, struct wined3d_shader_lconst, entry)
3253 if (constant->idx == ins->src[1].reg.idx[0].offset)
3255 control_values = constant->value;
3263 struct wined3d_shader_loop_control loop_control;
3264 loop_control.count = control_values[0];
3265 loop_control.start = control_values[1];
3266 loop_control.step = (int)control_values[2];
3268 if (loop_control.step > 0)
3270 shader_addline(buffer, "for (aL%u = %u; aL%u < (%u * %d + %u); aL%u += %d)\n{\n",
3271 loop_state->current_depth, loop_control.start,
3272 loop_state->current_depth, loop_control.count, loop_control.step, loop_control.start,
3273 loop_state->current_depth, loop_control.step);
3275 else if (loop_control.step < 0)
3277 shader_addline(buffer, "for (aL%u = %u; aL%u > (%u * %d + %u); aL%u += %d)\n{\n",
3278 loop_state->current_depth, loop_control.start,
3279 loop_state->current_depth, loop_control.count, loop_control.step, loop_control.start,
3280 loop_state->current_depth, loop_control.step);
3284 shader_addline(buffer, "for (aL%u = %u, tmpInt%u = 0; tmpInt%u < %u; tmpInt%u++)\n{\n",
3285 loop_state->current_depth, loop_control.start, loop_state->current_depth,
3286 loop_state->current_depth, loop_control.count,
3287 loop_state->current_depth);
3292 shader_addline(buffer, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z)\n{\n",
3293 loop_state->current_depth, loop_state->current_reg,
3294 src1_param.reg_name, loop_state->current_depth, src1_param.reg_name,
3295 loop_state->current_depth, loop_state->current_reg, src1_param.reg_name);
3298 ++loop_state->current_reg;
3302 shader_addline(buffer, "for (;;)\n{\n");
3305 ++loop_state->current_depth;
3308 static void shader_glsl_end(const struct wined3d_shader_instruction *ins)
3310 struct wined3d_shader_loop_state *loop_state = ins->ctx->loop_state;
3312 shader_addline(ins->ctx->buffer, "}\n");
3314 if (ins->handler_idx == WINED3DSIH_ENDLOOP)
3316 --loop_state->current_depth;
3317 --loop_state->current_reg;
3320 if (ins->handler_idx == WINED3DSIH_ENDREP)
3322 --loop_state->current_depth;
3326 static void shader_glsl_rep(const struct wined3d_shader_instruction *ins)
3328 const struct wined3d_shader *shader = ins->ctx->shader;
3329 struct wined3d_shader_loop_state *loop_state = ins->ctx->loop_state;
3330 const struct wined3d_shader_lconst *constant;
3331 struct glsl_src_param src0_param;
3332 const DWORD *control_values = NULL;
3334 /* Try to hardcode local values to help the GLSL compiler to unroll and optimize the loop */
3335 if (ins->src[0].reg.type == WINED3DSPR_CONSTINT)
3337 LIST_FOR_EACH_ENTRY(constant, &shader->constantsI, struct wined3d_shader_lconst, entry)
3339 if (constant->idx == ins->src[0].reg.idx[0].offset)
3341 control_values = constant->value;
3349 shader_addline(ins->ctx->buffer, "for (tmpInt%d = 0; tmpInt%d < %d; tmpInt%d++) {\n",
3350 loop_state->current_depth, loop_state->current_depth,
3351 control_values[0], loop_state->current_depth);
3355 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
3356 shader_addline(ins->ctx->buffer, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
3357 loop_state->current_depth, loop_state->current_depth,
3358 src0_param.param_str, loop_state->current_depth);
3361 ++loop_state->current_depth;
3364 static void shader_glsl_if(const struct wined3d_shader_instruction *ins)
3366 struct glsl_src_param src0_param;
3368 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
3369 shader_addline(ins->ctx->buffer, "if (%s) {\n", src0_param.param_str);
3372 static void shader_glsl_ifc(const struct wined3d_shader_instruction *ins)
3374 struct glsl_src_param src0_param;
3375 struct glsl_src_param src1_param;
3377 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
3378 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0, &src1_param);
3380 shader_addline(ins->ctx->buffer, "if (%s %s %s) {\n",
3381 src0_param.param_str, shader_glsl_get_rel_op(ins->flags), src1_param.param_str);
3384 static void shader_glsl_else(const struct wined3d_shader_instruction *ins)
3386 shader_addline(ins->ctx->buffer, "} else {\n");
3389 static void shader_glsl_emit(const struct wined3d_shader_instruction *ins)
3391 shader_addline(ins->ctx->buffer, "EmitVertex();\n");
3394 static void shader_glsl_break(const struct wined3d_shader_instruction *ins)
3396 shader_addline(ins->ctx->buffer, "break;\n");
3399 /* FIXME: According to MSDN the compare is done per component. */
3400 static void shader_glsl_breakc(const struct wined3d_shader_instruction *ins)
3402 struct glsl_src_param src0_param;
3403 struct glsl_src_param src1_param;
3405 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
3406 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0, &src1_param);
3408 shader_addline(ins->ctx->buffer, "if (%s %s %s) break;\n",
3409 src0_param.param_str, shader_glsl_get_rel_op(ins->flags), src1_param.param_str);
3412 static void shader_glsl_breakp(const struct wined3d_shader_instruction *ins)
3414 struct glsl_src_param src_param;
3416 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src_param);
3417 shader_addline(ins->ctx->buffer, "if (bool(%s)) break;\n", src_param.param_str);
3420 static void shader_glsl_label(const struct wined3d_shader_instruction *ins)
3422 shader_addline(ins->ctx->buffer, "}\n");
3423 shader_addline(ins->ctx->buffer, "void subroutine%u()\n{\n", ins->src[0].reg.idx[0].offset);
3426 static void shader_glsl_call(const struct wined3d_shader_instruction *ins)
3428 shader_addline(ins->ctx->buffer, "subroutine%u();\n", ins->src[0].reg.idx[0].offset);
3431 static void shader_glsl_callnz(const struct wined3d_shader_instruction *ins)
3433 struct glsl_src_param src1_param;
3435 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0, &src1_param);
3436 shader_addline(ins->ctx->buffer, "if (%s) subroutine%u();\n",
3437 src1_param.param_str, ins->src[0].reg.idx[0].offset);
3440 static void shader_glsl_ret(const struct wined3d_shader_instruction *ins)
3442 /* No-op. The closing } is written when a new function is started, and at the end of the shader. This
3443 * function only suppresses the unhandled instruction warning
3447 /*********************************************
3448 * Pixel Shader Specific Code begins here
3449 ********************************************/
3450 static void shader_glsl_tex(const struct wined3d_shader_instruction *ins)
3452 DWORD shader_version = WINED3D_SHADER_VERSION(ins->ctx->reg_maps->shader_version.major,
3453 ins->ctx->reg_maps->shader_version.minor);
3454 struct glsl_sample_function sample_function;
3455 DWORD sample_flags = 0;
3457 DWORD mask = 0, swizzle;
3458 const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
3460 /* 1.0-1.4: Use destination register as sampler source.
3461 * 2.0+: Use provided sampler source. */
3462 if (shader_version < WINED3D_SHADER_VERSION(2,0))
3463 sampler_idx = ins->dst[0].reg.idx[0].offset;
3465 sampler_idx = ins->src[1].reg.idx[0].offset;
3467 if (shader_version < WINED3D_SHADER_VERSION(1,4))
3469 DWORD flags = (priv->cur_ps_args->tex_transform >> sampler_idx * WINED3D_PSARGS_TEXTRANSFORM_SHIFT)
3470 & WINED3D_PSARGS_TEXTRANSFORM_MASK;
3471 enum wined3d_sampler_texture_type sampler_type = ins->ctx->reg_maps->sampler_type[sampler_idx];
3473 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
3474 if (flags & WINED3D_PSARGS_PROJECTED && sampler_type != WINED3DSTT_CUBE)
3476 sample_flags |= WINED3D_GLSL_SAMPLE_PROJECTED;
3477 switch (flags & ~WINED3D_PSARGS_PROJECTED)
3479 case WINED3D_TTFF_COUNT1:
3480 FIXME("WINED3D_TTFF_PROJECTED with WINED3D_TTFF_COUNT1?\n");
3482 case WINED3D_TTFF_COUNT2:
3483 mask = WINED3DSP_WRITEMASK_1;
3485 case WINED3D_TTFF_COUNT3:
3486 mask = WINED3DSP_WRITEMASK_2;
3488 case WINED3D_TTFF_COUNT4:
3489 case WINED3D_TTFF_DISABLE:
3490 mask = WINED3DSP_WRITEMASK_3;
3495 else if (shader_version < WINED3D_SHADER_VERSION(2,0))
3497 enum wined3d_shader_src_modifier src_mod = ins->src[0].modifiers;
3499 if (src_mod == WINED3DSPSM_DZ) {
3500 sample_flags |= WINED3D_GLSL_SAMPLE_PROJECTED;
3501 mask = WINED3DSP_WRITEMASK_2;
3502 } else if (src_mod == WINED3DSPSM_DW) {
3503 sample_flags |= WINED3D_GLSL_SAMPLE_PROJECTED;
3504 mask = WINED3DSP_WRITEMASK_3;
3507 if (ins->flags & WINED3DSI_TEXLD_PROJECT)
3509 /* ps 2.0 texldp instruction always divides by the fourth component. */
3510 sample_flags |= WINED3D_GLSL_SAMPLE_PROJECTED;
3511 mask = WINED3DSP_WRITEMASK_3;
3515 if (priv->cur_ps_args->np2_fixup & (1 << sampler_idx))
3516 sample_flags |= WINED3D_GLSL_SAMPLE_NPOT;
3518 shader_glsl_get_sample_function(ins->ctx, sampler_idx, sample_flags, &sample_function);
3519 mask |= sample_function.coord_mask;
3521 if (shader_version < WINED3D_SHADER_VERSION(2,0)) swizzle = WINED3DSP_NOSWIZZLE;
3522 else swizzle = ins->src[1].swizzle;
3524 /* 1.0-1.3: Use destination register as coordinate source.
3525 1.4+: Use provided coordinate source register. */
3526 if (shader_version < WINED3D_SHADER_VERSION(1,4))
3529 shader_glsl_write_mask_to_str(mask, coord_mask);
3530 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, NULL, NULL, NULL,
3531 "T%u%s", sampler_idx, coord_mask);
3535 struct glsl_src_param coord_param;
3536 shader_glsl_add_src_param(ins, &ins->src[0], mask, &coord_param);
3537 if (ins->flags & WINED3DSI_TEXLD_BIAS)
3539 struct glsl_src_param bias;
3540 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &bias);
3541 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, NULL, NULL, bias.param_str,
3542 "%s", coord_param.param_str);
3544 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, NULL, NULL, NULL,
3545 "%s", coord_param.param_str);
3550 static void shader_glsl_texldd(const struct wined3d_shader_instruction *ins)
3552 const struct wined3d_gl_info *gl_info = ins->ctx->gl_info;
3553 struct glsl_src_param coord_param, dx_param, dy_param;
3554 DWORD sample_flags = WINED3D_GLSL_SAMPLE_GRAD;
3555 struct glsl_sample_function sample_function;
3557 DWORD swizzle = ins->src[1].swizzle;
3558 const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
3560 if (!gl_info->supported[ARB_SHADER_TEXTURE_LOD] && !gl_info->supported[EXT_GPU_SHADER4])
3562 FIXME("texldd used, but not supported by hardware. Falling back to regular tex\n");
3563 shader_glsl_tex(ins);
3567 sampler_idx = ins->src[1].reg.idx[0].offset;
3568 if (priv->cur_ps_args->np2_fixup & (1 << sampler_idx))
3569 sample_flags |= WINED3D_GLSL_SAMPLE_NPOT;
3571 shader_glsl_get_sample_function(ins->ctx, sampler_idx, sample_flags, &sample_function);
3572 shader_glsl_add_src_param(ins, &ins->src[0], sample_function.coord_mask, &coord_param);
3573 shader_glsl_add_src_param(ins, &ins->src[2], sample_function.coord_mask, &dx_param);
3574 shader_glsl_add_src_param(ins, &ins->src[3], sample_function.coord_mask, &dy_param);
3576 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, dx_param.param_str, dy_param.param_str, NULL,
3577 "%s", coord_param.param_str);
3580 static void shader_glsl_texldl(const struct wined3d_shader_instruction *ins)
3582 const struct wined3d_gl_info *gl_info = ins->ctx->gl_info;
3583 struct glsl_src_param coord_param, lod_param;
3584 DWORD sample_flags = WINED3D_GLSL_SAMPLE_LOD;
3585 struct glsl_sample_function sample_function;
3587 DWORD swizzle = ins->src[1].swizzle;
3588 const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
3590 sampler_idx = ins->src[1].reg.idx[0].offset;
3591 if (ins->ctx->reg_maps->shader_version.type == WINED3D_SHADER_TYPE_PIXEL
3592 && priv->cur_ps_args->np2_fixup & (1 << sampler_idx))
3593 sample_flags |= WINED3D_GLSL_SAMPLE_NPOT;
3595 shader_glsl_get_sample_function(ins->ctx, sampler_idx, sample_flags, &sample_function);
3596 shader_glsl_add_src_param(ins, &ins->src[0], sample_function.coord_mask, &coord_param);
3598 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &lod_param);
3600 if (!gl_info->supported[ARB_SHADER_TEXTURE_LOD] && !gl_info->supported[EXT_GPU_SHADER4]
3601 && ins->ctx->reg_maps->shader_version.type == WINED3D_SHADER_TYPE_PIXEL)
3603 /* Plain GLSL only supports Lod sampling functions in vertex shaders.
3604 * However, the NVIDIA drivers allow them in fragment shaders as well,
3605 * even without the appropriate extension. */
3606 WARN("Using %s in fragment shader.\n", sample_function.name);
3608 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, NULL, NULL, lod_param.param_str,
3609 "%s", coord_param.param_str);
3612 static void shader_glsl_texcoord(const struct wined3d_shader_instruction *ins)
3614 /* FIXME: Make this work for more than just 2D textures */
3615 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3616 DWORD write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
3618 if (!(ins->ctx->reg_maps->shader_version.major == 1 && ins->ctx->reg_maps->shader_version.minor == 4))
3622 shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
3623 shader_addline(buffer, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
3624 ins->dst[0].reg.idx[0].offset, dst_mask);
3628 enum wined3d_shader_src_modifier src_mod = ins->src[0].modifiers;
3629 DWORD reg = ins->src[0].reg.idx[0].offset;
3630 char dst_swizzle[6];
3632 shader_glsl_get_swizzle(&ins->src[0], FALSE, write_mask, dst_swizzle);
3634 if (src_mod == WINED3DSPSM_DZ)
3636 unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
3637 struct glsl_src_param div_param;
3639 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_2, &div_param);
3641 if (mask_size > 1) {
3642 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
3644 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
3647 else if (src_mod == WINED3DSPSM_DW)
3649 unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
3650 struct glsl_src_param div_param;
3652 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &div_param);
3654 if (mask_size > 1) {
3655 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
3657 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
3660 shader_addline(buffer, "gl_TexCoord[%u]%s);\n", reg, dst_swizzle);
3665 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
3666 * Take a 3-component dot product of the TexCoord[dstreg] and src,
3667 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
3668 static void shader_glsl_texdp3tex(const struct wined3d_shader_instruction *ins)
3670 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3671 DWORD sampler_idx = ins->dst[0].reg.idx[0].offset;
3672 struct glsl_sample_function sample_function;
3673 struct glsl_src_param src0_param;
3676 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3678 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
3679 * scalar, and projected sampling would require 4.
3681 * It is a dependent read - not valid with conditional NP2 textures
3683 shader_glsl_get_sample_function(ins->ctx, sampler_idx, 0, &sample_function);
3684 mask_size = shader_glsl_get_write_mask_size(sample_function.coord_mask);
3689 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
3690 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx, src0_param.param_str);
3694 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
3695 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx, src0_param.param_str);
3699 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
3700 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx, src0_param.param_str);
3704 FIXME("Unexpected mask size %u\n", mask_size);
3709 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
3710 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
3711 static void shader_glsl_texdp3(const struct wined3d_shader_instruction *ins)
3713 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3714 DWORD dstreg = ins->dst[0].reg.idx[0].offset;
3715 struct glsl_src_param src0_param;
3717 unsigned int mask_size;
3719 dst_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
3720 mask_size = shader_glsl_get_write_mask_size(dst_mask);
3721 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3723 if (mask_size > 1) {
3724 shader_addline(ins->ctx->buffer, "vec%d(dot(T%u.xyz, %s)));\n", mask_size, dstreg, src0_param.param_str);
3726 shader_addline(ins->ctx->buffer, "dot(T%u.xyz, %s));\n", dstreg, src0_param.param_str);
3730 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
3731 * Calculate the depth as dst.x / dst.y */
3732 static void shader_glsl_texdepth(const struct wined3d_shader_instruction *ins)
3734 struct glsl_dst_param dst_param;
3736 shader_glsl_add_dst_param(ins, &ins->dst[0], &dst_param);
3738 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
3739 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
3740 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
3741 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
3744 shader_addline(ins->ctx->buffer, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
3745 dst_param.reg_name, dst_param.reg_name);
3748 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
3749 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
3750 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
3751 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
3753 static void shader_glsl_texm3x2depth(const struct wined3d_shader_instruction *ins)
3755 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3756 DWORD dstreg = ins->dst[0].reg.idx[0].offset;
3757 struct glsl_src_param src0_param;
3759 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3761 shader_addline(ins->ctx->buffer, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg, src0_param.param_str);
3762 shader_addline(ins->ctx->buffer, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
3765 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
3766 * Calculate the 1st of a 2-row matrix multiplication. */
3767 static void shader_glsl_texm3x2pad(const struct wined3d_shader_instruction *ins)
3769 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3770 DWORD reg = ins->dst[0].reg.idx[0].offset;
3771 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3772 struct glsl_src_param src0_param;
3774 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3775 shader_addline(buffer, "tmp0.x = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
3778 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
3779 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
3780 static void shader_glsl_texm3x3pad(const struct wined3d_shader_instruction *ins)
3782 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3783 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3784 struct wined3d_shader_tex_mx *tex_mx = ins->ctx->tex_mx;
3785 DWORD reg = ins->dst[0].reg.idx[0].offset;
3786 struct glsl_src_param src0_param;
3788 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3789 shader_addline(buffer, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + tex_mx->current_row, reg, src0_param.param_str);
3790 tex_mx->texcoord_w[tex_mx->current_row++] = reg;
3793 static void shader_glsl_texm3x2tex(const struct wined3d_shader_instruction *ins)
3795 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3796 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3797 struct glsl_sample_function sample_function;
3798 DWORD reg = ins->dst[0].reg.idx[0].offset;
3799 struct glsl_src_param src0_param;
3801 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3802 shader_addline(buffer, "tmp0.y = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
3804 shader_glsl_get_sample_function(ins->ctx, reg, 0, &sample_function);
3806 /* Sample the texture using the calculated coordinates */
3807 shader_glsl_gen_sample_code(ins, reg, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL, "tmp0.xy");
3810 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
3811 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
3812 static void shader_glsl_texm3x3tex(const struct wined3d_shader_instruction *ins)
3814 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3815 struct wined3d_shader_tex_mx *tex_mx = ins->ctx->tex_mx;
3816 struct glsl_sample_function sample_function;
3817 DWORD reg = ins->dst[0].reg.idx[0].offset;
3818 struct glsl_src_param src0_param;
3820 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3821 shader_addline(ins->ctx->buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
3823 /* Dependent read, not valid with conditional NP2 */
3824 shader_glsl_get_sample_function(ins->ctx, reg, 0, &sample_function);
3826 /* Sample the texture using the calculated coordinates */
3827 shader_glsl_gen_sample_code(ins, reg, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL, "tmp0.xyz");
3829 tex_mx->current_row = 0;
3832 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
3833 * Perform the 3rd row of a 3x3 matrix multiply */
3834 static void shader_glsl_texm3x3(const struct wined3d_shader_instruction *ins)
3836 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3837 struct wined3d_shader_tex_mx *tex_mx = ins->ctx->tex_mx;
3838 DWORD reg = ins->dst[0].reg.idx[0].offset;
3839 struct glsl_src_param src0_param;
3842 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3844 shader_glsl_append_dst(ins->ctx->buffer, ins);
3845 shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
3846 shader_addline(ins->ctx->buffer, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg, src0_param.param_str, dst_mask);
3848 tex_mx->current_row = 0;
3851 /* Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
3852 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3853 static void shader_glsl_texm3x3spec(const struct wined3d_shader_instruction *ins)
3855 struct glsl_src_param src0_param;
3856 struct glsl_src_param src1_param;
3857 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3858 struct wined3d_shader_tex_mx *tex_mx = ins->ctx->tex_mx;
3859 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3860 struct glsl_sample_function sample_function;
3861 DWORD reg = ins->dst[0].reg.idx[0].offset;
3864 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3865 shader_glsl_add_src_param(ins, &ins->src[1], src_mask, &src1_param);
3867 /* Perform the last matrix multiply operation */
3868 shader_addline(buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
3869 /* Reflection calculation */
3870 shader_addline(buffer, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param.param_str);
3872 /* Dependent read, not valid with conditional NP2 */
3873 shader_glsl_get_sample_function(ins->ctx, reg, 0, &sample_function);
3874 shader_glsl_write_mask_to_str(sample_function.coord_mask, coord_mask);
3876 /* Sample the texture */
3877 shader_glsl_gen_sample_code(ins, reg, &sample_function, WINED3DSP_NOSWIZZLE,
3878 NULL, NULL, NULL, "tmp0%s", coord_mask);
3880 tex_mx->current_row = 0;
3883 /* Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
3884 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3885 static void shader_glsl_texm3x3vspec(const struct wined3d_shader_instruction *ins)
3887 struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
3888 struct wined3d_shader_tex_mx *tex_mx = ins->ctx->tex_mx;
3889 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
3890 struct glsl_sample_function sample_function;
3891 DWORD reg = ins->dst[0].reg.idx[0].offset;
3892 struct glsl_src_param src0_param;
3895 shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
3897 /* Perform the last matrix multiply operation */
3898 shader_addline(buffer, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg, src0_param.param_str);
3900 /* Construct the eye-ray vector from w coordinates */
3901 shader_addline(buffer, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
3902 tex_mx->texcoord_w[0], tex_mx->texcoord_w[1], reg);
3903 shader_addline(buffer, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
3905 /* Dependent read, not valid with conditional NP2 */
3906 shader_glsl_get_sample_function(ins->ctx, reg, 0, &sample_function);
3907 shader_glsl_write_mask_to_str(sample_function.coord_mask, coord_mask);
3909 /* Sample the texture using the calculated coordinates */
3910 shader_glsl_gen_sample_code(ins, reg, &sample_function, WINED3DSP_NOSWIZZLE,
3911 NULL, NULL, NULL, "tmp0%s", coord_mask);
3913 tex_mx->current_row = 0;
3916 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
3917 * Apply a fake bump map transform.
3918 * texbem is pshader <= 1.3 only, this saves a few version checks
3920 static void shader_glsl_texbem(const struct wined3d_shader_instruction *ins)
3922 const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
3923 struct glsl_sample_function sample_function;
3924 struct glsl_src_param coord_param;
3930 sampler_idx = ins->dst[0].reg.idx[0].offset;
3931 flags = (priv->cur_ps_args->tex_transform >> sampler_idx * WINED3D_PSARGS_TEXTRANSFORM_SHIFT)
3932 & WINED3D_PSARGS_TEXTRANSFORM_MASK;
3934 /* Dependent read, not valid with conditional NP2 */
3935 shader_glsl_get_sample_function(ins->ctx, sampler_idx, 0, &sample_function);
3936 mask = sample_function.coord_mask;
3938 shader_glsl_write_mask_to_str(mask, coord_mask);
3940 /* With projected textures, texbem only divides the static texture coord,
3941 * not the displacement, so we can't let GL handle this. */
3942 if (flags & WINED3D_PSARGS_PROJECTED)
3945 char coord_div_mask[3];
3946 switch (flags & ~WINED3D_PSARGS_PROJECTED)
3948 case WINED3D_TTFF_COUNT1:
3949 FIXME("WINED3D_TTFF_PROJECTED with WINED3D_TTFF_COUNT1?\n");
3951 case WINED3D_TTFF_COUNT2:
3952 div_mask = WINED3DSP_WRITEMASK_1;
3954 case WINED3D_TTFF_COUNT3:
3955 div_mask = WINED3DSP_WRITEMASK_2;
3957 case WINED3D_TTFF_COUNT4:
3958 case WINED3D_TTFF_DISABLE:
3959 div_mask = WINED3DSP_WRITEMASK_3;
3962 shader_glsl_write_mask_to_str(div_mask, coord_div_mask);
3963 shader_addline(ins->ctx->buffer, "T%u%s /= T%u%s;\n", sampler_idx, coord_mask, sampler_idx, coord_div_mask);
3966 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &coord_param);
3968 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
3969 "T%u%s + vec4(bumpenv_mat%u * %s, 0.0, 0.0)%s", sampler_idx, coord_mask, sampler_idx,
3970 coord_param.param_str, coord_mask);
3972 if (ins->handler_idx == WINED3DSIH_TEXBEML)
3974 struct glsl_src_param luminance_param;
3975 struct glsl_dst_param dst_param;
3977 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_2, &luminance_param);
3978 shader_glsl_add_dst_param(ins, &ins->dst[0], &dst_param);
3980 shader_addline(ins->ctx->buffer, "%s%s *= (%s * bumpenv_lum_scale%u + bumpenv_lum_offset%u);\n",
3981 dst_param.reg_name, dst_param.mask_str,
3982 luminance_param.param_str, sampler_idx, sampler_idx);
3986 static void shader_glsl_bem(const struct wined3d_shader_instruction *ins)
3988 DWORD sampler_idx = ins->dst[0].reg.idx[0].offset;
3989 struct glsl_src_param src0_param, src1_param;
3991 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src0_param);
3992 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src1_param);
3994 shader_glsl_append_dst(ins->ctx->buffer, ins);
3995 shader_addline(ins->ctx->buffer, "%s + bumpenv_mat%u * %s);\n",
3996 src0_param.param_str, sampler_idx, src1_param.param_str);
3999 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
4000 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
4001 static void shader_glsl_texreg2ar(const struct wined3d_shader_instruction *ins)
4003 DWORD sampler_idx = ins->dst[0].reg.idx[0].offset;
4004 struct glsl_sample_function sample_function;
4005 struct glsl_src_param src0_param;
4007 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
4009 shader_glsl_get_sample_function(ins->ctx, sampler_idx, 0, &sample_function);
4010 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
4011 "%s.wx", src0_param.reg_name);
4014 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
4015 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
4016 static void shader_glsl_texreg2gb(const struct wined3d_shader_instruction *ins)
4018 DWORD sampler_idx = ins->dst[0].reg.idx[0].offset;
4019 struct glsl_sample_function sample_function;
4020 struct glsl_src_param src0_param;
4022 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
4024 shader_glsl_get_sample_function(ins->ctx, sampler_idx, 0, &sample_function);
4025 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
4026 "%s.yz", src0_param.reg_name);
4029 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
4030 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
4031 static void shader_glsl_texreg2rgb(const struct wined3d_shader_instruction *ins)
4033 DWORD sampler_idx = ins->dst[0].reg.idx[0].offset;
4034 struct glsl_sample_function sample_function;
4035 struct glsl_src_param src0_param;
4037 /* Dependent read, not valid with conditional NP2 */
4038 shader_glsl_get_sample_function(ins->ctx, sampler_idx, 0, &sample_function);
4039 shader_glsl_add_src_param(ins, &ins->src[0], sample_function.coord_mask, &src0_param);
4041 shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
4042 "%s", src0_param.param_str);
4045 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
4046 * If any of the first 3 components are < 0, discard this pixel */
4047 static void shader_glsl_texkill(const struct wined3d_shader_instruction *ins)
4049 struct glsl_dst_param dst_param;
4051 /* The argument is a destination parameter, and no writemasks are allowed */
4052 shader_glsl_add_dst_param(ins, &ins->dst[0], &dst_param);
4053 if (ins->ctx->reg_maps->shader_version.major >= 2)
4055 /* 2.0 shaders compare all 4 components in texkill */
4056 shader_addline(ins->ctx->buffer, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param.reg_name);
4058 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
4059 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
4060 * 4 components are defined, only the first 3 are used
4062 shader_addline(ins->ctx->buffer, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param.reg_name);
4066 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
4067 * dst = dot2(src0, src1) + src2 */
4068 static void shader_glsl_dp2add(const struct wined3d_shader_instruction *ins)
4070 struct glsl_src_param src0_param;
4071 struct glsl_src_param src1_param;
4072 struct glsl_src_param src2_param;
4074 unsigned int mask_size;
4076 write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
4077 mask_size = shader_glsl_get_write_mask_size(write_mask);
4079 shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src0_param);
4080 shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src1_param);
4081 shader_glsl_add_src_param(ins, &ins->src[2], WINED3DSP_WRITEMASK_0, &src2_param);
4083 if (mask_size > 1) {
4084 shader_addline(ins->ctx->buffer, "vec%d(dot(%s, %s) + %s));\n",
4085 mask_size, src0_param.param_str, src1_param.param_str, src2_param.param_str);
4087 shader_addline(ins->ctx->buffer, "dot(%s, %s) + %s);\n",
4088 src0_param.param_str, src1_param.param_str, src2_param.param_str);
4092 static void shader_glsl_input_pack(const struct wined3d_shader *shader, struct wined3d_shader_buffer *buffer,
4093 const struct wined3d_shader_signature_element *input_signature,
4094 const struct wined3d_shader_reg_maps *reg_maps,
4095 enum vertexprocessing_mode vertexprocessing)
4097 WORD map = reg_maps->input_registers;
4100 for (i = 0; map; map >>= 1, ++i)
4102 const char *semantic_name;
4107 if (!(map & 1)) continue;
4109 semantic_name = input_signature[i].semantic_name;
4110 semantic_idx = input_signature[i].semantic_idx;
4111 shader_glsl_write_mask_to_str(input_signature[i].mask, reg_mask);
4113 if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_TEXCOORD))
4115 if (semantic_idx < 8 && vertexprocessing == pretransformed)
4116 shader_addline(buffer, "ps_in[%u]%s = gl_TexCoord[%u]%s;\n",
4117 shader->u.ps.input_reg_map[i], reg_mask, semantic_idx, reg_mask);
4119 shader_addline(buffer, "ps_in[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
4120 shader->u.ps.input_reg_map[i], reg_mask, reg_mask);
4122 else if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_COLOR))
4125 shader_addline(buffer, "ps_in[%u]%s = vec4(gl_Color)%s;\n",
4126 shader->u.ps.input_reg_map[i], reg_mask, reg_mask);
4127 else if (semantic_idx == 1)
4128 shader_addline(buffer, "ps_in[%u]%s = vec4(gl_SecondaryColor)%s;\n",
4129 shader->u.ps.input_reg_map[i], reg_mask, reg_mask);
4131 shader_addline(buffer, "ps_in[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
4132 shader->u.ps.input_reg_map[i], reg_mask, 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);
4142 /*********************************************
4143 * Vertex Shader Specific Code begins here
4144 ********************************************/
4146 static void add_glsl_program_entry(struct shader_glsl_priv *priv, struct glsl_shader_prog_link *entry)
4148 struct glsl_program_key key;
4150 key.vs_id = entry->vs.id;
4151 key.gs_id = entry->gs.id;
4152 key.ps_id = entry->ps.id;
4154 if (wine_rb_put(&priv->program_lookup, &key, &entry->program_lookup_entry) == -1)
4156 ERR("Failed to insert program entry.\n");
4160 static struct glsl_shader_prog_link *get_glsl_program_entry(const struct shader_glsl_priv *priv,
4161 GLhandleARB vs_id, GLhandleARB gs_id, GLhandleARB ps_id)
4163 struct wine_rb_entry *entry;
4164 struct glsl_program_key key;
4170 entry = wine_rb_get(&priv->program_lookup, &key);
4171 return entry ? WINE_RB_ENTRY_VALUE(entry, struct glsl_shader_prog_link, program_lookup_entry) : NULL;
4174 /* Context activation is done by the caller. */
4175 static void delete_glsl_program_entry(struct shader_glsl_priv *priv, const struct wined3d_gl_info *gl_info,
4176 struct glsl_shader_prog_link *entry)
4178 struct glsl_program_key key;
4180 key.vs_id = entry->vs.id;
4181 key.gs_id = entry->gs.id;
4182 key.ps_id = entry->ps.id;
4183 wine_rb_remove(&priv->program_lookup, &key);
4185 GL_EXTCALL(glDeleteObjectARB(entry->programId));
4187 list_remove(&entry->vs.shader_entry);
4189 list_remove(&entry->gs.shader_entry);
4191 list_remove(&entry->ps.shader_entry);
4192 HeapFree(GetProcessHeap(), 0, entry->vs.uniform_f_locations);
4193 HeapFree(GetProcessHeap(), 0, entry->ps.uniform_f_locations);
4194 HeapFree(GetProcessHeap(), 0, entry);
4197 static void handle_ps3_input(struct wined3d_shader_buffer *buffer,
4198 const struct wined3d_gl_info *gl_info, const DWORD *map,
4199 const struct wined3d_shader_signature_element *input_signature,
4200 const struct wined3d_shader_reg_maps *reg_maps_in,
4201 const struct wined3d_shader_signature_element *output_signature,
4202 const struct wined3d_shader_reg_maps *reg_maps_out)
4205 const char *semantic_name_in;
4206 UINT semantic_idx_in;
4209 unsigned int in_count = vec4_varyings(3, gl_info);
4211 char destination[50];
4212 WORD input_map, output_map;
4214 set = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*set) * (in_count + 2));
4216 input_map = reg_maps_in->input_registers;
4217 for (i = 0; input_map; input_map >>= 1, ++i)
4219 if (!(input_map & 1)) continue;
4222 /* Declared, but not read register */
4223 if (in_idx == ~0U) continue;
4224 if (in_idx >= (in_count + 2))
4226 FIXME("More input varyings declared than supported, expect issues.\n");
4230 if (in_idx == in_count)
4231 sprintf(destination, "gl_FrontColor");
4232 else if (in_idx == in_count + 1)
4233 sprintf(destination, "gl_FrontSecondaryColor");
4235 sprintf(destination, "ps_in[%u]", in_idx);
4237 semantic_name_in = input_signature[i].semantic_name;
4238 semantic_idx_in = input_signature[i].semantic_idx;
4241 output_map = reg_maps_out->output_registers;
4242 for (j = 0; output_map; output_map >>= 1, ++j)
4246 if (!(output_map & 1)
4247 || semantic_idx_in != output_signature[j].semantic_idx
4248 || strcmp(semantic_name_in, output_signature[j].semantic_name)
4249 || !(mask = input_signature[i].mask & output_signature[j].mask))
4253 shader_glsl_write_mask_to_str(mask, reg_mask);
4255 shader_addline(buffer, "%s%s = vs_out[%u]%s;\n",
4256 destination, reg_mask, j, reg_mask);
4260 for (i = 0; i < in_count + 2; ++i)
4264 if (!set[i] || set[i] == WINED3DSP_WRITEMASK_ALL)
4267 if (set[i] == ~0U) set[i] = 0;
4270 if (!(set[i] & WINED3DSP_WRITEMASK_0)) reg_mask[size++] = 'x';
4271 if (!(set[i] & WINED3DSP_WRITEMASK_1)) reg_mask[size++] = 'y';
4272 if (!(set[i] & WINED3DSP_WRITEMASK_2)) reg_mask[size++] = 'z';
4273 if (!(set[i] & WINED3DSP_WRITEMASK_3)) reg_mask[size++] = 'w';
4274 reg_mask[size] = '\0';
4277 sprintf(destination, "gl_FrontColor");
4278 else if (i == in_count + 1)
4279 sprintf(destination, "gl_FrontSecondaryColor");
4281 sprintf(destination, "ps_in[%u]", i);
4283 if (size == 1) shader_addline(buffer, "%s.%s = 0.0;\n", destination, reg_mask);
4284 else shader_addline(buffer, "%s.%s = vec%u(0.0);\n", destination, reg_mask, size);
4287 HeapFree(GetProcessHeap(), 0, set);
4290 /* Context activation is done by the caller. */
4291 static GLhandleARB generate_param_reorder_function(struct wined3d_shader_buffer *buffer,
4292 const struct wined3d_shader *vs, const struct wined3d_shader *ps,
4293 const struct wined3d_gl_info *gl_info)
4295 GLhandleARB ret = 0;
4296 DWORD ps_major = ps ? ps->reg_maps.shader_version.major : 0;
4298 const char *semantic_name;
4301 const struct wined3d_shader_signature_element *output_signature = vs->output_signature;
4302 WORD map = vs->reg_maps.output_registers;
4304 shader_buffer_clear(buffer);
4306 shader_addline(buffer, "#version 120\n");
4310 shader_addline(buffer, "void order_ps_input(in vec4 vs_out[%u])\n{\n", vs->limits.packed_output);
4312 for (i = 0; map; map >>= 1, ++i)
4316 if (!(map & 1)) continue;
4318 semantic_name = output_signature[i].semantic_name;
4319 semantic_idx = output_signature[i].semantic_idx;
4320 write_mask = output_signature[i].mask;
4321 shader_glsl_write_mask_to_str(write_mask, reg_mask);
4323 if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_COLOR))
4326 shader_addline(buffer, "gl_FrontColor%s = vs_out[%u]%s;\n",
4327 reg_mask, i, reg_mask);
4328 else if (semantic_idx == 1)
4329 shader_addline(buffer, "gl_FrontSecondaryColor%s = vs_out[%u]%s;\n",
4330 reg_mask, i, reg_mask);
4332 else if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_POSITION))
4334 shader_addline(buffer, "gl_Position%s = vs_out[%u]%s;\n",
4335 reg_mask, i, reg_mask);
4337 else if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_TEXCOORD))
4339 if (semantic_idx < 8)
4341 if (!(gl_info->quirks & WINED3D_QUIRK_SET_TEXCOORD_W) || ps_major > 0)
4342 write_mask |= WINED3DSP_WRITEMASK_3;
4344 shader_addline(buffer, "gl_TexCoord[%u]%s = vs_out[%u]%s;\n",
4345 semantic_idx, reg_mask, i, reg_mask);
4346 if (!(write_mask & WINED3DSP_WRITEMASK_3))
4347 shader_addline(buffer, "gl_TexCoord[%u].w = 1.0;\n", semantic_idx);
4350 else if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_PSIZE))
4352 shader_addline(buffer, "gl_PointSize = vs_out[%u].%c;\n", i, reg_mask[1]);
4354 else if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_FOG))
4356 shader_addline(buffer, "gl_FogFragCoord = clamp(vs_out[%u].%c, 0.0, 1.0);\n", i, reg_mask[1]);
4359 shader_addline(buffer, "}\n");
4363 UINT in_count = min(vec4_varyings(ps_major, gl_info), ps->limits.packed_input);
4364 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
4365 shader_addline(buffer, "varying vec4 ps_in[%u];\n", in_count);
4366 shader_addline(buffer, "void order_ps_input(in vec4 vs_out[%u])\n{\n", vs->limits.packed_output);
4368 /* First, sort out position and point size. Those are not passed to the pixel shader */
4369 for (i = 0; map; map >>= 1, ++i)
4371 if (!(map & 1)) continue;
4373 semantic_name = output_signature[i].semantic_name;
4374 shader_glsl_write_mask_to_str(output_signature[i].mask, reg_mask);
4376 if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_POSITION))
4378 shader_addline(buffer, "gl_Position%s = vs_out[%u]%s;\n",
4379 reg_mask, i, reg_mask);
4381 else if (shader_match_semantic(semantic_name, WINED3D_DECL_USAGE_PSIZE))
4383 shader_addline(buffer, "gl_PointSize = vs_out[%u].%c;\n", i, reg_mask[1]);
4387 /* Then, fix the pixel shader input */
4388 handle_ps3_input(buffer, gl_info, ps->u.ps.input_reg_map, ps->input_signature,
4389 &ps->reg_maps, output_signature, &vs->reg_maps);
4391 shader_addline(buffer, "}\n");
4394 ret = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
4395 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
4396 shader_glsl_compile(gl_info, ret, buffer->buffer);
4401 static void shader_glsl_generate_srgb_write_correction(struct wined3d_shader_buffer *buffer)
4403 shader_addline(buffer, "tmp0.xyz = pow(gl_FragData[0].xyz, vec3(srgb_const0.x));\n");
4404 shader_addline(buffer, "tmp0.xyz = tmp0.xyz * vec3(srgb_const0.y) - vec3(srgb_const0.z);\n");
4405 shader_addline(buffer, "tmp1.xyz = gl_FragData[0].xyz * vec3(srgb_const0.w);\n");
4406 shader_addline(buffer, "bvec3 srgb_compare = lessThan(gl_FragData[0].xyz, vec3(srgb_const1.x));\n");
4407 shader_addline(buffer, "gl_FragData[0].xyz = mix(tmp0.xyz, tmp1.xyz, vec3(srgb_compare));\n");
4408 shader_addline(buffer, "gl_FragData[0] = clamp(gl_FragData[0], 0.0, 1.0);\n");
4411 static void shader_glsl_generate_fog_code(struct wined3d_shader_buffer *buffer, enum fogmode mode)
4419 /* Fog = (gl_Fog.end - gl_FogFragCoord) / (gl_Fog.end - gl_Fog.start) */
4420 shader_addline(buffer, "float Fog = (gl_Fog.end - gl_FogFragCoord) / (gl_Fog.end - gl_Fog.start);\n");
4424 /* Fog = e^-(gl_Fog.density * gl_FogFragCoord) */
4425 shader_addline(buffer, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4429 /* Fog = e^-((gl_Fog.density * gl_FogFragCoord)^2) */
4430 shader_addline(buffer, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4434 ERR("Invalid fog mode %#x.\n", mode);
4438 shader_addline(buffer, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, clamp(Fog, 0.0, 1.0));\n");
4441 /* Context activation is done by the caller. */
4442 static GLuint shader_glsl_generate_pshader(const struct wined3d_context *context,
4443 struct wined3d_shader_buffer *buffer, const struct wined3d_shader *shader,
4444 const struct ps_compile_args *args, struct ps_np2fixup_info *np2fixup_info)
4446 const struct wined3d_shader_reg_maps *reg_maps = &shader->reg_maps;
4447 const struct wined3d_gl_info *gl_info = context->gl_info;
4448 const DWORD *function = shader->function;
4449 struct shader_glsl_ctx_priv priv_ctx;
4451 /* Create the hw GLSL shader object and assign it as the shader->prgId */
4452 GLhandleARB shader_obj = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
4454 memset(&priv_ctx, 0, sizeof(priv_ctx));
4455 priv_ctx.cur_ps_args = args;
4456 priv_ctx.cur_np2fixup_info = np2fixup_info;
4458 shader_addline(buffer, "#version 120\n");
4460 if (gl_info->supported[ARB_SHADER_BIT_ENCODING])
4461 shader_addline(buffer, "#extension GL_ARB_shader_bit_encoding : enable\n");
4462 if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
4463 shader_addline(buffer, "#extension GL_ARB_shader_texture_lod : enable\n");
4464 /* The spec says that it doesn't have to be explicitly enabled, but the
4465 * nvidia drivers write a warning if we don't do so. */
4466 if (gl_info->supported[ARB_TEXTURE_RECTANGLE])
4467 shader_addline(buffer, "#extension GL_ARB_texture_rectangle : enable\n");
4468 if (gl_info->supported[EXT_GPU_SHADER4])
4469 shader_addline(buffer, "#extension GL_EXT_gpu_shader4 : enable\n");
4471 /* Base Declarations */
4472 shader_generate_glsl_declarations(context, buffer, shader, reg_maps, &priv_ctx);
4474 /* Pack 3.0 inputs */
4475 if (reg_maps->shader_version.major >= 3 && args->vp_mode != vertexshader)
4476 shader_glsl_input_pack(shader, buffer, shader->input_signature, reg_maps, args->vp_mode);
4478 /* Base Shader Body */
4479 shader_generate_main(shader, buffer, reg_maps, function, &priv_ctx);
4481 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
4482 if (reg_maps->shader_version.major < 2)
4484 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
4485 shader_addline(buffer, "gl_FragData[0] = R0;\n");
4488 if (args->srgb_correction)
4489 shader_glsl_generate_srgb_write_correction(buffer);
4491 /* SM < 3 does not replace the fog stage. */
4492 if (reg_maps->shader_version.major < 3)
4493 shader_glsl_generate_fog_code(buffer, args->fog);
4495 shader_addline(buffer, "}\n");
4497 TRACE("Compiling shader object %u\n", shader_obj);
4498 shader_glsl_compile(gl_info, shader_obj, buffer->buffer);
4500 /* Store the shader object */
4504 /* Context activation is done by the caller. */
4505 static GLuint shader_glsl_generate_vshader(const struct wined3d_context *context,
4506 struct wined3d_shader_buffer *buffer, const struct wined3d_shader *shader,
4507 const struct vs_compile_args *args)
4509 const struct wined3d_shader_reg_maps *reg_maps = &shader->reg_maps;
4510 const struct wined3d_gl_info *gl_info = context->gl_info;
4511 const DWORD *function = shader->function;
4512 struct shader_glsl_ctx_priv priv_ctx;
4514 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4515 GLhandleARB shader_obj = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
4517 shader_addline(buffer, "#version 120\n");
4519 if (gl_info->supported[ARB_SHADER_BIT_ENCODING])
4520 shader_addline(buffer, "#extension GL_ARB_shader_bit_encoding : enable\n");
4521 if (gl_info->supported[EXT_GPU_SHADER4])
4522 shader_addline(buffer, "#extension GL_EXT_gpu_shader4 : enable\n");
4524 memset(&priv_ctx, 0, sizeof(priv_ctx));
4525 priv_ctx.cur_vs_args = args;
4527 /* Base Declarations */
4528 shader_generate_glsl_declarations(context, buffer, shader, reg_maps, &priv_ctx);
4530 /* Base Shader Body */
4531 shader_generate_main(shader, buffer, reg_maps, function, &priv_ctx);
4533 /* Unpack outputs */
4534 shader_addline(buffer, "order_ps_input(vs_out);\n");
4536 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4537 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4538 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4539 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4541 if (args->fog_src == VS_FOG_Z)
4542 shader_addline(buffer, "gl_FogFragCoord = gl_Position.z;\n");
4543 else if (!reg_maps->fog)
4544 shader_addline(buffer, "gl_FogFragCoord = 0.0;\n");
4546 /* We always store the clipplanes without y inversion */
4547 if (args->clip_enabled)
4548 shader_addline(buffer, "gl_ClipVertex = gl_Position;\n");
4550 /* Write the final position.
4552 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4553 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4554 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4555 * contains 1.0 to allow a mad.
4557 shader_addline(buffer, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4558 shader_addline(buffer, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4560 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4562 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4563 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4564 * which is the same as z = z * 2 - w.
4566 shader_addline(buffer, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4568 shader_addline(buffer, "}\n");
4570 TRACE("Compiling shader object %u\n", shader_obj);
4571 shader_glsl_compile(gl_info, shader_obj, buffer->buffer);
4576 /* Context activation is done by the caller. */
4577 static GLhandleARB shader_glsl_generate_geometry_shader(const struct wined3d_context *context,
4578 struct wined3d_shader_buffer *buffer, const struct wined3d_shader *shader)
4580 const struct wined3d_shader_reg_maps *reg_maps = &shader->reg_maps;
4581 const struct wined3d_gl_info *gl_info = context->gl_info;
4582 const DWORD *function = shader->function;
4583 struct shader_glsl_ctx_priv priv_ctx;
4584 GLhandleARB shader_id;
4586 shader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_GEOMETRY_SHADER_ARB));
4588 shader_addline(buffer, "#version 120\n");
4590 if (gl_info->supported[ARB_GEOMETRY_SHADER4])
4591 shader_addline(buffer, "#extension GL_ARB_geometry_shader4 : enable\n");
4592 if (gl_info->supported[ARB_SHADER_BIT_ENCODING])
4593 shader_addline(buffer, "#extension GL_ARB_shader_bit_encoding : enable\n");
4594 if (gl_info->supported[EXT_GPU_SHADER4])
4595 shader_addline(buffer, "#extension GL_EXT_gpu_shader4 : enable\n");
4597 memset(&priv_ctx, 0, sizeof(priv_ctx));
4598 shader_generate_glsl_declarations(context, buffer, shader, reg_maps, &priv_ctx);
4599 shader_generate_main(shader, buffer, reg_maps, function, &priv_ctx);
4600 shader_addline(buffer, "}\n");
4602 TRACE("Compiling shader object %u.\n", shader_id);
4603 shader_glsl_compile(gl_info, shader_id, buffer->buffer);
4608 static GLhandleARB find_glsl_pshader(const struct wined3d_context *context,
4609 struct wined3d_shader_buffer *buffer, struct wined3d_shader *shader,
4610 const struct ps_compile_args *args, const struct ps_np2fixup_info **np2fixup_info)
4612 struct glsl_ps_compiled_shader *gl_shaders, *new_array;
4613 struct glsl_shader_private *shader_data;
4614 struct ps_np2fixup_info *np2fixup;
4619 if (!shader->backend_data)
4621 shader->backend_data = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*shader_data));
4622 if (!shader->backend_data)
4624 ERR("Failed to allocate backend data.\n");
4628 shader_data = shader->backend_data;
4629 gl_shaders = shader_data->gl_shaders.ps;
4631 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4632 * so a linear search is more performant than a hashmap or a binary search
4633 * (cache coherency etc)
4635 for (i = 0; i < shader_data->num_gl_shaders; ++i)
4637 if (!memcmp(&gl_shaders[i].args, args, sizeof(*args)))
4639 if (args->np2_fixup)
4640 *np2fixup_info = &gl_shaders[i].np2fixup;
4641 return gl_shaders[i].prgId;
4645 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader);
4646 if(shader_data->shader_array_size == shader_data->num_gl_shaders) {
4647 if (shader_data->num_gl_shaders)
4649 new_size = shader_data->shader_array_size + max(1, shader_data->shader_array_size / 2);
4650 new_array = HeapReAlloc(GetProcessHeap(), 0, shader_data->gl_shaders.ps,
4651 new_size * sizeof(*gl_shaders));
4655 new_array = HeapAlloc(GetProcessHeap(), 0, sizeof(*gl_shaders));
4660 ERR("Out of memory\n");
4663 shader_data->gl_shaders.ps = new_array;
4664 shader_data->shader_array_size = new_size;
4665 gl_shaders = new_array;
4668 gl_shaders[shader_data->num_gl_shaders].args = *args;
4670 np2fixup = &gl_shaders[shader_data->num_gl_shaders].np2fixup;
4671 memset(np2fixup, 0, sizeof(*np2fixup));
4672 *np2fixup_info = args->np2_fixup ? np2fixup : NULL;
4674 pixelshader_update_samplers(shader, args->tex_types);
4676 shader_buffer_clear(buffer);
4677 ret = shader_glsl_generate_pshader(context, buffer, shader, args, np2fixup);
4678 gl_shaders[shader_data->num_gl_shaders++].prgId = ret;
4683 static inline BOOL vs_args_equal(const struct vs_compile_args *stored, const struct vs_compile_args *new,
4684 const DWORD use_map) {
4685 if((stored->swizzle_map & use_map) != new->swizzle_map) return FALSE;
4686 if((stored->clip_enabled) != new->clip_enabled) return FALSE;
4687 return stored->fog_src == new->fog_src;
4690 static GLhandleARB find_glsl_vshader(const struct wined3d_context *context,
4691 struct wined3d_shader_buffer *buffer, struct wined3d_shader *shader,
4692 const struct vs_compile_args *args)
4696 DWORD use_map = shader->device->stream_info.use_map;
4697 struct glsl_vs_compiled_shader *gl_shaders, *new_array;
4698 struct glsl_shader_private *shader_data;
4701 if (!shader->backend_data)
4703 shader->backend_data = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*shader_data));
4704 if (!shader->backend_data)
4706 ERR("Failed to allocate backend data.\n");
4710 shader_data = shader->backend_data;
4711 gl_shaders = shader_data->gl_shaders.vs;
4713 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4714 * so a linear search is more performant than a hashmap or a binary search
4715 * (cache coherency etc)
4717 for (i = 0; i < shader_data->num_gl_shaders; ++i)
4719 if (vs_args_equal(&gl_shaders[i].args, args, use_map))
4720 return gl_shaders[i].prgId;
4723 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader);
4725 if(shader_data->shader_array_size == shader_data->num_gl_shaders) {
4726 if (shader_data->num_gl_shaders)
4728 new_size = shader_data->shader_array_size + max(1, shader_data->shader_array_size / 2);
4729 new_array = HeapReAlloc(GetProcessHeap(), 0, shader_data->gl_shaders.vs,
4730 new_size * sizeof(*gl_shaders));
4734 new_array = HeapAlloc(GetProcessHeap(), 0, sizeof(*gl_shaders));
4739 ERR("Out of memory\n");
4742 shader_data->gl_shaders.vs = new_array;
4743 shader_data->shader_array_size = new_size;
4744 gl_shaders = new_array;
4747 gl_shaders[shader_data->num_gl_shaders].args = *args;
4749 shader_buffer_clear(buffer);
4750 ret = shader_glsl_generate_vshader(context, buffer, shader, args);
4751 gl_shaders[shader_data->num_gl_shaders++].prgId = ret;
4756 static GLhandleARB find_glsl_geometry_shader(const struct wined3d_context *context,
4757 struct wined3d_shader_buffer *buffer, struct wined3d_shader *shader)
4759 struct glsl_gs_compiled_shader *gl_shaders;
4760 struct glsl_shader_private *shader_data;
4763 if (!shader->backend_data)
4765 if (!(shader->backend_data = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*shader_data))))
4767 ERR("Failed to allocate backend data.\n");
4771 shader_data = shader->backend_data;
4772 gl_shaders = shader_data->gl_shaders.gs;
4774 if (shader_data->num_gl_shaders)
4775 return gl_shaders[0].id;
4777 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader);
4779 if (!(shader_data->gl_shaders.gs = HeapAlloc(GetProcessHeap(), 0, sizeof(*gl_shaders))))
4781 ERR("Failed to allocate GL shader array.\n");
4784 shader_data->shader_array_size = 1;
4785 gl_shaders = shader_data->gl_shaders.gs;
4787 shader_buffer_clear(buffer);
4788 ret = shader_glsl_generate_geometry_shader(context, buffer, shader);
4789 gl_shaders[shader_data->num_gl_shaders++].id = ret;
4794 static const char *shader_glsl_get_ffp_fragment_op_arg(struct wined3d_shader_buffer *buffer,
4795 DWORD argnum, unsigned int stage, DWORD arg)
4799 if (arg == ARG_UNUSED)
4800 return "<unused arg>";
4802 switch (arg & WINED3DTA_SELECTMASK)
4804 case WINED3DTA_DIFFUSE:
4808 case WINED3DTA_CURRENT:
4815 case WINED3DTA_TEXTURE:
4818 case 0: ret = "tex0"; break;
4819 case 1: ret = "tex1"; break;
4820 case 2: ret = "tex2"; break;
4821 case 3: ret = "tex3"; break;
4822 case 4: ret = "tex4"; break;
4823 case 5: ret = "tex5"; break;
4824 case 6: ret = "tex6"; break;
4825 case 7: ret = "tex7"; break;
4827 ret = "<invalid texture>";
4832 case WINED3DTA_TFACTOR:
4836 case WINED3DTA_SPECULAR:
4837 ret = "gl_SecondaryColor";
4840 case WINED3DTA_TEMP:
4844 case WINED3DTA_CONSTANT:
4845 FIXME("Per-stage constants not implemented.\n");
4848 case 0: ret = "const0"; break;
4849 case 1: ret = "const1"; break;
4850 case 2: ret = "const2"; break;
4851 case 3: ret = "const3"; break;
4852 case 4: ret = "const4"; break;
4853 case 5: ret = "const5"; break;
4854 case 6: ret = "const6"; break;
4855 case 7: ret = "const7"; break;
4857 ret = "<invalid constant>";
4863 return "<unhandled arg>";
4866 if (arg & WINED3DTA_COMPLEMENT)
4868 shader_addline(buffer, "arg%u = vec4(1.0) - %s;\n", argnum, ret);
4871 else if (argnum == 1)
4873 else if (argnum == 2)
4877 if (arg & WINED3DTA_ALPHAREPLICATE)
4879 shader_addline(buffer, "arg%u = vec4(%s.w);\n", argnum, ret);
4882 else if (argnum == 1)
4884 else if (argnum == 2)
4891 static void shader_glsl_ffp_fragment_op(struct wined3d_shader_buffer *buffer, unsigned int stage, BOOL color,
4892 BOOL alpha, DWORD dst, DWORD op, DWORD dw_arg0, DWORD dw_arg1, DWORD dw_arg2)
4894 const char *dstmask, *dstreg, *arg0, *arg1, *arg2;
4904 dstreg = "temp_reg";
4908 arg0 = shader_glsl_get_ffp_fragment_op_arg(buffer, 0, stage, dw_arg0);
4909 arg1 = shader_glsl_get_ffp_fragment_op_arg(buffer, 1, stage, dw_arg1);
4910 arg2 = shader_glsl_get_ffp_fragment_op_arg(buffer, 2, stage, dw_arg2);
4914 case WINED3D_TOP_DISABLE:
4916 shader_addline(buffer, "%s%s = gl_Color%s;\n", dstreg, dstmask, dstmask);
4919 case WINED3D_TOP_SELECT_ARG1:
4920 shader_addline(buffer, "%s%s = %s%s;\n", dstreg, dstmask, arg1, dstmask);
4923 case WINED3D_TOP_SELECT_ARG2:
4924 shader_addline(buffer, "%s%s = %s%s;\n", dstreg, dstmask, arg2, dstmask);
4927 case WINED3D_TOP_MODULATE:
4928 shader_addline(buffer, "%s%s = %s%s * %s%s;\n", dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4931 case WINED3D_TOP_MODULATE_4X:
4932 shader_addline(buffer, "%s%s = clamp(%s%s * %s%s * 4.0, 0.0, 1.0);\n",
4933 dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4936 case WINED3D_TOP_MODULATE_2X:
4937 shader_addline(buffer, "%s%s = clamp(%s%s * %s%s * 2.0, 0.0, 1.0);\n",
4938 dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4941 case WINED3D_TOP_ADD:
4942 shader_addline(buffer, "%s%s = clamp(%s%s + %s%s, 0.0, 1.0);\n",
4943 dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4946 case WINED3D_TOP_ADD_SIGNED:
4947 shader_addline(buffer, "%s%s = clamp(%s%s + (%s - vec4(0.5))%s, 0.0, 1.0);\n",
4948 dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4951 case WINED3D_TOP_ADD_SIGNED_2X:
4952 shader_addline(buffer, "%s%s = clamp((%s%s + (%s - vec4(0.5))%s) * 2.0, 0.0, 1.0);\n",
4953 dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4956 case WINED3D_TOP_SUBTRACT:
4957 shader_addline(buffer, "%s%s = clamp(%s%s - %s%s, 0.0, 1.0);\n",
4958 dstreg, dstmask, arg1, dstmask, arg2, dstmask);
4961 case WINED3D_TOP_ADD_SMOOTH:
4962 shader_addline(buffer, "%s%s = clamp((vec4(1.0) - %s)%s * %s%s + %s%s, 0.0, 1.0);\n",
4963 dstreg, dstmask, arg1, dstmask, arg2, dstmask, arg1, dstmask);
4966 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA:
4967 arg0 = shader_glsl_get_ffp_fragment_op_arg(buffer, 0, stage, WINED3DTA_DIFFUSE);
4968 shader_addline(buffer, "%s%s = mix(%s%s, %s%s, %s.w);\n",
4969 dstreg, dstmask, arg2, dstmask, arg1, dstmask, arg0);
4972 case WINED3D_TOP_BLEND_TEXTURE_ALPHA:
4973 arg0 = shader_glsl_get_ffp_fragment_op_arg(buffer, 0, stage, WINED3DTA_TEXTURE);
4974 shader_addline(buffer, "%s%s = mix(%s%s, %s%s, %s.w);\n",
4975 dstreg, dstmask, arg2, dstmask, arg1, dstmask, arg0);
4978 case WINED3D_TOP_BLEND_FACTOR_ALPHA:
4979 arg0 = shader_glsl_get_ffp_fragment_op_arg(buffer, 0, stage, WINED3DTA_TFACTOR);
4980 shader_addline(buffer, "%s%s = mix(%s%s, %s%s, %s.w);\n",
4981 dstreg, dstmask, arg2, dstmask, arg1, dstmask, arg0);
4984 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM:
4985 arg0 = shader_glsl_get_ffp_fragment_op_arg(buffer, 0, stage, WINED3DTA_TEXTURE);
4986 shader_addline(buffer, "%s%s = clamp(%s%s * (1.0 - %s.w) + %s%s, 0.0, 1.0);\n",
4987 dstreg, dstmask, arg2, dstmask, arg0, arg1, dstmask);
4990 case WINED3D_TOP_BLEND_CURRENT_ALPHA:
4991 arg0 = shader_glsl_get_ffp_fragment_op_arg(buffer, 0, stage, WINED3DTA_CURRENT);
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_MODULATE_ALPHA_ADD_COLOR:
4997 shader_addline(buffer, "%s%s = clamp(%s%s * %s.w + %s%s, 0.0, 1.0);\n",
4998 dstreg, dstmask, arg2, dstmask, arg1, arg1, dstmask);
5001 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA:
5002 shader_addline(buffer, "%s%s = clamp(%s%s * %s%s + %s.w, 0.0, 1.0);\n",
5003 dstreg, dstmask, arg1, dstmask, arg2, dstmask, arg1);
5006 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR:
5007 shader_addline(buffer, "%s%s = clamp(%s%s * (1.0 - %s.w) + %s%s, 0.0, 1.0);\n",
5008 dstreg, dstmask, arg2, dstmask, arg1, arg1, dstmask);
5010 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA:
5011 shader_addline(buffer, "%s%s = clamp((vec4(1.0) - %s)%s * %s%s + %s.w, 0.0, 1.0);\n",
5012 dstreg, dstmask, arg1, dstmask, arg2, dstmask, arg1);
5015 case WINED3D_TOP_BUMPENVMAP:
5016 case WINED3D_TOP_BUMPENVMAP_LUMINANCE:
5017 /* These are handled in the first pass, nothing to do. */
5020 case WINED3D_TOP_DOTPRODUCT3:
5021 shader_addline(buffer, "%s%s = vec4(clamp(dot(%s.xyz - 0.5, %s.xyz - 0.5) * 4.0, 0.0, 1.0))%s;\n",
5022 dstreg, dstmask, arg1, arg2, dstmask);
5025 case WINED3D_TOP_MULTIPLY_ADD:
5026 shader_addline(buffer, "%s%s = clamp(%s%s * %s%s + %s%s, 0.0, 1.0);\n",
5027 dstreg, dstmask, arg1, dstmask, arg2, dstmask, arg0, dstmask);
5030 case WINED3D_TOP_LERP:
5031 /* MSDN isn't quite right here. */
5032 shader_addline(buffer, "%s%s = mix(%s%s, %s%s, %s%s);\n",
5033 dstreg, dstmask, arg2, dstmask, arg1, dstmask, arg0, dstmask);
5037 FIXME("Unhandled operation %#x.\n", op);
5042 /* Context activation is done by the caller. */
5043 static GLuint shader_glsl_generate_ffp_fragment_shader(struct wined3d_shader_buffer *buffer,
5044 const struct ffp_frag_settings *settings, const struct wined3d_gl_info *gl_info)
5046 BOOL tempreg_used = FALSE, tfactor_used = FALSE;
5047 BYTE lum_map = 0, bump_map = 0, tex_map = 0;
5048 const char *final_combiner_src = "ret";
5049 UINT lowest_disabled_stage;
5050 GLhandleARB shader_obj;
5051 DWORD arg0, arg1, arg2;
5054 shader_buffer_clear(buffer);
5056 /* Find out which textures are read */
5057 for (stage = 0; stage < MAX_TEXTURES; ++stage)
5059 if (settings->op[stage].cop == WINED3D_TOP_DISABLE)
5062 arg0 = settings->op[stage].carg0 & WINED3DTA_SELECTMASK;
5063 arg1 = settings->op[stage].carg1 & WINED3DTA_SELECTMASK;
5064 arg2 = settings->op[stage].carg2 & WINED3DTA_SELECTMASK;
5066 if (arg0 == WINED3DTA_TEXTURE || arg1 == WINED3DTA_TEXTURE || arg2 == WINED3DTA_TEXTURE)
5067 tex_map |= 1 << stage;
5068 if (arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR)
5069 tfactor_used = TRUE;
5070 if (arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP)
5071 tempreg_used = TRUE;
5072 if (settings->op[stage].dst == tempreg)
5073 tempreg_used = TRUE;
5075 switch (settings->op[stage].cop)
5077 case WINED3D_TOP_BUMPENVMAP_LUMINANCE:
5078 lum_map |= 1 << stage;
5080 case WINED3D_TOP_BUMPENVMAP:
5081 bump_map |= 1 << stage;
5083 case WINED3D_TOP_BLEND_TEXTURE_ALPHA:
5084 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM:
5085 tex_map |= 1 << stage;
5088 case WINED3D_TOP_BLEND_FACTOR_ALPHA:
5089 tfactor_used = TRUE;
5096 if (settings->op[stage].aop == WINED3D_TOP_DISABLE)
5099 arg0 = settings->op[stage].aarg0 & WINED3DTA_SELECTMASK;
5100 arg1 = settings->op[stage].aarg1 & WINED3DTA_SELECTMASK;
5101 arg2 = settings->op[stage].aarg2 & WINED3DTA_SELECTMASK;
5103 if (arg0 == WINED3DTA_TEXTURE || arg1 == WINED3DTA_TEXTURE || arg2 == WINED3DTA_TEXTURE)
5104 tex_map |= 1 << stage;
5105 if (arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR)
5106 tfactor_used = TRUE;
5107 if (arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP)
5108 tempreg_used = TRUE;
5110 lowest_disabled_stage = stage;
5112 shader_addline(buffer, "#version 120\n");
5114 if (gl_info->supported[ARB_TEXTURE_RECTANGLE])
5115 shader_addline(buffer, "#extension GL_ARB_texture_rectangle : enable\n");
5117 shader_addline(buffer, "vec4 tmp0, tmp1;\n");
5118 shader_addline(buffer, "vec4 ret;\n");
5119 if (tempreg_used || settings->sRGB_write)
5120 shader_addline(buffer, "vec4 temp_reg;\n");
5121 shader_addline(buffer, "vec4 arg0, arg1, arg2;\n");
5123 for (stage = 0; stage < MAX_TEXTURES; ++stage)
5125 if (!(tex_map & (1 << stage)))
5128 switch (settings->op[stage].tex_type)
5131 shader_addline(buffer, "uniform sampler1D ps_sampler%u;\n", stage);
5134 shader_addline(buffer, "uniform sampler2D ps_sampler%u;\n", stage);
5137 shader_addline(buffer, "uniform sampler3D ps_sampler%u;\n", stage);
5140 shader_addline(buffer, "uniform samplerCube ps_sampler%u;\n", stage);
5143 shader_addline(buffer, "uniform sampler2DRect ps_sampler%u;\n", stage);
5146 FIXME("Unhandled sampler type %#x.\n", settings->op[stage].tex_type);
5150 shader_addline(buffer, "vec4 tex%u;\n", stage);
5152 if (!(bump_map & (1 << stage)))
5154 shader_addline(buffer, "uniform mat2 bumpenv_mat%u;\n", stage);
5156 if (!(lum_map & (1 << stage)))
5158 shader_addline(buffer, "uniform float bumpenv_lum_scale%u;\n", stage);
5159 shader_addline(buffer, "uniform float bumpenv_lum_offset%u;\n", stage);
5162 shader_addline(buffer, "uniform vec4 tex_factor;\n");
5163 shader_addline(buffer, "uniform vec4 specular_enable;\n");
5165 if (settings->sRGB_write)
5167 shader_addline(buffer, "const vec4 srgb_const0 = vec4(%.8e, %.8e, %.8e, %.8e);\n",
5168 srgb_pow, srgb_mul_high, srgb_sub_high, srgb_mul_low);
5169 shader_addline(buffer, "const vec4 srgb_const1 = vec4(%.8e, 0.0, 0.0, 0.0);\n",
5173 shader_addline(buffer, "void main()\n{\n");
5175 if (lowest_disabled_stage < 7 && settings->emul_clipplanes)
5176 shader_addline(buffer, "if (any(lessThan(gl_texCoord[7], vec4(0.0)))) discard;\n");
5178 /* Generate texture sampling instructions) */
5179 for (stage = 0; stage < MAX_TEXTURES && settings->op[stage].cop != WINED3D_TOP_DISABLE; ++stage)
5181 const char *texture_function, *coord_mask;
5182 char tex_reg_name[8];
5185 if (!(tex_map & (1 << stage)))
5188 if (settings->op[stage].projected == proj_none)
5192 else if (settings->op[stage].projected == proj_count4
5193 || settings->op[stage].projected == proj_count3)
5199 FIXME("Unexpected projection mode %d\n", settings->op[stage].projected);
5203 if (settings->op[stage].cop == WINED3D_TOP_BUMPENVMAP
5204 || settings->op[stage].cop == WINED3D_TOP_BUMPENVMAP_LUMINANCE)
5209 switch (settings->op[stage].tex_type)
5214 texture_function = "texture1DProj";
5219 texture_function = "texture1D";
5226 texture_function = "texture2DProj";
5231 texture_function = "texture2D";
5238 texture_function = "texture3DProj";
5239 coord_mask = "xyzw";
5243 texture_function = "texture3D";
5248 texture_function = "textureCube";
5254 texture_function = "texture2DRectProj";
5259 texture_function = "texture2DRect";
5264 FIXME("Unhandled texture type %#x.\n", settings->op[stage].tex_type);
5265 texture_function = "";
5266 coord_mask = "xyzw";
5271 && (settings->op[stage - 1].cop == WINED3D_TOP_BUMPENVMAP
5272 || settings->op[stage - 1].cop == WINED3D_TOP_BUMPENVMAP_LUMINANCE))
5274 shader_addline(buffer, "ret.xy = bumpenv_mat%u * tex%u.xy;\n", stage - 1, stage - 1);
5276 /* With projective textures, texbem only divides the static
5277 * texture coord, not the displacement, so multiply the
5278 * displacement with the dividing parameter before passing it to
5280 if (settings->op[stage].projected != proj_none)
5282 if (settings->op[stage].projected == proj_count4)
5284 shader_addline(buffer, "ret.xy = (ret.xy * gl_TexCoord[%u].w) + gl_TexCoord[%u].xy;\n",
5286 shader_addline(buffer, "ret.zw = gl_TexCoord[%u].ww;\n", stage);
5290 shader_addline(buffer, "ret.xy = (ret.xy * gl_TexCoord[%u].z) + gl_TexCoord[%u].xy;\n",
5292 shader_addline(buffer, "ret.zw = gl_TexCoord[%u].zz;\n", stage);
5297 shader_addline(buffer, "ret = gl_TexCoord[%u] + ret.xyxy;\n", stage);
5301 shader_addline(buffer, "tex%u = clamp(%s(ps_sampler%u, ret.%s), 0.0, 1.0);\n",
5302 stage, texture_function, stage, coord_mask);
5304 shader_addline(buffer, "tex%u = %s(ps_sampler%u, ret.%s);\n",
5305 stage, texture_function, stage, coord_mask);
5307 if (settings->op[stage - 1].cop == WINED3D_TOP_BUMPENVMAP_LUMINANCE)
5308 shader_addline(buffer, "tex%u *= clamp(tex%u.z * bumpenv_lum_scale%u + bumpenv_lum_offset%u, 0.0, 1.0);\n",
5309 stage, stage - 1, stage - 1, stage - 1);
5311 else if (settings->op[stage].projected == proj_count3)
5314 shader_addline(buffer, "tex%u = clamp(%s(ps_sampler%u, gl_TexCoord[%u].xyz), 0.0, 1.0);\n",
5315 stage, texture_function, stage, stage);
5317 shader_addline(buffer, "tex%u = %s(ps_sampler%u, gl_TexCoord[%u].xyz);\n",
5318 stage, texture_function, stage, stage);
5323 shader_addline(buffer, "tex%u = clamp(%s(ps_sampler%u, gl_TexCoord[%u].%s), 0.0, 1.0);\n",
5324 stage, texture_function, stage, stage, coord_mask);
5326 shader_addline(buffer, "tex%u = %s(ps_sampler%u, gl_TexCoord[%u].%s);\n",
5327 stage, texture_function, stage, stage, coord_mask);
5330 sprintf(tex_reg_name, "tex%u", stage);
5331 shader_glsl_color_correction_ext(buffer, tex_reg_name, WINED3DSP_WRITEMASK_ALL,
5332 settings->op[stage].color_fixup);
5335 /* Generate the main shader */
5336 for (stage = 0; stage < MAX_TEXTURES; ++stage)
5340 if (settings->op[stage].cop == WINED3D_TOP_DISABLE)
5343 final_combiner_src = "gl_Color";
5347 if (settings->op[stage].cop == WINED3D_TOP_SELECT_ARG1
5348 && settings->op[stage].aop == WINED3D_TOP_SELECT_ARG1)
5349 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg1;
5350 else if (settings->op[stage].cop == WINED3D_TOP_SELECT_ARG1
5351 && settings->op[stage].aop == WINED3D_TOP_SELECT_ARG2)
5352 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg2;
5353 else if (settings->op[stage].cop == WINED3D_TOP_SELECT_ARG2
5354 && settings->op[stage].aop == WINED3D_TOP_SELECT_ARG1)
5355 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg1;
5356 else if (settings->op[stage].cop == WINED3D_TOP_SELECT_ARG2
5357 && settings->op[stage].aop == WINED3D_TOP_SELECT_ARG2)
5358 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg2;
5360 op_equal = settings->op[stage].aop == settings->op[stage].cop
5361 && settings->op[stage].carg0 == settings->op[stage].aarg0
5362 && settings->op[stage].carg1 == settings->op[stage].aarg1
5363 && settings->op[stage].carg2 == settings->op[stage].aarg2;
5365 if (settings->op[stage].aop == WINED3D_TOP_DISABLE)
5367 shader_glsl_ffp_fragment_op(buffer, stage, TRUE, FALSE, settings->op[stage].dst,
5368 settings->op[stage].cop, settings->op[stage].carg0,
5369 settings->op[stage].carg1, settings->op[stage].carg2);
5371 shader_addline(buffer, "ret.w = gl_Color.w;\n");
5375 shader_glsl_ffp_fragment_op(buffer, stage, TRUE, TRUE, settings->op[stage].dst,
5376 settings->op[stage].cop, settings->op[stage].carg0,
5377 settings->op[stage].carg1, settings->op[stage].carg2);
5381 shader_glsl_ffp_fragment_op(buffer, stage, TRUE, FALSE, settings->op[stage].dst,
5382 settings->op[stage].cop, settings->op[stage].carg0,
5383 settings->op[stage].carg1, settings->op[stage].carg2);
5384 shader_glsl_ffp_fragment_op(buffer, stage, FALSE, TRUE, settings->op[stage].dst,
5385 settings->op[stage].aop, settings->op[stage].aarg0,
5386 settings->op[stage].aarg1, settings->op[stage].aarg2);
5390 shader_addline(buffer, "gl_FragData[0] = gl_SecondaryColor * specular_enable + %s;\n", final_combiner_src);
5392 if (settings->sRGB_write)
5393 shader_glsl_generate_srgb_write_correction(buffer);
5395 shader_glsl_generate_fog_code(buffer, settings->fog);
5397 shader_addline(buffer, "}\n");
5399 shader_obj = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
5400 shader_glsl_compile(gl_info, shader_obj, buffer->buffer);
5404 static struct glsl_ffp_fragment_shader *shader_glsl_find_ffp_fragment_shader(struct shader_glsl_priv *priv,
5405 const struct wined3d_gl_info *gl_info, const struct ffp_frag_settings *args)
5407 struct glsl_ffp_fragment_shader *glsl_desc;
5408 const struct ffp_frag_desc *desc;
5410 if ((desc = find_ffp_frag_shader(&priv->ffp_fragment_shaders, args)))
5411 return CONTAINING_RECORD(desc, struct glsl_ffp_fragment_shader, entry);
5413 if (!(glsl_desc = HeapAlloc(GetProcessHeap(), 0, sizeof(*glsl_desc))))
5416 glsl_desc->entry.settings = *args;
5417 glsl_desc->id = shader_glsl_generate_ffp_fragment_shader(&priv->shader_buffer, args, gl_info);
5418 list_init(&glsl_desc->linked_programs);
5419 add_ffp_frag_shader(&priv->ffp_fragment_shaders, &glsl_desc->entry);
5425 static void shader_glsl_init_vs_uniform_locations(const struct wined3d_gl_info *gl_info,
5426 GLhandleARB program_id, struct glsl_vs_program *vs)
5431 vs->uniform_f_locations = HeapAlloc(GetProcessHeap(), 0,
5432 sizeof(GLhandleARB) * gl_info->limits.glsl_vs_float_constants);
5433 for (i = 0; i < gl_info->limits.glsl_vs_float_constants; ++i)
5435 snprintf(name, sizeof(name), "vs_c[%u]", i);
5436 vs->uniform_f_locations[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5439 for (i = 0; i < MAX_CONST_I; ++i)
5441 snprintf(name, sizeof(name), "vs_i[%u]", i);
5442 vs->uniform_i_locations[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5445 vs->pos_fixup_location = GL_EXTCALL(glGetUniformLocationARB(program_id, "posFixup"));
5448 static void shader_glsl_init_ps_uniform_locations(const struct wined3d_gl_info *gl_info,
5449 GLhandleARB program_id, struct glsl_ps_program *ps)
5454 ps->uniform_f_locations = HeapAlloc(GetProcessHeap(), 0,
5455 sizeof(GLhandleARB) * gl_info->limits.glsl_ps_float_constants);
5456 for (i = 0; i < gl_info->limits.glsl_ps_float_constants; ++i)
5458 snprintf(name, sizeof(name), "ps_c[%u]", i);
5459 ps->uniform_f_locations[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5462 for (i = 0; i < MAX_CONST_I; ++i)
5464 snprintf(name, sizeof(name), "ps_i[%u]", i);
5465 ps->uniform_i_locations[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5468 for (i = 0; i < MAX_TEXTURES; ++i)
5470 snprintf(name, sizeof(name), "bumpenv_mat%u", i);
5471 ps->bumpenv_mat_location[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5472 snprintf(name, sizeof(name), "bumpenv_lum_scale%u", i);
5473 ps->bumpenv_lum_scale_location[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5474 snprintf(name, sizeof(name), "bumpenv_lum_offset%u", i);
5475 ps->bumpenv_lum_offset_location[i] = GL_EXTCALL(glGetUniformLocationARB(program_id, name));
5478 ps->tex_factor_location = GL_EXTCALL(glGetUniformLocationARB(program_id, "tex_factor"));
5479 ps->specular_enable_location = GL_EXTCALL(glGetUniformLocationARB(program_id, "specular_enable"));
5480 ps->np2_fixup_location = GL_EXTCALL(glGetUniformLocationARB(program_id, "ps_samplerNP2Fixup"));
5481 ps->ycorrection_location = GL_EXTCALL(glGetUniformLocationARB(program_id, "ycorrection"));
5484 /* Context activation is done by the caller. */
5485 static void set_glsl_shader_program(const struct wined3d_context *context, struct wined3d_device *device,
5486 enum wined3d_shader_mode vertex_mode, enum wined3d_shader_mode fragment_mode)
5488 const struct wined3d_state *state = &device->stateBlock->state;
5489 const struct wined3d_gl_info *gl_info = context->gl_info;
5490 const struct ps_np2fixup_info *np2fixup_info = NULL;
5491 struct shader_glsl_priv *priv = device->shader_priv;
5492 struct glsl_shader_prog_link *entry = NULL;
5493 struct wined3d_shader *vshader = NULL;
5494 struct wined3d_shader *gshader = NULL;
5495 struct wined3d_shader *pshader = NULL;
5496 GLhandleARB programId = 0;
5497 GLhandleARB reorder_shader_id = 0;
5499 struct ps_compile_args ps_compile_args;
5500 struct vs_compile_args vs_compile_args;
5501 GLhandleARB vs_id, gs_id, ps_id;
5502 struct list *ps_list;
5504 if (vertex_mode == WINED3D_SHADER_MODE_SHADER)
5506 vshader = state->vertex_shader;
5507 find_vs_compile_args(state, vshader, &vs_compile_args);
5508 vs_id = find_glsl_vshader(context, &priv->shader_buffer, vshader, &vs_compile_args);
5510 if ((gshader = state->geometry_shader))
5511 gs_id = find_glsl_geometry_shader(context, &priv->shader_buffer, gshader);
5521 if (fragment_mode == WINED3D_SHADER_MODE_SHADER)
5523 pshader = state->pixel_shader;
5524 find_ps_compile_args(state, pshader, &ps_compile_args);
5525 ps_id = find_glsl_pshader(context, &priv->shader_buffer,
5526 pshader, &ps_compile_args, &np2fixup_info);
5527 ps_list = &pshader->linked_programs;
5529 else if (fragment_mode == WINED3D_SHADER_MODE_FFP && priv->fragment_pipe == &glsl_fragment_pipe)
5531 struct glsl_ffp_fragment_shader *ffp_shader;
5532 struct ffp_frag_settings settings;
5534 gen_ffp_frag_op(device, state, &settings, FALSE);
5535 ffp_shader = shader_glsl_find_ffp_fragment_shader(priv, gl_info, &settings);
5536 ps_id = ffp_shader->id;
5537 ps_list = &ffp_shader->linked_programs;
5544 if ((!vs_id && !gs_id && !ps_id) || (entry = get_glsl_program_entry(priv, vs_id, gs_id, ps_id)))
5546 priv->glsl_program = entry;
5550 /* If we get to this point, then no matching program exists, so we create one */
5551 programId = GL_EXTCALL(glCreateProgramObjectARB());
5552 TRACE("Created new GLSL shader program %u\n", programId);
5554 /* Create the entry */
5555 entry = HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link));
5556 entry->programId = programId;
5557 entry->vs.id = vs_id;
5558 entry->gs.id = gs_id;
5559 entry->ps.id = ps_id;
5560 entry->constant_version = 0;
5561 entry->ps.np2_fixup_info = np2fixup_info;
5562 /* Add the hash table entry */
5563 add_glsl_program_entry(priv, entry);
5565 /* Set the current program */
5566 priv->glsl_program = entry;
5568 /* Attach GLSL vshader */
5571 WORD map = vshader->reg_maps.input_registers;
5574 reorder_shader_id = generate_param_reorder_function(&priv->shader_buffer, vshader, pshader, gl_info);
5575 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id, programId);
5576 GL_EXTCALL(glAttachObjectARB(programId, reorder_shader_id));
5577 checkGLcall("glAttachObjectARB");
5578 /* Flag the reorder function for deletion, then it will be freed automatically when the program
5581 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id));
5583 TRACE("Attaching GLSL shader object %u to program %u.\n", vs_id, programId);
5584 GL_EXTCALL(glAttachObjectARB(programId, vs_id));
5585 checkGLcall("glAttachObjectARB");
5587 /* Bind vertex attributes to a corresponding index number to match
5588 * the same index numbers as ARB_vertex_programs (makes loading
5589 * vertex attributes simpler). With this method, we can use the
5590 * exact same code to load the attributes later for both ARB and
5593 * We have to do this here because we need to know the Program ID
5594 * in order to make the bindings work, and it has to be done prior
5595 * to linking the GLSL program. */
5596 for (i = 0; map; map >>= 1, ++i)
5598 if (!(map & 1)) continue;
5600 snprintf(tmp_name, sizeof(tmp_name), "vs_in%u", i);
5601 GL_EXTCALL(glBindAttribLocationARB(programId, i, tmp_name));
5603 checkGLcall("glBindAttribLocationARB");
5605 list_add_head(&vshader->linked_programs, &entry->vs.shader_entry);
5610 TRACE("Attaching GLSL geometry shader object %u to program %u.\n", gs_id, programId);
5611 GL_EXTCALL(glAttachObjectARB(programId, gs_id));
5612 checkGLcall("glAttachObjectARB");
5614 TRACE("input type %s, output type %s, vertices out %u.\n",
5615 debug_d3dprimitivetype(gshader->u.gs.input_type),
5616 debug_d3dprimitivetype(gshader->u.gs.output_type),
5617 gshader->u.gs.vertices_out);
5618 GL_EXTCALL(glProgramParameteriARB(programId, GL_GEOMETRY_INPUT_TYPE_ARB,
5619 gl_primitive_type_from_d3d(gshader->u.gs.input_type)));
5620 GL_EXTCALL(glProgramParameteriARB(programId, GL_GEOMETRY_OUTPUT_TYPE_ARB,
5621 gl_primitive_type_from_d3d(gshader->u.gs.output_type)));
5622 GL_EXTCALL(glProgramParameteriARB(programId, GL_GEOMETRY_VERTICES_OUT_ARB,
5623 gshader->u.gs.vertices_out));
5624 checkGLcall("glProgramParameteriARB");
5626 list_add_head(&gshader->linked_programs, &entry->gs.shader_entry);
5629 /* Attach GLSL pshader */
5632 TRACE("Attaching GLSL shader object %u to program %u.\n", ps_id, programId);
5633 GL_EXTCALL(glAttachObjectARB(programId, ps_id));
5634 checkGLcall("glAttachObjectARB");
5636 list_add_head(ps_list, &entry->ps.shader_entry);
5639 /* Link the program */
5640 TRACE("Linking GLSL shader program %u\n", programId);
5641 GL_EXTCALL(glLinkProgramARB(programId));
5642 shader_glsl_validate_link(gl_info, programId);
5644 shader_glsl_init_vs_uniform_locations(gl_info, programId, &entry->vs);
5645 shader_glsl_init_ps_uniform_locations(gl_info, programId, &entry->ps);
5646 checkGLcall("Find glsl program uniform locations");
5648 if (pshader && pshader->reg_maps.shader_version.major >= 3
5649 && pshader->u.ps.declared_in_count > vec4_varyings(3, gl_info))
5651 TRACE("Shader %d needs vertex color clamping disabled\n", programId);
5652 entry->vs.vertex_color_clamp = GL_FALSE;
5656 entry->vs.vertex_color_clamp = GL_FIXED_ONLY_ARB;
5659 /* Set the shader to allow uniform loading on it */
5660 GL_EXTCALL(glUseProgramObjectARB(programId));
5661 checkGLcall("glUseProgramObjectARB(programId)");
5663 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
5664 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
5665 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
5666 * vertex shader with fixed function pixel processing is used we make sure that the card
5667 * supports enough samplers to allow the max number of vertex samplers with all possible
5668 * fixed function fragment processing setups. So once the program is linked these samplers
5671 shader_glsl_load_vsamplers(gl_info, device->texUnitMap, programId);
5672 shader_glsl_load_psamplers(gl_info, device->texUnitMap, programId);
5675 /* Context activation is done by the caller. */
5676 static GLhandleARB create_glsl_blt_shader(const struct wined3d_gl_info *gl_info, enum tex_types tex_type, BOOL masked)
5678 GLhandleARB program_id;
5679 GLhandleARB vshader_id, pshader_id;
5680 const char *blt_pshader;
5682 static const char *blt_vshader =
5686 " gl_Position = gl_Vertex;\n"
5687 " gl_FrontColor = vec4(1.0);\n"
5688 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
5691 static const char * const blt_pshaders_full[tex_type_count] =
5697 "uniform sampler2D sampler;\n"
5700 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
5706 "uniform samplerCube sampler;\n"
5709 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
5713 "#extension GL_ARB_texture_rectangle : enable\n"
5714 "uniform sampler2DRect sampler;\n"
5717 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
5721 static const char * const blt_pshaders_masked[tex_type_count] =
5727 "uniform sampler2D sampler;\n"
5728 "uniform vec4 mask;\n"
5731 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
5732 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
5738 "uniform samplerCube sampler;\n"
5739 "uniform vec4 mask;\n"
5742 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
5743 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
5747 "#extension GL_ARB_texture_rectangle : enable\n"
5748 "uniform sampler2DRect sampler;\n"
5749 "uniform vec4 mask;\n"
5752 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
5753 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
5757 blt_pshader = masked ? blt_pshaders_masked[tex_type] : blt_pshaders_full[tex_type];
5760 FIXME("tex_type %#x not supported\n", tex_type);
5764 vshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
5765 shader_glsl_compile(gl_info, vshader_id, blt_vshader);
5767 pshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
5768 shader_glsl_compile(gl_info, pshader_id, blt_pshader);
5770 program_id = GL_EXTCALL(glCreateProgramObjectARB());
5771 GL_EXTCALL(glAttachObjectARB(program_id, vshader_id));
5772 GL_EXTCALL(glAttachObjectARB(program_id, pshader_id));
5773 GL_EXTCALL(glLinkProgramARB(program_id));
5775 shader_glsl_validate_link(gl_info, program_id);
5777 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
5780 GL_EXTCALL(glDeleteObjectARB(vshader_id));
5781 GL_EXTCALL(glDeleteObjectARB(pshader_id));
5785 /* Context activation is done by the caller. */
5786 static void shader_glsl_select(const struct wined3d_context *context, enum wined3d_shader_mode vertex_mode,
5787 enum wined3d_shader_mode fragment_mode)
5789 const struct wined3d_gl_info *gl_info = context->gl_info;
5790 struct wined3d_device *device = context->swapchain->device;
5791 struct shader_glsl_priv *priv = device->shader_priv;
5792 GLhandleARB program_id = 0;
5793 GLenum old_vertex_color_clamp, current_vertex_color_clamp;
5795 old_vertex_color_clamp = priv->glsl_program ? priv->glsl_program->vs.vertex_color_clamp : GL_FIXED_ONLY_ARB;
5796 set_glsl_shader_program(context, device, vertex_mode, fragment_mode);
5797 current_vertex_color_clamp = priv->glsl_program ? priv->glsl_program->vs.vertex_color_clamp : GL_FIXED_ONLY_ARB;
5798 if (old_vertex_color_clamp != current_vertex_color_clamp)
5800 if (gl_info->supported[ARB_COLOR_BUFFER_FLOAT])
5802 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB, current_vertex_color_clamp));
5803 checkGLcall("glClampColorARB");
5807 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
5811 program_id = priv->glsl_program ? priv->glsl_program->programId : 0;
5812 if (program_id) TRACE("Using GLSL program %u\n", program_id);
5813 GL_EXTCALL(glUseProgramObjectARB(program_id));
5814 checkGLcall("glUseProgramObjectARB");
5816 /* In case that NP2 texcoord fixup data is found for the selected program, trigger a reload of the
5817 * constants. This has to be done because it can't be guaranteed that sampler() (from state.c) is
5818 * called between selecting the shader and using it, which results in wrong fixup for some frames. */
5819 if (priv->glsl_program && priv->glsl_program->ps.np2_fixup_info)
5821 shader_glsl_load_np2fixup_constants(priv, gl_info, &device->stateBlock->state);
5825 /* Context activation is done by the caller. */
5826 static void shader_glsl_select_depth_blt(void *shader_priv, const struct wined3d_gl_info *gl_info,
5827 enum tex_types tex_type, const SIZE *ds_mask_size)
5829 BOOL masked = ds_mask_size->cx && ds_mask_size->cy;
5830 struct shader_glsl_priv *priv = shader_priv;
5831 GLhandleARB *blt_program;
5834 blt_program = masked ? &priv->depth_blt_program_masked[tex_type] : &priv->depth_blt_program_full[tex_type];
5837 *blt_program = create_glsl_blt_shader(gl_info, tex_type, masked);
5838 loc = GL_EXTCALL(glGetUniformLocationARB(*blt_program, "sampler"));
5839 GL_EXTCALL(glUseProgramObjectARB(*blt_program));
5840 GL_EXTCALL(glUniform1iARB(loc, 0));
5844 GL_EXTCALL(glUseProgramObjectARB(*blt_program));
5849 loc = GL_EXTCALL(glGetUniformLocationARB(*blt_program, "mask"));
5850 GL_EXTCALL(glUniform4fARB(loc, 0.0f, 0.0f, (float)ds_mask_size->cx, (float)ds_mask_size->cy));
5854 /* Context activation is done by the caller. */
5855 static void shader_glsl_deselect_depth_blt(void *shader_priv, const struct wined3d_gl_info *gl_info)
5857 struct shader_glsl_priv *priv = shader_priv;
5858 GLhandleARB program_id;
5860 program_id = priv->glsl_program ? priv->glsl_program->programId : 0;
5861 if (program_id) TRACE("Using GLSL program %u\n", program_id);
5863 GL_EXTCALL(glUseProgramObjectARB(program_id));
5864 checkGLcall("glUseProgramObjectARB");
5867 static void shader_glsl_destroy(struct wined3d_shader *shader)
5869 struct glsl_shader_private *shader_data = shader->backend_data;
5870 struct wined3d_device *device = shader->device;
5871 struct shader_glsl_priv *priv = device->shader_priv;
5872 const struct wined3d_gl_info *gl_info;
5873 const struct list *linked_programs;
5874 struct wined3d_context *context;
5876 if (!shader_data || !shader_data->num_gl_shaders)
5878 HeapFree(GetProcessHeap(), 0, shader_data);
5879 shader->backend_data = NULL;
5883 context = context_acquire(device, NULL);
5884 gl_info = context->gl_info;
5886 TRACE("Deleting linked programs.\n");
5887 linked_programs = &shader->linked_programs;
5888 if (linked_programs->next)
5890 struct glsl_shader_prog_link *entry, *entry2;
5893 switch (shader->reg_maps.shader_version.type)
5895 case WINED3D_SHADER_TYPE_PIXEL:
5897 struct glsl_ps_compiled_shader *gl_shaders = shader_data->gl_shaders.ps;
5899 LIST_FOR_EACH_ENTRY_SAFE(entry, entry2, linked_programs,
5900 struct glsl_shader_prog_link, ps.shader_entry)
5902 delete_glsl_program_entry(priv, gl_info, entry);
5905 for (i = 0; i < shader_data->num_gl_shaders; ++i)
5907 TRACE("Deleting pixel shader %u.\n", gl_shaders[i].prgId);
5908 if (priv->glsl_program && priv->glsl_program->ps.id == gl_shaders[i].prgId)
5909 shader_glsl_select(context, WINED3D_SHADER_MODE_NONE, WINED3D_SHADER_MODE_NONE);
5910 GL_EXTCALL(glDeleteObjectARB(gl_shaders[i].prgId));
5911 checkGLcall("glDeleteObjectARB");
5913 HeapFree(GetProcessHeap(), 0, shader_data->gl_shaders.ps);
5918 case WINED3D_SHADER_TYPE_VERTEX:
5920 struct glsl_vs_compiled_shader *gl_shaders = shader_data->gl_shaders.vs;
5922 LIST_FOR_EACH_ENTRY_SAFE(entry, entry2, linked_programs,
5923 struct glsl_shader_prog_link, vs.shader_entry)
5925 delete_glsl_program_entry(priv, gl_info, entry);
5928 for (i = 0; i < shader_data->num_gl_shaders; ++i)
5930 TRACE("Deleting vertex shader %u.\n", gl_shaders[i].prgId);
5931 if (priv->glsl_program && priv->glsl_program->vs.id == gl_shaders[i].prgId)
5932 shader_glsl_select(context, WINED3D_SHADER_MODE_NONE, WINED3D_SHADER_MODE_NONE);
5933 GL_EXTCALL(glDeleteObjectARB(gl_shaders[i].prgId));
5934 checkGLcall("glDeleteObjectARB");
5936 HeapFree(GetProcessHeap(), 0, shader_data->gl_shaders.vs);
5941 case WINED3D_SHADER_TYPE_GEOMETRY:
5943 struct glsl_gs_compiled_shader *gl_shaders = shader_data->gl_shaders.gs;
5945 LIST_FOR_EACH_ENTRY_SAFE(entry, entry2, linked_programs,
5946 struct glsl_shader_prog_link, gs.shader_entry)
5948 delete_glsl_program_entry(priv, gl_info, entry);
5951 for (i = 0; i < shader_data->num_gl_shaders; ++i)
5953 TRACE("Deleting geometry shader %u.\n", gl_shaders[i].id);
5954 if (priv->glsl_program && priv->glsl_program->gs.id == gl_shaders[i].id)
5955 shader_glsl_select(context, WINED3D_SHADER_MODE_NONE, WINED3D_SHADER_MODE_NONE);
5956 GL_EXTCALL(glDeleteObjectARB(gl_shaders[i].id));
5957 checkGLcall("glDeleteObjectARB");
5959 HeapFree(GetProcessHeap(), 0, shader_data->gl_shaders.gs);
5965 ERR("Unhandled shader type %#x.\n", shader->reg_maps.shader_version.type);
5970 HeapFree(GetProcessHeap(), 0, shader->backend_data);
5971 shader->backend_data = NULL;
5973 context_release(context);
5976 static int glsl_program_key_compare(const void *key, const struct wine_rb_entry *entry)
5978 const struct glsl_program_key *k = key;
5979 const struct glsl_shader_prog_link *prog = WINE_RB_ENTRY_VALUE(entry,
5980 const struct glsl_shader_prog_link, program_lookup_entry);
5982 if (k->vs_id > prog->vs.id) return 1;
5983 else if (k->vs_id < prog->vs.id) return -1;
5985 if (k->gs_id > prog->gs.id) return 1;
5986 else if (k->gs_id < prog->gs.id) return -1;
5988 if (k->ps_id > prog->ps.id) return 1;
5989 else if (k->ps_id < prog->ps.id) return -1;
5994 static BOOL constant_heap_init(struct constant_heap *heap, unsigned int constant_count)
5996 SIZE_T size = (constant_count + 1) * sizeof(*heap->entries) + constant_count * sizeof(*heap->positions);
5997 void *mem = HeapAlloc(GetProcessHeap(), 0, size);
6001 ERR("Failed to allocate memory\n");
6005 heap->entries = mem;
6006 heap->entries[1].version = 0;
6007 heap->positions = (unsigned int *)(heap->entries + constant_count + 1);
6013 static void constant_heap_free(struct constant_heap *heap)
6015 HeapFree(GetProcessHeap(), 0, heap->entries);
6018 static const struct wine_rb_functions wined3d_glsl_program_rb_functions =
6023 glsl_program_key_compare,
6026 static HRESULT shader_glsl_alloc(struct wined3d_device *device, const struct fragment_pipeline *fragment_pipe)
6028 const struct wined3d_gl_info *gl_info = &device->adapter->gl_info;
6029 struct shader_glsl_priv *priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_glsl_priv));
6030 SIZE_T stack_size = wined3d_log2i(max(gl_info->limits.glsl_vs_float_constants,
6031 gl_info->limits.glsl_ps_float_constants)) + 1;
6032 struct fragment_caps fragment_caps;
6033 void *fragment_priv;
6035 if (!(fragment_priv = fragment_pipe->alloc_private(&glsl_shader_backend, priv)))
6037 ERR("Failed to initialize fragment pipe.\n");
6038 HeapFree(GetProcessHeap(), 0, priv);
6042 if (!shader_buffer_init(&priv->shader_buffer))
6044 ERR("Failed to initialize shader buffer.\n");
6048 priv->stack = HeapAlloc(GetProcessHeap(), 0, stack_size * sizeof(*priv->stack));
6051 ERR("Failed to allocate memory.\n");
6055 if (!constant_heap_init(&priv->vconst_heap, gl_info->limits.glsl_vs_float_constants))
6057 ERR("Failed to initialize vertex shader constant heap\n");
6061 if (!constant_heap_init(&priv->pconst_heap, gl_info->limits.glsl_ps_float_constants))
6063 ERR("Failed to initialize pixel shader constant heap\n");
6067 if (wine_rb_init(&priv->program_lookup, &wined3d_glsl_program_rb_functions) == -1)
6069 ERR("Failed to initialize rbtree.\n");
6073 priv->next_constant_version = 1;
6074 fragment_pipe->get_caps(gl_info, &fragment_caps);
6075 priv->ffp_proj_control = fragment_caps.wined3d_caps & WINED3D_FRAGMENT_CAP_PROJ_CONTROL;
6076 device->fragment_priv = fragment_priv;
6077 priv->fragment_pipe = fragment_pipe;
6079 device->shader_priv = priv;
6083 constant_heap_free(&priv->pconst_heap);
6084 constant_heap_free(&priv->vconst_heap);
6085 HeapFree(GetProcessHeap(), 0, priv->stack);
6086 shader_buffer_free(&priv->shader_buffer);
6087 fragment_pipe->free_private(device);
6088 HeapFree(GetProcessHeap(), 0, priv);
6089 return E_OUTOFMEMORY;
6092 /* Context activation is done by the caller. */
6093 static void shader_glsl_free(struct wined3d_device *device)
6095 const struct wined3d_gl_info *gl_info = &device->adapter->gl_info;
6096 struct shader_glsl_priv *priv = device->shader_priv;
6099 for (i = 0; i < tex_type_count; ++i)
6101 if (priv->depth_blt_program_full[i])
6103 GL_EXTCALL(glDeleteObjectARB(priv->depth_blt_program_full[i]));
6105 if (priv->depth_blt_program_masked[i])
6107 GL_EXTCALL(glDeleteObjectARB(priv->depth_blt_program_masked[i]));
6111 wine_rb_destroy(&priv->program_lookup, NULL, NULL);
6112 constant_heap_free(&priv->pconst_heap);
6113 constant_heap_free(&priv->vconst_heap);
6114 HeapFree(GetProcessHeap(), 0, priv->stack);
6115 shader_buffer_free(&priv->shader_buffer);
6116 priv->fragment_pipe->free_private(device);
6118 HeapFree(GetProcessHeap(), 0, device->shader_priv);
6119 device->shader_priv = NULL;
6122 static void shader_glsl_context_destroyed(void *shader_priv, const struct wined3d_context *context) {}
6124 static void shader_glsl_get_caps(const struct wined3d_gl_info *gl_info, struct shader_caps *caps)
6128 if (gl_info->supported[EXT_GPU_SHADER4] && gl_info->supported[ARB_SHADER_BIT_ENCODING]
6129 && gl_info->supported[ARB_GEOMETRY_SHADER4] && gl_info->glsl_version >= MAKEDWORD_VERSION(1, 50)
6130 && gl_info->supported[ARB_DRAW_ELEMENTS_BASE_VERTEX] && gl_info->supported[ARB_DRAW_INSTANCED])
6132 /* ARB_shader_texture_lod or EXT_gpu_shader4 is required for the SM3
6133 * texldd and texldl instructions. */
6134 else if (gl_info->supported[ARB_SHADER_TEXTURE_LOD] || gl_info->supported[EXT_GPU_SHADER4])
6138 TRACE("Shader model %u.\n", shader_model);
6140 caps->vs_version = min(wined3d_settings.max_sm_vs, shader_model);
6141 caps->gs_version = min(wined3d_settings.max_sm_gs, shader_model);
6142 caps->ps_version = min(wined3d_settings.max_sm_ps, shader_model);
6144 caps->vs_uniform_count = gl_info->limits.glsl_vs_float_constants;
6145 caps->ps_uniform_count = gl_info->limits.glsl_ps_float_constants;
6147 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
6148 * Direct3D minimum requirement.
6150 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
6151 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
6153 * The problem is that the refrast clamps temporary results in the shader to
6154 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
6155 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
6156 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
6157 * offer a way to query this.
6159 caps->ps_1x_max_value = 8.0;
6161 /* Ideally we'd only set caps like sRGB writes here if supported by both
6162 * the shader backend and the fragment pipe, but we can get called before
6163 * shader_glsl_alloc(). */
6164 caps->wined3d_caps = WINED3D_SHADER_CAP_VS_CLIPPING
6165 | WINED3D_SHADER_CAP_SRGB_WRITE;
6168 static BOOL shader_glsl_color_fixup_supported(struct color_fixup_desc fixup)
6170 if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
6172 TRACE("Checking support for fixup:\n");
6173 dump_color_fixup_desc(fixup);
6176 /* We support everything except YUV conversions. */
6177 if (!is_complex_fixup(fixup))
6183 TRACE("[FAILED]\n");
6187 static const SHADER_HANDLER shader_glsl_instruction_handler_table[WINED3DSIH_TABLE_SIZE] =
6189 /* WINED3DSIH_ABS */ shader_glsl_map2gl,
6190 /* WINED3DSIH_ADD */ shader_glsl_binop,
6191 /* WINED3DSIH_AND */ shader_glsl_binop,
6192 /* WINED3DSIH_BEM */ shader_glsl_bem,
6193 /* WINED3DSIH_BREAK */ shader_glsl_break,
6194 /* WINED3DSIH_BREAKC */ shader_glsl_breakc,
6195 /* WINED3DSIH_BREAKP */ shader_glsl_breakp,
6196 /* WINED3DSIH_CALL */ shader_glsl_call,
6197 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz,
6198 /* WINED3DSIH_CMP */ shader_glsl_conditional_move,
6199 /* WINED3DSIH_CND */ shader_glsl_cnd,
6200 /* WINED3DSIH_CRS */ shader_glsl_cross,
6201 /* WINED3DSIH_CUT */ shader_glsl_cut,
6202 /* WINED3DSIH_DCL */ shader_glsl_nop,
6203 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_glsl_nop,
6204 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_glsl_nop,
6205 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_glsl_nop,
6206 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_glsl_nop,
6207 /* WINED3DSIH_DEF */ shader_glsl_nop,
6208 /* WINED3DSIH_DEFB */ shader_glsl_nop,
6209 /* WINED3DSIH_DEFI */ shader_glsl_nop,
6210 /* WINED3DSIH_DIV */ shader_glsl_binop,
6211 /* WINED3DSIH_DP2ADD */ shader_glsl_dp2add,
6212 /* WINED3DSIH_DP3 */ shader_glsl_dot,
6213 /* WINED3DSIH_DP4 */ shader_glsl_dot,
6214 /* WINED3DSIH_DST */ shader_glsl_dst,
6215 /* WINED3DSIH_DSX */ shader_glsl_map2gl,
6216 /* WINED3DSIH_DSY */ shader_glsl_map2gl,
6217 /* WINED3DSIH_ELSE */ shader_glsl_else,
6218 /* WINED3DSIH_EMIT */ shader_glsl_emit,
6219 /* WINED3DSIH_ENDIF */ shader_glsl_end,
6220 /* WINED3DSIH_ENDLOOP */ shader_glsl_end,
6221 /* WINED3DSIH_ENDREP */ shader_glsl_end,
6222 /* WINED3DSIH_EQ */ shader_glsl_relop,
6223 /* WINED3DSIH_EXP */ shader_glsl_map2gl,
6224 /* WINED3DSIH_EXPP */ shader_glsl_expp,
6225 /* WINED3DSIH_FRC */ shader_glsl_map2gl,
6226 /* WINED3DSIH_FTOI */ shader_glsl_to_int,
6227 /* WINED3DSIH_GE */ shader_glsl_relop,
6228 /* WINED3DSIH_IADD */ shader_glsl_binop,
6229 /* WINED3DSIH_IEQ */ NULL,
6230 /* WINED3DSIH_IF */ shader_glsl_if,
6231 /* WINED3DSIH_IFC */ shader_glsl_ifc,
6232 /* WINED3DSIH_IGE */ shader_glsl_relop,
6233 /* WINED3DSIH_IMUL */ shader_glsl_imul,
6234 /* WINED3DSIH_ITOF */ shader_glsl_to_float,
6235 /* WINED3DSIH_LABEL */ shader_glsl_label,
6236 /* WINED3DSIH_LD */ NULL,
6237 /* WINED3DSIH_LIT */ shader_glsl_lit,
6238 /* WINED3DSIH_LOG */ shader_glsl_log,
6239 /* WINED3DSIH_LOGP */ shader_glsl_log,
6240 /* WINED3DSIH_LOOP */ shader_glsl_loop,
6241 /* WINED3DSIH_LRP */ shader_glsl_lrp,
6242 /* WINED3DSIH_LT */ shader_glsl_relop,
6243 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn,
6244 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn,
6245 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn,
6246 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn,
6247 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn,
6248 /* WINED3DSIH_MAD */ shader_glsl_mad,
6249 /* WINED3DSIH_MAX */ shader_glsl_map2gl,
6250 /* WINED3DSIH_MIN */ shader_glsl_map2gl,
6251 /* WINED3DSIH_MOV */ shader_glsl_mov,
6252 /* WINED3DSIH_MOVA */ shader_glsl_mov,
6253 /* WINED3DSIH_MOVC */ shader_glsl_conditional_move,
6254 /* WINED3DSIH_MUL */ shader_glsl_binop,
6255 /* WINED3DSIH_NOP */ shader_glsl_nop,
6256 /* WINED3DSIH_NRM */ shader_glsl_nrm,
6257 /* WINED3DSIH_PHASE */ shader_glsl_nop,
6258 /* WINED3DSIH_POW */ shader_glsl_pow,
6259 /* WINED3DSIH_RCP */ shader_glsl_rcp,
6260 /* WINED3DSIH_REP */ shader_glsl_rep,
6261 /* WINED3DSIH_RET */ shader_glsl_ret,
6262 /* WINED3DSIH_ROUND_NI */ shader_glsl_map2gl,
6263 /* WINED3DSIH_RSQ */ shader_glsl_rsq,
6264 /* WINED3DSIH_SAMPLE */ NULL,
6265 /* WINED3DSIH_SAMPLE_GRAD */ NULL,
6266 /* WINED3DSIH_SAMPLE_LOD */ NULL,
6267 /* WINED3DSIH_SETP */ NULL,
6268 /* WINED3DSIH_SGE */ shader_glsl_compare,
6269 /* WINED3DSIH_SGN */ shader_glsl_sgn,
6270 /* WINED3DSIH_SINCOS */ shader_glsl_sincos,
6271 /* WINED3DSIH_SLT */ shader_glsl_compare,
6272 /* WINED3DSIH_SQRT */ NULL,
6273 /* WINED3DSIH_SUB */ shader_glsl_binop,
6274 /* WINED3DSIH_TEX */ shader_glsl_tex,
6275 /* WINED3DSIH_TEXBEM */ shader_glsl_texbem,
6276 /* WINED3DSIH_TEXBEML */ shader_glsl_texbem,
6277 /* WINED3DSIH_TEXCOORD */ shader_glsl_texcoord,
6278 /* WINED3DSIH_TEXDEPTH */ shader_glsl_texdepth,
6279 /* WINED3DSIH_TEXDP3 */ shader_glsl_texdp3,
6280 /* WINED3DSIH_TEXDP3TEX */ shader_glsl_texdp3tex,
6281 /* WINED3DSIH_TEXKILL */ shader_glsl_texkill,
6282 /* WINED3DSIH_TEXLDD */ shader_glsl_texldd,
6283 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl,
6284 /* WINED3DSIH_TEXM3x2DEPTH */ shader_glsl_texm3x2depth,
6285 /* WINED3DSIH_TEXM3x2PAD */ shader_glsl_texm3x2pad,
6286 /* WINED3DSIH_TEXM3x2TEX */ shader_glsl_texm3x2tex,
6287 /* WINED3DSIH_TEXM3x3 */ shader_glsl_texm3x3,
6288 /* WINED3DSIH_TEXM3x3DIFF */ NULL,
6289 /* WINED3DSIH_TEXM3x3PAD */ shader_glsl_texm3x3pad,
6290 /* WINED3DSIH_TEXM3x3SPEC */ shader_glsl_texm3x3spec,
6291 /* WINED3DSIH_TEXM3x3TEX */ shader_glsl_texm3x3tex,
6292 /* WINED3DSIH_TEXM3x3VSPEC */ shader_glsl_texm3x3vspec,
6293 /* WINED3DSIH_TEXREG2AR */ shader_glsl_texreg2ar,
6294 /* WINED3DSIH_TEXREG2GB */ shader_glsl_texreg2gb,
6295 /* WINED3DSIH_TEXREG2RGB */ shader_glsl_texreg2rgb,
6296 /* WINED3DSIH_UDIV */ shader_glsl_udiv,
6297 /* WINED3DSIH_USHR */ shader_glsl_binop,
6298 /* WINED3DSIH_UTOF */ shader_glsl_to_float,
6299 /* WINED3DSIH_XOR */ shader_glsl_binop,
6302 static void shader_glsl_handle_instruction(const struct wined3d_shader_instruction *ins) {
6303 SHADER_HANDLER hw_fct;
6305 /* Select handler */
6306 hw_fct = shader_glsl_instruction_handler_table[ins->handler_idx];
6308 /* Unhandled opcode */
6311 FIXME("Backend can't handle opcode %#x\n", ins->handler_idx);
6316 shader_glsl_add_instruction_modifiers(ins);
6319 static BOOL shader_glsl_has_ffp_proj_control(void *shader_priv)
6321 struct shader_glsl_priv *priv = shader_priv;
6323 return priv->ffp_proj_control;
6326 const struct wined3d_shader_backend_ops glsl_shader_backend =
6328 shader_glsl_handle_instruction,
6330 shader_glsl_select_depth_blt,
6331 shader_glsl_deselect_depth_blt,
6332 shader_glsl_update_float_vertex_constants,
6333 shader_glsl_update_float_pixel_constants,
6334 shader_glsl_load_constants,
6335 shader_glsl_load_np2fixup_constants,
6336 shader_glsl_destroy,
6339 shader_glsl_context_destroyed,
6340 shader_glsl_get_caps,
6341 shader_glsl_color_fixup_supported,
6342 shader_glsl_has_ffp_proj_control,
6345 static void glsl_fragment_pipe_enable(const struct wined3d_gl_info *gl_info, BOOL enable)
6347 /* Nothing to do. */
6350 static void glsl_fragment_pipe_get_caps(const struct wined3d_gl_info *gl_info, struct fragment_caps *caps)
6352 caps->wined3d_caps = WINED3D_FRAGMENT_CAP_PROJ_CONTROL
6353 | WINED3D_FRAGMENT_CAP_SRGB_WRITE;
6354 caps->PrimitiveMiscCaps = WINED3DPMISCCAPS_TSSARGTEMP;
6355 caps->TextureOpCaps = WINED3DTEXOPCAPS_DISABLE
6356 | WINED3DTEXOPCAPS_SELECTARG1
6357 | WINED3DTEXOPCAPS_SELECTARG2
6358 | WINED3DTEXOPCAPS_MODULATE4X
6359 | WINED3DTEXOPCAPS_MODULATE2X
6360 | WINED3DTEXOPCAPS_MODULATE
6361 | WINED3DTEXOPCAPS_ADDSIGNED2X
6362 | WINED3DTEXOPCAPS_ADDSIGNED
6363 | WINED3DTEXOPCAPS_ADD
6364 | WINED3DTEXOPCAPS_SUBTRACT
6365 | WINED3DTEXOPCAPS_ADDSMOOTH
6366 | WINED3DTEXOPCAPS_BLENDCURRENTALPHA
6367 | WINED3DTEXOPCAPS_BLENDFACTORALPHA
6368 | WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
6369 | WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
6370 | WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
6371 | WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
6372 | WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
6373 | WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
6374 | WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
6375 | WINED3DTEXOPCAPS_DOTPRODUCT3
6376 | WINED3DTEXOPCAPS_MULTIPLYADD
6377 | WINED3DTEXOPCAPS_LERP
6378 | WINED3DTEXOPCAPS_BUMPENVMAP
6379 | WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE;
6380 caps->MaxTextureBlendStages = 8;
6381 caps->MaxSimultaneousTextures = min(gl_info->limits.fragment_samplers, 8);
6384 static void *glsl_fragment_pipe_alloc(const struct wined3d_shader_backend_ops *shader_backend, void *shader_priv)
6386 struct shader_glsl_priv *priv;
6388 if (shader_backend == &glsl_shader_backend)
6392 if (wine_rb_init(&priv->ffp_fragment_shaders, &wined3d_ffp_frag_program_rb_functions) == -1)
6394 ERR("Failed to initialize rbtree.\n");
6401 FIXME("GLSL fragment pipe without GLSL shader backend not implemented.\n");
6406 struct glsl_ffp_destroy_ctx
6408 struct shader_glsl_priv *priv;
6409 const struct wined3d_gl_info *gl_info;
6412 static void shader_glsl_free_ffp_fragment_shader(struct wine_rb_entry *entry, void *context)
6414 struct glsl_ffp_fragment_shader *shader = WINE_RB_ENTRY_VALUE(entry,
6415 struct glsl_ffp_fragment_shader, entry.entry);
6416 struct glsl_shader_prog_link *program, *program2;
6417 struct glsl_ffp_destroy_ctx *ctx = context;
6419 LIST_FOR_EACH_ENTRY_SAFE(program, program2, &shader->linked_programs,
6420 struct glsl_shader_prog_link, ps.shader_entry)
6422 delete_glsl_program_entry(ctx->priv, ctx->gl_info, program);
6424 ctx->gl_info->gl_ops.ext.p_glDeleteObjectARB(shader->id);
6425 HeapFree(GetProcessHeap(), 0, shader);
6428 /* Context activation is done by the caller. */
6429 static void glsl_fragment_pipe_free(struct wined3d_device *device)
6431 struct shader_glsl_priv *priv = device->fragment_priv;
6432 struct glsl_ffp_destroy_ctx ctx;
6435 ctx.gl_info = &device->adapter->gl_info;
6436 wine_rb_destroy(&priv->ffp_fragment_shaders, shader_glsl_free_ffp_fragment_shader, &ctx);
6439 static void glsl_fragment_pipe_shader(struct wined3d_context *context,
6440 const struct wined3d_state *state, DWORD state_id)
6442 context->last_was_pshader = use_ps(state);
6444 context->select_shader = 1;
6445 context->load_constants = 1;
6448 static void glsl_fragment_pipe_fog(struct wined3d_context *context,
6449 const struct wined3d_state *state, DWORD state_id)
6451 BOOL use_vshader = use_vs(state);
6452 enum fogsource new_source;
6454 context->select_shader = 1;
6455 context->load_constants = 1;
6457 if (!state->render_states[WINED3D_RS_FOGENABLE])
6460 if (state->render_states[WINED3D_RS_FOGTABLEMODE] == WINED3D_FOG_NONE)
6463 new_source = FOGSOURCE_VS;
6464 else if (state->render_states[WINED3D_RS_FOGVERTEXMODE] == WINED3D_FOG_NONE || context->last_was_rhw)
6465 new_source = FOGSOURCE_COORD;
6467 new_source = FOGSOURCE_FFP;
6471 new_source = FOGSOURCE_FFP;
6474 if (new_source != context->fog_source)
6476 context->fog_source = new_source;
6477 state_fogstartend(context, state, STATE_RENDER(WINED3D_RS_FOGSTART));
6481 static void glsl_fragment_pipe_tex_transform(struct wined3d_context *context,
6482 const struct wined3d_state *state, DWORD state_id)
6484 context->select_shader = 1;
6485 context->load_constants = 1;
6488 static void glsl_fragment_pipe_invalidate_constants(struct wined3d_context *context,
6489 const struct wined3d_state *state, DWORD state_id)
6491 context->load_constants = 1;
6494 static const struct StateEntryTemplate glsl_fragment_pipe_state_template[] =
6496 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR), {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6497 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6498 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6499 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6500 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6501 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6502 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6503 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6504 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6505 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6506 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6507 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6508 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6509 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6510 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6511 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6512 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6513 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6514 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6515 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6516 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6517 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6518 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6519 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6520 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6521 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6522 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6523 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6524 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6525 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6526 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6527 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6528 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6529 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6530 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6531 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6532 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6533 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6534 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6535 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6536 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6537 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6538 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6539 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6540 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6541 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6542 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6543 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6544 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6545 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6546 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6547 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6548 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6549 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6550 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6551 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6552 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6553 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6554 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6555 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6556 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6557 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6558 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6559 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6560 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6561 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6562 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6563 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6564 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6565 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6566 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6567 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6568 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6569 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6570 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6571 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6572 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6573 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6574 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6575 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6576 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6577 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6578 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6579 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6580 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6581 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6582 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6583 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6584 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6585 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6586 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6587 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6588 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6589 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6590 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6591 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6592 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6593 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6594 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6595 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6596 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6597 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6598 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6599 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6600 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6601 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6602 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6603 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6604 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6605 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6606 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6607 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6608 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6609 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6610 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6611 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6612 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6613 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6614 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00), NULL }, WINED3D_GL_EXT_NONE },
6615 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6616 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE), NULL }, WINED3D_GL_EXT_NONE },
6617 {STATE_PIXELSHADER, {STATE_PIXELSHADER, glsl_fragment_pipe_shader }, WINED3D_GL_EXT_NONE },
6618 {STATE_RENDER(WINED3D_RS_FOGENABLE), {STATE_RENDER(WINED3D_RS_FOGENABLE), glsl_fragment_pipe_fog }, WINED3D_GL_EXT_NONE },
6619 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE), {STATE_RENDER(WINED3D_RS_FOGENABLE), NULL }, WINED3D_GL_EXT_NONE },
6620 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE), {STATE_RENDER(WINED3D_RS_FOGENABLE), NULL }, WINED3D_GL_EXT_NONE },
6621 {STATE_RENDER(WINED3D_RS_FOGSTART), {STATE_RENDER(WINED3D_RS_FOGSTART), state_fogstartend }, WINED3D_GL_EXT_NONE },
6622 {STATE_RENDER(WINED3D_RS_FOGEND), {STATE_RENDER(WINED3D_RS_FOGSTART), NULL }, WINED3D_GL_EXT_NONE },
6623 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE), {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE), state_srgbwrite }, ARB_FRAMEBUFFER_SRGB},
6624 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE), {STATE_PIXELSHADER, NULL }, WINED3D_GL_EXT_NONE },
6625 {STATE_RENDER(WINED3D_RS_FOGCOLOR), {STATE_RENDER(WINED3D_RS_FOGCOLOR), state_fogcolor }, WINED3D_GL_EXT_NONE },
6626 {STATE_RENDER(WINED3D_RS_FOGDENSITY), {STATE_RENDER(WINED3D_RS_FOGDENSITY), state_fogdensity }, WINED3D_GL_EXT_NONE },
6627 {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 },
6628 {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 },
6629 {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 },
6630 {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 },
6631 {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 },
6632 {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 },
6633 {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 },
6634 {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 },
6635 {STATE_RENDER(WINED3D_RS_SPECULARENABLE), {STATE_RENDER(WINED3D_RS_SPECULARENABLE), glsl_fragment_pipe_invalidate_constants}, WINED3D_GL_EXT_NONE },
6636 {0 /* Terminate */, {0, 0 }, WINED3D_GL_EXT_NONE },
6639 const struct fragment_pipeline glsl_fragment_pipe =
6641 glsl_fragment_pipe_enable,
6642 glsl_fragment_pipe_get_caps,
6643 glsl_fragment_pipe_alloc,
6644 glsl_fragment_pipe_free,
6645 shader_glsl_color_fixup_supported,
6646 glsl_fragment_pipe_state_template,