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
[MAX_CONST_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
[MAX_CONST_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
[MAX_CONST_I
][3];
212 struct arb_vs_compiled_shader
214 struct arb_vs_compile_args args
;
216 UINT int_consts
[MAX_CONST_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 GLuint depth_blt_vprogram_id
;
295 GLuint depth_blt_fprogram_id_full
[WINED3D_GL_RES_TYPE_COUNT
];
296 GLuint depth_blt_fprogram_id_masked
[WINED3D_GL_RES_TYPE_COUNT
];
297 BOOL use_arbfp_fixed_func
;
298 struct wine_rb_tree fragment_shaders
;
299 BOOL last_ps_const_clamped
;
300 BOOL last_vs_color_unclamp
;
302 struct wine_rb_tree signature_tree
;
305 unsigned int highest_dirty_ps_const
, highest_dirty_vs_const
;
306 char *vshader_const_dirty
, *pshader_const_dirty
;
307 const struct wined3d_context
*last_context
;
309 const struct wined3d_vertex_pipe_ops
*vertex_pipe
;
310 const struct fragment_pipeline
*fragment_pipe
;
311 BOOL ffp_proj_control
;
314 /* Context activation for state handlers is done by the caller. */
316 static BOOL
need_rel_addr_const(const struct arb_vshader_private
*shader_data
,
317 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
319 if (shader_data
->rel_offset
) return TRUE
;
320 if (!reg_maps
->usesmova
) return FALSE
;
321 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
324 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
325 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
327 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
328 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
331 static BOOL
need_helper_const(const struct arb_vshader_private
*shader_data
,
332 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
334 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) return TRUE
;
335 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
336 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
337 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) return TRUE
; /* Have to init texcoords. */
338 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
339 if (reg_maps
->usesnrm
) return TRUE
; /* 0.0 */
340 if (reg_maps
->usespow
) return TRUE
; /* EPS, 0.0 and 1.0 */
341 if (reg_maps
->fog
) return TRUE
; /* Clamping fog coord, 0.0 and 1.0 */
345 static unsigned int reserved_vs_const(const struct arb_vshader_private
*shader_data
,
346 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
348 unsigned int ret
= 1;
349 /* We use one PARAM for the pos fixup, and in some cases one to load
350 * some immediate values into the shader. */
351 if (need_helper_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
352 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
356 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
357 * When constant_list == NULL, it will load all the constants.
359 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
360 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
362 /* Context activation is done by the caller. */
363 static unsigned int shader_arb_load_constants_f(const struct wined3d_shader
*shader
,
364 const struct wined3d_gl_info
*gl_info
, GLuint target_type
, unsigned int max_constants
,
365 const struct wined3d_vec4
*constants
, char *dirty_consts
)
367 struct wined3d_shader_lconst
*lconst
;
368 unsigned int ret
, i
, j
;
370 if (TRACE_ON(d3d_constants
))
372 for (i
= 0; i
< max_constants
; ++i
)
374 if (!dirty_consts
[i
])
376 TRACE_(d3d_constants
)("Loading constant %u: %s.\n", i
, debug_vec4(&constants
[i
]));
382 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
383 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& shader
->reg_maps
.shader_version
.major
== 1)
386 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
387 * shaders, the first 8 constants are marked dirty for reload
389 for (; i
< min(8, max_constants
); ++i
)
391 if (!dirty_consts
[i
])
395 if (constants
[i
].x
> 1.0f
)
397 else if (constants
[i
].x
< -1.0f
)
398 lcl_const
[0] = -1.0f
;
400 lcl_const
[0] = constants
[i
].x
;
402 if (constants
[i
].y
> 1.0f
)
404 else if (constants
[i
].y
< -1.0f
)
405 lcl_const
[1] = -1.0f
;
407 lcl_const
[1] = constants
[i
].y
;
409 if (constants
[i
].z
> 1.0f
)
411 else if (constants
[i
].z
< -1.0f
)
412 lcl_const
[2] = -1.0f
;
414 lcl_const
[2] = constants
[i
].z
;
416 if (constants
[i
].w
> 1.0f
)
418 else if (constants
[i
].w
< -1.0f
)
419 lcl_const
[3] = -1.0f
;
421 lcl_const
[3] = constants
[i
].w
;
423 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
426 /* If further constants are dirty, reload them without clamping.
428 * The alternative is not to touch them, but then we cannot reset the dirty constant count
429 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
430 * above would always re-check the first 8 constants since max_constant remains at the init
435 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
437 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
438 * or just reloading *all* constants at once
440 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
442 for (; i
< max_constants
; ++i
)
444 if (!dirty_consts
[i
])
447 /* Find the next block of dirty constants */
450 for (++i
; (i
< max_constants
) && dirty_consts
[i
]; ++i
)
455 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, &constants
[j
].x
));
460 for (; i
< max_constants
; ++i
)
465 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, &constants
[i
].x
));
469 checkGLcall("glProgramEnvParameter4fvARB()");
471 /* Load immediate constants */
472 if (shader
->load_local_constsF
)
474 if (TRACE_ON(d3d_shader
))
476 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
478 GLfloat
* values
= (GLfloat
*)lconst
->value
;
479 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
480 values
[0], values
[1], values
[2], values
[3]);
483 /* Immediate constants are clamped for 1.X shaders at loading times */
485 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
487 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
488 ret
= max(ret
, lconst
->idx
+ 1);
489 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
491 checkGLcall("glProgramEnvParameter4fvARB()");
492 return ret
; /* The loaded immediate constants need reloading for the next shader */
494 return 0; /* No constants are dirty now */
498 /* Loads the texture dimensions for NP2 fixup into the currently set
499 * ARB_[vertex/fragment]_programs. */
500 static void shader_arb_load_np2fixup_constants(const struct arb_ps_np2fixup_info
*fixup
,
501 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
503 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
504 WORD active
= fixup
->super
.active
;
510 for (i
= 0; active
; active
>>= 1, ++i
)
512 const struct wined3d_texture
*tex
= state
->textures
[i
];
513 unsigned char idx
= fixup
->super
.idx
[i
];
514 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
521 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
527 tex_dim
[2] = tex
->pow2_matrix
[0];
528 tex_dim
[3] = tex
->pow2_matrix
[5];
532 tex_dim
[0] = tex
->pow2_matrix
[0];
533 tex_dim
[1] = tex
->pow2_matrix
[5];
537 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
)
539 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
540 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
544 /* Context activation is done by the caller. */
545 static void shader_arb_ps_local_constants(const struct arb_ps_compiled_shader
*gl_shader
,
546 const struct wined3d_context
*context
, const struct wined3d_state
*state
, UINT rt_height
)
548 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
551 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
553 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
555 /* The state manager takes care that this function is always called if the bump env matrix changes */
556 const float *data
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_MAT00
];
557 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
558 gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
560 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
562 /* WINED3D_TSS_BUMPENVLSCALE and WINED3D_TSS_BUMPENVLOFFSET are next to each other.
563 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
564 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
565 * (they're WINED3D_TSS_TEXTURETRANSFORMFLAGS and WINED3D_TSS_ADDRESSW, so most likely 0 or NaN
567 const float *scale
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_LSCALE
];
568 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
569 gl_shader
->luminanceconst
[i
].const_num
, scale
));
572 checkGLcall("Load bumpmap consts");
574 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
576 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
577 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
582 val
[0] = context
->render_offscreen
? 0.0f
: (float) rt_height
;
583 val
[1] = context
->render_offscreen
? 1.0f
: -1.0f
;
586 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
587 checkGLcall("y correction loading");
590 if (!gl_shader
->num_int_consts
) return;
592 for(i
= 0; i
< MAX_CONST_I
; i
++)
594 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
597 val
[0] = (float)state
->ps_consts_i
[4 * i
];
598 val
[1] = (float)state
->ps_consts_i
[4 * i
+ 1];
599 val
[2] = (float)state
->ps_consts_i
[4 * i
+ 2];
602 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
605 checkGLcall("Load ps int consts");
608 /* Context activation is done by the caller. */
609 static void shader_arb_vs_local_constants(const struct arb_vs_compiled_shader
*gl_shader
,
610 const struct wined3d_context
*context
, const struct wined3d_state
*state
)
612 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
613 float position_fixup
[4];
616 /* Upload the position fixup */
617 shader_get_position_fixup(context
, state
, position_fixup
);
618 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, position_fixup
));
620 if (!gl_shader
->num_int_consts
) return;
622 for(i
= 0; i
< MAX_CONST_I
; i
++)
624 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
627 val
[0] = (float)state
->vs_consts_i
[4 * i
];
628 val
[1] = (float)state
->vs_consts_i
[4 * i
+ 1];
629 val
[2] = (float)state
->vs_consts_i
[4 * i
+ 2];
632 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
635 checkGLcall("Load vs int consts");
638 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
639 const struct wined3d_state
*state
);
642 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
644 * We only support float constants in ARB at the moment, so don't
645 * worry about the Integers or Booleans
647 /* Context activation is done by the caller (state handler). */
648 static void shader_arb_load_constants_internal(struct shader_arb_priv
*priv
,
649 struct wined3d_context
*context
, const struct wined3d_state
*state
,
650 BOOL usePixelShader
, BOOL useVertexShader
, BOOL from_shader_select
)
652 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
653 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
655 if (!from_shader_select
)
657 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
658 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
661 && (vshader
->reg_maps
.boolean_constants
662 || (!gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
663 && (vshader
->reg_maps
.integer_constants
& ~vshader
->reg_maps
.local_int_consts
))))
665 TRACE("bool/integer vertex shader constants potentially modified, forcing shader reselection.\n");
666 shader_arb_select(priv
, context
, state
);
669 && (pshader
->reg_maps
.boolean_constants
670 || (!gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
]
671 && (pshader
->reg_maps
.integer_constants
& ~pshader
->reg_maps
.local_int_consts
))))
673 TRACE("bool/integer pixel shader constants potentially modified, forcing shader reselection.\n");
674 shader_arb_select(priv
, context
, state
);
678 if (context
!= priv
->last_context
)
680 memset(priv
->vshader_const_dirty
, 1,
681 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
682 priv
->highest_dirty_vs_const
= d3d_info
->limits
.vs_uniform_count
;
684 memset(priv
->pshader_const_dirty
, 1,
685 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
686 priv
->highest_dirty_ps_const
= d3d_info
->limits
.ps_uniform_count
;
688 priv
->last_context
= context
;
693 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
694 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
696 /* Load DirectX 9 float constants for vertex shader */
697 priv
->highest_dirty_vs_const
= shader_arb_load_constants_f(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
698 priv
->highest_dirty_vs_const
, state
->vs_consts_f
, priv
->vshader_const_dirty
);
699 shader_arb_vs_local_constants(gl_shader
, context
, state
);
704 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
705 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
706 UINT rt_height
= state
->fb
->render_targets
[0]->height
;
708 /* Load DirectX 9 float constants for pixel shader */
709 priv
->highest_dirty_ps_const
= shader_arb_load_constants_f(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
710 priv
->highest_dirty_ps_const
, state
->ps_consts_f
, priv
->pshader_const_dirty
);
711 shader_arb_ps_local_constants(gl_shader
, context
, state
, rt_height
);
713 if (context
->constant_update_mask
& WINED3D_SHADER_CONST_PS_NP2_FIXUP
)
714 shader_arb_load_np2fixup_constants(&gl_shader
->np2fixup_info
, gl_info
, state
);
718 static void shader_arb_load_constants(void *shader_priv
, struct wined3d_context
*context
,
719 const struct wined3d_state
*state
)
721 BOOL vs
= use_vs(state
);
722 BOOL ps
= use_ps(state
);
724 shader_arb_load_constants_internal(shader_priv
, context
, state
, ps
, vs
, FALSE
);
727 static void shader_arb_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
729 struct wined3d_context
*context
= context_get_current();
730 struct shader_arb_priv
*priv
= device
->shader_priv
;
733 for (i
= 0; i
< device
->context_count
; ++i
)
735 device
->contexts
[i
]->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
738 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
739 * context. On a context switch the old context will be fully dirtified */
740 if (!context
|| context
->swapchain
->device
!= device
) return;
742 memset(priv
->vshader_const_dirty
+ start
, 1, sizeof(*priv
->vshader_const_dirty
) * count
);
743 priv
->highest_dirty_vs_const
= max(priv
->highest_dirty_vs_const
, start
+ count
);
746 static void shader_arb_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
748 struct wined3d_context
*context
= context_get_current();
749 struct shader_arb_priv
*priv
= device
->shader_priv
;
752 for (i
= 0; i
< device
->context_count
; ++i
)
754 device
->contexts
[i
]->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
757 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
758 * context. On a context switch the old context will be fully dirtified */
759 if (!context
|| context
->swapchain
->device
!= device
) return;
761 memset(priv
->pshader_const_dirty
+ start
, 1, sizeof(*priv
->pshader_const_dirty
) * count
);
762 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, start
+ count
);
765 static void shader_arb_append_imm_vec4(struct wined3d_string_buffer
*buffer
, const float *values
)
769 wined3d_ftoa(values
[0], str
[0]);
770 wined3d_ftoa(values
[1], str
[1]);
771 wined3d_ftoa(values
[2], str
[2]);
772 wined3d_ftoa(values
[3], str
[3]);
773 shader_addline(buffer
, "{%s, %s, %s, %s}", str
[0], str
[1], str
[2], str
[3]);
776 /* Generate the variable & register declarations for the ARB_vertex_program output target */
777 static void shader_generate_arb_declarations(const struct wined3d_shader
*shader
,
778 const struct wined3d_shader_reg_maps
*reg_maps
, struct wined3d_string_buffer
*buffer
,
779 const struct wined3d_gl_info
*gl_info
, DWORD
*num_clipplanes
,
780 const struct shader_arb_ctx_priv
*ctx
)
783 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
784 const struct wined3d_shader_lconst
*lconst
;
785 unsigned max_constantsF
;
788 /* In pixel shaders, all private constants are program local, we don't need anything
789 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
790 * If we need a private constant the GL implementation will squeeze it in somewhere
792 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
793 * immediate values. The posFixup is loaded using program.env for now, so always
794 * subtract one from the number of constants. If the shader uses indirect addressing,
795 * account for the helper const too because we have to declare all available d3d constants
796 * and don't know which are actually used.
800 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
801 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
802 if (max_constantsF
< 24)
803 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
807 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
808 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
809 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
810 * Also prevents max_constantsF from becoming less than 0 and
812 if (max_constantsF
< 96)
813 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
815 if (reg_maps
->usesrelconstF
)
817 DWORD highest_constf
= 0, clip_limit
;
819 max_constantsF
-= reserved_vs_const(shader_data
, reg_maps
, gl_info
);
820 max_constantsF
-= wined3d_popcount(reg_maps
->integer_constants
);
821 max_constantsF
-= gl_info
->reserved_arb_constants
;
823 for (i
= 0; i
< shader
->limits
->constant_float
; ++i
)
826 DWORD shift
= i
& 0x1f;
827 if (reg_maps
->constf
[idx
] & (1u << shift
))
831 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
833 if(ctx
->cur_vs_args
->super
.clip_enabled
)
834 clip_limit
= gl_info
->limits
.clipplanes
;
840 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
841 clip_limit
= min(wined3d_popcount(mask
), 4);
843 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
844 max_constantsF
-= *num_clipplanes
;
845 if(*num_clipplanes
< clip_limit
)
847 WARN("Only %u clipplanes out of %u enabled\n", *num_clipplanes
, gl_info
->limits
.clipplanes
);
852 if (ctx
->target_version
>= NV2
) *num_clipplanes
= gl_info
->limits
.clipplanes
;
853 else *num_clipplanes
= min(gl_info
->limits
.clipplanes
, 4);
857 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
859 if (map
& 1) shader_addline(buffer
, "TEMP R%u;\n", i
);
862 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
864 if (map
& 1) shader_addline(buffer
, "ADDRESS A%u;\n", i
);
867 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
869 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
871 if (map
& 1) shader_addline(buffer
, "TEMP T%u;\n", i
);
875 if (!shader
->load_local_constsF
)
877 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
880 value
= (const float *)lconst
->value
;
881 shader_addline(buffer
, "PARAM C%u = ", lconst
->idx
);
882 shader_arb_append_imm_vec4(buffer
, value
);
883 shader_addline(buffer
, ";\n");
887 /* After subtracting privately used constants from the hardware limit(they are loaded as
888 * local constants), make sure the shader doesn't violate the env constant limit
892 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
896 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
899 /* Avoid declaring more constants than needed */
900 max_constantsF
= min(max_constantsF
, shader
->limits
->constant_float
);
902 /* we use the array-based constants array if the local constants are marked for loading,
903 * because then we use indirect addressing, or when the local constant list is empty,
904 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
905 * local constants do not declare the loaded constants as an array because ARB compilers usually
906 * do not optimize unused constants away
908 if (reg_maps
->usesrelconstF
)
910 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
911 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
912 max_constantsF
, max_constantsF
- 1);
916 for (i
= 0; i
< max_constantsF
; ++i
)
920 mask
= 1u << (i
& 0x1fu
);
921 if (!shader_constant_is_local(shader
, i
) && (reg_maps
->constf
[idx
] & mask
))
923 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
929 static const char * const shift_tab
[] = {
930 "dummy", /* 0 (none) */
931 "coefmul.x", /* 1 (x2) */
932 "coefmul.y", /* 2 (x4) */
933 "coefmul.z", /* 3 (x8) */
934 "coefmul.w", /* 4 (x16) */
935 "dummy", /* 5 (x32) */
936 "dummy", /* 6 (x64) */
937 "dummy", /* 7 (x128) */
938 "dummy", /* 8 (d256) */
939 "dummy", /* 9 (d128) */
940 "dummy", /* 10 (d64) */
941 "dummy", /* 11 (d32) */
942 "coefdiv.w", /* 12 (d16) */
943 "coefdiv.z", /* 13 (d8) */
944 "coefdiv.y", /* 14 (d4) */
945 "coefdiv.x" /* 15 (d2) */
948 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
949 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
951 char *ptr
= write_mask
;
953 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
956 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
957 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
958 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
959 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
965 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
967 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
968 * but addressed as "rgba". To fix this we need to swap the register's x
969 * and z components. */
970 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
971 char *ptr
= swizzle_str
;
973 /* swizzle bits fields: wwzzyyxx */
974 DWORD swizzle
= param
->swizzle
;
975 DWORD swizzle_x
= swizzle
& 0x03;
976 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
977 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
978 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
980 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
981 * generate a swizzle string. Unless we need to our own swizzling. */
982 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
985 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
986 *ptr
++ = swizzle_chars
[swizzle_x
];
988 *ptr
++ = swizzle_chars
[swizzle_x
];
989 *ptr
++ = swizzle_chars
[swizzle_y
];
990 *ptr
++ = swizzle_chars
[swizzle_z
];
991 *ptr
++ = swizzle_chars
[swizzle_w
];
998 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
1000 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1001 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1003 if (!strcmp(priv
->addr_reg
, src
)) return;
1005 strcpy(priv
->addr_reg
, src
);
1006 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
1009 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1010 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
1012 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
1013 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
1015 /* oPos, oFog and oPts in D3D */
1016 static const char * const rastout_reg_names
[] = {"TMP_OUT", "TMP_FOGCOORD", "result.pointsize"};
1017 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1018 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1019 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1020 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1026 case WINED3DSPR_TEMP
:
1027 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1030 case WINED3DSPR_INPUT
:
1033 if (reg_maps
->shader_version
.major
< 3)
1035 if (!reg
->idx
[0].offset
)
1036 strcpy(register_name
, "fragment.color.primary");
1038 strcpy(register_name
, "fragment.color.secondary");
1042 if (reg
->idx
[0].rel_addr
)
1045 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1047 if (!strcmp(rel_reg
, "**aL_emul**"))
1049 DWORD idx
= ctx
->aL
+ reg
->idx
[0].offset
;
1050 if(idx
< MAX_REG_INPUT
)
1052 strcpy(register_name
, ctx
->ps_input
[idx
]);
1056 ERR("Pixel shader input register out of bounds: %u\n", idx
);
1057 sprintf(register_name
, "out_of_bounds_%u", idx
);
1060 else if (reg_maps
->input_registers
& 0x0300)
1062 /* There are two ways basically:
1064 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
1065 * That means trouble if the loop also contains a breakc or if the control values
1066 * aren't local constants.
1067 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
1068 * source dynamically. The trouble is that we cannot simply read aL.y because it
1069 * is an ADDRESS register. We could however push it, load .zw with a value and use
1070 * ADAC to load the condition code register and pop it again afterwards
1072 FIXME("Relative input register addressing with more than 8 registers\n");
1074 /* This is better than nothing for now */
1075 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1077 else if(ctx
->cur_ps_args
->super
.vp_mode
!= vertexshader
)
1079 /* This is problematic because we'd have to consult the ctx->ps_input strings
1080 * for where to find the varying. Some may be "0.0", others can be texcoords or
1081 * colors. This needs either a pipeline replacement to make the vertex shader feed
1082 * proper varyings, or loop unrolling
1084 * For now use the texcoords and hope for the best
1086 FIXME("Non-vertex shader varying input with indirect addressing\n");
1087 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1091 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1092 * pulls GL_NV_fragment_program2 in
1094 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1099 if (reg
->idx
[0].offset
< MAX_REG_INPUT
)
1101 strcpy(register_name
, ctx
->ps_input
[reg
->idx
[0].offset
]);
1105 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
[0].offset
);
1106 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
[0].offset
);
1113 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1u << reg
->idx
[0].offset
))
1115 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
[0].offset
);
1119 case WINED3DSPR_CONST
:
1120 if (!pshader
&& reg
->idx
[0].rel_addr
)
1122 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1123 UINT rel_offset
= ctx
->target_version
== ARB
? shader_data
->rel_offset
: 0;
1126 if (reg_maps
->shader_version
.major
< 2)
1128 sprintf(rel_reg
, "A0.x");
1132 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1133 if (ctx
->target_version
== ARB
)
1135 if (!strcmp(rel_reg
, "**aL_emul**"))
1139 shader_arb_request_a0(ins
, rel_reg
);
1140 sprintf(rel_reg
, "A0.x");
1145 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
[0].offset
);
1146 else if (reg
->idx
[0].offset
>= rel_offset
)
1147 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
[0].offset
- rel_offset
);
1149 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
[0].offset
);
1153 if (reg_maps
->usesrelconstF
)
1154 sprintf(register_name
, "C[%u]", reg
->idx
[0].offset
);
1156 sprintf(register_name
, "C%u", reg
->idx
[0].offset
);
1160 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1163 if (reg_maps
->shader_version
.major
== 1
1164 && reg_maps
->shader_version
.minor
<= 3)
1165 /* In ps <= 1.3, Tx is a temporary register as destination
1166 * to all instructions, and as source to most instructions.
1167 * For some instructions it is the texcoord input. Those
1168 * instructions know about the special use. */
1169 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1171 /* In ps 1.4 and 2.x Tx is always a (read-only) varying. */
1172 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
[0].offset
);
1176 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1177 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1179 sprintf(register_name
, "A%u_SHADOW", reg
->idx
[0].offset
);
1183 case WINED3DSPR_COLOROUT
:
1184 if (ctx
->ps_post_process
&& !reg
->idx
[0].offset
)
1186 strcpy(register_name
, "TMP_COLOR");
1190 if (ctx
->cur_ps_args
->super
.srgb_correction
)
1191 FIXME("sRGB correction on higher render targets.\n");
1192 if (reg_maps
->rt_mask
> 1)
1193 sprintf(register_name
, "result.color[%u]", reg
->idx
[0].offset
);
1195 strcpy(register_name
, "result.color");
1199 case WINED3DSPR_RASTOUT
:
1200 if (reg
->idx
[0].offset
== 1)
1201 sprintf(register_name
, "%s", ctx
->fog_output
);
1203 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
[0].offset
]);
1206 case WINED3DSPR_DEPTHOUT
:
1207 strcpy(register_name
, "result.depth");
1210 case WINED3DSPR_ATTROUT
:
1211 /* case WINED3DSPR_OUTPUT: */
1213 sprintf(register_name
, "oD[%u]", reg
->idx
[0].offset
);
1215 strcpy(register_name
, ctx
->color_output
[reg
->idx
[0].offset
]);
1218 case WINED3DSPR_TEXCRDOUT
:
1220 sprintf(register_name
, "oT[%u]", reg
->idx
[0].offset
);
1221 else if (reg_maps
->shader_version
.major
< 3)
1222 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
[0].offset
]);
1224 strcpy(register_name
, ctx
->vs_output
[reg
->idx
[0].offset
]);
1227 case WINED3DSPR_LOOP
:
1228 if(ctx
->target_version
>= NV2
)
1230 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1231 if(pshader
) sprintf(register_name
, "A0.x");
1232 else sprintf(register_name
, "aL.y");
1236 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1237 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1238 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1239 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1242 sprintf(register_name
, "**aL_emul**");
1247 case WINED3DSPR_CONSTINT
:
1248 sprintf(register_name
, "I%u", reg
->idx
[0].offset
);
1251 case WINED3DSPR_MISCTYPE
:
1252 if (!reg
->idx
[0].offset
)
1253 sprintf(register_name
, "vpos");
1254 else if (reg
->idx
[0].offset
== 1)
1255 sprintf(register_name
, "fragment.facing.x");
1257 FIXME("Unknown MISCTYPE register index %u.\n", reg
->idx
[0].offset
);
1261 FIXME("Unhandled register type %#x[%u].\n", reg
->type
, reg
->idx
[0].offset
);
1262 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
[0].offset
);
1267 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1268 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1270 char register_name
[255];
1274 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1275 strcpy(str
, register_name
);
1277 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1278 strcat(str
, write_mask
);
1281 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1283 switch(channel_source
)
1285 case CHANNEL_SOURCE_ZERO
: return "0";
1286 case CHANNEL_SOURCE_ONE
: return "1";
1287 case CHANNEL_SOURCE_X
: return "x";
1288 case CHANNEL_SOURCE_Y
: return "y";
1289 case CHANNEL_SOURCE_Z
: return "z";
1290 case CHANNEL_SOURCE_W
: return "w";
1292 FIXME("Unhandled channel source %#x\n", channel_source
);
1297 struct color_fixup_masks
1303 static struct color_fixup_masks
calc_color_correction(struct color_fixup_desc fixup
, DWORD dst_mask
)
1305 struct color_fixup_masks masks
= {0, 0};
1307 if (is_complex_fixup(fixup
))
1309 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1310 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1314 if (fixup
.x_source
!= CHANNEL_SOURCE_X
)
1315 masks
.source
|= WINED3DSP_WRITEMASK_0
;
1316 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
)
1317 masks
.source
|= WINED3DSP_WRITEMASK_1
;
1318 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
)
1319 masks
.source
|= WINED3DSP_WRITEMASK_2
;
1320 if (fixup
.w_source
!= CHANNEL_SOURCE_W
)
1321 masks
.source
|= WINED3DSP_WRITEMASK_3
;
1322 masks
.source
&= dst_mask
;
1324 if (fixup
.x_sign_fixup
)
1325 masks
.sign
|= WINED3DSP_WRITEMASK_0
;
1326 if (fixup
.y_sign_fixup
)
1327 masks
.sign
|= WINED3DSP_WRITEMASK_1
;
1328 if (fixup
.z_sign_fixup
)
1329 masks
.sign
|= WINED3DSP_WRITEMASK_2
;
1330 if (fixup
.w_sign_fixup
)
1331 masks
.sign
|= WINED3DSP_WRITEMASK_3
;
1332 masks
.sign
&= dst_mask
;
1337 static void gen_color_correction(struct wined3d_string_buffer
*buffer
, const char *dst
,
1338 const char *src
, const char *one
, const char *two
,
1339 struct color_fixup_desc fixup
, struct color_fixup_masks masks
)
1341 const char *sign_fixup_src
= dst
;
1346 sign_fixup_src
= "TA";
1348 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", sign_fixup_src
, src
,
1349 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1350 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1352 else if (masks
.sign
)
1354 sign_fixup_src
= src
;
1360 char *ptr
= reg_mask
;
1362 if (masks
.sign
!= WINED3DSP_WRITEMASK_ALL
)
1365 if (masks
.sign
& WINED3DSP_WRITEMASK_0
)
1367 if (masks
.sign
& WINED3DSP_WRITEMASK_1
)
1369 if (masks
.sign
& WINED3DSP_WRITEMASK_2
)
1371 if (masks
.sign
& WINED3DSP_WRITEMASK_3
)
1376 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", dst
, reg_mask
, sign_fixup_src
, two
, one
);
1380 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1383 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1384 if (!ins
->dst_count
) return "";
1386 mod
= ins
->dst
[0].modifiers
;
1388 /* Silently ignore PARTIALPRECISION if its not supported */
1389 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1391 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1393 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1394 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1399 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1402 case WINED3DSPDM_SATURATE
:
1405 case WINED3DSPDM_PARTIALPRECISION
:
1412 FIXME("Unknown modifiers 0x%08x\n", mod
);
1417 #define TEX_PROJ 0x1
1418 #define TEX_BIAS 0x2
1420 #define TEX_DERIV 0x10
1422 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1423 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1425 enum wined3d_shader_resource_type resource_type
= ins
->ctx
->reg_maps
->resource_info
[sampler_idx
].type
;
1426 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1427 const char *tex_type
;
1428 BOOL np2_fixup
= FALSE
;
1429 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1431 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1432 const struct wined3d_shader
*shader
;
1433 const struct wined3d_device
*device
;
1434 const struct wined3d_gl_info
*gl_info
;
1435 const char *tex_dst
= dst_str
;
1436 struct color_fixup_masks masks
;
1438 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1439 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1441 switch (resource_type
)
1443 case WINED3D_SHADER_RESOURCE_TEXTURE_1D
:
1447 case WINED3D_SHADER_RESOURCE_TEXTURE_2D
:
1448 shader
= ins
->ctx
->shader
;
1449 device
= shader
->device
;
1450 gl_info
= &device
->adapter
->gl_info
;
1452 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1u << sampler_idx
)
1453 && gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1457 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1459 if (priv
->cur_np2fixup_info
->super
.active
& (1u << sampler_idx
))
1461 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1462 else np2_fixup
= TRUE
;
1467 case WINED3D_SHADER_RESOURCE_TEXTURE_3D
:
1471 case WINED3D_SHADER_RESOURCE_TEXTURE_CUBE
:
1476 ERR("Unexpected resource type %#x.\n", resource_type
);
1480 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1481 * so don't use shader_arb_get_modifier
1483 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1486 /* Fragment samplers always have indentity mapping */
1487 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1489 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1494 masks
= calc_color_correction(priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
],
1495 ins
->dst
[0].write_mask
);
1497 if (masks
.source
|| masks
.sign
)
1501 if (flags
& TEX_DERIV
)
1503 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1504 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1505 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1506 dsx
, dsy
, sampler_idx
, tex_type
);
1508 else if(flags
& TEX_LOD
)
1510 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1511 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1512 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1513 sampler_idx
, tex_type
);
1515 else if (flags
& TEX_BIAS
)
1517 /* Shouldn't be possible, but let's check for it */
1518 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1519 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1520 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1522 else if (flags
& TEX_PROJ
)
1524 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1530 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1531 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1532 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1534 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, tex_dst
, sampler_idx
, tex_type
);
1537 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1542 gen_color_correction(buffer
, dst_str
, tex_dst
,
1543 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1544 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1545 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
], masks
);
1549 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1550 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1552 /* Generate a line that does the input modifier computation and return the input register to use */
1553 BOOL is_color
= FALSE
, insert_line
;
1556 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1557 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1558 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1559 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1561 /* Assume a new line will be added */
1564 /* Get register name */
1565 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1566 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1568 switch (src
->modifiers
)
1570 case WINED3DSPSM_NONE
:
1571 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1572 insert_line
= FALSE
;
1574 case WINED3DSPSM_NEG
:
1575 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1576 insert_line
= FALSE
;
1578 case WINED3DSPSM_BIAS
:
1579 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1581 case WINED3DSPSM_BIASNEG
:
1582 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1584 case WINED3DSPSM_SIGN
:
1585 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1587 case WINED3DSPSM_SIGNNEG
:
1588 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1590 case WINED3DSPSM_COMP
:
1591 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1593 case WINED3DSPSM_X2
:
1594 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1596 case WINED3DSPSM_X2NEG
:
1597 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1599 case WINED3DSPSM_DZ
:
1600 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1601 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1603 case WINED3DSPSM_DW
:
1604 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1605 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1607 case WINED3DSPSM_ABS
:
1608 if(ctx
->target_version
>= NV2
) {
1609 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1610 insert_line
= FALSE
;
1612 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1615 case WINED3DSPSM_ABSNEG
:
1616 if(ctx
->target_version
>= NV2
) {
1617 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1619 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1620 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1622 insert_line
= FALSE
;
1625 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1626 insert_line
= FALSE
;
1629 /* Return modified or original register, with swizzle */
1631 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1634 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1636 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1637 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1638 DWORD sampler_code
= dst
->reg
.idx
[0].offset
;
1640 char src_name
[2][50];
1642 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1644 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1646 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1647 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1650 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1651 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1652 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1653 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1654 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1656 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1657 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1660 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1665 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1666 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1667 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1668 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1669 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1670 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1671 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1672 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1673 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1674 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1675 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1676 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1677 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1679 FIXME("Unknown modifier %u\n", mod
);
1683 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1685 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1686 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1688 char src_name
[3][50];
1689 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1690 ins
->ctx
->reg_maps
->shader_version
.minor
);
1692 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1693 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1695 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
1696 && ins
->dst
->write_mask
!= WINED3DSP_WRITEMASK_3
)
1698 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1702 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1705 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1706 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1708 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1709 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1710 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1711 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1712 dst_name
, src_name
[1], src_name
[2]);
1716 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1718 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1719 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1721 char src_name
[3][50];
1723 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1725 /* Generate input register names (with modifiers) */
1726 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1727 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1728 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1730 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1731 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1734 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1735 * dst = dot2(src0, src1) + src2 */
1736 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1738 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1739 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1741 char src_name
[3][50];
1742 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1744 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1745 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1746 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1748 if(ctx
->target_version
>= NV3
)
1750 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1751 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1752 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1753 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1755 else if(ctx
->target_version
>= NV2
)
1757 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1758 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1759 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1760 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1762 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1764 * .xyxy and other swizzles that we could get with this are not valid in
1765 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1767 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1768 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1769 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1771 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1773 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1774 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1778 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1779 /* 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]
1780 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1782 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1783 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1784 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1785 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1789 /* Map the opcode 1-to-1 to the GL code */
1790 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1792 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1793 const char *instruction
;
1794 char arguments
[256], dst_str
[50];
1796 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1798 switch (ins
->handler_idx
)
1800 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1801 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1802 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1803 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1804 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1805 case WINED3DSIH_DST
: instruction
= "DST"; break;
1806 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1807 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1808 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1809 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1810 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1811 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1812 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1813 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1814 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1815 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1816 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1817 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1818 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1819 default: instruction
= "";
1820 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
1824 /* Note that shader_arb_add_dst_param() adds spaces. */
1825 arguments
[0] = '\0';
1826 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1827 for (i
= 0; i
< ins
->src_count
; ++i
)
1830 strcat(arguments
, ", ");
1831 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1832 strcat(arguments
, operand
);
1834 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1837 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1839 static DWORD
shader_arb_select_component(DWORD swizzle
, DWORD component
)
1841 return ((swizzle
>> 2 * component
) & 0x3) * 0x55;
1844 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1846 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1847 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1848 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1849 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1850 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1851 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1852 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1854 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1855 char src0_param
[256];
1857 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1859 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1861 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1863 if(ctx
->target_version
>= NV2
) {
1864 shader_hw_map2gl(ins
);
1867 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1868 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1870 /* This implements the mova formula used in GLSL. The first two instructions
1871 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1875 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1877 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1878 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1880 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1881 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1883 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1884 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1885 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1886 if (shader_data
->rel_offset
)
1888 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1890 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1892 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1894 else if (reg_maps
->shader_version
.major
== 1
1895 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1896 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1898 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1899 src0_param
[0] = '\0';
1901 if (shader_data
->rel_offset
&& ctx
->target_version
== ARB
)
1903 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1904 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1905 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1906 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1910 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1911 * with more than one component. Thus replicate the first source argument over all
1912 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1913 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1914 tmp_src
.swizzle
= shader_arb_select_component(tmp_src
.swizzle
, 0);
1915 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1916 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1919 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1921 if (ctx
->ps_post_process
&& shader
->u
.ps
.color0_mov
)
1923 shader_addline(buffer
, "#mov handled in srgb write or fog code\n");
1926 shader_hw_map2gl(ins
);
1930 shader_hw_map2gl(ins
);
1934 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1936 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1937 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1940 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1941 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1943 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1945 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1947 const char *kilsrc
= "TA";
1950 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1951 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1957 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1958 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1959 * masked out components to 0(won't kill)
1961 char x
= '0', y
= '0', z
= '0', w
= '0';
1962 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1963 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1964 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1965 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1966 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1968 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1972 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1973 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1975 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1976 * or pass in any temporary register(in shader phase 2)
1978 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1979 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1981 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1982 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1983 shader_addline(buffer
, "KIL TA;\n");
1987 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1989 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1990 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1991 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1992 ins
->ctx
->reg_maps
->shader_version
.minor
);
1993 struct wined3d_shader_src_param src
;
1997 DWORD reg_sampler_code
;
1999 BOOL swizzle_coord
= FALSE
;
2001 /* All versions have a destination register */
2002 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
2004 /* 1.0-1.4: Use destination register number as texture code.
2005 2.0+: Use provided sampler number as texture code. */
2006 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
2007 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
2009 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
2011 /* 1.0-1.3: Use the texcoord varying.
2012 1.4+: Use provided coordinate source register. */
2013 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2014 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
2016 /* TEX is the only instruction that can handle DW and DZ natively */
2018 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
2019 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
2020 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
2024 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
2025 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
2026 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
2028 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2031 if (reg_sampler_code
< MAX_TEXTURES
)
2032 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2033 if (flags
& WINED3D_PSARGS_PROJECTED
)
2035 myflags
|= TEX_PROJ
;
2036 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
2037 swizzle_coord
= TRUE
;
2040 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
2042 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
2043 if (src_mod
== WINED3DSPSM_DZ
)
2045 swizzle_coord
= TRUE
;
2046 myflags
|= TEX_PROJ
;
2047 } else if(src_mod
== WINED3DSPSM_DW
) {
2048 myflags
|= TEX_PROJ
;
2051 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
2052 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
2057 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
2058 * reg_coord is a read-only varying register, so we need a temp reg */
2059 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
2060 strcpy(reg_coord
, "TA");
2063 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
2066 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
2068 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2069 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2070 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2071 ins
->ctx
->reg_maps
->shader_version
.minor
);
2074 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2076 DWORD reg
= dst
->reg
.idx
[0].offset
;
2078 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2079 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2083 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2084 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2085 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2089 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2091 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2094 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2098 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2099 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2100 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2101 /* Move .x first in case src_str is "TA" */
2102 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2103 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2104 if (reg1
< MAX_TEXTURES
)
2106 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2107 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2109 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2112 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2114 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2116 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2120 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2121 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2122 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2123 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2124 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2125 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2128 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2130 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2134 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2135 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2136 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2137 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2140 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2142 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2143 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2144 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2145 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2146 DWORD reg_dest_code
;
2148 /* All versions have a destination register. The Tx where the texture coordinates come
2149 * from is the varying incarnation of the texture register
2151 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2152 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2153 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2154 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2156 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2157 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2159 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2160 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2162 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2163 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2166 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2167 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2168 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2169 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2171 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2172 * so we can't let the GL handle this.
2174 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2175 & WINED3D_PSARGS_PROJECTED
)
2177 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2178 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2179 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2181 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2184 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2186 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2188 /* No src swizzles are allowed, so this is ok */
2189 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2190 src_reg
, reg_dest_code
, reg_dest_code
);
2191 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2195 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2197 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2198 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2199 char src0_name
[50], dst_name
[50];
2201 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2203 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2204 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2205 * T<reg+1> register. Use this register to store the calculated vector
2207 tmp_reg
.idx
[0].offset
= reg
+ 1;
2208 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2209 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2212 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2214 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2216 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2217 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2223 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2224 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2226 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2227 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2228 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2229 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2230 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2233 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2235 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2236 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2237 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2238 char src0_name
[50], dst_name
[50];
2239 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2242 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2243 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2244 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2246 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2247 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2249 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2250 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2251 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2252 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2255 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2257 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2258 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2260 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2261 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2263 char src0_name
[50], dst_name
[50];
2266 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2267 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2268 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2270 /* Sample the texture using the calculated coordinates */
2271 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2272 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2273 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2274 tex_mx
->current_row
= 0;
2277 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2279 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2280 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2282 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2283 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2289 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2290 * components for temporary data storage
2292 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2293 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2294 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2296 /* Construct the eye-ray vector from w coordinates */
2297 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2298 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2299 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2301 /* Calculate reflection vector
2303 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2304 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2305 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2306 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2307 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2308 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2309 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2311 /* Sample the texture using the calculated coordinates */
2312 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2313 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2314 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2315 tex_mx
->current_row
= 0;
2318 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2320 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2321 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2323 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2324 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2331 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2332 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2333 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2334 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2335 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2337 /* Calculate reflection vector.
2340 * dst_reg.xyz = 2 * --------- * N - E
2343 * Which normalizes the normal vector
2345 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2346 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2347 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2348 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2349 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2350 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2352 /* Sample the texture using the calculated coordinates */
2353 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2354 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2355 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2356 tex_mx
->current_row
= 0;
2359 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2361 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2362 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2364 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2365 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2367 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2368 * which is essentially an input, is the destination register because it is the first
2369 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2370 * here(writemasks/swizzles are not valid on texdepth)
2372 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2374 /* According to the msdn, the source register(must be r5) is unusable after
2375 * the texdepth instruction, so we're free to modify it
2377 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2379 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2380 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2381 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2383 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2384 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2385 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2386 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2389 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2390 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2391 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2392 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2394 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2395 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2399 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2400 shader_addline(buffer
, "MOV TB, 0.0;\n");
2401 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2403 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2404 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2407 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2408 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2409 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2411 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2414 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2416 /* Handle output register */
2417 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2418 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2419 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2422 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2423 * Perform the 3rd row of a 3x3 matrix multiply */
2424 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2426 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2427 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2428 char dst_str
[50], dst_name
[50];
2432 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2433 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2434 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2435 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2436 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2439 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2440 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2441 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2442 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2444 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2446 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2447 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2448 char src0
[50], dst_name
[50];
2450 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2451 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2453 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2454 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2455 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2457 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2458 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2459 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2461 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2462 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2463 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2464 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2467 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2468 Vertex/Pixel shaders to ARB_vertex_program codes */
2469 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2472 int nComponents
= 0;
2473 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2474 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2475 struct wined3d_shader_instruction tmp_ins
;
2477 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2479 /* Set constants for the temporary argument */
2480 tmp_ins
.ctx
= ins
->ctx
;
2481 tmp_ins
.dst_count
= 1;
2482 tmp_ins
.dst
= &tmp_dst
;
2483 tmp_ins
.src_count
= 2;
2484 tmp_ins
.src
= tmp_src
;
2486 switch(ins
->handler_idx
)
2488 case WINED3DSIH_M4x4
:
2490 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2492 case WINED3DSIH_M4x3
:
2494 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2496 case WINED3DSIH_M3x4
:
2498 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2500 case WINED3DSIH_M3x3
:
2502 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2504 case WINED3DSIH_M3x2
:
2506 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2509 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2513 tmp_dst
= ins
->dst
[0];
2514 tmp_src
[0] = ins
->src
[0];
2515 tmp_src
[1] = ins
->src
[1];
2516 for (i
= 0; i
< nComponents
; ++i
)
2518 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2519 shader_hw_map2gl(&tmp_ins
);
2520 ++tmp_src
[1].reg
.idx
[0].offset
;
2524 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2530 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2531 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2532 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2533 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2534 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2535 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2536 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2537 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2538 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2539 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2540 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2541 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2542 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2544 FIXME("Unknown modifier %u\n", mod
);
2548 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2550 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2551 const char *instruction
;
2552 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2553 BOOL need_abs
= FALSE
;
2558 switch(ins
->handler_idx
)
2560 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2561 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2562 case WINED3DSIH_EXPP
:
2563 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2565 instruction
= "EXP";
2569 case WINED3DSIH_EXP
:
2570 instruction
= "EX2";
2572 case WINED3DSIH_LOG
:
2573 case WINED3DSIH_LOGP
:
2574 /* The precision requirements suggest that LOGP matches ARBvp's LOG
2575 * instruction, but notice that the output of those instructions is
2577 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2578 instruction
= "LG2";
2580 default: instruction
= "";
2581 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2585 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2587 src0_copy
.swizzle
= shader_arb_select_component(src0_copy
.swizzle
, 3);
2589 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2590 shader_arb_get_src_param(ins
, &src0_copy
, 0, src
);
2594 shader_addline(buffer
, "ABS TA.w, %s;\n", src
);
2595 shader_addline(buffer
, "%s%s %s, TA.w;\n", instruction
, shader_arb_get_modifier(ins
), dst
);
2599 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2604 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2606 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2609 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2610 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2611 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2613 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2614 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2616 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2617 * otherwise NRM or RSQ would return NaN */
2618 if(pshader
&& priv
->target_version
>= NV3
)
2620 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2622 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2624 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2625 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2626 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2628 else if(priv
->target_version
>= NV2
)
2630 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2631 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2632 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2637 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2639 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2640 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2641 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2643 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2644 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2646 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2647 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2648 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2653 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2655 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2657 char src_name
[3][50];
2659 /* ARB_fragment_program has a convenient LRP instruction */
2660 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2661 shader_hw_map2gl(ins
);
2665 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2666 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2667 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2668 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2670 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2671 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2672 dst_name
, src_name
[0], src_name
[2]);
2675 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2677 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2678 * must contain fixed constants. So we need a separate function to filter those constants and
2681 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2682 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2683 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2685 char src_name0
[50], src_name1
[50], src_name2
[50];
2688 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2689 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2690 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2691 /* No modifiers are supported on SCS */
2692 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2694 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2696 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2697 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2699 } else if(priv
->target_version
>= NV2
) {
2700 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2702 /* Sincos writemask must be .x, .y or .xy */
2703 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2704 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2705 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2706 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2708 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2709 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2711 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2712 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2714 * The constants we get are:
2716 * +1 +1, -1 -1 +1 +1 -1 -1
2717 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2718 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2720 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2724 * (x/2)^4 = x^4 / 16
2725 * (x/2)^5 = x^5 / 32
2728 * To get the final result:
2729 * sin(x) = 2 * sin(x/2) * cos(x/2)
2730 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2731 * (from sin(x+y) and cos(x+y) rules)
2733 * As per MSDN, dst.z is undefined after the operation, and so is
2734 * dst.x and dst.y if they're masked out by the writemask. Ie
2735 * sincos dst.y, src1, c0, c1
2736 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2737 * vsa.exe also stops with an error if the dest register is the same register as the source
2738 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2739 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2741 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2742 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2743 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2745 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2746 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2747 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2748 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2749 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2750 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2754 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2755 * properly merge that with MULs in the code above?
2756 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2757 * we can merge the sine and cosine MAD rows to calculate them together.
2759 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2760 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2761 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2762 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2765 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2766 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2767 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2769 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2771 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2772 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2774 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2776 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2777 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2782 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2784 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2787 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2789 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2790 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2792 /* SGN is only valid in vertex shaders */
2793 if(ctx
->target_version
>= NV2
) {
2794 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2798 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2799 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2801 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2802 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2804 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2805 * Then use TA, and calculate the final result
2807 * Not reading from TA? Store the first result in TA to avoid overwriting the
2808 * destination if src reg = dst reg
2810 if(strstr(src_name
, "TA"))
2812 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2813 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2814 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2818 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2819 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2820 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2825 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2827 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2833 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2834 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2835 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2837 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2838 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2841 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2843 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2844 char src0
[50], src1
[50], dst
[50];
2845 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2846 BOOL need_abs
= FALSE
;
2847 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2848 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2850 /* POW operates on the absolute value of the input */
2851 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2853 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2854 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2855 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2858 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2860 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2862 if (priv
->target_version
>= NV2
)
2864 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2865 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2866 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2870 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2871 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2873 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2874 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2875 /* Possibly add flt_eps to avoid getting float special values */
2876 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2877 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2878 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2879 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2883 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2885 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2887 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2890 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2894 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2895 struct list
*e
= list_head(&priv
->control_frames
);
2896 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2898 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2899 /* The constant loader makes sure to load -1 into iX.w */
2900 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2901 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2902 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2906 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2910 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2912 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2914 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2916 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2918 /* The constant loader makes sure to load -1 into iX.w */
2921 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2922 struct list
*e
= list_head(&priv
->control_frames
);
2923 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2925 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2927 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2928 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2929 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2933 shader_addline(buffer
, "REP %s;\n", src_name
);
2937 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2939 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2940 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2944 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2945 struct list
*e
= list_head(&priv
->control_frames
);
2946 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2948 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2949 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2950 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2952 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2956 shader_addline(buffer
, "ENDLOOP;\n");
2960 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2962 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2963 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2967 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2968 struct list
*e
= list_head(&priv
->control_frames
);
2969 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2971 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2972 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2973 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2975 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2979 shader_addline(buffer
, "ENDREP;\n");
2983 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2985 struct control_frame
*control_frame
;
2987 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2989 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2991 ERR("Could not find loop for break\n");
2995 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2997 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2998 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2999 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3003 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
3007 shader_addline(buffer
, "BRK;\n");
3011 static const char *get_compare(enum wined3d_shader_rel_op op
)
3015 case WINED3D_SHADER_REL_OP_GT
: return "GT";
3016 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
3017 case WINED3D_SHADER_REL_OP_GE
: return "GE";
3018 case WINED3D_SHADER_REL_OP_LT
: return "LT";
3019 case WINED3D_SHADER_REL_OP_NE
: return "NE";
3020 case WINED3D_SHADER_REL_OP_LE
: return "LE";
3022 FIXME("Unrecognized operator %#x.\n", op
);
3027 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
3031 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
3032 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
3033 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
3034 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
3035 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
3036 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
3038 FIXME("Unrecognized operator %#x.\n", op
);
3043 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
3045 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3046 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3047 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
3050 const char *comp
= get_compare(ins
->flags
);
3052 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3053 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3057 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
3058 * away the subtraction result
3060 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3061 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3065 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3066 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3070 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3072 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3073 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3074 struct list
*e
= list_head(&priv
->control_frames
);
3075 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3079 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3081 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3082 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3086 /* Invert the flag. We jump to the else label if the condition is NOT true */
3087 comp
= get_compare(invert_compare(ins
->flags
));
3088 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3089 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3093 comp
= get_compare(ins
->flags
);
3094 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3095 shader_addline(buffer
, "IF %s.x;\n", comp
);
3099 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3101 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3102 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3103 struct list
*e
= list_head(&priv
->control_frames
);
3104 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3105 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3109 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3110 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3111 control_frame
->had_else
= TRUE
;
3115 shader_addline(buffer
, "ELSE;\n");
3119 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3121 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3122 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3123 struct list
*e
= list_head(&priv
->control_frames
);
3124 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3125 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3129 if(control_frame
->had_else
)
3131 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3135 shader_addline(buffer
, "#No else branch. else is endif\n");
3136 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3141 shader_addline(buffer
, "ENDIF;\n");
3145 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3147 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3149 char reg_src
[3][40];
3150 WORD flags
= TEX_DERIV
;
3152 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3153 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3154 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3155 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3157 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3158 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3160 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3163 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3165 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3168 WORD flags
= TEX_LOD
;
3170 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3171 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3173 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3174 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3176 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3179 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3181 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3182 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3184 priv
->in_main_func
= FALSE
;
3185 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3186 * subroutine, don't generate a label that will make GL complain
3188 if(priv
->target_version
== ARB
) return;
3190 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3193 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3194 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3195 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3196 struct wined3d_string_buffer
*buffer
)
3200 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3201 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3202 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3203 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3205 if (args
->super
.fog_src
== VS_FOG_Z
)
3207 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3213 /* posFixup.x is always 1.0, so we can safely use it */
3214 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3218 /* Clamp fogcoord */
3219 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3220 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3222 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3223 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3227 /* Clipplanes are always stored without y inversion */
3228 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3230 if (args
->super
.clip_enabled
)
3232 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3234 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3238 else if (args
->clip
.boolclip
.clip_texcoord
)
3240 static const char component
[4] = {'x', 'y', 'z', 'w'};
3241 unsigned int cur_clip
= 0;
3242 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3244 for (i
= 0; i
< gl_info
->limits
.clipplanes
; ++i
)
3246 if (args
->clip
.boolclip
.clipplane_mask
& (1u << i
))
3248 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3249 component
[cur_clip
++], i
);
3255 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3258 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3261 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3264 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3267 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3268 args
->clip
.boolclip
.clip_texcoord
- 1);
3271 /* Write the final position.
3273 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3274 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3275 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3276 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3278 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3279 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3280 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3282 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3283 * and the glsl equivalent
3285 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3287 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3288 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3292 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3293 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3296 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3298 priv_ctx
->footer_written
= TRUE
;
3301 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3303 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3304 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3305 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3306 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3308 if(priv
->target_version
== ARB
) return;
3312 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3313 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, ins
->ctx
->gl_info
, buffer
);
3316 shader_addline(buffer
, "RET;\n");
3319 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3321 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3322 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3325 static BOOL
shader_arb_compile(const struct wined3d_gl_info
*gl_info
, GLenum target
, const char *src
)
3327 const char *ptr
, *line
;
3330 if (TRACE_ON(d3d_shader
))
3333 while ((line
= get_line(&ptr
))) TRACE_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3336 GL_EXTCALL(glProgramStringARB(target
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(src
), src
));
3337 checkGLcall("glProgramStringARB()");
3339 if (FIXME_ON(d3d_shader
))
3341 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3344 FIXME_(d3d_shader
)("Program error at position %d: %s\n\n", pos
,
3345 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3347 while ((line
= get_line(&ptr
))) FIXME_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3348 FIXME_(d3d_shader
)("\n");
3354 if (WARN_ON(d3d_perf
))
3356 GL_EXTCALL(glGetProgramivARB(target
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3357 checkGLcall("glGetProgramivARB()");
3359 WARN_(d3d_perf
)("Program exceeds native resource limits.\n");
3365 /* Context activation is done by the caller. */
3366 static GLuint
create_arb_blt_vertex_program(const struct wined3d_gl_info
*gl_info
)
3368 GLuint program_id
= 0;
3370 static const char blt_vprogram
[] =
3372 "PARAM c[1] = { { 1, 0.5 } };\n"
3373 "MOV result.position, vertex.position;\n"
3374 "MOV result.color, c[0].x;\n"
3375 "MOV result.texcoord[0], vertex.texcoord[0];\n"
3378 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3379 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
3380 shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, blt_vprogram
);
3385 /* Context activation is done by the caller. */
3386 static GLuint
create_arb_blt_fragment_program(const struct wined3d_gl_info
*gl_info
,
3387 enum wined3d_gl_resource_type tex_type
, BOOL masked
)
3389 GLuint program_id
= 0;
3390 const char *fprogram
;
3392 static const char * const blt_fprograms_full
[WINED3D_GL_RES_TYPE_COUNT
] =
3394 /* WINED3D_GL_RES_TYPE_TEX_1D */
3396 /* WINED3D_GL_RES_TYPE_TEX_2D */
3399 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3400 "MOV result.depth.z, R0.x;\n"
3402 /* WINED3D_GL_RES_TYPE_TEX_3D */
3404 /* WINED3D_GL_RES_TYPE_TEX_CUBE */
3407 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3408 "MOV result.depth.z, R0.x;\n"
3410 /* WINED3D_GL_RES_TYPE_TEX_RECT */
3413 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3414 "MOV result.depth.z, R0.x;\n"
3416 /* WINED3D_GL_RES_TYPE_BUFFER */
3418 /* WINED3D_GL_RES_TYPE_RB */
3422 static const char * const blt_fprograms_masked
[WINED3D_GL_RES_TYPE_COUNT
] =
3424 /* WINED3D_GL_RES_TYPE_TEX_1D */
3426 /* WINED3D_GL_RES_TYPE_TEX_2D */
3428 "PARAM mask = program.local[0];\n"
3430 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3431 "MUL R0.x, R0.x, R0.y;\n"
3433 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3434 "MOV result.depth.z, R0.x;\n"
3436 /* WINED3D_GL_RES_TYPE_TEX_3D */
3438 /* WINED3D_GL_RES_TYPE_TEX_CUBE */
3440 "PARAM mask = program.local[0];\n"
3442 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3443 "MUL R0.x, R0.x, R0.y;\n"
3445 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3446 "MOV result.depth.z, R0.x;\n"
3448 /* WINED3D_GL_RES_TYPE_TEX_RECT */
3450 "PARAM mask = program.local[0];\n"
3452 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3453 "MUL R0.x, R0.x, R0.y;\n"
3455 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3456 "MOV result.depth.z, R0.x;\n"
3458 /* WINED3D_GL_RES_TYPE_BUFFER */
3460 /* WINED3D_GL_RES_TYPE_RB */
3464 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3467 FIXME("tex_type %#x not supported, falling back to 2D\n", tex_type
);
3468 tex_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
3469 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3472 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3473 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
3474 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, fprogram
);
3479 static void arbfp_add_sRGB_correction(struct wined3d_string_buffer
*buffer
, const char *fragcolor
,
3480 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3482 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3486 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3487 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.x;\n", tmp1
, fragcolor
);
3488 /* Calculate the > 0.0031308 case */
3489 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3490 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3491 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3492 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts0.y;\n", fragcolor
, fragcolor
);
3493 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts0.z;\n", fragcolor
, fragcolor
);
3494 /* Calculate the < case */
3495 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts0.w, %s;\n", fragcolor
, fragcolor
);
3499 /* Calculate the > 0.0031308 case */
3500 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts0.x;\n", tmp1
, fragcolor
);
3501 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts0.x;\n", tmp1
, fragcolor
);
3502 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts0.x;\n", tmp1
, fragcolor
);
3503 shader_addline(buffer
, "MUL %s, %s, srgb_consts0.y;\n", tmp1
, tmp1
);
3504 shader_addline(buffer
, "SUB %s, %s, srgb_consts0.z;\n", tmp1
, tmp1
);
3505 /* Calculate the < case */
3506 shader_addline(buffer
, "MUL %s, srgb_consts0.w, %s;\n", tmp2
, fragcolor
);
3507 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3508 shader_addline(buffer
, "SLT %s, srgb_consts1.x, %s;\n", tmp3
, fragcolor
);
3509 shader_addline(buffer
, "SGE %s, srgb_consts1.x, %s;\n", tmp4
, fragcolor
);
3510 /* Store the components > 0.0031308 in the destination */
3511 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3512 /* Add the components that are < 0.0031308 */
3513 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3514 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3515 * result.color writes(.rgb first, then .a), or handle overwriting already written
3516 * components. The assembler uses a temporary register in this case, which is usually
3517 * not allocated from one of our registers that were used earlier.
3520 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3523 static const DWORD
*find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3525 const struct wined3d_shader_lconst
*constant
;
3527 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3529 if (constant
->idx
== idx
)
3531 return constant
->value
;
3537 static void init_ps_input(const struct wined3d_shader
*shader
,
3538 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3540 static const char * const texcoords
[8] =
3542 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3543 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3546 const struct wined3d_shader_signature_element
*input
;
3547 const char *semantic_name
;
3550 switch(args
->super
.vp_mode
)
3552 case pretransformed
:
3554 /* The pixelshader has to collect the varyings on its own. In any case properly load
3555 * color0 and color1. In the case of pretransformed vertices also load texcoords. Set
3556 * other attribs to 0.0.
3558 * For fixedfunction this behavior is correct, according to the tests. For pretransformed
3559 * we'd either need a replacement shader that can load other attribs like BINORMAL, or
3560 * load the texcoord attrib pointers to match the pixel shader signature
3562 for (i
= 0; i
< shader
->input_signature
.element_count
; ++i
)
3564 input
= &shader
->input_signature
.elements
[i
];
3565 if (!(semantic_name
= input
->semantic_name
))
3567 semantic_idx
= input
->semantic_idx
;
3569 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3572 priv
->ps_input
[input
->register_idx
] = "fragment.color.primary";
3573 else if (semantic_idx
== 1)
3574 priv
->ps_input
[input
->register_idx
] = "fragment.color.secondary";
3576 priv
->ps_input
[input
->register_idx
] = "0.0";
3578 else if (args
->super
.vp_mode
== fixedfunction
)
3580 priv
->ps_input
[input
->register_idx
] = "0.0";
3582 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3584 if (semantic_idx
< 8)
3585 priv
->ps_input
[input
->register_idx
] = texcoords
[semantic_idx
];
3587 priv
->ps_input
[input
->register_idx
] = "0.0";
3589 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3592 priv
->ps_input
[input
->register_idx
] = "fragment.fogcoord";
3594 priv
->ps_input
[input
->register_idx
] = "0.0";
3598 priv
->ps_input
[input
->register_idx
] = "0.0";
3601 TRACE("v%u, semantic %s%u is %s\n", input
->register_idx
,
3602 semantic_name
, semantic_idx
, priv
->ps_input
[input
->register_idx
]);
3607 /* That one is easy. The vertex shaders provide v0-v7 in fragment.texcoord and v8 and v9 in
3610 for(i
= 0; i
< 8; i
++)
3612 priv
->ps_input
[i
] = texcoords
[i
];
3614 priv
->ps_input
[8] = "fragment.color.primary";
3615 priv
->ps_input
[9] = "fragment.color.secondary";
3620 static void arbfp_add_linear_fog(struct wined3d_string_buffer
*buffer
,
3621 const char *fragcolor
, const char *tmp
)
3623 shader_addline(buffer
, "SUB %s.x, state.fog.params.z, fragment.fogcoord.x;\n", tmp
);
3624 shader_addline(buffer
, "MUL_SAT %s.x, %s.x, state.fog.params.w;\n", tmp
, tmp
);
3625 shader_addline(buffer
, "LRP %s.rgb, %s.x, %s, state.fog.color;\n", fragcolor
, tmp
, fragcolor
);
3628 /* Context activation is done by the caller. */
3629 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3630 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
3631 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3633 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3634 const DWORD
*function
= shader
->function
;
3637 DWORD next_local
= 0;
3638 struct shader_arb_ctx_priv priv_ctx
;
3639 BOOL dcl_td
= FALSE
;
3640 BOOL want_nv_prog
= FALSE
;
3641 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3643 BOOL custom_linear_fog
= FALSE
;
3647 unsigned int i
, found
= 0;
3649 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3652 || (shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3653 || (reg_maps
->shader_version
.major
< 2 && !i
))
3656 sprintf(srgbtmp
[found
], "R%u", i
);
3658 if (found
== 4) break;
3663 sprintf(srgbtmp
[0], "TA");
3664 sprintf(srgbtmp
[1], "TB");
3665 sprintf(srgbtmp
[2], "TC");
3666 sprintf(srgbtmp
[3], "TD");
3670 sprintf(srgbtmp
[1], "TA");
3671 sprintf(srgbtmp
[2], "TB");
3672 sprintf(srgbtmp
[3], "TC");
3675 sprintf(srgbtmp
[2], "TA");
3676 sprintf(srgbtmp
[3], "TB");
3679 sprintf(srgbtmp
[3], "TA");
3685 /* Create the hw ARB shader */
3686 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3687 priv_ctx
.cur_ps_args
= args
;
3688 priv_ctx
.compiled_fprog
= compiled
;
3689 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3690 init_ps_input(shader
, args
, &priv_ctx
);
3691 list_init(&priv_ctx
.control_frames
);
3692 priv_ctx
.ps_post_process
= args
->super
.srgb_correction
;
3694 /* Avoid enabling NV_fragment_program* if we do not need it.
3696 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3697 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3698 * is faster than what we gain from using higher native instructions. There are some things though
3699 * that cannot be emulated. In that case enable the extensions.
3700 * If the extension is enabled, instruction handlers that support both ways will use it.
3702 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3703 * So enable the best we can get.
3705 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3706 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3708 want_nv_prog
= TRUE
;
3711 shader_addline(buffer
, "!!ARBfp1.0\n");
3712 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3714 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3715 priv_ctx
.target_version
= NV3
;
3717 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3719 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3720 priv_ctx
.target_version
= NV2
;
3724 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3727 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3730 priv_ctx
.target_version
= ARB
;
3733 if (reg_maps
->rt_mask
> 1)
3735 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3738 if (reg_maps
->shader_version
.major
< 3)
3740 switch (args
->super
.fog
)
3742 case WINED3D_FFP_PS_FOG_OFF
:
3744 case WINED3D_FFP_PS_FOG_LINEAR
:
3745 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
3747 custom_linear_fog
= TRUE
;
3748 priv_ctx
.ps_post_process
= TRUE
;
3751 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3753 case WINED3D_FFP_PS_FOG_EXP
:
3754 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3756 case WINED3D_FFP_PS_FOG_EXP2
:
3757 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3762 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3763 * unused temps away(but occupies them for the whole shader if they're used once). Always
3764 * declaring them avoids tricky bookkeeping work
3766 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3767 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3768 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3769 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3770 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3771 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3772 wined3d_ftoa(eps
, ftoa_tmp
);
3773 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %s, 0.0 };\n", ftoa_tmp
);
3775 if (reg_maps
->shader_version
.major
< 2)
3777 strcpy(fragcolor
, "R0");
3781 if (priv_ctx
.ps_post_process
)
3783 if (shader
->u
.ps
.color0_mov
)
3785 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3789 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3790 strcpy(fragcolor
, "TMP_COLOR");
3793 strcpy(fragcolor
, "result.color");
3797 if (args
->super
.srgb_correction
)
3799 shader_addline(buffer
, "PARAM srgb_consts0 = ");
3800 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const0
);
3801 shader_addline(buffer
, ";\n");
3802 shader_addline(buffer
, "PARAM srgb_consts1 = ");
3803 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const1
);
3804 shader_addline(buffer
, ";\n");
3807 /* Base Declarations */
3808 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3810 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3812 unsigned char bump_const
;
3814 if (!(map
& 1)) continue;
3816 bump_const
= compiled
->numbumpenvmatconsts
;
3817 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3818 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3819 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3820 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3822 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3823 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3824 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3825 * textures due to conditional NP2 restrictions)
3827 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3828 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3829 * their location is shader dependent anyway and they cannot be loaded globally.
3831 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3832 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3833 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3834 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3836 if (!(reg_maps
->luminanceparams
& (1u << i
)))
3839 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3840 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3841 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3844 for(i
= 0; i
< MAX_CONST_I
; i
++)
3846 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3847 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
3849 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
3853 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3854 control_values
[0], control_values
[1], control_values
[2]);
3858 compiled
->int_consts
[i
] = next_local
;
3859 compiled
->num_int_consts
++;
3860 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3865 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3867 compiled
->ycorrection
= next_local
;
3868 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3872 shader_addline(buffer
, "TEMP vpos;\n");
3873 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3874 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3875 * ycorrection.z: 1.0
3876 * ycorrection.w: 0.0
3878 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3879 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3884 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3887 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3888 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3889 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3890 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3891 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3892 * shader compilation errors and the subsequent errors when drawing with this shader. */
3893 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3894 unsigned char cur_fixup_sampler
= 0;
3896 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3897 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3898 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3900 fixup
->offset
= next_local
;
3901 fixup
->super
.active
= 0;
3903 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
)
3905 if (!(map
& (1u << i
)))
3908 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
)
3910 fixup
->super
.active
|= (1u << i
);
3911 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3915 FIXME("No free constant found to load NP2 fixup data into shader. "
3916 "Sampling from this texture will probably look wrong.\n");
3921 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3922 if (fixup
->super
.num_consts
) {
3923 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3924 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3928 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3930 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3933 /* Base Shader Body */
3934 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
3936 if(args
->super
.srgb_correction
) {
3937 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3938 priv_ctx
.target_version
>= NV2
);
3941 if (custom_linear_fog
)
3942 arbfp_add_linear_fog(buffer
, fragcolor
, "TA");
3944 if(strcmp(fragcolor
, "result.color")) {
3945 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3947 shader_addline(buffer
, "END\n");
3949 /* TODO: change to resource.glObjectHandle or something like that */
3950 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3952 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3953 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3955 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3956 if (!shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
->buffer
))
3962 static int compare_sig(const struct wined3d_shader_signature
*sig1
, const struct wined3d_shader_signature
*sig2
)
3967 if (sig1
->element_count
!= sig2
->element_count
)
3968 return sig1
->element_count
< sig2
->element_count
? -1 : 1;
3970 for (i
= 0; i
< sig1
->element_count
; ++i
)
3972 const struct wined3d_shader_signature_element
*e1
, *e2
;
3974 e1
= &sig1
->elements
[i
];
3975 e2
= &sig2
->elements
[i
];
3977 if (!e1
->semantic_name
|| !e2
->semantic_name
)
3979 /* Compare pointers, not contents. One string is NULL (element
3980 * does not exist), the other one is not NULL. */
3981 if (e1
->semantic_name
!= e2
->semantic_name
)
3982 return e1
->semantic_name
< e2
->semantic_name
? -1 : 1;
3986 if ((ret
= strcmp(e1
->semantic_name
, e2
->semantic_name
)))
3988 if (e1
->semantic_idx
!= e2
->semantic_idx
)
3989 return e1
->semantic_idx
< e2
->semantic_idx
? -1 : 1;
3990 if (e1
->sysval_semantic
!= e2
->sysval_semantic
)
3991 return e1
->sysval_semantic
< e2
->sysval_semantic
? -1 : 1;
3992 if (e1
->component_type
!= e2
->component_type
)
3993 return e1
->component_type
< e2
->component_type
? -1 : 1;
3994 if (e1
->register_idx
!= e2
->register_idx
)
3995 return e1
->register_idx
< e2
->register_idx
? -1 : 1;
3996 if (e1
->mask
!= e2
->mask
)
3997 return e1
->mask
< e2
->mask
? -1 : 1;
4002 static void clone_sig(struct wined3d_shader_signature
*new, const struct wined3d_shader_signature
*sig
)
4007 new->element_count
= sig
->element_count
;
4008 new->elements
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new->elements
) * new->element_count
);
4009 for (i
= 0; i
< sig
->element_count
; ++i
)
4011 new->elements
[i
] = sig
->elements
[i
];
4013 if (!new->elements
[i
].semantic_name
)
4016 /* Clone the semantic string */
4017 name
= HeapAlloc(GetProcessHeap(), 0, strlen(sig
->elements
[i
].semantic_name
) + 1);
4018 strcpy(name
, sig
->elements
[i
].semantic_name
);
4019 new->elements
[i
].semantic_name
= name
;
4023 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature
*sig
)
4025 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
4026 struct ps_signature
*found_sig
;
4030 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4031 TRACE("Found existing signature %u\n", found_sig
->idx
);
4032 return found_sig
->idx
;
4034 found_sig
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*found_sig
));
4035 clone_sig(&found_sig
->sig
, sig
);
4036 found_sig
->idx
= priv
->ps_sig_number
++;
4037 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
4038 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
4040 ERR("Failed to insert program entry.\n");
4042 return found_sig
->idx
;
4045 static void init_output_registers(const struct wined3d_shader
*shader
,
4046 const struct wined3d_shader_signature
*ps_input_sig
,
4047 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
4050 static const char * const texcoords
[8] =
4052 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
4053 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
4055 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
4056 * and varying 9 to result.color.secondary
4058 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
4060 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
4061 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
4062 "result.color.primary", "result.color.secondary"
4067 TRACE("Pixel shader uses builtin varyings\n");
4068 /* Map builtins to builtins */
4069 for(i
= 0; i
< 8; i
++)
4071 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
4073 priv_ctx
->color_output
[0] = "result.color.primary";
4074 priv_ctx
->color_output
[1] = "result.color.secondary";
4075 priv_ctx
->fog_output
= "TMP_FOGCOORD";
4077 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
4078 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4080 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4082 if (!output
->semantic_name
)
4085 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
))
4087 TRACE("o%u is TMP_OUT\n", output
->register_idx
);
4088 if (!output
->semantic_idx
)
4089 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4091 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4093 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
4095 TRACE("o%u is result.pointsize\n", output
->register_idx
);
4096 if (!output
->semantic_idx
)
4097 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4099 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4101 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4103 TRACE("o%u is result.color.?, idx %u\n", output
->register_idx
, output
->semantic_idx
);
4104 if (!output
->semantic_idx
)
4105 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.primary";
4106 else if (output
->semantic_idx
== 1)
4107 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.secondary";
4108 else priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4110 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4112 TRACE("o%u is result.texcoord[%u]\n", output
->register_idx
, output
->semantic_idx
);
4113 if (output
->semantic_idx
>= 8)
4114 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4116 priv_ctx
->vs_output
[output
->register_idx
] = texcoords
[output
->semantic_idx
];
4118 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4120 TRACE("o%u is result.fogcoord\n", output
->register_idx
);
4121 if (output
->semantic_idx
> 0)
4122 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4124 priv_ctx
->vs_output
[output
->register_idx
] = "result.fogcoord";
4128 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4134 TRACE("Pixel shader uses declared varyings\n");
4136 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4137 for(i
= 0; i
< 8; i
++)
4139 priv_ctx
->texcrd_output
[i
] = "TA";
4141 priv_ctx
->color_output
[0] = "TA";
4142 priv_ctx
->color_output
[1] = "TA";
4143 priv_ctx
->fog_output
= "TA";
4145 for (i
= 0; i
< ps_input_sig
->element_count
; ++i
)
4147 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[i
];
4149 if (!input
->semantic_name
)
4152 /* If a declared input register is not written by builtin arguments, don't write to it.
4153 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4155 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4156 * to TMP_OUT in any case
4158 if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4160 if (input
->semantic_idx
< 8)
4161 priv_ctx
->texcrd_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4163 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4165 if (input
->semantic_idx
< 2)
4166 priv_ctx
->color_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4168 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4170 if (!input
->semantic_idx
)
4171 priv_ctx
->fog_output
= decl_idx_to_string
[input
->register_idx
];
4178 if (!strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.primary")
4179 || !strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.secondary"))
4181 compiled
->need_color_unclamp
= TRUE
;
4185 /* Map declared to declared */
4186 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4188 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4190 /* Write unread output to TA to throw them away */
4191 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4193 if (!output
->semantic_name
)
4196 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
) && !output
->semantic_idx
)
4198 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4201 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
) && !output
->semantic_idx
)
4203 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4207 for (j
= 0; j
< ps_input_sig
->element_count
; ++j
)
4209 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[j
];
4211 if (!input
->semantic_name
)
4214 if (!strcmp(input
->semantic_name
, output
->semantic_name
)
4215 && input
->semantic_idx
== output
->semantic_idx
)
4217 priv_ctx
->vs_output
[output
->register_idx
] = decl_idx_to_string
[input
->register_idx
];
4219 if (!strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.primary")
4220 || !strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.secondary"))
4222 compiled
->need_color_unclamp
= TRUE
;
4229 /* Context activation is done by the caller. */
4230 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4231 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
4232 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4233 const struct wined3d_shader_signature
*ps_input_sig
)
4235 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4236 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4237 struct shader_arb_priv
*priv
= shader
->device
->shader_priv
;
4238 const DWORD
*function
= shader
->function
;
4240 DWORD next_local
= 0;
4241 struct shader_arb_ctx_priv priv_ctx
;
4244 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4245 priv_ctx
.cur_vs_args
= args
;
4246 list_init(&priv_ctx
.control_frames
);
4247 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4249 /* Create the hw ARB shader */
4250 shader_addline(buffer
, "!!ARBvp1.0\n");
4252 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4253 * mesurable performance penalty, and we can always make use of it for clipplanes.
4255 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4257 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4258 priv_ctx
.target_version
= NV3
;
4259 shader_addline(buffer
, "ADDRESS aL;\n");
4261 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4263 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4264 priv_ctx
.target_version
= NV2
;
4265 shader_addline(buffer
, "ADDRESS aL;\n");
4267 priv_ctx
.target_version
= ARB
;
4270 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4272 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4273 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4276 wined3d_ftoa(eps
, ftoa_tmp
);
4277 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %s};\n", ftoa_tmp
);
4279 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4281 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4282 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4285 shader_addline(buffer
, "TEMP TA;\n");
4286 shader_addline(buffer
, "TEMP TB;\n");
4288 /* Base Declarations */
4289 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4290 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4292 for(i
= 0; i
< MAX_CONST_I
; i
++)
4294 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4295 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
4297 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
4301 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4302 control_values
[0], control_values
[1], control_values
[2]);
4306 compiled
->int_consts
[i
] = next_local
;
4307 compiled
->num_int_consts
++;
4308 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4313 /* We need a constant to fixup the final position */
4314 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4315 compiled
->pos_fixup
= next_local
++;
4317 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4318 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4319 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4320 * a replacement shader depend on the texcoord.w being set properly.
4322 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4323 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4324 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4325 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4326 * this can eat a number of instructions, so skip it unless this cap is set as well
4328 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4330 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4331 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4333 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !priv
->ffp_proj_control
)
4336 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4337 for(i
= 0; i
< MAX_REG_TEXCRD
; i
++)
4339 if (reg_maps
->u
.texcoord_mask
[i
] && reg_maps
->u
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
)
4340 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4345 /* The shader starts with the main function */
4346 priv_ctx
.in_main_func
= TRUE
;
4347 /* Base Shader Body */
4348 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4350 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4351 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4353 shader_addline(buffer
, "END\n");
4355 /* TODO: change to resource.glObjectHandle or something like that */
4356 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4358 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4359 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4361 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4362 if (!shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, buffer
->buffer
))
4368 /* Context activation is done by the caller. */
4369 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_shader
*shader
,
4370 const struct arb_ps_compile_args
*args
)
4372 struct wined3d_device
*device
= shader
->device
;
4373 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4374 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4377 struct arb_ps_compiled_shader
*new_array
;
4378 struct wined3d_string_buffer buffer
;
4379 struct arb_pshader_private
*shader_data
;
4382 if (!shader
->backend_data
)
4384 struct shader_arb_priv
*priv
= device
->shader_priv
;
4386 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4387 shader_data
= shader
->backend_data
;
4388 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4390 if (shader
->reg_maps
.shader_version
.major
< 3)
4391 shader_data
->input_signature_idx
= ~0U;
4393 shader_data
->input_signature_idx
= find_input_signature(priv
, &shader
->input_signature
);
4395 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4397 if (!d3d_info
->vs_clipping
)
4398 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4399 d3d_info
->limits
.ffp_blend_stages
- 1);
4401 shader_data
->clipplane_emulation
= ~0U;
4403 shader_data
= shader
->backend_data
;
4405 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4406 * so a linear search is more performant than a hashmap or a binary search
4407 * (cache coherency etc)
4409 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4411 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4412 return &shader_data
->gl_shaders
[i
];
4415 TRACE("No matching GL shader found, compiling a new shader\n");
4416 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4417 if (shader_data
->num_gl_shaders
)
4419 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4420 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4421 new_size
* sizeof(*shader_data
->gl_shaders
));
4423 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4428 ERR("Out of memory\n");
4431 shader_data
->gl_shaders
= new_array
;
4432 shader_data
->shader_array_size
= new_size
;
4435 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4437 pixelshader_update_resource_types(shader
, args
->super
.tex_types
);
4439 if (!string_buffer_init(&buffer
))
4441 ERR("Failed to initialize shader buffer.\n");
4445 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4446 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4447 string_buffer_free(&buffer
);
4448 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4450 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4453 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4454 const DWORD use_map
, BOOL skip_int
) {
4455 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4456 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4457 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4458 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4459 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4460 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4461 if(skip_int
) return TRUE
;
4463 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4466 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4467 const struct wined3d_gl_info
*gl_info
, DWORD use_map
, const struct arb_vs_compile_args
*args
,
4468 const struct wined3d_shader_signature
*ps_input_sig
)
4472 struct arb_vs_compiled_shader
*new_array
;
4473 struct wined3d_string_buffer buffer
;
4474 struct arb_vshader_private
*shader_data
;
4477 if (!shader
->backend_data
)
4479 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4481 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4482 shader_data
= shader
->backend_data
;
4484 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4485 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4487 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4489 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4490 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4491 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4493 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4494 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4495 else if (reg_maps
->max_rel_offset
> 63)
4496 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4499 shader_data
= shader
->backend_data
;
4501 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4502 * so a linear search is more performant than a hashmap or a binary search
4503 * (cache coherency etc)
4505 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4506 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4507 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4509 return &shader_data
->gl_shaders
[i
];
4513 TRACE("No matching GL shader found, compiling a new shader\n");
4515 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4516 if (shader_data
->num_gl_shaders
)
4518 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4519 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4520 new_size
* sizeof(*shader_data
->gl_shaders
));
4522 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4527 ERR("Out of memory\n");
4530 shader_data
->gl_shaders
= new_array
;
4531 shader_data
->shader_array_size
= new_size
;
4534 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4536 if (!string_buffer_init(&buffer
))
4538 ERR("Failed to initialize shader buffer.\n");
4542 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4543 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4545 string_buffer_free(&buffer
);
4546 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4548 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4551 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4552 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4553 struct arb_ps_compile_args
*args
)
4555 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4556 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4560 find_ps_compile_args(state
, shader
, context
->stream_info
.position_transformed
, &args
->super
, context
);
4562 /* This forces all local boolean constants to 1 to make them stateblock independent */
4563 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4565 for(i
= 0; i
< MAX_CONST_B
; i
++)
4567 if (state
->ps_consts_b
[i
])
4568 args
->bools
|= ( 1u << i
);
4571 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4572 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4573 * duplicate the shader than have a no-op KIL instruction in every shader
4575 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4576 && state
->render_states
[WINED3D_RS_CLIPPING
]
4577 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4582 /* Skip if unused or local, or supported natively */
4583 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4584 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4586 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4590 for (i
= 0; i
< MAX_CONST_I
; ++i
)
4592 if (int_skip
& (1u << i
))
4594 args
->loop_ctrl
[i
][0] = 0;
4595 args
->loop_ctrl
[i
][1] = 0;
4596 args
->loop_ctrl
[i
][2] = 0;
4600 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
* 4];
4601 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
* 4 + 1];
4602 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
* 4 + 2];
4607 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4608 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4609 struct arb_vs_compile_args
*args
)
4611 const struct wined3d_device
*device
= shader
->device
;
4612 const struct wined3d_adapter
*adapter
= device
->adapter
;
4613 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4614 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4618 find_vs_compile_args(state
, shader
, context
->stream_info
.swizzle_map
, &args
->super
, d3d_info
);
4620 args
->clip
.boolclip_compare
= 0;
4623 const struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4624 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4625 args
->ps_signature
= shader_priv
->input_signature_idx
;
4627 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4631 args
->ps_signature
= ~0;
4632 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4633 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(context
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4634 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4637 if (args
->clip
.boolclip
.clip_texcoord
)
4639 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4640 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4641 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4644 /* This forces all local boolean constants to 1 to make them stateblock independent */
4645 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4646 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4647 for(i
= 0; i
< MAX_CONST_B
; i
++)
4649 if (state
->vs_consts_b
[i
])
4650 args
->clip
.boolclip
.bools
|= (1u << i
);
4653 args
->vertex
.samplers
[0] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 0];
4654 args
->vertex
.samplers
[1] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 1];
4655 args
->vertex
.samplers
[2] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 2];
4656 args
->vertex
.samplers
[3] = 0;
4658 /* Skip if unused or local */
4659 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4660 /* This is about flow control, not clipping. */
4661 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4663 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4667 for(i
= 0; i
< MAX_CONST_I
; i
++)
4669 if (int_skip
& (1u << i
))
4671 args
->loop_ctrl
[i
][0] = 0;
4672 args
->loop_ctrl
[i
][1] = 0;
4673 args
->loop_ctrl
[i
][2] = 0;
4677 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
* 4];
4678 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
* 4 + 1];
4679 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
* 4 + 2];
4684 /* Context activation is done by the caller. */
4685 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
4686 const struct wined3d_state
*state
)
4688 struct shader_arb_priv
*priv
= shader_priv
;
4689 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4692 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4695 struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4696 struct arb_ps_compile_args compile_args
;
4697 struct arb_ps_compiled_shader
*compiled
;
4699 TRACE("Using pixel shader %p.\n", ps
);
4700 find_arb_ps_compile_args(state
, context
, ps
, &compile_args
);
4701 compiled
= find_arb_pshader(ps
, &compile_args
);
4702 priv
->current_fprogram_id
= compiled
->prgId
;
4703 priv
->compiled_fprog
= compiled
;
4705 /* Bind the fragment program */
4706 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4707 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4709 if (!priv
->use_arbfp_fixed_func
)
4710 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4712 /* Enable OpenGL fragment programs. */
4713 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4714 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4716 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", priv
->current_fprogram_id
);
4718 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4719 * a 1.x and newer shader, reload the first 8 constants
4721 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4723 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4724 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4725 for(i
= 0; i
< 8; i
++)
4727 priv
->pshader_const_dirty
[i
] = 1;
4729 /* Also takes care of loading local constants */
4730 shader_arb_load_constants_internal(shader_priv
, context
, state
, TRUE
, FALSE
, TRUE
);
4734 UINT rt_height
= state
->fb
->render_targets
[0]->height
;
4735 shader_arb_ps_local_constants(compiled
, context
, state
, rt_height
);
4738 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4739 if (compiled
->np2fixup_info
.super
.active
)
4740 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_NP2_FIXUP
;
4742 if (ps
->load_local_constsF
)
4743 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
4747 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4749 /* Disable only if we're not using arbfp fixed function fragment
4750 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4751 * enabled, and the fixed function pipeline will bind the fixed
4752 * function replacement shader. */
4753 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4754 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4755 priv
->current_fprogram_id
= 0;
4757 priv
->fragment_pipe
->enable_extension(gl_info
, TRUE
);
4762 struct wined3d_shader
*vs
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
4763 struct arb_vs_compile_args compile_args
;
4764 struct arb_vs_compiled_shader
*compiled
;
4765 const struct wined3d_shader_signature
*ps_input_sig
;
4767 TRACE("Using vertex shader %p\n", vs
);
4768 find_arb_vs_compile_args(state
, context
, vs
, &compile_args
);
4770 /* Instead of searching for the signature in the signature list, read the one from the
4771 * current pixel shader. It's maybe not the shader where the signature came from, but it
4772 * is the same signature and faster to find. */
4773 if (compile_args
.ps_signature
== ~0U)
4774 ps_input_sig
= NULL
;
4776 ps_input_sig
= &state
->shader
[WINED3D_SHADER_TYPE_PIXEL
]->input_signature
;
4778 compiled
= find_arb_vshader(vs
, context
->gl_info
, context
->stream_info
.use_map
,
4779 &compile_args
, ps_input_sig
);
4780 priv
->current_vprogram_id
= compiled
->prgId
;
4781 priv
->compiled_vprog
= compiled
;
4783 /* Bind the vertex program */
4784 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4785 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4787 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4789 /* Enable OpenGL vertex programs */
4790 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4791 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4792 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", priv
->current_vprogram_id
);
4793 shader_arb_vs_local_constants(compiled
, context
, state
);
4795 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4796 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4798 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4800 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4801 checkGLcall("glClampColorARB");
4803 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4807 if (vs
->load_local_constsF
)
4808 context
->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
4812 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4814 priv
->current_vprogram_id
= 0;
4815 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4816 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4818 priv
->vertex_pipe
->vp_enable(gl_info
, TRUE
);
4823 /* Context activation is done by the caller. */
4824 static void shader_arb_disable(void *shader_priv
, struct wined3d_context
*context
)
4826 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4827 struct shader_arb_priv
*priv
= shader_priv
;
4829 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4831 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4832 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4833 priv
->current_fprogram_id
= 0;
4835 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4837 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4839 priv
->current_vprogram_id
= 0;
4840 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4841 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4843 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4845 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4847 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4848 checkGLcall("glClampColorARB");
4849 priv
->last_vs_color_unclamp
= FALSE
;
4852 context
->shader_update_mask
= (1u << WINED3D_SHADER_TYPE_PIXEL
)
4853 | (1u << WINED3D_SHADER_TYPE_VERTEX
)
4854 | (1u << WINED3D_SHADER_TYPE_GEOMETRY
)
4855 | (1u << WINED3D_SHADER_TYPE_HULL
)
4856 | (1u << WINED3D_SHADER_TYPE_DOMAIN
);
4859 /* Context activation is done by the caller. */
4860 static void shader_arb_select_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
,
4861 enum wined3d_gl_resource_type tex_type
, const SIZE
*ds_mask_size
)
4863 const float mask
[] = {0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
};
4864 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4865 struct shader_arb_priv
*priv
= shader_priv
;
4866 GLuint
*blt_fprogram
;
4868 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
4869 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
4870 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4872 blt_fprogram
= masked
? &priv
->depth_blt_fprogram_id_masked
[tex_type
] : &priv
->depth_blt_fprogram_id_full
[tex_type
];
4873 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
, masked
);
4874 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
4875 if (masked
) GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, mask
));
4876 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4879 /* Context activation is done by the caller. */
4880 static void shader_arb_deselect_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
)
4882 struct shader_arb_priv
*priv
= shader_priv
;
4884 if (priv
->current_vprogram_id
) {
4885 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4886 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
4888 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB.\n", priv
->current_vprogram_id
);
4892 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4893 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4896 if (priv
->current_fprogram_id
) {
4897 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4898 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
4900 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB.\n", priv
->current_fprogram_id
);
4902 else if(!priv
->use_arbfp_fixed_func
)
4904 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4905 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4909 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4911 struct wined3d_device
*device
= shader
->device
;
4912 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4914 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4916 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4919 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4921 if (shader_data
->num_gl_shaders
)
4923 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4925 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4927 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4928 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4931 context_release(context
);
4934 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4935 HeapFree(GetProcessHeap(), 0, shader_data
);
4936 shader
->backend_data
= NULL
;
4940 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4943 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4945 if (shader_data
->num_gl_shaders
)
4947 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4949 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4951 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4952 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4955 context_release(context
);
4958 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4959 HeapFree(GetProcessHeap(), 0, shader_data
);
4960 shader
->backend_data
= NULL
;
4964 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4966 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4967 return compare_sig(key
, &e
->sig
);
4970 static const struct wine_rb_functions sig_tree_functions
=
4978 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4979 const struct fragment_pipeline
*fragment_pipe
)
4981 struct shader_arb_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
));
4982 struct fragment_caps fragment_caps
;
4983 void *vertex_priv
, *fragment_priv
;
4984 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4986 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4988 ERR("Failed to initialize vertex pipe.\n");
4989 HeapFree(GetProcessHeap(), 0, priv
);
4993 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4995 ERR("Failed to initialize fragment pipe.\n");
4996 vertex_pipe
->vp_free(device
);
4997 HeapFree(GetProcessHeap(), 0, priv
);
5001 priv
->vshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
5002 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
5003 if (!priv
->vshader_const_dirty
)
5005 memset(priv
->vshader_const_dirty
, 1,
5006 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
5008 priv
->pshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
5009 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
5010 if (!priv
->pshader_const_dirty
)
5012 memset(priv
->pshader_const_dirty
, 1,
5013 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
5015 if(wine_rb_init(&priv
->signature_tree
, &sig_tree_functions
) == -1)
5017 ERR("RB tree init failed\n");
5021 priv
->vertex_pipe
= vertex_pipe
;
5022 priv
->fragment_pipe
= fragment_pipe
;
5023 fragment_pipe
->get_caps(&device
->adapter
->gl_info
, &fragment_caps
);
5024 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
5026 device
->vertex_priv
= vertex_priv
;
5027 device
->fragment_priv
= fragment_priv
;
5028 device
->shader_priv
= priv
;
5033 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
5034 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
5035 fragment_pipe
->free_private(device
);
5036 vertex_pipe
->vp_free(device
);
5037 HeapFree(GetProcessHeap(), 0, priv
);
5038 return E_OUTOFMEMORY
;
5041 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
5043 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
5046 for (i
= 0; i
< sig
->sig
.element_count
; ++i
)
5048 HeapFree(GetProcessHeap(), 0, (char *)sig
->sig
.elements
[i
].semantic_name
);
5050 HeapFree(GetProcessHeap(), 0, sig
->sig
.elements
);
5051 HeapFree(GetProcessHeap(), 0, sig
);
5054 /* Context activation is done by the caller. */
5055 static void shader_arb_free(struct wined3d_device
*device
)
5057 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
5058 struct shader_arb_priv
*priv
= device
->shader_priv
;
5061 if (priv
->depth_blt_vprogram_id
)
5062 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
5064 for (i
= 0; i
< WINED3D_GL_RES_TYPE_COUNT
; ++i
)
5066 if (priv
->depth_blt_fprogram_id_full
[i
])
5068 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_full
[i
]));
5070 if (priv
->depth_blt_fprogram_id_masked
[i
])
5072 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_masked
[i
]));
5076 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
5077 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
5078 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
5079 priv
->fragment_pipe
->free_private(device
);
5080 priv
->vertex_pipe
->vp_free(device
);
5081 HeapFree(GetProcessHeap(), 0, device
->shader_priv
);
5084 static BOOL
shader_arb_allocate_context_data(struct wined3d_context
*context
)
5089 static void shader_arb_free_context_data(struct wined3d_context
*context
)
5091 struct shader_arb_priv
*priv
;
5093 if (!context
->swapchain
)
5096 priv
= context
->swapchain
->device
->shader_priv
;
5097 if (priv
->last_context
== context
)
5098 priv
->last_context
= NULL
;
5101 static void shader_arb_init_context_state(struct wined3d_context
*context
) {}
5103 static void shader_arb_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
5105 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
5110 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
5111 * for vertex programs. If the native limit is less than that it's
5112 * not very useful, and e.g. Mesa swrast returns 0, probably to
5113 * indicate it's a software implementation. */
5114 if (gl_info
->limits
.arb_vs_native_constants
< 96)
5115 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
5117 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
5119 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
5122 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
5124 else if (vs_consts
>= 256)
5126 /* Shader Model 2.0 requires at least 256 vertex shader constants */
5128 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
5133 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
5135 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
5136 caps
->vs_uniform_count
= vs_consts
;
5140 caps
->vs_version
= 0;
5141 caps
->vs_uniform_count
= 0;
5144 caps
->hs_version
= 0;
5145 caps
->ds_version
= 0;
5146 caps
->gs_version
= 0;
5148 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
5153 /* Similar as above for vertex programs, but the minimum for fragment
5154 * programs is 24. */
5155 if (gl_info
->limits
.arb_ps_native_constants
< 24)
5156 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
5158 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
5160 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
5163 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
5165 else if (ps_consts
>= 32)
5167 /* Shader Model 2.0 requires at least 32 pixel shader constants */
5169 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
5174 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
5176 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
5177 caps
->ps_uniform_count
= ps_consts
;
5178 caps
->ps_1x_max_value
= 8.0f
;
5182 caps
->ps_version
= 0;
5183 caps
->ps_uniform_count
= 0;
5184 caps
->ps_1x_max_value
= 0.0f
;
5187 caps
->varying_count
= 0;
5188 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
5189 if (use_nv_clip(gl_info
))
5190 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
5193 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
5195 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
5197 TRACE("Checking support for color_fixup:\n");
5198 dump_color_fixup_desc(fixup
);
5201 /* We support everything except complex conversions. */
5202 if (!is_complex_fixup(fixup
))
5208 TRACE("[FAILED]\n");
5212 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
5214 char write_mask
[20], regstr
[50];
5215 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5216 BOOL is_color
= FALSE
;
5217 const struct wined3d_shader_dst_param
*dst
;
5219 if (!ins
->dst_count
) return;
5223 if (!shift
) return; /* Saturate alone is handled by the instructions */
5225 shader_arb_get_write_mask(ins
, dst
, write_mask
);
5226 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
5228 /* Generate a line that does the output modifier computation
5229 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
5230 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
5232 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
5233 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
5236 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
5238 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
5239 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
5240 /* WINED3DSIH_AND */ NULL
,
5241 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5242 /* WINED3DSIH_BREAK */ shader_hw_break
,
5243 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5244 /* WINED3DSIH_BREAKP */ NULL
,
5245 /* WINED3DSIH_CALL */ shader_hw_call
,
5246 /* WINED3DSIH_CALLNZ */ NULL
,
5247 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5248 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5249 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5250 /* WINED3DSIH_CUT */ NULL
,
5251 /* WINED3DSIH_DCL */ shader_hw_nop
,
5252 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5253 /* WINED3DSIH_DCL_GLOBAL_FLAGS */ NULL
,
5254 /* WINED3DSIH_DCL_HS_FORK_PHASE_INSTANCE_COUNT */ NULL
,
5255 /* WINED3DSIH_DCL_HS_MAX_TESSFACTOR */ NULL
,
5256 /* WINED3DSIH_DCL_IMMEDIATE_CONSTANT_BUFFER */ NULL
,
5257 /* WINED3DSIH_DCL_INPUT */ NULL
,
5258 /* WINED3DSIH_DCL_INPUT_CONTROL_POINT_COUNT */ NULL
,
5259 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5260 /* WINED3DSIH_DCL_INPUT_PS */ NULL
,
5261 /* WINED3DSIH_DCL_INPUT_PS_SGV */ NULL
,
5262 /* WINED3DSIH_DCL_INPUT_PS_SIV */ NULL
,
5263 /* WINED3DSIH_DCL_INPUT_SGV */ NULL
,
5264 /* WINED3DSIH_DCL_INPUT_SIV */ NULL
,
5265 /* WINED3DSIH_DCL_OUTPUT */ NULL
,
5266 /* WINED3DSIH_DCL_OUTPUT_CONTROL_POINT_COUNT */ NULL
,
5267 /* WINED3DSIH_DCL_OUTPUT_SIV */ NULL
,
5268 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5269 /* WINED3DSIH_DCL_RESOURCE_STRUCTURED */ NULL
,
5270 /* WINED3DSIH_DCL_SAMPLER */ NULL
,
5271 /* WINED3DSIH_DCL_TEMPS */ NULL
,
5272 /* WINED3DSIH_DCL_TESSELLATOR_DOMAIN */ NULL
,
5273 /* WINED3DSIH_DCL_TESSELLATOR_OUTPUT_PRIMITIVE */ NULL
,
5274 /* WINED3DSIH_DCL_TESSELLATOR_PARTITIONING */ NULL
,
5275 /* WINED3DSIH_DCL_UAV_TYPED */ NULL
,
5276 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5277 /* WINED3DSIH_DEF */ shader_hw_nop
,
5278 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5279 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5280 /* WINED3DSIH_DIV */ NULL
,
5281 /* WINED3DSIH_DP2 */ NULL
,
5282 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5283 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5284 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5285 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5286 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5287 /* WINED3DSIH_DSX_COARSE */ NULL
,
5288 /* WINED3DSIH_DSX_FINE */ NULL
,
5289 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5290 /* WINED3DSIH_DSY_COARSE */ NULL
,
5291 /* WINED3DSIH_DSY_FINE */ NULL
,
5292 /* WINED3DSIH_ELSE */ shader_hw_else
,
5293 /* WINED3DSIH_EMIT */ NULL
,
5294 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5295 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5296 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5297 /* WINED3DSIH_EQ */ NULL
,
5298 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5299 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5300 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5301 /* WINED3DSIH_FTOI */ NULL
,
5302 /* WINED3DSIH_FTOU */ NULL
,
5303 /* WINED3DSIH_GE */ NULL
,
5304 /* WINED3DSIH_HS_CONTROL_POINT_PHASE */ NULL
,
5305 /* WINED3DSIH_HS_DECLS */ NULL
,
5306 /* WINED3DSIH_HS_FORK_PHASE */ NULL
,
5307 /* WINED3DSIH_HS_JOIN_PHASE */ NULL
,
5308 /* WINED3DSIH_IADD */ NULL
,
5309 /* WINED3DSIH_IEQ */ NULL
,
5310 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5311 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5312 /* WINED3DSIH_IGE */ NULL
,
5313 /* WINED3DSIH_ILT */ NULL
,
5314 /* WINED3DSIH_IMAD */ NULL
,
5315 /* WINED3DSIH_IMAX */ NULL
,
5316 /* WINED3DSIH_IMIN */ NULL
,
5317 /* WINED3DSIH_IMUL */ NULL
,
5318 /* WINED3DSIH_INE */ NULL
,
5319 /* WINED3DSIH_INEG */ NULL
,
5320 /* WINED3DSIH_ISHL */ NULL
,
5321 /* WINED3DSIH_ITOF */ NULL
,
5322 /* WINED3DSIH_LABEL */ shader_hw_label
,
5323 /* WINED3DSIH_LD */ NULL
,
5324 /* WINED3DSIH_LD2DMS */ NULL
,
5325 /* WINED3DSIH_LD_STRUCTURED */ NULL
,
5326 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5327 /* WINED3DSIH_LOG */ shader_hw_scalar_op
,
5328 /* WINED3DSIH_LOGP */ shader_hw_scalar_op
,
5329 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5330 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5331 /* WINED3DSIH_LT */ NULL
,
5332 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5333 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5334 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5335 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5336 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5337 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5338 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5339 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5340 /* WINED3DSIH_MOV */ shader_hw_mov
,
5341 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5342 /* WINED3DSIH_MOVC */ NULL
,
5343 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5344 /* WINED3DSIH_NE */ NULL
,
5345 /* WINED3DSIH_NOP */ shader_hw_nop
,
5346 /* WINED3DSIH_NOT */ NULL
,
5347 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5348 /* WINED3DSIH_OR */ NULL
,
5349 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5350 /* WINED3DSIH_POW */ shader_hw_pow
,
5351 /* WINED3DSIH_RCP */ shader_hw_scalar_op
,
5352 /* WINED3DSIH_REP */ shader_hw_rep
,
5353 /* WINED3DSIH_RESINFO */ NULL
,
5354 /* WINED3DSIH_RET */ shader_hw_ret
,
5355 /* WINED3DSIH_ROUND_NI */ NULL
,
5356 /* WINED3DSIH_ROUND_PI */ NULL
,
5357 /* WINED3DSIH_ROUND_Z */ NULL
,
5358 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5359 /* WINED3DSIH_SAMPLE */ NULL
,
5360 /* WINED3DSIH_SAMPLE_B */ NULL
,
5361 /* WINED3DSIH_SAMPLE_C */ NULL
,
5362 /* WINED3DSIH_SAMPLE_C_LZ */ NULL
,
5363 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5364 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5365 /* WINED3DSIH_SETP */ NULL
,
5366 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5367 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5368 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5369 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5370 /* WINED3DSIH_SQRT */ NULL
,
5371 /* WINED3DSIH_STORE_UAV_TYPED */ NULL
,
5372 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5373 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5374 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5375 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5376 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5377 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5378 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5379 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5380 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5381 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5382 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5383 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5384 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5385 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5386 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5387 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5388 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5389 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5390 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5391 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5392 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5393 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5394 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5395 /* WINED3DSIH_UDIV */ NULL
,
5396 /* WINED3DSIH_UGE */ NULL
,
5397 /* WINED3DSIH_ULT */ NULL
,
5398 /* WINED3DSIH_USHR */ NULL
,
5399 /* WINED3DSIH_UTOF */ NULL
,
5400 /* WINED3DSIH_XOR */ NULL
,
5403 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5404 const struct wined3d_shader
*shader
, DWORD idx
)
5406 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5407 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5408 const struct wined3d_shader_lconst
*constant
;
5410 WORD flag
= (1u << idx
);
5411 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5413 if (reg_maps
->local_bool_consts
& flag
)
5415 /* What good is an if(bool) with a hardcoded local constant? I don't know, but handle it */
5416 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5418 if (constant
->idx
== idx
)
5420 return constant
->value
[0];
5423 ERR("Local constant not found\n");
5428 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5429 else bools
= priv
->cur_ps_args
->bools
;
5430 return bools
& flag
;
5434 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5435 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5437 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5438 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5440 /* Integer constants can either be a local constant, or they can be stored in the shader
5441 * type specific compile args. */
5442 if (reg_maps
->local_int_consts
& (1u << idx
))
5444 const struct wined3d_shader_lconst
*constant
;
5446 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5448 if (constant
->idx
== idx
)
5450 loop_control
->count
= constant
->value
[0];
5451 loop_control
->start
= constant
->value
[1];
5452 /* Step is signed. */
5453 loop_control
->step
= (int)constant
->value
[2];
5457 /* If this happens the flag was set incorrectly */
5458 ERR("Local constant not found\n");
5459 loop_control
->count
= 0;
5460 loop_control
->start
= 0;
5461 loop_control
->step
= 0;
5465 switch (reg_maps
->shader_version
.type
)
5467 case WINED3D_SHADER_TYPE_VERTEX
:
5468 /* Count and aL start value are unsigned */
5469 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5470 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5471 /* Step is signed. */
5472 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5475 case WINED3D_SHADER_TYPE_PIXEL
:
5476 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5477 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5478 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5482 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5487 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5490 struct wined3d_shader_dst_param
*dst_param
;
5491 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
;
5492 struct recorded_instruction
*rec
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*rec
));
5495 ERR("Out of memory\n");
5500 dst_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
));
5501 if(!dst_param
) goto free
;
5502 *dst_param
= *ins
->dst
;
5503 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5505 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5508 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5509 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5511 rec
->ins
.dst
= dst_param
;
5513 src_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*src_param
) * ins
->src_count
);
5516 for (i
= 0; i
< ins
->src_count
; ++i
)
5518 src_param
[i
] = ins
->src
[i
];
5519 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5521 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5524 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5525 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5528 rec
->ins
.src
= src_param
;
5529 list_add_tail(list
, &rec
->entry
);
5533 ERR("Out of memory\n");
5536 HeapFree(GetProcessHeap(), 0, (void *)dst_param
->reg
.idx
[0].rel_addr
);
5537 HeapFree(GetProcessHeap(), 0, dst_param
);
5541 for(i
= 0; i
< ins
->src_count
; i
++)
5543 HeapFree(GetProcessHeap(), 0, (void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5545 HeapFree(GetProcessHeap(), 0, src_param
);
5547 HeapFree(GetProcessHeap(), 0, rec
);
5550 static void free_recorded_instruction(struct list
*list
)
5552 struct recorded_instruction
*rec_ins
, *entry2
;
5555 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5557 list_remove(&rec_ins
->entry
);
5558 if (rec_ins
->ins
.dst
)
5560 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5561 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
);
5563 if (rec_ins
->ins
.src
)
5565 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5567 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5569 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
);
5571 HeapFree(GetProcessHeap(), 0, rec_ins
);
5575 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5577 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5578 struct control_frame
*control_frame
;
5580 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5582 struct list
*e
= list_head(&priv
->control_frames
);
5583 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5584 list_remove(&control_frame
->entry
);
5585 HeapFree(GetProcessHeap(), 0, control_frame
);
5588 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5590 /* Non-ifc ENDIFs were already handled previously. */
5591 struct list
*e
= list_head(&priv
->control_frames
);
5592 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5593 list_remove(&control_frame
->entry
);
5594 HeapFree(GetProcessHeap(), 0, control_frame
);
5598 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5599 SHADER_HANDLER hw_fct
;
5600 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5601 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5602 struct control_frame
*control_frame
;
5603 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5606 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5608 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5609 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5611 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5612 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5614 if(priv
->target_version
>= NV2
)
5616 control_frame
->no
.loop
= priv
->num_loops
++;
5621 /* Don't bother recording when we're in a not used if branch */
5627 if(!priv
->recording
)
5629 list_init(&priv
->record
);
5630 priv
->recording
= TRUE
;
5631 control_frame
->outer_loop
= TRUE
;
5632 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5633 return; /* Instruction is handled */
5635 /* Record this loop in the outer loop's recording */
5638 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5640 if(priv
->target_version
>= NV2
)
5642 /* Nothing to do. The control frame is popped after the HW instr handler */
5646 struct list
*e
= list_head(&priv
->control_frames
);
5647 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5648 list_remove(&control_frame
->entry
);
5650 if(control_frame
->outer_loop
)
5652 unsigned int iteration
;
5656 /* Turn off recording before playback */
5657 priv
->recording
= FALSE
;
5659 /* Move the recorded instructions to a separate list and get them out of the private data
5660 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5661 * be recorded again, thus priv->record might be overwritten
5664 list_move_tail(©
, &priv
->record
);
5665 list_init(&priv
->record
);
5667 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5669 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5670 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5671 control_frame
->loop_control
.step
);
5672 aL
= control_frame
->loop_control
.start
;
5676 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5679 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5681 struct recorded_instruction
*rec_ins
;
5682 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5685 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5689 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5692 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5694 shader_arb_handle_instruction(&rec_ins
->ins
);
5697 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5699 aL
+= control_frame
->loop_control
.step
;
5702 shader_addline(buffer
, "#end loop/rep\n");
5704 free_recorded_instruction(©
);
5705 HeapFree(GetProcessHeap(), 0, control_frame
);
5706 return; /* Instruction is handled */
5710 /* This is a nested loop. Proceed to the normal recording function */
5711 HeapFree(GetProcessHeap(), 0, control_frame
);
5718 record_instruction(&priv
->record
, ins
);
5723 if(ins
->handler_idx
== WINED3DSIH_IF
)
5725 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5726 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5727 control_frame
->type
= IF
;
5729 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5730 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5731 bool_const
= !bool_const
;
5732 if (!priv
->muted
&& !bool_const
)
5734 shader_addline(buffer
, "#if(FALSE){\n");
5736 control_frame
->muting
= TRUE
;
5738 else shader_addline(buffer
, "#if(TRUE) {\n");
5740 return; /* Instruction is handled */
5742 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5744 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5745 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5746 control_frame
->type
= IFC
;
5747 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5748 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5750 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5752 struct list
*e
= list_head(&priv
->control_frames
);
5753 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5755 if(control_frame
->type
== IF
)
5757 shader_addline(buffer
, "#} else {\n");
5758 if(!priv
->muted
&& !control_frame
->muting
)
5761 control_frame
->muting
= TRUE
;
5763 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5764 return; /* Instruction is handled. */
5766 /* In case of an ifc, generate a HW shader instruction */
5767 if (control_frame
->type
!= IFC
)
5768 ERR("Control frame does not match.\n");
5770 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5772 struct list
*e
= list_head(&priv
->control_frames
);
5773 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5775 if(control_frame
->type
== IF
)
5777 shader_addline(buffer
, "#} endif\n");
5778 if(control_frame
->muting
) priv
->muted
= FALSE
;
5779 list_remove(&control_frame
->entry
);
5780 HeapFree(GetProcessHeap(), 0, control_frame
);
5781 return; /* Instruction is handled */
5783 /* In case of an ifc, generate a HW shader instruction */
5784 if (control_frame
->type
!= IFC
)
5785 ERR("Control frame does not match.\n");
5790 pop_control_frame(ins
);
5794 /* Select handler */
5795 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5797 /* Unhandled opcode */
5800 FIXME("Backend can't handle opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
5805 pop_control_frame(ins
);
5807 shader_arb_add_instruction_modifiers(ins
);
5810 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5812 struct shader_arb_priv
*priv
= shader_priv
;
5814 return priv
->ffp_proj_control
;
5817 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5819 shader_arb_handle_instruction
,
5822 shader_arb_select_depth_blt
,
5823 shader_arb_deselect_depth_blt
,
5824 shader_arb_update_float_vertex_constants
,
5825 shader_arb_update_float_pixel_constants
,
5826 shader_arb_load_constants
,
5830 shader_arb_allocate_context_data
,
5831 shader_arb_free_context_data
,
5832 shader_arb_init_context_state
,
5833 shader_arb_get_caps
,
5834 shader_arb_color_fixup_supported
,
5835 shader_arb_has_ffp_proj_control
,
5838 /* ARB_fragment_program fixed function pipeline replacement definitions */
5839 #define ARB_FFP_CONST_TFACTOR 0
5840 #define ARB_FFP_CONST_COLOR_KEY_LOW ((ARB_FFP_CONST_TFACTOR) + 1)
5841 #define ARB_FFP_CONST_COLOR_KEY_HIGH ((ARB_FFP_CONST_COLOR_KEY_LOW) + 1)
5842 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_COLOR_KEY_HIGH) + 1)
5843 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5844 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5845 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5847 struct arbfp_ffp_desc
5849 struct ffp_frag_desc parent
;
5853 /* Context activation is done by the caller. */
5854 static void arbfp_enable(const struct wined3d_gl_info
*gl_info
, BOOL enable
)
5858 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5859 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5863 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5864 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5868 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5870 struct shader_arb_priv
*priv
;
5872 /* Share private data between the shader backend and the pipeline
5873 * replacement, if both are the arb implementation. This is needed to
5874 * figure out whether ARBfp should be disabled if no pixel shader is bound
5876 if (shader_backend
== &arb_program_shader_backend
)
5878 else if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
5881 if (wine_rb_init(&priv
->fragment_shaders
, &wined3d_ffp_frag_program_rb_functions
) == -1)
5883 ERR("Failed to initialize rbtree.\n");
5884 if (priv
!= shader_priv
)
5885 HeapFree(GetProcessHeap(), 0, priv
);
5888 priv
->use_arbfp_fixed_func
= TRUE
;
5893 /* Context activation is done by the caller. */
5894 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5896 const struct wined3d_gl_info
*gl_info
= context
;
5897 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5899 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5900 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5901 HeapFree(GetProcessHeap(), 0, entry_arb
);
5904 /* Context activation is done by the caller. */
5905 static void arbfp_free(struct wined3d_device
*device
)
5907 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5909 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &device
->adapter
->gl_info
);
5910 priv
->use_arbfp_fixed_func
= FALSE
;
5912 if (device
->shader_backend
!= &arb_program_shader_backend
)
5914 HeapFree(GetProcessHeap(), 0, device
->fragment_priv
);
5918 static void arbfp_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5920 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5921 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
5922 | WINED3D_FRAGMENT_CAP_COLOR_KEY
;
5923 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5924 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5925 WINED3DTEXOPCAPS_SELECTARG1
|
5926 WINED3DTEXOPCAPS_SELECTARG2
|
5927 WINED3DTEXOPCAPS_MODULATE4X
|
5928 WINED3DTEXOPCAPS_MODULATE2X
|
5929 WINED3DTEXOPCAPS_MODULATE
|
5930 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5931 WINED3DTEXOPCAPS_ADDSIGNED
|
5932 WINED3DTEXOPCAPS_ADD
|
5933 WINED3DTEXOPCAPS_SUBTRACT
|
5934 WINED3DTEXOPCAPS_ADDSMOOTH
|
5935 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5936 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5937 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5938 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5939 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5940 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5941 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5942 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5943 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5944 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5945 WINED3DTEXOPCAPS_MULTIPLYADD
|
5946 WINED3DTEXOPCAPS_LERP
|
5947 WINED3DTEXOPCAPS_BUMPENVMAP
|
5948 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5950 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5952 caps
->MaxTextureBlendStages
= 8;
5953 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.fragment_samplers
, 8);
5956 static DWORD
arbfp_get_emul_mask(const struct wined3d_gl_info
*gl_info
)
5958 return GL_EXT_EMUL_ARB_MULTITEXTURE
| GL_EXT_EMUL_EXT_FOG_COORD
;
5961 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5962 const struct wined3d_state
*state
, DWORD state_id
)
5964 struct wined3d_device
*device
= context
->swapchain
->device
;
5965 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5966 struct wined3d_color color
;
5968 /* Don't load the parameter if we're using an arbfp pixel shader,
5969 * otherwise we'll overwrite application provided constants. */
5970 if (device
->shader_backend
== &arb_program_shader_backend
)
5972 struct shader_arb_priv
*priv
;
5974 if (use_ps(state
)) return;
5976 priv
= device
->shader_priv
;
5977 priv
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5978 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5981 wined3d_color_from_d3dcolor(&color
, state
->render_states
[WINED3D_RS_TEXTUREFACTOR
]);
5982 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, &color
.r
));
5983 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, &color.r)");
5986 static void state_arb_specularenable(struct wined3d_context
*context
,
5987 const struct wined3d_state
*state
, DWORD state_id
)
5989 struct wined3d_device
*device
= context
->swapchain
->device
;
5990 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5993 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5994 * application provided constants
5996 if (device
->shader_backend
== &arb_program_shader_backend
)
5998 struct shader_arb_priv
*priv
;
6000 if (use_ps(state
)) return;
6002 priv
= device
->shader_priv
;
6003 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
6004 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
6007 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
6009 /* The specular color has no alpha */
6010 col
[0] = 1.0f
; col
[1] = 1.0f
;
6011 col
[2] = 1.0f
; col
[3] = 0.0f
;
6013 col
[0] = 0.0f
; col
[1] = 0.0f
;
6014 col
[2] = 0.0f
; col
[3] = 0.0f
;
6016 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
6017 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
6020 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6022 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
6023 struct wined3d_device
*device
= context
->swapchain
->device
;
6024 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6027 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
6029 if (device
->shader_backend
== &arb_program_shader_backend
)
6031 struct shader_arb_priv
*priv
= device
->shader_priv
;
6033 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
6037 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
6038 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
6041 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
6042 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
6043 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
6044 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
6046 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
6047 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
6050 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
6051 const struct wined3d_state
*state
, DWORD state_id
)
6053 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
6054 struct wined3d_device
*device
= context
->swapchain
->device
;
6055 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6058 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
6060 if (device
->shader_backend
== &arb_program_shader_backend
)
6062 struct shader_arb_priv
*priv
= device
->shader_priv
;
6064 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
6068 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
6069 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
6072 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
6073 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
6077 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
6078 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
6081 static void alpha_test_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6083 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6087 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6089 if (state
->render_states
[WINED3D_RS_ALPHATESTENABLE
])
6091 gl_info
->gl_ops
.gl
.p_glEnable(GL_ALPHA_TEST
);
6092 checkGLcall("glEnable GL_ALPHA_TEST");
6096 gl_info
->gl_ops
.gl
.p_glDisable(GL_ALPHA_TEST
);
6097 checkGLcall("glDisable GL_ALPHA_TEST");
6101 ref
= ((float)state
->render_states
[WINED3D_RS_ALPHAREF
]) / 255.0f
;
6102 glParm
= wined3d_gl_compare_func(state
->render_states
[WINED3D_RS_ALPHAFUNC
]);
6106 gl_info
->gl_ops
.gl
.p_glAlphaFunc(glParm
, ref
);
6107 checkGLcall("glAlphaFunc");
6111 static void color_key_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6113 struct wined3d_device
*device
= context
->swapchain
->device
;
6114 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6115 struct wined3d_color float_key
[2];
6116 const struct wined3d_texture
*texture
= state
->textures
[0];
6121 /* Don't load the parameter if we're using an arbfp pixel shader,
6122 * otherwise we'll overwrite application provided constants. */
6123 if (device
->shader_backend
== &arb_program_shader_backend
)
6125 struct shader_arb_priv
*priv
;
6130 priv
= device
->shader_priv
;
6131 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_LOW
] = 1;
6132 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_HIGH
] = 1;
6133 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_COLOR_KEY_HIGH
+ 1);
6136 wined3d_format_get_float_color_key(texture
->resource
.format
, &texture
->async
.src_blt_color_key
, float_key
);
6138 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_LOW
, &float_key
[0].r
));
6139 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_LOW, &float_key[0].r)");
6140 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_HIGH
, &float_key
[1].r
));
6141 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_HIGH, &float_key[1].r)");
6144 static const char *get_argreg(struct wined3d_string_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
6148 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
6150 switch(arg
& WINED3DTA_SELECTMASK
) {
6151 case WINED3DTA_DIFFUSE
:
6152 ret
= "fragment.color.primary"; break;
6154 case WINED3DTA_CURRENT
:
6158 case WINED3DTA_TEXTURE
:
6160 case 0: ret
= "tex0"; break;
6161 case 1: ret
= "tex1"; break;
6162 case 2: ret
= "tex2"; break;
6163 case 3: ret
= "tex3"; break;
6164 case 4: ret
= "tex4"; break;
6165 case 5: ret
= "tex5"; break;
6166 case 6: ret
= "tex6"; break;
6167 case 7: ret
= "tex7"; break;
6168 default: ret
= "unknown texture";
6172 case WINED3DTA_TFACTOR
:
6173 ret
= "tfactor"; break;
6175 case WINED3DTA_SPECULAR
:
6176 ret
= "fragment.color.secondary"; break;
6178 case WINED3DTA_TEMP
:
6179 ret
= "tempreg"; break;
6181 case WINED3DTA_CONSTANT
:
6182 FIXME("Implement perstage constants\n");
6184 case 0: ret
= "const0"; break;
6185 case 1: ret
= "const1"; break;
6186 case 2: ret
= "const2"; break;
6187 case 3: ret
= "const3"; break;
6188 case 4: ret
= "const4"; break;
6189 case 5: ret
= "const5"; break;
6190 case 6: ret
= "const6"; break;
6191 case 7: ret
= "const7"; break;
6192 default: ret
= "unknown constant";
6200 if(arg
& WINED3DTA_COMPLEMENT
) {
6201 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
6202 if(argnum
== 0) ret
= "arg0";
6203 if(argnum
== 1) ret
= "arg1";
6204 if(argnum
== 2) ret
= "arg2";
6206 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
6207 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
6208 if(argnum
== 0) ret
= "arg0";
6209 if(argnum
== 1) ret
= "arg1";
6210 if(argnum
== 2) ret
= "arg2";
6215 static void gen_ffp_instr(struct wined3d_string_buffer
*buffer
, unsigned int stage
, BOOL color
,
6216 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
6218 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6219 unsigned int mul
= 1;
6221 if(color
&& alpha
) dstmask
= "";
6222 else if(color
) dstmask
= ".xyz";
6223 else dstmask
= ".w";
6225 if(dst
== tempreg
) dstreg
= "tempreg";
6226 else dstreg
= "ret";
6228 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6229 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6230 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6234 case WINED3D_TOP_DISABLE
:
6237 case WINED3D_TOP_SELECT_ARG2
:
6240 case WINED3D_TOP_SELECT_ARG1
:
6241 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6244 case WINED3D_TOP_MODULATE_4X
:
6247 case WINED3D_TOP_MODULATE_2X
:
6250 case WINED3D_TOP_MODULATE
:
6251 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6254 case WINED3D_TOP_ADD_SIGNED_2X
:
6257 case WINED3D_TOP_ADD_SIGNED
:
6258 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6261 case WINED3D_TOP_ADD
:
6262 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6265 case WINED3D_TOP_SUBTRACT
:
6266 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6269 case WINED3D_TOP_ADD_SMOOTH
:
6270 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6271 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6274 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6275 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6276 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6278 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6279 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6280 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6282 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6283 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6284 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6286 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6287 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6288 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6291 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6292 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6293 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6294 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6297 /* D3DTOP_PREMODULATE ???? */
6299 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6300 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6301 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6303 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6304 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6306 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6307 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6308 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6310 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6311 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6314 case WINED3D_TOP_DOTPRODUCT3
:
6316 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6317 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6318 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6321 case WINED3D_TOP_MULTIPLY_ADD
:
6322 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6325 case WINED3D_TOP_LERP
:
6326 /* The msdn is not quite right here */
6327 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6330 case WINED3D_TOP_BUMPENVMAP
:
6331 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6332 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6336 FIXME("Unhandled texture op %08x\n", op
);
6340 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6342 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6345 static const char *arbfp_texture_target(enum wined3d_gl_resource_type type
)
6349 case WINED3D_GL_RES_TYPE_TEX_1D
:
6351 case WINED3D_GL_RES_TYPE_TEX_2D
:
6353 case WINED3D_GL_RES_TYPE_TEX_3D
:
6355 case WINED3D_GL_RES_TYPE_TEX_CUBE
:
6357 case WINED3D_GL_RES_TYPE_TEX_RECT
:
6360 return "unexpected_resource_type";
6364 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6367 struct wined3d_string_buffer buffer
;
6368 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6369 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6370 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6371 UINT lowest_disabled_stage
;
6372 const char *textype
;
6374 char colorcor_dst
[8];
6376 DWORD arg0
, arg1
, arg2
;
6377 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6379 BOOL custom_linear_fog
= FALSE
;
6380 struct color_fixup_masks masks
;
6382 if (!string_buffer_init(&buffer
))
6384 ERR("Failed to initialize shader buffer.\n");
6388 shader_addline(&buffer
, "!!ARBfp1.0\n");
6390 if (settings
->color_key_enabled
)
6392 shader_addline(&buffer
, "PARAM color_key_low = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_LOW
);
6393 shader_addline(&buffer
, "PARAM color_key_high = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_HIGH
);
6397 /* Find out which textures are read */
6398 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6400 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6402 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6403 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6404 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6405 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6406 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6407 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6409 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA
)
6410 tex_read
[stage
] = TRUE
;
6411 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
)
6412 tex_read
[stage
] = TRUE
;
6413 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP
)
6415 bump_used
[stage
] = TRUE
;
6416 tex_read
[stage
] = TRUE
;
6418 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6420 bump_used
[stage
] = TRUE
;
6421 tex_read
[stage
] = TRUE
;
6422 luminance_used
[stage
] = TRUE
;
6424 else if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_FACTOR_ALPHA
)
6426 tfactor_used
= TRUE
;
6429 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6430 tfactor_used
= TRUE
;
6433 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
6434 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6435 tempreg_used
= TRUE
;
6438 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6440 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6441 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6442 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6443 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6444 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6445 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6447 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6448 tempreg_used
= TRUE
;
6450 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6451 tfactor_used
= TRUE
;
6454 lowest_disabled_stage
= stage
;
6456 switch (settings
->fog
)
6458 case WINED3D_FFP_PS_FOG_OFF
: break;
6459 case WINED3D_FFP_PS_FOG_LINEAR
:
6460 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
6462 custom_linear_fog
= TRUE
;
6465 shader_addline(&buffer
, "OPTION ARB_fog_linear;\n");
6468 case WINED3D_FFP_PS_FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6469 case WINED3D_FFP_PS_FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6470 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6473 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6474 shader_addline(&buffer
, "TEMP TMP;\n");
6475 shader_addline(&buffer
, "TEMP ret;\n");
6476 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6477 shader_addline(&buffer
, "TEMP arg0;\n");
6478 shader_addline(&buffer
, "TEMP arg1;\n");
6479 shader_addline(&buffer
, "TEMP arg2;\n");
6480 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
6481 if(!tex_read
[stage
]) continue;
6482 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6483 if(!bump_used
[stage
]) continue;
6484 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6485 if(!luminance_used
[stage
]) continue;
6486 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6489 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6491 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6493 if (settings
->sRGB_write
)
6495 shader_addline(&buffer
, "PARAM srgb_consts0 = ");
6496 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const0
);
6497 shader_addline(&buffer
, ";\n");
6498 shader_addline(&buffer
, "PARAM srgb_consts1 = ");
6499 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const1
);
6500 shader_addline(&buffer
, ";\n");
6503 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6504 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6506 if (tempreg_used
|| settings
->sRGB_write
)
6507 shader_addline(&buffer
, "MOV tempreg, 0.0;\n");
6509 /* Generate texture sampling instructions */
6510 for (stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6512 if (!tex_read
[stage
])
6515 textype
= arbfp_texture_target(settings
->op
[stage
].tex_type
);
6517 if(settings
->op
[stage
].projected
== proj_none
) {
6519 } else if(settings
->op
[stage
].projected
== proj_count4
||
6520 settings
->op
[stage
].projected
== proj_count3
) {
6523 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6528 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6529 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6531 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6532 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6533 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6534 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6536 /* with projective textures, texbem only divides the static texture coord, not the displacement,
6537 * so multiply the displacement with the dividing parameter before passing it to TXP
6539 if (settings
->op
[stage
].projected
!= proj_none
) {
6540 if(settings
->op
[stage
].projected
== proj_count4
) {
6541 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6542 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
6544 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6545 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
6548 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6551 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6552 instr
, stage
, stage
, textype
);
6553 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6555 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6556 stage
- 1, stage
- 1, stage
- 1);
6557 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6559 } else if(settings
->op
[stage
].projected
== proj_count3
) {
6560 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6561 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6562 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6563 instr
, stage
, stage
, textype
);
6565 shader_addline(&buffer
, "%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6566 instr
, stage
, stage
, stage
, textype
);
6569 sprintf(colorcor_dst
, "tex%u", stage
);
6570 masks
= calc_color_correction(settings
->op
[stage
].color_fixup
, WINED3DSP_WRITEMASK_ALL
);
6571 gen_color_correction(&buffer
, colorcor_dst
, colorcor_dst
, "const.x", "const.y",
6572 settings
->op
[stage
].color_fixup
, masks
);
6575 if (settings
->color_key_enabled
)
6577 shader_addline(&buffer
, "SLT TMP, tex0, color_key_low;\n"); /* below low key */
6578 shader_addline(&buffer
, "SGE ret, tex0, color_key_high;\n"); /* above high key */
6579 shader_addline(&buffer
, "ADD TMP, TMP, ret;\n"); /* or */
6580 shader_addline(&buffer
, "DP4 TMP.b, TMP, TMP;\n"); /* on any channel */
6581 shader_addline(&buffer
, "SGE TMP, -TMP.b, 0.0;\n"); /* logical not */
6582 shader_addline(&buffer
, "KIL -TMP;\n"); /* discard if true */
6585 shader_addline(&buffer
, "MOV ret, fragment.color.primary;\n");
6587 /* Generate the main shader */
6588 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6590 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6593 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6594 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6595 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6596 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6597 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6598 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6599 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6600 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6601 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6602 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6603 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6604 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6606 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6607 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6608 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6609 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6611 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6613 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6614 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6615 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6619 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
6620 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6621 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6623 else if (settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP
6624 && settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6626 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6627 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6628 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6629 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
6630 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6631 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6635 if (settings
->sRGB_write
|| custom_linear_fog
)
6637 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, ret;\n");
6638 if (settings
->sRGB_write
)
6639 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6640 if (custom_linear_fog
)
6641 arbfp_add_linear_fog(&buffer
, "ret", "arg0");
6642 shader_addline(&buffer
, "MOV result.color, ret;\n");
6646 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, ret;\n");
6650 shader_addline(&buffer
, "END\n");
6652 /* Generate the shader */
6653 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6654 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6655 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
6657 string_buffer_free(&buffer
);
6661 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6663 const struct wined3d_device
*device
= context
->swapchain
->device
;
6664 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6665 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6666 BOOL use_pshader
= use_ps(state
);
6667 struct ffp_frag_settings settings
;
6668 const struct arbfp_ffp_desc
*desc
;
6671 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6673 if (isStateDirty(context
, STATE_RENDER(WINED3D_RS_FOGENABLE
)))
6675 if (!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6677 /* Reload fixed function constants since they collide with the
6678 * pixel shader constants. */
6679 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6681 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6683 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6684 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6685 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6687 else if (use_pshader
)
6689 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6696 /* Find or create a shader implementing the fixed function pipeline
6697 * settings, then activate it. */
6698 gen_ffp_frag_op(context
, state
, &settings
, FALSE
);
6699 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6701 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
6704 ERR("Out of memory\n");
6708 new_desc
->parent
.settings
= settings
;
6709 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6710 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6711 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6715 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6716 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6719 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6720 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6721 priv
->current_fprogram_id
= desc
->shader
;
6723 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6725 /* Reload fixed function constants since they collide with the
6726 * pixel shader constants. */
6727 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6729 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6731 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6732 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6734 context
->last_was_pshader
= FALSE
;
6736 else if (!context
->last_was_pshader
)
6738 if (device
->shader_backend
== &arb_program_shader_backend
)
6739 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
6740 context
->last_was_pshader
= TRUE
;
6743 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6746 /* We can't link the fog states to the fragment state directly since the
6747 * vertex pipeline links them to FOGENABLE. A different linking in different
6748 * pipeline parts can't be expressed in the combined state table, so we need
6749 * to handle that with a forwarding function. The other invisible side effect
6750 * is that changing the fog start and fog end (which links to FOGENABLE in
6751 * vertex) results in the fragment_prog_arbfp function being called because
6752 * FOGENABLE is dirty, which calls this function here. */
6753 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6755 enum fogsource new_source
;
6756 DWORD fogstart
= state
->render_states
[WINED3D_RS_FOGSTART
];
6757 DWORD fogend
= state
->render_states
[WINED3D_RS_FOGEND
];
6759 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6761 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6762 fragment_prog_arbfp(context
, state
, state_id
);
6764 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6767 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6771 new_source
= FOGSOURCE_VS
;
6775 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6776 new_source
= FOGSOURCE_COORD
;
6778 new_source
= FOGSOURCE_FFP
;
6783 new_source
= FOGSOURCE_FFP
;
6786 if (new_source
!= context
->fog_source
|| fogstart
== fogend
)
6788 context
->fog_source
= new_source
;
6789 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6793 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6795 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6796 fragment_prog_arbfp(context
, state
, state_id
);
6799 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] =
6801 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6802 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6803 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6804 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6805 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6806 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6807 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6808 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6809 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6810 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6811 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6812 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6813 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6814 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6815 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6816 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6817 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6818 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6819 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6820 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6821 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6822 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6823 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6824 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6825 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6826 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6827 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6828 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6829 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6830 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6831 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6832 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6833 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6834 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6835 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6836 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6837 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6838 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6839 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6840 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6841 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6842 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6843 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6844 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6845 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6846 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6847 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6848 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6849 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6850 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6851 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6852 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6853 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6854 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6855 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6856 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6857 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6858 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6859 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6860 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6861 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6862 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6863 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6864 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6865 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6866 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6867 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6868 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6869 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6870 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6871 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6872 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6873 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6874 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6875 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6876 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6877 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6878 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6879 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6880 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6881 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6882 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6883 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6884 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6885 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6886 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6887 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6888 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6889 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6890 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6891 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6892 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6893 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6894 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6895 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6896 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6897 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6898 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6899 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6900 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6901 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6902 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6903 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6904 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6905 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6906 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6907 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6908 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6909 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6910 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6911 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6912 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6913 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6914 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6915 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6916 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6917 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6918 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6919 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6920 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6921 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6922 {STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6923 {STATE_RENDER(WINED3D_RS_ALPHAFUNC
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6924 {STATE_RENDER(WINED3D_RS_ALPHAREF
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6925 {STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), alpha_test_arbfp
}, WINED3D_GL_EXT_NONE
},
6926 {STATE_RENDER(WINED3D_RS_COLORKEYENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6927 {STATE_COLOR_KEY
, { STATE_COLOR_KEY
, color_key_arbfp
}, WINED3D_GL_EXT_NONE
},
6928 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6929 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6930 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6931 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6932 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6933 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6934 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6935 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6936 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6937 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6938 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6939 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6940 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6941 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6942 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6943 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6944 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6945 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6946 {STATE_RENDER(WINED3D_RS_SHADEMODE
), { STATE_RENDER(WINED3D_RS_SHADEMODE
), state_shademode
}, WINED3D_GL_EXT_NONE
},
6947 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6950 static BOOL
arbfp_alloc_context_data(struct wined3d_context
*context
)
6955 static void arbfp_free_context_data(struct wined3d_context
*context
)
6959 const struct fragment_pipeline arbfp_fragment_pipeline
= {
6962 arbfp_get_emul_mask
,
6965 arbfp_alloc_context_data
,
6966 arbfp_free_context_data
,
6967 shader_arb_color_fixup_supported
,
6968 arbfp_fragmentstate_template
,
6971 struct arbfp_blit_type
6973 enum complex_fixup fixup
: 4;
6974 enum wined3d_gl_resource_type res_type
: 3;
6975 DWORD use_color_key
: 1;
6979 struct arbfp_blit_desc
6982 struct arbfp_blit_type type
;
6983 struct wine_rb_entry entry
;
6986 #define ARBFP_BLIT_PARAM_SIZE 0
6987 #define ARBFP_BLIT_PARAM_COLOR_KEY_LOW 1
6988 #define ARBFP_BLIT_PARAM_COLOR_KEY_HIGH 2
6990 struct arbfp_blit_priv
6992 struct wine_rb_tree shaders
;
6993 GLuint palette_texture
;
6996 static int arbfp_blit_type_compare(const void *key
, const struct wine_rb_entry
*entry
)
6998 const struct arbfp_blit_type
*ka
= key
;
6999 const struct arbfp_blit_type
*kb
= &WINE_RB_ENTRY_VALUE(entry
, const struct arbfp_blit_desc
, entry
)->type
;
7001 return memcmp(ka
, kb
, sizeof(*ka
));
7004 /* Context activation is done by the caller. */
7005 static void arbfp_free_blit_shader(struct wine_rb_entry
*entry
, void *context
)
7007 const struct wined3d_gl_info
*gl_info
= context
;
7008 struct arbfp_blit_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7010 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
7011 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
7012 HeapFree(GetProcessHeap(), 0, entry_arb
);
7015 static const struct wine_rb_functions wined3d_arbfp_blit_rb_functions
=
7020 arbfp_blit_type_compare
,
7023 static HRESULT
arbfp_blit_alloc(struct wined3d_device
*device
)
7025 struct arbfp_blit_priv
*priv
;
7027 if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
7028 return E_OUTOFMEMORY
;
7030 if (wine_rb_init(&priv
->shaders
, &wined3d_arbfp_blit_rb_functions
) == -1)
7032 ERR("Failed to initialize rbtree.\n");
7033 HeapFree(GetProcessHeap(), 0, priv
);
7034 return E_OUTOFMEMORY
;
7037 device
->blit_priv
= priv
;
7042 /* Context activation is done by the caller. */
7043 static void arbfp_blit_free(struct wined3d_device
*device
)
7045 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
7046 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
7048 wine_rb_destroy(&priv
->shaders
, arbfp_free_blit_shader
, &device
->adapter
->gl_info
);
7049 checkGLcall("Delete blit programs");
7051 if (priv
->palette_texture
)
7052 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &priv
->palette_texture
);
7054 HeapFree(GetProcessHeap(), 0, device
->blit_priv
);
7055 device
->blit_priv
= NULL
;
7058 static BOOL
gen_planar_yuv_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7062 const char *tex
, *texinstr
= "TXP";
7064 if (type
->fixup
== COMPLEX_FIXUP_UYVY
)
7075 tex
= arbfp_texture_target(type
->res_type
);
7076 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_RECT
)
7079 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
7080 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
7081 * filtering when we sample the texture.
7083 * These are the rules for reading the chroma:
7089 * So we have to get the sampling x position in non-normalized coordinates in integers
7091 if (type
->res_type
!= WINED3D_GL_RES_TYPE_TEX_RECT
)
7093 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
7094 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
7098 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7100 /* We must not allow filtering between pixel x and x+1, this would mix U and V
7101 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7104 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7105 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
7107 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
7108 * even and odd pixels respectively
7110 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
7111 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
7113 /* Sample Pixel 1 */
7114 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7116 /* Put the value into either of the chroma values */
7117 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7118 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
7119 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7120 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
7122 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
7123 * the pixel right to the current one. Otherwise, sample the left pixel.
7124 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
7126 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
7127 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
7128 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7130 /* Put the value into the other chroma */
7131 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7132 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
7133 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7134 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
7136 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
7137 * the current one and lerp the two U and V values
7140 /* This gives the correctly filtered luminance value */
7141 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
7146 static BOOL
gen_yv12_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7150 static const float yv12_coef
[]
7151 = {2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
};
7153 tex
= arbfp_texture_target(type
->res_type
);
7155 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7156 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
7157 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
7158 * pitch of the luminance plane, the packing into the gl texture is a bit
7159 * unfortunate. If the whole texture is interpreted as luminance data it looks
7160 * approximately like this:
7162 * +----------------------------------+----
7174 * +----------------+-----------------+----
7176 * | V even rows | V odd rows |
7178 * +----------------+------------------ -
7180 * | U even rows | U odd rows |
7182 * +----------------+-----------------+----
7186 * So it appears as if there are 4 chroma images, but in fact the odd rows
7187 * in the chroma images are in the same row as the even ones. So its is
7188 * kinda tricky to read
7190 * When reading from rectangle textures, keep in mind that the input y coordinates
7191 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
7193 shader_addline(buffer
, "PARAM yv12_coef = ");
7194 shader_arb_append_imm_vec4(buffer
, yv12_coef
);
7195 shader_addline(buffer
, ";\n");
7197 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7198 /* the chroma planes have only half the width */
7199 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7201 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
7202 * the coordinate. Also read the right side of the image when reading odd lines
7204 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
7207 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7210 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
7212 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
7214 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
7215 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
7217 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7218 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7219 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7220 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7221 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7223 /* clamp, keep the half pixel origin in mind */
7224 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
7225 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7226 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
7227 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7231 /* Read from [size - size+size/4] */
7232 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7233 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
7235 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7236 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7237 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7238 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7239 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
7240 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7242 /* Make sure to read exactly from the pixel center */
7243 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7244 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
7247 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
7248 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7249 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7250 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7251 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7253 /* Read the texture, put the result into the output register */
7254 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7255 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7257 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
7258 * No need to clamp because we're just reusing the already clamped value from above
7260 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7261 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
7263 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
7264 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7265 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7267 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7268 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7269 * values due to filtering
7271 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7272 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7274 /* Multiply the y coordinate by 2/3 and clamp it */
7275 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
7276 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
7277 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7278 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7282 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7283 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7286 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7287 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7288 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7295 static BOOL
gen_nv12_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7299 static const float nv12_coef
[]
7300 = {2.0f
/ 3.0f
, 1.0f
/ 3.0f
, 1.0f
, 1.0f
};
7302 tex
= arbfp_texture_target(type
->res_type
);
7304 /* NV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7305 * sized plane where each component is an UV pair. So the effective
7306 * bitdepth is 12 bits per pixel If the whole texture is interpreted as luminance
7307 * data it looks approximately like this:
7309 * +----------------------------------+----
7321 * +----------------------------------+----
7322 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7323 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7329 * +----------------------------------+----
7331 * When reading from rectangle textures, keep in mind that the input y coordinates
7332 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height. */
7334 shader_addline(buffer
, "PARAM nv12_coef = ");
7335 shader_arb_append_imm_vec4(buffer
, nv12_coef
);
7336 shader_addline(buffer
, ";\n");
7338 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7339 /* We only have half the number of chroma pixels. */
7340 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7342 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7344 shader_addline(buffer
, "RCP chroma.w, size.x;\n");
7345 shader_addline(buffer
, "RCP chroma.z, size.y;\n");
7347 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, nv12_coef.y, nv12_coef.x;\n");
7349 /* We must not allow filtering horizontally, this would mix U and V.
7350 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7351 * 0.5, so add 0.5. */
7353 /* Convert to non-normalized coordinates so we can find the
7354 * individual pixel. */
7355 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, size.x;\n");
7356 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7357 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7358 * add 0.5 to hit the center of the pixel. */
7359 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7361 /* Convert back to normalized coordinates. */
7362 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, chroma.w;\n");
7364 /* Clamp, keep the half pixel origin in mind. */
7365 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.z, nv12_coef.x;\n");
7366 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7367 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.z, nv12_coef.z;\n");
7368 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7372 /* Read from [size - size+size/2] */
7373 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.y, size.y;\n");
7375 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7376 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7377 * add 0.5 to hit the center of the pixel. */
7378 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7381 shader_addline(buffer
, "MAD temp.y, size.y, coef.y, size.y;\n");
7382 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7383 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7384 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7385 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7387 /* Read the texture, put the result into the output register. */
7388 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7389 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7391 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7394 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, chroma.w;\n");
7399 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.x;\n");
7401 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7402 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7404 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7405 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7406 * values due to filtering. */
7407 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7408 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7410 /* Multiply the y coordinate by 2/3 and clamp it */
7411 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, nv12_coef.x;\n");
7412 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, nv12_coef.x;\n");
7413 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7414 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7418 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7419 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7422 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7423 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7424 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7431 /* Context activation is done by the caller. */
7432 static GLuint
gen_p8_shader(struct arbfp_blit_priv
*priv
,
7433 const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7436 struct wined3d_string_buffer buffer
;
7437 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7439 /* This should not happen because we only use this conversion for
7440 * present blits which don't use color keying. */
7441 if (type
->use_color_key
)
7442 FIXME("Implement P8 color keying.\n");
7445 if (!string_buffer_init(&buffer
))
7447 ERR("Failed to initialize shader buffer.\n");
7451 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7452 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7455 string_buffer_free(&buffer
);
7459 shader_addline(&buffer
, "!!ARBfp1.0\n");
7460 shader_addline(&buffer
, "TEMP index;\n");
7462 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7463 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7465 /* The alpha-component contains the palette index */
7466 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7468 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7469 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7471 /* Use the alpha-component as an index in the palette to get the final color */
7472 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7473 shader_addline(&buffer
, "END\n");
7475 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7477 string_buffer_free(&buffer
);
7482 /* Context activation is done by the caller. */
7483 static void upload_palette(const struct wined3d_texture
*texture
, struct wined3d_context
*context
)
7485 const struct wined3d_palette
*palette
= texture
->swapchain
? texture
->swapchain
->palette
: NULL
;
7486 struct wined3d_device
*device
= texture
->resource
.device
;
7487 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7488 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
7490 if (!priv
->palette_texture
)
7491 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &priv
->palette_texture
);
7493 GL_EXTCALL(glActiveTexture(GL_TEXTURE1
));
7494 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, priv
->palette_texture
);
7496 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7498 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7499 /* Make sure we have discrete color levels. */
7500 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7501 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7502 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7505 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 256, 0, GL_BGRA
,
7506 GL_UNSIGNED_INT_8_8_8_8_REV
, palette
->colors
);
7510 static const DWORD black
;
7511 FIXME("P8 surface loaded without a palette.\n");
7512 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 1, 0, GL_BGRA
,
7513 GL_UNSIGNED_INT_8_8_8_8_REV
, &black
);
7516 /* Switch back to unit 0 in which the 2D texture will be stored. */
7517 context_active_texture(context
, gl_info
, 0);
7520 /* Context activation is done by the caller. */
7521 static GLuint
gen_yuv_shader(struct arbfp_blit_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
7522 const struct arbfp_blit_type
*type
)
7525 struct wined3d_string_buffer buffer
;
7526 char luminance_component
;
7528 if (type
->use_color_key
)
7529 FIXME("Implement YUV color keying.\n");
7532 if (!string_buffer_init(&buffer
))
7534 ERR("Failed to initialize shader buffer.\n");
7538 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7539 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7540 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7541 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7544 string_buffer_free(&buffer
);
7548 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7549 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7550 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7551 * each single pixel it contains, and one U and one V value shared between both
7554 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7555 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7556 * take the format into account when generating the read swizzles
7558 * Reading the Y value is straightforward - just sample the texture. The hardware
7559 * takes care of filtering in the horizontal and vertical direction.
7561 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7562 * because that would mix the U and V values of one pixel or two adjacent pixels.
7563 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7564 * regardless of the filtering setting. Vertical filtering works automatically
7565 * though - the U and V values of two rows are mixed nicely.
7567 * Apart of avoiding filtering issues, the code has to know which value it just
7568 * read, and where it can find the other one. To determine this, it checks if
7569 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7571 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7572 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7574 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7575 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7576 * in an unfiltered situation. Finding the luminance on the other hand requires
7577 * finding out if it is an odd or even pixel. The real drawback of this approach
7578 * is filtering. This would have to be emulated completely in the shader, reading
7579 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7580 * vertically. Beyond that it would require adjustments to the texture handling
7581 * code to deal with the width scaling
7583 shader_addline(&buffer
, "!!ARBfp1.0\n");
7584 shader_addline(&buffer
, "TEMP luminance;\n");
7585 shader_addline(&buffer
, "TEMP temp;\n");
7586 shader_addline(&buffer
, "TEMP chroma;\n");
7587 shader_addline(&buffer
, "TEMP texcrd;\n");
7588 shader_addline(&buffer
, "TEMP texcrd2;\n");
7589 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7590 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7591 shader_addline(&buffer
, "PARAM size = program.local[%u];\n", ARBFP_BLIT_PARAM_SIZE
);
7593 switch (type
->fixup
)
7595 case COMPLEX_FIXUP_UYVY
:
7596 case COMPLEX_FIXUP_YUY2
:
7597 if (!gen_planar_yuv_read(&buffer
, type
, &luminance_component
))
7599 string_buffer_free(&buffer
);
7604 case COMPLEX_FIXUP_YV12
:
7605 if (!gen_yv12_read(&buffer
, type
, &luminance_component
))
7607 string_buffer_free(&buffer
);
7612 case COMPLEX_FIXUP_NV12
:
7613 if (!gen_nv12_read(&buffer
, type
, &luminance_component
))
7615 string_buffer_free(&buffer
);
7621 FIXME("Unsupported YUV fixup %#x\n", type
->fixup
);
7622 string_buffer_free(&buffer
);
7626 /* Calculate the final result. Formula is taken from
7627 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7628 * ranges from -0.5 to 0.5
7630 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7632 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7633 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7634 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7635 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7636 shader_addline(&buffer
, "END\n");
7638 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7640 string_buffer_free(&buffer
);
7645 /* Context activation is done by the caller. */
7646 static GLuint
arbfp_gen_plain_shader(struct arbfp_blit_priv
*priv
,
7647 const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7650 struct wined3d_string_buffer buffer
;
7651 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7654 if (!string_buffer_init(&buffer
))
7656 ERR("Failed to initialize shader buffer.\n");
7660 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7663 string_buffer_free(&buffer
);
7666 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7668 shader_addline(&buffer
, "!!ARBfp1.0\n");
7670 if (type
->use_color_key
)
7672 shader_addline(&buffer
, "TEMP color;\n");
7673 shader_addline(&buffer
, "TEMP less, greater;\n");
7674 shader_addline(&buffer
, "PARAM color_key_low = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_LOW
);
7675 shader_addline(&buffer
, "PARAM color_key_high = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
);
7676 shader_addline(&buffer
, "TEX color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7677 shader_addline(&buffer
, "SLT less, color, color_key_low;\n"); /* below low key */
7678 shader_addline(&buffer
, "SGE greater, color, color_key_high;\n"); /* above high key */
7679 shader_addline(&buffer
, "ADD less, less, greater;\n"); /* or */
7680 shader_addline(&buffer
, "DP4 less.b, less, less;\n"); /* on any channel */
7681 shader_addline(&buffer
, "SGE less, -less.b, 0.0;\n"); /* logical not */
7682 shader_addline(&buffer
, "KIL -less;\n"); /* discard if true */
7683 shader_addline(&buffer
, "MOV result.color, color;\n");
7687 shader_addline(&buffer
, "TEX result.color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7690 shader_addline(&buffer
, "END\n");
7692 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7694 string_buffer_free(&buffer
);
7699 /* Context activation is done by the caller. */
7700 static HRESULT
arbfp_blit_set(void *blit_priv
, struct wined3d_context
*context
, const struct wined3d_surface
*surface
,
7701 const struct wined3d_color_key
*color_key
)
7703 const struct wined3d_texture
*texture
= surface
->container
;
7704 struct arbfp_blit_priv
*priv
= blit_priv
;
7705 enum complex_fixup fixup
;
7706 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7707 struct wine_rb_entry
*entry
;
7708 struct arbfp_blit_type type
;
7709 struct arbfp_blit_desc
*desc
;
7710 struct wined3d_color float_color_key
[2];
7711 struct wined3d_vec4 size
;
7714 size
.x
= wined3d_texture_get_level_pow2_width(texture
, surface
->texture_level
);
7715 size
.y
= wined3d_texture_get_level_pow2_height(texture
, surface
->texture_level
);
7719 if (is_complex_fixup(texture
->resource
.format
->color_fixup
))
7720 fixup
= get_complex_fixup(texture
->resource
.format
->color_fixup
);
7722 fixup
= COMPLEX_FIXUP_NONE
;
7724 switch (texture
->target
)
7727 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_1D
;
7731 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7735 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_3D
;
7738 case GL_TEXTURE_CUBE_MAP_ARB
:
7739 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_CUBE
;
7742 case GL_TEXTURE_RECTANGLE_ARB
:
7743 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_RECT
;
7747 ERR("Unexpected GL texture type %#x.\n", texture
->target
);
7748 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7751 type
.use_color_key
= !!color_key
;
7754 entry
= wine_rb_get(&priv
->shaders
, &type
);
7757 desc
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7758 shader
= desc
->shader
;
7764 case COMPLEX_FIXUP_NONE
:
7765 if (!is_identity_fixup(texture
->resource
.format
->color_fixup
))
7766 FIXME("Implement support for sign or swizzle fixups.\n");
7767 shader
= arbfp_gen_plain_shader(priv
, gl_info
, &type
);
7770 case COMPLEX_FIXUP_P8
:
7771 shader
= gen_p8_shader(priv
, gl_info
, &type
);
7774 case COMPLEX_FIXUP_YUY2
:
7775 case COMPLEX_FIXUP_UYVY
:
7776 case COMPLEX_FIXUP_YV12
:
7777 case COMPLEX_FIXUP_NV12
:
7778 shader
= gen_yuv_shader(priv
, gl_info
, &type
);
7784 FIXME("Unsupported complex fixup %#x, not setting a shader\n", fixup
);
7788 desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*desc
));
7793 desc
->shader
= shader
;
7794 if (wine_rb_put(&priv
->shaders
, &desc
->type
, &desc
->entry
) == -1)
7797 ERR("Out of memory\n");
7798 GL_EXTCALL(glDeleteProgramsARB(1, &shader
));
7799 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader))");
7800 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, 0));
7801 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, 0)");
7802 HeapFree(GetProcessHeap(), 0, desc
);
7803 return E_OUTOFMEMORY
;
7807 if (fixup
== COMPLEX_FIXUP_P8
)
7808 upload_palette(texture
, context
);
7810 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7811 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7812 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7813 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7814 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARBFP_BLIT_PARAM_SIZE
, &size
.x
));
7815 checkGLcall("glProgramLocalParameter4fvARB");
7816 if (type
.use_color_key
)
7818 wined3d_format_get_float_color_key(texture
->resource
.format
, color_key
, float_color_key
);
7819 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7820 ARBFP_BLIT_PARAM_COLOR_KEY_LOW
, &float_color_key
[0].r
));
7821 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7822 ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
, &float_color_key
[1].r
));
7823 checkGLcall("glProgramLocalParameter4fvARB");
7829 /* Context activation is done by the caller. */
7830 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7832 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7833 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7836 static BOOL
arbfp_blit_supported(const struct wined3d_gl_info
*gl_info
,
7837 const struct wined3d_d3d_info
*d3d_info
, enum wined3d_blit_op blit_op
,
7838 const RECT
*src_rect
, DWORD src_usage
, enum wined3d_pool src_pool
, const struct wined3d_format
*src_format
,
7839 const RECT
*dst_rect
, DWORD dst_usage
, enum wined3d_pool dst_pool
, const struct wined3d_format
*dst_format
)
7841 enum complex_fixup src_fixup
;
7843 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7848 case WINED3D_BLIT_OP_COLOR_BLIT_CKEY
:
7849 if (!d3d_info
->shader_color_key
)
7851 /* The conversion modifies the alpha channel so the color key might no longer match. */
7852 TRACE("Color keying not supported with converted textures.\n");
7855 case WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
:
7856 case WINED3D_BLIT_OP_COLOR_BLIT
:
7860 TRACE("Unsupported blit_op=%d\n", blit_op
);
7864 if (src_pool
== WINED3D_POOL_SYSTEM_MEM
|| dst_pool
== WINED3D_POOL_SYSTEM_MEM
)
7867 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7868 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7870 TRACE("Checking support for fixup:\n");
7871 dump_color_fixup_desc(src_format
->color_fixup
);
7874 if (!is_identity_fixup(dst_format
->color_fixup
))
7876 TRACE("Destination fixups are not supported\n");
7880 if (is_identity_fixup(src_format
->color_fixup
))
7886 /* We only support YUV conversions. */
7887 if (!is_complex_fixup(src_format
->color_fixup
))
7889 TRACE("[FAILED]\n");
7895 case COMPLEX_FIXUP_YUY2
:
7896 case COMPLEX_FIXUP_UYVY
:
7897 case COMPLEX_FIXUP_YV12
:
7898 case COMPLEX_FIXUP_NV12
:
7899 case COMPLEX_FIXUP_P8
:
7904 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7905 TRACE("[FAILED]\n");
7910 static void arbfp_blit_surface(struct wined3d_device
*device
, enum wined3d_blit_op op
, DWORD filter
,
7911 struct wined3d_surface
*src_surface
, const RECT
*src_rect_in
,
7912 struct wined3d_surface
*dst_surface
, const RECT
*dst_rect_in
,
7913 const struct wined3d_color_key
*color_key
)
7915 unsigned int dst_sub_resource_idx
= surface_get_sub_resource_idx(dst_surface
);
7916 struct wined3d_texture
*src_texture
= src_surface
->container
;
7917 struct wined3d_texture
*dst_texture
= dst_surface
->container
;
7918 struct wined3d_context
*context
;
7919 RECT src_rect
= *src_rect_in
;
7920 RECT dst_rect
= *dst_rect_in
;
7921 struct wined3d_color_key alpha_test_key
;
7923 /* Activate the destination context, set it up for blitting */
7924 context
= context_acquire(device
, dst_surface
);
7926 /* Now load the surface */
7927 if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7928 && (surface_get_sub_resource(src_surface
)->locations
7929 & (WINED3D_LOCATION_TEXTURE_RGB
| WINED3D_LOCATION_DRAWABLE
))
7930 == WINED3D_LOCATION_DRAWABLE
7931 && !wined3d_resource_is_offscreen(&src_texture
->resource
))
7933 /* Without FBO blits transferring from the drawable to the texture is
7934 * expensive, because we have to flip the data in sysmem. Since we can
7935 * flip in the blitter, we don't actually need that flip anyway. So we
7936 * use the surface's texture as scratch texture, and flip the source
7937 * rectangle instead. */
7938 surface_load_fb_texture(src_surface
, FALSE
, context
);
7940 src_rect
.top
= wined3d_texture_get_level_height(src_texture
, src_surface
->texture_level
) - src_rect
.top
;
7941 src_rect
.bottom
= wined3d_texture_get_level_height(src_texture
, src_surface
->texture_level
) - src_rect
.bottom
;
7944 wined3d_texture_load(src_texture
, context
, FALSE
);
7946 context_apply_blit_state(context
, device
);
7948 if (!wined3d_resource_is_offscreen(&dst_texture
->resource
))
7949 surface_translate_drawable_coords(dst_surface
, context
->win_handle
, &dst_rect
);
7951 if (op
== WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
)
7953 const struct wined3d_format
*fmt
= src_texture
->resource
.format
;
7954 alpha_test_key
.color_space_low_value
= 0;
7955 alpha_test_key
.color_space_high_value
= ~(((1u << fmt
->alpha_size
) - 1) << fmt
->alpha_offset
);
7956 color_key
= &alpha_test_key
;
7959 arbfp_blit_set(device
->blit_priv
, context
, src_surface
, color_key
);
7961 /* Draw a textured quad */
7962 draw_textured_quad(src_surface
, context
, &src_rect
, &dst_rect
, filter
);
7964 /* Leave the opengl state valid for blitting */
7965 arbfp_blit_unset(context
->gl_info
);
7967 if (wined3d_settings
.strict_draw_ordering
7968 || (dst_texture
->swapchain
&& (dst_texture
->swapchain
->front_buffer
== dst_texture
)))
7969 context
->gl_info
->gl_ops
.gl
.p_glFlush(); /* Flush to ensure ordering across contexts. */
7971 context_release(context
);
7973 wined3d_texture_validate_location(dst_texture
, dst_sub_resource_idx
, dst_texture
->resource
.draw_binding
);
7974 wined3d_texture_invalidate_location(dst_texture
, dst_sub_resource_idx
, ~dst_texture
->resource
.draw_binding
);
7977 static HRESULT
arbfp_blit_color_fill(struct wined3d_device
*device
, struct wined3d_rendertarget_view
*view
,
7978 const RECT
*rect
, const struct wined3d_color
*color
)
7980 FIXME("Color filling not implemented by arbfp_blit\n");
7981 return WINED3DERR_INVALIDCALL
;
7984 static HRESULT
arbfp_blit_depth_fill(struct wined3d_device
*device
, struct wined3d_rendertarget_view
*view
,
7985 const RECT
*rect
, DWORD clear_flags
, float depth
, DWORD stencil
)
7987 FIXME("Depth/stencil filling not implemented by arbfp_blit.\n");
7988 return WINED3DERR_INVALIDCALL
;
7991 const struct blit_shader arbfp_blit
= {
7996 arbfp_blit_supported
,
7997 arbfp_blit_color_fill
,
7998 arbfp_blit_depth_fill
,