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"
36 #include "wined3d_private.h"
38 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
41 WINE_DECLARE_DEBUG_CHANNEL(d3d_perf
);
43 static BOOL
shader_is_pshader_version(enum wined3d_shader_type type
)
45 return type
== WINED3D_SHADER_TYPE_PIXEL
;
48 static BOOL
shader_is_vshader_version(enum wined3d_shader_type type
)
50 return type
== WINED3D_SHADER_TYPE_VERTEX
;
53 static const char *get_line(const char **ptr
)
58 if (!(q
= strstr(p
, "\n")))
80 static const char *arb_get_helper_value(enum wined3d_shader_type shader
, enum arb_helper_value value
)
82 if (shader
!= WINED3D_SHADER_TYPE_VERTEX
&& shader
!= WINED3D_SHADER_TYPE_PIXEL
)
84 ERR("Unsupported shader type '%s'.\n", debug_shader_type(shader
));
88 if (shader
== WINED3D_SHADER_TYPE_PIXEL
)
92 case ARB_ZERO
: return "ps_helper_const.x";
93 case ARB_ONE
: return "ps_helper_const.y";
94 case ARB_TWO
: return "coefmul.x";
95 case ARB_0001
: return "ps_helper_const.xxxy";
96 case ARB_EPS
: return "ps_helper_const.z";
104 case ARB_ZERO
: return "helper_const.x";
105 case ARB_ONE
: return "helper_const.y";
106 case ARB_TWO
: return "helper_const.z";
107 case ARB_EPS
: return "helper_const.w";
108 case ARB_0001
: return "helper_const.xxxy";
109 case ARB_VS_REL_OFFSET
: return "rel_addr_const.y";
112 FIXME("Unmanaged %s shader helper constant requested: %u.\n",
113 shader
== WINED3D_SHADER_TYPE_PIXEL
? "pixel" : "vertex", value
);
116 case ARB_ZERO
: return "0.0";
117 case ARB_ONE
: return "1.0";
118 case ARB_TWO
: return "2.0";
119 case ARB_0001
: return "{0.0, 0.0, 0.0, 1.0}";
120 case ARB_EPS
: return "1e-8";
121 default: return "bad";
125 static inline BOOL
ffp_clip_emul(const struct wined3d_context
*context
)
127 return context
->lowest_disabled_stage
< 7;
130 /* ARB_program_shader private data */
149 struct wined3d_shader_loop_control loop_control
;
153 struct arb_ps_np2fixup_info
155 struct ps_np2fixup_info super
;
156 /* For ARB we need an offset value:
157 * With both GLSL and ARB mode the NP2 fixup information (the texture dimensions) are stored in a
158 * consecutive way (GLSL uses a uniform array). Since ARB doesn't know the notion of a "standalone"
159 * array we need an offset to the index inside the program local parameter array. */
163 struct arb_ps_compile_args
165 struct ps_compile_args super
;
167 WORD clip
; /* only a boolean, use a WORD for alignment */
168 unsigned char loop_ctrl
[WINED3D_MAX_CONSTS_I
][3];
171 struct stb_const_desc
173 unsigned char texunit
;
177 struct arb_ps_compiled_shader
179 struct arb_ps_compile_args args
;
180 struct arb_ps_np2fixup_info np2fixup_info
;
181 struct stb_const_desc bumpenvmatconst
[MAX_TEXTURES
];
182 struct stb_const_desc luminanceconst
[MAX_TEXTURES
];
183 UINT int_consts
[WINED3D_MAX_CONSTS_I
];
186 unsigned char numbumpenvmatconsts
;
190 struct arb_vs_compile_args
192 struct vs_compile_args super
;
198 unsigned char clip_texcoord
;
199 unsigned char clipplane_mask
;
201 DWORD boolclip_compare
;
206 unsigned char samplers
[4];
207 DWORD samplers_compare
;
209 unsigned char loop_ctrl
[WINED3D_MAX_CONSTS_I
][3];
212 struct arb_vs_compiled_shader
214 struct arb_vs_compile_args args
;
216 UINT int_consts
[WINED3D_MAX_CONSTS_I
];
218 char need_color_unclamp
;
222 struct recorded_instruction
224 struct wined3d_shader_instruction ins
;
228 struct shader_arb_ctx_priv
233 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
235 /* GL_NV_vertex_program2_option or GL_NV_fragment_program_option */
237 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
241 const struct arb_vs_compile_args
*cur_vs_args
;
242 const struct arb_ps_compile_args
*cur_ps_args
;
243 const struct arb_ps_compiled_shader
*compiled_fprog
;
244 const struct arb_vs_compiled_shader
*compiled_vprog
;
245 struct arb_ps_np2fixup_info
*cur_np2fixup_info
;
246 struct list control_frames
;
250 unsigned int num_loops
, loop_depth
, num_ifcs
;
252 BOOL ps_post_process
;
254 unsigned int vs_clipplanes
;
258 /* For 3.0 vertex shaders */
259 const char *vs_output
[MAX_REG_OUTPUT
];
260 /* For 2.x and earlier vertex shaders */
261 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
263 /* 3.0 pshader input for compatibility with fixed function */
264 const char *ps_input
[MAX_REG_INPUT
];
269 struct wined3d_shader_signature sig
;
271 struct wine_rb_entry entry
;
274 struct arb_pshader_private
{
275 struct arb_ps_compiled_shader
*gl_shaders
;
276 UINT num_gl_shaders
, shader_array_size
;
277 DWORD input_signature_idx
;
278 DWORD clipplane_emulation
;
282 struct arb_vshader_private
{
283 struct arb_vs_compiled_shader
*gl_shaders
;
284 UINT num_gl_shaders
, shader_array_size
;
288 struct shader_arb_priv
290 GLuint current_vprogram_id
;
291 GLuint current_fprogram_id
;
292 const struct arb_ps_compiled_shader
*compiled_fprog
;
293 const struct arb_vs_compiled_shader
*compiled_vprog
;
294 BOOL use_arbfp_fixed_func
;
295 struct wine_rb_tree fragment_shaders
;
296 BOOL last_ps_const_clamped
;
297 BOOL last_vs_color_unclamp
;
299 struct wine_rb_tree signature_tree
;
302 unsigned int highest_dirty_ps_const
, highest_dirty_vs_const
;
303 char vshader_const_dirty
[WINED3D_MAX_VS_CONSTS_F
];
304 char pshader_const_dirty
[WINED3D_MAX_PS_CONSTS_F
];
305 const struct wined3d_context
*last_context
;
307 const struct wined3d_vertex_pipe_ops
*vertex_pipe
;
308 const struct fragment_pipeline
*fragment_pipe
;
309 BOOL ffp_proj_control
;
312 /* Context activation for state handlers is done by the caller. */
314 static BOOL
need_rel_addr_const(const struct arb_vshader_private
*shader_data
,
315 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
317 if (shader_data
->rel_offset
) return TRUE
;
318 if (!reg_maps
->usesmova
) return FALSE
;
319 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
322 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
323 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
325 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
326 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
329 static BOOL
need_helper_const(const struct arb_vshader_private
*shader_data
,
330 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
332 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) return TRUE
;
333 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
334 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
335 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) return TRUE
; /* Have to init texcoords. */
336 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
337 if (reg_maps
->usesnrm
) return TRUE
; /* 0.0 */
338 if (reg_maps
->usespow
) return TRUE
; /* EPS, 0.0 and 1.0 */
339 if (reg_maps
->fog
) return TRUE
; /* Clamping fog coord, 0.0 and 1.0 */
343 static unsigned int reserved_vs_const(const struct arb_vshader_private
*shader_data
,
344 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
346 unsigned int ret
= 1;
347 /* We use one PARAM for the pos fixup, and in some cases one to load
348 * some immediate values into the shader. */
349 if (need_helper_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
350 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
354 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
355 * When constant_list == NULL, it will load all the constants.
357 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
358 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
360 /* Context activation is done by the caller. */
361 static unsigned int shader_arb_load_constants_f(const struct wined3d_shader
*shader
,
362 const struct wined3d_gl_info
*gl_info
, GLuint target_type
, unsigned int max_constants
,
363 const struct wined3d_vec4
*constants
, char *dirty_consts
)
365 struct wined3d_shader_lconst
*lconst
;
366 unsigned int ret
, i
, j
;
368 if (TRACE_ON(d3d_constants
))
370 for (i
= 0; i
< max_constants
; ++i
)
372 if (!dirty_consts
[i
])
374 TRACE_(d3d_constants
)("Loading constant %u: %s.\n", i
, debug_vec4(&constants
[i
]));
380 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
381 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& shader
->reg_maps
.shader_version
.major
== 1)
384 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
385 * shaders, the first 8 constants are marked dirty for reload
387 for (; i
< min(8, max_constants
); ++i
)
389 if (!dirty_consts
[i
])
393 if (constants
[i
].x
> 1.0f
)
395 else if (constants
[i
].x
< -1.0f
)
396 lcl_const
[0] = -1.0f
;
398 lcl_const
[0] = constants
[i
].x
;
400 if (constants
[i
].y
> 1.0f
)
402 else if (constants
[i
].y
< -1.0f
)
403 lcl_const
[1] = -1.0f
;
405 lcl_const
[1] = constants
[i
].y
;
407 if (constants
[i
].z
> 1.0f
)
409 else if (constants
[i
].z
< -1.0f
)
410 lcl_const
[2] = -1.0f
;
412 lcl_const
[2] = constants
[i
].z
;
414 if (constants
[i
].w
> 1.0f
)
416 else if (constants
[i
].w
< -1.0f
)
417 lcl_const
[3] = -1.0f
;
419 lcl_const
[3] = constants
[i
].w
;
421 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
424 /* If further constants are dirty, reload them without clamping.
426 * The alternative is not to touch them, but then we cannot reset the dirty constant count
427 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
428 * above would always re-check the first 8 constants since max_constant remains at the init
433 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
435 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
436 * or just reloading *all* constants at once
438 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
440 for (; i
< max_constants
; ++i
)
442 if (!dirty_consts
[i
])
445 /* Find the next block of dirty constants */
448 for (++i
; (i
< max_constants
) && dirty_consts
[i
]; ++i
)
453 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, &constants
[j
].x
));
458 for (; i
< max_constants
; ++i
)
463 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, &constants
[i
].x
));
467 checkGLcall("glProgramEnvParameter4fvARB()");
469 /* Load immediate constants */
470 if (shader
->load_local_constsF
)
472 if (TRACE_ON(d3d_shader
))
474 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
476 GLfloat
* values
= (GLfloat
*)lconst
->value
;
477 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
478 values
[0], values
[1], values
[2], values
[3]);
481 /* Immediate constants are clamped for 1.X shaders at loading times */
483 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
485 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
486 ret
= max(ret
, lconst
->idx
+ 1);
487 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
489 checkGLcall("glProgramEnvParameter4fvARB()");
490 return ret
; /* The loaded immediate constants need reloading for the next shader */
492 return 0; /* No constants are dirty now */
496 /* Loads the texture dimensions for NP2 fixup into the currently set
497 * ARB_[vertex/fragment]_programs. */
498 static void shader_arb_load_np2fixup_constants(const struct arb_ps_np2fixup_info
*fixup
,
499 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
501 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
502 WORD active
= fixup
->super
.active
;
508 for (i
= 0; active
; active
>>= 1, ++i
)
510 const struct wined3d_texture
*tex
= state
->textures
[i
];
511 unsigned char idx
= fixup
->super
.idx
[i
];
512 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
519 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
525 tex_dim
[2] = tex
->pow2_matrix
[0];
526 tex_dim
[3] = tex
->pow2_matrix
[5];
530 tex_dim
[0] = tex
->pow2_matrix
[0];
531 tex_dim
[1] = tex
->pow2_matrix
[5];
535 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
)
537 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
538 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
542 /* Context activation is done by the caller. */
543 static void shader_arb_ps_local_constants(const struct arb_ps_compiled_shader
*gl_shader
,
544 const struct wined3d_context
*context
, const struct wined3d_state
*state
, UINT rt_height
)
546 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
549 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
551 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
553 /* The state manager takes care that this function is always called if the bump env matrix changes */
554 const float *data
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_MAT00
];
555 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
556 gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
558 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
560 /* WINED3D_TSS_BUMPENVLSCALE and WINED3D_TSS_BUMPENVLOFFSET are next to each other.
561 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
562 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
563 * (they're WINED3D_TSS_TEXTURETRANSFORMFLAGS and WINED3D_TSS_ADDRESSW, so most likely 0 or NaN
565 const float *scale
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_LSCALE
];
566 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
567 gl_shader
->luminanceconst
[i
].const_num
, scale
));
570 checkGLcall("Load bumpmap consts");
572 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
574 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
575 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
580 val
[0] = context
->render_offscreen
? 0.0f
: (float) rt_height
;
581 val
[1] = context
->render_offscreen
? 1.0f
: -1.0f
;
584 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
585 checkGLcall("y correction loading");
588 if (!gl_shader
->num_int_consts
) return;
590 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
592 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
595 val
[0] = (float)state
->ps_consts_i
[i
].x
;
596 val
[1] = (float)state
->ps_consts_i
[i
].y
;
597 val
[2] = (float)state
->ps_consts_i
[i
].z
;
600 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
603 checkGLcall("Load ps int consts");
606 /* Context activation is done by the caller. */
607 static void shader_arb_vs_local_constants(const struct arb_vs_compiled_shader
*gl_shader
,
608 const struct wined3d_context
*context
, const struct wined3d_state
*state
)
610 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
611 float position_fixup
[4];
614 /* Upload the position fixup */
615 shader_get_position_fixup(context
, state
, position_fixup
);
616 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, position_fixup
));
618 if (!gl_shader
->num_int_consts
) return;
620 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
622 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
625 val
[0] = (float)state
->vs_consts_i
[i
].x
;
626 val
[1] = (float)state
->vs_consts_i
[i
].y
;
627 val
[2] = (float)state
->vs_consts_i
[i
].z
;
630 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
633 checkGLcall("Load vs int consts");
636 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
637 const struct wined3d_state
*state
);
640 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
642 * We only support float constants in ARB at the moment, so don't
643 * worry about the Integers or Booleans
645 /* Context activation is done by the caller (state handler). */
646 static void shader_arb_load_constants_internal(struct shader_arb_priv
*priv
,
647 struct wined3d_context
*context
, const struct wined3d_state
*state
,
648 BOOL usePixelShader
, BOOL useVertexShader
, BOOL from_shader_select
)
650 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
651 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
653 if (!from_shader_select
)
655 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
656 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
659 && (vshader
->reg_maps
.boolean_constants
660 || (!gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
661 && (vshader
->reg_maps
.integer_constants
& ~vshader
->reg_maps
.local_int_consts
))))
663 TRACE("bool/integer vertex shader constants potentially modified, forcing shader reselection.\n");
664 shader_arb_select(priv
, context
, state
);
667 && (pshader
->reg_maps
.boolean_constants
668 || (!gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
]
669 && (pshader
->reg_maps
.integer_constants
& ~pshader
->reg_maps
.local_int_consts
))))
671 TRACE("bool/integer pixel shader constants potentially modified, forcing shader reselection.\n");
672 shader_arb_select(priv
, context
, state
);
676 if (context
!= priv
->last_context
)
678 memset(priv
->vshader_const_dirty
, 1,
679 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
680 priv
->highest_dirty_vs_const
= d3d_info
->limits
.vs_uniform_count
;
682 memset(priv
->pshader_const_dirty
, 1,
683 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
684 priv
->highest_dirty_ps_const
= d3d_info
->limits
.ps_uniform_count
;
686 priv
->last_context
= context
;
691 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
692 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
694 /* Load DirectX 9 float constants for vertex shader */
695 priv
->highest_dirty_vs_const
= shader_arb_load_constants_f(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
696 priv
->highest_dirty_vs_const
, state
->vs_consts_f
, priv
->vshader_const_dirty
);
697 shader_arb_vs_local_constants(gl_shader
, context
, state
);
702 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
703 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
704 UINT rt_height
= state
->fb
->render_targets
[0]->height
;
706 /* Load DirectX 9 float constants for pixel shader */
707 priv
->highest_dirty_ps_const
= shader_arb_load_constants_f(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
708 priv
->highest_dirty_ps_const
, state
->ps_consts_f
, priv
->pshader_const_dirty
);
709 shader_arb_ps_local_constants(gl_shader
, context
, state
, rt_height
);
711 if (context
->constant_update_mask
& WINED3D_SHADER_CONST_PS_NP2_FIXUP
)
712 shader_arb_load_np2fixup_constants(&gl_shader
->np2fixup_info
, gl_info
, state
);
716 static void shader_arb_load_constants(void *shader_priv
, struct wined3d_context
*context
,
717 const struct wined3d_state
*state
)
719 BOOL vs
= use_vs(state
);
720 BOOL ps
= use_ps(state
);
722 shader_arb_load_constants_internal(shader_priv
, context
, state
, ps
, vs
, FALSE
);
725 static void shader_arb_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
727 struct wined3d_context
*context
= context_get_current();
728 struct shader_arb_priv
*priv
= device
->shader_priv
;
730 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
731 * context. On a context switch the old context will be fully dirtified */
732 if (!context
|| context
->device
!= device
)
735 memset(priv
->vshader_const_dirty
+ start
, 1, sizeof(*priv
->vshader_const_dirty
) * count
);
736 priv
->highest_dirty_vs_const
= max(priv
->highest_dirty_vs_const
, start
+ count
);
739 static void shader_arb_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
741 struct wined3d_context
*context
= context_get_current();
742 struct shader_arb_priv
*priv
= device
->shader_priv
;
744 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
745 * context. On a context switch the old context will be fully dirtified */
746 if (!context
|| context
->device
!= device
)
749 memset(priv
->pshader_const_dirty
+ start
, 1, sizeof(*priv
->pshader_const_dirty
) * count
);
750 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, start
+ count
);
753 static void shader_arb_append_imm_vec4(struct wined3d_string_buffer
*buffer
, const float *values
)
757 wined3d_ftoa(values
[0], str
[0]);
758 wined3d_ftoa(values
[1], str
[1]);
759 wined3d_ftoa(values
[2], str
[2]);
760 wined3d_ftoa(values
[3], str
[3]);
761 shader_addline(buffer
, "{%s, %s, %s, %s}", str
[0], str
[1], str
[2], str
[3]);
764 /* Generate the variable & register declarations for the ARB_vertex_program output target */
765 static void shader_generate_arb_declarations(const struct wined3d_shader
*shader
,
766 const struct wined3d_shader_reg_maps
*reg_maps
, struct wined3d_string_buffer
*buffer
,
767 const struct wined3d_gl_info
*gl_info
, DWORD
*num_clipplanes
,
768 const struct shader_arb_ctx_priv
*ctx
)
771 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
772 const struct wined3d_shader_lconst
*lconst
;
773 unsigned max_constantsF
;
776 /* In pixel shaders, all private constants are program local, we don't need anything
777 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
778 * If we need a private constant the GL implementation will squeeze it in somewhere
780 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
781 * immediate values. The posFixup is loaded using program.env for now, so always
782 * subtract one from the number of constants. If the shader uses indirect addressing,
783 * account for the helper const too because we have to declare all available d3d constants
784 * and don't know which are actually used.
788 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
789 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
790 if (max_constantsF
< 24)
791 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
795 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
796 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
797 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
798 * Also prevents max_constantsF from becoming less than 0 and
800 if (max_constantsF
< 96)
801 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
803 if (reg_maps
->usesrelconstF
)
805 DWORD highest_constf
= 0, clip_limit
;
807 max_constantsF
-= reserved_vs_const(shader_data
, reg_maps
, gl_info
);
808 max_constantsF
-= wined3d_popcount(reg_maps
->integer_constants
);
809 max_constantsF
-= gl_info
->reserved_arb_constants
;
811 for (i
= 0; i
< shader
->limits
->constant_float
; ++i
)
814 DWORD shift
= i
& 0x1f;
815 if (reg_maps
->constf
[idx
] & (1u << shift
))
819 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
821 if(ctx
->cur_vs_args
->super
.clip_enabled
)
822 clip_limit
= gl_info
->limits
.user_clip_distances
;
828 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
829 clip_limit
= min(wined3d_popcount(mask
), 4);
831 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
832 max_constantsF
-= *num_clipplanes
;
833 if(*num_clipplanes
< clip_limit
)
835 WARN("Only %u clip planes out of %u enabled.\n", *num_clipplanes
,
836 gl_info
->limits
.user_clip_distances
);
841 if (ctx
->target_version
>= NV2
)
842 *num_clipplanes
= gl_info
->limits
.user_clip_distances
;
844 *num_clipplanes
= min(gl_info
->limits
.user_clip_distances
, 4);
848 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
850 if (map
& 1) shader_addline(buffer
, "TEMP R%u;\n", i
);
853 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
855 if (map
& 1) shader_addline(buffer
, "ADDRESS A%u;\n", i
);
858 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
860 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
862 if (map
& 1) shader_addline(buffer
, "TEMP T%u;\n", i
);
866 if (!shader
->load_local_constsF
)
868 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
871 value
= (const float *)lconst
->value
;
872 shader_addline(buffer
, "PARAM C%u = ", lconst
->idx
);
873 shader_arb_append_imm_vec4(buffer
, value
);
874 shader_addline(buffer
, ";\n");
878 /* After subtracting privately used constants from the hardware limit(they are loaded as
879 * local constants), make sure the shader doesn't violate the env constant limit
883 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
887 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
890 /* Avoid declaring more constants than needed */
891 max_constantsF
= min(max_constantsF
, shader
->limits
->constant_float
);
893 /* we use the array-based constants array if the local constants are marked for loading,
894 * because then we use indirect addressing, or when the local constant list is empty,
895 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
896 * local constants do not declare the loaded constants as an array because ARB compilers usually
897 * do not optimize unused constants away
899 if (reg_maps
->usesrelconstF
)
901 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
902 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
903 max_constantsF
, max_constantsF
- 1);
907 for (i
= 0; i
< max_constantsF
; ++i
)
911 mask
= 1u << (i
& 0x1fu
);
912 if (!shader_constant_is_local(shader
, i
) && (reg_maps
->constf
[idx
] & mask
))
914 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
920 static const char * const shift_tab
[] = {
921 "dummy", /* 0 (none) */
922 "coefmul.x", /* 1 (x2) */
923 "coefmul.y", /* 2 (x4) */
924 "coefmul.z", /* 3 (x8) */
925 "coefmul.w", /* 4 (x16) */
926 "dummy", /* 5 (x32) */
927 "dummy", /* 6 (x64) */
928 "dummy", /* 7 (x128) */
929 "dummy", /* 8 (d256) */
930 "dummy", /* 9 (d128) */
931 "dummy", /* 10 (d64) */
932 "dummy", /* 11 (d32) */
933 "coefdiv.w", /* 12 (d16) */
934 "coefdiv.z", /* 13 (d8) */
935 "coefdiv.y", /* 14 (d4) */
936 "coefdiv.x" /* 15 (d2) */
939 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
940 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
942 char *ptr
= write_mask
;
944 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
947 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
948 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
949 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
950 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
956 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
958 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
959 * but addressed as "rgba". To fix this we need to swap the register's x
960 * and z components. */
961 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
962 char *ptr
= swizzle_str
;
964 /* swizzle bits fields: wwzzyyxx */
965 DWORD swizzle
= param
->swizzle
;
966 DWORD swizzle_x
= swizzle
& 0x03;
967 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
968 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
969 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
971 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
972 * generate a swizzle string. Unless we need to our own swizzling. */
973 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
976 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
977 *ptr
++ = swizzle_chars
[swizzle_x
];
979 *ptr
++ = swizzle_chars
[swizzle_x
];
980 *ptr
++ = swizzle_chars
[swizzle_y
];
981 *ptr
++ = swizzle_chars
[swizzle_z
];
982 *ptr
++ = swizzle_chars
[swizzle_w
];
989 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
991 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
992 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
994 if (!strcmp(priv
->addr_reg
, src
)) return;
996 strcpy(priv
->addr_reg
, src
);
997 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
1000 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1001 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
1003 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
1004 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
1006 /* oPos, oFog and oPts in D3D */
1007 static const char * const rastout_reg_names
[] = {"TMP_OUT", "TMP_FOGCOORD", "result.pointsize"};
1008 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1009 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1010 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1011 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1017 case WINED3DSPR_TEMP
:
1018 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1021 case WINED3DSPR_INPUT
:
1024 if (reg_maps
->shader_version
.major
< 3)
1026 if (!reg
->idx
[0].offset
)
1027 strcpy(register_name
, "fragment.color.primary");
1029 strcpy(register_name
, "fragment.color.secondary");
1033 if (reg
->idx
[0].rel_addr
)
1036 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1038 if (!strcmp(rel_reg
, "**aL_emul**"))
1040 DWORD idx
= ctx
->aL
+ reg
->idx
[0].offset
;
1041 if(idx
< MAX_REG_INPUT
)
1043 strcpy(register_name
, ctx
->ps_input
[idx
]);
1047 ERR("Pixel shader input register out of bounds: %u\n", idx
);
1048 sprintf(register_name
, "out_of_bounds_%u", idx
);
1051 else if (reg_maps
->input_registers
& 0x0300)
1053 /* There are two ways basically:
1055 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
1056 * That means trouble if the loop also contains a breakc or if the control values
1057 * aren't local constants.
1058 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
1059 * source dynamically. The trouble is that we cannot simply read aL.y because it
1060 * is an ADDRESS register. We could however push it, load .zw with a value and use
1061 * ADAC to load the condition code register and pop it again afterwards
1063 FIXME("Relative input register addressing with more than 8 registers\n");
1065 /* This is better than nothing for now */
1066 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1068 else if(ctx
->cur_ps_args
->super
.vp_mode
!= vertexshader
)
1070 /* This is problematic because we'd have to consult the ctx->ps_input strings
1071 * for where to find the varying. Some may be "0.0", others can be texcoords or
1072 * colors. This needs either a pipeline replacement to make the vertex shader feed
1073 * proper varyings, or loop unrolling
1075 * For now use the texcoords and hope for the best
1077 FIXME("Non-vertex shader varying input with indirect addressing\n");
1078 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1082 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1083 * pulls GL_NV_fragment_program2 in
1085 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1090 if (reg
->idx
[0].offset
< MAX_REG_INPUT
)
1092 strcpy(register_name
, ctx
->ps_input
[reg
->idx
[0].offset
]);
1096 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
[0].offset
);
1097 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
[0].offset
);
1104 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1u << reg
->idx
[0].offset
))
1106 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
[0].offset
);
1110 case WINED3DSPR_CONST
:
1111 if (!pshader
&& reg
->idx
[0].rel_addr
)
1113 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1114 UINT rel_offset
= ctx
->target_version
== ARB
? shader_data
->rel_offset
: 0;
1117 if (reg_maps
->shader_version
.major
< 2)
1119 sprintf(rel_reg
, "A0.x");
1123 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1124 if (ctx
->target_version
== ARB
)
1126 if (!strcmp(rel_reg
, "**aL_emul**"))
1130 shader_arb_request_a0(ins
, rel_reg
);
1131 sprintf(rel_reg
, "A0.x");
1136 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
[0].offset
);
1137 else if (reg
->idx
[0].offset
>= rel_offset
)
1138 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
[0].offset
- rel_offset
);
1140 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
[0].offset
);
1144 if (reg_maps
->usesrelconstF
)
1145 sprintf(register_name
, "C[%u]", reg
->idx
[0].offset
);
1147 sprintf(register_name
, "C%u", reg
->idx
[0].offset
);
1151 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1154 if (reg_maps
->shader_version
.major
== 1
1155 && reg_maps
->shader_version
.minor
<= 3)
1156 /* In ps <= 1.3, Tx is a temporary register as destination
1157 * to all instructions, and as source to most instructions.
1158 * For some instructions it is the texcoord input. Those
1159 * instructions know about the special use. */
1160 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1162 /* In ps 1.4 and 2.x Tx is always a (read-only) varying. */
1163 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
[0].offset
);
1167 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1168 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1170 sprintf(register_name
, "A%u_SHADOW", reg
->idx
[0].offset
);
1174 case WINED3DSPR_COLOROUT
:
1175 if (ctx
->ps_post_process
&& !reg
->idx
[0].offset
)
1177 strcpy(register_name
, "TMP_COLOR");
1181 if (ctx
->cur_ps_args
->super
.srgb_correction
)
1182 FIXME("sRGB correction on higher render targets.\n");
1183 if (reg_maps
->rt_mask
> 1)
1184 sprintf(register_name
, "result.color[%u]", reg
->idx
[0].offset
);
1186 strcpy(register_name
, "result.color");
1190 case WINED3DSPR_RASTOUT
:
1191 if (reg
->idx
[0].offset
== 1)
1192 sprintf(register_name
, "%s", ctx
->fog_output
);
1194 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
[0].offset
]);
1197 case WINED3DSPR_DEPTHOUT
:
1198 strcpy(register_name
, "result.depth");
1201 case WINED3DSPR_ATTROUT
:
1202 /* case WINED3DSPR_OUTPUT: */
1204 sprintf(register_name
, "oD[%u]", reg
->idx
[0].offset
);
1206 strcpy(register_name
, ctx
->color_output
[reg
->idx
[0].offset
]);
1209 case WINED3DSPR_TEXCRDOUT
:
1211 sprintf(register_name
, "oT[%u]", reg
->idx
[0].offset
);
1212 else if (reg_maps
->shader_version
.major
< 3)
1213 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
[0].offset
]);
1215 strcpy(register_name
, ctx
->vs_output
[reg
->idx
[0].offset
]);
1218 case WINED3DSPR_LOOP
:
1219 if(ctx
->target_version
>= NV2
)
1221 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1222 if(pshader
) sprintf(register_name
, "A0.x");
1223 else sprintf(register_name
, "aL.y");
1227 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1228 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1229 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1230 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1233 sprintf(register_name
, "**aL_emul**");
1238 case WINED3DSPR_CONSTINT
:
1239 sprintf(register_name
, "I%u", reg
->idx
[0].offset
);
1242 case WINED3DSPR_MISCTYPE
:
1243 if (!reg
->idx
[0].offset
)
1244 sprintf(register_name
, "vpos");
1245 else if (reg
->idx
[0].offset
== 1)
1246 sprintf(register_name
, "fragment.facing.x");
1248 FIXME("Unknown MISCTYPE register index %u.\n", reg
->idx
[0].offset
);
1252 FIXME("Unhandled register type %#x[%u].\n", reg
->type
, reg
->idx
[0].offset
);
1253 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
[0].offset
);
1258 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1259 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1261 char register_name
[255];
1265 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1266 strcpy(str
, register_name
);
1268 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1269 strcat(str
, write_mask
);
1272 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1274 switch(channel_source
)
1276 case CHANNEL_SOURCE_ZERO
: return "0";
1277 case CHANNEL_SOURCE_ONE
: return "1";
1278 case CHANNEL_SOURCE_X
: return "x";
1279 case CHANNEL_SOURCE_Y
: return "y";
1280 case CHANNEL_SOURCE_Z
: return "z";
1281 case CHANNEL_SOURCE_W
: return "w";
1283 FIXME("Unhandled channel source %#x\n", channel_source
);
1288 struct color_fixup_masks
1294 static struct color_fixup_masks
calc_color_correction(struct color_fixup_desc fixup
, DWORD dst_mask
)
1296 struct color_fixup_masks masks
= {0, 0};
1298 if (is_complex_fixup(fixup
))
1300 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1301 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1305 if (fixup
.x_source
!= CHANNEL_SOURCE_X
)
1306 masks
.source
|= WINED3DSP_WRITEMASK_0
;
1307 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
)
1308 masks
.source
|= WINED3DSP_WRITEMASK_1
;
1309 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
)
1310 masks
.source
|= WINED3DSP_WRITEMASK_2
;
1311 if (fixup
.w_source
!= CHANNEL_SOURCE_W
)
1312 masks
.source
|= WINED3DSP_WRITEMASK_3
;
1313 masks
.source
&= dst_mask
;
1315 if (fixup
.x_sign_fixup
)
1316 masks
.sign
|= WINED3DSP_WRITEMASK_0
;
1317 if (fixup
.y_sign_fixup
)
1318 masks
.sign
|= WINED3DSP_WRITEMASK_1
;
1319 if (fixup
.z_sign_fixup
)
1320 masks
.sign
|= WINED3DSP_WRITEMASK_2
;
1321 if (fixup
.w_sign_fixup
)
1322 masks
.sign
|= WINED3DSP_WRITEMASK_3
;
1323 masks
.sign
&= dst_mask
;
1328 static void gen_color_correction(struct wined3d_string_buffer
*buffer
, const char *dst
,
1329 const char *src
, const char *one
, const char *two
,
1330 struct color_fixup_desc fixup
, struct color_fixup_masks masks
)
1332 const char *sign_fixup_src
= dst
;
1337 sign_fixup_src
= "TA";
1339 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", sign_fixup_src
, src
,
1340 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1341 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1343 else if (masks
.sign
)
1345 sign_fixup_src
= src
;
1351 char *ptr
= reg_mask
;
1353 if (masks
.sign
!= WINED3DSP_WRITEMASK_ALL
)
1356 if (masks
.sign
& WINED3DSP_WRITEMASK_0
)
1358 if (masks
.sign
& WINED3DSP_WRITEMASK_1
)
1360 if (masks
.sign
& WINED3DSP_WRITEMASK_2
)
1362 if (masks
.sign
& WINED3DSP_WRITEMASK_3
)
1367 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", dst
, reg_mask
, sign_fixup_src
, two
, one
);
1371 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1374 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1375 if (!ins
->dst_count
) return "";
1377 mod
= ins
->dst
[0].modifiers
;
1379 /* Silently ignore PARTIALPRECISION if it's not supported */
1380 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1382 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1384 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1385 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1390 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1393 case WINED3DSPDM_SATURATE
:
1396 case WINED3DSPDM_PARTIALPRECISION
:
1403 FIXME("Unknown modifiers 0x%08x\n", mod
);
1408 #define TEX_PROJ 0x1
1409 #define TEX_BIAS 0x2
1411 #define TEX_DERIV 0x10
1413 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1414 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1416 enum wined3d_shader_resource_type resource_type
= ins
->ctx
->reg_maps
->resource_info
[sampler_idx
].type
;
1417 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1418 const char *tex_type
;
1419 BOOL np2_fixup
= FALSE
;
1420 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1422 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1423 const struct wined3d_shader
*shader
;
1424 const struct wined3d_device
*device
;
1425 const struct wined3d_gl_info
*gl_info
;
1426 const char *tex_dst
= dst_str
;
1427 struct color_fixup_masks masks
;
1429 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1430 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1432 switch (resource_type
)
1434 case WINED3D_SHADER_RESOURCE_TEXTURE_1D
:
1438 case WINED3D_SHADER_RESOURCE_TEXTURE_2D
:
1439 shader
= ins
->ctx
->shader
;
1440 device
= shader
->device
;
1441 gl_info
= &device
->adapter
->gl_info
;
1443 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1u << sampler_idx
)
1444 && gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1448 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1450 if (priv
->cur_np2fixup_info
->super
.active
& (1u << sampler_idx
))
1452 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1453 else np2_fixup
= TRUE
;
1458 case WINED3D_SHADER_RESOURCE_TEXTURE_3D
:
1462 case WINED3D_SHADER_RESOURCE_TEXTURE_CUBE
:
1467 ERR("Unexpected resource type %#x.\n", resource_type
);
1471 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1472 * so don't use shader_arb_get_modifier
1474 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1477 /* Fragment samplers always have indentity mapping */
1478 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1480 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1485 masks
= calc_color_correction(priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
],
1486 ins
->dst
[0].write_mask
);
1488 if (masks
.source
|| masks
.sign
)
1492 if (flags
& TEX_DERIV
)
1494 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1495 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1496 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1497 dsx
, dsy
, sampler_idx
, tex_type
);
1499 else if(flags
& TEX_LOD
)
1501 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1502 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1503 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1504 sampler_idx
, tex_type
);
1506 else if (flags
& TEX_BIAS
)
1508 /* Shouldn't be possible, but let's check for it */
1509 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1510 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1511 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1513 else if (flags
& TEX_PROJ
)
1515 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1521 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1522 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1523 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1525 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, tex_dst
, sampler_idx
, tex_type
);
1528 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1533 gen_color_correction(buffer
, dst_str
, tex_dst
,
1534 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1535 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1536 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
], masks
);
1540 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1541 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1543 /* Generate a line that does the input modifier computation and return the input register to use */
1544 BOOL is_color
= FALSE
, insert_line
;
1547 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1548 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1549 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1550 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1552 /* Assume a new line will be added */
1555 /* Get register name */
1556 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1557 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1559 switch (src
->modifiers
)
1561 case WINED3DSPSM_NONE
:
1562 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1563 insert_line
= FALSE
;
1565 case WINED3DSPSM_NEG
:
1566 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1567 insert_line
= FALSE
;
1569 case WINED3DSPSM_BIAS
:
1570 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1572 case WINED3DSPSM_BIASNEG
:
1573 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1575 case WINED3DSPSM_SIGN
:
1576 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1578 case WINED3DSPSM_SIGNNEG
:
1579 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1581 case WINED3DSPSM_COMP
:
1582 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1584 case WINED3DSPSM_X2
:
1585 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1587 case WINED3DSPSM_X2NEG
:
1588 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1590 case WINED3DSPSM_DZ
:
1591 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1592 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1594 case WINED3DSPSM_DW
:
1595 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1596 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1598 case WINED3DSPSM_ABS
:
1599 if(ctx
->target_version
>= NV2
) {
1600 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1601 insert_line
= FALSE
;
1603 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1606 case WINED3DSPSM_ABSNEG
:
1607 if(ctx
->target_version
>= NV2
) {
1608 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1610 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1611 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1613 insert_line
= FALSE
;
1616 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1617 insert_line
= FALSE
;
1620 /* Return modified or original register, with swizzle */
1622 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1625 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1627 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1628 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1629 DWORD sampler_code
= dst
->reg
.idx
[0].offset
;
1631 char src_name
[2][50];
1633 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1635 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1637 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1638 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1641 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1642 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1643 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1644 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1645 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1647 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1648 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1651 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1656 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1657 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1658 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1659 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1660 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1661 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1662 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1663 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1664 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1665 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1666 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1667 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1668 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1670 FIXME("Unknown modifier %u\n", mod
);
1674 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1676 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1677 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1679 char src_name
[3][50];
1680 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1681 ins
->ctx
->reg_maps
->shader_version
.minor
);
1683 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1684 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1686 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
1687 && ins
->dst
->write_mask
!= WINED3DSP_WRITEMASK_3
)
1689 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1693 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1696 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1697 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1699 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1700 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1701 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1702 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1703 dst_name
, src_name
[1], src_name
[2]);
1707 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1709 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1710 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1712 char src_name
[3][50];
1714 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1716 /* Generate input register names (with modifiers) */
1717 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1718 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1719 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1721 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1722 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1725 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1726 * dst = dot2(src0, src1) + src2 */
1727 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1729 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1730 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1732 char src_name
[3][50];
1733 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1735 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1736 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1737 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1739 if(ctx
->target_version
>= NV3
)
1741 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1742 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1743 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1744 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1746 else if(ctx
->target_version
>= NV2
)
1748 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1749 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1750 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1751 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1753 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1755 * .xyxy and other swizzles that we could get with this are not valid in
1756 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1758 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1759 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1760 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1762 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1764 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1765 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1769 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1770 /* 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]
1771 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1773 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1774 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1775 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1776 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1780 /* Map the opcode 1-to-1 to the GL code */
1781 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1783 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1784 const char *instruction
;
1785 char arguments
[256], dst_str
[50];
1787 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1789 switch (ins
->handler_idx
)
1791 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1792 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1793 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1794 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1795 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1796 case WINED3DSIH_DST
: instruction
= "DST"; break;
1797 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1798 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1799 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1800 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1801 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1802 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1803 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1804 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1805 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1806 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1807 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1808 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1809 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1810 default: instruction
= "";
1811 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
1815 /* Note that shader_arb_add_dst_param() adds spaces. */
1816 arguments
[0] = '\0';
1817 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1818 for (i
= 0; i
< ins
->src_count
; ++i
)
1821 strcat(arguments
, ", ");
1822 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1823 strcat(arguments
, operand
);
1825 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1828 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1830 static DWORD
shader_arb_select_component(DWORD swizzle
, DWORD component
)
1832 return ((swizzle
>> 2 * component
) & 0x3) * 0x55;
1835 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1837 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1838 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1839 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1840 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1841 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1842 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1843 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1845 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1846 char src0_param
[256];
1848 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1850 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1852 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1854 if(ctx
->target_version
>= NV2
) {
1855 shader_hw_map2gl(ins
);
1858 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1859 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1861 /* This implements the mova formula used in GLSL. The first two instructions
1862 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1866 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1868 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1869 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1871 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1872 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1874 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1875 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1876 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1877 if (shader_data
->rel_offset
)
1879 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1881 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1883 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1885 else if (reg_maps
->shader_version
.major
== 1
1886 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1887 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1889 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1890 src0_param
[0] = '\0';
1892 if (shader_data
->rel_offset
&& ctx
->target_version
== ARB
)
1894 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1895 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1896 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1897 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1901 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1902 * with more than one component. Thus replicate the first source argument over all
1903 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1904 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1905 tmp_src
.swizzle
= shader_arb_select_component(tmp_src
.swizzle
, 0);
1906 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1907 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1910 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1912 if (ctx
->ps_post_process
&& shader
->u
.ps
.color0_mov
)
1914 shader_addline(buffer
, "#mov handled in srgb write or fog code\n");
1917 shader_hw_map2gl(ins
);
1921 shader_hw_map2gl(ins
);
1925 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1927 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1928 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1931 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1932 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1934 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1936 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1938 const char *kilsrc
= "TA";
1941 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1942 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1948 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1949 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1950 * masked out components to 0(won't kill)
1952 char x
= '0', y
= '0', z
= '0', w
= '0';
1953 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1954 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1955 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1956 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1957 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1959 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1963 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1964 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1966 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1967 * or pass in any temporary register(in shader phase 2)
1969 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1970 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1972 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1973 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1974 shader_addline(buffer
, "KIL TA;\n");
1978 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1980 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1981 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1982 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1983 ins
->ctx
->reg_maps
->shader_version
.minor
);
1984 struct wined3d_shader_src_param src
;
1988 DWORD reg_sampler_code
;
1990 BOOL swizzle_coord
= FALSE
;
1992 /* All versions have a destination register */
1993 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1995 /* 1.0-1.4: Use destination register number as texture code.
1996 2.0+: Use provided sampler number as texture code. */
1997 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1998 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
2000 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
2002 /* 1.0-1.3: Use the texcoord varying.
2003 1.4+: Use provided coordinate source register. */
2004 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2005 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
2007 /* TEX is the only instruction that can handle DW and DZ natively */
2009 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
2010 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
2011 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
2015 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
2016 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
2017 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
2019 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2022 if (reg_sampler_code
< MAX_TEXTURES
)
2023 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2024 if (flags
& WINED3D_PSARGS_PROJECTED
)
2026 myflags
|= TEX_PROJ
;
2027 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
2028 swizzle_coord
= TRUE
;
2031 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
2033 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
2034 if (src_mod
== WINED3DSPSM_DZ
)
2036 swizzle_coord
= TRUE
;
2037 myflags
|= TEX_PROJ
;
2038 } else if(src_mod
== WINED3DSPSM_DW
) {
2039 myflags
|= TEX_PROJ
;
2042 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
2043 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
2048 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
2049 * reg_coord is a read-only varying register, so we need a temp reg */
2050 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
2051 strcpy(reg_coord
, "TA");
2054 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
2057 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
2059 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2060 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2061 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2062 ins
->ctx
->reg_maps
->shader_version
.minor
);
2065 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2067 DWORD reg
= dst
->reg
.idx
[0].offset
;
2069 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2070 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2074 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2075 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2076 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2080 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2082 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2085 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2089 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2090 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2091 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2092 /* Move .x first in case src_str is "TA" */
2093 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2094 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2095 if (reg1
< MAX_TEXTURES
)
2097 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2098 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2100 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2103 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2105 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2107 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2111 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2112 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2113 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2114 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2115 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2116 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2119 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2121 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2125 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2126 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2127 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2128 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2131 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2133 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2134 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2135 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2136 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2137 DWORD reg_dest_code
;
2139 /* All versions have a destination register. The Tx where the texture coordinates come
2140 * from is the varying incarnation of the texture register
2142 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2143 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2144 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2145 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2147 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2148 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2150 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2151 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2153 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2154 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2157 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2158 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2159 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2160 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2162 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2163 * so we can't let the GL handle this.
2165 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2166 & WINED3D_PSARGS_PROJECTED
)
2168 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2169 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2170 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2172 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2175 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2177 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2179 /* No src swizzles are allowed, so this is ok */
2180 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2181 src_reg
, reg_dest_code
, reg_dest_code
);
2182 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2186 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2188 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2189 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2190 char src0_name
[50], dst_name
[50];
2192 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2194 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2195 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2196 * T<reg+1> register. Use this register to store the calculated vector
2198 tmp_reg
.idx
[0].offset
= reg
+ 1;
2199 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2200 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2203 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2205 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2207 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2208 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2214 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2215 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2217 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2218 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2219 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2220 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2221 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2224 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2226 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2227 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2228 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2229 char src0_name
[50], dst_name
[50];
2230 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2233 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2234 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2235 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2237 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2238 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2240 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2241 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2242 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2243 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2246 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2248 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2249 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2251 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2252 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2254 char src0_name
[50], dst_name
[50];
2257 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2258 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2259 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2261 /* Sample the texture using the calculated coordinates */
2262 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2263 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2264 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2265 tex_mx
->current_row
= 0;
2268 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2270 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2271 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2273 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2274 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2280 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2281 * components for temporary data storage
2283 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2284 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2285 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2287 /* Construct the eye-ray vector from w coordinates */
2288 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2289 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2290 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2292 /* Calculate reflection vector
2294 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2295 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2296 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2297 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2298 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2299 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2300 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2302 /* Sample the texture using the calculated coordinates */
2303 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2304 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2305 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2306 tex_mx
->current_row
= 0;
2309 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2311 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2312 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2314 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2315 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2322 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2323 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2324 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2325 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2326 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2328 /* Calculate reflection vector.
2331 * dst_reg.xyz = 2 * --------- * N - E
2334 * Which normalizes the normal vector
2336 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2337 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2338 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2339 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2340 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2341 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2343 /* Sample the texture using the calculated coordinates */
2344 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2345 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2346 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2347 tex_mx
->current_row
= 0;
2350 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2352 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2353 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2355 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2356 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2358 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2359 * which is essentially an input, is the destination register because it is the first
2360 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2361 * here(writemasks/swizzles are not valid on texdepth)
2363 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2365 /* According to the msdn, the source register(must be r5) is unusable after
2366 * the texdepth instruction, so we're free to modify it
2368 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2370 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2371 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2372 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2374 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2375 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2376 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2377 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2380 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2381 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2382 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2383 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2385 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2386 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2390 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2391 shader_addline(buffer
, "MOV TB, 0.0;\n");
2392 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2394 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2395 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2398 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2399 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2400 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2402 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2405 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2407 /* Handle output register */
2408 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2409 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2410 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2413 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2414 * Perform the 3rd row of a 3x3 matrix multiply */
2415 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2417 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2418 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2419 char dst_str
[50], dst_name
[50];
2423 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2424 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2425 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2426 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2427 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2430 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2431 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2432 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2433 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2435 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2437 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2438 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2439 char src0
[50], dst_name
[50];
2441 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2442 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2444 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2445 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2446 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2448 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2449 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2450 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2452 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2453 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2454 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2455 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2458 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2459 Vertex/Pixel shaders to ARB_vertex_program codes */
2460 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2463 int nComponents
= 0;
2464 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2465 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2466 struct wined3d_shader_instruction tmp_ins
;
2468 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2470 /* Set constants for the temporary argument */
2471 tmp_ins
.ctx
= ins
->ctx
;
2472 tmp_ins
.dst_count
= 1;
2473 tmp_ins
.dst
= &tmp_dst
;
2474 tmp_ins
.src_count
= 2;
2475 tmp_ins
.src
= tmp_src
;
2477 switch(ins
->handler_idx
)
2479 case WINED3DSIH_M4x4
:
2481 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2483 case WINED3DSIH_M4x3
:
2485 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2487 case WINED3DSIH_M3x4
:
2489 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2491 case WINED3DSIH_M3x3
:
2493 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2495 case WINED3DSIH_M3x2
:
2497 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2500 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2504 tmp_dst
= ins
->dst
[0];
2505 tmp_src
[0] = ins
->src
[0];
2506 tmp_src
[1] = ins
->src
[1];
2507 for (i
= 0; i
< nComponents
; ++i
)
2509 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2510 shader_hw_map2gl(&tmp_ins
);
2511 ++tmp_src
[1].reg
.idx
[0].offset
;
2515 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2521 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2522 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2523 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2524 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2525 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2526 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2527 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2528 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2529 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2530 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2531 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2532 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2533 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2535 FIXME("Unknown modifier %u\n", mod
);
2539 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2541 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2542 const char *instruction
;
2543 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2544 BOOL need_abs
= FALSE
;
2549 switch(ins
->handler_idx
)
2551 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2552 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2553 case WINED3DSIH_EXPP
:
2554 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2556 instruction
= "EXP";
2560 case WINED3DSIH_EXP
:
2561 instruction
= "EX2";
2563 case WINED3DSIH_LOG
:
2564 case WINED3DSIH_LOGP
:
2565 /* The precision requirements suggest that LOGP matches ARBvp's LOG
2566 * instruction, but notice that the output of those instructions is
2568 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2569 instruction
= "LG2";
2571 default: instruction
= "";
2572 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2576 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2578 src0_copy
.swizzle
= shader_arb_select_component(src0_copy
.swizzle
, 3);
2580 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2581 shader_arb_get_src_param(ins
, &src0_copy
, 0, src
);
2585 shader_addline(buffer
, "ABS TA.w, %s;\n", src
);
2586 shader_addline(buffer
, "%s%s %s, TA.w;\n", instruction
, shader_arb_get_modifier(ins
), dst
);
2590 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2595 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2597 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2600 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2601 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2602 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2604 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2605 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2607 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2608 * otherwise NRM or RSQ would return NaN */
2609 if(pshader
&& priv
->target_version
>= NV3
)
2611 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2613 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2615 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2616 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2617 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2619 else if(priv
->target_version
>= NV2
)
2621 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2622 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2623 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2628 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2630 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2631 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2632 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2634 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2635 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2637 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2638 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2639 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2644 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2646 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2648 char src_name
[3][50];
2650 /* ARB_fragment_program has a convenient LRP instruction */
2651 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2652 shader_hw_map2gl(ins
);
2656 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2657 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2658 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2659 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2661 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2662 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2663 dst_name
, src_name
[0], src_name
[2]);
2666 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2668 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2669 * must contain fixed constants. So we need a separate function to filter those constants and
2672 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2673 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2674 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2676 char src_name0
[50], src_name1
[50], src_name2
[50];
2679 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2680 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2681 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2682 /* No modifiers are supported on SCS */
2683 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2685 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2687 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2688 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2690 } else if(priv
->target_version
>= NV2
) {
2691 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2693 /* Sincos writemask must be .x, .y or .xy */
2694 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2695 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2696 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2697 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2699 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2700 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2702 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2703 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2705 * The constants we get are:
2707 * +1 +1, -1 -1 +1 +1 -1 -1
2708 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2709 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2711 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2715 * (x/2)^4 = x^4 / 16
2716 * (x/2)^5 = x^5 / 32
2719 * To get the final result:
2720 * sin(x) = 2 * sin(x/2) * cos(x/2)
2721 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2722 * (from sin(x+y) and cos(x+y) rules)
2724 * As per MSDN, dst.z is undefined after the operation, and so is
2725 * dst.x and dst.y if they're masked out by the writemask. Ie
2726 * sincos dst.y, src1, c0, c1
2727 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2728 * vsa.exe also stops with an error if the dest register is the same register as the source
2729 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2730 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2732 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2733 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2734 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2736 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2737 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2738 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2739 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2740 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2741 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2745 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2746 * properly merge that with MULs in the code above?
2747 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2748 * we can merge the sine and cosine MAD rows to calculate them together.
2750 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2751 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2752 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2753 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2756 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2757 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2758 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2760 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2762 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2763 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2765 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2767 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2768 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2773 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2775 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2778 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2780 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2781 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2783 /* SGN is only valid in vertex shaders */
2784 if(ctx
->target_version
>= NV2
) {
2785 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2789 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2790 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2792 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2793 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2795 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2796 * Then use TA, and calculate the final result
2798 * Not reading from TA? Store the first result in TA to avoid overwriting the
2799 * destination if src reg = dst reg
2801 if(strstr(src_name
, "TA"))
2803 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2804 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2805 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2809 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2810 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2811 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2816 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2818 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2824 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2825 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2826 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2828 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2829 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2832 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2834 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2835 char src0
[50], src1
[50], dst
[50];
2836 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2837 BOOL need_abs
= FALSE
;
2838 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2839 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2841 /* POW operates on the absolute value of the input */
2842 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2844 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2845 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2846 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2849 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2851 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2853 if (priv
->target_version
>= NV2
)
2855 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2856 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2857 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2861 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2862 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2864 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2865 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2866 /* Possibly add flt_eps to avoid getting float special values */
2867 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2868 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2869 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2870 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2874 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2876 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2878 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2881 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2885 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2886 struct list
*e
= list_head(&priv
->control_frames
);
2887 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2889 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2890 /* The constant loader makes sure to load -1 into iX.w */
2891 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2892 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2893 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2897 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2901 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2903 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2905 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2907 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2909 /* The constant loader makes sure to load -1 into iX.w */
2912 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2913 struct list
*e
= list_head(&priv
->control_frames
);
2914 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2916 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2918 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2919 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2920 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2924 shader_addline(buffer
, "REP %s;\n", src_name
);
2928 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2930 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2931 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2935 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2936 struct list
*e
= list_head(&priv
->control_frames
);
2937 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2939 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2940 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2941 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2943 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2947 shader_addline(buffer
, "ENDLOOP;\n");
2951 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2953 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2954 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2958 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2959 struct list
*e
= list_head(&priv
->control_frames
);
2960 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2962 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2963 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2964 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2966 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2970 shader_addline(buffer
, "ENDREP;\n");
2974 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2976 struct control_frame
*control_frame
;
2978 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2980 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2982 ERR("Could not find loop for break\n");
2986 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2988 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2989 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2990 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2994 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2998 shader_addline(buffer
, "BRK;\n");
3002 static const char *get_compare(enum wined3d_shader_rel_op op
)
3006 case WINED3D_SHADER_REL_OP_GT
: return "GT";
3007 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
3008 case WINED3D_SHADER_REL_OP_GE
: return "GE";
3009 case WINED3D_SHADER_REL_OP_LT
: return "LT";
3010 case WINED3D_SHADER_REL_OP_NE
: return "NE";
3011 case WINED3D_SHADER_REL_OP_LE
: return "LE";
3013 FIXME("Unrecognized operator %#x.\n", op
);
3018 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
3022 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
3023 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
3024 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
3025 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
3026 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
3027 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
3029 FIXME("Unrecognized operator %#x.\n", op
);
3034 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
3036 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3037 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3038 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
3041 const char *comp
= get_compare(ins
->flags
);
3043 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3044 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3048 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
3049 * away the subtraction result
3051 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3052 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3056 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3057 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3061 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3063 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3064 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3065 struct list
*e
= list_head(&priv
->control_frames
);
3066 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3070 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3072 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3073 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3077 /* Invert the flag. We jump to the else label if the condition is NOT true */
3078 comp
= get_compare(invert_compare(ins
->flags
));
3079 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3080 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3084 comp
= get_compare(ins
->flags
);
3085 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3086 shader_addline(buffer
, "IF %s.x;\n", comp
);
3090 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3092 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3093 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3094 struct list
*e
= list_head(&priv
->control_frames
);
3095 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3096 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3100 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3101 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3102 control_frame
->had_else
= TRUE
;
3106 shader_addline(buffer
, "ELSE;\n");
3110 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3112 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3113 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3114 struct list
*e
= list_head(&priv
->control_frames
);
3115 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3116 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3120 if(control_frame
->had_else
)
3122 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3126 shader_addline(buffer
, "#No else branch. else is endif\n");
3127 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3132 shader_addline(buffer
, "ENDIF;\n");
3136 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3138 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3140 char reg_src
[3][40];
3141 WORD flags
= TEX_DERIV
;
3143 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3144 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3145 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3146 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3148 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3149 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3151 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3154 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3156 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3159 WORD flags
= TEX_LOD
;
3161 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3162 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3164 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3165 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3167 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3170 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3172 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3173 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3175 priv
->in_main_func
= FALSE
;
3176 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3177 * subroutine, don't generate a label that will make GL complain
3179 if(priv
->target_version
== ARB
) return;
3181 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3184 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3185 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3186 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3187 struct wined3d_string_buffer
*buffer
)
3191 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3192 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3193 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3194 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3196 if (args
->super
.fog_src
== VS_FOG_Z
)
3198 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3204 /* posFixup.x is always 1.0, so we can safely use it */
3205 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3209 /* Clamp fogcoord */
3210 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3211 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3213 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3214 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3218 /* Clipplanes are always stored without y inversion */
3219 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3221 if (args
->super
.clip_enabled
)
3223 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3225 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3229 else if (args
->clip
.boolclip
.clip_texcoord
)
3231 static const char component
[4] = {'x', 'y', 'z', 'w'};
3232 unsigned int cur_clip
= 0;
3233 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3235 for (i
= 0; i
< gl_info
->limits
.user_clip_distances
; ++i
)
3237 if (args
->clip
.boolclip
.clipplane_mask
& (1u << i
))
3239 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3240 component
[cur_clip
++], i
);
3246 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3249 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3252 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3255 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3258 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3259 args
->clip
.boolclip
.clip_texcoord
- 1);
3262 /* Write the final position.
3264 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3265 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3266 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3267 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3269 if (!gl_info
->supported
[ARB_CLIP_CONTROL
])
3271 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3272 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3273 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3275 /* Z coord [0;1]->[-1;1] mapping, see comment in
3276 * get_projection_matrix() in utils.c. */
3277 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3279 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3280 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3284 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3285 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3289 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3291 priv_ctx
->footer_written
= TRUE
;
3294 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3296 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3297 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3298 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3299 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3301 if(priv
->target_version
== ARB
) return;
3305 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3306 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, ins
->ctx
->gl_info
, buffer
);
3309 shader_addline(buffer
, "RET;\n");
3312 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3314 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3315 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3318 static BOOL
shader_arb_compile(const struct wined3d_gl_info
*gl_info
, GLenum target
, const char *src
)
3320 const char *ptr
, *line
;
3323 if (TRACE_ON(d3d_shader
))
3326 while ((line
= get_line(&ptr
))) TRACE_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3329 GL_EXTCALL(glProgramStringARB(target
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(src
), src
));
3330 checkGLcall("glProgramStringARB()");
3332 if (FIXME_ON(d3d_shader
))
3334 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3337 FIXME_(d3d_shader
)("Program error at position %d: %s\n\n", pos
,
3338 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3340 while ((line
= get_line(&ptr
))) FIXME_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3341 FIXME_(d3d_shader
)("\n");
3347 if (WARN_ON(d3d_perf
))
3349 GL_EXTCALL(glGetProgramivARB(target
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3350 checkGLcall("glGetProgramivARB()");
3352 WARN_(d3d_perf
)("Program exceeds native resource limits.\n");
3358 static void arbfp_add_sRGB_correction(struct wined3d_string_buffer
*buffer
, const char *fragcolor
,
3359 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3361 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3365 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3366 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.x;\n", tmp1
, fragcolor
);
3367 /* Calculate the > 0.0031308 case */
3368 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3369 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3370 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3371 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts0.y;\n", fragcolor
, fragcolor
);
3372 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts0.z;\n", fragcolor
, fragcolor
);
3373 /* Calculate the < case */
3374 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts0.w, %s;\n", fragcolor
, fragcolor
);
3378 /* Calculate the > 0.0031308 case */
3379 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts0.x;\n", tmp1
, fragcolor
);
3380 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts0.x;\n", tmp1
, fragcolor
);
3381 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts0.x;\n", tmp1
, fragcolor
);
3382 shader_addline(buffer
, "MUL %s, %s, srgb_consts0.y;\n", tmp1
, tmp1
);
3383 shader_addline(buffer
, "SUB %s, %s, srgb_consts0.z;\n", tmp1
, tmp1
);
3384 /* Calculate the < case */
3385 shader_addline(buffer
, "MUL %s, srgb_consts0.w, %s;\n", tmp2
, fragcolor
);
3386 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3387 shader_addline(buffer
, "SLT %s, srgb_consts1.x, %s;\n", tmp3
, fragcolor
);
3388 shader_addline(buffer
, "SGE %s, srgb_consts1.x, %s;\n", tmp4
, fragcolor
);
3389 /* Store the components > 0.0031308 in the destination */
3390 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3391 /* Add the components that are < 0.0031308 */
3392 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3393 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3394 * result.color writes(.rgb first, then .a), or handle overwriting already written
3395 * components. The assembler uses a temporary register in this case, which is usually
3396 * not allocated from one of our registers that were used earlier.
3399 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3402 static const DWORD
*find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3404 const struct wined3d_shader_lconst
*constant
;
3406 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3408 if (constant
->idx
== idx
)
3410 return constant
->value
;
3416 static void init_ps_input(const struct wined3d_shader
*shader
,
3417 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3419 static const char * const texcoords
[8] =
3421 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3422 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3425 const struct wined3d_shader_signature_element
*input
;
3426 const char *semantic_name
;
3429 switch(args
->super
.vp_mode
)
3431 case pretransformed
:
3433 /* The pixelshader has to collect the varyings on its own. In any case properly load
3434 * color0 and color1. In the case of pretransformed vertices also load texcoords. Set
3435 * other attribs to 0.0.
3437 * For fixedfunction this behavior is correct, according to the tests. For pretransformed
3438 * we'd either need a replacement shader that can load other attribs like BINORMAL, or
3439 * load the texcoord attrib pointers to match the pixel shader signature
3441 for (i
= 0; i
< shader
->input_signature
.element_count
; ++i
)
3443 input
= &shader
->input_signature
.elements
[i
];
3444 if (!(semantic_name
= input
->semantic_name
))
3446 semantic_idx
= input
->semantic_idx
;
3448 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3451 priv
->ps_input
[input
->register_idx
] = "fragment.color.primary";
3452 else if (semantic_idx
== 1)
3453 priv
->ps_input
[input
->register_idx
] = "fragment.color.secondary";
3455 priv
->ps_input
[input
->register_idx
] = "0.0";
3457 else if (args
->super
.vp_mode
== fixedfunction
)
3459 priv
->ps_input
[input
->register_idx
] = "0.0";
3461 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3463 if (semantic_idx
< 8)
3464 priv
->ps_input
[input
->register_idx
] = texcoords
[semantic_idx
];
3466 priv
->ps_input
[input
->register_idx
] = "0.0";
3468 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3471 priv
->ps_input
[input
->register_idx
] = "fragment.fogcoord";
3473 priv
->ps_input
[input
->register_idx
] = "0.0";
3477 priv
->ps_input
[input
->register_idx
] = "0.0";
3480 TRACE("v%u, semantic %s%u is %s\n", input
->register_idx
,
3481 semantic_name
, semantic_idx
, priv
->ps_input
[input
->register_idx
]);
3486 /* That one is easy. The vertex shaders provide v0-v7 in fragment.texcoord and v8 and v9 in
3489 for(i
= 0; i
< 8; i
++)
3491 priv
->ps_input
[i
] = texcoords
[i
];
3493 priv
->ps_input
[8] = "fragment.color.primary";
3494 priv
->ps_input
[9] = "fragment.color.secondary";
3499 static void arbfp_add_linear_fog(struct wined3d_string_buffer
*buffer
,
3500 const char *fragcolor
, const char *tmp
)
3502 shader_addline(buffer
, "SUB %s.x, state.fog.params.z, fragment.fogcoord.x;\n", tmp
);
3503 shader_addline(buffer
, "MUL_SAT %s.x, %s.x, state.fog.params.w;\n", tmp
, tmp
);
3504 shader_addline(buffer
, "LRP %s.rgb, %s.x, %s, state.fog.color;\n", fragcolor
, tmp
, fragcolor
);
3507 /* Context activation is done by the caller. */
3508 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3509 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
3510 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3512 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3515 DWORD next_local
= 0;
3516 struct shader_arb_ctx_priv priv_ctx
;
3517 BOOL dcl_td
= FALSE
;
3518 BOOL want_nv_prog
= FALSE
;
3519 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3521 BOOL custom_linear_fog
= FALSE
;
3525 unsigned int i
, found
= 0;
3527 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3530 || (shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3531 || (reg_maps
->shader_version
.major
< 2 && !i
))
3534 sprintf(srgbtmp
[found
], "R%u", i
);
3536 if (found
== 4) break;
3541 sprintf(srgbtmp
[0], "TA");
3542 sprintf(srgbtmp
[1], "TB");
3543 sprintf(srgbtmp
[2], "TC");
3544 sprintf(srgbtmp
[3], "TD");
3548 sprintf(srgbtmp
[1], "TA");
3549 sprintf(srgbtmp
[2], "TB");
3550 sprintf(srgbtmp
[3], "TC");
3553 sprintf(srgbtmp
[2], "TA");
3554 sprintf(srgbtmp
[3], "TB");
3557 sprintf(srgbtmp
[3], "TA");
3563 /* Create the hw ARB shader */
3564 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3565 priv_ctx
.cur_ps_args
= args
;
3566 priv_ctx
.compiled_fprog
= compiled
;
3567 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3568 init_ps_input(shader
, args
, &priv_ctx
);
3569 list_init(&priv_ctx
.control_frames
);
3570 priv_ctx
.ps_post_process
= args
->super
.srgb_correction
;
3572 /* Avoid enabling NV_fragment_program* if we do not need it.
3574 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3575 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3576 * is faster than what we gain from using higher native instructions. There are some things though
3577 * that cannot be emulated. In that case enable the extensions.
3578 * If the extension is enabled, instruction handlers that support both ways will use it.
3580 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3581 * So enable the best we can get.
3583 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3584 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3586 want_nv_prog
= TRUE
;
3589 shader_addline(buffer
, "!!ARBfp1.0\n");
3590 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3592 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3593 priv_ctx
.target_version
= NV3
;
3595 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3597 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3598 priv_ctx
.target_version
= NV2
;
3602 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3605 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3608 priv_ctx
.target_version
= ARB
;
3611 if (reg_maps
->rt_mask
> 1)
3613 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3616 if (reg_maps
->shader_version
.major
< 3)
3618 switch (args
->super
.fog
)
3620 case WINED3D_FFP_PS_FOG_OFF
:
3622 case WINED3D_FFP_PS_FOG_LINEAR
:
3623 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
3625 custom_linear_fog
= TRUE
;
3626 priv_ctx
.ps_post_process
= TRUE
;
3629 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3631 case WINED3D_FFP_PS_FOG_EXP
:
3632 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3634 case WINED3D_FFP_PS_FOG_EXP2
:
3635 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3640 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3641 * unused temps away(but occupies them for the whole shader if they're used once). Always
3642 * declaring them avoids tricky bookkeeping work
3644 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3645 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3646 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3647 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3648 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3649 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3650 wined3d_ftoa(eps
, ftoa_tmp
);
3651 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %s, 0.0 };\n", ftoa_tmp
);
3653 if (reg_maps
->shader_version
.major
< 2)
3655 strcpy(fragcolor
, "R0");
3659 if (priv_ctx
.ps_post_process
)
3661 if (shader
->u
.ps
.color0_mov
)
3663 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3667 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3668 strcpy(fragcolor
, "TMP_COLOR");
3671 strcpy(fragcolor
, "result.color");
3675 if (args
->super
.srgb_correction
)
3677 shader_addline(buffer
, "PARAM srgb_consts0 = ");
3678 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const0
);
3679 shader_addline(buffer
, ";\n");
3680 shader_addline(buffer
, "PARAM srgb_consts1 = ");
3681 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const1
);
3682 shader_addline(buffer
, ";\n");
3685 /* Base Declarations */
3686 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3688 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3690 unsigned char bump_const
;
3692 if (!(map
& 1)) continue;
3694 bump_const
= compiled
->numbumpenvmatconsts
;
3695 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3696 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3697 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3698 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3700 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3701 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3702 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3703 * textures due to conditional NP2 restrictions)
3705 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3706 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3707 * their location is shader dependent anyway and they cannot be loaded globally.
3709 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3710 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3711 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3712 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3714 if (!(reg_maps
->luminanceparams
& (1u << i
)))
3717 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3718 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3719 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3722 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
3724 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3725 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
3727 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
3731 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3732 control_values
[0], control_values
[1], control_values
[2]);
3736 compiled
->int_consts
[i
] = next_local
;
3737 compiled
->num_int_consts
++;
3738 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3743 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3745 compiled
->ycorrection
= next_local
;
3746 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3750 shader_addline(buffer
, "TEMP vpos;\n");
3751 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3752 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3753 * ycorrection.z: 1.0
3754 * ycorrection.w: 0.0
3756 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3757 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3762 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3765 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3766 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3767 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3768 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3769 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3770 * shader compilation errors and the subsequent errors when drawing with this shader. */
3771 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3772 unsigned char cur_fixup_sampler
= 0;
3774 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3775 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3776 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3778 fixup
->offset
= next_local
;
3779 fixup
->super
.active
= 0;
3781 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
)
3783 if (!(map
& (1u << i
)))
3786 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
)
3788 fixup
->super
.active
|= (1u << i
);
3789 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3793 FIXME("No free constant found to load NP2 fixup data into shader. "
3794 "Sampling from this texture will probably look wrong.\n");
3799 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3800 if (fixup
->super
.num_consts
) {
3801 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3802 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3806 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3808 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3811 /* Base Shader Body */
3812 if (FAILED(shader_generate_main(shader
, buffer
, reg_maps
, &priv_ctx
)))
3815 if(args
->super
.srgb_correction
) {
3816 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3817 priv_ctx
.target_version
>= NV2
);
3820 if (custom_linear_fog
)
3821 arbfp_add_linear_fog(buffer
, fragcolor
, "TA");
3823 if(strcmp(fragcolor
, "result.color")) {
3824 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3826 shader_addline(buffer
, "END\n");
3828 /* TODO: change to resource.glObjectHandle or something like that */
3829 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3831 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3832 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3834 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3835 if (!shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
->buffer
))
3841 static int compare_sig(const struct wined3d_shader_signature
*sig1
, const struct wined3d_shader_signature
*sig2
)
3846 if (sig1
->element_count
!= sig2
->element_count
)
3847 return sig1
->element_count
< sig2
->element_count
? -1 : 1;
3849 for (i
= 0; i
< sig1
->element_count
; ++i
)
3851 const struct wined3d_shader_signature_element
*e1
, *e2
;
3853 e1
= &sig1
->elements
[i
];
3854 e2
= &sig2
->elements
[i
];
3856 if (!e1
->semantic_name
|| !e2
->semantic_name
)
3858 /* Compare pointers, not contents. One string is NULL (element
3859 * does not exist), the other one is not NULL. */
3860 if (e1
->semantic_name
!= e2
->semantic_name
)
3861 return e1
->semantic_name
< e2
->semantic_name
? -1 : 1;
3865 if ((ret
= strcmp(e1
->semantic_name
, e2
->semantic_name
)))
3867 if (e1
->semantic_idx
!= e2
->semantic_idx
)
3868 return e1
->semantic_idx
< e2
->semantic_idx
? -1 : 1;
3869 if (e1
->sysval_semantic
!= e2
->sysval_semantic
)
3870 return e1
->sysval_semantic
< e2
->sysval_semantic
? -1 : 1;
3871 if (e1
->component_type
!= e2
->component_type
)
3872 return e1
->component_type
< e2
->component_type
? -1 : 1;
3873 if (e1
->register_idx
!= e2
->register_idx
)
3874 return e1
->register_idx
< e2
->register_idx
? -1 : 1;
3875 if (e1
->mask
!= e2
->mask
)
3876 return e1
->mask
< e2
->mask
? -1 : 1;
3881 static void clone_sig(struct wined3d_shader_signature
*new, const struct wined3d_shader_signature
*sig
)
3886 new->element_count
= sig
->element_count
;
3887 new->elements
= wined3d_calloc(new->element_count
, sizeof(*new->elements
));
3888 for (i
= 0; i
< sig
->element_count
; ++i
)
3890 new->elements
[i
] = sig
->elements
[i
];
3892 if (!new->elements
[i
].semantic_name
)
3895 /* Clone the semantic string */
3896 name
= HeapAlloc(GetProcessHeap(), 0, strlen(sig
->elements
[i
].semantic_name
) + 1);
3897 strcpy(name
, sig
->elements
[i
].semantic_name
);
3898 new->elements
[i
].semantic_name
= name
;
3902 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature
*sig
)
3904 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3905 struct ps_signature
*found_sig
;
3909 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3910 TRACE("Found existing signature %u\n", found_sig
->idx
);
3911 return found_sig
->idx
;
3913 found_sig
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*found_sig
));
3914 clone_sig(&found_sig
->sig
, sig
);
3915 found_sig
->idx
= priv
->ps_sig_number
++;
3916 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
3917 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
3919 ERR("Failed to insert program entry.\n");
3921 return found_sig
->idx
;
3924 static void init_output_registers(const struct wined3d_shader
*shader
,
3925 const struct wined3d_shader_signature
*ps_input_sig
,
3926 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
3929 static const char * const texcoords
[8] =
3931 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3932 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
3934 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
3935 * and varying 9 to result.color.secondary
3937 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
3939 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3940 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
3941 "result.color.primary", "result.color.secondary"
3946 TRACE("Pixel shader uses builtin varyings\n");
3947 /* Map builtins to builtins */
3948 for(i
= 0; i
< 8; i
++)
3950 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
3952 priv_ctx
->color_output
[0] = "result.color.primary";
3953 priv_ctx
->color_output
[1] = "result.color.secondary";
3954 priv_ctx
->fog_output
= "TMP_FOGCOORD";
3956 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
3957 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
3959 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
3961 if (!output
->semantic_name
)
3964 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
))
3966 TRACE("o%u is TMP_OUT\n", output
->register_idx
);
3967 if (!output
->semantic_idx
)
3968 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
3970 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3972 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
3974 TRACE("o%u is result.pointsize\n", output
->register_idx
);
3975 if (!output
->semantic_idx
)
3976 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
3978 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3980 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3982 TRACE("o%u is result.color.?, idx %u\n", output
->register_idx
, output
->semantic_idx
);
3983 if (!output
->semantic_idx
)
3984 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.primary";
3985 else if (output
->semantic_idx
== 1)
3986 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.secondary";
3987 else priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3989 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3991 TRACE("o%u is result.texcoord[%u]\n", output
->register_idx
, output
->semantic_idx
);
3992 if (output
->semantic_idx
>= 8)
3993 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3995 priv_ctx
->vs_output
[output
->register_idx
] = texcoords
[output
->semantic_idx
];
3997 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
3999 TRACE("o%u is result.fogcoord\n", output
->register_idx
);
4000 if (output
->semantic_idx
> 0)
4001 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4003 priv_ctx
->vs_output
[output
->register_idx
] = "result.fogcoord";
4007 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4013 TRACE("Pixel shader uses declared varyings\n");
4015 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4016 for(i
= 0; i
< 8; i
++)
4018 priv_ctx
->texcrd_output
[i
] = "TA";
4020 priv_ctx
->color_output
[0] = "TA";
4021 priv_ctx
->color_output
[1] = "TA";
4022 priv_ctx
->fog_output
= "TA";
4024 for (i
= 0; i
< ps_input_sig
->element_count
; ++i
)
4026 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[i
];
4028 if (!input
->semantic_name
)
4031 /* If a declared input register is not written by builtin arguments, don't write to it.
4032 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4034 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4035 * to TMP_OUT in any case
4037 if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4039 if (input
->semantic_idx
< 8)
4040 priv_ctx
->texcrd_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4042 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4044 if (input
->semantic_idx
< 2)
4045 priv_ctx
->color_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4047 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4049 if (!input
->semantic_idx
)
4050 priv_ctx
->fog_output
= decl_idx_to_string
[input
->register_idx
];
4057 if (!strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.primary")
4058 || !strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.secondary"))
4060 compiled
->need_color_unclamp
= TRUE
;
4064 /* Map declared to declared */
4065 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4067 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4069 /* Write unread output to TA to throw them away */
4070 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4072 if (!output
->semantic_name
)
4075 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
) && !output
->semantic_idx
)
4077 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4080 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
) && !output
->semantic_idx
)
4082 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4086 for (j
= 0; j
< ps_input_sig
->element_count
; ++j
)
4088 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[j
];
4090 if (!input
->semantic_name
)
4093 if (!strcmp(input
->semantic_name
, output
->semantic_name
)
4094 && input
->semantic_idx
== output
->semantic_idx
)
4096 priv_ctx
->vs_output
[output
->register_idx
] = decl_idx_to_string
[input
->register_idx
];
4098 if (!strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.primary")
4099 || !strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.secondary"))
4101 compiled
->need_color_unclamp
= TRUE
;
4108 /* Context activation is done by the caller. */
4109 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4110 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
4111 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4112 const struct wined3d_shader_signature
*ps_input_sig
)
4114 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4115 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4116 struct shader_arb_priv
*priv
= shader
->device
->shader_priv
;
4118 DWORD next_local
= 0;
4119 struct shader_arb_ctx_priv priv_ctx
;
4122 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4123 priv_ctx
.cur_vs_args
= args
;
4124 list_init(&priv_ctx
.control_frames
);
4125 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4127 /* Create the hw ARB shader */
4128 shader_addline(buffer
, "!!ARBvp1.0\n");
4130 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4131 * mesurable performance penalty, and we can always make use of it for clipplanes.
4133 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4135 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4136 priv_ctx
.target_version
= NV3
;
4137 shader_addline(buffer
, "ADDRESS aL;\n");
4139 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4141 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4142 priv_ctx
.target_version
= NV2
;
4143 shader_addline(buffer
, "ADDRESS aL;\n");
4145 priv_ctx
.target_version
= ARB
;
4148 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4150 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4151 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4154 wined3d_ftoa(eps
, ftoa_tmp
);
4155 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %s};\n", ftoa_tmp
);
4157 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4159 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4160 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4163 shader_addline(buffer
, "TEMP TA;\n");
4164 shader_addline(buffer
, "TEMP TB;\n");
4166 /* Base Declarations */
4167 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4168 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4170 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4172 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4173 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
4175 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
4179 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4180 control_values
[0], control_values
[1], control_values
[2]);
4184 compiled
->int_consts
[i
] = next_local
;
4185 compiled
->num_int_consts
++;
4186 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4191 /* We need a constant to fixup the final position */
4192 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4193 compiled
->pos_fixup
= next_local
++;
4195 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4196 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4197 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4198 * a replacement shader depend on the texcoord.w being set properly.
4200 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4201 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4202 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4203 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4204 * this can eat a number of instructions, so skip it unless this cap is set as well
4206 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4208 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4209 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4211 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !priv
->ffp_proj_control
)
4214 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4215 for(i
= 0; i
< MAX_REG_TEXCRD
; i
++)
4217 if (reg_maps
->u
.texcoord_mask
[i
] && reg_maps
->u
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
)
4218 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4223 /* The shader starts with the main function */
4224 priv_ctx
.in_main_func
= TRUE
;
4225 /* Base Shader Body */
4226 if (FAILED(shader_generate_main(shader
, buffer
, reg_maps
, &priv_ctx
)))
4229 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4230 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4232 shader_addline(buffer
, "END\n");
4234 /* TODO: change to resource.glObjectHandle or something like that */
4235 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4237 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4238 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4240 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4241 if (!shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, buffer
->buffer
))
4247 /* Context activation is done by the caller. */
4248 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_shader
*shader
,
4249 const struct arb_ps_compile_args
*args
)
4251 struct wined3d_device
*device
= shader
->device
;
4252 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4253 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4256 struct arb_ps_compiled_shader
*new_array
;
4257 struct wined3d_string_buffer buffer
;
4258 struct arb_pshader_private
*shader_data
;
4261 if (!shader
->backend_data
)
4263 struct shader_arb_priv
*priv
= device
->shader_priv
;
4265 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4266 shader_data
= shader
->backend_data
;
4267 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4269 if (shader
->reg_maps
.shader_version
.major
< 3)
4270 shader_data
->input_signature_idx
= ~0U;
4272 shader_data
->input_signature_idx
= find_input_signature(priv
, &shader
->input_signature
);
4274 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4276 if (!d3d_info
->vs_clipping
)
4277 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4278 d3d_info
->limits
.ffp_blend_stages
- 1);
4280 shader_data
->clipplane_emulation
= ~0U;
4282 shader_data
= shader
->backend_data
;
4284 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4285 * so a linear search is more performant than a hashmap or a binary search
4286 * (cache coherency etc)
4288 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4290 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4291 return &shader_data
->gl_shaders
[i
];
4294 TRACE("No matching GL shader found, compiling a new shader\n");
4295 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4296 if (shader_data
->num_gl_shaders
)
4298 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4299 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4300 new_size
* sizeof(*shader_data
->gl_shaders
));
4302 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4307 ERR("Out of memory\n");
4310 shader_data
->gl_shaders
= new_array
;
4311 shader_data
->shader_array_size
= new_size
;
4314 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4316 pixelshader_update_resource_types(shader
, args
->super
.tex_types
);
4318 if (!string_buffer_init(&buffer
))
4320 ERR("Failed to initialize shader buffer.\n");
4324 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4325 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4326 string_buffer_free(&buffer
);
4327 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4329 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4332 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4333 const DWORD use_map
, BOOL skip_int
) {
4334 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4335 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4336 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4337 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4338 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4339 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4340 if(skip_int
) return TRUE
;
4342 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4345 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4346 const struct wined3d_gl_info
*gl_info
, DWORD use_map
, const struct arb_vs_compile_args
*args
,
4347 const struct wined3d_shader_signature
*ps_input_sig
)
4351 struct arb_vs_compiled_shader
*new_array
;
4352 struct wined3d_string_buffer buffer
;
4353 struct arb_vshader_private
*shader_data
;
4356 if (!shader
->backend_data
)
4358 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4360 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4361 shader_data
= shader
->backend_data
;
4363 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4364 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4366 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4368 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4369 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4370 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4372 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4373 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4374 else if (reg_maps
->max_rel_offset
> 63)
4375 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4378 shader_data
= shader
->backend_data
;
4380 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4381 * so a linear search is more performant than a hashmap or a binary search
4382 * (cache coherency etc)
4384 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4385 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4386 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4388 return &shader_data
->gl_shaders
[i
];
4392 TRACE("No matching GL shader found, compiling a new shader\n");
4394 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4395 if (shader_data
->num_gl_shaders
)
4397 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4398 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4399 new_size
* sizeof(*shader_data
->gl_shaders
));
4401 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4406 ERR("Out of memory\n");
4409 shader_data
->gl_shaders
= new_array
;
4410 shader_data
->shader_array_size
= new_size
;
4413 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4415 if (!string_buffer_init(&buffer
))
4417 ERR("Failed to initialize shader buffer.\n");
4421 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4422 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4424 string_buffer_free(&buffer
);
4425 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4427 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4430 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4431 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4432 struct arb_ps_compile_args
*args
)
4434 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4435 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4439 find_ps_compile_args(state
, shader
, context
->stream_info
.position_transformed
, &args
->super
, context
);
4441 /* This forces all local boolean constants to 1 to make them stateblock independent */
4442 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4444 for (i
= 0; i
< WINED3D_MAX_CONSTS_B
; ++i
)
4446 if (state
->ps_consts_b
[i
])
4447 args
->bools
|= ( 1u << i
);
4450 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4451 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4452 * duplicate the shader than have a no-op KIL instruction in every shader
4454 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4455 && state
->render_states
[WINED3D_RS_CLIPPING
]
4456 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4461 /* Skip if unused or local, or supported natively */
4462 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4463 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4465 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4469 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4471 if (int_skip
& (1u << i
))
4473 args
->loop_ctrl
[i
][0] = 0;
4474 args
->loop_ctrl
[i
][1] = 0;
4475 args
->loop_ctrl
[i
][2] = 0;
4479 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
].x
;
4480 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
].y
;
4481 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
].z
;
4486 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4487 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4488 struct arb_vs_compile_args
*args
)
4490 const struct wined3d_device
*device
= shader
->device
;
4491 const struct wined3d_adapter
*adapter
= device
->adapter
;
4492 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4493 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4497 find_vs_compile_args(state
, shader
, context
->stream_info
.swizzle_map
, &args
->super
, d3d_info
);
4499 args
->clip
.boolclip_compare
= 0;
4502 const struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4503 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4504 args
->ps_signature
= shader_priv
->input_signature_idx
;
4506 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4510 args
->ps_signature
= ~0;
4511 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4512 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(context
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4513 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4516 if (args
->clip
.boolclip
.clip_texcoord
)
4518 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4519 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4520 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4523 /* This forces all local boolean constants to 1 to make them stateblock independent */
4524 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4525 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4526 for (i
= 0; i
< WINED3D_MAX_CONSTS_B
; ++i
)
4528 if (state
->vs_consts_b
[i
])
4529 args
->clip
.boolclip
.bools
|= (1u << i
);
4532 args
->vertex
.samplers
[0] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 0];
4533 args
->vertex
.samplers
[1] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 1];
4534 args
->vertex
.samplers
[2] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 2];
4535 args
->vertex
.samplers
[3] = 0;
4537 /* Skip if unused or local */
4538 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4539 /* This is about flow control, not clipping. */
4540 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4542 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4546 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4548 if (int_skip
& (1u << i
))
4550 args
->loop_ctrl
[i
][0] = 0;
4551 args
->loop_ctrl
[i
][1] = 0;
4552 args
->loop_ctrl
[i
][2] = 0;
4556 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
].x
;
4557 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
].y
;
4558 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
].z
;
4563 /* Context activation is done by the caller. */
4564 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
4565 const struct wined3d_state
*state
)
4567 struct shader_arb_priv
*priv
= shader_priv
;
4568 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4571 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4574 struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4575 struct arb_ps_compile_args compile_args
;
4576 struct arb_ps_compiled_shader
*compiled
;
4578 TRACE("Using pixel shader %p.\n", ps
);
4579 find_arb_ps_compile_args(state
, context
, ps
, &compile_args
);
4580 compiled
= find_arb_pshader(ps
, &compile_args
);
4581 priv
->current_fprogram_id
= compiled
->prgId
;
4582 priv
->compiled_fprog
= compiled
;
4584 /* Bind the fragment program */
4585 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4586 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4588 if (!priv
->use_arbfp_fixed_func
)
4589 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4591 /* Enable OpenGL fragment programs. */
4592 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4593 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4595 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", priv
->current_fprogram_id
);
4597 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4598 * a 1.x and newer shader, reload the first 8 constants
4600 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4602 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4603 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4604 for(i
= 0; i
< 8; i
++)
4606 priv
->pshader_const_dirty
[i
] = 1;
4608 /* Also takes care of loading local constants */
4609 shader_arb_load_constants_internal(shader_priv
, context
, state
, TRUE
, FALSE
, TRUE
);
4613 UINT rt_height
= state
->fb
->render_targets
[0]->height
;
4614 shader_arb_ps_local_constants(compiled
, context
, state
, rt_height
);
4617 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4618 if (compiled
->np2fixup_info
.super
.active
)
4619 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_NP2_FIXUP
;
4621 if (ps
->load_local_constsF
)
4622 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
4626 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4628 /* Disable only if we're not using arbfp fixed function fragment
4629 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4630 * enabled, and the fixed function pipeline will bind the fixed
4631 * function replacement shader. */
4632 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4633 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4634 priv
->current_fprogram_id
= 0;
4636 priv
->fragment_pipe
->enable_extension(gl_info
, TRUE
);
4641 struct wined3d_shader
*vs
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
4642 struct arb_vs_compile_args compile_args
;
4643 struct arb_vs_compiled_shader
*compiled
;
4644 const struct wined3d_shader_signature
*ps_input_sig
;
4646 TRACE("Using vertex shader %p\n", vs
);
4647 find_arb_vs_compile_args(state
, context
, vs
, &compile_args
);
4649 /* Instead of searching for the signature in the signature list, read the one from the
4650 * current pixel shader. It's maybe not the shader where the signature came from, but it
4651 * is the same signature and faster to find. */
4652 if (compile_args
.ps_signature
== ~0U)
4653 ps_input_sig
= NULL
;
4655 ps_input_sig
= &state
->shader
[WINED3D_SHADER_TYPE_PIXEL
]->input_signature
;
4657 compiled
= find_arb_vshader(vs
, context
->gl_info
, context
->stream_info
.use_map
,
4658 &compile_args
, ps_input_sig
);
4659 priv
->current_vprogram_id
= compiled
->prgId
;
4660 priv
->compiled_vprog
= compiled
;
4662 /* Bind the vertex program */
4663 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4664 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4666 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4668 /* Enable OpenGL vertex programs */
4669 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4670 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4671 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", priv
->current_vprogram_id
);
4672 shader_arb_vs_local_constants(compiled
, context
, state
);
4674 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4675 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4677 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4679 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4680 checkGLcall("glClampColorARB");
4682 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4686 if (vs
->load_local_constsF
)
4687 context
->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
4691 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4693 priv
->current_vprogram_id
= 0;
4694 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4695 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4697 priv
->vertex_pipe
->vp_enable(gl_info
, TRUE
);
4701 static void shader_arb_select_compute(void *shader_priv
, struct wined3d_context
*context
,
4702 const struct wined3d_state
*state
)
4704 ERR("Compute pipeline not supported by the ARB shader backend.\n");
4707 /* Context activation is done by the caller. */
4708 static void shader_arb_disable(void *shader_priv
, struct wined3d_context
*context
)
4710 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4711 struct shader_arb_priv
*priv
= shader_priv
;
4713 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4715 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4716 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4717 priv
->current_fprogram_id
= 0;
4719 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4721 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4723 priv
->current_vprogram_id
= 0;
4724 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4725 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4727 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4729 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4731 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4732 checkGLcall("glClampColorARB");
4733 priv
->last_vs_color_unclamp
= FALSE
;
4736 context
->shader_update_mask
= (1u << WINED3D_SHADER_TYPE_PIXEL
)
4737 | (1u << WINED3D_SHADER_TYPE_VERTEX
)
4738 | (1u << WINED3D_SHADER_TYPE_GEOMETRY
)
4739 | (1u << WINED3D_SHADER_TYPE_HULL
)
4740 | (1u << WINED3D_SHADER_TYPE_DOMAIN
)
4741 | (1u << WINED3D_SHADER_TYPE_COMPUTE
);
4744 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4746 struct wined3d_device
*device
= shader
->device
;
4747 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4749 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4751 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4754 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4756 if (shader_data
->num_gl_shaders
)
4758 struct wined3d_context
*context
= context_acquire(device
, NULL
, 0);
4760 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4762 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4763 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4766 context_release(context
);
4769 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4770 HeapFree(GetProcessHeap(), 0, shader_data
);
4771 shader
->backend_data
= NULL
;
4775 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4778 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4780 if (shader_data
->num_gl_shaders
)
4782 struct wined3d_context
*context
= context_acquire(device
, NULL
, 0);
4784 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4786 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4787 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4790 context_release(context
);
4793 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4794 HeapFree(GetProcessHeap(), 0, shader_data
);
4795 shader
->backend_data
= NULL
;
4799 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4801 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4802 return compare_sig(key
, &e
->sig
);
4805 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4806 const struct fragment_pipeline
*fragment_pipe
)
4808 struct shader_arb_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
));
4809 struct fragment_caps fragment_caps
;
4810 void *vertex_priv
, *fragment_priv
;
4811 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4813 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4815 ERR("Failed to initialize vertex pipe.\n");
4816 HeapFree(GetProcessHeap(), 0, priv
);
4820 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4822 ERR("Failed to initialize fragment pipe.\n");
4823 vertex_pipe
->vp_free(device
);
4824 HeapFree(GetProcessHeap(), 0, priv
);
4828 memset(priv
->vshader_const_dirty
, 1,
4829 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4830 memset(priv
->pshader_const_dirty
, 1,
4831 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4833 wine_rb_init(&priv
->signature_tree
, sig_tree_compare
);
4835 priv
->vertex_pipe
= vertex_pipe
;
4836 priv
->fragment_pipe
= fragment_pipe
;
4837 fragment_pipe
->get_caps(&device
->adapter
->gl_info
, &fragment_caps
);
4838 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
4840 device
->vertex_priv
= vertex_priv
;
4841 device
->fragment_priv
= fragment_priv
;
4842 device
->shader_priv
= priv
;
4847 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
4849 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4852 for (i
= 0; i
< sig
->sig
.element_count
; ++i
)
4854 HeapFree(GetProcessHeap(), 0, (char *)sig
->sig
.elements
[i
].semantic_name
);
4856 HeapFree(GetProcessHeap(), 0, sig
->sig
.elements
);
4857 HeapFree(GetProcessHeap(), 0, sig
);
4860 /* Context activation is done by the caller. */
4861 static void shader_arb_free(struct wined3d_device
*device
)
4863 struct shader_arb_priv
*priv
= device
->shader_priv
;
4865 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
4866 priv
->fragment_pipe
->free_private(device
);
4867 priv
->vertex_pipe
->vp_free(device
);
4868 HeapFree(GetProcessHeap(), 0, device
->shader_priv
);
4871 static BOOL
shader_arb_allocate_context_data(struct wined3d_context
*context
)
4876 static void shader_arb_free_context_data(struct wined3d_context
*context
)
4878 struct shader_arb_priv
*priv
;
4880 priv
= context
->device
->shader_priv
;
4881 if (priv
->last_context
== context
)
4882 priv
->last_context
= NULL
;
4885 static void shader_arb_init_context_state(struct wined3d_context
*context
) {}
4887 static void shader_arb_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
4889 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4894 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
4895 * for vertex programs. If the native limit is less than that it's
4896 * not very useful, and e.g. Mesa swrast returns 0, probably to
4897 * indicate it's a software implementation. */
4898 if (gl_info
->limits
.arb_vs_native_constants
< 96)
4899 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
4901 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
4903 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4906 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
4908 else if (vs_consts
>= 256)
4910 /* Shader Model 2.0 requires at least 256 vertex shader constants */
4912 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
4917 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
4919 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
4920 caps
->vs_uniform_count
= min(WINED3D_MAX_VS_CONSTS_F
, vs_consts
);
4924 caps
->vs_version
= 0;
4925 caps
->vs_uniform_count
= 0;
4928 caps
->hs_version
= 0;
4929 caps
->ds_version
= 0;
4930 caps
->gs_version
= 0;
4931 caps
->cs_version
= 0;
4933 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4938 /* Similar as above for vertex programs, but the minimum for fragment
4939 * programs is 24. */
4940 if (gl_info
->limits
.arb_ps_native_constants
< 24)
4941 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
4943 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
4945 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
4948 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
4950 else if (ps_consts
>= 32)
4952 /* Shader Model 2.0 requires at least 32 pixel shader constants */
4954 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
4959 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
4961 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
4962 caps
->ps_uniform_count
= min(WINED3D_MAX_PS_CONSTS_F
, ps_consts
);
4963 caps
->ps_1x_max_value
= 8.0f
;
4967 caps
->ps_version
= 0;
4968 caps
->ps_uniform_count
= 0;
4969 caps
->ps_1x_max_value
= 0.0f
;
4972 caps
->varying_count
= 0;
4973 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
4974 if (use_nv_clip(gl_info
))
4975 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
4978 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
4980 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4982 TRACE("Checking support for color_fixup:\n");
4983 dump_color_fixup_desc(fixup
);
4986 /* We support everything except complex conversions. */
4987 if (!is_complex_fixup(fixup
))
4993 TRACE("[FAILED]\n");
4997 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
4999 char write_mask
[20], regstr
[50];
5000 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5001 BOOL is_color
= FALSE
;
5002 const struct wined3d_shader_dst_param
*dst
;
5004 if (!ins
->dst_count
) return;
5008 if (!shift
) return; /* Saturate alone is handled by the instructions */
5010 shader_arb_get_write_mask(ins
, dst
, write_mask
);
5011 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
5013 /* Generate a line that does the output modifier computation
5014 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
5015 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
5017 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
5018 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
5021 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
5023 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
5024 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
5025 /* WINED3DSIH_AND */ NULL
,
5026 /* WINED3DSIH_ATOMIC_AND */ NULL
,
5027 /* WINED3DSIH_ATOMIC_CMP_STORE */ NULL
,
5028 /* WINED3DSIH_ATOMIC_IADD */ NULL
,
5029 /* WINED3DSIH_ATOMIC_IMAX */ NULL
,
5030 /* WINED3DSIH_ATOMIC_IMIN */ NULL
,
5031 /* WINED3DSIH_ATOMIC_OR */ NULL
,
5032 /* WINED3DSIH_ATOMIC_UMAX */ NULL
,
5033 /* WINED3DSIH_ATOMIC_UMIN */ NULL
,
5034 /* WINED3DSIH_ATOMIC_XOR */ NULL
,
5035 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5036 /* WINED3DSIH_BFI */ NULL
,
5037 /* WINED3DSIH_BFREV */ NULL
,
5038 /* WINED3DSIH_BREAK */ shader_hw_break
,
5039 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5040 /* WINED3DSIH_BREAKP */ NULL
,
5041 /* WINED3DSIH_BUFINFO */ NULL
,
5042 /* WINED3DSIH_CALL */ shader_hw_call
,
5043 /* WINED3DSIH_CALLNZ */ NULL
,
5044 /* WINED3DSIH_CASE */ NULL
,
5045 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5046 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5047 /* WINED3DSIH_CONTINUE */ NULL
,
5048 /* WINED3DSIH_COUNTBITS */ NULL
,
5049 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5050 /* WINED3DSIH_CUT */ NULL
,
5051 /* WINED3DSIH_CUT_STREAM */ NULL
,
5052 /* WINED3DSIH_DCL */ shader_hw_nop
,
5053 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5054 /* WINED3DSIH_DCL_FUNCTION_BODY */ NULL
,
5055 /* WINED3DSIH_DCL_FUNCTION_TABLE */ NULL
,
5056 /* WINED3DSIH_DCL_GLOBAL_FLAGS */ NULL
,
5057 /* WINED3DSIH_DCL_HS_FORK_PHASE_INSTANCE_COUNT */ NULL
,
5058 /* WINED3DSIH_DCL_HS_JOIN_PHASE_INSTANCE_COUNT */ NULL
,
5059 /* WINED3DSIH_DCL_HS_MAX_TESSFACTOR */ NULL
,
5060 /* WINED3DSIH_DCL_IMMEDIATE_CONSTANT_BUFFER */ NULL
,
5061 /* WINED3DSIH_DCL_INDEX_RANGE */ NULL
,
5062 /* WINED3DSIH_DCL_INDEXABLE_TEMP */ NULL
,
5063 /* WINED3DSIH_DCL_INPUT */ NULL
,
5064 /* WINED3DSIH_DCL_INPUT_CONTROL_POINT_COUNT */ NULL
,
5065 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5066 /* WINED3DSIH_DCL_INPUT_PS */ NULL
,
5067 /* WINED3DSIH_DCL_INPUT_PS_SGV */ NULL
,
5068 /* WINED3DSIH_DCL_INPUT_PS_SIV */ NULL
,
5069 /* WINED3DSIH_DCL_INPUT_SGV */ NULL
,
5070 /* WINED3DSIH_DCL_INPUT_SIV */ NULL
,
5071 /* WINED3DSIH_DCL_INTERFACE */ NULL
,
5072 /* WINED3DSIH_DCL_OUTPUT */ NULL
,
5073 /* WINED3DSIH_DCL_OUTPUT_CONTROL_POINT_COUNT */ NULL
,
5074 /* WINED3DSIH_DCL_OUTPUT_SIV */ NULL
,
5075 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5076 /* WINED3DSIH_DCL_RESOURCE_RAW */ NULL
,
5077 /* WINED3DSIH_DCL_RESOURCE_STRUCTURED */ NULL
,
5078 /* WINED3DSIH_DCL_SAMPLER */ NULL
,
5079 /* WINED3DSIH_DCL_STREAM */ NULL
,
5080 /* WINED3DSIH_DCL_TEMPS */ NULL
,
5081 /* WINED3DSIH_DCL_TESSELLATOR_DOMAIN */ NULL
,
5082 /* WINED3DSIH_DCL_TESSELLATOR_OUTPUT_PRIMITIVE */ NULL
,
5083 /* WINED3DSIH_DCL_TESSELLATOR_PARTITIONING */ NULL
,
5084 /* WINED3DSIH_DCL_TGSM_RAW */ NULL
,
5085 /* WINED3DSIH_DCL_TGSM_STRUCTURED */ NULL
,
5086 /* WINED3DSIH_DCL_THREAD_GROUP */ NULL
,
5087 /* WINED3DSIH_DCL_UAV_RAW */ NULL
,
5088 /* WINED3DSIH_DCL_UAV_STRUCTURED */ NULL
,
5089 /* WINED3DSIH_DCL_UAV_TYPED */ NULL
,
5090 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5091 /* WINED3DSIH_DEF */ shader_hw_nop
,
5092 /* WINED3DSIH_DEFAULT */ NULL
,
5093 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5094 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5095 /* WINED3DSIH_DIV */ NULL
,
5096 /* WINED3DSIH_DP2 */ NULL
,
5097 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5098 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5099 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5100 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5101 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5102 /* WINED3DSIH_DSX_COARSE */ NULL
,
5103 /* WINED3DSIH_DSX_FINE */ NULL
,
5104 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5105 /* WINED3DSIH_DSY_COARSE */ NULL
,
5106 /* WINED3DSIH_DSY_FINE */ NULL
,
5107 /* WINED3DSIH_ELSE */ shader_hw_else
,
5108 /* WINED3DSIH_EMIT */ NULL
,
5109 /* WINED3DSIH_EMIT_STREAM */ NULL
,
5110 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5111 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5112 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5113 /* WINED3DSIH_ENDSWITCH */ NULL
,
5114 /* WINED3DSIH_EQ */ NULL
,
5115 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5116 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5117 /* WINED3DSIH_F16TOF32 */ NULL
,
5118 /* WINED3DSIH_F32TOF16 */ NULL
,
5119 /* WINED3DSIH_FCALL */ NULL
,
5120 /* WINED3DSIH_FIRSTBIT_HI */ NULL
,
5121 /* WINED3DSIH_FIRSTBIT_LO */ NULL
,
5122 /* WINED3DSIH_FIRSTBIT_SHI */ NULL
,
5123 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5124 /* WINED3DSIH_FTOI */ NULL
,
5125 /* WINED3DSIH_FTOU */ NULL
,
5126 /* WINED3DSIH_GATHER4 */ NULL
,
5127 /* WINED3DSIH_GATHER4_C */ NULL
,
5128 /* WINED3DSIH_GE */ NULL
,
5129 /* WINED3DSIH_HS_CONTROL_POINT_PHASE */ NULL
,
5130 /* WINED3DSIH_HS_DECLS */ NULL
,
5131 /* WINED3DSIH_HS_FORK_PHASE */ NULL
,
5132 /* WINED3DSIH_HS_JOIN_PHASE */ NULL
,
5133 /* WINED3DSIH_IADD */ NULL
,
5134 /* WINED3DSIH_IEQ */ NULL
,
5135 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5136 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5137 /* WINED3DSIH_IGE */ NULL
,
5138 /* WINED3DSIH_ILT */ NULL
,
5139 /* WINED3DSIH_IMAD */ NULL
,
5140 /* WINED3DSIH_IMAX */ NULL
,
5141 /* WINED3DSIH_IMIN */ NULL
,
5142 /* WINED3DSIH_IMM_ATOMIC_ALLOC */ NULL
,
5143 /* WINED3DSIH_IMM_ATOMIC_AND */ NULL
,
5144 /* WINED3DSIH_IMM_ATOMIC_CMP_EXCH */ NULL
,
5145 /* WINED3DSIH_IMM_ATOMIC_CONSUME */ NULL
,
5146 /* WINED3DSIH_IMM_ATOMIC_EXCH */ NULL
,
5147 /* WINED3DSIH_IMM_ATOMIC_IADD */ NULL
,
5148 /* WINED3DSIH_IMM_ATOMIC_IMAX */ NULL
,
5149 /* WINED3DSIH_IMM_ATOMIC_IMIN */ NULL
,
5150 /* WINED3DSIH_IMM_ATOMIC_OR */ NULL
,
5151 /* WINED3DSIH_IMM_ATOMIC_UMAX */ NULL
,
5152 /* WINED3DSIH_IMM_ATOMIC_UMIN */ NULL
,
5153 /* WINED3DSIH_IMM_ATOMIC_XOR */ NULL
,
5154 /* WINED3DSIH_IMUL */ NULL
,
5155 /* WINED3DSIH_INE */ NULL
,
5156 /* WINED3DSIH_INEG */ NULL
,
5157 /* WINED3DSIH_ISHL */ NULL
,
5158 /* WINED3DSIH_ISHR */ NULL
,
5159 /* WINED3DSIH_ITOF */ NULL
,
5160 /* WINED3DSIH_LABEL */ shader_hw_label
,
5161 /* WINED3DSIH_LD */ NULL
,
5162 /* WINED3DSIH_LD2DMS */ NULL
,
5163 /* WINED3DSIH_LD_RAW */ NULL
,
5164 /* WINED3DSIH_LD_STRUCTURED */ NULL
,
5165 /* WINED3DSIH_LD_UAV_TYPED */ NULL
,
5166 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5167 /* WINED3DSIH_LOD */ NULL
,
5168 /* WINED3DSIH_LOG */ shader_hw_scalar_op
,
5169 /* WINED3DSIH_LOGP */ shader_hw_scalar_op
,
5170 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5171 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5172 /* WINED3DSIH_LT */ NULL
,
5173 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5174 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5175 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5176 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5177 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5178 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5179 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5180 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5181 /* WINED3DSIH_MOV */ shader_hw_mov
,
5182 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5183 /* WINED3DSIH_MOVC */ NULL
,
5184 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5185 /* WINED3DSIH_NE */ NULL
,
5186 /* WINED3DSIH_NOP */ shader_hw_nop
,
5187 /* WINED3DSIH_NOT */ NULL
,
5188 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5189 /* WINED3DSIH_OR */ NULL
,
5190 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5191 /* WINED3DSIH_POW */ shader_hw_pow
,
5192 /* WINED3DSIH_RCP */ shader_hw_scalar_op
,
5193 /* WINED3DSIH_REP */ shader_hw_rep
,
5194 /* WINED3DSIH_RESINFO */ NULL
,
5195 /* WINED3DSIH_RET */ shader_hw_ret
,
5196 /* WINED3DSIH_ROUND_NE */ NULL
,
5197 /* WINED3DSIH_ROUND_NI */ NULL
,
5198 /* WINED3DSIH_ROUND_PI */ NULL
,
5199 /* WINED3DSIH_ROUND_Z */ NULL
,
5200 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5201 /* WINED3DSIH_SAMPLE */ NULL
,
5202 /* WINED3DSIH_SAMPLE_B */ NULL
,
5203 /* WINED3DSIH_SAMPLE_C */ NULL
,
5204 /* WINED3DSIH_SAMPLE_C_LZ */ NULL
,
5205 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5206 /* WINED3DSIH_SAMPLE_INFO */ NULL
,
5207 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5208 /* WINED3DSIH_SAMPLE_POS */ NULL
,
5209 /* WINED3DSIH_SETP */ NULL
,
5210 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5211 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5212 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5213 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5214 /* WINED3DSIH_SQRT */ NULL
,
5215 /* WINED3DSIH_STORE_RAW */ NULL
,
5216 /* WINED3DSIH_STORE_STRUCTURED */ NULL
,
5217 /* WINED3DSIH_STORE_UAV_TYPED */ NULL
,
5218 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5219 /* WINED3DSIH_SWAPC */ NULL
,
5220 /* WINED3DSIH_SWITCH */ NULL
,
5221 /* WINED3DSIH_SYNC */ NULL
,
5222 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5223 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5224 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5225 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5226 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5227 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5228 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5229 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5230 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5231 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5232 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5233 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5234 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5235 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5236 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5237 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5238 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5239 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5240 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5241 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5242 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5243 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5244 /* WINED3DSIH_UBFE */ NULL
,
5245 /* WINED3DSIH_UDIV */ NULL
,
5246 /* WINED3DSIH_UGE */ NULL
,
5247 /* WINED3DSIH_ULT */ NULL
,
5248 /* WINED3DSIH_UMAX */ NULL
,
5249 /* WINED3DSIH_UMIN */ NULL
,
5250 /* WINED3DSIH_UMUL */ NULL
,
5251 /* WINED3DSIH_USHR */ NULL
,
5252 /* WINED3DSIH_UTOF */ NULL
,
5253 /* WINED3DSIH_XOR */ NULL
,
5256 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5257 const struct wined3d_shader
*shader
, DWORD idx
)
5259 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5260 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5261 const struct wined3d_shader_lconst
*constant
;
5263 WORD flag
= (1u << idx
);
5264 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5266 if (reg_maps
->local_bool_consts
& flag
)
5268 /* What good is an if(bool) with a hardcoded local constant? I don't know, but handle it */
5269 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5271 if (constant
->idx
== idx
)
5273 return constant
->value
[0];
5276 ERR("Local constant not found\n");
5281 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5282 else bools
= priv
->cur_ps_args
->bools
;
5283 return bools
& flag
;
5287 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5288 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5290 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5291 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5293 /* Integer constants can either be a local constant, or they can be stored in the shader
5294 * type specific compile args. */
5295 if (reg_maps
->local_int_consts
& (1u << idx
))
5297 const struct wined3d_shader_lconst
*constant
;
5299 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5301 if (constant
->idx
== idx
)
5303 loop_control
->count
= constant
->value
[0];
5304 loop_control
->start
= constant
->value
[1];
5305 /* Step is signed. */
5306 loop_control
->step
= (int)constant
->value
[2];
5310 /* If this happens the flag was set incorrectly */
5311 ERR("Local constant not found\n");
5312 loop_control
->count
= 0;
5313 loop_control
->start
= 0;
5314 loop_control
->step
= 0;
5318 switch (reg_maps
->shader_version
.type
)
5320 case WINED3D_SHADER_TYPE_VERTEX
:
5321 /* Count and aL start value are unsigned */
5322 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5323 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5324 /* Step is signed. */
5325 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5328 case WINED3D_SHADER_TYPE_PIXEL
:
5329 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5330 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5331 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5335 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5340 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5343 struct wined3d_shader_dst_param
*dst_param
;
5344 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
;
5345 struct recorded_instruction
*rec
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*rec
));
5348 ERR("Out of memory\n");
5353 dst_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
));
5354 if(!dst_param
) goto free
;
5355 *dst_param
= *ins
->dst
;
5356 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5358 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5361 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5362 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5364 rec
->ins
.dst
= dst_param
;
5366 if (!(src_param
= wined3d_calloc(ins
->src_count
, sizeof(*src_param
))))
5368 for (i
= 0; i
< ins
->src_count
; ++i
)
5370 src_param
[i
] = ins
->src
[i
];
5371 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5373 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5376 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5377 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5380 rec
->ins
.src
= src_param
;
5381 list_add_tail(list
, &rec
->entry
);
5385 ERR("Out of memory\n");
5388 HeapFree(GetProcessHeap(), 0, (void *)dst_param
->reg
.idx
[0].rel_addr
);
5389 HeapFree(GetProcessHeap(), 0, dst_param
);
5393 for(i
= 0; i
< ins
->src_count
; i
++)
5395 HeapFree(GetProcessHeap(), 0, (void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5397 HeapFree(GetProcessHeap(), 0, src_param
);
5399 HeapFree(GetProcessHeap(), 0, rec
);
5402 static void free_recorded_instruction(struct list
*list
)
5404 struct recorded_instruction
*rec_ins
, *entry2
;
5407 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5409 list_remove(&rec_ins
->entry
);
5410 if (rec_ins
->ins
.dst
)
5412 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5413 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
);
5415 if (rec_ins
->ins
.src
)
5417 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5419 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5421 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
);
5423 HeapFree(GetProcessHeap(), 0, rec_ins
);
5427 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5429 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5430 struct control_frame
*control_frame
;
5432 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5434 struct list
*e
= list_head(&priv
->control_frames
);
5435 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5436 list_remove(&control_frame
->entry
);
5437 HeapFree(GetProcessHeap(), 0, control_frame
);
5440 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5442 /* Non-ifc ENDIFs were already handled previously. */
5443 struct list
*e
= list_head(&priv
->control_frames
);
5444 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5445 list_remove(&control_frame
->entry
);
5446 HeapFree(GetProcessHeap(), 0, control_frame
);
5450 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5451 SHADER_HANDLER hw_fct
;
5452 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5453 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5454 struct control_frame
*control_frame
;
5455 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5458 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5460 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5461 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5463 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5464 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5466 if(priv
->target_version
>= NV2
)
5468 control_frame
->no
.loop
= priv
->num_loops
++;
5473 /* Don't bother recording when we're in a not used if branch */
5479 if(!priv
->recording
)
5481 list_init(&priv
->record
);
5482 priv
->recording
= TRUE
;
5483 control_frame
->outer_loop
= TRUE
;
5484 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5485 return; /* Instruction is handled */
5487 /* Record this loop in the outer loop's recording */
5490 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5492 if(priv
->target_version
>= NV2
)
5494 /* Nothing to do. The control frame is popped after the HW instr handler */
5498 struct list
*e
= list_head(&priv
->control_frames
);
5499 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5500 list_remove(&control_frame
->entry
);
5502 if(control_frame
->outer_loop
)
5504 unsigned int iteration
;
5508 /* Turn off recording before playback */
5509 priv
->recording
= FALSE
;
5511 /* Move the recorded instructions to a separate list and get them out of the private data
5512 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5513 * be recorded again, thus priv->record might be overwritten
5516 list_move_tail(©
, &priv
->record
);
5517 list_init(&priv
->record
);
5519 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5521 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5522 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5523 control_frame
->loop_control
.step
);
5524 aL
= control_frame
->loop_control
.start
;
5528 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5531 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5533 struct recorded_instruction
*rec_ins
;
5534 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5537 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5541 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5544 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5546 shader_arb_handle_instruction(&rec_ins
->ins
);
5549 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5551 aL
+= control_frame
->loop_control
.step
;
5554 shader_addline(buffer
, "#end loop/rep\n");
5556 free_recorded_instruction(©
);
5557 HeapFree(GetProcessHeap(), 0, control_frame
);
5558 return; /* Instruction is handled */
5562 /* This is a nested loop. Proceed to the normal recording function */
5563 HeapFree(GetProcessHeap(), 0, control_frame
);
5570 record_instruction(&priv
->record
, ins
);
5575 if(ins
->handler_idx
== WINED3DSIH_IF
)
5577 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5578 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5579 control_frame
->type
= IF
;
5581 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5582 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5583 bool_const
= !bool_const
;
5584 if (!priv
->muted
&& !bool_const
)
5586 shader_addline(buffer
, "#if(FALSE){\n");
5588 control_frame
->muting
= TRUE
;
5590 else shader_addline(buffer
, "#if(TRUE) {\n");
5592 return; /* Instruction is handled */
5594 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5596 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5597 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5598 control_frame
->type
= IFC
;
5599 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5600 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5602 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5604 struct list
*e
= list_head(&priv
->control_frames
);
5605 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5607 if(control_frame
->type
== IF
)
5609 shader_addline(buffer
, "#} else {\n");
5610 if(!priv
->muted
&& !control_frame
->muting
)
5613 control_frame
->muting
= TRUE
;
5615 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5616 return; /* Instruction is handled. */
5618 /* In case of an ifc, generate a HW shader instruction */
5619 if (control_frame
->type
!= IFC
)
5620 ERR("Control frame does not match.\n");
5622 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5624 struct list
*e
= list_head(&priv
->control_frames
);
5625 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5627 if(control_frame
->type
== IF
)
5629 shader_addline(buffer
, "#} endif\n");
5630 if(control_frame
->muting
) priv
->muted
= FALSE
;
5631 list_remove(&control_frame
->entry
);
5632 HeapFree(GetProcessHeap(), 0, control_frame
);
5633 return; /* Instruction is handled */
5635 /* In case of an ifc, generate a HW shader instruction */
5636 if (control_frame
->type
!= IFC
)
5637 ERR("Control frame does not match.\n");
5642 pop_control_frame(ins
);
5646 /* Select handler */
5647 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5649 /* Unhandled opcode */
5652 FIXME("Backend can't handle opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
5657 pop_control_frame(ins
);
5659 shader_arb_add_instruction_modifiers(ins
);
5662 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5664 struct shader_arb_priv
*priv
= shader_priv
;
5666 return priv
->ffp_proj_control
;
5669 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5671 shader_arb_handle_instruction
,
5673 shader_arb_select_compute
,
5675 shader_arb_update_float_vertex_constants
,
5676 shader_arb_update_float_pixel_constants
,
5677 shader_arb_load_constants
,
5681 shader_arb_allocate_context_data
,
5682 shader_arb_free_context_data
,
5683 shader_arb_init_context_state
,
5684 shader_arb_get_caps
,
5685 shader_arb_color_fixup_supported
,
5686 shader_arb_has_ffp_proj_control
,
5689 /* ARB_fragment_program fixed function pipeline replacement definitions */
5690 #define ARB_FFP_CONST_TFACTOR 0
5691 #define ARB_FFP_CONST_COLOR_KEY_LOW ((ARB_FFP_CONST_TFACTOR) + 1)
5692 #define ARB_FFP_CONST_COLOR_KEY_HIGH ((ARB_FFP_CONST_COLOR_KEY_LOW) + 1)
5693 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_COLOR_KEY_HIGH) + 1)
5694 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5695 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5696 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5698 struct arbfp_ffp_desc
5700 struct ffp_frag_desc parent
;
5704 /* Context activation is done by the caller. */
5705 static void arbfp_enable(const struct wined3d_gl_info
*gl_info
, BOOL enable
)
5709 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5710 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5714 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5715 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5719 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5721 struct shader_arb_priv
*priv
;
5723 /* Share private data between the shader backend and the pipeline
5724 * replacement, if both are the arb implementation. This is needed to
5725 * figure out whether ARBfp should be disabled if no pixel shader is bound
5727 if (shader_backend
== &arb_program_shader_backend
)
5729 else if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
5732 wine_rb_init(&priv
->fragment_shaders
, wined3d_ffp_frag_program_key_compare
);
5733 priv
->use_arbfp_fixed_func
= TRUE
;
5738 /* Context activation is done by the caller. */
5739 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5741 const struct wined3d_gl_info
*gl_info
= context
;
5742 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5744 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5745 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5746 HeapFree(GetProcessHeap(), 0, entry_arb
);
5749 /* Context activation is done by the caller. */
5750 static void arbfp_free(struct wined3d_device
*device
)
5752 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5754 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &device
->adapter
->gl_info
);
5755 priv
->use_arbfp_fixed_func
= FALSE
;
5757 if (device
->shader_backend
!= &arb_program_shader_backend
)
5759 HeapFree(GetProcessHeap(), 0, device
->fragment_priv
);
5763 static void arbfp_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5765 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5766 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
5767 | WINED3D_FRAGMENT_CAP_COLOR_KEY
;
5768 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5769 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5770 WINED3DTEXOPCAPS_SELECTARG1
|
5771 WINED3DTEXOPCAPS_SELECTARG2
|
5772 WINED3DTEXOPCAPS_MODULATE4X
|
5773 WINED3DTEXOPCAPS_MODULATE2X
|
5774 WINED3DTEXOPCAPS_MODULATE
|
5775 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5776 WINED3DTEXOPCAPS_ADDSIGNED
|
5777 WINED3DTEXOPCAPS_ADD
|
5778 WINED3DTEXOPCAPS_SUBTRACT
|
5779 WINED3DTEXOPCAPS_ADDSMOOTH
|
5780 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5781 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5782 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5783 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5784 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5785 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5786 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5787 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5788 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5789 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5790 WINED3DTEXOPCAPS_MULTIPLYADD
|
5791 WINED3DTEXOPCAPS_LERP
|
5792 WINED3DTEXOPCAPS_BUMPENVMAP
|
5793 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5795 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5797 caps
->MaxTextureBlendStages
= MAX_TEXTURES
;
5798 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.samplers
[WINED3D_SHADER_TYPE_PIXEL
], MAX_TEXTURES
);
5801 static DWORD
arbfp_get_emul_mask(const struct wined3d_gl_info
*gl_info
)
5803 return GL_EXT_EMUL_ARB_MULTITEXTURE
| GL_EXT_EMUL_EXT_FOG_COORD
;
5806 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5807 const struct wined3d_state
*state
, DWORD state_id
)
5809 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5810 struct wined3d_device
*device
= context
->device
;
5811 struct wined3d_color color
;
5813 if (device
->shader_backend
== &arb_program_shader_backend
)
5815 struct shader_arb_priv
*priv
;
5817 /* Don't load the parameter if we're using an arbfp pixel shader,
5818 * otherwise we'll overwrite application provided constants. */
5822 priv
= device
->shader_priv
;
5823 priv
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5824 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5827 wined3d_color_from_d3dcolor(&color
, state
->render_states
[WINED3D_RS_TEXTUREFACTOR
]);
5828 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, &color
.r
));
5829 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, &color.r)");
5832 static void state_tss_constant_arbfp(struct wined3d_context
*context
,
5833 const struct wined3d_state
*state
, DWORD state_id
)
5835 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5836 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5837 struct wined3d_device
*device
= context
->device
;
5838 struct wined3d_color color
;
5840 if (device
->shader_backend
== &arb_program_shader_backend
)
5842 struct shader_arb_priv
*priv
;
5844 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5845 * application provided constants.
5850 priv
= device
->shader_priv
;
5851 priv
->pshader_const_dirty
[ARB_FFP_CONST_CONSTANT(stage
)] = 1;
5852 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_CONSTANT(stage
) + 1);
5855 wined3d_color_from_d3dcolor(&color
, state
->texture_states
[stage
][WINED3D_TSS_CONSTANT
]);
5856 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_CONSTANT(stage
), &color
.r
));
5857 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_CONSTANT(stage), &color.r)");
5860 static void state_arb_specularenable(struct wined3d_context
*context
,
5861 const struct wined3d_state
*state
, DWORD state_id
)
5863 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5864 struct wined3d_device
*device
= context
->device
;
5867 if (device
->shader_backend
== &arb_program_shader_backend
)
5869 struct shader_arb_priv
*priv
;
5871 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5872 * application provided constants.
5877 priv
= device
->shader_priv
;
5878 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5879 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5882 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
5884 /* The specular color has no alpha */
5885 col
[0] = 1.0f
; col
[1] = 1.0f
;
5886 col
[2] = 1.0f
; col
[3] = 0.0f
;
5888 col
[0] = 0.0f
; col
[1] = 0.0f
;
5889 col
[2] = 0.0f
; col
[3] = 0.0f
;
5891 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5892 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5895 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5897 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5898 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5899 struct wined3d_device
*device
= context
->device
;
5902 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5904 if (device
->shader_backend
== &arb_program_shader_backend
)
5906 struct shader_arb_priv
*priv
= device
->shader_priv
;
5908 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5912 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5913 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5916 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
5917 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
5918 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
5919 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
5921 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5922 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5925 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
5926 const struct wined3d_state
*state
, DWORD state_id
)
5928 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5929 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5930 struct wined3d_device
*device
= context
->device
;
5933 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5935 if (device
->shader_backend
== &arb_program_shader_backend
)
5937 struct shader_arb_priv
*priv
= device
->shader_priv
;
5939 /* Exit now, don't set the luminance below, otherwise we may overwrite pixel shader constants. */
5943 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5944 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5947 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
5948 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
5952 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5953 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5956 static void alpha_test_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5958 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5962 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
5964 if (state
->render_states
[WINED3D_RS_ALPHATESTENABLE
])
5966 gl_info
->gl_ops
.gl
.p_glEnable(GL_ALPHA_TEST
);
5967 checkGLcall("glEnable GL_ALPHA_TEST");
5971 gl_info
->gl_ops
.gl
.p_glDisable(GL_ALPHA_TEST
);
5972 checkGLcall("glDisable GL_ALPHA_TEST");
5976 ref
= ((float)state
->render_states
[WINED3D_RS_ALPHAREF
]) / 255.0f
;
5977 glParm
= wined3d_gl_compare_func(state
->render_states
[WINED3D_RS_ALPHAFUNC
]);
5981 gl_info
->gl_ops
.gl
.p_glAlphaFunc(glParm
, ref
);
5982 checkGLcall("glAlphaFunc");
5986 static void color_key_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5988 const struct wined3d_texture
*texture
= state
->textures
[0];
5989 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5990 struct wined3d_device
*device
= context
->device
;
5991 struct wined3d_color float_key
[2];
5996 if (device
->shader_backend
== &arb_program_shader_backend
)
5998 struct shader_arb_priv
*priv
;
6000 /* Don't load the parameter if we're using an arbfp pixel shader,
6001 * otherwise we'll overwrite application provided constants. */
6005 priv
= device
->shader_priv
;
6006 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_LOW
] = 1;
6007 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_HIGH
] = 1;
6008 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_COLOR_KEY_HIGH
+ 1);
6011 wined3d_format_get_float_color_key(texture
->resource
.format
, &texture
->async
.src_blt_color_key
, float_key
);
6013 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_LOW
, &float_key
[0].r
));
6014 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_LOW, &float_key[0].r)");
6015 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_HIGH
, &float_key
[1].r
));
6016 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_HIGH, &float_key[1].r)");
6019 static const char *get_argreg(struct wined3d_string_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
6023 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
6025 switch(arg
& WINED3DTA_SELECTMASK
) {
6026 case WINED3DTA_DIFFUSE
:
6027 ret
= "fragment.color.primary"; break;
6029 case WINED3DTA_CURRENT
:
6033 case WINED3DTA_TEXTURE
:
6035 case 0: ret
= "tex0"; break;
6036 case 1: ret
= "tex1"; break;
6037 case 2: ret
= "tex2"; break;
6038 case 3: ret
= "tex3"; break;
6039 case 4: ret
= "tex4"; break;
6040 case 5: ret
= "tex5"; break;
6041 case 6: ret
= "tex6"; break;
6042 case 7: ret
= "tex7"; break;
6043 default: ret
= "unknown texture";
6047 case WINED3DTA_TFACTOR
:
6048 ret
= "tfactor"; break;
6050 case WINED3DTA_SPECULAR
:
6051 ret
= "fragment.color.secondary"; break;
6053 case WINED3DTA_TEMP
:
6054 ret
= "tempreg"; break;
6056 case WINED3DTA_CONSTANT
:
6058 case 0: ret
= "const0"; break;
6059 case 1: ret
= "const1"; break;
6060 case 2: ret
= "const2"; break;
6061 case 3: ret
= "const3"; break;
6062 case 4: ret
= "const4"; break;
6063 case 5: ret
= "const5"; break;
6064 case 6: ret
= "const6"; break;
6065 case 7: ret
= "const7"; break;
6066 default: ret
= "unknown constant";
6074 if(arg
& WINED3DTA_COMPLEMENT
) {
6075 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
6076 if(argnum
== 0) ret
= "arg0";
6077 if(argnum
== 1) ret
= "arg1";
6078 if(argnum
== 2) ret
= "arg2";
6080 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
6081 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
6082 if(argnum
== 0) ret
= "arg0";
6083 if(argnum
== 1) ret
= "arg1";
6084 if(argnum
== 2) ret
= "arg2";
6089 static void gen_ffp_instr(struct wined3d_string_buffer
*buffer
, unsigned int stage
, BOOL color
,
6090 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
6092 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6093 unsigned int mul
= 1;
6095 if(color
&& alpha
) dstmask
= "";
6096 else if(color
) dstmask
= ".xyz";
6097 else dstmask
= ".w";
6099 if(dst
== tempreg
) dstreg
= "tempreg";
6100 else dstreg
= "ret";
6102 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6103 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6104 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6108 case WINED3D_TOP_DISABLE
:
6111 case WINED3D_TOP_SELECT_ARG2
:
6114 case WINED3D_TOP_SELECT_ARG1
:
6115 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6118 case WINED3D_TOP_MODULATE_4X
:
6121 case WINED3D_TOP_MODULATE_2X
:
6124 case WINED3D_TOP_MODULATE
:
6125 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6128 case WINED3D_TOP_ADD_SIGNED_2X
:
6131 case WINED3D_TOP_ADD_SIGNED
:
6132 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6135 case WINED3D_TOP_ADD
:
6136 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6139 case WINED3D_TOP_SUBTRACT
:
6140 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6143 case WINED3D_TOP_ADD_SMOOTH
:
6144 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6145 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6148 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6149 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6150 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6152 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6153 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6154 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6156 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6157 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6158 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6160 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6161 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6162 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6165 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6166 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6167 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6168 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6171 /* D3DTOP_PREMODULATE ???? */
6173 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6174 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6175 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6177 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6178 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6180 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6181 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6182 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6184 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6185 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6188 case WINED3D_TOP_DOTPRODUCT3
:
6190 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6191 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6192 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6195 case WINED3D_TOP_MULTIPLY_ADD
:
6196 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6199 case WINED3D_TOP_LERP
:
6200 /* The msdn is not quite right here */
6201 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6204 case WINED3D_TOP_BUMPENVMAP
:
6205 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6206 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6210 FIXME("Unhandled texture op %08x\n", op
);
6214 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6216 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6219 static const char *arbfp_texture_target(enum wined3d_gl_resource_type type
)
6223 case WINED3D_GL_RES_TYPE_TEX_1D
:
6225 case WINED3D_GL_RES_TYPE_TEX_2D
:
6227 case WINED3D_GL_RES_TYPE_TEX_3D
:
6229 case WINED3D_GL_RES_TYPE_TEX_CUBE
:
6231 case WINED3D_GL_RES_TYPE_TEX_RECT
:
6234 return "unexpected_resource_type";
6238 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6240 BYTE tex_read
= 0, bump_used
= 0, luminance_used
= 0, constant_used
= 0;
6241 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6242 unsigned int stage
, lowest_disabled_stage
;
6243 struct wined3d_string_buffer buffer
;
6244 struct color_fixup_masks masks
;
6245 BOOL custom_linear_fog
= FALSE
;
6246 const char *textype
, *instr
;
6247 DWORD arg0
, arg1
, arg2
;
6248 char colorcor_dst
[8];
6252 if (!string_buffer_init(&buffer
))
6254 ERR("Failed to initialize shader buffer.\n");
6258 shader_addline(&buffer
, "!!ARBfp1.0\n");
6260 if (settings
->color_key_enabled
)
6262 shader_addline(&buffer
, "PARAM color_key_low = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_LOW
);
6263 shader_addline(&buffer
, "PARAM color_key_high = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_HIGH
);
6267 /* Find out which textures are read */
6268 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6270 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6273 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6274 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6275 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6277 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
6278 tex_read
|= 1u << stage
;
6279 if (settings
->op
[stage
].dst
== tempreg
)
6280 tempreg_used
= TRUE
;
6281 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
6282 tempreg_used
= TRUE
;
6283 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
6284 tfactor_used
= TRUE
;
6285 if (arg0
== WINED3DTA_CONSTANT
|| arg1
== WINED3DTA_CONSTANT
|| arg2
== WINED3DTA_CONSTANT
)
6286 constant_used
|= 1u << stage
;
6288 switch (settings
->op
[stage
].cop
)
6290 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6291 luminance_used
|= 1u << stage
;
6293 case WINED3D_TOP_BUMPENVMAP
:
6294 bump_used
|= 1u << stage
;
6296 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6297 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6298 tex_read
|= 1u << stage
;
6301 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6302 tfactor_used
= TRUE
;
6309 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6312 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6313 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6314 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6316 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
6317 tex_read
|= 1u << stage
;
6318 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
6319 tempreg_used
= TRUE
;
6320 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
6321 tfactor_used
= TRUE
;
6322 if (arg0
== WINED3DTA_CONSTANT
|| arg1
== WINED3DTA_CONSTANT
|| arg2
== WINED3DTA_CONSTANT
)
6323 constant_used
|= 1u << stage
;
6325 lowest_disabled_stage
= stage
;
6327 switch (settings
->fog
)
6329 case WINED3D_FFP_PS_FOG_OFF
: break;
6330 case WINED3D_FFP_PS_FOG_LINEAR
:
6331 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
6333 custom_linear_fog
= TRUE
;
6336 shader_addline(&buffer
, "OPTION ARB_fog_linear;\n");
6339 case WINED3D_FFP_PS_FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6340 case WINED3D_FFP_PS_FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6341 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6344 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6345 shader_addline(&buffer
, "TEMP TMP;\n");
6346 shader_addline(&buffer
, "TEMP ret;\n");
6347 if (tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6348 shader_addline(&buffer
, "TEMP arg0;\n");
6349 shader_addline(&buffer
, "TEMP arg1;\n");
6350 shader_addline(&buffer
, "TEMP arg2;\n");
6351 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6353 if (constant_used
& (1u << stage
))
6354 shader_addline(&buffer
, "PARAM const%u = program.env[%u];\n", stage
, ARB_FFP_CONST_CONSTANT(stage
));
6356 if (!(tex_read
& (1u << stage
)))
6359 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6361 if (!(bump_used
& (1u << stage
)))
6363 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6365 if (!(luminance_used
& (1u << stage
)))
6367 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6370 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6371 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6373 if (settings
->sRGB_write
)
6375 shader_addline(&buffer
, "PARAM srgb_consts0 = ");
6376 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const0
);
6377 shader_addline(&buffer
, ";\n");
6378 shader_addline(&buffer
, "PARAM srgb_consts1 = ");
6379 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const1
);
6380 shader_addline(&buffer
, ";\n");
6383 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6384 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6386 if (tempreg_used
|| settings
->sRGB_write
)
6387 shader_addline(&buffer
, "MOV tempreg, 0.0;\n");
6389 /* Generate texture sampling instructions */
6390 for (stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6392 if (!(tex_read
& (1u << stage
)))
6395 textype
= arbfp_texture_target(settings
->op
[stage
].tex_type
);
6397 if(settings
->op
[stage
].projected
== proj_none
) {
6399 } else if(settings
->op
[stage
].projected
== proj_count4
||
6400 settings
->op
[stage
].projected
== proj_count3
) {
6403 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6408 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6409 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6411 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6412 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6413 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6414 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6416 /* with projective textures, texbem only divides the static texture coord, not the displacement,
6417 * so multiply the displacement with the dividing parameter before passing it to TXP
6419 if (settings
->op
[stage
].projected
!= proj_none
) {
6420 if(settings
->op
[stage
].projected
== proj_count4
) {
6421 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6422 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
6424 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6425 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
6428 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6431 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6432 instr
, stage
, stage
, textype
);
6433 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6435 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6436 stage
- 1, stage
- 1, stage
- 1);
6437 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6439 } else if(settings
->op
[stage
].projected
== proj_count3
) {
6440 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6441 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6442 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6443 instr
, stage
, stage
, textype
);
6445 shader_addline(&buffer
, "%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6446 instr
, stage
, stage
, stage
, textype
);
6449 sprintf(colorcor_dst
, "tex%u", stage
);
6450 masks
= calc_color_correction(settings
->op
[stage
].color_fixup
, WINED3DSP_WRITEMASK_ALL
);
6451 gen_color_correction(&buffer
, colorcor_dst
, colorcor_dst
, "const.x", "const.y",
6452 settings
->op
[stage
].color_fixup
, masks
);
6455 if (settings
->color_key_enabled
)
6457 shader_addline(&buffer
, "SLT TMP, tex0, color_key_low;\n"); /* below low key */
6458 shader_addline(&buffer
, "SGE ret, tex0, color_key_high;\n"); /* above high key */
6459 shader_addline(&buffer
, "ADD TMP, TMP, ret;\n"); /* or */
6460 shader_addline(&buffer
, "DP4 TMP.b, TMP, TMP;\n"); /* on any channel */
6461 shader_addline(&buffer
, "SGE TMP, -TMP.b, 0.0;\n"); /* logical not */
6462 shader_addline(&buffer
, "KIL -TMP;\n"); /* discard if true */
6465 shader_addline(&buffer
, "MOV ret, fragment.color.primary;\n");
6467 /* Generate the main shader */
6468 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6470 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6473 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6474 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6475 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6476 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6477 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6478 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6479 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6480 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6481 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6482 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6483 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6484 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6486 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6487 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6488 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6489 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6491 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6493 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6494 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6495 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6499 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
6500 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6501 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6503 else if (settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP
6504 && settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6506 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6507 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6508 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6509 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
6510 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6511 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6515 if (settings
->sRGB_write
|| custom_linear_fog
)
6517 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, ret;\n");
6518 if (settings
->sRGB_write
)
6519 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6520 if (custom_linear_fog
)
6521 arbfp_add_linear_fog(&buffer
, "ret", "arg0");
6522 shader_addline(&buffer
, "MOV result.color, ret;\n");
6526 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, ret;\n");
6530 shader_addline(&buffer
, "END\n");
6532 /* Generate the shader */
6533 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6534 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6535 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
6537 string_buffer_free(&buffer
);
6541 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6543 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6544 const struct wined3d_device
*device
= context
->device
;
6545 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6546 BOOL use_pshader
= use_ps(state
);
6547 struct ffp_frag_settings settings
;
6548 const struct arbfp_ffp_desc
*desc
;
6551 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6553 if (isStateDirty(context
, STATE_RENDER(WINED3D_RS_FOGENABLE
)))
6555 if (!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6557 /* Reload fixed function constants since they collide with the
6558 * pixel shader constants. */
6559 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6561 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6562 state_tss_constant_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_CONSTANT
));
6564 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6565 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6566 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6568 else if (use_pshader
)
6570 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6577 /* Find or create a shader implementing the fixed function pipeline
6578 * settings, then activate it. */
6579 gen_ffp_frag_op(context
, state
, &settings
, FALSE
);
6580 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6583 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
6586 ERR("Out of memory\n");
6590 new_desc
->parent
.settings
= settings
;
6591 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6592 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6593 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6597 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6598 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6601 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6602 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6603 priv
->current_fprogram_id
= desc
->shader
;
6605 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6607 /* Reload fixed function constants since they collide with the
6608 * pixel shader constants. */
6609 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6611 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6612 state_tss_constant_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_CONSTANT
));
6614 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6615 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6616 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6618 context
->last_was_pshader
= FALSE
;
6620 else if (!context
->last_was_pshader
)
6622 if (device
->shader_backend
== &arb_program_shader_backend
)
6623 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
6624 context
->last_was_pshader
= TRUE
;
6627 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6630 /* We can't link the fog states to the fragment state directly since the
6631 * vertex pipeline links them to FOGENABLE. A different linking in different
6632 * pipeline parts can't be expressed in the combined state table, so we need
6633 * to handle that with a forwarding function. The other invisible side effect
6634 * is that changing the fog start and fog end (which links to FOGENABLE in
6635 * vertex) results in the fragment_prog_arbfp function being called because
6636 * FOGENABLE is dirty, which calls this function here. */
6637 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6639 enum fogsource new_source
;
6640 DWORD fogstart
= state
->render_states
[WINED3D_RS_FOGSTART
];
6641 DWORD fogend
= state
->render_states
[WINED3D_RS_FOGEND
];
6643 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6645 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6646 fragment_prog_arbfp(context
, state
, state_id
);
6648 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6651 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6655 new_source
= FOGSOURCE_VS
;
6659 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6660 new_source
= FOGSOURCE_COORD
;
6662 new_source
= FOGSOURCE_FFP
;
6667 new_source
= FOGSOURCE_FFP
;
6670 if (new_source
!= context
->fog_source
|| fogstart
== fogend
)
6672 context
->fog_source
= new_source
;
6673 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6677 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6679 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6680 fragment_prog_arbfp(context
, state
, state_id
);
6683 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] =
6685 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6686 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6687 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6688 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6689 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6690 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6691 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6692 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6693 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6694 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6695 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6696 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6697 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6698 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6699 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6700 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6701 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6702 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6703 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6704 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6705 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6706 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6707 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6708 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6709 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6710 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6711 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6712 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6713 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6714 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6715 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6716 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6717 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6718 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6719 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6720 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6721 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6722 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6723 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6724 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6725 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6726 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6727 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6728 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6729 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6730 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6731 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6732 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6733 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6734 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6735 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6736 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6737 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6738 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6739 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6740 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6741 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6742 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6743 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6744 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6745 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6746 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6747 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6748 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6749 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6750 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6751 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6752 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6753 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6754 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6755 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6756 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6757 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6758 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6759 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6760 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6761 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6762 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6763 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6764 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6765 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6766 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6767 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6768 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6769 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6770 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6771 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6772 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6773 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6774 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6775 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6776 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6777 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6778 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6779 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6780 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6781 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6782 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6783 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6784 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6785 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6786 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6787 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6788 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6789 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6790 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6791 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6792 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6793 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6794 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6795 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6796 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6797 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6798 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6799 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6800 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6801 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6802 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6803 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6804 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6805 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6806 {STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6807 {STATE_RENDER(WINED3D_RS_ALPHAFUNC
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6808 {STATE_RENDER(WINED3D_RS_ALPHAREF
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6809 {STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), alpha_test_arbfp
}, WINED3D_GL_EXT_NONE
},
6810 {STATE_RENDER(WINED3D_RS_COLORKEYENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6811 {STATE_COLOR_KEY
, { STATE_COLOR_KEY
, color_key_arbfp
}, WINED3D_GL_EXT_NONE
},
6812 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6813 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6814 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6815 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6816 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6817 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6818 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6819 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6820 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6821 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6822 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6823 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6824 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6825 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6826 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6827 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6828 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6829 {STATE_TEXTURESTAGE(0, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6830 {STATE_TEXTURESTAGE(1, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6831 {STATE_TEXTURESTAGE(2, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6832 {STATE_TEXTURESTAGE(3, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6833 {STATE_TEXTURESTAGE(4, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6834 {STATE_TEXTURESTAGE(5, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6835 {STATE_TEXTURESTAGE(6, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6836 {STATE_TEXTURESTAGE(7, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6837 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6838 {STATE_RENDER(WINED3D_RS_SHADEMODE
), { STATE_RENDER(WINED3D_RS_SHADEMODE
), state_shademode
}, WINED3D_GL_EXT_NONE
},
6839 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6842 static BOOL
arbfp_alloc_context_data(struct wined3d_context
*context
)
6847 static void arbfp_free_context_data(struct wined3d_context
*context
)
6851 const struct fragment_pipeline arbfp_fragment_pipeline
= {
6854 arbfp_get_emul_mask
,
6857 arbfp_alloc_context_data
,
6858 arbfp_free_context_data
,
6859 shader_arb_color_fixup_supported
,
6860 arbfp_fragmentstate_template
,
6863 struct arbfp_blit_type
6865 enum complex_fixup fixup
: 4;
6866 enum wined3d_gl_resource_type res_type
: 3;
6867 DWORD use_color_key
: 1;
6871 struct arbfp_blit_desc
6874 struct arbfp_blit_type type
;
6875 struct wine_rb_entry entry
;
6878 #define ARBFP_BLIT_PARAM_SIZE 0
6879 #define ARBFP_BLIT_PARAM_COLOR_KEY_LOW 1
6880 #define ARBFP_BLIT_PARAM_COLOR_KEY_HIGH 2
6882 struct arbfp_blit_priv
6884 struct wine_rb_tree shaders
;
6885 GLuint palette_texture
;
6888 static int arbfp_blit_type_compare(const void *key
, const struct wine_rb_entry
*entry
)
6890 const struct arbfp_blit_type
*ka
= key
;
6891 const struct arbfp_blit_type
*kb
= &WINE_RB_ENTRY_VALUE(entry
, const struct arbfp_blit_desc
, entry
)->type
;
6893 return memcmp(ka
, kb
, sizeof(*ka
));
6896 /* Context activation is done by the caller. */
6897 static void arbfp_free_blit_shader(struct wine_rb_entry
*entry
, void *context
)
6899 const struct wined3d_gl_info
*gl_info
= context
;
6900 struct arbfp_blit_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
6902 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
6903 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
6904 HeapFree(GetProcessHeap(), 0, entry_arb
);
6907 static HRESULT
arbfp_blit_alloc(struct wined3d_device
*device
)
6909 struct arbfp_blit_priv
*priv
;
6911 if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
6912 return E_OUTOFMEMORY
;
6914 wine_rb_init(&priv
->shaders
, arbfp_blit_type_compare
);
6916 device
->blit_priv
= priv
;
6921 /* Context activation is done by the caller. */
6922 static void arbfp_blit_free(struct wined3d_device
*device
)
6924 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6925 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6927 wine_rb_destroy(&priv
->shaders
, arbfp_free_blit_shader
, &device
->adapter
->gl_info
);
6928 checkGLcall("Delete blit programs");
6930 if (priv
->palette_texture
)
6931 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &priv
->palette_texture
);
6933 HeapFree(GetProcessHeap(), 0, device
->blit_priv
);
6934 device
->blit_priv
= NULL
;
6937 static BOOL
gen_planar_yuv_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
6941 const char *tex
, *texinstr
= "TXP";
6943 if (type
->fixup
== COMPLEX_FIXUP_UYVY
)
6954 tex
= arbfp_texture_target(type
->res_type
);
6955 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_RECT
)
6958 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6959 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6960 * filtering when we sample the texture.
6962 * These are the rules for reading the chroma:
6968 * So we have to get the sampling x position in non-normalized coordinates in integers
6970 if (type
->res_type
!= WINED3D_GL_RES_TYPE_TEX_RECT
)
6972 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6973 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6977 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6979 /* We must not allow filtering between pixel x and x+1, this would mix U and V
6980 * Vertical filtering is ok. However, bear in mind that the pixel center is at
6983 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
6984 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
6986 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
6987 * even and odd pixels respectively
6989 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
6990 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
6992 /* Sample Pixel 1 */
6993 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6995 /* Put the value into either of the chroma values */
6996 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6997 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
6998 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6999 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
7001 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
7002 * the pixel right to the current one. Otherwise, sample the left pixel.
7003 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
7005 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
7006 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
7007 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7009 /* Put the value into the other chroma */
7010 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7011 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
7012 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7013 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
7015 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
7016 * the current one and lerp the two U and V values
7019 /* This gives the correctly filtered luminance value */
7020 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
7025 static BOOL
gen_yv12_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7029 static const float yv12_coef
[]
7030 = {2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
};
7032 tex
= arbfp_texture_target(type
->res_type
);
7034 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7035 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
7036 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
7037 * pitch of the luminance plane, the packing into the gl texture is a bit
7038 * unfortunate. If the whole texture is interpreted as luminance data it looks
7039 * approximately like this:
7041 * +----------------------------------+----
7053 * +----------------+-----------------+----
7055 * | V even rows | V odd rows |
7057 * +----------------+------------------ -
7059 * | U even rows | U odd rows |
7061 * +----------------+-----------------+----
7065 * So it appears as if there are 4 chroma images, but in fact the odd rows
7066 * in the chroma images are in the same row as the even ones. So it is
7067 * kinda tricky to read
7069 * When reading from rectangle textures, keep in mind that the input y coordinates
7070 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
7072 shader_addline(buffer
, "PARAM yv12_coef = ");
7073 shader_arb_append_imm_vec4(buffer
, yv12_coef
);
7074 shader_addline(buffer
, ";\n");
7076 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7077 /* the chroma planes have only half the width */
7078 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7080 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
7081 * the coordinate. Also read the right side of the image when reading odd lines
7083 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
7086 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7089 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
7091 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
7093 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
7094 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
7096 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7097 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7098 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7099 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7100 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7102 /* clamp, keep the half pixel origin in mind */
7103 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
7104 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7105 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
7106 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7110 /* Read from [size - size+size/4] */
7111 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7112 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
7114 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7115 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7116 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7117 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7118 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
7119 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7121 /* Make sure to read exactly from the pixel center */
7122 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7123 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
7126 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
7127 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7128 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7129 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7130 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7132 /* Read the texture, put the result into the output register */
7133 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7134 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7136 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
7137 * No need to clamp because we're just reusing the already clamped value from above
7139 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7140 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
7142 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
7143 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7144 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7146 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7147 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7148 * values due to filtering
7150 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7151 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7153 /* Multiply the y coordinate by 2/3 and clamp it */
7154 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
7155 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
7156 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7157 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7161 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7162 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7165 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7166 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7167 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7174 static BOOL
gen_nv12_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7178 static const float nv12_coef
[]
7179 = {2.0f
/ 3.0f
, 1.0f
/ 3.0f
, 1.0f
, 1.0f
};
7181 tex
= arbfp_texture_target(type
->res_type
);
7183 /* NV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7184 * sized plane where each component is an UV pair. So the effective
7185 * bitdepth is 12 bits per pixel If the whole texture is interpreted as luminance
7186 * data it looks approximately like this:
7188 * +----------------------------------+----
7200 * +----------------------------------+----
7201 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7202 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7208 * +----------------------------------+----
7210 * When reading from rectangle textures, keep in mind that the input y coordinates
7211 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height. */
7213 shader_addline(buffer
, "PARAM nv12_coef = ");
7214 shader_arb_append_imm_vec4(buffer
, nv12_coef
);
7215 shader_addline(buffer
, ";\n");
7217 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7218 /* We only have half the number of chroma pixels. */
7219 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7221 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7223 shader_addline(buffer
, "RCP chroma.w, size.x;\n");
7224 shader_addline(buffer
, "RCP chroma.z, size.y;\n");
7226 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, nv12_coef.y, nv12_coef.x;\n");
7228 /* We must not allow filtering horizontally, this would mix U and V.
7229 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7230 * 0.5, so add 0.5. */
7232 /* Convert to non-normalized coordinates so we can find the
7233 * individual pixel. */
7234 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, size.x;\n");
7235 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7236 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7237 * add 0.5 to hit the center of the pixel. */
7238 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7240 /* Convert back to normalized coordinates. */
7241 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, chroma.w;\n");
7243 /* Clamp, keep the half pixel origin in mind. */
7244 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.z, nv12_coef.x;\n");
7245 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7246 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.z, nv12_coef.z;\n");
7247 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7251 /* Read from [size - size+size/2] */
7252 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.y, size.y;\n");
7254 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7255 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7256 * add 0.5 to hit the center of the pixel. */
7257 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7260 shader_addline(buffer
, "MAD temp.y, size.y, coef.y, size.y;\n");
7261 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7262 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7263 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7264 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7266 /* Read the texture, put the result into the output register. */
7267 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7268 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7270 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7273 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, chroma.w;\n");
7278 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.x;\n");
7280 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7281 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7283 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7284 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7285 * values due to filtering. */
7286 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7287 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7289 /* Multiply the y coordinate by 2/3 and clamp it */
7290 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, nv12_coef.x;\n");
7291 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, nv12_coef.x;\n");
7292 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7293 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7297 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7298 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7301 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7302 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7303 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7310 /* Context activation is done by the caller. */
7311 static GLuint
gen_p8_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7314 struct wined3d_string_buffer buffer
;
7315 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7317 /* This should not happen because we only use this conversion for
7318 * present blits which don't use color keying. */
7319 if (type
->use_color_key
)
7320 FIXME("Implement P8 color keying.\n");
7323 if (!string_buffer_init(&buffer
))
7325 ERR("Failed to initialize shader buffer.\n");
7329 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7330 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7333 string_buffer_free(&buffer
);
7337 shader_addline(&buffer
, "!!ARBfp1.0\n");
7338 shader_addline(&buffer
, "TEMP index;\n");
7340 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7341 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7343 /* The alpha-component contains the palette index */
7344 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7346 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7347 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7349 /* Use the alpha-component as an index in the palette to get the final color */
7350 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7351 shader_addline(&buffer
, "END\n");
7353 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7355 string_buffer_free(&buffer
);
7360 /* Context activation is done by the caller. */
7361 static void upload_palette(const struct wined3d_texture
*texture
, struct wined3d_context
*context
)
7363 const struct wined3d_palette
*palette
= texture
->swapchain
? texture
->swapchain
->palette
: NULL
;
7364 struct wined3d_device
*device
= texture
->resource
.device
;
7365 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7366 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
7368 if (!priv
->palette_texture
)
7369 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &priv
->palette_texture
);
7371 GL_EXTCALL(glActiveTexture(GL_TEXTURE1
));
7372 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, priv
->palette_texture
);
7374 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7376 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7377 /* Make sure we have discrete color levels. */
7378 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7379 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7380 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7383 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 256, 0, GL_BGRA
,
7384 GL_UNSIGNED_INT_8_8_8_8_REV
, palette
->colors
);
7388 static const DWORD black
;
7389 FIXME("P8 surface loaded without a palette.\n");
7390 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 1, 0, GL_BGRA
,
7391 GL_UNSIGNED_INT_8_8_8_8_REV
, &black
);
7394 /* Switch back to unit 0 in which the 2D texture will be stored. */
7395 context_active_texture(context
, gl_info
, 0);
7398 /* Context activation is done by the caller. */
7399 static GLuint
gen_yuv_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7402 struct wined3d_string_buffer buffer
;
7403 char luminance_component
;
7405 if (type
->use_color_key
)
7406 FIXME("Implement YUV color keying.\n");
7409 if (!string_buffer_init(&buffer
))
7411 ERR("Failed to initialize shader buffer.\n");
7415 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7416 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7417 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7418 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7421 string_buffer_free(&buffer
);
7425 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7426 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7427 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7428 * each single pixel it contains, and one U and one V value shared between both
7431 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7432 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7433 * take the format into account when generating the read swizzles
7435 * Reading the Y value is straightforward - just sample the texture. The hardware
7436 * takes care of filtering in the horizontal and vertical direction.
7438 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7439 * because that would mix the U and V values of one pixel or two adjacent pixels.
7440 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7441 * regardless of the filtering setting. Vertical filtering works automatically
7442 * though - the U and V values of two rows are mixed nicely.
7444 * Apart of avoiding filtering issues, the code has to know which value it just
7445 * read, and where it can find the other one. To determine this, it checks if
7446 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7448 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7449 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7451 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7452 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7453 * in an unfiltered situation. Finding the luminance on the other hand requires
7454 * finding out if it is an odd or even pixel. The real drawback of this approach
7455 * is filtering. This would have to be emulated completely in the shader, reading
7456 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7457 * vertically. Beyond that it would require adjustments to the texture handling
7458 * code to deal with the width scaling
7460 shader_addline(&buffer
, "!!ARBfp1.0\n");
7461 shader_addline(&buffer
, "TEMP luminance;\n");
7462 shader_addline(&buffer
, "TEMP temp;\n");
7463 shader_addline(&buffer
, "TEMP chroma;\n");
7464 shader_addline(&buffer
, "TEMP texcrd;\n");
7465 shader_addline(&buffer
, "TEMP texcrd2;\n");
7466 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7467 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7468 shader_addline(&buffer
, "PARAM size = program.local[%u];\n", ARBFP_BLIT_PARAM_SIZE
);
7470 switch (type
->fixup
)
7472 case COMPLEX_FIXUP_UYVY
:
7473 case COMPLEX_FIXUP_YUY2
:
7474 if (!gen_planar_yuv_read(&buffer
, type
, &luminance_component
))
7476 string_buffer_free(&buffer
);
7481 case COMPLEX_FIXUP_YV12
:
7482 if (!gen_yv12_read(&buffer
, type
, &luminance_component
))
7484 string_buffer_free(&buffer
);
7489 case COMPLEX_FIXUP_NV12
:
7490 if (!gen_nv12_read(&buffer
, type
, &luminance_component
))
7492 string_buffer_free(&buffer
);
7498 FIXME("Unsupported YUV fixup %#x\n", type
->fixup
);
7499 string_buffer_free(&buffer
);
7503 /* Calculate the final result. Formula is taken from
7504 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7505 * ranges from -0.5 to 0.5
7507 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7509 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7510 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7511 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7512 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7513 shader_addline(&buffer
, "END\n");
7515 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7517 string_buffer_free(&buffer
);
7522 /* Context activation is done by the caller. */
7523 static GLuint
arbfp_gen_plain_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7526 struct wined3d_string_buffer buffer
;
7527 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7530 if (!string_buffer_init(&buffer
))
7532 ERR("Failed to initialize shader buffer.\n");
7536 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7539 string_buffer_free(&buffer
);
7542 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7544 shader_addline(&buffer
, "!!ARBfp1.0\n");
7546 if (type
->use_color_key
)
7548 shader_addline(&buffer
, "TEMP color;\n");
7549 shader_addline(&buffer
, "TEMP less, greater;\n");
7550 shader_addline(&buffer
, "PARAM color_key_low = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_LOW
);
7551 shader_addline(&buffer
, "PARAM color_key_high = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
);
7552 shader_addline(&buffer
, "TEX color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7553 shader_addline(&buffer
, "SLT less, color, color_key_low;\n"); /* below low key */
7554 shader_addline(&buffer
, "SGE greater, color, color_key_high;\n"); /* above high key */
7555 shader_addline(&buffer
, "ADD less, less, greater;\n"); /* or */
7556 shader_addline(&buffer
, "DP4 less.b, less, less;\n"); /* on any channel */
7557 shader_addline(&buffer
, "SGE less, -less.b, 0.0;\n"); /* logical not */
7558 shader_addline(&buffer
, "KIL -less;\n"); /* discard if true */
7559 shader_addline(&buffer
, "MOV result.color, color;\n");
7563 shader_addline(&buffer
, "TEX result.color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7566 shader_addline(&buffer
, "END\n");
7568 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7570 string_buffer_free(&buffer
);
7575 /* Context activation is done by the caller. */
7576 static HRESULT
arbfp_blit_set(void *blit_priv
, struct wined3d_context
*context
, const struct wined3d_surface
*surface
,
7577 const struct wined3d_color_key
*color_key
)
7579 const struct wined3d_texture
*texture
= surface
->container
;
7580 struct arbfp_blit_priv
*priv
= blit_priv
;
7581 enum complex_fixup fixup
;
7582 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7583 struct wine_rb_entry
*entry
;
7584 struct arbfp_blit_type type
;
7585 struct arbfp_blit_desc
*desc
;
7586 struct wined3d_color float_color_key
[2];
7587 struct wined3d_vec4 size
;
7590 size
.x
= wined3d_texture_get_level_pow2_width(texture
, surface
->texture_level
);
7591 size
.y
= wined3d_texture_get_level_pow2_height(texture
, surface
->texture_level
);
7595 if (is_complex_fixup(texture
->resource
.format
->color_fixup
))
7596 fixup
= get_complex_fixup(texture
->resource
.format
->color_fixup
);
7598 fixup
= COMPLEX_FIXUP_NONE
;
7600 switch (texture
->target
)
7603 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_1D
;
7607 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7611 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_3D
;
7614 case GL_TEXTURE_CUBE_MAP_ARB
:
7615 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_CUBE
;
7618 case GL_TEXTURE_RECTANGLE_ARB
:
7619 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_RECT
;
7623 ERR("Unexpected GL texture type %#x.\n", texture
->target
);
7624 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7627 type
.use_color_key
= !!color_key
;
7630 entry
= wine_rb_get(&priv
->shaders
, &type
);
7633 desc
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7634 shader
= desc
->shader
;
7640 case COMPLEX_FIXUP_NONE
:
7641 if (!is_identity_fixup(texture
->resource
.format
->color_fixup
))
7642 FIXME("Implement support for sign or swizzle fixups.\n");
7643 shader
= arbfp_gen_plain_shader(gl_info
, &type
);
7646 case COMPLEX_FIXUP_P8
:
7647 shader
= gen_p8_shader(gl_info
, &type
);
7650 case COMPLEX_FIXUP_YUY2
:
7651 case COMPLEX_FIXUP_UYVY
:
7652 case COMPLEX_FIXUP_YV12
:
7653 case COMPLEX_FIXUP_NV12
:
7654 shader
= gen_yuv_shader(gl_info
, &type
);
7660 FIXME("Unsupported complex fixup %#x, not setting a shader\n", fixup
);
7664 desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*desc
));
7669 desc
->shader
= shader
;
7670 if (wine_rb_put(&priv
->shaders
, &desc
->type
, &desc
->entry
) == -1)
7673 ERR("Out of memory\n");
7674 GL_EXTCALL(glDeleteProgramsARB(1, &shader
));
7675 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader))");
7676 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, 0));
7677 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, 0)");
7678 HeapFree(GetProcessHeap(), 0, desc
);
7679 return E_OUTOFMEMORY
;
7683 if (fixup
== COMPLEX_FIXUP_P8
)
7684 upload_palette(texture
, context
);
7686 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7687 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7688 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7689 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7690 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARBFP_BLIT_PARAM_SIZE
, &size
.x
));
7691 checkGLcall("glProgramLocalParameter4fvARB");
7692 if (type
.use_color_key
)
7694 wined3d_format_get_float_color_key(texture
->resource
.format
, color_key
, float_color_key
);
7695 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7696 ARBFP_BLIT_PARAM_COLOR_KEY_LOW
, &float_color_key
[0].r
));
7697 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7698 ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
, &float_color_key
[1].r
));
7699 checkGLcall("glProgramLocalParameter4fvARB");
7705 /* Context activation is done by the caller. */
7706 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7708 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7709 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7712 static BOOL
arbfp_blit_supported(const struct wined3d_gl_info
*gl_info
,
7713 const struct wined3d_d3d_info
*d3d_info
, enum wined3d_blit_op blit_op
,
7714 const RECT
*src_rect
, DWORD src_usage
, enum wined3d_pool src_pool
, const struct wined3d_format
*src_format
,
7715 const RECT
*dst_rect
, DWORD dst_usage
, enum wined3d_pool dst_pool
, const struct wined3d_format
*dst_format
)
7717 enum complex_fixup src_fixup
;
7720 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7725 case WINED3D_BLIT_OP_COLOR_BLIT_CKEY
:
7726 if (!d3d_info
->shader_color_key
)
7728 /* The conversion modifies the alpha channel so the color key might no longer match. */
7729 TRACE("Color keying not supported with converted textures.\n");
7732 case WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
:
7733 case WINED3D_BLIT_OP_COLOR_BLIT
:
7737 TRACE("Unsupported blit_op=%d\n", blit_op
);
7741 decompress
= src_format
&& (src_format
->flags
[WINED3D_GL_RES_TYPE_TEX_2D
] & WINED3DFMT_FLAG_COMPRESSED
)
7742 && !(dst_format
->flags
[WINED3D_GL_RES_TYPE_TEX_2D
] & WINED3DFMT_FLAG_COMPRESSED
);
7743 if (!decompress
&& (dst_pool
== WINED3D_POOL_SYSTEM_MEM
|| src_pool
== WINED3D_POOL_SYSTEM_MEM
))
7746 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7747 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7749 TRACE("Checking support for fixup:\n");
7750 dump_color_fixup_desc(src_format
->color_fixup
);
7753 if (!is_identity_fixup(dst_format
->color_fixup
))
7755 TRACE("Destination fixups are not supported\n");
7759 if (is_identity_fixup(src_format
->color_fixup
))
7765 /* We only support YUV conversions. */
7766 if (!is_complex_fixup(src_format
->color_fixup
))
7768 TRACE("[FAILED]\n");
7774 case COMPLEX_FIXUP_YUY2
:
7775 case COMPLEX_FIXUP_UYVY
:
7776 case COMPLEX_FIXUP_YV12
:
7777 case COMPLEX_FIXUP_NV12
:
7778 case COMPLEX_FIXUP_P8
:
7783 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7784 TRACE("[FAILED]\n");
7789 static void arbfp_blit_surface(struct wined3d_device
*device
, enum wined3d_blit_op op
, struct wined3d_context
*context
,
7790 struct wined3d_surface
*src_surface
, const RECT
*src_rect
,
7791 struct wined3d_surface
*dst_surface
, DWORD dst_location
, const RECT
*dst_rect
,
7792 const struct wined3d_color_key
*color_key
, enum wined3d_texture_filter_type filter
)
7794 struct wined3d_texture
*src_texture
= src_surface
->container
;
7795 struct wined3d_texture
*dst_texture
= dst_surface
->container
;
7796 struct wined3d_color_key alpha_test_key
;
7799 /* Now load the surface */
7800 if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7801 && (surface_get_sub_resource(src_surface
)->locations
7802 & (WINED3D_LOCATION_TEXTURE_RGB
| WINED3D_LOCATION_DRAWABLE
))
7803 == WINED3D_LOCATION_DRAWABLE
7804 && !wined3d_resource_is_offscreen(&src_texture
->resource
))
7806 /* Without FBO blits transferring from the drawable to the texture is
7807 * expensive, because we have to flip the data in sysmem. Since we can
7808 * flip in the blitter, we don't actually need that flip anyway. So we
7809 * use the surface's texture as scratch texture, and flip the source
7810 * rectangle instead. */
7811 surface_load_fb_texture(src_surface
, FALSE
, context
);
7814 s
.top
= wined3d_texture_get_level_height(src_texture
, src_surface
->texture_level
) - s
.top
;
7815 s
.bottom
= wined3d_texture_get_level_height(src_texture
, src_surface
->texture_level
) - s
.bottom
;
7819 wined3d_texture_load(src_texture
, context
, FALSE
);
7821 context_apply_blit_state(context
, device
);
7823 if (dst_location
== WINED3D_LOCATION_DRAWABLE
)
7826 surface_translate_drawable_coords(dst_surface
, context
->win_handle
, &d
);
7830 if (wined3d_settings
.offscreen_rendering_mode
== ORM_FBO
)
7834 if (dst_location
== WINED3D_LOCATION_DRAWABLE
)
7836 TRACE("Destination surface %p is onscreen.\n", dst_surface
);
7837 buffer
= wined3d_texture_get_gl_buffer(dst_texture
);
7841 TRACE("Destination surface %p is offscreen.\n", dst_surface
);
7842 buffer
= GL_COLOR_ATTACHMENT0
;
7844 context_apply_fbo_state_blit(context
, GL_DRAW_FRAMEBUFFER
, dst_surface
, NULL
, dst_location
);
7845 context_set_draw_buffer(context
, buffer
);
7846 context_check_fbo_status(context
, GL_DRAW_FRAMEBUFFER
);
7847 context_invalidate_state(context
, STATE_FRAMEBUFFER
);
7850 if (op
== WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
)
7852 const struct wined3d_format
*fmt
= src_texture
->resource
.format
;
7853 alpha_test_key
.color_space_low_value
= 0;
7854 alpha_test_key
.color_space_high_value
= ~(((1u << fmt
->alpha_size
) - 1) << fmt
->alpha_offset
);
7855 color_key
= &alpha_test_key
;
7858 arbfp_blit_set(device
->blit_priv
, context
, src_surface
, color_key
);
7860 /* Draw a textured quad */
7861 draw_textured_quad(src_surface
, context
, src_rect
, dst_rect
, filter
);
7863 /* Leave the opengl state valid for blitting */
7864 arbfp_blit_unset(context
->gl_info
);
7866 if (wined3d_settings
.strict_draw_ordering
7867 || (dst_texture
->swapchain
&& (dst_texture
->swapchain
->front_buffer
== dst_texture
)))
7868 context
->gl_info
->gl_ops
.gl
.p_glFlush(); /* Flush to ensure ordering across contexts. */
7871 static HRESULT
arbfp_blit_color_fill(struct wined3d_device
*device
, struct wined3d_rendertarget_view
*view
,
7872 const RECT
*rect
, const struct wined3d_color
*color
)
7874 FIXME("Color filling not implemented by arbfp_blit\n");
7875 return WINED3DERR_INVALIDCALL
;
7878 static HRESULT
arbfp_blit_depth_fill(struct wined3d_device
*device
, struct wined3d_rendertarget_view
*view
,
7879 const RECT
*rect
, DWORD clear_flags
, float depth
, DWORD stencil
)
7881 FIXME("Depth/stencil filling not implemented by arbfp_blit.\n");
7882 return WINED3DERR_INVALIDCALL
;
7885 const struct wined3d_blitter_ops arbfp_blit
=
7889 arbfp_blit_supported
,
7890 arbfp_blit_color_fill
,
7891 arbfp_blit_depth_fill
,