2 * Pixel and vertex shaders implementation using ARB_vertex_program
3 * and ARB_fragment_program GL extensions.
5 * Copyright 2002-2003 Jason Edmeades
6 * Copyright 2002-2003 Raphael Junqueira
7 * Copyright 2004 Christian Costa
8 * Copyright 2005 Oliver Stieber
9 * Copyright 2006 Ivan Gyurdiev
10 * Copyright 2006 Jason Green
11 * Copyright 2006 Henri Verbeet
12 * Copyright 2007-2011, 2013-2014 Stefan Dösinger for CodeWeavers
13 * Copyright 2009 Henri Verbeet for CodeWeavers
15 * This library is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU Lesser General Public
17 * License as published by the Free Software Foundation; either
18 * version 2.1 of the License, or (at your option) any later version.
20 * This library is distributed in the hope that it will be useful,
21 * but WITHOUT ANY WARRANTY; without even the implied warranty of
22 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
23 * Lesser General Public License for more details.
25 * You should have received a copy of the GNU Lesser General Public
26 * License along with this library; if not, write to the Free Software
27 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
31 #include "wine/port.h"
35 #include "wined3d_private.h"
37 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d_perf
);
42 static BOOL
shader_is_pshader_version(enum wined3d_shader_type type
)
44 return type
== WINED3D_SHADER_TYPE_PIXEL
;
47 static BOOL
shader_is_vshader_version(enum wined3d_shader_type type
)
49 return type
== WINED3D_SHADER_TYPE_VERTEX
;
52 static const char *get_line(const char **ptr
)
57 if (!(q
= strstr(p
, "\n")))
79 static const char *arb_get_helper_value(enum wined3d_shader_type shader
, enum arb_helper_value value
)
81 if (shader
!= WINED3D_SHADER_TYPE_VERTEX
&& shader
!= WINED3D_SHADER_TYPE_PIXEL
)
83 ERR("Unsupported shader type '%s'.\n", debug_shader_type(shader
));
87 if (shader
== WINED3D_SHADER_TYPE_PIXEL
)
91 case ARB_ZERO
: return "ps_helper_const.x";
92 case ARB_ONE
: return "ps_helper_const.y";
93 case ARB_TWO
: return "coefmul.x";
94 case ARB_0001
: return "ps_helper_const.xxxy";
95 case ARB_EPS
: return "ps_helper_const.z";
103 case ARB_ZERO
: return "helper_const.x";
104 case ARB_ONE
: return "helper_const.y";
105 case ARB_TWO
: return "helper_const.z";
106 case ARB_EPS
: return "helper_const.w";
107 case ARB_0001
: return "helper_const.xxxy";
108 case ARB_VS_REL_OFFSET
: return "rel_addr_const.y";
111 FIXME("Unmanaged %s shader helper constant requested: %u.\n",
112 shader
== WINED3D_SHADER_TYPE_PIXEL
? "pixel" : "vertex", value
);
115 case ARB_ZERO
: return "0.0";
116 case ARB_ONE
: return "1.0";
117 case ARB_TWO
: return "2.0";
118 case ARB_0001
: return "{0.0, 0.0, 0.0, 1.0}";
119 case ARB_EPS
: return "1e-8";
120 default: return "bad";
124 static inline BOOL
ffp_clip_emul(const struct wined3d_context
*context
)
126 return context
->lowest_disabled_stage
< 7;
129 /* ARB_program_shader private data */
148 struct wined3d_shader_loop_control loop_control
;
152 struct arb_ps_np2fixup_info
154 struct ps_np2fixup_info super
;
155 /* For ARB we need an offset value:
156 * With both GLSL and ARB mode the NP2 fixup information (the texture dimensions) are stored in a
157 * consecutive way (GLSL uses a uniform array). Since ARB doesn't know the notion of a "standalone"
158 * array we need an offset to the index inside the program local parameter array. */
162 struct arb_ps_compile_args
164 struct ps_compile_args super
;
166 WORD clip
; /* only a boolean, use a WORD for alignment */
167 unsigned char loop_ctrl
[WINED3D_MAX_CONSTS_I
][3];
170 struct stb_const_desc
172 unsigned char texunit
;
176 struct arb_ps_compiled_shader
178 struct arb_ps_compile_args args
;
179 struct arb_ps_np2fixup_info np2fixup_info
;
180 struct stb_const_desc bumpenvmatconst
[MAX_TEXTURES
];
181 struct stb_const_desc luminanceconst
[MAX_TEXTURES
];
182 UINT int_consts
[WINED3D_MAX_CONSTS_I
];
185 unsigned char numbumpenvmatconsts
;
189 struct arb_vs_compile_args
191 struct vs_compile_args super
;
197 unsigned char clip_texcoord
;
198 unsigned char clipplane_mask
;
200 DWORD boolclip_compare
;
205 unsigned char samplers
[4];
206 DWORD samplers_compare
;
208 unsigned char loop_ctrl
[WINED3D_MAX_CONSTS_I
][3];
211 struct arb_vs_compiled_shader
213 struct arb_vs_compile_args args
;
215 UINT int_consts
[WINED3D_MAX_CONSTS_I
];
217 char need_color_unclamp
;
221 struct recorded_instruction
223 struct wined3d_shader_instruction ins
;
227 struct shader_arb_ctx_priv
232 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
234 /* GL_NV_vertex_program2_option or GL_NV_fragment_program_option */
236 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
240 const struct arb_vs_compile_args
*cur_vs_args
;
241 const struct arb_ps_compile_args
*cur_ps_args
;
242 const struct arb_ps_compiled_shader
*compiled_fprog
;
243 const struct arb_vs_compiled_shader
*compiled_vprog
;
244 struct arb_ps_np2fixup_info
*cur_np2fixup_info
;
245 struct list control_frames
;
249 unsigned int num_loops
, loop_depth
, num_ifcs
;
251 BOOL ps_post_process
;
253 unsigned int vs_clipplanes
;
257 /* For 3.0 vertex shaders */
258 const char *vs_output
[MAX_REG_OUTPUT
];
259 /* For 2.x and earlier vertex shaders */
260 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
262 /* 3.0 pshader input for compatibility with fixed function */
263 const char *ps_input
[MAX_REG_INPUT
];
268 struct wined3d_shader_signature sig
;
270 struct wine_rb_entry entry
;
273 struct arb_pshader_private
{
274 struct arb_ps_compiled_shader
*gl_shaders
;
275 UINT num_gl_shaders
, shader_array_size
;
276 DWORD input_signature_idx
;
277 DWORD clipplane_emulation
;
281 struct arb_vshader_private
{
282 struct arb_vs_compiled_shader
*gl_shaders
;
283 UINT num_gl_shaders
, shader_array_size
;
287 struct shader_arb_priv
289 GLuint current_vprogram_id
;
290 GLuint current_fprogram_id
;
291 const struct arb_ps_compiled_shader
*compiled_fprog
;
292 const struct arb_vs_compiled_shader
*compiled_vprog
;
293 BOOL use_arbfp_fixed_func
;
294 struct wine_rb_tree fragment_shaders
;
295 BOOL last_ps_const_clamped
;
296 BOOL last_vs_color_unclamp
;
298 struct wine_rb_tree signature_tree
;
301 unsigned int highest_dirty_ps_const
, highest_dirty_vs_const
;
302 char vshader_const_dirty
[WINED3D_MAX_VS_CONSTS_F
];
303 char pshader_const_dirty
[WINED3D_MAX_PS_CONSTS_F
];
304 const struct wined3d_context
*last_context
;
306 const struct wined3d_vertex_pipe_ops
*vertex_pipe
;
307 const struct fragment_pipeline
*fragment_pipe
;
308 BOOL ffp_proj_control
;
311 /* Context activation for state handlers is done by the caller. */
313 static BOOL
need_rel_addr_const(const struct arb_vshader_private
*shader_data
,
314 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
316 if (shader_data
->rel_offset
) return TRUE
;
317 if (!reg_maps
->usesmova
) return FALSE
;
318 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
321 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
322 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
324 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
325 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
328 static BOOL
need_helper_const(const struct arb_vshader_private
*shader_data
,
329 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
331 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) return TRUE
;
332 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
333 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
334 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) return TRUE
; /* Have to init texcoords. */
335 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
336 if (reg_maps
->usesnrm
) return TRUE
; /* 0.0 */
337 if (reg_maps
->usespow
) return TRUE
; /* EPS, 0.0 and 1.0 */
338 if (reg_maps
->fog
) return TRUE
; /* Clamping fog coord, 0.0 and 1.0 */
342 static unsigned int reserved_vs_const(const struct arb_vshader_private
*shader_data
,
343 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
345 unsigned int ret
= 1;
346 /* We use one PARAM for the pos fixup, and in some cases one to load
347 * some immediate values into the shader. */
348 if (need_helper_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
349 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
353 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
354 * When constant_list == NULL, it will load all the constants.
356 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
357 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
359 /* Context activation is done by the caller. */
360 static unsigned int shader_arb_load_constants_f(const struct wined3d_shader
*shader
,
361 const struct wined3d_gl_info
*gl_info
, GLuint target_type
, unsigned int max_constants
,
362 const struct wined3d_vec4
*constants
, char *dirty_consts
)
364 struct wined3d_shader_lconst
*lconst
;
365 unsigned int ret
, i
, j
;
367 if (TRACE_ON(d3d_constants
))
369 for (i
= 0; i
< max_constants
; ++i
)
371 if (!dirty_consts
[i
])
373 TRACE_(d3d_constants
)("Loading constant %u: %s.\n", i
, debug_vec4(&constants
[i
]));
379 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
380 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& shader
->reg_maps
.shader_version
.major
== 1)
383 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
384 * shaders, the first 8 constants are marked dirty for reload
386 for (; i
< min(8, max_constants
); ++i
)
388 if (!dirty_consts
[i
])
392 if (constants
[i
].x
> 1.0f
)
394 else if (constants
[i
].x
< -1.0f
)
395 lcl_const
[0] = -1.0f
;
397 lcl_const
[0] = constants
[i
].x
;
399 if (constants
[i
].y
> 1.0f
)
401 else if (constants
[i
].y
< -1.0f
)
402 lcl_const
[1] = -1.0f
;
404 lcl_const
[1] = constants
[i
].y
;
406 if (constants
[i
].z
> 1.0f
)
408 else if (constants
[i
].z
< -1.0f
)
409 lcl_const
[2] = -1.0f
;
411 lcl_const
[2] = constants
[i
].z
;
413 if (constants
[i
].w
> 1.0f
)
415 else if (constants
[i
].w
< -1.0f
)
416 lcl_const
[3] = -1.0f
;
418 lcl_const
[3] = constants
[i
].w
;
420 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
423 /* If further constants are dirty, reload them without clamping.
425 * The alternative is not to touch them, but then we cannot reset the dirty constant count
426 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
427 * above would always re-check the first 8 constants since max_constant remains at the init
432 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
434 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
435 * or just reloading *all* constants at once
437 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
439 for (; i
< max_constants
; ++i
)
441 if (!dirty_consts
[i
])
444 /* Find the next block of dirty constants */
447 for (++i
; (i
< max_constants
) && dirty_consts
[i
]; ++i
)
452 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, &constants
[j
].x
));
457 for (; i
< max_constants
; ++i
)
462 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, &constants
[i
].x
));
466 checkGLcall("glProgramEnvParameter4fvARB()");
468 /* Load immediate constants */
469 if (shader
->load_local_constsF
)
471 if (TRACE_ON(d3d_shader
))
473 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
475 GLfloat
* values
= (GLfloat
*)lconst
->value
;
476 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
477 values
[0], values
[1], values
[2], values
[3]);
480 /* Immediate constants are clamped for 1.X shaders at loading times */
482 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
484 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
485 ret
= max(ret
, lconst
->idx
+ 1);
486 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
488 checkGLcall("glProgramEnvParameter4fvARB()");
489 return ret
; /* The loaded immediate constants need reloading for the next shader */
491 return 0; /* No constants are dirty now */
495 /* Loads the texture dimensions for NP2 fixup into the currently set
496 * ARB_[vertex/fragment]_programs. */
497 static void shader_arb_load_np2fixup_constants(const struct arb_ps_np2fixup_info
*fixup
,
498 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
500 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
501 WORD active
= fixup
->super
.active
;
507 for (i
= 0; active
; active
>>= 1, ++i
)
509 const struct wined3d_texture
*tex
= state
->textures
[i
];
510 unsigned char idx
= fixup
->super
.idx
[i
];
511 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
518 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
524 tex_dim
[2] = tex
->pow2_matrix
[0];
525 tex_dim
[3] = tex
->pow2_matrix
[5];
529 tex_dim
[0] = tex
->pow2_matrix
[0];
530 tex_dim
[1] = tex
->pow2_matrix
[5];
534 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
)
536 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
537 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
541 /* Context activation is done by the caller. */
542 static void shader_arb_ps_local_constants(const struct arb_ps_compiled_shader
*gl_shader
,
543 const struct wined3d_context
*context
, const struct wined3d_state
*state
, UINT rt_height
)
545 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
548 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
550 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
552 /* The state manager takes care that this function is always called if the bump env matrix changes */
553 const float *data
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_MAT00
];
554 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
555 gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
557 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
559 /* WINED3D_TSS_BUMPENVLSCALE and WINED3D_TSS_BUMPENVLOFFSET are next to each other.
560 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
561 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
562 * (they're WINED3D_TSS_TEXTURETRANSFORMFLAGS and WINED3D_TSS_ADDRESSW, so most likely 0 or NaN
564 const float *scale
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_LSCALE
];
565 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
566 gl_shader
->luminanceconst
[i
].const_num
, scale
));
569 checkGLcall("Load bumpmap consts");
571 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
573 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
574 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
579 val
[0] = context
->render_offscreen
? 0.0f
: (float) rt_height
;
580 val
[1] = context
->render_offscreen
? 1.0f
: -1.0f
;
583 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
584 checkGLcall("y correction loading");
587 if (!gl_shader
->num_int_consts
) return;
589 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
591 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
594 val
[0] = (float)state
->ps_consts_i
[i
].x
;
595 val
[1] = (float)state
->ps_consts_i
[i
].y
;
596 val
[2] = (float)state
->ps_consts_i
[i
].z
;
599 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
602 checkGLcall("Load ps int consts");
605 /* Context activation is done by the caller. */
606 static void shader_arb_vs_local_constants(const struct arb_vs_compiled_shader
*gl_shader
,
607 const struct wined3d_context
*context
, const struct wined3d_state
*state
)
609 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
610 float position_fixup
[4];
613 /* Upload the position fixup */
614 shader_get_position_fixup(context
, state
, 1, position_fixup
);
615 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, position_fixup
));
617 if (!gl_shader
->num_int_consts
) return;
619 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
621 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
624 val
[0] = (float)state
->vs_consts_i
[i
].x
;
625 val
[1] = (float)state
->vs_consts_i
[i
].y
;
626 val
[2] = (float)state
->vs_consts_i
[i
].z
;
629 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
632 checkGLcall("Load vs int consts");
635 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
636 const struct wined3d_state
*state
);
639 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
641 * We only support float constants in ARB at the moment, so don't
642 * worry about the Integers or Booleans
644 /* Context activation is done by the caller (state handler). */
645 static void shader_arb_load_constants_internal(struct shader_arb_priv
*priv
,
646 struct wined3d_context
*context
, const struct wined3d_state
*state
,
647 BOOL usePixelShader
, BOOL useVertexShader
, BOOL from_shader_select
)
649 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
650 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
652 if (!from_shader_select
)
654 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
655 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
658 && (vshader
->reg_maps
.boolean_constants
659 || (!gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
660 && (vshader
->reg_maps
.integer_constants
& ~vshader
->reg_maps
.local_int_consts
))))
662 TRACE("bool/integer vertex shader constants potentially modified, forcing shader reselection.\n");
663 shader_arb_select(priv
, context
, state
);
666 && (pshader
->reg_maps
.boolean_constants
667 || (!gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
]
668 && (pshader
->reg_maps
.integer_constants
& ~pshader
->reg_maps
.local_int_consts
))))
670 TRACE("bool/integer pixel shader constants potentially modified, forcing shader reselection.\n");
671 shader_arb_select(priv
, context
, state
);
675 if (context
!= priv
->last_context
)
677 memset(priv
->vshader_const_dirty
, 1,
678 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
679 priv
->highest_dirty_vs_const
= d3d_info
->limits
.vs_uniform_count
;
681 memset(priv
->pshader_const_dirty
, 1,
682 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
683 priv
->highest_dirty_ps_const
= d3d_info
->limits
.ps_uniform_count
;
685 priv
->last_context
= context
;
690 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
691 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
693 /* Load DirectX 9 float constants for vertex shader */
694 priv
->highest_dirty_vs_const
= shader_arb_load_constants_f(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
695 priv
->highest_dirty_vs_const
, state
->vs_consts_f
, priv
->vshader_const_dirty
);
696 shader_arb_vs_local_constants(gl_shader
, context
, state
);
701 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
702 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
703 UINT rt_height
= state
->fb
->render_targets
[0]->height
;
705 /* Load DirectX 9 float constants for pixel shader */
706 priv
->highest_dirty_ps_const
= shader_arb_load_constants_f(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
707 priv
->highest_dirty_ps_const
, state
->ps_consts_f
, priv
->pshader_const_dirty
);
708 shader_arb_ps_local_constants(gl_shader
, context
, state
, rt_height
);
710 if (context
->constant_update_mask
& WINED3D_SHADER_CONST_PS_NP2_FIXUP
)
711 shader_arb_load_np2fixup_constants(&gl_shader
->np2fixup_info
, gl_info
, state
);
715 static void shader_arb_load_constants(void *shader_priv
, struct wined3d_context
*context
,
716 const struct wined3d_state
*state
)
718 BOOL vs
= use_vs(state
);
719 BOOL ps
= use_ps(state
);
721 shader_arb_load_constants_internal(shader_priv
, context
, state
, ps
, vs
, FALSE
);
724 static void shader_arb_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
726 struct wined3d_context
*context
= context_get_current();
727 struct shader_arb_priv
*priv
= device
->shader_priv
;
729 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
730 * context. On a context switch the old context will be fully dirtified */
731 if (!context
|| context
->device
!= device
)
734 memset(priv
->vshader_const_dirty
+ start
, 1, sizeof(*priv
->vshader_const_dirty
) * count
);
735 priv
->highest_dirty_vs_const
= max(priv
->highest_dirty_vs_const
, start
+ count
);
738 static void shader_arb_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
740 struct wined3d_context
*context
= context_get_current();
741 struct shader_arb_priv
*priv
= device
->shader_priv
;
743 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
744 * context. On a context switch the old context will be fully dirtified */
745 if (!context
|| context
->device
!= device
)
748 memset(priv
->pshader_const_dirty
+ start
, 1, sizeof(*priv
->pshader_const_dirty
) * count
);
749 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, start
+ count
);
752 static void shader_arb_append_imm_vec4(struct wined3d_string_buffer
*buffer
, const float *values
)
756 wined3d_ftoa(values
[0], str
[0]);
757 wined3d_ftoa(values
[1], str
[1]);
758 wined3d_ftoa(values
[2], str
[2]);
759 wined3d_ftoa(values
[3], str
[3]);
760 shader_addline(buffer
, "{%s, %s, %s, %s}", str
[0], str
[1], str
[2], str
[3]);
763 /* Generate the variable & register declarations for the ARB_vertex_program output target */
764 static void shader_generate_arb_declarations(const struct wined3d_shader
*shader
,
765 const struct wined3d_shader_reg_maps
*reg_maps
, struct wined3d_string_buffer
*buffer
,
766 const struct wined3d_gl_info
*gl_info
, DWORD
*num_clipplanes
,
767 const struct shader_arb_ctx_priv
*ctx
)
770 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
771 const struct wined3d_shader_lconst
*lconst
;
772 unsigned max_constantsF
;
775 /* In pixel shaders, all private constants are program local, we don't need anything
776 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
777 * If we need a private constant the GL implementation will squeeze it in somewhere
779 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
780 * immediate values. The posFixup is loaded using program.env for now, so always
781 * subtract one from the number of constants. If the shader uses indirect addressing,
782 * account for the helper const too because we have to declare all available d3d constants
783 * and don't know which are actually used.
787 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
788 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
789 if (max_constantsF
< 24)
790 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
794 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
795 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
796 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
797 * Also prevents max_constantsF from becoming less than 0 and
799 if (max_constantsF
< 96)
800 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
802 if (reg_maps
->usesrelconstF
)
804 DWORD highest_constf
= 0, clip_limit
;
806 max_constantsF
-= reserved_vs_const(shader_data
, reg_maps
, gl_info
);
807 max_constantsF
-= wined3d_popcount(reg_maps
->integer_constants
);
808 max_constantsF
-= gl_info
->reserved_arb_constants
;
810 for (i
= 0; i
< shader
->limits
->constant_float
; ++i
)
813 DWORD shift
= i
& 0x1f;
814 if (reg_maps
->constf
[idx
] & (1u << shift
))
818 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
820 if(ctx
->cur_vs_args
->super
.clip_enabled
)
821 clip_limit
= gl_info
->limits
.user_clip_distances
;
827 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
828 clip_limit
= min(wined3d_popcount(mask
), 4);
830 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
831 max_constantsF
-= *num_clipplanes
;
832 if(*num_clipplanes
< clip_limit
)
834 WARN("Only %u clip planes out of %u enabled.\n", *num_clipplanes
,
835 gl_info
->limits
.user_clip_distances
);
840 if (ctx
->target_version
>= NV2
)
841 *num_clipplanes
= gl_info
->limits
.user_clip_distances
;
843 *num_clipplanes
= min(gl_info
->limits
.user_clip_distances
, 4);
847 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
849 if (map
& 1) shader_addline(buffer
, "TEMP R%u;\n", i
);
852 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
854 if (map
& 1) shader_addline(buffer
, "ADDRESS A%u;\n", i
);
857 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
859 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
861 if (map
& 1) shader_addline(buffer
, "TEMP T%u;\n", i
);
865 if (!shader
->load_local_constsF
)
867 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
870 value
= (const float *)lconst
->value
;
871 shader_addline(buffer
, "PARAM C%u = ", lconst
->idx
);
872 shader_arb_append_imm_vec4(buffer
, value
);
873 shader_addline(buffer
, ";\n");
877 /* After subtracting privately used constants from the hardware limit(they are loaded as
878 * local constants), make sure the shader doesn't violate the env constant limit
882 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
886 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
889 /* Avoid declaring more constants than needed */
890 max_constantsF
= min(max_constantsF
, shader
->limits
->constant_float
);
892 /* we use the array-based constants array if the local constants are marked for loading,
893 * because then we use indirect addressing, or when the local constant list is empty,
894 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
895 * local constants do not declare the loaded constants as an array because ARB compilers usually
896 * do not optimize unused constants away
898 if (reg_maps
->usesrelconstF
)
900 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
901 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
902 max_constantsF
, max_constantsF
- 1);
906 for (i
= 0; i
< max_constantsF
; ++i
)
908 if (!shader_constant_is_local(shader
, i
) && wined3d_extract_bits(reg_maps
->constf
, i
, 1))
910 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
916 static const char * const shift_tab
[] = {
917 "dummy", /* 0 (none) */
918 "coefmul.x", /* 1 (x2) */
919 "coefmul.y", /* 2 (x4) */
920 "coefmul.z", /* 3 (x8) */
921 "coefmul.w", /* 4 (x16) */
922 "dummy", /* 5 (x32) */
923 "dummy", /* 6 (x64) */
924 "dummy", /* 7 (x128) */
925 "dummy", /* 8 (d256) */
926 "dummy", /* 9 (d128) */
927 "dummy", /* 10 (d64) */
928 "dummy", /* 11 (d32) */
929 "coefdiv.w", /* 12 (d16) */
930 "coefdiv.z", /* 13 (d8) */
931 "coefdiv.y", /* 14 (d4) */
932 "coefdiv.x" /* 15 (d2) */
935 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
936 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
938 char *ptr
= write_mask
;
940 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
943 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
944 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
945 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
946 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
952 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
954 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
955 * but addressed as "rgba". To fix this we need to swap the register's x
956 * and z components. */
957 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
958 char *ptr
= swizzle_str
;
960 /* swizzle bits fields: wwzzyyxx */
961 DWORD swizzle
= param
->swizzle
;
962 DWORD swizzle_x
= swizzle
& 0x03;
963 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
964 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
965 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
967 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
968 * generate a swizzle string. Unless we need to our own swizzling. */
969 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
972 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
973 *ptr
++ = swizzle_chars
[swizzle_x
];
975 *ptr
++ = swizzle_chars
[swizzle_x
];
976 *ptr
++ = swizzle_chars
[swizzle_y
];
977 *ptr
++ = swizzle_chars
[swizzle_z
];
978 *ptr
++ = swizzle_chars
[swizzle_w
];
985 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
987 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
988 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
990 if (!strcmp(priv
->addr_reg
, src
))
993 snprintf(priv
->addr_reg
, sizeof(priv
->addr_reg
), "%s", src
);
994 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
997 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
998 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
1000 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
1001 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
1003 /* oPos, oFog and oPts in D3D */
1004 static const char * const rastout_reg_names
[] = {"TMP_OUT", "TMP_FOGCOORD", "result.pointsize"};
1005 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1006 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1007 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1008 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1014 case WINED3DSPR_TEMP
:
1015 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1018 case WINED3DSPR_INPUT
:
1021 if (reg_maps
->shader_version
.major
< 3)
1023 if (!reg
->idx
[0].offset
)
1024 strcpy(register_name
, "fragment.color.primary");
1026 strcpy(register_name
, "fragment.color.secondary");
1030 if (reg
->idx
[0].rel_addr
)
1033 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1035 if (!strcmp(rel_reg
, "**aL_emul**"))
1037 DWORD idx
= ctx
->aL
+ reg
->idx
[0].offset
;
1038 if(idx
< MAX_REG_INPUT
)
1040 strcpy(register_name
, ctx
->ps_input
[idx
]);
1044 ERR("Pixel shader input register out of bounds: %u\n", idx
);
1045 sprintf(register_name
, "out_of_bounds_%u", idx
);
1048 else if (reg_maps
->input_registers
& 0x0300)
1050 /* There are two ways basically:
1052 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
1053 * That means trouble if the loop also contains a breakc or if the control values
1054 * aren't local constants.
1055 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
1056 * source dynamically. The trouble is that we cannot simply read aL.y because it
1057 * is an ADDRESS register. We could however push it, load .zw with a value and use
1058 * ADAC to load the condition code register and pop it again afterwards
1060 FIXME("Relative input register addressing with more than 8 registers\n");
1062 /* This is better than nothing for now */
1063 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1065 else if(ctx
->cur_ps_args
->super
.vp_mode
!= WINED3D_VP_MODE_SHADER
)
1067 /* This is problematic because we'd have to consult the ctx->ps_input strings
1068 * for where to find the varying. Some may be "0.0", others can be texcoords or
1069 * colors. This needs either a pipeline replacement to make the vertex shader feed
1070 * proper varyings, or loop unrolling
1072 * For now use the texcoords and hope for the best
1074 FIXME("Non-vertex shader varying input with indirect addressing\n");
1075 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1079 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1080 * pulls GL_NV_fragment_program2 in
1082 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1087 if (reg
->idx
[0].offset
< MAX_REG_INPUT
)
1089 strcpy(register_name
, ctx
->ps_input
[reg
->idx
[0].offset
]);
1093 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
[0].offset
);
1094 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
[0].offset
);
1101 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1u << reg
->idx
[0].offset
))
1103 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
[0].offset
);
1107 case WINED3DSPR_CONST
:
1108 if (!pshader
&& reg
->idx
[0].rel_addr
)
1110 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1111 UINT rel_offset
= ctx
->target_version
== ARB
? shader_data
->rel_offset
: 0;
1114 if (reg_maps
->shader_version
.major
< 2)
1116 sprintf(rel_reg
, "A0.x");
1120 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1121 if (ctx
->target_version
== ARB
)
1123 if (!strcmp(rel_reg
, "**aL_emul**"))
1127 shader_arb_request_a0(ins
, rel_reg
);
1128 sprintf(rel_reg
, "A0.x");
1133 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
[0].offset
);
1134 else if (reg
->idx
[0].offset
>= rel_offset
)
1135 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
[0].offset
- rel_offset
);
1137 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
[0].offset
);
1141 if (reg_maps
->usesrelconstF
)
1142 sprintf(register_name
, "C[%u]", reg
->idx
[0].offset
);
1144 sprintf(register_name
, "C%u", reg
->idx
[0].offset
);
1148 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1151 if (reg_maps
->shader_version
.major
== 1
1152 && reg_maps
->shader_version
.minor
<= 3)
1153 /* In ps <= 1.3, Tx is a temporary register as destination
1154 * to all instructions, and as source to most instructions.
1155 * For some instructions it is the texcoord input. Those
1156 * instructions know about the special use. */
1157 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1159 /* In ps 1.4 and 2.x Tx is always a (read-only) varying. */
1160 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
[0].offset
);
1164 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1165 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1167 sprintf(register_name
, "A%u_SHADOW", reg
->idx
[0].offset
);
1171 case WINED3DSPR_COLOROUT
:
1172 if (ctx
->ps_post_process
&& !reg
->idx
[0].offset
)
1174 strcpy(register_name
, "TMP_COLOR");
1178 if (ctx
->cur_ps_args
->super
.srgb_correction
)
1179 FIXME("sRGB correction on higher render targets.\n");
1180 if (reg_maps
->rt_mask
> 1)
1181 sprintf(register_name
, "result.color[%u]", reg
->idx
[0].offset
);
1183 strcpy(register_name
, "result.color");
1187 case WINED3DSPR_RASTOUT
:
1188 if (reg
->idx
[0].offset
== 1)
1189 sprintf(register_name
, "%s", ctx
->fog_output
);
1191 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
[0].offset
]);
1194 case WINED3DSPR_DEPTHOUT
:
1195 strcpy(register_name
, "result.depth");
1198 case WINED3DSPR_ATTROUT
:
1199 /* case WINED3DSPR_OUTPUT: */
1201 sprintf(register_name
, "oD[%u]", reg
->idx
[0].offset
);
1203 strcpy(register_name
, ctx
->color_output
[reg
->idx
[0].offset
]);
1206 case WINED3DSPR_TEXCRDOUT
:
1208 sprintf(register_name
, "oT[%u]", reg
->idx
[0].offset
);
1209 else if (reg_maps
->shader_version
.major
< 3)
1210 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
[0].offset
]);
1212 strcpy(register_name
, ctx
->vs_output
[reg
->idx
[0].offset
]);
1215 case WINED3DSPR_LOOP
:
1216 if(ctx
->target_version
>= NV2
)
1218 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1219 if(pshader
) sprintf(register_name
, "A0.x");
1220 else sprintf(register_name
, "aL.y");
1224 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1225 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1226 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1227 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1230 sprintf(register_name
, "**aL_emul**");
1235 case WINED3DSPR_CONSTINT
:
1236 sprintf(register_name
, "I%u", reg
->idx
[0].offset
);
1239 case WINED3DSPR_MISCTYPE
:
1240 if (!reg
->idx
[0].offset
)
1241 sprintf(register_name
, "vpos");
1242 else if (reg
->idx
[0].offset
== 1)
1243 sprintf(register_name
, "fragment.facing.x");
1245 FIXME("Unknown MISCTYPE register index %u.\n", reg
->idx
[0].offset
);
1249 FIXME("Unhandled register type %#x[%u].\n", reg
->type
, reg
->idx
[0].offset
);
1250 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
[0].offset
);
1255 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1256 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1258 char register_name
[255];
1262 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1263 strcpy(str
, register_name
);
1265 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1266 strcat(str
, write_mask
);
1269 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1271 switch(channel_source
)
1273 case CHANNEL_SOURCE_ZERO
: return "0";
1274 case CHANNEL_SOURCE_ONE
: return "1";
1275 case CHANNEL_SOURCE_X
: return "x";
1276 case CHANNEL_SOURCE_Y
: return "y";
1277 case CHANNEL_SOURCE_Z
: return "z";
1278 case CHANNEL_SOURCE_W
: return "w";
1280 FIXME("Unhandled channel source %#x\n", channel_source
);
1285 struct color_fixup_masks
1291 static struct color_fixup_masks
calc_color_correction(struct color_fixup_desc fixup
, DWORD dst_mask
)
1293 struct color_fixup_masks masks
= {0, 0};
1295 if (is_complex_fixup(fixup
))
1297 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1298 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1302 if (fixup
.x_source
!= CHANNEL_SOURCE_X
)
1303 masks
.source
|= WINED3DSP_WRITEMASK_0
;
1304 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
)
1305 masks
.source
|= WINED3DSP_WRITEMASK_1
;
1306 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
)
1307 masks
.source
|= WINED3DSP_WRITEMASK_2
;
1308 if (fixup
.w_source
!= CHANNEL_SOURCE_W
)
1309 masks
.source
|= WINED3DSP_WRITEMASK_3
;
1310 masks
.source
&= dst_mask
;
1312 if (fixup
.x_sign_fixup
)
1313 masks
.sign
|= WINED3DSP_WRITEMASK_0
;
1314 if (fixup
.y_sign_fixup
)
1315 masks
.sign
|= WINED3DSP_WRITEMASK_1
;
1316 if (fixup
.z_sign_fixup
)
1317 masks
.sign
|= WINED3DSP_WRITEMASK_2
;
1318 if (fixup
.w_sign_fixup
)
1319 masks
.sign
|= WINED3DSP_WRITEMASK_3
;
1320 masks
.sign
&= dst_mask
;
1325 static void gen_color_correction(struct wined3d_string_buffer
*buffer
, const char *dst
,
1326 const char *src
, const char *one
, const char *two
,
1327 struct color_fixup_desc fixup
, struct color_fixup_masks masks
)
1329 const char *sign_fixup_src
= dst
;
1334 sign_fixup_src
= "TA";
1336 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", sign_fixup_src
, src
,
1337 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1338 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1340 else if (masks
.sign
)
1342 sign_fixup_src
= src
;
1348 char *ptr
= reg_mask
;
1350 if (masks
.sign
!= WINED3DSP_WRITEMASK_ALL
)
1353 if (masks
.sign
& WINED3DSP_WRITEMASK_0
)
1355 if (masks
.sign
& WINED3DSP_WRITEMASK_1
)
1357 if (masks
.sign
& WINED3DSP_WRITEMASK_2
)
1359 if (masks
.sign
& WINED3DSP_WRITEMASK_3
)
1364 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", dst
, reg_mask
, sign_fixup_src
, two
, one
);
1368 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1371 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1372 if (!ins
->dst_count
) return "";
1374 mod
= ins
->dst
[0].modifiers
;
1376 /* Silently ignore PARTIALPRECISION if it's not supported */
1377 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1379 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1381 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1382 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1387 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1390 case WINED3DSPDM_SATURATE
:
1393 case WINED3DSPDM_PARTIALPRECISION
:
1400 FIXME("Unknown modifiers 0x%08x\n", mod
);
1405 #define TEX_PROJ 0x1
1406 #define TEX_BIAS 0x2
1408 #define TEX_DERIV 0x10
1410 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1411 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1413 enum wined3d_shader_resource_type resource_type
= ins
->ctx
->reg_maps
->resource_info
[sampler_idx
].type
;
1414 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1415 const char *tex_type
;
1416 BOOL np2_fixup
= FALSE
;
1417 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1419 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1420 const char *tex_dst
= dst_str
;
1421 struct color_fixup_masks masks
;
1423 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1424 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1426 switch (resource_type
)
1428 case WINED3D_SHADER_RESOURCE_TEXTURE_1D
:
1432 case WINED3D_SHADER_RESOURCE_TEXTURE_2D
:
1433 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1u << sampler_idx
)
1434 && ins
->ctx
->gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1438 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1440 if (priv
->cur_np2fixup_info
->super
.active
& (1u << sampler_idx
))
1442 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1443 else np2_fixup
= TRUE
;
1448 case WINED3D_SHADER_RESOURCE_TEXTURE_3D
:
1452 case WINED3D_SHADER_RESOURCE_TEXTURE_CUBE
:
1457 ERR("Unexpected resource type %#x.\n", resource_type
);
1461 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1462 * so don't use shader_arb_get_modifier
1464 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1467 /* Fragment samplers always have indentity mapping */
1468 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1470 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1475 masks
= calc_color_correction(priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
],
1476 ins
->dst
[0].write_mask
);
1478 if (masks
.source
|| masks
.sign
)
1482 if (flags
& TEX_DERIV
)
1484 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1485 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1486 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1487 dsx
, dsy
, sampler_idx
, tex_type
);
1489 else if(flags
& TEX_LOD
)
1491 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1492 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1493 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1494 sampler_idx
, tex_type
);
1496 else if (flags
& TEX_BIAS
)
1498 /* Shouldn't be possible, but let's check for it */
1499 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1500 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1501 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1503 else if (flags
& TEX_PROJ
)
1505 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1511 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1512 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1513 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1515 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, tex_dst
, sampler_idx
, tex_type
);
1518 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1523 gen_color_correction(buffer
, dst_str
, tex_dst
,
1524 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1525 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1526 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
], masks
);
1530 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1531 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1533 /* Generate a line that does the input modifier computation and return the input register to use */
1534 BOOL is_color
= FALSE
, insert_line
;
1537 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1538 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1539 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1540 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1542 /* Assume a new line will be added */
1545 /* Get register name */
1546 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1547 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1549 switch (src
->modifiers
)
1551 case WINED3DSPSM_NONE
:
1552 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1553 insert_line
= FALSE
;
1555 case WINED3DSPSM_NEG
:
1556 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1557 insert_line
= FALSE
;
1559 case WINED3DSPSM_BIAS
:
1560 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1562 case WINED3DSPSM_BIASNEG
:
1563 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1565 case WINED3DSPSM_SIGN
:
1566 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1568 case WINED3DSPSM_SIGNNEG
:
1569 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1571 case WINED3DSPSM_COMP
:
1572 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1574 case WINED3DSPSM_X2
:
1575 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1577 case WINED3DSPSM_X2NEG
:
1578 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1580 case WINED3DSPSM_DZ
:
1581 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1582 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1584 case WINED3DSPSM_DW
:
1585 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1586 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1588 case WINED3DSPSM_ABS
:
1589 if(ctx
->target_version
>= NV2
) {
1590 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1591 insert_line
= FALSE
;
1593 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1596 case WINED3DSPSM_ABSNEG
:
1597 if(ctx
->target_version
>= NV2
) {
1598 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1600 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1601 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1603 insert_line
= FALSE
;
1606 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1607 insert_line
= FALSE
;
1610 /* Return modified or original register, with swizzle */
1612 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1615 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1617 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1618 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1619 DWORD sampler_code
= dst
->reg
.idx
[0].offset
;
1621 char src_name
[2][50];
1623 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1625 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1627 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1628 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1631 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1632 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1633 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1634 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1635 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1637 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1638 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1641 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1646 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1647 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1648 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1649 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1650 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1651 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1652 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1653 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1654 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1655 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1656 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1657 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1658 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1660 FIXME("Unknown modifier %u\n", mod
);
1664 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1666 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1667 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1669 char src_name
[3][50];
1670 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1671 ins
->ctx
->reg_maps
->shader_version
.minor
);
1673 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1674 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1676 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
1677 && ins
->dst
->write_mask
!= WINED3DSP_WRITEMASK_3
)
1679 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1683 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1686 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1687 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1689 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1690 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1691 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1692 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1693 dst_name
, src_name
[1], src_name
[2]);
1697 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1699 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1700 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1702 char src_name
[3][50];
1704 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1706 /* Generate input register names (with modifiers) */
1707 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1708 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1709 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1711 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1712 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1715 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1716 * dst = dot2(src0, src1) + src2 */
1717 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1719 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1720 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1722 char src_name
[3][50];
1723 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1725 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1726 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1727 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1729 if(ctx
->target_version
>= NV3
)
1731 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1732 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1733 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1734 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1736 else if(ctx
->target_version
>= NV2
)
1738 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1739 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1740 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1741 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1743 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1745 * .xyxy and other swizzles that we could get with this are not valid in
1746 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1748 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1749 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1750 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1752 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1754 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1755 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1759 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1760 /* Emulate a DP2 with a DP3 and 0.0. Don't use the dest as temp register, it could be src[1] or src[2]
1761 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1763 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1764 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1765 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1766 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1770 /* Map the opcode 1-to-1 to the GL code */
1771 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1773 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1774 const char *instruction
;
1775 char arguments
[256], dst_str
[50];
1777 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1779 switch (ins
->handler_idx
)
1781 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1782 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1783 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1784 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1785 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1786 case WINED3DSIH_DST
: instruction
= "DST"; break;
1787 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1788 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1789 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1790 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1791 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1792 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1793 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1794 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1795 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1796 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1797 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1798 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1799 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1800 default: instruction
= "";
1801 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
1805 /* Note that shader_arb_add_dst_param() adds spaces. */
1806 arguments
[0] = '\0';
1807 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1808 for (i
= 0; i
< ins
->src_count
; ++i
)
1811 strcat(arguments
, ", ");
1812 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1813 strcat(arguments
, operand
);
1815 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1818 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1820 static DWORD
shader_arb_select_component(DWORD swizzle
, DWORD component
)
1822 return ((swizzle
>> 2 * component
) & 0x3) * 0x55;
1825 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1827 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1828 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1829 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1830 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1831 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1832 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1833 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1835 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1836 char src0_param
[256];
1838 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1840 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1842 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1844 if(ctx
->target_version
>= NV2
) {
1845 shader_hw_map2gl(ins
);
1848 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1849 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1851 /* This implements the mova formula used in GLSL. The first two instructions
1852 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1856 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1858 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1859 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1861 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1862 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1864 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1865 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1866 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1867 if (shader_data
->rel_offset
)
1869 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1871 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1873 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1875 else if (reg_maps
->shader_version
.major
== 1
1876 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1877 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1879 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1880 src0_param
[0] = '\0';
1882 if (shader_data
->rel_offset
&& ctx
->target_version
== ARB
)
1884 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1885 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1886 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1887 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1891 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1892 * with more than one component. Thus replicate the first source argument over all
1893 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1894 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1895 tmp_src
.swizzle
= shader_arb_select_component(tmp_src
.swizzle
, 0);
1896 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1897 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1900 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1902 if (ctx
->ps_post_process
&& shader
->u
.ps
.color0_mov
)
1904 shader_addline(buffer
, "#mov handled in srgb write or fog code\n");
1907 shader_hw_map2gl(ins
);
1911 shader_hw_map2gl(ins
);
1915 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1917 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1918 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1921 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1922 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1924 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1926 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1928 const char *kilsrc
= "TA";
1931 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1932 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1938 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1939 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1940 * masked out components to 0(won't kill)
1942 char x
= '0', y
= '0', z
= '0', w
= '0';
1943 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1944 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1945 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1946 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1947 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1949 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1953 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1954 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1956 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1957 * or pass in any temporary register(in shader phase 2)
1959 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1960 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1962 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1963 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1964 shader_addline(buffer
, "KIL TA;\n");
1968 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1970 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1971 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1972 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1973 ins
->ctx
->reg_maps
->shader_version
.minor
);
1974 struct wined3d_shader_src_param src
;
1978 DWORD reg_sampler_code
;
1980 BOOL swizzle_coord
= FALSE
;
1982 /* All versions have a destination register */
1983 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1985 /* 1.0-1.4: Use destination register number as texture code.
1986 2.0+: Use provided sampler number as texture code. */
1987 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1988 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
1990 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
1992 /* 1.0-1.3: Use the texcoord varying.
1993 1.4+: Use provided coordinate source register. */
1994 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1995 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
1997 /* TEX is the only instruction that can handle DW and DZ natively */
1999 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
2000 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
2001 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
2005 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
2006 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
2007 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
2009 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2012 if (reg_sampler_code
< MAX_TEXTURES
)
2013 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2014 if (flags
& WINED3D_PSARGS_PROJECTED
)
2016 myflags
|= TEX_PROJ
;
2017 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
2018 swizzle_coord
= TRUE
;
2021 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
2023 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
2024 if (src_mod
== WINED3DSPSM_DZ
)
2026 swizzle_coord
= TRUE
;
2027 myflags
|= TEX_PROJ
;
2028 } else if(src_mod
== WINED3DSPSM_DW
) {
2029 myflags
|= TEX_PROJ
;
2032 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
2033 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
2038 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
2039 * reg_coord is a read-only varying register, so we need a temp reg */
2040 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
2041 strcpy(reg_coord
, "TA");
2044 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
2047 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
2049 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2050 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2051 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2052 ins
->ctx
->reg_maps
->shader_version
.minor
);
2055 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2057 DWORD reg
= dst
->reg
.idx
[0].offset
;
2059 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2060 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2064 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2065 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2066 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2070 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2072 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2075 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2079 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2080 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2081 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2082 /* Move .x first in case src_str is "TA" */
2083 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2084 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2085 if (reg1
< MAX_TEXTURES
)
2087 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2088 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2090 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2093 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2095 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2097 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2101 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2102 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2103 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2104 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2105 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2106 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2109 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2111 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2115 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2116 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2117 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2118 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2121 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2123 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2124 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2125 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2126 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2127 DWORD reg_dest_code
;
2129 /* All versions have a destination register. The Tx where the texture coordinates come
2130 * from is the varying incarnation of the texture register
2132 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2133 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2134 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2135 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2137 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2138 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2140 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2141 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2143 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2144 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2147 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2148 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2149 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2150 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2152 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2153 * so we can't let the GL handle this.
2155 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2156 & WINED3D_PSARGS_PROJECTED
)
2158 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2159 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2160 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2162 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2165 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2167 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2169 /* No src swizzles are allowed, so this is ok */
2170 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2171 src_reg
, reg_dest_code
, reg_dest_code
);
2172 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2176 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2178 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2179 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2180 char src0_name
[50], dst_name
[50];
2182 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2184 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2185 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2186 * T<reg+1> register. Use this register to store the calculated vector
2188 tmp_reg
.idx
[0].offset
= reg
+ 1;
2189 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2190 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2193 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2195 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2197 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2198 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2204 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2205 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2207 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2208 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2209 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2210 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2211 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2214 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2216 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2217 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2218 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2219 char src0_name
[50], dst_name
[50];
2220 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2223 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2224 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2225 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2227 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2228 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2230 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2231 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2232 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2233 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2236 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2238 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2239 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2241 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2242 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2244 char src0_name
[50], dst_name
[50];
2247 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2248 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2249 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2251 /* Sample the texture using the calculated coordinates */
2252 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2253 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2254 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2255 tex_mx
->current_row
= 0;
2258 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2260 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2261 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2263 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2264 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2270 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2271 * components for temporary data storage
2273 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2274 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2275 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2277 /* Construct the eye-ray vector from w coordinates */
2278 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2279 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2280 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2282 /* Calculate reflection vector
2284 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2285 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2286 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2287 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2288 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2289 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2290 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2292 /* Sample the texture using the calculated coordinates */
2293 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2294 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2295 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2296 tex_mx
->current_row
= 0;
2299 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2301 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2302 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2304 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2305 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2312 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2313 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2314 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2315 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2316 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2318 /* Calculate reflection vector.
2321 * dst_reg.xyz = 2 * --------- * N - E
2324 * Which normalizes the normal vector
2326 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2327 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2328 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2329 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2330 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2331 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2333 /* Sample the texture using the calculated coordinates */
2334 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2335 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2336 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2337 tex_mx
->current_row
= 0;
2340 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2342 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2343 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2345 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2346 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2348 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2349 * which is essentially an input, is the destination register because it is the first
2350 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2351 * here(writemasks/swizzles are not valid on texdepth)
2353 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2355 /* According to the msdn, the source register(must be r5) is unusable after
2356 * the texdepth instruction, so we're free to modify it
2358 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2360 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2361 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2362 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2364 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2365 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2366 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2367 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2370 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2371 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2372 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2373 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2375 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2376 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2380 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2381 shader_addline(buffer
, "MOV TB, 0.0;\n");
2382 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2384 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2385 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2388 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2389 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2390 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2392 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2395 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2397 /* Handle output register */
2398 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2399 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2400 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2403 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2404 * Perform the 3rd row of a 3x3 matrix multiply */
2405 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2407 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2408 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2409 char dst_str
[50], dst_name
[50];
2413 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2414 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2415 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2416 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2417 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2420 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2421 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2422 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2423 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2425 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2427 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2428 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2429 char src0
[50], dst_name
[50];
2431 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2432 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2434 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2435 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2436 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2438 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2439 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2440 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2442 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2443 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2444 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2445 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2448 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2449 Vertex/Pixel shaders to ARB_vertex_program codes */
2450 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2453 int nComponents
= 0;
2454 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2455 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2456 struct wined3d_shader_instruction tmp_ins
;
2458 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2460 /* Set constants for the temporary argument */
2461 tmp_ins
.ctx
= ins
->ctx
;
2462 tmp_ins
.dst_count
= 1;
2463 tmp_ins
.dst
= &tmp_dst
;
2464 tmp_ins
.src_count
= 2;
2465 tmp_ins
.src
= tmp_src
;
2467 switch(ins
->handler_idx
)
2469 case WINED3DSIH_M4x4
:
2471 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2473 case WINED3DSIH_M4x3
:
2475 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2477 case WINED3DSIH_M3x4
:
2479 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2481 case WINED3DSIH_M3x3
:
2483 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2485 case WINED3DSIH_M3x2
:
2487 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2490 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2494 tmp_dst
= ins
->dst
[0];
2495 tmp_src
[0] = ins
->src
[0];
2496 tmp_src
[1] = ins
->src
[1];
2497 for (i
= 0; i
< nComponents
; ++i
)
2499 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2500 shader_hw_map2gl(&tmp_ins
);
2501 ++tmp_src
[1].reg
.idx
[0].offset
;
2505 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2511 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2512 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2513 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2514 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2515 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2516 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2517 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2518 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2519 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2520 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2521 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2522 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2523 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2525 FIXME("Unknown modifier %u\n", mod
);
2529 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2531 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2532 const char *instruction
;
2533 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2534 BOOL need_abs
= FALSE
;
2539 switch(ins
->handler_idx
)
2541 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2542 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2543 case WINED3DSIH_EXPP
:
2544 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2546 instruction
= "EXP";
2550 case WINED3DSIH_EXP
:
2551 instruction
= "EX2";
2553 case WINED3DSIH_LOG
:
2554 case WINED3DSIH_LOGP
:
2555 /* The precision requirements suggest that LOGP matches ARBvp's LOG
2556 * instruction, but notice that the output of those instructions is
2558 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2559 instruction
= "LG2";
2561 default: instruction
= "";
2562 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2566 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2568 src0_copy
.swizzle
= shader_arb_select_component(src0_copy
.swizzle
, 3);
2570 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2571 shader_arb_get_src_param(ins
, &src0_copy
, 0, src
);
2575 shader_addline(buffer
, "ABS TA.w, %s;\n", src
);
2576 shader_addline(buffer
, "%s%s %s, TA.w;\n", instruction
, shader_arb_get_modifier(ins
), dst
);
2580 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2585 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2587 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2590 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2591 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2592 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2594 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2595 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2597 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2598 * otherwise NRM or RSQ would return NaN */
2599 if(pshader
&& priv
->target_version
>= NV3
)
2601 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2603 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2605 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2606 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2607 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2609 else if(priv
->target_version
>= NV2
)
2611 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2612 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2613 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2618 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2620 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2621 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2622 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2624 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2625 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2627 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2628 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2629 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2634 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2636 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2638 char src_name
[3][50];
2640 /* ARB_fragment_program has a convenient LRP instruction */
2641 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2642 shader_hw_map2gl(ins
);
2646 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2647 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2648 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2649 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2651 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2652 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2653 dst_name
, src_name
[0], src_name
[2]);
2656 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2658 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2659 * must contain fixed constants. So we need a separate function to filter those constants and
2662 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2663 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2664 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2666 char src_name0
[50], src_name1
[50], src_name2
[50];
2669 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2670 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2671 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2672 /* No modifiers are supported on SCS */
2673 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2675 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2677 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2678 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2680 } else if(priv
->target_version
>= NV2
) {
2681 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2683 /* Sincos writemask must be .x, .y or .xy */
2684 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2685 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2686 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2687 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2689 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2690 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2692 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2693 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2695 * The constants we get are:
2697 * +1 +1, -1 -1 +1 +1 -1 -1
2698 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2699 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2701 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2705 * (x/2)^4 = x^4 / 16
2706 * (x/2)^5 = x^5 / 32
2709 * To get the final result:
2710 * sin(x) = 2 * sin(x/2) * cos(x/2)
2711 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2712 * (from sin(x+y) and cos(x+y) rules)
2714 * As per MSDN, dst.z is undefined after the operation, and so is
2715 * dst.x and dst.y if they're masked out by the writemask. Ie
2716 * sincos dst.y, src1, c0, c1
2717 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2718 * vsa.exe also stops with an error if the dest register is the same register as the source
2719 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2720 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2722 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2723 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2724 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2726 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2727 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2728 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2729 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2730 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2731 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2735 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2736 * properly merge that with MULs in the code above?
2737 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2738 * we can merge the sine and cosine MAD rows to calculate them together.
2740 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2741 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2742 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2743 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2746 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2747 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2748 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2750 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2752 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2753 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2755 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2757 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2758 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2763 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2765 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2768 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2770 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2771 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2773 /* SGN is only valid in vertex shaders */
2774 if(ctx
->target_version
>= NV2
) {
2775 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2779 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2780 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2782 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2783 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2785 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2786 * Then use TA, and calculate the final result
2788 * Not reading from TA? Store the first result in TA to avoid overwriting the
2789 * destination if src reg = dst reg
2791 if(strstr(src_name
, "TA"))
2793 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2794 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2795 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2799 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2800 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2801 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2806 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2808 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2814 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2815 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2816 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2818 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2819 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2822 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2824 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2825 char src0
[50], src1
[50], dst
[50];
2826 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2827 BOOL need_abs
= FALSE
;
2828 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2829 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2831 /* POW operates on the absolute value of the input */
2832 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2834 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2835 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2836 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2839 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2841 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2843 if (priv
->target_version
>= NV2
)
2845 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2846 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2847 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2851 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2852 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2854 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2855 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2856 /* Possibly add flt_eps to avoid getting float special values */
2857 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2858 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2859 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2860 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2864 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2866 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2868 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2871 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2875 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2876 struct list
*e
= list_head(&priv
->control_frames
);
2877 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2879 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2880 /* The constant loader makes sure to load -1 into iX.w */
2881 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2882 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2883 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2887 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2891 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2893 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2895 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2897 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2899 /* The constant loader makes sure to load -1 into iX.w */
2902 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2903 struct list
*e
= list_head(&priv
->control_frames
);
2904 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2906 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2908 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2909 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2910 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2914 shader_addline(buffer
, "REP %s;\n", src_name
);
2918 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2920 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2921 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2925 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2926 struct list
*e
= list_head(&priv
->control_frames
);
2927 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2929 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2930 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2931 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2933 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2937 shader_addline(buffer
, "ENDLOOP;\n");
2941 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2943 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2944 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2948 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2949 struct list
*e
= list_head(&priv
->control_frames
);
2950 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2952 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2953 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2954 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2956 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2960 shader_addline(buffer
, "ENDREP;\n");
2964 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2966 struct control_frame
*control_frame
;
2968 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2970 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2972 ERR("Could not find loop for break\n");
2976 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2978 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2979 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2980 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2984 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2988 shader_addline(buffer
, "BRK;\n");
2992 static const char *get_compare(enum wined3d_shader_rel_op op
)
2996 case WINED3D_SHADER_REL_OP_GT
: return "GT";
2997 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
2998 case WINED3D_SHADER_REL_OP_GE
: return "GE";
2999 case WINED3D_SHADER_REL_OP_LT
: return "LT";
3000 case WINED3D_SHADER_REL_OP_NE
: return "NE";
3001 case WINED3D_SHADER_REL_OP_LE
: return "LE";
3003 FIXME("Unrecognized operator %#x.\n", op
);
3008 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
3012 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
3013 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
3014 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
3015 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
3016 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
3017 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
3019 FIXME("Unrecognized operator %#x.\n", op
);
3024 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
3026 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3027 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3028 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
3031 const char *comp
= get_compare(ins
->flags
);
3033 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3034 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3038 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
3039 * away the subtraction result
3041 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3042 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3046 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3047 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3051 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3053 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3054 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3055 struct list
*e
= list_head(&priv
->control_frames
);
3056 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3060 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3062 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3063 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3067 /* Invert the flag. We jump to the else label if the condition is NOT true */
3068 comp
= get_compare(invert_compare(ins
->flags
));
3069 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3070 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3074 comp
= get_compare(ins
->flags
);
3075 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3076 shader_addline(buffer
, "IF %s.x;\n", comp
);
3080 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3082 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3083 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3084 struct list
*e
= list_head(&priv
->control_frames
);
3085 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3086 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3090 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3091 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3092 control_frame
->had_else
= TRUE
;
3096 shader_addline(buffer
, "ELSE;\n");
3100 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3102 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3103 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3104 struct list
*e
= list_head(&priv
->control_frames
);
3105 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3106 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3110 if(control_frame
->had_else
)
3112 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3116 shader_addline(buffer
, "#No else branch. else is endif\n");
3117 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3122 shader_addline(buffer
, "ENDIF;\n");
3126 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3128 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3130 char reg_src
[3][40];
3131 WORD flags
= TEX_DERIV
;
3133 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3134 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3135 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3136 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3138 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3139 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3141 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3144 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3146 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3149 WORD flags
= TEX_LOD
;
3151 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3152 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3154 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3155 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3157 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3160 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3162 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3163 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3165 priv
->in_main_func
= FALSE
;
3166 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3167 * subroutine, don't generate a label that will make GL complain
3169 if(priv
->target_version
== ARB
) return;
3171 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3174 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3175 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3176 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3177 struct wined3d_string_buffer
*buffer
)
3181 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3182 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3183 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3184 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3186 if (args
->super
.fog_src
== VS_FOG_Z
)
3188 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3194 /* posFixup.x is always 1.0, so we can safely use it */
3195 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3199 /* Clamp fogcoord */
3200 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3201 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3203 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3204 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3208 /* Clipplanes are always stored without y inversion */
3209 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3211 if (args
->super
.clip_enabled
)
3213 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3215 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3219 else if (args
->clip
.boolclip
.clip_texcoord
)
3221 static const char component
[4] = {'x', 'y', 'z', 'w'};
3222 unsigned int cur_clip
= 0;
3223 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3225 for (i
= 0; i
< gl_info
->limits
.user_clip_distances
; ++i
)
3227 if (args
->clip
.boolclip
.clipplane_mask
& (1u << i
))
3229 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3230 component
[cur_clip
++], i
);
3236 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3239 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3242 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3245 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3248 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3249 args
->clip
.boolclip
.clip_texcoord
- 1);
3252 /* Write the final position.
3254 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3255 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3256 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3257 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3259 if (!gl_info
->supported
[ARB_CLIP_CONTROL
])
3261 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3262 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3263 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3265 /* Z coord [0;1]->[-1;1] mapping, see comment in
3266 * get_projection_matrix() in utils.c. */
3267 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3269 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3270 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3274 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3275 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3279 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3281 priv_ctx
->footer_written
= TRUE
;
3284 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3286 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3287 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3288 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3289 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3291 if(priv
->target_version
== ARB
) return;
3295 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3296 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, ins
->ctx
->gl_info
, buffer
);
3299 shader_addline(buffer
, "RET;\n");
3302 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3304 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3305 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3308 static BOOL
shader_arb_compile(const struct wined3d_gl_info
*gl_info
, GLenum target
, const char *src
)
3310 const char *ptr
, *line
;
3313 if (TRACE_ON(d3d_shader
))
3316 while ((line
= get_line(&ptr
))) TRACE_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3319 GL_EXTCALL(glProgramStringARB(target
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(src
), src
));
3320 checkGLcall("glProgramStringARB()");
3322 if (FIXME_ON(d3d_shader
))
3324 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3327 FIXME_(d3d_shader
)("Program error at position %d: %s\n\n", pos
,
3328 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3330 while ((line
= get_line(&ptr
))) FIXME_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3331 FIXME_(d3d_shader
)("\n");
3337 if (WARN_ON(d3d_perf
))
3339 GL_EXTCALL(glGetProgramivARB(target
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3340 checkGLcall("glGetProgramivARB()");
3342 WARN_(d3d_perf
)("Program exceeds native resource limits.\n");
3348 static void arbfp_add_sRGB_correction(struct wined3d_string_buffer
*buffer
, const char *fragcolor
,
3349 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3351 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3355 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3356 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.x;\n", tmp1
, fragcolor
);
3357 /* Calculate the > 0.0031308 case */
3358 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3359 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3360 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3361 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts0.y;\n", fragcolor
, fragcolor
);
3362 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts0.z;\n", fragcolor
, fragcolor
);
3363 /* Calculate the < case */
3364 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts0.w, %s;\n", fragcolor
, fragcolor
);
3368 /* Calculate the > 0.0031308 case */
3369 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts0.x;\n", tmp1
, fragcolor
);
3370 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts0.x;\n", tmp1
, fragcolor
);
3371 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts0.x;\n", tmp1
, fragcolor
);
3372 shader_addline(buffer
, "MUL %s, %s, srgb_consts0.y;\n", tmp1
, tmp1
);
3373 shader_addline(buffer
, "SUB %s, %s, srgb_consts0.z;\n", tmp1
, tmp1
);
3374 /* Calculate the < case */
3375 shader_addline(buffer
, "MUL %s, srgb_consts0.w, %s;\n", tmp2
, fragcolor
);
3376 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3377 shader_addline(buffer
, "SLT %s, srgb_consts1.x, %s;\n", tmp3
, fragcolor
);
3378 shader_addline(buffer
, "SGE %s, srgb_consts1.x, %s;\n", tmp4
, fragcolor
);
3379 /* Store the components > 0.0031308 in the destination */
3380 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3381 /* Add the components that are < 0.0031308 */
3382 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3383 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3384 * result.color writes(.rgb first, then .a), or handle overwriting already written
3385 * components. The assembler uses a temporary register in this case, which is usually
3386 * not allocated from one of our registers that were used earlier.
3389 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3392 static const DWORD
*find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3394 const struct wined3d_shader_lconst
*constant
;
3396 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3398 if (constant
->idx
== idx
)
3400 return constant
->value
;
3406 static void init_ps_input(const struct wined3d_shader
*shader
,
3407 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3409 static const char * const texcoords
[8] =
3411 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3412 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3415 const struct wined3d_shader_signature_element
*input
;
3416 const char *semantic_name
;
3419 if (args
->super
.vp_mode
== WINED3D_VP_MODE_SHADER
)
3421 /* That one is easy. The vertex shaders provide v0-v7 in
3422 * fragment.texcoord and v8 and v9 in fragment.color. */
3423 for (i
= 0; i
< 8; ++i
)
3425 priv
->ps_input
[i
] = texcoords
[i
];
3427 priv
->ps_input
[8] = "fragment.color.primary";
3428 priv
->ps_input
[9] = "fragment.color.secondary";
3432 /* The fragment shader has to collect the varyings on its own. In any case
3433 * properly load color0 and color1. In the case of pre-transformed
3434 * vertices also load texture coordinates. Set other attributes to 0.0.
3436 * For fixed-function this behavior is correct, according to the tests.
3437 * For pre-transformed we'd either need a replacement shader that can load
3438 * other attributes like BINORMAL, or load the texture coordinate
3439 * attribute pointers to match the fragment shader signature. */
3440 for (i
= 0; i
< shader
->input_signature
.element_count
; ++i
)
3442 input
= &shader
->input_signature
.elements
[i
];
3443 if (!(semantic_name
= input
->semantic_name
))
3445 semantic_idx
= input
->semantic_idx
;
3447 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3450 priv
->ps_input
[input
->register_idx
] = "fragment.color.primary";
3451 else if (semantic_idx
== 1)
3452 priv
->ps_input
[input
->register_idx
] = "fragment.color.secondary";
3454 priv
->ps_input
[input
->register_idx
] = "0.0";
3456 else if (args
->super
.vp_mode
== WINED3D_VP_MODE_FF
)
3458 priv
->ps_input
[input
->register_idx
] = "0.0";
3460 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3462 if (semantic_idx
< 8)
3463 priv
->ps_input
[input
->register_idx
] = texcoords
[semantic_idx
];
3465 priv
->ps_input
[input
->register_idx
] = "0.0";
3467 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3470 priv
->ps_input
[input
->register_idx
] = "fragment.fogcoord";
3472 priv
->ps_input
[input
->register_idx
] = "0.0";
3476 priv
->ps_input
[input
->register_idx
] = "0.0";
3479 TRACE("v%u, semantic %s%u is %s\n", input
->register_idx
,
3480 semantic_name
, semantic_idx
, priv
->ps_input
[input
->register_idx
]);
3484 static void arbfp_add_linear_fog(struct wined3d_string_buffer
*buffer
,
3485 const char *fragcolor
, const char *tmp
)
3487 shader_addline(buffer
, "SUB %s.x, state.fog.params.z, fragment.fogcoord.x;\n", tmp
);
3488 shader_addline(buffer
, "MUL_SAT %s.x, %s.x, state.fog.params.w;\n", tmp
, tmp
);
3489 shader_addline(buffer
, "LRP %s.rgb, %s.x, %s, state.fog.color;\n", fragcolor
, tmp
, fragcolor
);
3492 /* Context activation is done by the caller. */
3493 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3494 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
3495 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3497 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3500 DWORD next_local
= 0;
3501 struct shader_arb_ctx_priv priv_ctx
;
3502 BOOL dcl_td
= FALSE
;
3503 BOOL want_nv_prog
= FALSE
;
3504 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3506 BOOL custom_linear_fog
= FALSE
;
3510 unsigned int i
, found
= 0;
3512 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3515 || (shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3516 || (reg_maps
->shader_version
.major
< 2 && !i
))
3519 sprintf(srgbtmp
[found
], "R%u", i
);
3521 if (found
== 4) break;
3526 sprintf(srgbtmp
[0], "TA");
3527 sprintf(srgbtmp
[1], "TB");
3528 sprintf(srgbtmp
[2], "TC");
3529 sprintf(srgbtmp
[3], "TD");
3533 sprintf(srgbtmp
[1], "TA");
3534 sprintf(srgbtmp
[2], "TB");
3535 sprintf(srgbtmp
[3], "TC");
3538 sprintf(srgbtmp
[2], "TA");
3539 sprintf(srgbtmp
[3], "TB");
3542 sprintf(srgbtmp
[3], "TA");
3548 /* Create the hw ARB shader */
3549 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3550 priv_ctx
.cur_ps_args
= args
;
3551 priv_ctx
.compiled_fprog
= compiled
;
3552 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3553 init_ps_input(shader
, args
, &priv_ctx
);
3554 list_init(&priv_ctx
.control_frames
);
3555 priv_ctx
.ps_post_process
= args
->super
.srgb_correction
;
3557 /* Avoid enabling NV_fragment_program* if we do not need it.
3559 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3560 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3561 * is faster than what we gain from using higher native instructions. There are some things though
3562 * that cannot be emulated. In that case enable the extensions.
3563 * If the extension is enabled, instruction handlers that support both ways will use it.
3565 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3566 * So enable the best we can get.
3568 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3569 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3571 want_nv_prog
= TRUE
;
3574 shader_addline(buffer
, "!!ARBfp1.0\n");
3575 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3577 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3578 priv_ctx
.target_version
= NV3
;
3580 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3582 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3583 priv_ctx
.target_version
= NV2
;
3587 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3590 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3593 priv_ctx
.target_version
= ARB
;
3596 if (reg_maps
->rt_mask
> 1)
3598 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3601 if (reg_maps
->shader_version
.major
< 3)
3603 switch (args
->super
.fog
)
3605 case WINED3D_FFP_PS_FOG_OFF
:
3607 case WINED3D_FFP_PS_FOG_LINEAR
:
3608 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
3610 custom_linear_fog
= TRUE
;
3611 priv_ctx
.ps_post_process
= TRUE
;
3614 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3616 case WINED3D_FFP_PS_FOG_EXP
:
3617 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3619 case WINED3D_FFP_PS_FOG_EXP2
:
3620 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3625 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3626 * unused temps away(but occupies them for the whole shader if they're used once). Always
3627 * declaring them avoids tricky bookkeeping work
3629 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3630 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3631 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3632 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3633 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3634 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3635 wined3d_ftoa(eps
, ftoa_tmp
);
3636 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %s, 0.0 };\n", ftoa_tmp
);
3638 if (reg_maps
->shader_version
.major
< 2)
3640 strcpy(fragcolor
, "R0");
3644 if (priv_ctx
.ps_post_process
)
3646 if (shader
->u
.ps
.color0_mov
)
3648 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3652 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3653 strcpy(fragcolor
, "TMP_COLOR");
3656 strcpy(fragcolor
, "result.color");
3660 if (args
->super
.srgb_correction
)
3662 shader_addline(buffer
, "PARAM srgb_consts0 = ");
3663 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const0
);
3664 shader_addline(buffer
, ";\n");
3665 shader_addline(buffer
, "PARAM srgb_consts1 = ");
3666 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const1
);
3667 shader_addline(buffer
, ";\n");
3670 /* Base Declarations */
3671 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3673 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3675 unsigned char bump_const
;
3677 if (!(map
& 1)) continue;
3679 bump_const
= compiled
->numbumpenvmatconsts
;
3680 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3681 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3682 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3683 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3685 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3686 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3687 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3688 * textures due to conditional NP2 restrictions)
3690 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3691 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3692 * their location is shader dependent anyway and they cannot be loaded globally.
3694 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3695 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3696 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3697 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3699 if (!(reg_maps
->luminanceparams
& (1u << i
)))
3702 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3703 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3704 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3707 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
3709 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3710 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
3712 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
3716 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3717 control_values
[0], control_values
[1], control_values
[2]);
3721 compiled
->int_consts
[i
] = next_local
;
3722 compiled
->num_int_consts
++;
3723 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3728 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3730 compiled
->ycorrection
= next_local
;
3731 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3735 shader_addline(buffer
, "TEMP vpos;\n");
3736 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3737 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3738 * ycorrection.z: 1.0
3739 * ycorrection.w: 0.0
3741 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3742 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3747 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3750 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3751 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3752 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3753 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3754 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3755 * shader compilation errors and the subsequent errors when drawing with this shader. */
3756 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3757 unsigned char cur_fixup_sampler
= 0;
3759 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3760 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3761 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3763 fixup
->offset
= next_local
;
3764 fixup
->super
.active
= 0;
3766 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
)
3768 if (!(map
& (1u << i
)))
3771 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
)
3773 fixup
->super
.active
|= (1u << i
);
3774 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3778 FIXME("No free constant found to load NP2 fixup data into shader. "
3779 "Sampling from this texture will probably look wrong.\n");
3784 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3785 if (fixup
->super
.num_consts
) {
3786 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3787 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3791 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3793 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3796 /* Base Shader Body */
3797 if (FAILED(shader_generate_code(shader
, buffer
, reg_maps
, &priv_ctx
, NULL
, NULL
)))
3800 if(args
->super
.srgb_correction
) {
3801 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3802 priv_ctx
.target_version
>= NV2
);
3805 if (custom_linear_fog
)
3806 arbfp_add_linear_fog(buffer
, fragcolor
, "TA");
3808 if(strcmp(fragcolor
, "result.color")) {
3809 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3811 shader_addline(buffer
, "END\n");
3813 /* TODO: change to resource.glObjectHandle or something like that */
3814 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3816 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3817 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3819 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3820 if (!shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
->buffer
))
3826 static int compare_sig(const struct wined3d_shader_signature
*sig1
, const struct wined3d_shader_signature
*sig2
)
3831 if (sig1
->element_count
!= sig2
->element_count
)
3832 return sig1
->element_count
< sig2
->element_count
? -1 : 1;
3834 for (i
= 0; i
< sig1
->element_count
; ++i
)
3836 const struct wined3d_shader_signature_element
*e1
, *e2
;
3838 e1
= &sig1
->elements
[i
];
3839 e2
= &sig2
->elements
[i
];
3841 if (!e1
->semantic_name
|| !e2
->semantic_name
)
3843 /* Compare pointers, not contents. One string is NULL (element
3844 * does not exist), the other one is not NULL. */
3845 if (e1
->semantic_name
!= e2
->semantic_name
)
3846 return e1
->semantic_name
< e2
->semantic_name
? -1 : 1;
3850 if ((ret
= strcmp(e1
->semantic_name
, e2
->semantic_name
)))
3852 if (e1
->semantic_idx
!= e2
->semantic_idx
)
3853 return e1
->semantic_idx
< e2
->semantic_idx
? -1 : 1;
3854 if (e1
->sysval_semantic
!= e2
->sysval_semantic
)
3855 return e1
->sysval_semantic
< e2
->sysval_semantic
? -1 : 1;
3856 if (e1
->component_type
!= e2
->component_type
)
3857 return e1
->component_type
< e2
->component_type
? -1 : 1;
3858 if (e1
->register_idx
!= e2
->register_idx
)
3859 return e1
->register_idx
< e2
->register_idx
? -1 : 1;
3860 if (e1
->mask
!= e2
->mask
)
3861 return e1
->mask
< e2
->mask
? -1 : 1;
3866 static void clone_sig(struct wined3d_shader_signature
*new, const struct wined3d_shader_signature
*sig
)
3871 new->element_count
= sig
->element_count
;
3872 new->elements
= heap_calloc(new->element_count
, sizeof(*new->elements
));
3873 for (i
= 0; i
< sig
->element_count
; ++i
)
3875 new->elements
[i
] = sig
->elements
[i
];
3877 if (!new->elements
[i
].semantic_name
)
3880 /* Clone the semantic string */
3881 name
= heap_alloc(strlen(sig
->elements
[i
].semantic_name
) + 1);
3882 strcpy(name
, sig
->elements
[i
].semantic_name
);
3883 new->elements
[i
].semantic_name
= name
;
3887 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature
*sig
)
3889 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3890 struct ps_signature
*found_sig
;
3894 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3895 TRACE("Found existing signature %u\n", found_sig
->idx
);
3896 return found_sig
->idx
;
3898 found_sig
= heap_alloc_zero(sizeof(*found_sig
));
3899 clone_sig(&found_sig
->sig
, sig
);
3900 found_sig
->idx
= priv
->ps_sig_number
++;
3901 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
3902 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
3904 ERR("Failed to insert program entry.\n");
3906 return found_sig
->idx
;
3909 static void init_output_registers(const struct wined3d_shader
*shader
,
3910 const struct wined3d_shader_signature
*ps_input_sig
,
3911 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
3914 static const char * const texcoords
[8] =
3916 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3917 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
3919 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
3920 * and varying 9 to result.color.secondary
3922 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
3924 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3925 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
3926 "result.color.primary", "result.color.secondary"
3931 TRACE("Pixel shader uses builtin varyings\n");
3932 /* Map builtins to builtins */
3933 for(i
= 0; i
< 8; i
++)
3935 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
3937 priv_ctx
->color_output
[0] = "result.color.primary";
3938 priv_ctx
->color_output
[1] = "result.color.secondary";
3939 priv_ctx
->fog_output
= "TMP_FOGCOORD";
3941 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
3942 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
3944 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
3946 if (!output
->semantic_name
)
3949 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
))
3951 TRACE("o%u is TMP_OUT\n", output
->register_idx
);
3952 if (!output
->semantic_idx
)
3953 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
3955 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3957 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
3959 TRACE("o%u is result.pointsize\n", output
->register_idx
);
3960 if (!output
->semantic_idx
)
3961 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
3963 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3965 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3967 TRACE("o%u is result.color.?, idx %u\n", output
->register_idx
, output
->semantic_idx
);
3968 if (!output
->semantic_idx
)
3969 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.primary";
3970 else if (output
->semantic_idx
== 1)
3971 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.secondary";
3972 else priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3974 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3976 TRACE("o%u is result.texcoord[%u]\n", output
->register_idx
, output
->semantic_idx
);
3977 if (output
->semantic_idx
>= 8)
3978 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3980 priv_ctx
->vs_output
[output
->register_idx
] = texcoords
[output
->semantic_idx
];
3982 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
3984 TRACE("o%u is result.fogcoord\n", output
->register_idx
);
3985 if (output
->semantic_idx
> 0)
3986 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3988 priv_ctx
->vs_output
[output
->register_idx
] = "result.fogcoord";
3992 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3998 TRACE("Pixel shader uses declared varyings\n");
4000 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4001 for(i
= 0; i
< 8; i
++)
4003 priv_ctx
->texcrd_output
[i
] = "TA";
4005 priv_ctx
->color_output
[0] = "TA";
4006 priv_ctx
->color_output
[1] = "TA";
4007 priv_ctx
->fog_output
= "TA";
4009 for (i
= 0; i
< ps_input_sig
->element_count
; ++i
)
4011 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[i
];
4013 if (!input
->semantic_name
)
4016 /* If a declared input register is not written by builtin arguments, don't write to it.
4017 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4019 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4020 * to TMP_OUT in any case
4022 if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4024 if (input
->semantic_idx
< 8)
4025 priv_ctx
->texcrd_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4027 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4029 if (input
->semantic_idx
< 2)
4030 priv_ctx
->color_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4032 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4034 if (!input
->semantic_idx
)
4035 priv_ctx
->fog_output
= decl_idx_to_string
[input
->register_idx
];
4042 if (!strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.primary")
4043 || !strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.secondary"))
4045 compiled
->need_color_unclamp
= TRUE
;
4049 /* Map declared to declared */
4050 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4052 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4054 /* Write unread output to TA to throw them away */
4055 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4057 if (!output
->semantic_name
)
4060 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
) && !output
->semantic_idx
)
4062 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4065 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
) && !output
->semantic_idx
)
4067 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4071 for (j
= 0; j
< ps_input_sig
->element_count
; ++j
)
4073 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[j
];
4075 if (!input
->semantic_name
)
4078 if (!strcmp(input
->semantic_name
, output
->semantic_name
)
4079 && input
->semantic_idx
== output
->semantic_idx
)
4081 priv_ctx
->vs_output
[output
->register_idx
] = decl_idx_to_string
[input
->register_idx
];
4083 if (!strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.primary")
4084 || !strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.secondary"))
4086 compiled
->need_color_unclamp
= TRUE
;
4093 /* Context activation is done by the caller. */
4094 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4095 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
4096 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4097 const struct wined3d_shader_signature
*ps_input_sig
)
4099 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4100 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4101 struct shader_arb_priv
*priv
= shader
->device
->shader_priv
;
4103 DWORD next_local
= 0;
4104 struct shader_arb_ctx_priv priv_ctx
;
4107 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4108 priv_ctx
.cur_vs_args
= args
;
4109 list_init(&priv_ctx
.control_frames
);
4110 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4112 /* Create the hw ARB shader */
4113 shader_addline(buffer
, "!!ARBvp1.0\n");
4115 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4116 * mesurable performance penalty, and we can always make use of it for clipplanes.
4118 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4120 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4121 priv_ctx
.target_version
= NV3
;
4122 shader_addline(buffer
, "ADDRESS aL;\n");
4124 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4126 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4127 priv_ctx
.target_version
= NV2
;
4128 shader_addline(buffer
, "ADDRESS aL;\n");
4130 priv_ctx
.target_version
= ARB
;
4133 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4135 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4136 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4139 wined3d_ftoa(eps
, ftoa_tmp
);
4140 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %s};\n", ftoa_tmp
);
4142 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4144 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4145 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4148 shader_addline(buffer
, "TEMP TA;\n");
4149 shader_addline(buffer
, "TEMP TB;\n");
4151 /* Base Declarations */
4152 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4153 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4155 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4157 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4158 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
4160 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
4164 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4165 control_values
[0], control_values
[1], control_values
[2]);
4169 compiled
->int_consts
[i
] = next_local
;
4170 compiled
->num_int_consts
++;
4171 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4176 /* We need a constant to fixup the final position */
4177 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4178 compiled
->pos_fixup
= next_local
++;
4180 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4181 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4182 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4183 * a replacement shader depend on the texcoord.w being set properly.
4185 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4186 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4187 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4188 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4189 * this can eat a number of instructions, so skip it unless this cap is set as well
4191 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4193 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4194 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4196 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !priv
->ffp_proj_control
)
4199 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4200 for(i
= 0; i
< MAX_REG_TEXCRD
; i
++)
4202 if (reg_maps
->u
.texcoord_mask
[i
] && reg_maps
->u
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
)
4203 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4208 /* The shader starts with the main function */
4209 priv_ctx
.in_main_func
= TRUE
;
4210 /* Base Shader Body */
4211 if (FAILED(shader_generate_code(shader
, buffer
, reg_maps
, &priv_ctx
, NULL
, NULL
)))
4214 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4215 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4217 shader_addline(buffer
, "END\n");
4219 /* TODO: change to resource.glObjectHandle or something like that */
4220 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4222 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4223 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4225 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4226 if (!shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, buffer
->buffer
))
4232 /* Context activation is done by the caller. */
4233 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_context
*context
,
4234 struct wined3d_shader
*shader
, const struct arb_ps_compile_args
*args
)
4236 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4237 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4238 struct wined3d_device
*device
= shader
->device
;
4241 struct arb_ps_compiled_shader
*new_array
;
4242 struct wined3d_string_buffer buffer
;
4243 struct arb_pshader_private
*shader_data
;
4246 if (!shader
->backend_data
)
4248 struct shader_arb_priv
*priv
= device
->shader_priv
;
4250 shader
->backend_data
= heap_alloc_zero(sizeof(*shader_data
));
4251 shader_data
= shader
->backend_data
;
4252 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4254 if (shader
->reg_maps
.shader_version
.major
< 3)
4255 shader_data
->input_signature_idx
= ~0U;
4257 shader_data
->input_signature_idx
= find_input_signature(priv
, &shader
->input_signature
);
4259 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4261 if (!d3d_info
->vs_clipping
)
4262 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4263 d3d_info
->limits
.ffp_blend_stages
- 1);
4265 shader_data
->clipplane_emulation
= ~0U;
4267 shader_data
= shader
->backend_data
;
4269 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4270 * so a linear search is more performant than a hashmap or a binary search
4271 * (cache coherency etc)
4273 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4275 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4276 return &shader_data
->gl_shaders
[i
];
4279 TRACE("No matching GL shader found, compiling a new shader\n");
4280 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4281 if (shader_data
->num_gl_shaders
)
4283 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4284 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4285 new_size
* sizeof(*shader_data
->gl_shaders
));
4289 new_array
= heap_alloc_zero(sizeof(*shader_data
->gl_shaders
));
4294 ERR("Out of memory\n");
4297 shader_data
->gl_shaders
= new_array
;
4298 shader_data
->shader_array_size
= new_size
;
4301 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4303 pixelshader_update_resource_types(shader
, args
->super
.tex_types
);
4305 if (!string_buffer_init(&buffer
))
4307 ERR("Failed to initialize shader buffer.\n");
4311 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4312 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4313 string_buffer_free(&buffer
);
4314 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4316 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4319 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4320 const DWORD use_map
, BOOL skip_int
) {
4321 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4322 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4323 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4324 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4325 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4326 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4327 if(skip_int
) return TRUE
;
4329 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4332 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4333 const struct wined3d_gl_info
*gl_info
, DWORD use_map
, const struct arb_vs_compile_args
*args
,
4334 const struct wined3d_shader_signature
*ps_input_sig
)
4338 struct arb_vs_compiled_shader
*new_array
;
4339 struct wined3d_string_buffer buffer
;
4340 struct arb_vshader_private
*shader_data
;
4343 if (!shader
->backend_data
)
4345 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4347 shader
->backend_data
= heap_alloc_zero(sizeof(*shader_data
));
4348 shader_data
= shader
->backend_data
;
4350 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4351 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4353 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4355 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4356 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4357 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4359 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4360 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4361 else if (reg_maps
->max_rel_offset
> 63)
4362 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4365 shader_data
= shader
->backend_data
;
4367 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4368 * so a linear search is more performant than a hashmap or a binary search
4369 * (cache coherency etc)
4371 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4372 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4373 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4375 return &shader_data
->gl_shaders
[i
];
4379 TRACE("No matching GL shader found, compiling a new shader\n");
4381 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4382 if (shader_data
->num_gl_shaders
)
4384 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4385 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4386 new_size
* sizeof(*shader_data
->gl_shaders
));
4390 new_array
= heap_alloc_zero(sizeof(*shader_data
->gl_shaders
));
4395 ERR("Out of memory\n");
4398 shader_data
->gl_shaders
= new_array
;
4399 shader_data
->shader_array_size
= new_size
;
4402 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4404 if (!string_buffer_init(&buffer
))
4406 ERR("Failed to initialize shader buffer.\n");
4410 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4411 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4413 string_buffer_free(&buffer
);
4414 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4416 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4419 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4420 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4421 struct arb_ps_compile_args
*args
)
4423 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4424 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4428 find_ps_compile_args(state
, shader
, context
->stream_info
.position_transformed
, &args
->super
, context
);
4430 /* This forces all local boolean constants to 1 to make them stateblock independent */
4431 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4433 for (i
= 0; i
< WINED3D_MAX_CONSTS_B
; ++i
)
4435 if (state
->ps_consts_b
[i
])
4436 args
->bools
|= ( 1u << i
);
4439 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4440 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4441 * duplicate the shader than have a no-op KIL instruction in every shader
4443 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4444 && state
->render_states
[WINED3D_RS_CLIPPING
]
4445 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4450 /* Skip if unused or local, or supported natively */
4451 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4452 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4454 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4458 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4460 if (int_skip
& (1u << i
))
4462 args
->loop_ctrl
[i
][0] = 0;
4463 args
->loop_ctrl
[i
][1] = 0;
4464 args
->loop_ctrl
[i
][2] = 0;
4468 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
].x
;
4469 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
].y
;
4470 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
].z
;
4475 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4476 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4477 struct arb_vs_compile_args
*args
)
4479 const struct wined3d_device
*device
= shader
->device
;
4480 const struct wined3d_adapter
*adapter
= device
->adapter
;
4481 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4482 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4486 find_vs_compile_args(state
, shader
, context
->stream_info
.swizzle_map
, &args
->super
, context
);
4488 args
->clip
.boolclip_compare
= 0;
4491 const struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4492 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4493 args
->ps_signature
= shader_priv
->input_signature_idx
;
4495 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4499 args
->ps_signature
= ~0;
4500 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4501 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(context
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4502 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4505 if (args
->clip
.boolclip
.clip_texcoord
)
4507 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4508 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4509 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4512 /* This forces all local boolean constants to 1 to make them stateblock independent */
4513 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4514 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4515 for (i
= 0; i
< WINED3D_MAX_CONSTS_B
; ++i
)
4517 if (state
->vs_consts_b
[i
])
4518 args
->clip
.boolclip
.bools
|= (1u << i
);
4521 args
->vertex
.samplers
[0] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 0];
4522 args
->vertex
.samplers
[1] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 1];
4523 args
->vertex
.samplers
[2] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 2];
4524 args
->vertex
.samplers
[3] = 0;
4526 /* Skip if unused or local */
4527 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4528 /* This is about flow control, not clipping. */
4529 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4531 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4535 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4537 if (int_skip
& (1u << i
))
4539 args
->loop_ctrl
[i
][0] = 0;
4540 args
->loop_ctrl
[i
][1] = 0;
4541 args
->loop_ctrl
[i
][2] = 0;
4545 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
].x
;
4546 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
].y
;
4547 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
].z
;
4552 /* Context activation is done by the caller. */
4553 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
4554 const struct wined3d_state
*state
)
4556 struct shader_arb_priv
*priv
= shader_priv
;
4557 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4560 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4563 struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4564 struct arb_ps_compile_args compile_args
;
4565 struct arb_ps_compiled_shader
*compiled
;
4567 TRACE("Using pixel shader %p.\n", ps
);
4568 find_arb_ps_compile_args(state
, context
, ps
, &compile_args
);
4569 compiled
= find_arb_pshader(context
, ps
, &compile_args
);
4570 priv
->current_fprogram_id
= compiled
->prgId
;
4571 priv
->compiled_fprog
= compiled
;
4573 /* Bind the fragment program */
4574 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4575 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4577 if (!priv
->use_arbfp_fixed_func
)
4578 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4580 /* Enable OpenGL fragment programs. */
4581 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4582 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4584 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", priv
->current_fprogram_id
);
4586 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4587 * a 1.x and newer shader, reload the first 8 constants
4589 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4591 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4592 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4593 for(i
= 0; i
< 8; i
++)
4595 priv
->pshader_const_dirty
[i
] = 1;
4597 /* Also takes care of loading local constants */
4598 shader_arb_load_constants_internal(shader_priv
, context
, state
, TRUE
, FALSE
, TRUE
);
4602 UINT rt_height
= state
->fb
->render_targets
[0]->height
;
4603 shader_arb_ps_local_constants(compiled
, context
, state
, rt_height
);
4606 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4607 if (compiled
->np2fixup_info
.super
.active
)
4608 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_NP2_FIXUP
;
4610 if (ps
->load_local_constsF
)
4611 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
4615 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4617 /* Disable only if we're not using arbfp fixed function fragment
4618 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4619 * enabled, and the fixed function pipeline will bind the fixed
4620 * function replacement shader. */
4621 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4622 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4623 priv
->current_fprogram_id
= 0;
4625 priv
->fragment_pipe
->enable_extension(gl_info
, TRUE
);
4630 struct wined3d_shader
*vs
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
4631 struct arb_vs_compile_args compile_args
;
4632 struct arb_vs_compiled_shader
*compiled
;
4633 const struct wined3d_shader_signature
*ps_input_sig
;
4635 TRACE("Using vertex shader %p\n", vs
);
4636 find_arb_vs_compile_args(state
, context
, vs
, &compile_args
);
4638 /* Instead of searching for the signature in the signature list, read the one from the
4639 * current pixel shader. It's maybe not the shader where the signature came from, but it
4640 * is the same signature and faster to find. */
4641 if (compile_args
.ps_signature
== ~0U)
4642 ps_input_sig
= NULL
;
4644 ps_input_sig
= &state
->shader
[WINED3D_SHADER_TYPE_PIXEL
]->input_signature
;
4646 compiled
= find_arb_vshader(vs
, context
->gl_info
, context
->stream_info
.use_map
,
4647 &compile_args
, ps_input_sig
);
4648 priv
->current_vprogram_id
= compiled
->prgId
;
4649 priv
->compiled_vprog
= compiled
;
4651 /* Bind the vertex program */
4652 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4653 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4655 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4657 /* Enable OpenGL vertex programs */
4658 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4659 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4660 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", priv
->current_vprogram_id
);
4661 shader_arb_vs_local_constants(compiled
, context
, state
);
4663 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4664 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4666 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4668 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4669 checkGLcall("glClampColorARB");
4671 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4675 if (vs
->load_local_constsF
)
4676 context
->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
4680 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4682 priv
->current_vprogram_id
= 0;
4683 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4684 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4686 priv
->vertex_pipe
->vp_enable(gl_info
, TRUE
);
4690 static void shader_arb_select_compute(void *shader_priv
, struct wined3d_context
*context
,
4691 const struct wined3d_state
*state
)
4693 ERR("Compute pipeline not supported by the ARB shader backend.\n");
4696 /* Context activation is done by the caller. */
4697 static void shader_arb_disable(void *shader_priv
, struct wined3d_context
*context
)
4699 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4700 struct shader_arb_priv
*priv
= shader_priv
;
4702 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4704 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4705 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4706 priv
->current_fprogram_id
= 0;
4708 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4710 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4712 priv
->current_vprogram_id
= 0;
4713 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4714 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4716 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4718 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4720 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4721 checkGLcall("glClampColorARB");
4722 priv
->last_vs_color_unclamp
= FALSE
;
4725 context
->shader_update_mask
= (1u << WINED3D_SHADER_TYPE_PIXEL
)
4726 | (1u << WINED3D_SHADER_TYPE_VERTEX
)
4727 | (1u << WINED3D_SHADER_TYPE_GEOMETRY
)
4728 | (1u << WINED3D_SHADER_TYPE_HULL
)
4729 | (1u << WINED3D_SHADER_TYPE_DOMAIN
)
4730 | (1u << WINED3D_SHADER_TYPE_COMPUTE
);
4733 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4735 struct wined3d_device
*device
= shader
->device
;
4736 const struct wined3d_gl_info
*gl_info
;
4737 struct wined3d_context
*context
;
4740 /* This can happen if a shader was never compiled */
4741 if (!shader
->backend_data
)
4744 context
= context_acquire(device
, NULL
, 0);
4745 gl_info
= context
->gl_info
;
4747 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4749 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4751 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4752 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4754 heap_free(shader_data
->gl_shaders
);
4758 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4760 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4761 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4763 heap_free(shader_data
->gl_shaders
);
4766 checkGLcall("delete programs");
4768 context_release(context
);
4770 heap_free(shader
->backend_data
);
4771 shader
->backend_data
= NULL
;
4774 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4776 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4777 return compare_sig(key
, &e
->sig
);
4780 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4781 const struct fragment_pipeline
*fragment_pipe
)
4783 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4784 struct fragment_caps fragment_caps
;
4785 void *vertex_priv
, *fragment_priv
;
4786 struct shader_arb_priv
*priv
;
4788 if (!(priv
= heap_alloc_zero(sizeof(*priv
))))
4789 return E_OUTOFMEMORY
;
4791 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4793 ERR("Failed to initialize vertex pipe.\n");
4798 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4800 ERR("Failed to initialize fragment pipe.\n");
4801 vertex_pipe
->vp_free(device
);
4806 memset(priv
->vshader_const_dirty
, 1,
4807 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4808 memset(priv
->pshader_const_dirty
, 1,
4809 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4811 wine_rb_init(&priv
->signature_tree
, sig_tree_compare
);
4813 priv
->vertex_pipe
= vertex_pipe
;
4814 priv
->fragment_pipe
= fragment_pipe
;
4815 fragment_pipe
->get_caps(&device
->adapter
->gl_info
, &fragment_caps
);
4816 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
4818 device
->vertex_priv
= vertex_priv
;
4819 device
->fragment_priv
= fragment_priv
;
4820 device
->shader_priv
= priv
;
4825 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
4827 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4830 for (i
= 0; i
< sig
->sig
.element_count
; ++i
)
4832 heap_free((char *)sig
->sig
.elements
[i
].semantic_name
);
4834 heap_free(sig
->sig
.elements
);
4838 /* Context activation is done by the caller. */
4839 static void shader_arb_free(struct wined3d_device
*device
)
4841 struct shader_arb_priv
*priv
= device
->shader_priv
;
4843 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
4844 priv
->fragment_pipe
->free_private(device
);
4845 priv
->vertex_pipe
->vp_free(device
);
4846 heap_free(device
->shader_priv
);
4849 static BOOL
shader_arb_allocate_context_data(struct wined3d_context
*context
)
4854 static void shader_arb_free_context_data(struct wined3d_context
*context
)
4856 struct shader_arb_priv
*priv
;
4858 priv
= context
->device
->shader_priv
;
4859 if (priv
->last_context
== context
)
4860 priv
->last_context
= NULL
;
4863 static void shader_arb_init_context_state(struct wined3d_context
*context
) {}
4865 static void shader_arb_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
4867 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4872 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
4873 * for vertex programs. If the native limit is less than that it's
4874 * not very useful, and e.g. Mesa swrast returns 0, probably to
4875 * indicate it's a software implementation. */
4876 if (gl_info
->limits
.arb_vs_native_constants
< 96)
4877 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
4879 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
4881 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4884 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
4886 else if (vs_consts
>= 256)
4888 /* Shader Model 2.0 requires at least 256 vertex shader constants */
4890 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
4895 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
4897 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
4898 caps
->vs_uniform_count
= min(WINED3D_MAX_VS_CONSTS_F
, vs_consts
);
4902 caps
->vs_version
= 0;
4903 caps
->vs_uniform_count
= 0;
4906 caps
->hs_version
= 0;
4907 caps
->ds_version
= 0;
4908 caps
->gs_version
= 0;
4909 caps
->cs_version
= 0;
4911 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4916 /* Similar as above for vertex programs, but the minimum for fragment
4917 * programs is 24. */
4918 if (gl_info
->limits
.arb_ps_native_constants
< 24)
4919 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
4921 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
4923 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
4926 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
4928 else if (ps_consts
>= 32)
4930 /* Shader Model 2.0 requires at least 32 pixel shader constants */
4932 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
4937 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
4939 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
4940 caps
->ps_uniform_count
= min(WINED3D_MAX_PS_CONSTS_F
, ps_consts
);
4941 caps
->ps_1x_max_value
= 8.0f
;
4945 caps
->ps_version
= 0;
4946 caps
->ps_uniform_count
= 0;
4947 caps
->ps_1x_max_value
= 0.0f
;
4950 caps
->varying_count
= 0;
4951 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
4952 if (use_nv_clip(gl_info
))
4953 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
4956 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
4958 /* We support everything except complex conversions. */
4959 return !is_complex_fixup(fixup
);
4962 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
4964 char write_mask
[20], regstr
[50];
4965 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
4966 BOOL is_color
= FALSE
;
4967 const struct wined3d_shader_dst_param
*dst
;
4969 if (!ins
->dst_count
) return;
4973 if (!shift
) return; /* Saturate alone is handled by the instructions */
4975 shader_arb_get_write_mask(ins
, dst
, write_mask
);
4976 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
4978 /* Generate a line that does the output modifier computation
4979 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
4980 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
4982 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
4983 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
4986 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4988 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
4989 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
4990 /* WINED3DSIH_AND */ NULL
,
4991 /* WINED3DSIH_ATOMIC_AND */ NULL
,
4992 /* WINED3DSIH_ATOMIC_CMP_STORE */ NULL
,
4993 /* WINED3DSIH_ATOMIC_IADD */ NULL
,
4994 /* WINED3DSIH_ATOMIC_IMAX */ NULL
,
4995 /* WINED3DSIH_ATOMIC_IMIN */ NULL
,
4996 /* WINED3DSIH_ATOMIC_OR */ NULL
,
4997 /* WINED3DSIH_ATOMIC_UMAX */ NULL
,
4998 /* WINED3DSIH_ATOMIC_UMIN */ NULL
,
4999 /* WINED3DSIH_ATOMIC_XOR */ NULL
,
5000 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5001 /* WINED3DSIH_BFI */ NULL
,
5002 /* WINED3DSIH_BFREV */ NULL
,
5003 /* WINED3DSIH_BREAK */ shader_hw_break
,
5004 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5005 /* WINED3DSIH_BREAKP */ NULL
,
5006 /* WINED3DSIH_BUFINFO */ NULL
,
5007 /* WINED3DSIH_CALL */ shader_hw_call
,
5008 /* WINED3DSIH_CALLNZ */ NULL
,
5009 /* WINED3DSIH_CASE */ NULL
,
5010 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5011 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5012 /* WINED3DSIH_CONTINUE */ NULL
,
5013 /* WINED3DSIH_CONTINUEP */ NULL
,
5014 /* WINED3DSIH_COUNTBITS */ NULL
,
5015 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5016 /* WINED3DSIH_CUT */ NULL
,
5017 /* WINED3DSIH_CUT_STREAM */ NULL
,
5018 /* WINED3DSIH_DCL */ shader_hw_nop
,
5019 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5020 /* WINED3DSIH_DCL_FUNCTION_BODY */ NULL
,
5021 /* WINED3DSIH_DCL_FUNCTION_TABLE */ NULL
,
5022 /* WINED3DSIH_DCL_GLOBAL_FLAGS */ NULL
,
5023 /* WINED3DSIH_DCL_GS_INSTANCES */ NULL
,
5024 /* WINED3DSIH_DCL_HS_FORK_PHASE_INSTANCE_COUNT */ NULL
,
5025 /* WINED3DSIH_DCL_HS_JOIN_PHASE_INSTANCE_COUNT */ NULL
,
5026 /* WINED3DSIH_DCL_HS_MAX_TESSFACTOR */ NULL
,
5027 /* WINED3DSIH_DCL_IMMEDIATE_CONSTANT_BUFFER */ NULL
,
5028 /* WINED3DSIH_DCL_INDEX_RANGE */ NULL
,
5029 /* WINED3DSIH_DCL_INDEXABLE_TEMP */ NULL
,
5030 /* WINED3DSIH_DCL_INPUT */ NULL
,
5031 /* WINED3DSIH_DCL_INPUT_CONTROL_POINT_COUNT */ NULL
,
5032 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5033 /* WINED3DSIH_DCL_INPUT_PS */ NULL
,
5034 /* WINED3DSIH_DCL_INPUT_PS_SGV */ NULL
,
5035 /* WINED3DSIH_DCL_INPUT_PS_SIV */ NULL
,
5036 /* WINED3DSIH_DCL_INPUT_SGV */ NULL
,
5037 /* WINED3DSIH_DCL_INPUT_SIV */ NULL
,
5038 /* WINED3DSIH_DCL_INTERFACE */ NULL
,
5039 /* WINED3DSIH_DCL_OUTPUT */ NULL
,
5040 /* WINED3DSIH_DCL_OUTPUT_CONTROL_POINT_COUNT */ NULL
,
5041 /* WINED3DSIH_DCL_OUTPUT_SIV */ NULL
,
5042 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5043 /* WINED3DSIH_DCL_RESOURCE_RAW */ NULL
,
5044 /* WINED3DSIH_DCL_RESOURCE_STRUCTURED */ NULL
,
5045 /* WINED3DSIH_DCL_SAMPLER */ NULL
,
5046 /* WINED3DSIH_DCL_STREAM */ NULL
,
5047 /* WINED3DSIH_DCL_TEMPS */ NULL
,
5048 /* WINED3DSIH_DCL_TESSELLATOR_DOMAIN */ NULL
,
5049 /* WINED3DSIH_DCL_TESSELLATOR_OUTPUT_PRIMITIVE */ NULL
,
5050 /* WINED3DSIH_DCL_TESSELLATOR_PARTITIONING */ NULL
,
5051 /* WINED3DSIH_DCL_TGSM_RAW */ NULL
,
5052 /* WINED3DSIH_DCL_TGSM_STRUCTURED */ NULL
,
5053 /* WINED3DSIH_DCL_THREAD_GROUP */ NULL
,
5054 /* WINED3DSIH_DCL_UAV_RAW */ NULL
,
5055 /* WINED3DSIH_DCL_UAV_STRUCTURED */ NULL
,
5056 /* WINED3DSIH_DCL_UAV_TYPED */ NULL
,
5057 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5058 /* WINED3DSIH_DEF */ shader_hw_nop
,
5059 /* WINED3DSIH_DEFAULT */ NULL
,
5060 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5061 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5062 /* WINED3DSIH_DIV */ NULL
,
5063 /* WINED3DSIH_DP2 */ NULL
,
5064 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5065 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5066 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5067 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5068 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5069 /* WINED3DSIH_DSX_COARSE */ NULL
,
5070 /* WINED3DSIH_DSX_FINE */ NULL
,
5071 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5072 /* WINED3DSIH_DSY_COARSE */ NULL
,
5073 /* WINED3DSIH_DSY_FINE */ NULL
,
5074 /* WINED3DSIH_ELSE */ shader_hw_else
,
5075 /* WINED3DSIH_EMIT */ NULL
,
5076 /* WINED3DSIH_EMIT_STREAM */ NULL
,
5077 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5078 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5079 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5080 /* WINED3DSIH_ENDSWITCH */ NULL
,
5081 /* WINED3DSIH_EQ */ NULL
,
5082 /* WINED3DSIH_EVAL_SAMPLE_INDEX */ NULL
,
5083 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5084 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5085 /* WINED3DSIH_F16TOF32 */ NULL
,
5086 /* WINED3DSIH_F32TOF16 */ NULL
,
5087 /* WINED3DSIH_FCALL */ NULL
,
5088 /* WINED3DSIH_FIRSTBIT_HI */ NULL
,
5089 /* WINED3DSIH_FIRSTBIT_LO */ NULL
,
5090 /* WINED3DSIH_FIRSTBIT_SHI */ NULL
,
5091 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5092 /* WINED3DSIH_FTOI */ NULL
,
5093 /* WINED3DSIH_FTOU */ NULL
,
5094 /* WINED3DSIH_GATHER4 */ NULL
,
5095 /* WINED3DSIH_GATHER4_C */ NULL
,
5096 /* WINED3DSIH_GATHER4_PO */ NULL
,
5097 /* WINED3DSIH_GATHER4_PO_C */ NULL
,
5098 /* WINED3DSIH_GE */ NULL
,
5099 /* WINED3DSIH_HS_CONTROL_POINT_PHASE */ NULL
,
5100 /* WINED3DSIH_HS_DECLS */ NULL
,
5101 /* WINED3DSIH_HS_FORK_PHASE */ NULL
,
5102 /* WINED3DSIH_HS_JOIN_PHASE */ NULL
,
5103 /* WINED3DSIH_IADD */ NULL
,
5104 /* WINED3DSIH_IBFE */ NULL
,
5105 /* WINED3DSIH_IEQ */ NULL
,
5106 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5107 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5108 /* WINED3DSIH_IGE */ NULL
,
5109 /* WINED3DSIH_ILT */ NULL
,
5110 /* WINED3DSIH_IMAD */ NULL
,
5111 /* WINED3DSIH_IMAX */ NULL
,
5112 /* WINED3DSIH_IMIN */ NULL
,
5113 /* WINED3DSIH_IMM_ATOMIC_ALLOC */ NULL
,
5114 /* WINED3DSIH_IMM_ATOMIC_AND */ NULL
,
5115 /* WINED3DSIH_IMM_ATOMIC_CMP_EXCH */ NULL
,
5116 /* WINED3DSIH_IMM_ATOMIC_CONSUME */ NULL
,
5117 /* WINED3DSIH_IMM_ATOMIC_EXCH */ NULL
,
5118 /* WINED3DSIH_IMM_ATOMIC_IADD */ NULL
,
5119 /* WINED3DSIH_IMM_ATOMIC_IMAX */ NULL
,
5120 /* WINED3DSIH_IMM_ATOMIC_IMIN */ NULL
,
5121 /* WINED3DSIH_IMM_ATOMIC_OR */ NULL
,
5122 /* WINED3DSIH_IMM_ATOMIC_UMAX */ NULL
,
5123 /* WINED3DSIH_IMM_ATOMIC_UMIN */ NULL
,
5124 /* WINED3DSIH_IMM_ATOMIC_XOR */ NULL
,
5125 /* WINED3DSIH_IMUL */ NULL
,
5126 /* WINED3DSIH_INE */ NULL
,
5127 /* WINED3DSIH_INEG */ NULL
,
5128 /* WINED3DSIH_ISHL */ NULL
,
5129 /* WINED3DSIH_ISHR */ NULL
,
5130 /* WINED3DSIH_ITOF */ NULL
,
5131 /* WINED3DSIH_LABEL */ shader_hw_label
,
5132 /* WINED3DSIH_LD */ NULL
,
5133 /* WINED3DSIH_LD2DMS */ NULL
,
5134 /* WINED3DSIH_LD_RAW */ NULL
,
5135 /* WINED3DSIH_LD_STRUCTURED */ NULL
,
5136 /* WINED3DSIH_LD_UAV_TYPED */ NULL
,
5137 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5138 /* WINED3DSIH_LOD */ NULL
,
5139 /* WINED3DSIH_LOG */ shader_hw_scalar_op
,
5140 /* WINED3DSIH_LOGP */ shader_hw_scalar_op
,
5141 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5142 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5143 /* WINED3DSIH_LT */ NULL
,
5144 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5145 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5146 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5147 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5148 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5149 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5150 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5151 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5152 /* WINED3DSIH_MOV */ shader_hw_mov
,
5153 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5154 /* WINED3DSIH_MOVC */ NULL
,
5155 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5156 /* WINED3DSIH_NE */ NULL
,
5157 /* WINED3DSIH_NOP */ shader_hw_nop
,
5158 /* WINED3DSIH_NOT */ NULL
,
5159 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5160 /* WINED3DSIH_OR */ NULL
,
5161 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5162 /* WINED3DSIH_POW */ shader_hw_pow
,
5163 /* WINED3DSIH_RCP */ shader_hw_scalar_op
,
5164 /* WINED3DSIH_REP */ shader_hw_rep
,
5165 /* WINED3DSIH_RESINFO */ NULL
,
5166 /* WINED3DSIH_RET */ shader_hw_ret
,
5167 /* WINED3DSIH_RETP */ NULL
,
5168 /* WINED3DSIH_ROUND_NE */ NULL
,
5169 /* WINED3DSIH_ROUND_NI */ NULL
,
5170 /* WINED3DSIH_ROUND_PI */ NULL
,
5171 /* WINED3DSIH_ROUND_Z */ NULL
,
5172 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5173 /* WINED3DSIH_SAMPLE */ NULL
,
5174 /* WINED3DSIH_SAMPLE_B */ NULL
,
5175 /* WINED3DSIH_SAMPLE_C */ NULL
,
5176 /* WINED3DSIH_SAMPLE_C_LZ */ NULL
,
5177 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5178 /* WINED3DSIH_SAMPLE_INFO */ NULL
,
5179 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5180 /* WINED3DSIH_SAMPLE_POS */ NULL
,
5181 /* WINED3DSIH_SETP */ NULL
,
5182 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5183 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5184 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5185 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5186 /* WINED3DSIH_SQRT */ NULL
,
5187 /* WINED3DSIH_STORE_RAW */ NULL
,
5188 /* WINED3DSIH_STORE_STRUCTURED */ NULL
,
5189 /* WINED3DSIH_STORE_UAV_TYPED */ NULL
,
5190 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5191 /* WINED3DSIH_SWAPC */ NULL
,
5192 /* WINED3DSIH_SWITCH */ NULL
,
5193 /* WINED3DSIH_SYNC */ NULL
,
5194 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5195 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5196 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5197 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5198 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5199 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5200 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5201 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5202 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5203 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5204 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5205 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5206 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5207 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5208 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5209 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5210 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5211 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5212 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5213 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5214 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5215 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5216 /* WINED3DSIH_UBFE */ NULL
,
5217 /* WINED3DSIH_UDIV */ NULL
,
5218 /* WINED3DSIH_UGE */ NULL
,
5219 /* WINED3DSIH_ULT */ NULL
,
5220 /* WINED3DSIH_UMAX */ NULL
,
5221 /* WINED3DSIH_UMIN */ NULL
,
5222 /* WINED3DSIH_UMUL */ NULL
,
5223 /* WINED3DSIH_USHR */ NULL
,
5224 /* WINED3DSIH_UTOF */ NULL
,
5225 /* WINED3DSIH_XOR */ NULL
,
5228 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5229 const struct wined3d_shader
*shader
, DWORD idx
)
5231 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5232 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5233 const struct wined3d_shader_lconst
*constant
;
5235 WORD flag
= (1u << idx
);
5236 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5238 if (reg_maps
->local_bool_consts
& flag
)
5240 /* What good is an if(bool) with a hardcoded local constant? I don't know, but handle it */
5241 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5243 if (constant
->idx
== idx
)
5245 return constant
->value
[0];
5248 ERR("Local constant not found\n");
5253 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5254 else bools
= priv
->cur_ps_args
->bools
;
5255 return bools
& flag
;
5259 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5260 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5262 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5263 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5265 /* Integer constants can either be a local constant, or they can be stored in the shader
5266 * type specific compile args. */
5267 if (reg_maps
->local_int_consts
& (1u << idx
))
5269 const struct wined3d_shader_lconst
*constant
;
5271 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5273 if (constant
->idx
== idx
)
5275 loop_control
->count
= constant
->value
[0];
5276 loop_control
->start
= constant
->value
[1];
5277 /* Step is signed. */
5278 loop_control
->step
= (int)constant
->value
[2];
5282 /* If this happens the flag was set incorrectly */
5283 ERR("Local constant not found\n");
5284 loop_control
->count
= 0;
5285 loop_control
->start
= 0;
5286 loop_control
->step
= 0;
5290 switch (reg_maps
->shader_version
.type
)
5292 case WINED3D_SHADER_TYPE_VERTEX
:
5293 /* Count and aL start value are unsigned */
5294 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5295 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5296 /* Step is signed. */
5297 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5300 case WINED3D_SHADER_TYPE_PIXEL
:
5301 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5302 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5303 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5307 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5312 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5314 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
;
5315 struct wined3d_shader_dst_param
*dst_param
;
5316 struct recorded_instruction
*rec
;
5319 if (!(rec
= heap_alloc_zero(sizeof(*rec
))))
5321 ERR("Out of memory\n");
5326 if (!(dst_param
= heap_alloc(sizeof(*dst_param
))))
5328 *dst_param
= *ins
->dst
;
5329 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5331 if (!(rel_addr
= heap_alloc(sizeof(*rel_addr
))))
5333 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5334 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5336 rec
->ins
.dst
= dst_param
;
5338 if (!(src_param
= heap_calloc(ins
->src_count
, sizeof(*src_param
))))
5340 for (i
= 0; i
< ins
->src_count
; ++i
)
5342 src_param
[i
] = ins
->src
[i
];
5343 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5345 if (!(rel_addr
= heap_alloc(sizeof(*rel_addr
))))
5347 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5348 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5351 rec
->ins
.src
= src_param
;
5352 list_add_tail(list
, &rec
->entry
);
5356 ERR("Out of memory\n");
5359 heap_free((void *)dst_param
->reg
.idx
[0].rel_addr
);
5360 heap_free(dst_param
);
5364 for (i
= 0; i
< ins
->src_count
; ++i
)
5366 heap_free((void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5368 heap_free(src_param
);
5373 static void free_recorded_instruction(struct list
*list
)
5375 struct recorded_instruction
*rec_ins
, *entry2
;
5378 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5380 list_remove(&rec_ins
->entry
);
5381 if (rec_ins
->ins
.dst
)
5383 heap_free((void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5384 heap_free((void *)rec_ins
->ins
.dst
);
5386 if (rec_ins
->ins
.src
)
5388 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5390 heap_free((void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5392 heap_free((void *)rec_ins
->ins
.src
);
5398 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5400 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5401 struct control_frame
*control_frame
;
5403 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5405 struct list
*e
= list_head(&priv
->control_frames
);
5406 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5407 list_remove(&control_frame
->entry
);
5408 heap_free(control_frame
);
5411 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5413 /* Non-ifc ENDIFs were already handled previously. */
5414 struct list
*e
= list_head(&priv
->control_frames
);
5415 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5416 list_remove(&control_frame
->entry
);
5417 heap_free(control_frame
);
5421 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5422 SHADER_HANDLER hw_fct
;
5423 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5424 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5425 struct control_frame
*control_frame
;
5426 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5429 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5431 control_frame
= heap_alloc_zero(sizeof(*control_frame
));
5432 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5434 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5435 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5437 if(priv
->target_version
>= NV2
)
5439 control_frame
->no
.loop
= priv
->num_loops
++;
5444 /* Don't bother recording when we're in a not used if branch */
5450 if(!priv
->recording
)
5452 list_init(&priv
->record
);
5453 priv
->recording
= TRUE
;
5454 control_frame
->outer_loop
= TRUE
;
5455 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5456 return; /* Instruction is handled */
5458 /* Record this loop in the outer loop's recording */
5461 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5463 if(priv
->target_version
>= NV2
)
5465 /* Nothing to do. The control frame is popped after the HW instr handler */
5469 struct list
*e
= list_head(&priv
->control_frames
);
5470 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5471 list_remove(&control_frame
->entry
);
5473 if(control_frame
->outer_loop
)
5475 unsigned int iteration
;
5479 /* Turn off recording before playback */
5480 priv
->recording
= FALSE
;
5482 /* Move the recorded instructions to a separate list and get them out of the private data
5483 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5484 * be recorded again, thus priv->record might be overwritten
5487 list_move_tail(©
, &priv
->record
);
5488 list_init(&priv
->record
);
5490 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5492 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5493 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5494 control_frame
->loop_control
.step
);
5495 aL
= control_frame
->loop_control
.start
;
5499 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5502 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5504 struct recorded_instruction
*rec_ins
;
5505 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5508 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5512 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5515 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5517 shader_arb_handle_instruction(&rec_ins
->ins
);
5520 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5522 aL
+= control_frame
->loop_control
.step
;
5525 shader_addline(buffer
, "#end loop/rep\n");
5527 free_recorded_instruction(©
);
5528 heap_free(control_frame
);
5529 return; /* Instruction is handled */
5533 /* This is a nested loop. Proceed to the normal recording function */
5534 heap_free(control_frame
);
5541 record_instruction(&priv
->record
, ins
);
5546 if(ins
->handler_idx
== WINED3DSIH_IF
)
5548 control_frame
= heap_alloc_zero(sizeof(*control_frame
));
5549 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5550 control_frame
->type
= IF
;
5552 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5553 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5554 bool_const
= !bool_const
;
5555 if (!priv
->muted
&& !bool_const
)
5557 shader_addline(buffer
, "#if(FALSE){\n");
5559 control_frame
->muting
= TRUE
;
5561 else shader_addline(buffer
, "#if(TRUE) {\n");
5563 return; /* Instruction is handled */
5565 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5567 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5568 control_frame
= heap_alloc_zero(sizeof(*control_frame
));
5569 control_frame
->type
= IFC
;
5570 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5571 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5573 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5575 struct list
*e
= list_head(&priv
->control_frames
);
5576 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5578 if(control_frame
->type
== IF
)
5580 shader_addline(buffer
, "#} else {\n");
5581 if(!priv
->muted
&& !control_frame
->muting
)
5584 control_frame
->muting
= TRUE
;
5586 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5587 return; /* Instruction is handled. */
5589 /* In case of an ifc, generate a HW shader instruction */
5590 if (control_frame
->type
!= IFC
)
5591 ERR("Control frame does not match.\n");
5593 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5595 struct list
*e
= list_head(&priv
->control_frames
);
5596 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5598 if(control_frame
->type
== IF
)
5600 shader_addline(buffer
, "#} endif\n");
5601 if(control_frame
->muting
) priv
->muted
= FALSE
;
5602 list_remove(&control_frame
->entry
);
5603 heap_free(control_frame
);
5604 return; /* Instruction is handled */
5606 /* In case of an ifc, generate a HW shader instruction */
5607 if (control_frame
->type
!= IFC
)
5608 ERR("Control frame does not match.\n");
5613 pop_control_frame(ins
);
5617 /* Select handler */
5618 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5620 /* Unhandled opcode */
5623 FIXME("Backend can't handle opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
5628 pop_control_frame(ins
);
5630 shader_arb_add_instruction_modifiers(ins
);
5633 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5635 struct shader_arb_priv
*priv
= shader_priv
;
5637 return priv
->ffp_proj_control
;
5640 static void shader_arb_precompile(void *shader_priv
, struct wined3d_shader
*shader
) {}
5642 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5644 shader_arb_handle_instruction
,
5645 shader_arb_precompile
,
5647 shader_arb_select_compute
,
5649 shader_arb_update_float_vertex_constants
,
5650 shader_arb_update_float_pixel_constants
,
5651 shader_arb_load_constants
,
5655 shader_arb_allocate_context_data
,
5656 shader_arb_free_context_data
,
5657 shader_arb_init_context_state
,
5658 shader_arb_get_caps
,
5659 shader_arb_color_fixup_supported
,
5660 shader_arb_has_ffp_proj_control
,
5663 /* ARB_fragment_program fixed function pipeline replacement definitions */
5664 #define ARB_FFP_CONST_TFACTOR 0
5665 #define ARB_FFP_CONST_COLOR_KEY_LOW ((ARB_FFP_CONST_TFACTOR) + 1)
5666 #define ARB_FFP_CONST_COLOR_KEY_HIGH ((ARB_FFP_CONST_COLOR_KEY_LOW) + 1)
5667 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_COLOR_KEY_HIGH) + 1)
5668 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5669 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5670 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5672 struct arbfp_ffp_desc
5674 struct ffp_frag_desc parent
;
5678 /* Context activation is done by the caller. */
5679 static void arbfp_enable(const struct wined3d_gl_info
*gl_info
, BOOL enable
)
5683 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5684 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5688 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5689 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5693 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5695 struct shader_arb_priv
*priv
;
5697 /* Share private data between the shader backend and the pipeline
5698 * replacement, if both are the arb implementation. This is needed to
5699 * figure out whether ARBfp should be disabled if no pixel shader is bound
5701 if (shader_backend
== &arb_program_shader_backend
)
5703 else if (!(priv
= heap_alloc_zero(sizeof(*priv
))))
5706 wine_rb_init(&priv
->fragment_shaders
, wined3d_ffp_frag_program_key_compare
);
5707 priv
->use_arbfp_fixed_func
= TRUE
;
5712 /* Context activation is done by the caller. */
5713 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5715 const struct wined3d_gl_info
*gl_info
= context
;
5716 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5718 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5719 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5720 heap_free(entry_arb
);
5723 /* Context activation is done by the caller. */
5724 static void arbfp_free(struct wined3d_device
*device
)
5726 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5728 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &device
->adapter
->gl_info
);
5729 priv
->use_arbfp_fixed_func
= FALSE
;
5731 if (device
->shader_backend
!= &arb_program_shader_backend
)
5732 heap_free(device
->fragment_priv
);
5735 static void arbfp_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5737 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5738 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
5739 | WINED3D_FRAGMENT_CAP_COLOR_KEY
;
5740 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5741 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5742 WINED3DTEXOPCAPS_SELECTARG1
|
5743 WINED3DTEXOPCAPS_SELECTARG2
|
5744 WINED3DTEXOPCAPS_MODULATE4X
|
5745 WINED3DTEXOPCAPS_MODULATE2X
|
5746 WINED3DTEXOPCAPS_MODULATE
|
5747 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5748 WINED3DTEXOPCAPS_ADDSIGNED
|
5749 WINED3DTEXOPCAPS_ADD
|
5750 WINED3DTEXOPCAPS_SUBTRACT
|
5751 WINED3DTEXOPCAPS_ADDSMOOTH
|
5752 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5753 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5754 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5755 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5756 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5757 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5758 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5759 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5760 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5761 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5762 WINED3DTEXOPCAPS_MULTIPLYADD
|
5763 WINED3DTEXOPCAPS_LERP
|
5764 WINED3DTEXOPCAPS_BUMPENVMAP
|
5765 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5767 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5769 caps
->MaxTextureBlendStages
= MAX_TEXTURES
;
5770 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.samplers
[WINED3D_SHADER_TYPE_PIXEL
], MAX_TEXTURES
);
5773 static DWORD
arbfp_get_emul_mask(const struct wined3d_gl_info
*gl_info
)
5775 return GL_EXT_EMUL_ARB_MULTITEXTURE
| GL_EXT_EMUL_EXT_FOG_COORD
;
5778 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5779 const struct wined3d_state
*state
, DWORD state_id
)
5781 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5782 struct wined3d_device
*device
= context
->device
;
5783 struct wined3d_color color
;
5785 if (device
->shader_backend
== &arb_program_shader_backend
)
5787 struct shader_arb_priv
*priv
;
5789 /* Don't load the parameter if we're using an arbfp pixel shader,
5790 * otherwise we'll overwrite application provided constants. */
5794 priv
= device
->shader_priv
;
5795 priv
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5796 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5799 wined3d_color_from_d3dcolor(&color
, state
->render_states
[WINED3D_RS_TEXTUREFACTOR
]);
5800 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, &color
.r
));
5801 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, &color.r)");
5804 static void state_tss_constant_arbfp(struct wined3d_context
*context
,
5805 const struct wined3d_state
*state
, DWORD state_id
)
5807 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5808 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5809 struct wined3d_device
*device
= context
->device
;
5810 struct wined3d_color color
;
5812 if (device
->shader_backend
== &arb_program_shader_backend
)
5814 struct shader_arb_priv
*priv
;
5816 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5817 * application provided constants.
5822 priv
= device
->shader_priv
;
5823 priv
->pshader_const_dirty
[ARB_FFP_CONST_CONSTANT(stage
)] = 1;
5824 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_CONSTANT(stage
) + 1);
5827 wined3d_color_from_d3dcolor(&color
, state
->texture_states
[stage
][WINED3D_TSS_CONSTANT
]);
5828 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_CONSTANT(stage
), &color
.r
));
5829 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_CONSTANT(stage), &color.r)");
5832 static void state_arb_specularenable(struct wined3d_context
*context
,
5833 const struct wined3d_state
*state
, DWORD state_id
)
5835 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5836 struct wined3d_device
*device
= context
->device
;
5839 if (device
->shader_backend
== &arb_program_shader_backend
)
5841 struct shader_arb_priv
*priv
;
5843 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5844 * application provided constants.
5849 priv
= device
->shader_priv
;
5850 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5851 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5854 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
5856 /* The specular color has no alpha */
5857 col
[0] = 1.0f
; col
[1] = 1.0f
;
5858 col
[2] = 1.0f
; col
[3] = 0.0f
;
5860 col
[0] = 0.0f
; col
[1] = 0.0f
;
5861 col
[2] = 0.0f
; col
[3] = 0.0f
;
5863 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5864 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5867 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5869 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5870 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5871 struct wined3d_device
*device
= context
->device
;
5874 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5876 if (device
->shader_backend
== &arb_program_shader_backend
)
5878 struct shader_arb_priv
*priv
= device
->shader_priv
;
5880 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5884 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5885 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5888 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
5889 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
5890 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
5891 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
5893 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5894 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5897 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
5898 const struct wined3d_state
*state
, DWORD state_id
)
5900 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5901 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5902 struct wined3d_device
*device
= context
->device
;
5905 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5907 if (device
->shader_backend
== &arb_program_shader_backend
)
5909 struct shader_arb_priv
*priv
= device
->shader_priv
;
5911 /* Exit now, don't set the luminance below, otherwise we may overwrite pixel shader constants. */
5915 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5916 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5919 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
5920 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
5924 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5925 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5928 static void alpha_test_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5930 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5934 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
5936 if (state
->render_states
[WINED3D_RS_ALPHATESTENABLE
])
5938 gl_info
->gl_ops
.gl
.p_glEnable(GL_ALPHA_TEST
);
5939 checkGLcall("glEnable GL_ALPHA_TEST");
5943 gl_info
->gl_ops
.gl
.p_glDisable(GL_ALPHA_TEST
);
5944 checkGLcall("glDisable GL_ALPHA_TEST");
5948 ref
= ((float)state
->render_states
[WINED3D_RS_ALPHAREF
]) / 255.0f
;
5949 glParm
= wined3d_gl_compare_func(state
->render_states
[WINED3D_RS_ALPHAFUNC
]);
5953 gl_info
->gl_ops
.gl
.p_glAlphaFunc(glParm
, ref
);
5954 checkGLcall("glAlphaFunc");
5958 static void color_key_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5960 const struct wined3d_texture
*texture
= state
->textures
[0];
5961 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5962 struct wined3d_device
*device
= context
->device
;
5963 struct wined3d_color float_key
[2];
5968 if (device
->shader_backend
== &arb_program_shader_backend
)
5970 struct shader_arb_priv
*priv
;
5972 /* Don't load the parameter if we're using an arbfp pixel shader,
5973 * otherwise we'll overwrite application provided constants. */
5977 priv
= device
->shader_priv
;
5978 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_LOW
] = 1;
5979 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_HIGH
] = 1;
5980 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_COLOR_KEY_HIGH
+ 1);
5983 wined3d_format_get_float_color_key(texture
->resource
.format
, &texture
->async
.src_blt_color_key
, float_key
);
5985 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_LOW
, &float_key
[0].r
));
5986 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_LOW, &float_key[0].r)");
5987 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_HIGH
, &float_key
[1].r
));
5988 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_HIGH, &float_key[1].r)");
5991 static const char *get_argreg(struct wined3d_string_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
5995 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
5997 switch(arg
& WINED3DTA_SELECTMASK
) {
5998 case WINED3DTA_DIFFUSE
:
5999 ret
= "fragment.color.primary"; break;
6001 case WINED3DTA_CURRENT
:
6005 case WINED3DTA_TEXTURE
:
6007 case 0: ret
= "tex0"; break;
6008 case 1: ret
= "tex1"; break;
6009 case 2: ret
= "tex2"; break;
6010 case 3: ret
= "tex3"; break;
6011 case 4: ret
= "tex4"; break;
6012 case 5: ret
= "tex5"; break;
6013 case 6: ret
= "tex6"; break;
6014 case 7: ret
= "tex7"; break;
6015 default: ret
= "unknown texture";
6019 case WINED3DTA_TFACTOR
:
6020 ret
= "tfactor"; break;
6022 case WINED3DTA_SPECULAR
:
6023 ret
= "fragment.color.secondary"; break;
6025 case WINED3DTA_TEMP
:
6026 ret
= "tempreg"; break;
6028 case WINED3DTA_CONSTANT
:
6030 case 0: ret
= "const0"; break;
6031 case 1: ret
= "const1"; break;
6032 case 2: ret
= "const2"; break;
6033 case 3: ret
= "const3"; break;
6034 case 4: ret
= "const4"; break;
6035 case 5: ret
= "const5"; break;
6036 case 6: ret
= "const6"; break;
6037 case 7: ret
= "const7"; break;
6038 default: ret
= "unknown constant";
6046 if(arg
& WINED3DTA_COMPLEMENT
) {
6047 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
6048 if(argnum
== 0) ret
= "arg0";
6049 if(argnum
== 1) ret
= "arg1";
6050 if(argnum
== 2) ret
= "arg2";
6052 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
6053 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
6054 if(argnum
== 0) ret
= "arg0";
6055 if(argnum
== 1) ret
= "arg1";
6056 if(argnum
== 2) ret
= "arg2";
6061 static void gen_ffp_instr(struct wined3d_string_buffer
*buffer
, unsigned int stage
, BOOL color
,
6062 BOOL alpha
, BOOL tmp_dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
6064 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6065 unsigned int mul
= 1;
6074 dstreg
= tmp_dst
? "tempreg" : "ret";
6076 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6077 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6078 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6082 case WINED3D_TOP_DISABLE
:
6085 case WINED3D_TOP_SELECT_ARG2
:
6088 case WINED3D_TOP_SELECT_ARG1
:
6089 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6092 case WINED3D_TOP_MODULATE_4X
:
6095 case WINED3D_TOP_MODULATE_2X
:
6098 case WINED3D_TOP_MODULATE
:
6099 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6102 case WINED3D_TOP_ADD_SIGNED_2X
:
6105 case WINED3D_TOP_ADD_SIGNED
:
6106 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6109 case WINED3D_TOP_ADD
:
6110 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6113 case WINED3D_TOP_SUBTRACT
:
6114 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6117 case WINED3D_TOP_ADD_SMOOTH
:
6118 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6119 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6122 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6123 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6124 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6126 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6127 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6128 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6130 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6131 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6132 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6134 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6135 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6136 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6139 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6140 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6141 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6142 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6145 /* D3DTOP_PREMODULATE ???? */
6147 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6148 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6149 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6151 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6152 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6154 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6155 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6156 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6158 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6159 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6162 case WINED3D_TOP_DOTPRODUCT3
:
6164 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6165 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6166 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6169 case WINED3D_TOP_MULTIPLY_ADD
:
6170 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6173 case WINED3D_TOP_LERP
:
6174 /* The msdn is not quite right here */
6175 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6178 case WINED3D_TOP_BUMPENVMAP
:
6179 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6180 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6184 FIXME("Unhandled texture op %08x\n", op
);
6188 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6190 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6193 static const char *arbfp_texture_target(enum wined3d_gl_resource_type type
)
6197 case WINED3D_GL_RES_TYPE_TEX_1D
:
6199 case WINED3D_GL_RES_TYPE_TEX_2D
:
6201 case WINED3D_GL_RES_TYPE_TEX_3D
:
6203 case WINED3D_GL_RES_TYPE_TEX_CUBE
:
6205 case WINED3D_GL_RES_TYPE_TEX_RECT
:
6208 return "unexpected_resource_type";
6212 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6214 BYTE tex_read
= 0, bump_used
= 0, luminance_used
= 0, constant_used
= 0;
6215 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6216 unsigned int stage
, lowest_disabled_stage
;
6217 struct wined3d_string_buffer buffer
;
6218 struct color_fixup_masks masks
;
6219 BOOL custom_linear_fog
= FALSE
;
6220 const char *textype
, *instr
;
6221 DWORD arg0
, arg1
, arg2
;
6222 char colorcor_dst
[8];
6226 if (!string_buffer_init(&buffer
))
6228 ERR("Failed to initialize shader buffer.\n");
6232 shader_addline(&buffer
, "!!ARBfp1.0\n");
6234 if (settings
->color_key_enabled
)
6236 shader_addline(&buffer
, "PARAM color_key_low = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_LOW
);
6237 shader_addline(&buffer
, "PARAM color_key_high = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_HIGH
);
6241 /* Find out which textures are read */
6242 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6244 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6247 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6248 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6249 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6251 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
6252 tex_read
|= 1u << stage
;
6253 if (settings
->op
[stage
].tmp_dst
)
6254 tempreg_used
= TRUE
;
6255 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
6256 tempreg_used
= TRUE
;
6257 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
6258 tfactor_used
= TRUE
;
6259 if (arg0
== WINED3DTA_CONSTANT
|| arg1
== WINED3DTA_CONSTANT
|| arg2
== WINED3DTA_CONSTANT
)
6260 constant_used
|= 1u << stage
;
6262 switch (settings
->op
[stage
].cop
)
6264 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6265 luminance_used
|= 1u << stage
;
6267 case WINED3D_TOP_BUMPENVMAP
:
6268 bump_used
|= 1u << stage
;
6270 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6271 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6272 tex_read
|= 1u << stage
;
6275 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6276 tfactor_used
= TRUE
;
6283 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6286 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6287 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6288 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6290 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
6291 tex_read
|= 1u << stage
;
6292 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
6293 tempreg_used
= TRUE
;
6294 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
6295 tfactor_used
= TRUE
;
6296 if (arg0
== WINED3DTA_CONSTANT
|| arg1
== WINED3DTA_CONSTANT
|| arg2
== WINED3DTA_CONSTANT
)
6297 constant_used
|= 1u << stage
;
6299 lowest_disabled_stage
= stage
;
6301 switch (settings
->fog
)
6303 case WINED3D_FFP_PS_FOG_OFF
: break;
6304 case WINED3D_FFP_PS_FOG_LINEAR
:
6305 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
6307 custom_linear_fog
= TRUE
;
6310 shader_addline(&buffer
, "OPTION ARB_fog_linear;\n");
6313 case WINED3D_FFP_PS_FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6314 case WINED3D_FFP_PS_FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6315 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6318 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6319 shader_addline(&buffer
, "TEMP TMP;\n");
6320 shader_addline(&buffer
, "TEMP ret;\n");
6321 if (tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6322 shader_addline(&buffer
, "TEMP arg0;\n");
6323 shader_addline(&buffer
, "TEMP arg1;\n");
6324 shader_addline(&buffer
, "TEMP arg2;\n");
6325 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6327 if (constant_used
& (1u << stage
))
6328 shader_addline(&buffer
, "PARAM const%u = program.env[%u];\n", stage
, ARB_FFP_CONST_CONSTANT(stage
));
6330 if (!(tex_read
& (1u << stage
)))
6333 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6335 if (!(bump_used
& (1u << stage
)))
6337 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6339 if (!(luminance_used
& (1u << stage
)))
6341 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6344 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6345 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6347 if (settings
->sRGB_write
)
6349 shader_addline(&buffer
, "PARAM srgb_consts0 = ");
6350 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const0
);
6351 shader_addline(&buffer
, ";\n");
6352 shader_addline(&buffer
, "PARAM srgb_consts1 = ");
6353 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const1
);
6354 shader_addline(&buffer
, ";\n");
6357 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6358 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6360 if (tempreg_used
|| settings
->sRGB_write
)
6361 shader_addline(&buffer
, "MOV tempreg, 0.0;\n");
6363 /* Generate texture sampling instructions */
6364 for (stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6366 if (!(tex_read
& (1u << stage
)))
6369 textype
= arbfp_texture_target(settings
->op
[stage
].tex_type
);
6371 if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_NONE
)
6375 else if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT4
6376 || settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT3
)
6382 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6387 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6388 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6390 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6391 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6392 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6393 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6395 /* With projective textures, texbem only divides the static
6396 * texture coordinate, not the displacement, so multiply the
6397 * displacement with the dividing parameter before passing it to
6399 if (settings
->op
[stage
].projected
!= WINED3D_PROJECTION_NONE
)
6401 if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT4
)
6403 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6404 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n",
6409 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6410 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n",
6416 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6419 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6420 instr
, stage
, stage
, textype
);
6421 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6423 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6424 stage
- 1, stage
- 1, stage
- 1);
6425 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6428 else if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT3
)
6430 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6431 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6432 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6433 instr
, stage
, stage
, textype
);
6437 shader_addline(&buffer
, "%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6438 instr
, stage
, stage
, stage
, textype
);
6441 sprintf(colorcor_dst
, "tex%u", stage
);
6442 masks
= calc_color_correction(settings
->op
[stage
].color_fixup
, WINED3DSP_WRITEMASK_ALL
);
6443 gen_color_correction(&buffer
, colorcor_dst
, colorcor_dst
, "const.x", "const.y",
6444 settings
->op
[stage
].color_fixup
, masks
);
6447 if (settings
->color_key_enabled
)
6449 shader_addline(&buffer
, "SLT TMP, tex0, color_key_low;\n"); /* below low key */
6450 shader_addline(&buffer
, "SGE ret, tex0, color_key_high;\n"); /* above high key */
6451 shader_addline(&buffer
, "ADD TMP, TMP, ret;\n"); /* or */
6452 shader_addline(&buffer
, "DP4 TMP.b, TMP, TMP;\n"); /* on any channel */
6453 shader_addline(&buffer
, "SGE TMP, -TMP.b, 0.0;\n"); /* logical not */
6454 shader_addline(&buffer
, "KIL -TMP;\n"); /* discard if true */
6457 shader_addline(&buffer
, "MOV ret, fragment.color.primary;\n");
6459 /* Generate the main shader */
6460 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6462 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6465 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6466 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6467 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6468 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6469 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6470 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6471 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6472 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6473 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6474 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6475 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6476 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6478 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6479 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6480 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6481 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6483 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6485 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].tmp_dst
,
6486 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6487 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6491 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].tmp_dst
,
6492 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6493 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6495 else if (settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP
6496 && settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6498 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].tmp_dst
,
6499 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6500 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6501 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].tmp_dst
,
6502 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6503 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6507 if (settings
->sRGB_write
|| custom_linear_fog
)
6509 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, ret;\n");
6510 if (settings
->sRGB_write
)
6511 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6512 if (custom_linear_fog
)
6513 arbfp_add_linear_fog(&buffer
, "ret", "arg0");
6514 shader_addline(&buffer
, "MOV result.color, ret;\n");
6518 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, ret;\n");
6522 shader_addline(&buffer
, "END\n");
6524 /* Generate the shader */
6525 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6526 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6527 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
6529 string_buffer_free(&buffer
);
6533 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6535 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6536 const struct wined3d_device
*device
= context
->device
;
6537 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6538 BOOL use_pshader
= use_ps(state
);
6539 struct ffp_frag_settings settings
;
6540 const struct arbfp_ffp_desc
*desc
;
6543 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6545 if (isStateDirty(context
, STATE_RENDER(WINED3D_RS_FOGENABLE
)))
6547 if (!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6549 /* Reload fixed function constants since they collide with the
6550 * pixel shader constants. */
6551 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6553 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6554 state_tss_constant_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_CONSTANT
));
6556 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6557 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6558 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6560 else if (use_pshader
)
6562 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6569 /* Find or create a shader implementing the fixed function pipeline
6570 * settings, then activate it. */
6571 gen_ffp_frag_op(context
, state
, &settings
, FALSE
);
6572 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6575 struct arbfp_ffp_desc
*new_desc
;
6577 if (!(new_desc
= heap_alloc(sizeof(*new_desc
))))
6579 ERR("Out of memory\n");
6583 new_desc
->parent
.settings
= settings
;
6584 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6585 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6586 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6590 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6591 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6594 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6595 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6596 priv
->current_fprogram_id
= desc
->shader
;
6598 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6600 /* Reload fixed function constants since they collide with the
6601 * pixel shader constants. */
6602 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6604 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6605 state_tss_constant_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_CONSTANT
));
6607 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6608 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6609 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6611 context
->last_was_pshader
= FALSE
;
6613 else if (!context
->last_was_pshader
)
6615 if (device
->shader_backend
== &arb_program_shader_backend
)
6616 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
6617 context
->last_was_pshader
= TRUE
;
6620 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6623 /* We can't link the fog states to the fragment state directly since the
6624 * vertex pipeline links them to FOGENABLE. A different linking in different
6625 * pipeline parts can't be expressed in the combined state table, so we need
6626 * to handle that with a forwarding function. The other invisible side effect
6627 * is that changing the fog start and fog end (which links to FOGENABLE in
6628 * vertex) results in the fragment_prog_arbfp function being called because
6629 * FOGENABLE is dirty, which calls this function here. */
6630 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6632 enum fogsource new_source
;
6633 DWORD fogstart
= state
->render_states
[WINED3D_RS_FOGSTART
];
6634 DWORD fogend
= state
->render_states
[WINED3D_RS_FOGEND
];
6636 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6638 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6639 fragment_prog_arbfp(context
, state
, state_id
);
6641 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6644 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6648 new_source
= FOGSOURCE_VS
;
6652 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6653 new_source
= FOGSOURCE_COORD
;
6655 new_source
= FOGSOURCE_FFP
;
6660 new_source
= FOGSOURCE_FFP
;
6663 if (new_source
!= context
->fog_source
|| fogstart
== fogend
)
6665 context
->fog_source
= new_source
;
6666 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6670 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6672 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6673 fragment_prog_arbfp(context
, state
, state_id
);
6676 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] =
6678 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6679 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6680 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6681 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6682 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6683 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6684 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6685 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6686 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6687 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6688 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6689 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6690 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6691 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6692 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6693 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6694 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6695 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6696 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6697 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6698 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6699 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6700 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6701 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6702 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6703 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6704 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6705 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6706 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6707 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6708 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6709 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6710 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6711 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6712 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6713 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6714 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6715 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6716 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6717 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6718 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6719 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6720 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6721 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6722 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6723 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6724 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6725 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6726 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6727 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6728 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6729 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6730 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6731 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6732 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6733 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6734 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6735 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6736 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6737 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6738 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6739 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6740 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6741 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6742 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6743 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6744 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6745 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6746 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6747 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6748 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6749 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6750 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6751 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6752 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6753 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6754 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6755 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6756 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6757 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6758 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6759 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6760 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6761 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6762 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6763 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6764 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6765 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6766 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6767 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6768 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6769 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6770 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6771 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6772 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6773 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6774 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6775 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6776 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6777 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6778 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6779 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6780 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6781 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6782 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6783 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6784 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6785 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6786 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6787 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6788 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6789 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6790 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6791 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6792 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6793 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6794 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6795 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6796 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6797 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6798 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6799 {STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6800 {STATE_RENDER(WINED3D_RS_ALPHAFUNC
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6801 {STATE_RENDER(WINED3D_RS_ALPHAREF
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6802 {STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), alpha_test_arbfp
}, WINED3D_GL_EXT_NONE
},
6803 {STATE_RENDER(WINED3D_RS_COLORKEYENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6804 {STATE_COLOR_KEY
, { STATE_COLOR_KEY
, color_key_arbfp
}, WINED3D_GL_EXT_NONE
},
6805 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6806 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6807 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6808 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6809 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6810 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6811 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6812 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6813 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6814 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6815 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6816 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6817 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6818 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6819 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6820 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6821 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6822 {STATE_TEXTURESTAGE(0, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6823 {STATE_TEXTURESTAGE(1, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6824 {STATE_TEXTURESTAGE(2, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6825 {STATE_TEXTURESTAGE(3, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6826 {STATE_TEXTURESTAGE(4, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6827 {STATE_TEXTURESTAGE(5, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6828 {STATE_TEXTURESTAGE(6, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6829 {STATE_TEXTURESTAGE(7, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6830 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6831 {STATE_RENDER(WINED3D_RS_SHADEMODE
), { STATE_RENDER(WINED3D_RS_SHADEMODE
), state_shademode
}, WINED3D_GL_EXT_NONE
},
6832 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6835 static BOOL
arbfp_alloc_context_data(struct wined3d_context
*context
)
6840 static void arbfp_free_context_data(struct wined3d_context
*context
)
6844 const struct fragment_pipeline arbfp_fragment_pipeline
= {
6847 arbfp_get_emul_mask
,
6850 arbfp_alloc_context_data
,
6851 arbfp_free_context_data
,
6852 shader_arb_color_fixup_supported
,
6853 arbfp_fragmentstate_template
,
6856 struct arbfp_blit_type
6858 enum complex_fixup fixup
: 4;
6859 enum wined3d_gl_resource_type res_type
: 3;
6860 DWORD use_color_key
: 1;
6864 struct arbfp_blit_desc
6867 struct arbfp_blit_type type
;
6868 struct wine_rb_entry entry
;
6871 #define ARBFP_BLIT_PARAM_SIZE 0
6872 #define ARBFP_BLIT_PARAM_COLOR_KEY_LOW 1
6873 #define ARBFP_BLIT_PARAM_COLOR_KEY_HIGH 2
6875 struct wined3d_arbfp_blitter
6877 struct wined3d_blitter blitter
;
6878 struct wine_rb_tree shaders
;
6879 GLuint palette_texture
;
6882 static int arbfp_blit_type_compare(const void *key
, const struct wine_rb_entry
*entry
)
6884 const struct arbfp_blit_type
*ka
= key
;
6885 const struct arbfp_blit_type
*kb
= &WINE_RB_ENTRY_VALUE(entry
, const struct arbfp_blit_desc
, entry
)->type
;
6887 return memcmp(ka
, kb
, sizeof(*ka
));
6890 /* Context activation is done by the caller. */
6891 static void arbfp_free_blit_shader(struct wine_rb_entry
*entry
, void *ctx
)
6893 struct arbfp_blit_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
6894 const struct wined3d_gl_info
*gl_info
;
6895 struct wined3d_context
*context
;
6898 gl_info
= context
->gl_info
;
6900 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
6901 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
6902 heap_free(entry_arb
);
6905 /* Context activation is done by the caller. */
6906 static void arbfp_blitter_destroy(struct wined3d_blitter
*blitter
, struct wined3d_context
*context
)
6908 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6909 struct wined3d_arbfp_blitter
*arbfp_blitter
;
6910 struct wined3d_blitter
*next
;
6912 if ((next
= blitter
->next
))
6913 next
->ops
->blitter_destroy(next
, context
);
6915 arbfp_blitter
= CONTAINING_RECORD(blitter
, struct wined3d_arbfp_blitter
, blitter
);
6917 wine_rb_destroy(&arbfp_blitter
->shaders
, arbfp_free_blit_shader
, context
);
6918 checkGLcall("Delete blit programs");
6920 if (arbfp_blitter
->palette_texture
)
6921 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &arbfp_blitter
->palette_texture
);
6923 heap_free(arbfp_blitter
);
6926 static void gen_packed_yuv_read(struct wined3d_string_buffer
*buffer
,
6927 const struct arbfp_blit_type
*type
, char *luminance
)
6930 const char *tex
, *texinstr
= "TXP";
6932 if (type
->fixup
== COMPLEX_FIXUP_UYVY
)
6943 tex
= arbfp_texture_target(type
->res_type
);
6944 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_RECT
)
6947 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6948 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6949 * filtering when we sample the texture.
6951 * These are the rules for reading the chroma:
6957 * So we have to get the sampling x position in non-normalized coordinates in integers
6959 if (type
->res_type
!= WINED3D_GL_RES_TYPE_TEX_RECT
)
6961 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6962 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6966 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6968 /* We must not allow filtering between pixel x and x+1, this would mix U and V
6969 * Vertical filtering is ok. However, bear in mind that the pixel center is at
6972 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
6973 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
6975 /* Multiply the x coordinate by 0.5 and get the fraction. This gives 0.25
6976 * and 0.75 for the even and odd pixels respectively. */
6977 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
6978 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
6980 /* Sample Pixel 1. */
6981 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6983 /* Put the value into either of the chroma values */
6984 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6985 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
6986 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6987 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
6989 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
6990 * the pixel right to the current one. Otherwise, sample the left pixel.
6991 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
6993 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
6994 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
6995 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6997 /* Put the value into the other chroma */
6998 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6999 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
7000 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7001 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
7003 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
7004 * the current one and lerp the two U and V values
7007 /* This gives the correctly filtered luminance value */
7008 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
7011 static void gen_yv12_read(struct wined3d_string_buffer
*buffer
,
7012 const struct arbfp_blit_type
*type
, char *luminance
)
7015 static const float yv12_coef
[]
7016 = {2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
};
7018 tex
= arbfp_texture_target(type
->res_type
);
7020 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7021 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
7022 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
7023 * pitch of the luminance plane, the packing into the gl texture is a bit
7024 * unfortunate. If the whole texture is interpreted as luminance data it looks
7025 * approximately like this:
7027 * +----------------------------------+----
7039 * +----------------+-----------------+----
7041 * | V even rows | V odd rows |
7043 * +----------------+------------------ -
7045 * | U even rows | U odd rows |
7047 * +----------------+-----------------+----
7051 * So it appears as if there are 4 chroma images, but in fact the odd rows
7052 * in the chroma images are in the same row as the even ones. So it is
7053 * kinda tricky to read
7055 * When reading from rectangle textures, keep in mind that the input y coordinates
7056 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
7058 shader_addline(buffer
, "PARAM yv12_coef = ");
7059 shader_arb_append_imm_vec4(buffer
, yv12_coef
);
7060 shader_addline(buffer
, ";\n");
7062 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7063 /* the chroma planes have only half the width */
7064 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7066 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
7067 * the coordinate. Also read the right side of the image when reading odd lines
7069 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
7072 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7074 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
7076 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
7078 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
7079 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
7081 /* Read odd lines from the right side (add size * 0.5 to the x coordinate). */
7082 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7083 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7084 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7085 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7087 /* clamp, keep the half pixel origin in mind */
7088 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
7089 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7090 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
7091 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7095 /* The y coordinate for V is in the range [size, size + size / 4). */
7096 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7097 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
7099 /* Read odd lines from the right side (add size * 0.5 to the x coordinate). */
7100 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7101 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7102 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7103 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
7104 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7106 /* Make sure to read exactly from the pixel center */
7107 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7108 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
7111 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
7112 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7113 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7114 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7115 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7117 /* Read the texture, put the result into the output register */
7118 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7119 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7121 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
7122 * No need to clamp because we're just reusing the already clamped value from above
7124 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7125 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
7127 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
7128 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7129 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7131 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7132 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7133 * values due to filtering
7135 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7136 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7138 /* Multiply the y coordinate by 2/3 and clamp it */
7139 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
7140 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
7141 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7142 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7146 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7147 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7150 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7151 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7152 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7157 static void gen_nv12_read(struct wined3d_string_buffer
*buffer
,
7158 const struct arbfp_blit_type
*type
, char *luminance
)
7161 static const float nv12_coef
[]
7162 = {2.0f
/ 3.0f
, 1.0f
/ 3.0f
, 1.0f
, 1.0f
};
7164 tex
= arbfp_texture_target(type
->res_type
);
7166 /* NV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7167 * sized plane where each component is an UV pair. So the effective
7168 * bitdepth is 12 bits per pixel If the whole texture is interpreted as luminance
7169 * data it looks approximately like this:
7171 * +----------------------------------+----
7183 * +----------------------------------+----
7184 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7185 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7191 * +----------------------------------+----
7193 * When reading from rectangle textures, keep in mind that the input y coordinates
7194 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height. */
7196 shader_addline(buffer
, "PARAM nv12_coef = ");
7197 shader_arb_append_imm_vec4(buffer
, nv12_coef
);
7198 shader_addline(buffer
, ";\n");
7200 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7201 /* We only have half the number of chroma pixels. */
7202 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7204 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7206 shader_addline(buffer
, "RCP chroma.w, size.x;\n");
7207 shader_addline(buffer
, "RCP chroma.z, size.y;\n");
7209 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, nv12_coef.y, nv12_coef.x;\n");
7211 /* We must not allow filtering horizontally, this would mix U and V.
7212 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7213 * 0.5, so add 0.5. */
7215 /* Convert to non-normalized coordinates so we can find the
7216 * individual pixel. */
7217 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, size.x;\n");
7218 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7219 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7220 * add 0.5 to hit the center of the pixel. */
7221 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7223 /* Convert back to normalized coordinates. */
7224 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, chroma.w;\n");
7226 /* Clamp, keep the half pixel origin in mind. */
7227 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.z, nv12_coef.x;\n");
7228 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7229 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.z, nv12_coef.z;\n");
7230 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7234 /* The y coordinate for chroma is in the range [size, size + size / 2). */
7235 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.y, size.y;\n");
7237 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7238 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7239 * add 0.5 to hit the center of the pixel. */
7240 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7243 shader_addline(buffer
, "MAD temp.y, size.y, coef.y, size.y;\n");
7244 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7245 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7246 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7247 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7249 /* Read the texture, put the result into the output register. */
7250 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7251 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7253 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7256 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, chroma.w;\n");
7261 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.x;\n");
7263 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7264 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7266 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7267 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7268 * values due to filtering. */
7269 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7270 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7272 /* Multiply the y coordinate by 2/3 and clamp it */
7273 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, nv12_coef.x;\n");
7274 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, nv12_coef.x;\n");
7275 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7276 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7280 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7281 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7284 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7285 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7286 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7291 /* Context activation is done by the caller. */
7292 static GLuint
gen_p8_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7295 struct wined3d_string_buffer buffer
;
7296 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7298 /* This should not happen because we only use this conversion for
7299 * present blits which don't use color keying. */
7300 if (type
->use_color_key
)
7301 FIXME("Implement P8 color keying.\n");
7304 if (!string_buffer_init(&buffer
))
7306 ERR("Failed to initialize shader buffer.\n");
7310 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7311 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7314 string_buffer_free(&buffer
);
7318 shader_addline(&buffer
, "!!ARBfp1.0\n");
7319 shader_addline(&buffer
, "TEMP index;\n");
7321 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7322 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7324 /* The alpha-component contains the palette index */
7325 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7327 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7328 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7330 /* Use the alpha-component as an index in the palette to get the final color */
7331 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7332 shader_addline(&buffer
, "END\n");
7334 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7336 string_buffer_free(&buffer
);
7341 /* Context activation is done by the caller. */
7342 static void upload_palette(struct wined3d_arbfp_blitter
*blitter
,
7343 const struct wined3d_texture
*texture
, struct wined3d_context
*context
)
7345 const struct wined3d_palette
*palette
= texture
->swapchain
? texture
->swapchain
->palette
: NULL
;
7346 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7348 if (!blitter
->palette_texture
)
7349 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &blitter
->palette_texture
);
7351 GL_EXTCALL(glActiveTexture(GL_TEXTURE1
));
7352 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, blitter
->palette_texture
);
7354 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7356 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7357 /* Make sure we have discrete color levels. */
7358 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7359 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7360 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7363 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 256, 0, GL_BGRA
,
7364 GL_UNSIGNED_INT_8_8_8_8_REV
, palette
->colors
);
7368 static const DWORD black
;
7369 FIXME("P8 surface loaded without a palette.\n");
7370 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 1, 0, GL_BGRA
,
7371 GL_UNSIGNED_INT_8_8_8_8_REV
, &black
);
7374 /* Switch back to unit 0 in which the 2D texture will be stored. */
7375 context_active_texture(context
, gl_info
, 0);
7378 /* Context activation is done by the caller. */
7379 static GLuint
gen_yuv_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7382 struct wined3d_string_buffer buffer
;
7383 char luminance_component
;
7385 if (type
->use_color_key
)
7386 FIXME("Implement YUV color keying.\n");
7389 if (!string_buffer_init(&buffer
))
7391 ERR("Failed to initialize shader buffer.\n");
7395 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7396 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7397 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7398 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7401 string_buffer_free(&buffer
);
7405 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7406 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7407 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7408 * each single pixel it contains, and one U and one V value shared between both
7411 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7412 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7413 * take the format into account when generating the read swizzles
7415 * Reading the Y value is straightforward - just sample the texture. The hardware
7416 * takes care of filtering in the horizontal and vertical direction.
7418 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7419 * because that would mix the U and V values of one pixel or two adjacent pixels.
7420 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7421 * regardless of the filtering setting. Vertical filtering works automatically
7422 * though - the U and V values of two rows are mixed nicely.
7424 * Apart of avoiding filtering issues, the code has to know which value it just
7425 * read, and where it can find the other one. To determine this, it checks if
7426 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7428 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7429 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7431 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7432 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7433 * in an unfiltered situation. Finding the luminance on the other hand requires
7434 * finding out if it is an odd or even pixel. The real drawback of this approach
7435 * is filtering. This would have to be emulated completely in the shader, reading
7436 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7437 * vertically. Beyond that it would require adjustments to the texture handling
7438 * code to deal with the width scaling
7440 shader_addline(&buffer
, "!!ARBfp1.0\n");
7441 shader_addline(&buffer
, "TEMP luminance;\n");
7442 shader_addline(&buffer
, "TEMP temp;\n");
7443 shader_addline(&buffer
, "TEMP chroma;\n");
7444 shader_addline(&buffer
, "TEMP texcrd;\n");
7445 shader_addline(&buffer
, "TEMP texcrd2;\n");
7446 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7447 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7448 shader_addline(&buffer
, "PARAM size = program.local[%u];\n", ARBFP_BLIT_PARAM_SIZE
);
7450 switch (type
->fixup
)
7452 case COMPLEX_FIXUP_UYVY
:
7453 case COMPLEX_FIXUP_YUY2
:
7454 gen_packed_yuv_read(&buffer
, type
, &luminance_component
);
7457 case COMPLEX_FIXUP_YV12
:
7458 gen_yv12_read(&buffer
, type
, &luminance_component
);
7461 case COMPLEX_FIXUP_NV12
:
7462 gen_nv12_read(&buffer
, type
, &luminance_component
);
7466 FIXME("Unsupported YUV fixup %#x\n", type
->fixup
);
7467 string_buffer_free(&buffer
);
7471 /* Calculate the final result. Formula is taken from
7472 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7473 * ranges from -0.5 to 0.5
7475 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7477 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7478 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7479 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7480 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7481 shader_addline(&buffer
, "END\n");
7483 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7485 string_buffer_free(&buffer
);
7490 /* Context activation is done by the caller. */
7491 static GLuint
arbfp_gen_plain_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7494 struct wined3d_string_buffer buffer
;
7495 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7498 if (!string_buffer_init(&buffer
))
7500 ERR("Failed to initialize shader buffer.\n");
7504 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7507 string_buffer_free(&buffer
);
7510 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7512 shader_addline(&buffer
, "!!ARBfp1.0\n");
7514 if (type
->use_color_key
)
7516 shader_addline(&buffer
, "TEMP color;\n");
7517 shader_addline(&buffer
, "TEMP less, greater;\n");
7518 shader_addline(&buffer
, "PARAM color_key_low = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_LOW
);
7519 shader_addline(&buffer
, "PARAM color_key_high = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
);
7520 shader_addline(&buffer
, "TEX color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7521 shader_addline(&buffer
, "SLT less, color, color_key_low;\n"); /* below low key */
7522 shader_addline(&buffer
, "SGE greater, color, color_key_high;\n"); /* above high key */
7523 shader_addline(&buffer
, "ADD less, less, greater;\n"); /* or */
7524 shader_addline(&buffer
, "DP4 less.b, less, less;\n"); /* on any channel */
7525 shader_addline(&buffer
, "SGE less, -less.b, 0.0;\n"); /* logical not */
7526 shader_addline(&buffer
, "KIL -less;\n"); /* discard if true */
7527 shader_addline(&buffer
, "MOV result.color, color;\n");
7531 shader_addline(&buffer
, "TEX result.color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7534 shader_addline(&buffer
, "END\n");
7536 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7538 string_buffer_free(&buffer
);
7543 /* Context activation is done by the caller. */
7544 static HRESULT
arbfp_blit_set(struct wined3d_arbfp_blitter
*blitter
, struct wined3d_context
*context
,
7545 const struct wined3d_texture
*texture
, unsigned int sub_resource_idx
,
7546 const struct wined3d_color_key
*color_key
)
7548 enum complex_fixup fixup
;
7549 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7550 struct wine_rb_entry
*entry
;
7551 struct arbfp_blit_type type
;
7552 struct arbfp_blit_desc
*desc
;
7553 struct wined3d_color float_color_key
[2];
7554 struct wined3d_vec4 size
;
7558 level
= sub_resource_idx
% texture
->level_count
;
7559 size
.x
= wined3d_texture_get_level_pow2_width(texture
, level
);
7560 size
.y
= wined3d_texture_get_level_pow2_height(texture
, level
);
7564 if (is_complex_fixup(texture
->resource
.format
->color_fixup
))
7565 fixup
= get_complex_fixup(texture
->resource
.format
->color_fixup
);
7567 fixup
= COMPLEX_FIXUP_NONE
;
7569 switch (texture
->target
)
7572 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_1D
;
7576 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7580 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_3D
;
7583 case GL_TEXTURE_CUBE_MAP_ARB
:
7584 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_CUBE
;
7587 case GL_TEXTURE_RECTANGLE_ARB
:
7588 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_RECT
;
7592 ERR("Unexpected GL texture type %#x.\n", texture
->target
);
7593 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7596 type
.use_color_key
= !!color_key
;
7599 if ((entry
= wine_rb_get(&blitter
->shaders
, &type
)))
7601 desc
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7602 shader
= desc
->shader
;
7608 case COMPLEX_FIXUP_NONE
:
7609 if (!is_identity_fixup(texture
->resource
.format
->color_fixup
))
7610 FIXME("Implement support for sign or swizzle fixups.\n");
7611 shader
= arbfp_gen_plain_shader(gl_info
, &type
);
7614 case COMPLEX_FIXUP_P8
:
7615 shader
= gen_p8_shader(gl_info
, &type
);
7618 case COMPLEX_FIXUP_YUY2
:
7619 case COMPLEX_FIXUP_UYVY
:
7620 case COMPLEX_FIXUP_YV12
:
7621 case COMPLEX_FIXUP_NV12
:
7622 shader
= gen_yuv_shader(gl_info
, &type
);
7628 FIXME("Unsupported complex fixup %#x, not setting a shader\n", fixup
);
7632 if (!(desc
= heap_alloc(sizeof(*desc
))))
7636 desc
->shader
= shader
;
7637 if (wine_rb_put(&blitter
->shaders
, &desc
->type
, &desc
->entry
) == -1)
7640 ERR("Out of memory\n");
7641 GL_EXTCALL(glDeleteProgramsARB(1, &shader
));
7642 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader))");
7643 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, 0));
7644 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, 0)");
7646 return E_OUTOFMEMORY
;
7650 if (fixup
== COMPLEX_FIXUP_P8
)
7651 upload_palette(blitter
, texture
, context
);
7653 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7654 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7655 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7656 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7657 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARBFP_BLIT_PARAM_SIZE
, &size
.x
));
7658 checkGLcall("glProgramLocalParameter4fvARB");
7659 if (type
.use_color_key
)
7661 wined3d_format_get_float_color_key(texture
->resource
.format
, color_key
, float_color_key
);
7662 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7663 ARBFP_BLIT_PARAM_COLOR_KEY_LOW
, &float_color_key
[0].r
));
7664 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7665 ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
, &float_color_key
[1].r
));
7666 checkGLcall("glProgramLocalParameter4fvARB");
7672 /* Context activation is done by the caller. */
7673 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7675 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7676 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7679 static BOOL
arbfp_blit_supported(enum wined3d_blit_op blit_op
, const struct wined3d_context
*context
,
7680 const struct wined3d_resource
*src_resource
, DWORD src_location
,
7681 const struct wined3d_resource
*dst_resource
, DWORD dst_location
)
7683 const struct wined3d_format
*src_format
= src_resource
->format
;
7684 const struct wined3d_format
*dst_format
= dst_resource
->format
;
7685 enum complex_fixup src_fixup
;
7688 if (src_resource
->type
!= WINED3D_RTYPE_TEXTURE_2D
)
7691 if (blit_op
== WINED3D_BLIT_OP_RAW_BLIT
&& dst_format
->id
== src_format
->id
)
7693 if (dst_format
->flags
[WINED3D_GL_RES_TYPE_TEX_2D
] & (WINED3DFMT_FLAG_DEPTH
| WINED3DFMT_FLAG_STENCIL
))
7694 blit_op
= WINED3D_BLIT_OP_DEPTH_BLIT
;
7696 blit_op
= WINED3D_BLIT_OP_COLOR_BLIT
;
7701 case WINED3D_BLIT_OP_COLOR_BLIT_CKEY
:
7702 if (!context
->d3d_info
->shader_color_key
)
7704 /* The conversion modifies the alpha channel so the color key might no longer match. */
7705 TRACE("Color keying not supported with converted textures.\n");
7708 case WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
:
7709 case WINED3D_BLIT_OP_COLOR_BLIT
:
7713 TRACE("Unsupported blit_op=%d\n", blit_op
);
7717 decompress
= src_format
&& (src_format
->flags
[WINED3D_GL_RES_TYPE_TEX_2D
] & WINED3DFMT_FLAG_COMPRESSED
)
7718 && !(dst_format
->flags
[WINED3D_GL_RES_TYPE_TEX_2D
] & WINED3DFMT_FLAG_COMPRESSED
);
7719 if (!decompress
&& !(src_resource
->access
& dst_resource
->access
& WINED3D_RESOURCE_ACCESS_GPU
))
7722 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7723 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7725 TRACE("Checking support for fixup:\n");
7726 dump_color_fixup_desc(src_format
->color_fixup
);
7729 if (!is_identity_fixup(dst_format
->color_fixup
)
7730 && (dst_format
->id
!= src_format
->id
|| dst_location
!= WINED3D_LOCATION_DRAWABLE
))
7732 TRACE("Destination fixups are not supported\n");
7736 if (is_identity_fixup(src_format
->color_fixup
))
7742 /* We only support YUV conversions. */
7743 if (!is_complex_fixup(src_format
->color_fixup
))
7745 if (wined3d_settings
.offscreen_rendering_mode
== ORM_BACKBUFFER
)
7747 WARN("Claiming fixup support because of ORM_BACKBUFFER.\n");
7751 TRACE("[FAILED]\n");
7757 case COMPLEX_FIXUP_YUY2
:
7758 case COMPLEX_FIXUP_UYVY
:
7759 case COMPLEX_FIXUP_YV12
:
7760 case COMPLEX_FIXUP_NV12
:
7761 case COMPLEX_FIXUP_P8
:
7766 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7767 TRACE("[FAILED]\n");
7772 static DWORD
arbfp_blitter_blit(struct wined3d_blitter
*blitter
, enum wined3d_blit_op op
,
7773 struct wined3d_context
*context
, struct wined3d_texture
*src_texture
, unsigned int src_sub_resource_idx
,
7774 DWORD src_location
, const RECT
*src_rect
, struct wined3d_texture
*dst_texture
,
7775 unsigned int dst_sub_resource_idx
, DWORD dst_location
, const RECT
*dst_rect
,
7776 const struct wined3d_color_key
*color_key
, enum wined3d_texture_filter_type filter
)
7778 struct wined3d_device
*device
= dst_texture
->resource
.device
;
7779 struct wined3d_texture
*staging_texture
= NULL
;
7780 struct wined3d_arbfp_blitter
*arbfp_blitter
;
7781 struct wined3d_color_key alpha_test_key
;
7782 struct wined3d_blitter
*next
;
7783 unsigned int src_level
;
7786 TRACE("blitter %p, op %#x, context %p, src_texture %p, src_sub_resource_idx %u, src_location %s, src_rect %s, "
7787 "dst_texture %p, dst_sub_resource_idx %u, dst_location %s, dst_rect %s, colour_key %p, filter %s.\n",
7788 blitter
, op
, context
, src_texture
, src_sub_resource_idx
, wined3d_debug_location(src_location
),
7789 wine_dbgstr_rect(src_rect
), dst_texture
, dst_sub_resource_idx
, wined3d_debug_location(dst_location
),
7790 wine_dbgstr_rect(dst_rect
), color_key
, debug_d3dtexturefiltertype(filter
));
7792 if (!arbfp_blit_supported(op
, context
, &src_texture
->resource
, src_location
,
7793 &dst_texture
->resource
, dst_location
))
7795 if (!(next
= blitter
->next
))
7797 ERR("No blitter to handle blit op %#x.\n", op
);
7798 return dst_location
;
7801 TRACE("Forwarding to blitter %p.\n", next
);
7802 return next
->ops
->blitter_blit(next
, op
, context
, src_texture
, src_sub_resource_idx
, src_location
,
7803 src_rect
, dst_texture
, dst_sub_resource_idx
, dst_location
, dst_rect
, color_key
, filter
);
7806 arbfp_blitter
= CONTAINING_RECORD(blitter
, struct wined3d_arbfp_blitter
, blitter
);
7808 if (!(src_texture
->resource
.access
& WINED3D_RESOURCE_ACCESS_GPU
))
7810 struct wined3d_resource_desc desc
;
7811 struct wined3d_box upload_box
;
7814 TRACE("Source texture is not GPU accessible, creating a staging texture.\n");
7816 src_level
= src_sub_resource_idx
% src_texture
->level_count
;
7817 desc
.resource_type
= WINED3D_RTYPE_TEXTURE_2D
;
7818 desc
.format
= src_texture
->resource
.format
->id
;
7819 desc
.multisample_type
= src_texture
->resource
.multisample_type
;
7820 desc
.multisample_quality
= src_texture
->resource
.multisample_quality
;
7821 desc
.usage
= WINED3DUSAGE_PRIVATE
;
7822 desc
.access
= WINED3D_RESOURCE_ACCESS_GPU
;
7823 desc
.width
= wined3d_texture_get_level_width(src_texture
, src_level
);
7824 desc
.height
= wined3d_texture_get_level_height(src_texture
, src_level
);
7828 if (FAILED(hr
= wined3d_texture_create(device
, &desc
, 1, 1, 0,
7829 NULL
, NULL
, &wined3d_null_parent_ops
, &staging_texture
)))
7831 ERR("Failed to create staging texture, hr %#x.\n", hr
);
7832 return dst_location
;
7835 wined3d_box_set(&upload_box
, 0, 0, desc
.width
, desc
.height
, 0, desc
.depth
);
7836 wined3d_texture_upload_from_texture(staging_texture
, 0, 0, 0, 0,
7837 src_texture
, src_sub_resource_idx
, &upload_box
);
7839 src_texture
= staging_texture
;
7840 src_sub_resource_idx
= 0;
7842 else if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7843 && (src_texture
->sub_resources
[src_sub_resource_idx
].locations
7844 & (WINED3D_LOCATION_TEXTURE_RGB
| WINED3D_LOCATION_DRAWABLE
)) == WINED3D_LOCATION_DRAWABLE
7845 && !wined3d_resource_is_offscreen(&src_texture
->resource
))
7848 /* Without FBO blits transferring from the drawable to the texture is
7849 * expensive, because we have to flip the data in sysmem. Since we can
7850 * flip in the blitter, we don't actually need that flip anyway. So we
7851 * use the surface's texture as scratch texture, and flip the source
7852 * rectangle instead. */
7853 texture2d_load_fb_texture(src_texture
, src_sub_resource_idx
, FALSE
, context
);
7856 src_level
= src_sub_resource_idx
% src_texture
->level_count
;
7857 s
.top
= wined3d_texture_get_level_height(src_texture
, src_level
) - s
.top
;
7858 s
.bottom
= wined3d_texture_get_level_height(src_texture
, src_level
) - s
.bottom
;
7863 wined3d_texture_load(src_texture
, context
, FALSE
);
7866 context_apply_ffp_blit_state(context
, device
);
7868 if (dst_location
== WINED3D_LOCATION_DRAWABLE
)
7871 wined3d_texture_translate_drawable_coords(dst_texture
, context
->win_handle
, &d
);
7875 if (wined3d_settings
.offscreen_rendering_mode
== ORM_FBO
)
7879 if (dst_location
== WINED3D_LOCATION_DRAWABLE
)
7881 TRACE("Destination texture %p is onscreen.\n", dst_texture
);
7882 buffer
= wined3d_texture_get_gl_buffer(dst_texture
);
7886 TRACE("Destination texture %p is offscreen.\n", dst_texture
);
7887 buffer
= GL_COLOR_ATTACHMENT0
;
7889 context_apply_fbo_state_blit(context
, GL_DRAW_FRAMEBUFFER
,
7890 &dst_texture
->resource
, dst_sub_resource_idx
, NULL
, 0, dst_location
);
7891 context_set_draw_buffer(context
, buffer
);
7892 context_check_fbo_status(context
, GL_DRAW_FRAMEBUFFER
);
7893 context_invalidate_state(context
, STATE_FRAMEBUFFER
);
7896 if (op
== WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
)
7898 const struct wined3d_format
*fmt
= src_texture
->resource
.format
;
7899 alpha_test_key
.color_space_low_value
= 0;
7900 alpha_test_key
.color_space_high_value
= ~(((1u << fmt
->alpha_size
) - 1) << fmt
->alpha_offset
);
7901 color_key
= &alpha_test_key
;
7904 arbfp_blit_set(arbfp_blitter
, context
, src_texture
, src_sub_resource_idx
, color_key
);
7906 /* Draw a textured quad */
7907 context_draw_textured_quad(context
, src_texture
, src_sub_resource_idx
, src_rect
, dst_rect
, filter
);
7909 /* Leave the opengl state valid for blitting */
7910 arbfp_blit_unset(context
->gl_info
);
7912 if (dst_texture
->swapchain
&& (dst_texture
->swapchain
->front_buffer
== dst_texture
))
7913 context
->gl_info
->gl_ops
.gl
.p_glFlush();
7915 if (staging_texture
)
7916 wined3d_texture_decref(staging_texture
);
7918 return dst_location
;
7921 static void arbfp_blitter_clear(struct wined3d_blitter
*blitter
, struct wined3d_device
*device
,
7922 unsigned int rt_count
, const struct wined3d_fb_state
*fb
, unsigned int rect_count
, const RECT
*clear_rects
,
7923 const RECT
*draw_rect
, DWORD flags
, const struct wined3d_color
*colour
, float depth
, DWORD stencil
)
7925 struct wined3d_blitter
*next
;
7927 if ((next
= blitter
->next
))
7928 next
->ops
->blitter_clear(next
, device
, rt_count
, fb
, rect_count
,
7929 clear_rects
, draw_rect
, flags
, colour
, depth
, stencil
);
7932 static const struct wined3d_blitter_ops arbfp_blitter_ops
=
7934 arbfp_blitter_destroy
,
7935 arbfp_blitter_clear
,
7939 void wined3d_arbfp_blitter_create(struct wined3d_blitter
**next
, const struct wined3d_device
*device
)
7941 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
7942 struct wined3d_arbfp_blitter
*blitter
;
7944 if (device
->shader_backend
!= &arb_program_shader_backend
7945 && device
->shader_backend
!= &glsl_shader_backend
)
7948 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7951 if (!gl_info
->supported
[WINED3D_GL_LEGACY_CONTEXT
])
7954 if (!(blitter
= heap_alloc(sizeof(*blitter
))))
7956 ERR("Failed to allocate blitter.\n");
7960 TRACE("Created blitter %p.\n", blitter
);
7962 blitter
->blitter
.ops
= &arbfp_blitter_ops
;
7963 blitter
->blitter
.next
= *next
;
7964 wine_rb_init(&blitter
->shaders
, arbfp_blit_type_compare
);
7965 blitter
->palette_texture
= 0;
7966 *next
= &blitter
->blitter
;